Introduction to

Microcontroller
Programming for
Power Electronics
Control Applications
 Introduction to
Microcontroller
Programming for
Power Electronics
Control Applications
Coding with MATLAB® and
Simulink®

Mattia Rossi
Nicola Toscani
Marco Mauri
Francesco Castelli Dezza
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks
does not warrant the accuracy of the text or exercises in this book. This book’s use or discussion of
MATLAB® software or related products does not constitute endorsement or sponsorship by The
MathWorks of a particular pedagogical approach or particular use of the MATLAB® software.

First edition published 2022

by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742

and by CRC Press

2 Park Square, Milton Park, Abingdon, Oxon, OX14 4RN

© 2022 Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, LLC

Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of their use.
The authors and publishers have attempted to trace the copyright holders of all material reproduced in
this publication and apologize to copyright holders if permission to publish in this form has not been
obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, access www.copyright.com
or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. For works that are not available on CCC please contact mpkbookspermissions@tandf.
co.uk

Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Library of Congress Cataloging‑in‑Publication Data

ISBN: 978-0-367-70985-3 (hbk)

ISBN: 978-1-032-05303-5 (pbk)
ISBN: 978-1-003-19693-8 (ebk)
eBook + ISBN: 978-1-032-05465-0

DOI: 10.1201/9781003196938

Typeset in LM Roman
by KnowledgeWorks Global Ltd.

eResources are available for this title at www.routledge.com/9780367709853

a Daniela
Contents

Foreword xiii

Preface xvii

Acknowledgments xix

Biographies xxi

1 Advances in Firmware Design for

Power Electronics Control Platforms 1
1.1 Embedded Control System . . . . . . . . . . . . . . . . . . . 1
1.2 Selecting a Development Board . . . . . . . . . . . . . . . . . 3
1.2.1 Key elements of a microcontroller . . . . . . . . . . . . 4
1.2.2 Programming microcontrollers . . . . . . . . . . . . . 5
1.3 The C2000™ Family of MCU from Texas Instruments™ . . . 7
1.4 Scheme of a Power Electronics Control Problem . . . . . . . 9

I Embedded Development:
Hardware Kits and Coding 13
2 Automatic Code Generation through MATLAB® 15
2.1 Model-Based Design and Rapid Prototyping . . . . . . . . . 16
2.2 Workflow for Automatic Code Generation . . . . . . . . . . . 18
2.3 Generate Code for C2000™ Microcontrollers . . . . . . . . . 22
2.4 TI C2000™ Processors Block-set . . . . . . . . . . . . . . . . 24

3 Texas Instruments™ Development Kit 27

3.1 TI C2000™ LaunchPad™ : F28069M Piccolo . . . . . . . . . 27
3.1.1 Features . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.2 Pin muxing . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.3 Power connectivity . . . . . . . . . . . . . . . . . . . . 32
3.1.4 Serial connectivity . . . . . . . . . . . . . . . . . . . . 33
3.1.5 Boot options . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 TI BOOSTXL-DRV8301 BoosterPack . . . . . . . . . . . . 34
3.2.1 BoosterPack PWM signals . . . . . . . . . . . . . . . . 35
3.2.2 BoosterPack GPIO signals . . . . . . . . . . . . . . . . 36
3.2.3 DC bus and phase voltage sense . . . . . . . . . . . . 37

vii
viii Contents

3.2.4 Low-side shunt-based current sense . . . . . . . . . . . 37

4 Software Installation 39
4.1 TI Support Packages:
Code Composer™ Studio and ControlSUITE™ . . . . . . . . 39
4.2 MATLAB® Support Package:
Embedded Coder for Texas Instruments C2000 Processors . 41
4.3 Installation Procedure . . . . . . . . . . . . . . . . . . . . . . 41

II Review of Control Theory: Closing the Loop 47

Introduction 49

5 Designing a Closed-Loop Control System 51

5.1 Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.1 Mathematical laws . . . . . . . . . . . . . . . . . . . . 52
5.1.2 Dynamical systems in electrical applications . . . . . . 54
5.2 Design a PI Controller in Continuous-Time Domain . . . . . 54
5.2.1 Serial/parallel form . . . . . . . . . . . . . . . . . . . . 55
5.2.2 Characterization of the closed-loop dynamics F (s) . . 55
5.3 Derive a PI Controller in Discrete-Time Domain . . . . . . . 60
5.3.1 General properties of the discretization process . . . . 60
5.3.2 Characterization of the closed-loop dynamics F (z) . . 62

6 Design Example: PI-Based Current Control of an RL Load 65

6.1 Simulink® Simulation . . . . . . . . . . . . . . . . . . . . . . 67
6.1.1 Use of standard blocks (continuous/discrete) . . . . . 71
6.1.2 Use of Simscape™ (specialized power systems) . . . . 73
6.1.3 Controller performances . . . . . . . . . . . . . . . . . 74
6.2 Derive an Anti-Windup PI Controller Scheme . . . . . . . . 77
6.3 Design Summary . . . . . . . . . . . . . . . . . . . . . . . . . 82

7 Manipulate the Variables Format: Data Types 85

7.1 Fixed Point vs Floating Point Representation . . . . . . . . . 85
7.2 Single vs Double Precision . . . . . . . . . . . . . . . . . . . 88
7.3 Use of Scaling in Fixed Point Representation . . . . . . . . 91
7.4 Converting from Decimal Representation to Single Format . 93
7.5 Processing the Data: Implementation Hints . . . . . . . . . . 95

III Real-Time Control in Power Electronics:

Peripherals Settings 97
Introduction 99
Contents ix

8 Basic Settings:
Serial Communication COM and Hardware Target 101
8.1 Virtual Serial Communication through COM port . . . . . . 101

9 Simulink® Configuration 105

9.1 Simulink® Environments: Firmware vs Testing . . . . . . . . 107
9.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . 107
9.1.2 Execution in Simulink® . . . . . . . . . . . . . . . . . 108
9.2 MCUs and Real-Time Control with Simulink® . . . . . . . . 109

10 Serial Communication Interface (SCI) Peripheral 111

10.1 Hardware Details . . . . . . . . . . . . . . . . . . . . . . . . 112
10.2 Firmware Environment:
Send and Receive Data through Serial Communication . . . 113
10.2.1 C2806x SCI receive . . . . . . . . . . . . . . . . . . . . 113
10.2.2 C2806x SCI transmit . . . . . . . . . . . . . . . . . . . 116
10.3 Testing Environment:
Send/Receive Data through Serial Communication . . . . . . 117
10.3.1 Serial configuration . . . . . . . . . . . . . . . . . . . . 117
10.3.2 Serial send . . . . . . . . . . . . . . . . . . . . . . . . 118
10.3.3 Serial receive . . . . . . . . . . . . . . . . . . . . . . . 119
10.4 Time Variable Settings (Sample Rates) . . . . . . . . . . . . 120
10.5 Examples on Serial Communication . . . . . . . . . . . . . . 122

11 GPIO Peripheral—Digital Input/Output 131

11.1 Hardware Details . . . . . . . . . . . . . . . . . . . . . . . . 131
11.2 Firmware Environment: GPIO Peripherals . . . . . . . . . . 133
11.2.1 C2806x GPIO digital input (GPIO DI) . . . . . . . . 133
11.2.2 C2806x digital output (GPIO DO) . . . . . . . . . . . 134
11.3 Examples with GPIO blocks . . . . . . . . . . . . . . . . . . 135

12 Analog to Digital Converter Peripheral 149

12.1 Operating Principle . . . . . . . . . . . . . . . . . . . . . . . 149
12.1.1 Sample & hold . . . . . . . . . . . . . . . . . . . . . . 150
12.1.2 Analog to digital converter . . . . . . . . . . . . . . . 150
12.2 Hardware Details . . . . . . . . . . . . . . . . . . . . . . . . 151
12.2.1 Difference between acquisition window
and sample time . . . . . . . . . . . . . . . . . . . . . 153
12.3 Firmware Environment: ADC Peripheral . . . . . . . . . . . 153
12.3.1 C2806x ADC . . . . . . . . . . . . . . . . . . . . . . . 153
12.4 Example with ADC block . . . . . . . . . . . . . . . . . . . . 155
12.5 Synchronization between ADC modules . . . . . . . . . . . . 160
x Contents

13 Pulse Width Modulator Peripheral 163

13.1 Operating Principle . . . . . . . . . . . . . . . . . . . . . . . 164
13.2 Hardware Details . . . . . . . . . . . . . . . . . . . . . . . . 166
13.2.1 ePWM sub-modules . . . . . . . . . . . . . . . . . . . 169
13.3 Generation of PWM signals . . . . . . . . . . . . . . . . . . . 171
13.3.1 Counting modes . . . . . . . . . . . . . . . . . . . . . 172
13.3.2 ePWMxA and ePWMxB sub-modules . . . . . . . . . 174
13.3.3 Setting dead bands . . . . . . . . . . . . . . . . . . . . 175
13.4 Firmware Environment: ePWM Peripheral . . . . . . . . . . 178
13.4.1 C2806x ePWM . . . . . . . . . . . . . . . . . . . . . . 178
13.5 Example with ePWM block . . . . . . . . . . . . . . . . . . . 186
13.6 DAC Peripheral—Filtered PWM . . . . . . . . . . . . . . . . 191
13.7 Examples with DAC Peripherals . . . . . . . . . . . . . . . . 192
13.8 Synchronization between Multiple ePWM Modules . . . . . . 197
13.9 Synchronization between ADC and ePWM Modules:
Average Measurements . . . . . . . . . . . . . . . . . . . . . 202
13.10Events Execution within Sample Time . . . . . . . . . . . . 204

14 Encoder Peripheral 207

14.1 Operating Principle of Incremental Encoders . . . . . . . . . 207
14.2 Hardware Details . . . . . . . . . . . . . . . . . . . . . . . . 209
14.3 Optical Rotary Encoder LPD3806 . . . . . . . . . . . . . . . 210
14.4 Speed Computation . . . . . . . . . . . . . . . . . . . . . . . 211
14.5 Firmware Environment: eQEP Peripheral . . . . . . . . . . . 213
14.5.1 C2806x eQEP . . . . . . . . . . . . . . . . . . . . . . . 213
14.6 Example with eQEP block . . . . . . . . . . . . . . . . . . . 215

IV Real-Time Control in Power Electronics:

Applications 219
15 Open Loop Control of a Permanent Magnet DC Motor 221
15.1 Required Hardware . . . . . . . . . . . . . . . . . . . . . . . 221
15.2 Linear Model of a PMDC Motor . . . . . . . . . . . . . . . . 222
15.3 System Simulations . . . . . . . . . . . . . . . . . . . . . . . 226
15.4 Half-Bridge Configuration . . . . . . . . . . . . . . . . . . . . 227
15.4.1 Control implementation . . . . . . . . . . . . . . . . . 231
15.5 Full-Bridge Configuration . . . . . . . . . . . . . . . . . . . . 234
15.5.1 Modulation strategies . . . . . . . . . . . . . . . . . . 235
15.5.2 Unipolar voltage switching . . . . . . . . . . . . . . . 236
15.5.3 Bipolar voltage switching . . . . . . . . . . . . . . . . 241
15.5.4 Control implementation . . . . . . . . . . . . . . . . . 246

16 Low-Side Shunt Current Sensing 251

16.1 Sensor Characterization: Theoretical Approach . . . . . . . . 252
16.2 Locked Rotor Test . . . . . . . . . . . . . . . . . . . . . . . . 254
16.3 Sensor Characterization: Experimental Approach . . . . . . . 260
Contents xi

17 Current Control of an RL Load 267

17.1 Required Hardware . . . . . . . . . . . . . . . . . . . . . . . 267
17.2 Linear Average Model and Controller Design . . . . . . . . . 269
17.3 System Simulations . . . . . . . . . . . . . . . . . . . . . . . 271
17.3.1 Detailed modeling of the actuation variables . . . . . . 271
17.4 Half-Bridge Configuration . . . . . . . . . . . . . . . . . . . . 273
17.4.1 Control implementation . . . . . . . . . . . . . . . . . 280
17.5 Variation of Load Parameters . . . . . . . . . . . . . . . . . 288
17.5.1 Effects on the transient response . . . . . . . . . . . . 288
17.5.2 Parameters estimation . . . . . . . . . . . . . . . . . . 290

18 Voltage Control of an RLC load 293

18.1 Required Hardware . . . . . . . . . . . . . . . . . . . . . . . 293
18.2 Guidelines for the Hardware Design of a RLC Load . . . . . 296
18.3 General State-Space Average Modeling Method . . . . . . . . 300
18.3.1 Linear average model and controller design . . . . . . 303
18.4 System Simulations . . . . . . . . . . . . . . . . . . . . . . . 306
18.5 Half-Bridge Configuration . . . . . . . . . . . . . . . . . . . . 306
18.5.1 Control implementation . . . . . . . . . . . . . . . . . 314
18.6 Variations of LC Filter Parameters . . . . . . . . . . . . . . 322

19 Cascade Speed Control of a Permanent Magnet DC Motor 325

19.1 Required Hardware . . . . . . . . . . . . . . . . . . . . . . . 327
19.2 Linear Model of a PMDC Motor . . . . . . . . . . . . . . . . 328
19.3 Cascade Control Architecture and Design . . . . . . . . . . . 330
19.4 System Simulations . . . . . . . . . . . . . . . . . . . . . . . 333
19.5 Full-Bridge Configuration . . . . . . . . . . . . . . . . . . . . 334
19.5.1 Model reference adaptive system (MRAS) observer . . 344
19.6 Single Motor Configuration . . . . . . . . . . . . . . . . . . . 347
19.6.1 Parameter identification . . . . . . . . . . . . . . . . . 348
19.6.2 Control implementation . . . . . . . . . . . . . . . . . 350
19.7 Back-to-Back (B2B) Configuration . . . . . . . . . . . . . . . 361
19.7.1 Parameter identification . . . . . . . . . . . . . . . . . 363
19.7.2 Control implementation . . . . . . . . . . . . . . . . . 364

V Real-Time Control in Power Electronics:

Load Emulation 371
20 Debugging Tools and Firmware Profiling 373
20.1 Processor-in-the-loop with Simulink® . . . . . . . . . . . . . 373
20.1.1 PMDC motor control implementation through PIL . . 375
20.2 External Mode Execution with Simulink® . . . . . . . . . . . 380
20.2.1 Simulink® setup for external mode execution . . . . . 381
xii Contents

21 Electric Propulsion Case Studies 385

21.1 Urban Tramway . . . . . . . . . . . . . . . . . . . . . . . . . 385
21.2 Electric Racing Car . . . . . . . . . . . . . . . . . . . . . . . 390

A Appendix A: Basics of C 401

A.1 Operations between numbers . . . . . . . . . . . . . . . . . . 401
A.1.1 Sum and differences . . . . . . . . . . . . . . . . . . . 401
A.1.2 Shift operation . . . . . . . . . . . . . . . . . . . . . . 401
A.1.3 Multiplication . . . . . . . . . . . . . . . . . . . . . . . 402
A.1.4 Division . . . . . . . . . . . . . . . . . . . . . . . . . . 402
A.2 Structure of a C program . . . . . . . . . . . . . . . . . . . 403

B Appendix B: Custom Expansion Boards and Hardware Kits 405

Bibliography 423
Index 427
Foreword

The new book Introduction to Microcontroller Programming for Power Elec-

tronics Control Applications contains the fundamental subjects of the interdis-
ciplinary field of power electronic based systems, which draws knowledge from
circuit and control theory, (digital) signal processing for embedded implemen-
tation, electrical machines/drives, and power semiconductor devices. Written
for students and practicing engineers, this book introduces the analysis and de-
sign of motor control systems and their implementation on microcontrollers.
The requirements and capabilities of the latter influence the structure and
design choice of the closed-loop control scheme, a subject particularly rele-
vant for laboratory activities both at university and industry level. This book
presents state-of-the-art techniques to implement modulation schemes and
control algorithms in a commercial microcontroller (MCU) suitable for rapid
prototyping approach, and hint for designing analog circuits, such as low-
voltage converter, output filters/load. MATLAB/Simulink® is introduced and
used to solve example problems.
The book presents a concise workflow for the reader by using a specific
embedded target, which is not a limiting factor for the validity of the sug-
gested approach. The latter can be extended even to different boards. It is
valuable to every graduate student, serving as a textbook for classes (looking
to create teamwork) and as a starting point for more advanced studies, for
industry professionals, researchers, and academics willing to study the broad
field of power electronics. The contents of the book are easy to read and pre-
sented in an interesting way with good illustrations and solid background on
the underlying theory. Solved problems (built-in files) are presented to help
the readers.

Introduction to Microcontroller Programming for Power Electronics Control

Applications is unique in the synthesis of the main characteristics of electri-
cal drive behavior and in their link to the main implementation aspect into
modern MCU, trying to fill the gap between system/circuit design, control
techniques, and digital signal processing. The authors are power electronics
and drives specialists from the Electrical Machines, Drives and Power Elec-
tronics Research Group of Politecnico di Milano, Italy. They collected in the
book the several years of teaching and subjects of the laboratory activities for

xiii
xiv Foreword

the M.Sc. in Automation and Control Engineering and Electrical Engineering

at the same university, with the collaboration and contribution of industrial
partner. As industrial partners, we retain this book a great achievement of
the last four years collaboration with the authors.

Angelo Strati — Field Application Engineer Italy

Domenico Santoro — Field Application Engineer Italy
. Giuseppe Ballarin — Team Leader FAE Italy
Würth Elektronik Group
[email protected]

Most people go about their day blissfully unaware of the electric motors that
are spinning the world around us. We wake up staring upwards at a ceiling
fan, silently rotating in a circle. We jump into our car and rely on up to
40 motors—pumps, fans, locks, and lifts—to get us to our destination. We
power up our laptop computers and hear the soft whine of fans working to
keep the electronics cool. Motors are everywhere because they are one of the
main ways that an electronic circuit can interact with the real world, i.e. a
power electronic-based system. They are “lectromechanical, turning analog
and digital signals into real and visible mechanical motion. It is estimated
that electric motors consume 45 percent of the total worldwide electricity—
this is a stunning statistic! As we look to reducing energy consumption and
enabling a greener future, electric motors present a huge potential for efficiency
improvements.
Few engineering students are aware of the impact of electric motors on
the world around them, and even less are versed in the design and control of
motor systems. This is a problem! We need engineers growing in competency
in this field to create better and more efficient motor drive systems.
Motor drive and control is an incredibly multidisciplinary field. Real-time
digital processing is implemented in microcontrollers to be the “brain” behind
the motor system; controlling speed, power, and efficiency from the digital
domain. A wealth of analog components from power management (voltage
regulators & gate drivers) and signal chain (amplifiers & sensors) interface
the microcontroller to the motor through a power converter while providing
sensing, safe operation, and support for the system. Texas Instruments has
over 25 years of experience in the field of real-time control and also provides

Texas Instruments Incorporated (Nasdaq: TXN) is a global semiconductor company

that designs, manufactures, tests and sells analog and embedded processing chips for mar-
kets such as industrial, automotive, personal electronics, communications equipment and
enterprise systems. Our passion to create a better world by making electronics more afford-
able through semiconductors is alive today, as each generation of innovation builds upon
the last to make our technology smaller, more efficient, more reliable and more affordable—
making it possible for semiconductors to go into electronics everywhere. We think of this
as Engineering Progress. It is what we do and have been doing for decades. Learn more at
www.ti.com.
Foreword xv

a comprehensive analog portfolio covering every block of the motor drive and
control system.
This book presents very practical and important lessons to engineers and
engineering students alike on the topics of motor drive and control, covering
not only general concepts but details on how to create a motor drive system.
It provides an excellent resource to encourage the next generation of engineers
to grow and develop skills in the area of electric motors and power electronics,
introducing them the tools they need to make an impact on the world.
Politecnico di Milano is an outstanding academic partner, and the focus
of the Electrical Machines, Drives and Power Electronics Research Group on
cutting-edge power electronic-based technologies helps shape quality engineer-
ing minds. We wish the best of success to this publication and to the continued
collaboration between industry and academia.

Olivier Monnier — Marketing Manager, C2000 Real-Time MCU

Matt Hein — Applications Manager, Brushless-DC Motor Drives
. Antonio Faggio — Field Application Engineer
Texas Instruments Inc.
[email protected]

Can you write the 100 million lines of code that are needed to build an average
modern car? The answer is pretty obvious: of course you can, it’s just a matter
of time. And how would you compare the complexity of this problem to writing
the 4501 lines of assembly code needed to build the first version of UNIX in
1971? While both tasks appear to be at a similar level of dauntlessness, the half
century separating them has witnessed the emergence of high-level languages
that enable programmers to address highly complex problems on their own
while reusing the legacy of their peers.
At MathWorks Inc., we relentlessly work on providing the best high-level
programming tools to automate the implementation of your ideas into embed-
ded systems. Simulink allows you to design and simulate complex algorithms
that you can translate into thousands of lines of embedded code with a click
of a button via our code generation technology.
The book Introduction to Microcontroller Programming for Power Elec-
tronics Control Applications will teach you how to use these modern techniques
to create control algorithms for systems involving complex physics. The re-
markable work of Mattia Rossi, Nicola Toscani, Marco Mauri and Francesco
Castelli Dezza from Politecnico di Milano, Italy, clearly explains deep concepts
to the reader in the field of embedded programming for power electronics ap-
plications using Model-Based Design.

The copyrighted material included in this book is reprinted with permission of The
MathWorks, Inc.
xvi Foreword

While the shift to digital is now largely dominating the industry of mo-
tor control, this revolution is just starting for power conversion applications.
The material in this book provides state of art techniques to train the many
engineers that the world needs tomorrow in a field that is at the core of the
indispensable transition to clean energy.
In recent conversations with Mattia and Nicola, while they politely thanked
us for our help, it was clear to us that the quality and the amount of effort
in this book deserved much more thanking from our side. With this foreword,
we extend all our gratitude to this outstanding contribution to accelerating
the pace of engineering and science, our core mission.

Antonin Ancelle — Embedded Targets Development Manager

John Kluza — Embedded Systems Partner Manager
Tom Erkkinen — Code Generation Marketing Manager
.
Brian McKay — Embedded Systems Partner Manager
MathWorks Inc.
[email protected]

www.we-online.com www.ti.com www.mathworks.com

Preface

Power electronics-based systems are the key enabling technology to meet most
of the future sustainable challenges from grid to motor applications.
Standard textbooks and courses about power electronics and electrical ma-
chines deal with analysis in continuous-time, averaged modeling of switched-
mode power converters, and continuous-time control theory. Nevertheless, real
control algorithms and management functions around power converters are
implemented digitally, thus, extending the field of fundamentals studies to
discrete-time modeling and digital control concepts specific to power electron-
ics. The necessary background is achieved by combining specific textbooks
and courses from both power electronics and digital control theory. However,
students who approach the design of digitally controlled power converters for
the first time may not fully understand and successfully practice for a targeted
problem due to such fragmented references.
In this book, we attempt to fill this gap by treating the fundamental as-
pects of digital control implementation for power electronics based systems
in a systematic and rigorous manner. Our objectives are to put the reader
in the position to understand, analyze, model, design, and implement digital
feedback loops around power converters, from system-level transfer function
formulations to understand which coding tool may be used when working
with microcontroller (MCU or µC) platforms. In particular, the latter be-
longs to Texas Instruments™ C2000™ family, which is specifically designed
for real-time closed loop control such as power supplies, industrial drives,
and solar inverters applications. The Simulink® environment is able to au-
tomatically generate ANSI/ISO C/C++ code tailored for specific embedded
targets through a model-based workflow. Given the settings which enable a
background usage of the Code Composer™ Studio IDE, a Simulink® scheme
can be directly compiled and executed on C2000™ MCUs. Such automated
build and execution procedure speed up the control algorithms implementa-
tion, thus, the code generation of software interfaces and MCU peripherals
(e.g. ADC, digital I/O, PWM), which can be tested with execution profiling.
This makes the reader working in a rapid prototyping manner.
This book is oriented to graduate students of electrical and automation
and control engineering pursuing a curriculum in power electronics and drives.
Moreover, it aims to be a reference for engineers and researchers who seek to
expand on the expertize in design-oriented knowledge for the aforementioned
applications. It is assumed that the reader is well acquainted with funda-
mentals of electrical machines and power converters, along with associated

xvii
xviii Preface

continuous-time modeling and control techniques. Familiarity with sampled-

data, discrete-time system analysis and embedded design topics is helpful but
not absolutely essential since the scope of the book is to provide a basic knowl-
edge even to whom is approaching these topics for the first time. Key concepts
are developed from scratches, including a brief review of control theory and
detailed description of the hardware and test bench used. Either custom ex-
pansion boards and assembled kits are open to users community. Project files
can be shared and/or pre-assembled boards/kit can be directly shipped.
There is no standard notation to cover all of the topics covered in this
book. We tried to use the most familiar notation from the literature whenever
possible to help the reader’s understanding through the chapters.
Acknowledgments

Most of the projects shown in this book have been funded by the Department
of Mechanical Engineering of Politecnico di Milano, Italy, with the particular
contribution of its Laboratory of Electrical Drives and Power Electronics.
Special thanks goes to the university rector Prof. Ferrucio Resta and the
department director Prof. Marco Bocciolone for their support to this initiative.
We are grateful to our graduate students Marco Gerosa, Matteo Scandella,
Andrea Polastri, Matteo Sposito, and Luca Grittini for the precious work
in supporting the hardware development, the boards testing and the many
suggestions they made. We also thank all the Ph.D. students and researchers
who contributed to this book. In particular, Dr. Khaled ElShawarby and Dr.
Alberto Bolzoni, who were supporting the project from day zero.

This book is the final outcome of a collaboration between Politecnico di

Milano, Italy, Würth Elektronik™ Group Texas Instruments™ Inc. and
MathWorks® Inc. . The authors are grateful to them for their support, help
and vision. In particular, Tristan De Cande for his encouragement to tight
the collaboration between university and company; Antonio Faggio and the
US C2000™ product line of Texas Instruments™ Inc. for their commitment to
invest in young generation growing at the university; Angelo Strati, Giuseppe
Ballarin, Domenico Santoro, Andrea De Gruttola from Würth Elektronik™
Group for helping in components selection and boards design; Antonin
Ancelle, John Kluza and their teams from MathWorks® Inc. for their help
on embedded development. Moreover, we thank also MathWorks® Inc. for let-
ting us being part of their Book Program.

The authors would like to specially thank Prof. Petros Karamanakos from
Tampere University, Finland, and Prof. Ralph Kennel from Technical Univer-
sity Munich, Germany, for their guidance, long discussions and availability to
share their high expertize in this field.

We are grateful to Nora Konopka and CRC Press LLC from Taylor and Fran-
cis Group for publishing this book. Special thanks goes to Prachi Mishra for
her guidance and support.

Finally, we acknowledge the inspiration, patience, and support of our fami-

lies during the preparation of this book, who allowed us to work during long
nights, weekends, and holidays. The book is dedicated to Daniela.

xix
Biographies

Mattia Rossi is a Research Assistant at Politecnico di

Milano, Italy. He received the B.Sc. and M.Sc. degrees in
Automation and Control Engineering from Politecnico di
Milano, Italy, in 2013 and 2015, respectively. In 2015, he
was at the ABB MV Drives, Switzerland, working on en-
hanced motor control design to reduce mechanical vibrations
in motor-load couplings. Since 2016, he is a Ph.D. student
in Electrical Engineering at Politecnico di Milano, Italy, in
collaboration with Tampere University, Finland. In 2019, he
was a visiting Ph.D. student at the Technical University of
Munich, Germany. His main research activities cover model predictive control
(MPC) algorithms for multilevel medium voltage power electronic-based sys-
tems and their embedded implementation, aiming to improve power conversion
efficiency and system components reliability. He received the Best Student Pa-
per Award at the 2019 IEEE International Symposium on Predictive Control
of Electrical Drives and Power Electronics, and the Jorma Luomi Student Fo-
rum Award at the 2016 International Conference on Electrical Machines.
Email: [email protected]

Nicola Toscani received the Bachelor (B.Sc.), Master of

Science (M.Sc.), and Doctor of Philosophy (Ph.D.) degrees
in Electrical Engineering from the Politecnico di Milano,
Milan, Italy, in 2013, 2015, and 2019, respectively. He is
currently working as a Postdoctoral Research Fellow in the
Department of Mechanical Engineering of Politecnico di Mi-
lano. His last research activities deal with the development
and the control of electrical machines for high-performance
vehicles and wireless power transfer. His research inter-
ests also include modeling strategies for Electromagnetic
Compatibility problems, Power Electronics and Electrical Drives. Email:
[email protected]

xxi
xxii Biographies

Marco Mauri is an Assistant Professor in Electrical Ma-

chines and Drives at Politecnico di Milano, Italy. He received
the Master of Science (M.Sc.) and Doctor of Philosophy
(Ph.D.) degrees in Electrical Engineering from Politecnico
di Milano in 1998 and 2002. His research interests mainly
include the control of electrical machines and modeling prin-
ciple of electrical drives and electromagnetic effects. He is a
member of the IEEE Power Electronics and Industrial Elec-
tronics societies Email: [email protected]

Francesco Castelli Dezza is a Full Professor in Electri-

cal Machines and Drives at Politecnico di Milano, Italy.
He received the Master of Science (M.Sc.) and Doctor of
Philosophy (Ph.D.) degrees in electrical engineering from
the Politecnico di Milano, Milano, Italy, in 1986 and 1990,
respectively. He is currently the head of the Electrical Ma-
chines, Drives and Power Electronics research group of Po-
litecnico di Milano, Italy, which is mainly located at the De-
partment of Mechanical Engineering of the same university.
His research interests include studies on dynamic behavior of
electrical machines, electrical drives control and design, and power electronics
for energy flow management. He is a member of the IEEE Power Electronics
and Industrial Electronics societies. Email: [email protected]
1
Advances in Firmware Design for Power
Electronics Control Platforms

The rising complexity and tighter requirements of power electronics-based sys-

tems lead to a great complexity of digital control algorithms, which has to be
tested in many working conditions, i.e., looking for reliability. Hence, firmware
design and testing routines have become increasingly time-consuming. To
speed up the proof of concepts and to improve firmware quality assurance, the
simulation analysis is combined with several automatic code generation proce-
dures to achieve rapid prototyping. Such approach is aimed to test the control
algorithm and the related generated code on real hardware through mini-
mal modifications. This approach also helps to identify coding and platform-
specific configuration errors, as well as to get rid of them. In particular, this
book refers to microcontrollers (MCU or µC).
In this first, introductory chapter, basics of embedded systems and corre-
sponding coding approaches are presented in detail.

1.1 Embedded Control System

Everyone who encounter these book subjects for the first time may have a
very fundamental question: what is an embedded system?
To provide a clear answer, it is necessary to underline the major difference
between a microprocessor and a microcontroller first.
A microprocessor is the computational intelligence of complex systems,
i.e., the central processing unit (CPU), which is optimized to do logical-
mathematical computations. It usually has mainly on-board communica-
tions peripherals to interact with the rest of the system. However, in many
fields, including power electronics-based systems, there is the need for com-
municating with rather complex peripherals like Analog-to-Digital Conver-
sion (ADC) channels, Digital-to-Analog Conversion (DAC) modules, Pulse-
Width-Modulation (PWM) peripheral, and Inter Integrated Circuit (I2C)
communication.
A microcontroller (MCU) is designed to include such peripherals. Hence,
a MCU is a device built around a microprocessor (or CPU), where the

DOI: 10.1201/9781003196938-1 1
2 Advances in Firmware Design for Power Electronics Control Platforms

Standard microprocessor Microcontroller

(PCs) (Power electronics)
• High computational power • “Low” computational power
• Several communication peripheral • Communication, actuation and
sensing (measurements) peripherals
• General usage (e.g., software sim- • Target usage: optimizing perfor-
ulations, document editing) mances and costs

number of peripherals as well as their type and accuracy/resolution are related

to the final market price other than the specific task they have to carry out.
Figure 1.1 shows an example of architecture of MCU board. Note that, MCU
refers to microprocessor for which the computational power is significantly
lower compared to those used inside a personal computer. This is the reason
why most of the MCUs for the industrial markets are able to manage a re-
duced number of computations even if they drive high-speed/high-resolution
peripherals.
Therefore, an embedded system can be generally defined as a control plat-
form based on a programmable logic (i.e., microprocessor), where the control
algorithm and the peripheral interfaces are dedicated to specific tasks/func-
tions or a group of them. All the application requirements, e.g., from ADC
resolution to modulator dead time, are given from the beginning. This allows
every user to choose the best trade-off between computational power, type of
peripherals, and unitary cost, optimizing its design. The algorithm is executed
in real-time, generally without using an on-board operating systems (OS). In-
deed a custom (light) OS might be needed just for the most demanding cases.
Nowadays, the modern definition of embedded system mainly refers to mi-
crocontrollers. Nevertheless, even a single microprocessor which is specialized

CPU
A/D
CONVERTER RAM

PROGRAM
CLOCK
MEMORY
MCU

Figure 1.1 Example of the architecture of a MCU board [11].

Selecting a Development Board 3

Figure 1.2 Examples of power electronic-based applications [8].

in certain class of computations, or even custom designed for the applica-

tion at hand, may be considered as an embedded system. This is the case
of digital signal processors (DSPs), which are out of the scope of this book.
The wide-ranging insertion of power electronics in many applications with
different requirements still creates market for microcontrollers. Some of such
applications are reported in Figure 1.2. As an example, the MCU usage in the
automotive field is grown about +11% from 2006 to 2015. The popularity of
MCUs depends on their ability to work with floating point variables (i.e., to
include a floating point unit) as well as their benefit from mass production.
Nevertheless, some applications require cost-effective devices,1 thus, working
with integer variables (e.g., 8–16 bit), targeting low cost MCUs. Furthermore,
in order to extend the potential market of control platforms, today microcon-
trollers can be interfaced with different peripheral on the same pins, which
can be chosen by setting different values on the corresponding registers. Thus,
that allows to have more peripheral channels than pins.

1.2 Selecting a Development Board

Power electronics-based systems like power management and motor drive are
going on with their fast evolution toward high-performance and high-efficiency
solutions. Likewise, the demand for digital control algorithms managing the
electronic component operations and the overall system has been rising, and
embedded systems have been increasing in complexity reaching fast execution
times. Therefore, it has become a standard engineering practice to test the
control algorithm through real-time simulations and to verify its feasibility
based on the hardware at disposal. This is the reason why in both research
centers and academia development board must be updated according to the
new emerging platforms on the market.
The selection of the development board represents the very first
task of a project. In particular, development boards are designed to be
efficient, portable, and sensored according to a specific applications field.

1 From the customer’s point of view the “best” microcontroller is the one which matches

all the application requirements at the lowest price.

 4 Advances in Firmware Design for Power Electronics Control Platforms

Since this book focuses on electrical power conversion case studies, from
now on the text refers to MCU platforms targeted to power electronics-based
applications. Even if this last statement reduces the candidate list, there still
are several suitable solutions available on the market which share a common
goal of being compact and versatile. Since the definition of a criterion to cat-
egorize each board would not be practical (i.e., there might be even deep
technology differences), it is recommended to follow a system-level approach
like the one presented here:
1. Consider all the components and peripherals that are necessary to
run all of the required features. If the board lacks any of them, it is
important to identify some supported expansions to include them.
2. Consider the supported programming languages and the level of
competence of the final user. Moreover, investigate if any automatic
code generation procedure is available as well as the quality of the
documentation at disposal for the adopted board. Indeed, commu-
nity and support are factors of great importance since they are the
primary resources when designing a project.
3. Evaluate costs versus adopted components for the considered spe-
cific application: is it worth paying for them?

1.2.1 Key elements of a microcontroller

The main features of microcontroller boards that set their performances are
reported below:

• CPU and clock speed: these two values affect the overall performance of
the board. Namely, how fast it can perform computations. It should be noted
that clock speed comparisons between CPUs coming from different families
may not lead to meaningful considerations. Other factors, such as instruction
cycles, instruction sets, and pipeline depth, also affect MCU performances.
• RAM: the size of this memory affects the number of tasks that can be run
simultaneously. It also impacts how fast data can be processed, as swapping
it from RAM to nonvolatile storage incurs large performance overheads.
• Graphical processing unit (GPU): it allows development board to run
video output (e.g., VGA/HDMI). High-performance GPUs are needed while
processing video/images with the development board.
• Data memory: it affects the size of programs, operating systems, and gen-
erated/downloaded data that can be stored on development boards.
• General purpose input/output pins: these pins are used to connect
external components to the development board in use. Hence, more pins
typically means more possible simultaneous connections. These pins are
usually assigned specific functionalities by the manufacturer being compli-
ant with some standards, such as integrated circuit (I2C), serial peripheral
 Selecting a Development Board 5

interface (SPI), and universal asynchronous receiver/transmitter (UART).

These standards facilitate interoperability with third-party devices such as
displays or sensors.
• Analog-in pins: these pins are necessary for any data acquisition from
sensors and they are classified through the resolution and sampling rate of
the peripheral to which they are connected. In particular, resolution refers
to the number of discrete levels to which the input signal is quantized, while
sampling rate is the number of data points that can be obtained in a reference
time interval. High resolution is required for accurate measurements, while
high sampling rates are needed for the acquisition of fast changing signals.
• Pulse width modulator: pulse width modulation is a kind of digital signal
that is ideal for mimicking analog signals. Such signals are generated through
modulators which vary the duty cycle of square waves depending on the
desired output.
• Power supply and consumption: both of them can play a major role in
design choices. For portable products, the runtime requirements and power
demand of the board/associated components must be considered when se-
lecting an energy source. For example, a development board serving as a
data logger in a remote location must run uninterrupted for months, while
another one working as a mobile personal computer only needs to run for sev-
eral hours before recharging. Power consumption can be difficult to quantify
as many boards have varying modes of operation, allowing some of them
to work in very low power consumption modes. When considering power
consumption, the designer must take the following factors into account: re-
quired computational power, maximum run time without recharging, and
the cost/size of the energy source. The size of power sources is easier to
determine since it depends on the rated values of the adopted board

1.2.2 Programming microcontrollers

The development environment of boards affects how the designer interacts
with the hardware. The support for multiple operating systems, languages,
and integrated development environments creates a rich programming frame-
work and more appealing development board as well.
Depending on the specific application, all the onboard peripherals as well
as the main algorithm features have to be correctly set while programming
the MCU for executing required tasks.
Across the years, different coding approaches have been developed:

1. Low-level, based on machine code (e.g., Assembler language);

2. Intermediate-level, based on a combination of C-code (to define
algorithm features and peripheral settings) and the usage of an In-
tegrated Development Environment (IDE, e.g., Code Composer™
6 Advances in Firmware Design for Power Electronics Control Platforms

Studio from Texas Instruments™ ) to compile, link, debug, and down-

load the script on target;
3. High-level, based on a combination of objective-oriented languages
(e.g., Simulink® , to define algorithm features and peripherals set-
tings) and the usage of specific target toolbox to generate C-code,
compile, link, debug and download the algorithm into platform.
C code is considered as the universal language for microcontroller program-
ming, with C++ following closely behind. In addition, more and more boards
have custom IDEs associated with them, providing pre-configured device sup-
port and libraries, aimed at offering an all-around user-friendly workspace.
More versatile IDEs (i.e., those not designed for a specific development board)
allow users to explore alternative programming languages.
Each of the previously mentioned approaches has advantages and draw-
backs according to the end-user type. Some of the main differences between
low and high level coding are listed here in the following:
• Platform Dependencies
Low-level programming languages are platform-dependent, which means
that programs are written in such way they can run on the one hardware
configuration only. Instead, high-level programming languages are platform-
independent, that means programs written in such way can run on different
hardware configurations.
Remark: platform-independent does not mean operating system-
independent. Indeed, hardware configuration may change, but OS (if
present) should be the same for every new setup.
• Speed
Low-level language programs are faster than those written in high-level lan-
guage since they do not need to convert any algorithm in code executable on
the hardware. In addition, low-level languages have less syntax, functions,
keywords, and class libraries compared to other coding approaches.
• Easy Programming and Flexibility
Low-level languages are not as easy and user-friendly to manage as high-
level ones. Indeed, there are only two low-level programming languages,
which are Binary and Assembly. Binary language foresees codes made by
zeros and ones only, whereas Assembly requires some symbols knows as
mnemonics which are difficult to type. On the other hand, high-level lan-
guage programs are easy to write, read, modify, and understand. Moreover,
high-level languages have huge libraries with a rich set of functions, allowing
the development of algorithms for several applications with low effort.
• Performance
Since low-level language programs are faster than any other, their perfor-
mance are for sure better than those of high-level ones.
 The C2000™ Family of MCU from Texas Instruments™ 7

• Translation
Programs written in Binary code does not need any translation as this lan-
guage is a machine code already. Namely, the hardware is capable of un-
derstanding them without any translation. Instead, Assembly codes need
an Assembler to translate programs to their equivalent counterpart in Ma-
chine Code. High-level languages are always translated by compilers or in-
terpreters. Some of them required both compilers and interpreters to get the
Object/Binary file.

• Support
Low-level languages have less support than high-level ones. There may be
lower number of communities for low-level languages than for high-level ones.

This book focuses on high-level programming languages which are user-

friendly both for students and for everyone who is approaching microcontroller
programming for the first time. In particular, coding/programming through
Simulink® is proposed in the following chapters.

1.3 The C2000™ Family of MCU from Texas

Instruments™
Texas Instruments™ (TI) offers free software development tools for
LaunchPad™ boards, such as Code Composer™ Studio (IDE) and Energia2 .
In addition, TI allows free use (limited size) of cloud-based development tools.
All these tools accept C/C++ code for programming.

Texas Instruments™ has a wide range of embedded processors from very

low-cost, limited performance up to high-cost, very high performance. Its main
product families are:

• MSP43x™ ultra-low-power microcontroller family;

• C2000™ microcontroller family;

• ARM-based microcontroller family;

• C5000™ low-power signal-processing DSPs;
• C6000™ high-perfomance signal-processing DSPs.

Among them, the C2000™ microcontroller family (also known as the

TMS320C2000™ family) is a product line aimed at high-performance control

2 Energia is a rapid prototyping platform based on Arduino IDE.

 8 Advances in Firmware Design for Power Electronics Control Platforms

applications in the fields of motor control, digital power supplies, lighting, re-
newable energy, and smart grids. This family is made up of several subfamilies,
from which it is worth mentioning:

• C24xx: a 16-bit microcontroller that evolved from the TMS320C2x family

of digital signal processors.
• C28xx: a 32-bit microcontroller, fixed or floating point, with a robust set
of peripherals and I/Os which match the classical needs typical of power
electronics applications.

In particular, the C28xx chips are from low to high performance MCU.
Piccolo™ (which main features and applications are summarized in Figure 1.3)
and Delfino™ are the families for low and high performance microcontrollers,
respectively. Their main characteristics are reported here in the following:
1. Piccolo™
•MCU with floating-point unit;
•CPU frequency: from 40 to 120 MHz;
•Core: 1xC28x;
•Memory:
from 60 kbit up to 512 kbit flash;
from 12 kbit up to 100 kbit SRAM;
•Main peripherals: ADC, PWM, QEP, DMA, SPI, UART, I2C,
CAN, USB.
2. Delfino™
•MCU with floating-point unit;
•CPU frequency: from 100 to 200 MHz;
•Core: from 1xC28x up to 2xC28x + 2xCLA + ARM Cortex-
M4;
•Memory:
from 512 kB up to 1.5 MB flash;
from 68 kB up to 338 kB SRAM;
•Main peripherals: ADC, PWM, QEP, DMA, SPI, UART, I2C,
CAN, EMIF.
In particular, the following families of C28xx MCUs (which are also available
as LaunchPad™ development kits) are supported with a dedicated library
available in Matlab® Simulink® : F2802x Piccolo™ , F2803x Piccolo™ , F2805x
Piccolo™ , F2807x Piccolo™ , F2806x Piccolo™ , F2837xS Delfino™ , F2837xD
Delfino™ , F28004x, F2823x Delfino™ , F28M3x.
From Chapter 2 on, this book refers to a specific model of development board,
that is Texas Instruments™ LaunchXL F28069M Piccolo™ , which is shown in
 Scheme of a Power Electronics Control Problem 9

Figure 1.3 LaunchPad™ Piccolo MCU framework [10].

Figure 1.4. The reason behind this choice is its low-cost, low-power, and simple
development environment of the board. Moreover, it has both a 256 kbit flash
memory and a 96 kbit RAM. Finally, the community and project support for
this device is relatively sparse compared to other boards. TI offers several plug-
in expansion boards to expand the capabilities of the F28069M LaunchPad™
Piccolo™ .

1.4 Scheme of a Power Electronics Control Problem

Most of the MCU-based closed loop control schemes for power electronics
applications can be summarized in a structure like the one reported in Figure
1.5, which refers specifically to C2000™ processors. More in details, the key
elements/features of this scheme (i.e., the ones necessary to close a loop) are:

• Reference: the controller must act to let the system follow a reference
signal. The latter can be both internally generated on the MCU or provided
as an external signal.
• Processing: this relates to the control logic implemented on the MCU.
The control structure is the main part of the scheme, since it is aimed at
generating the desired output based on the error between the reference and
the feedback signals. This stage may also comprise scaling routines.

Figure 1.4 F28069M LaunchPad™ Piccolo™ board.

 10 Advances in Firmware Design for Power Electronics Control Platforms

Figure 1.5 Classical structure of a closed-loop control scheme for implemen-

tations on C2000™ -based boards and its main elements [13].

• Actuation: the controller output must be translated into a signal which

must be suitable and consistent with the actuation logic used to operate the
power converter. The modulator stage is the key element of the actuation
block, which also comprise the safety routines.
• Power electronics: this refer to the switching devices used to realize a
energy conversion. This system is used to drive different loads based on the
reference signal. Generally, this stage is not embedded into MCU boards.
Typically, they are built up on expansion boards.
• Sensing: sensors may be embedded on the MCU board or physically in-
stalled in close proximity of the switching device. In both case, analog sig-
nals must be acquired and digitalized through analog to digital converters
(included in the architecture of the MCU), being available to be processed
for closing the loop.

Note that, the C2000™ MCU family includes several parts of this struc-
ture representing a complex ecosystem. Its main features are summarized in
Figure 1.6.

Figure 1.6 C2000™ MCU feature highlights [13].

 Scheme of a Power Electronics Control Problem 11

Figure 1.7 Traction converter and motor control for an efficient conversion DC
to AC to drive an electric motor [9].

Examples of fields of application for MCU boards

The wide-ranging integration of power electronics in many applications with
different requirements creates many control challenges. Several constraints
means large embedded platform capabilities (in addition to ad-hoc power elec-
tronics design) which is reflected in the demand of specific number/type of
peripherals, resolution/accuracy, computational power, safety target, power
absorbition, etc . Some examples of power electronic-based applications that
can be controlled by an MCU-based control scheme are:
• Industrial drives: they are aimed to control a wide variety of motors, e.g.,
AC and servo drives, equipped with sensing technologies (e.g., encoder or
sensor-less solutions) and communications networks (like real-time Ethernet
communications and functional safety topologies), which may be included
in a complex production system (such as robotic lines or rolling mills) or
targeted to a specific industrial task.
• Electric vehicles (auxiliaries and traction systems): advanced power
control technology is needed for the electrification of vehicles. In particu-
lar, a car environment comprises motor control technologies, e.g., traction
or propulsion motors, auxiliary motors, power steering, and digital power
 12 Advances in Firmware Design for Power Electronics Control Platforms

Figure 1.8 Example of energy collection from PV panels [8].

technology, like charging stations (AC/DC), DC/DC power conversion, on-

board charging, and AC output stage. In order to achieve high performance
while maximizing driving range and minimizing energy loss, more attention
is needed on the choice of the right motor and the most efficient control
techniques. As more electric vehicles enter the market, it is challenging to
provide on-board charging solutions that can support multiple regions with
different power grid infrastructures [27].
Several aspects and subsystems may be defined and analyzed, thus, electric
vehicles represent a complex environment, as shown in Figure 1.7.
• Renewable and digital power: this kind of applications addresses the
growing need for high efficiency, smaller form factor, and adaptability in
digitally controlled power electronic converters driving energy productions,
such as photovoltaic panels and wind turbines.
Digital power management and control capabilities provide real-time intel-
ligence allowing for the development of adaptable high frequency switching
power supplies that automatically tune to their environment, improving ef-
ficiency and performance. This automatic adjustment allows for changes in
input voltage, output load current and system temperature, delivering en-
ergy savings with dynamic voltage scaling, and advanced control techniques
for optimal system performance [39].
Several aspects and subsystems (like batteries and power conversion substa-
tions) may be defined and analyzed, as reported in the renewable energy-
based distribution system shown in Figure 1.8.
 Part I

Embedded Development:
Hardware Kits and Coding
 2
Automatic Code Generation through
MATLAB®

Any time engineers design digital controls for power electronic-based applica-
tions, there are many good reasons to perform modeling and simulation:
• Test system behavior an possible variations, e.g., in the topology, power
supply, load.
• Test different passive (e.g. resistors, capacitors, inductors) and active (e.g.
semiconductor technology) elements to find suitable components.
• Test if the feedback control algorithms are able to meet the currents/volt-
ages/speeds regulation requirements.
Then, move to the coding stage aimed to embedded implementation for a
specific target. MathWorks® provides tools which bring the simulation stage
together with the implementation one, creating a powerful ecosystem which
allow to speed up the workflow from idea to practice.
In this book, the concepts of rapid prototyping and digital control tech-
niques for power electronics-based systems are explained by programming
a TI C2000™ based MCU platform through the MathWorks® MATLAB®
and Simulink® frameworks. Both MATLAB® and Simulink® are commonly
used for the analysis, design, simulation and optimization of models, includ-
ing power electronic circuits. For the latter, the Simscape™ Power Systems1
toolbox allows to model all the parts of power networks and to take into
account the realistic behavior of each component. Moreover, MathWorks®
along with various MCU manufacturers, such as Texas Instruments® , devel-
oped several Simulink® toolboxes aimed at automatically generating C/C++
targeted code for a specific CPU. Such toolboxes work together with the
IDE of the microcontoller supplier, i.e., TI Code Composer™ Studio in this
case.

1 Simscape™ toolbox enables a quick creation of models of physical systems within the

Simulink® environment. With Simscape™ , component models based on physical connections

are built directly integrated with block diagrams and other modeling paradigms. Thus, it
is possible to model systems such as electric motors, bridge rectifiers other than hydraulic
actuators. Their interactions can be included in the same environment.

DOI: 10.1201/9781003196938-2 15
16 Automatic Code Generation through MATLAB®

Figure 2.1 Model-Based Design workflow [34].

2.1 Model-Based Design and Rapid Prototyping

Rather than relying on physical prototypes and textual specifications, Model-
Based Design approach proposed by MathWorks® in [34] uses a system model
as an executable specification throughout development. It supports system-
and component-level design and simulation, automatic code generation, and
continuous test and verification.
Anyway, the best way to understand what a Model-Based Design is, it is
through an example (the reader is referred to [34]):
A team of engineers sets out to build a motor control unit for an AC
electrical drive. Because they are using the Model-Based Design approach,
they begin by settling an architecture model from the system requirements;
in this case, the system comprises the AC electrical machine, the power
converter, the sensing/actuation stage and the control unit. A simula-
tion/design model is then derived. This high-level model includes portions
of the controls software that will be running in the control unit, and the
load–in this case, the motor and the power converter. The team performs
initial system and integration tests by simulating this high-level model un-
der various scenarios to verify that the system is represented correctly and
that it properly responds to input signals. They add detail to the model,
continuously testing and verifying the system-level behavior against spec-
ifications. If the system is large and complex, the engineers can develop
and test individual components independently but still test them frequently
in a full system simulation. Ultimately, they build a detailed model of the
 Model-Based Design and Rapid Prototyping 17

Figure 2.2 Model-Based Design workflow specified to power electronic-based

applications [33].

system and the specific environment in which it operates. This model cap-
tures the accumulated knowledge about the control unit (e.g. considering
the peripheral of the targeted MCU). The engineers generate code auto-
matically from the model of the control algorithms for firmware testing and
verification. Then, they download the generated code onto production hard-
ware (e.g., MCU) for testing in an real hardware.

As this example shows, Model-Based Design uses the same elements as tradi-
tional development workflows, but with two key differences:
• A system model is at the heart of development, from requirements capture
through design, implementation, and testing;
• It requires to follow this modeling approach to enable the automated rou-
tines, e.g., automatic code generation.
The workflow reported in Figure 2.1 can be specified for power electronic-
based applications, as shown in Figure 2.2.
 18 Automatic Code Generation through MATLAB®

Simulations allow to analyze system performances in conditions otherwise

too expensive, risky, or time-consuming to consider. This aspect combined
with the possibility to automatically generate C, C++, HDL, or Structured
Text from the model (the generated code can be optimized and combined with
hand-written code) defines the rapid prototyping approach.

Goals of Real-Time Rapid Prototyping

Assuming that the functional requirements, the system hardware and the
targeted control platform are given, rapid prototyping (in terms of real-time
simulation, testing and coding) can be used to:

• Refine and verify the functional operations of control system design with
the exploited hardware by rapidly iterating between algorithm design and
prototyping;
• Continuously explore and test new ideas using a flexible, scalable platform;
validate whether a component can adequately control the physical system
in real time;
• Evaluate system performance, investigate scenarios and hardware interac-
tions that are complex, expensive, or dangerous to perform with production
hardware (i.e. before laying out hardware, coding production software);
• Test hardware cutting the development time from idea to practice, to avoid
costly design flaws by detecting errors early when they are still cost-effective
to correct.

For more information on this topic, the reader is referenced to [34].

2.2 Workflow for Automatic Code Generation

Making changes to the controller code during hardware testing can be time-
consuming. Indeed, modifying the code, recompiling and deploying it to the
microcontroller frequently may take long time depending on the extent of the
required improvements. As an alternative, Simulink® can generate C/C++
code for a specific MCU starting from the scheme which models the controller,
achieving rapid prototyping. By simulating the complete power electronic sys-
tem (controller included) over all its possible operating and fault conditions,
the ability to handle those same scenarios in the real system is increased.
Assuming to consider a generic power electronics-based systems, e.g., a
two-level converter driving a combined resistive-inductive (RL) load, it is
quite easy (and fast) to simulate the system behavior by representing it as
a transfer function or by means of physical components from the Simscape™
Workflow for Automatic Code Generation 19
Micro controller, discrete time

y ∗ (k)
Reference + e (k) PI u (k) PWM u (t) Power v (t) Electrical
− controller modulator electronics system

y (k) y (t)
Sensor

Figure 2.3 Block diagram of a microcontroller-based closed-loop system for

power electronics applications.

Power Systems library. In both cases, the system receives an actuating vari-
able/signal u(t) and it returns an output variable/signal y(t). Supposing to
design a closed-loop control, e.g., a PI-based current control, the regulator
processes the error y ∗ (t) − y(t) and it returns the control input u(t). The
whole control scheme is reported in Figure 2.3.
Due to the switching nature of power converters (i.e., discrete on-off behav-
ior), a modulation stage such as pulse width modulation (PWM), translates
the control input u∗ (t) into the actuating signal u(t) by a suitable switch-
ing pattern of the power converter. In this framework, the controller has to
be implemented into the processor of the MCU which cannot handle con-
tinuous signals. Thus, the control algorithm must be discretized. Namely, its
input/output signals are sampled at discrete time instances, i.e., u(k) and
y(k), while the discrete controller form is derived from the continuous-time
one through a discretization method. These latter aspects are deeply discussed
in Part II of this book.
The overall control implementation process can be summarized in three
main steps:
1. The Simulink® file is used to test and to optimize the controller
through simulation before its deployment on the selected hardware2 ;
2. Once the control design is ready, the control input and output are
substituted by the related MCU peripherals which are given in
Simulink® as block-set. According to the peripheral requirements,
it may be necessary to edit the data type of the signals;
3. Finally, the overall scheme (controller + I/O block-sets) can be
deployed on the MCU, i.e., translated into binary code and uploaded
into the MCU. The whole procedure is summarized in Figure 2.4.
This model translation is not simple since it involves several hidden steps.
Simulink® converts the model to a C programming code through MathWorks®

2 Given that the converter behavior is assumed to be faster than the load dynamics,

it is common practice to design the controller without taking the converter dynamics into
account
20 Automatic Code Generation through MATLAB®
Specifcations
I m pro v e m e nts

s
Simulation Bug
Firmware
Design of the
execution
control logic
E rrors
Discretization

Firmware
Automatic Deploy
Peripheral setting
code generation to hardware
Data type setting

Figure 2.4 Workflow for firmware design.

Embedded Coder® . Then, the executable C code is fed to the IDE Code Com-
poser Studio™ in which it is:
1. Sequentially compiled to assembly language exploiting the Texas
Instruments libraries;
2. Assembled (e.g., ASM source code);
3. Lik-edited;
4. Downloaded on the TI C2000™ MCU flash memory.
Figure 2.5 shows such steps with a flowchart.

Benefits
The key advantage of such rapid prototyping technique is a seamless integra-
tion capability over multiple processors. This can be achieved by just replacing

Simulink®
Embedded
Coder

C-code

Code Composer Studio™

Compiler

Assembly Linker OBJ to Hex

OBJ fle HEX fle
language converter

Download

Hardware

Figure 2.5 Simulink® workflow for firmware deployment on MCU boards.

 Workflow for Automatic Code Generation 21

the processor specific block-set, making the necessary changes in their config-
uration rather than rewriting or rebuilding the whole model. This is valid not
only for hardware made by the same manufacturer, but by different producers
as well. Hence, designers do not have to worry about the compatibility of the
code. To validate changes made on the controller, it is enough to run the simu-
lation model first and, then, to verify that no errors are generated. Hence, this
approach is naturally oriented to research and development activities (i.e., for
academic and industries).

Drawbacks
It is important to underline how such procedure may imply performance bot-
tlenecks. Skipping the effort of low-level coding may limit the computational
efficiency of the generated code. The resulting C/C++ code is numerically
equivalent to the previously validated algorithms in Simulink® , but these lat-
ter has to be prepared for code generation, e.g., introducing implementation
considerations needed for low-level C code and using functions for code gen-
eration support.
To clarify such concept, the code generation of a simple MATLAB® func-
tion which multiplies two inputs is investigated here in the following:

function c = Prod(a,b) #include “Prod.h”

% multiply two inputs double Prod(double a, double b)
c = a * b; {
AAreturn a * b;
}
(a) MATLAB® source code (b) generated C code

Given two scalar inputs, the automatically generated C code maps clearly
back to the MATLAB® environment, as shown above.
Nevertheless, as any MATLAB® algorithms intended for code generation,
implementation constraints due to the differences between the two program-
ming languages must be considered. These mainly include:

• Memory allocation: in MATLAB® , memory allocation is automatic while

in C code is manual. Namely, it may be allocated either statically (using
static), dynamically (using malloc), or on the stack (using local variables).
• Array-based language: MATLAB® provides a rich set of array operations
that allow concise coding of numerical algorithms; C code requires explicit
for-loops to express the same algorithms.
• Data type: MATLAB® automatically determines the data types and sizes
as the code runs; C language requires explicit type declarations on all the
exploited variables and functions (e.g., int, uint, single, double).
 22 Automatic Code Generation through MATLAB®

• Polymorphism: MATLAB® functions can support many different input

types and they adapt the use of proper operators, while C code requires
clear data type declarations.

Given the purpose of this book, the last two points require particular at-
tention. The polymorphism can give a single line of MATLAB® code different
meanings depending on the inputs. For example, the function shown previ-
ously could mean scalar multiplication, dot product, or matrix multiplication.
In fact, the inputs could be of different data types (logical, integer, floating-
point, fixed-point) or either real or complex numbers. If two matrices are
multiplied, the automatic procedure produce many lines of C code, even with
3 for-loops, as shown here in the following:
void Prod ( const double a [12] ,
const double b [20] ,
double c [15])
{
int i0 ; int i1 ; int i2 ;
for ( i0 = 0; i0 <3; i0 ++)
{
for ( i1 = 0; i1 <5; i1 ++)
{
c [ i0 + 3 * i1 ] = 0.0;
for ( i2 = 0; i2 <4; i2 ++)
{
c [ i0 + 3 * i1 ] =...
... a [ i0 + 3 * i2 ] * b [ i2 + ( i1 << 2)];
}
}
}
}

Thus, this piece of code looks quite different from that one reported before.
For further information on this topic, the reader is referenced to [5].

2.3 Generate Code for C2000™ Microcontrollers

MATLAB® Coder™ , Simulink® Coder™ , and Embedded Coder® generate
ANSI/ISO C/C++ code that can be compiled and executed on Texas
Instruments™ (TI) C2000™ microcontrollers (MCUs) using Code Composer™
Studio IDE. Embedded Coder® allows an easy configuration of the code gen-
eration from MATLAB® and Simulink® algorithms to control software inter-
faces, optimize execution performance, and minimize memory consumption.
 Generate Code for C2000™ Microcontrollers 23

Figure 2.6 The Embedded Coder® Support Package for Texas Instruments™
C2000™ Processors Simulink® library.

In addition, Embedded Coder® Support Package for Texas Instruments™

C2000™ Processors provides additional support for the Piccolo™ and the
Delfino™ F28x 32-bit microcontrollers that includes:

• Automated build and execution;

• Block libraries for on-chip and on-board peripherals such as ADC, digital
I/O, ePWM, SPI, I2C, and more;
• Real-time parameter tuning and logging using the external mode execution;
• Processor optimized code including DMC and IQMath libraries;
• Ability to perform processor-in-the-loop (PIL) tests with execution profiling;
• Examples for PMSM FOC motor control and DC/DC buck converter.

Embedded Coder® provides additional support packages for the TI proces-

sors, IDEs, and real time operating system (RTOS) listed previously. The fea-
tures vary within each support package and they may include automated build
and execution, processor-optimized code, the ability to perform processor-
in-the-loop (PIL) tests with execution profiling, block libraries for on-chip and
on-board peripherals, and deployment support using bare-board or RTOS. To
learn more about the specific processors and evaluation boards supported
for each MathWorks® release, it is necessary to examine the settings in the
Hardware Implementation section of the Code Generation Configuration Pa-
rameters provided with Embedded Coder® .
24 Automatic Code Generation through MATLAB®

Figure 2.7 Example of a rapid-prototyping on a motor control test bench

based on the LaunchPad™ F28069M Piccolo™ combined with TI BOOSTXL-
DRV8301 and custom extPot3 board (B2B-PMDC hardware kit, see Appendix
B). This hardware is driven through Simulink® .

2.4 TI C2000™ Processors Block-set

The installation of the Embedded Coder® Support Package for Texas
Instruments™ C2000™ Processors results in the availability of a new Simulink®
block-set library in the Library Browser, as shown in Figure 2.6. To assess

the Library Browser, click on the icon in the simulation bar of the

Simulink® window. Since this manuscript focuses on implementations with

the LaunchPad™ F28069M, the block of interest for this book lay in the
C2806x subfamily. This set includes all the functionalities of the main periph-
erals of the MCU in a user-friendly manner, i.e., through a graphical user
interface (GUI). As an example, the C/C++ implementation of a PWM pe-
ripheral is already included in the built-in ePWM block. Figure 2.7 reports
an example of two Simulink schemes that can be used to control a motor
test bench and set the communication with the host pc. Nevertheless, their
use requires an essential know-how on the block configurations, otherwise the
TI C2000™ Processors Block-set 25

coding simplification is useless. Part III shows in detail how to set the most
exploited blocks in power electronics control problems. They can be used to-
gether with the standard Simulink® blocks, generating complete control algo-
rithms and real-time interfaces. An example of the given workflow is denoted in
Figure 2.7.
3
Texas Instruments™ Development Kit

3.1 TI C2000™ LaunchPad™ : F28069M Piccolo

The C2000™ LaunchPad™ F28069M Piccolo™ is a complete low-cost develop-
ment board produced by Texas Instruments™ . This board is shown in Figure
3.1 and it is equipped with a TMS320F28069M 32-bit MCU featuring a 90
MHz C28x CPU (as reported in Figure 3.2), 256 KB Flash, 12-bit ADC, 2
encoder interfaces (eQEP) and 16 PWM pins. It offers an on-board JTAG em-
ulation1 tool which provides direct interface to a PC for easy programming,
debugging and evaluation of the firmware. In addition, the USB interface pro-
vides a UART2 serial connection from MCU to the host PC.
The LaunchPad™ includes also two further synchronous serial communica-
tion protocols called I2 C and SPI. Those protocols are not used for PC-device
communication but rather for communication between single modules and
sensors. In particular, SPI is a synchronous full- duplex serial communica-
tion protocol used for the communications between all the modules mounted
on the LaunchPad™ board. Its registers contains useful information about
the working status of the board as long as many protection parameters (e.g.,
the current protection modes). However, these registers cannot be easily ac-
cessed from the Simulink® environment. For example, the access to nFault
GPIO (which notifies whether the LaunchPad ™ experienced a fault or not)
in Simulink® environment is enabled. Instead, the information on the na-
ture/origin of the fault is stored in different internal SPI registers, which are
not easily accessible through Simulink® blocks.

1 JTAG emulators are the “umbilical cord” between PC software tools and DSP boards.

It is a common hardware interface that provides the computer with a way to communicate
directly with the chips on a board. It was originally developed by a consortium, the Joint
(European) Test Access Group, in the mid-80s to address the increasing difficulty of testing
printed circuit boards (PCBs).
2 UART stands for Universal Asynchronous Receiver/Transmitter. It is not a commu-

nication protocol like SPI and I2 C, but it is a physical circuit in a microcontroller, or a

stand-alone IC. The main purpose of a UART is to transmit and receive serial data.

DOI: 10.1201/9781003196938-3 27
 28 Texas Instruments™ Development Kit

Figure 3.1 LaunchPad™ F28069 Piccolo™ board: main characteristics [12], top
and bottom view.

3.1.1 Features
The LaunchPad™ F28069M Piccolo™ allows applications to easily migrate to
lower cost devices. As a matter of fact, this board shows the features reported
here in the following and in Figure 3.3:

• XDS100v2 On Board Emulator: it is the native JTAG emulator. It enables

JTAG debugging/programming as well as it provides serial communication
back to the PC. The XDS100 can also supply the target MCU through the
USB supply and in Figure 3.3;
• Power Domain: the LaunchPad™ F28069M has different power domains to
enable JTAG isolation. Jumpers JP1, JP2, JP3, JP4, JP5 can be arranged
to reconfigure the power flow on the board as reported in Table 3.1;

Table 3.1 Jumper functions for the LAUNCHXL F28069M [16].

Jumper Power Domain

JP1 Enable 3.3 V from USB (disable isolation)
JP2 Enable GND from USB (disable isolation)
Enable 5 V switcher (powered o˙ 3.3 V supply of
JP3
target device)
Connects target MCU 3.3 V to second set of
JP4
BoosterPack headers
Connects target MCU 5 V to second set of
JP5
BoosterPack headers
 TI C2000™ LaunchPad™ : F28069M Piccolo 29

Figure 3.2 The C2000 MCU uses the C28x core as the main processing unit.
This is a 32-bit floating point (single precision) core with dedicated instruc-
tions tailored to real-time control applications. Complementing the C28x core
is a Control Law Accelerator (CLA) a 32-bit floating point co-processor capa-
ble of independent code execution increasing the system bandwidth versus a
C28x core alone. There are both dual and single core implementations across
the C2000 MCU family of devices [15].

• Electrically Isolated PC Interface: when the LAUNCHXL F28069M is exter-

nally supplied with power through the headers of other expansion boards,
jumpers JP1 and JP2 may be removed to enable electrical isolation of the
LaunchPad™ from the PC. Hence, the board will not be supplied through
the USB cable any more but only by the DC supply connected to the ex-
pansion board, provided that Vs > 5.5 V;
• Connecting a BoosterPack: all the pins mounted on the the LAUNCHXL
F28069M are aligned in a 0.1 in = 2.54 mm grid to allow easy and inex-
pensive plugging of add-on expansion boards called BoosterPacks. These
satellite boards can access all the GPIOs and analog channels;
• Two user LEDs: D9/D10 are two LEDs (blue and red) driven by GPIO34
and GPIO39 channels. Both of them work with negative logic, i.e., GPIO34
HIGH →LED D9 OFF, GPIO34 LOW →LED D9 ON;
• Device reset push button;

• InstaSPIN library in ROM allowing implementation of InstaSPIN-MOTION

and InstaSPIN-FOC solutions;
• Dual 5 V enhanced quadrature encoder (eQEP) interfaces: the eQEP module
is used for direct interface with a linear or rotary incremental encoder to get
position, direction, and speed information from a rotating machine for use
in a high-performance motion and position-control system;
 30 Texas Instruments™ Development Kit

Figure 3.3 Overview of the main components and parts of the LAUNCHXL
F28069M [11].

• CAN Interface with integrated transceiver;

• Boot selection switches, that is, switches whose combination let the user
select the Boot Mode of the board.

3.1.2 Pin muxing

The C2000™ Piccolo™ LAUNCHXL-F28069M has more functions than pins.
Hence, an hardware muxing through jumper selection is used to exploit all pin
functionalities. The pinout of the LAUNCHXL F28069M is reported in Figure
3.4, whereas the GPIO-PIN matching is best explained by the schematics
reported in Figure 3.5. More information can be found by looking at the board
schematics (Figure 3.5 shows the connector pinout) and from the function
summary reported in Figure 3.6.

Figure 3.4 Pinout of LaunchPad™ F28069M [11].

TI C2000™ LaunchPad™ : F28069M Piccolo 31
JP4 JP5 +5V

+3V3
+3V3 +5V
PWR 41 PWR 61

1
2

1
2
ANA 42 GND 62 PWR 1 PWR 21
FAULTn 43 J7.3 Vdc 63 ADCINB7 ANA 2 ADCINA6 GND 22
OCTWn 44 J7.4 V_A2 64 ADCINB4 FAULTn 3 J1.3 Vdc 23 ADCINA7
45 GPIO20 V_B2 65 ADCINA5 OCTWn 4 J1.4 V_A1 24 ADCINB1
ANA 46 V_C2 66 ADCINB5 GPIO 5 GPIO12 V_B1 25 ADCINA2
2SPI_CLK 47 GPIO14 I_A2 67 ADCINA3 ANA 6 ADCINB6 V_C1 26 ADCINB2
48 GPIO21 I_B2 68 ADCINB3 1SPI_CLK 7 GPIO18 I_A1 27 ADCINA0
SCL 49 GPIO23 I_C2 69 ADCINA4 GPIO 8 GPIO22 I_B1 28 ADCINB0
SDA 50 GPIO54 70 9 GPIO33 I_C1 29 ADCINA1
10 GPIO32 30
J5 J7 GND J1 J3
GND
2PWM_AH 80 GPIO6 GND 60 1PWM_AH 40 GPIO0 GND 20
2PWM_AL 79 GPIO7 2SPI_CS 59 GPIO27 1PWM_AL 39 GPIO1 1SPI_CS 19 GPIO19
2PWM_BH 78 GPIO8 GPIO 58 GPIO26 1PWM_BH 38 GPIO2 GPIO 18 GPIO44
2PWM_BL 77 GPIO9 57 1PWM_BL 37 GPIO3 17
2PWM_CH 76 GPIO10 RESET 56 RESET# 1PWM_CH 36 GPIO4 RESET 16 RESET#
2PWM_CL 75 GPIO11 2SPI_SI 55 GPIO24 1PWM_CL 35 GPIO5 1SPI_SI 15 GPIO16
74 2SPI_SO 54 GPIO25 GPIO 34 GPIO13 1SPI_SO 14 GPIO17
73 2EN_GATE 53 GPIO52 33 1EN_GATE 13 GPIO50
PWM_DAC 72 DAC3 2DC_CAL 52 GPIO53 PWM_DAC 32 DAC1 1DC_CAL 12 GPIO51
PWM_DAC 71 DAC4 51 GPIO56 PWM_DAC 31 DAC2 11 GPIO55
J8 J6 J4 J2
GND GND

Figure 3.5 Correspondence between pin and GPIO channels for the
LaunchPad™ F28069M [16].

Figure 3.6 Pin multiplexing on LaunchPad™ F28069M [11].

 32 Texas Instruments™ Development Kit

Figure 3.7 Connection between the LAUNCHXL F28069M and a host PC.

3.1.3 Power connectivity

The F28069M LaunchPad™ has several different power domains to enable
JTAG isolation, see Figure 3.7. Jumpers JP1, JP2, JP3, JP4, and JP5 are
adopted to configure the power flow through the board. It is important to
note that the board is powered through the BoosterPack only if JP1 and JP2
are removed (see Table 3.1). This solution makes the operation with the board
safe. However, safety comes with a couple of severe drawbacks on the Piccolo™
F2806M:
• Resetting the board entails not only the wipe-out of all data collected, but
also of the whole firmware. Further deployment of the firmware is necessary;
• If the power on thr BoosterPack is cut off, the firmware is erased as well.
Thus, this means that the firmware can operate in stand alone operations
without JP1 and JP2 (Section 9 explains how to set the board for such opera-
tion) only if it is constantly supplied; any fault/power interruption may erase
the whole firmware.

Table 3.2 Configurations and functions of jumpers JP6/JP7 [16].

Jumper Function
JP6 ON / USB/UART: GPIO15-58; FAULT/OCTW: GPIO28-29;
JP7 OFF J7.3-J7.4: Hi-Z
JP6 OFF / USB/UART: GPIO15-58; FAULT/OCTW: GPIO28-29;
JP7 OFF J7.3-J7.4: Hi-Z
JP6 OFF / USB/UART: GPIO28-29; J1.3-J1.4 – Hi-Z; J7.3-J7.4 –
JP7 ON GPIO15-58
JP6 ON / USB/UART Disabled; J1.3-J1.4 – GPIO28-29; J7.3-J7.4
JP7 ON – GPIO15-58
TI C2000™ LaunchPad™ : F28069M Piccolo 33

Table 3.3 Boot modes and corresponding switch arrangements [16].

S1-Switch 1 S1-Switch 2 S1-Switch 3

(GPIO34) (GPIO37/TDO) (TRSTn)
Boot Mode
H=Pulled to 1 H=Pulled to 1 H=XDS100v2
L=Pulled to 0 L=Pulled to 0 L=Tied to 0
Emulation Boot L H H
Parallel IO L L L
SCI H L L
Wait L H L
GetMode H H L

3.1.4 Serial connectivity

The F28069M device on this LaunchPad™ contains two serial communica-
tion interface (SCI) peripherals. Because of this, a serial connectivity mux
has been added to the board to make configuration of the SCI routing easy.
Routing is configured via two jumpers (JP6 and JP7). These jumpers should
be configured as shown in Table 3.2 to achieve the desired serial connectivity.
• JP6 ON/JP7 OFF means SCI channel B ON, SCI channel A OFF
• JP6 OFF/JP7 OFF means SCI channel B OFF, SCI channel A OFF
• JP6 OFF/JP7 ON means SCI channel B OFF, SCI channel A ON
• JP6 ON/JP7 ON means SCI channel B ON, SCI channel A ON

3.1.5 Boot options

The LaunchPad™ F28069M includes a boot ROM that performs some basic
start-up checks and it allows the device to boot in many different ways. Most

(a) (b)

Figure 3.8 Details of jumper configuration for communication on channel SCIA

(a), e.g., JP7 is ON (SCI A is ON) and JP6 is OFF (SCI B is OFF), and boot
switches in high state (b).
34 Texas Instruments™ Development Kit

(a) (b)

Figure 3.9 The Motor Drive BoosterPack BOOSTXL-DRV8301 (a) and other
BoosterPack boards stacked one over the other [12].

users may either want to perform an emulation boot or a boot to flash if they
are running the application standalone. The three switches reported in S1 (see
Figures 3.8 (b) and Table 3.3) which allow users to easily configure the pins
that the boot ROM checks before the start-up of the board.
Remark: in digital circuits, a high impedance (or Hi-Z) output is not
driven to any defined logic level by the output circuit. The signal is neither
driven to a logical high nor low level; this third condition leads to the descrip-
tion “tri-stated”. Such a signal can be seen as an open circuit or “floating”.

3.2 TI BOOSTXL-DRV8301 BoosterPack

The Motor Drive BoosterPack BOOSTXL-DRV8301 reported in Figure 3.9 (a)
is a complete drive stage for 3-phase applications based on the DRV8301 three-
phase gate driver and CSD18533Q5A N-channel NexFET™ power MOSFETs,
which allows independent control of each switch through dedicated PWM
signal and 100% duty cycle operation. The BOOSTXL-DRV8301 can be fed
by a power supply in a range from 6 V to 60 V and the board can bear up
to 10 A continuous current and 14 A peak current. This BoosterPack includes
3 low-side current sense amplifiers for current measurement scaled for ±14 A
peak operation. Thus, the integrated amplifiers support bidirectional current
sensing and they provide an adjustable output offset up to 3.3 V. Current
is sensed on the low-side of each leg using a 0.01 Ω current sense resistor.
Individual phase and DC bus voltage dividers are available to measure these
voltages and are scaled for 6 − 60 V operation.
This expansion board requires a compatible LaunchPad™ to receive the
appropriate control signals and a power supply suitable to load needs.
The BOOSTXL-DRV8301 Motor Drive BoosterPack supplies 3.3 V power to
the LaunchPad™ through the onboard 1.5 A step-down buck converter. TI
recommends removing the jumper JP1 on the LaunchPad™ to separate the
TI BOOSTXL-DRV8301 BoosterPack 35

Figure 3.10 TI BOOSTXL-DRV8301 converter mounted on the left-hand side

of the TI LaunchXL F28069M Piccolo™ . On the right-hand side the extPot3
board is installed. The reader is referenced to Appendix B for further infor-
mation on this custom board.

3.3 V power supply from the USB connection (see Table 3.1). Figures 3.9 (b)
and 3.10 show how to plug the converter BoosterPack on the LaunchXL-
F28069M board, i.e., it is connected to pins in rows from J1 to J4 (the
reader shuld refer to Figures 3.9 (b) and 3.5 for the pinout of the adopted
LaunchPad™ ).

3.2.1 BoosterPack PWM signals

The BOOSTXL-DRV8301 can be configured for two different PWM control
modes outlined further in the DRV8301 device datasheet. A possible motor
control example is reported in the block scheme shown in Figure 3.11. These
control modes give PWM control through 3 or 6 independent inputs. Six-
input PWM mode gives independent control of the low-side and high-side
gates for each converter leg. Three-input PWM mode gives control only of the
3 high side gates of the legs, while low-side switches are driven by an internally
generated complementary signal with minimum internal dead time.
PWM modules on the three phase legs (which will be denoted from now on
as Leg A, B, C) can be accessed by using the ePWM block in Simulink® and
by selecting module 1, 2, or 3. Further details on the management of PWM
signals will be discussed in Chapter 13.

Figure 3.11 Block scheme of the interactions between a LaunchPad™ and the
BOOSTXL-DRV8301 aimed to a motor control application [14].
36 Texas Instruments™ Development Kit

Figure 3.12 Scheme showing the possible interactions between a LaunchPad™

and the BOOSTXL-DRV8301 [14].

3.2.2 BoosterPack GPIO signals

This board exploits the interface with the input/output peripherals of the
LaunchPad™ to output diagnostic signals or to set some enables, see Figure
3.12 and 3.13. Namely, these signals are:
1. nFAULT – Fault report indicator. Specific fault status can be ob-
tained through the status registers
2. nOCTW – Over-current, over-temperature, or both warning indi-
cator. Specific OCTW status can obtained through the SPI status
registers
3. EN_GATE – Enables gate driver and current shunt amplifiers of
the converter board. The corresponding pin on the LaunchPad™
must be set high to use the expansion board
4. DC_CAL – When high, the device shorts inputs of shunt

Figure 3.13 Top and bottom view of the BOOSTXL-DRV8301 BoosterPack

[22].
TI BOOSTXL-DRV8301 BoosterPack 37

Figure 3.14 Pinout of the BOOSTXL-DRV8301 converter board [22].

amplifiers and disconnects possible loads so that the LaunchPad™

can calibrate DC offsets
In addition, the BOOSTXL-DRV8301 nFAULT and nOCTW signals share
the GPIO channels with the UART-RX and UART-TX signals on the
LaunchPad™ . To see the proper nFAULT and nOCTW signals reporting on
the onboard LEDs and MCU peripheral, disconnect the UART lines from the
emulation side of the LaunchPad™ to the microcontroller. To this aim, refer to
Table 3.2 to set the LaunchPad™ F28069M accordingly. Software Enable:
it must be noted that the pin of the enable gate (EN-GATE, see Figure 3.14)
of this expansion board must be set high to perform any operation with the
converter. This is carried out by setting the voltage on this pin to the logic
state 1 through a suitable peripheral.

3.2.3 DC bus and phase voltage sense

The BOOSTXL-DRV8301 is designed with voltage sense circuits on the DC
bus (PVDD) and each half-bridge outputs (phases A, B, and C). These circuits
consist of a voltage divider with a filtering capacitor to reduce high-frequency
noise on the ADC pins. These circuits are scaled for 26.314 V, with a filter
pole location at 364.692 Hz. The high-side resistors for the phase outputs are
located near the output header J11, whereas the low side resistors and filtering
capacitors are located near the ADC inputs J3 (for improved noise reduction).

3.2.4 Low-side shunt-based current sense

Low side current shunt resistors are installed on each leg of the converter (i.e.,
on phase A, B, and C). The current sense setup takes advantage of the dual
38 Texas Instruments™ Development Kit

shunt current amplifiers (phases A and B) of DRV8301 driver and an exter-

nal TI OPA2374 CMOS operation amplifier (phase C). Each amplifier senses
voltage across a 0.01 Ω sense resistor through a differential connection. The
current measurement is scaled on ±16.5 A. Further details on low-side current
shunt sense are reported in Chapter 16.

Remark: all the examples proposed in this book refer to the hardware pre-
sented in this chapter. It must be said that Texas Instruments™ started
the production of a new LaunchPad™ F280049C and new BoosterPacks
BOOSTXL-DRV8320RS, BOOSTXL-DRV8323RS, which will be available
along with the LaunchPad™ F28069M and BOOSTXL-DRV8301 on the mar-
ket. Even if a new hardware will imply differences in the settings of the pro-
gramming environment as well as some modifications in the pinout of the
adopted boards, the proposed design approach/workflow of closed-loop con-
trol strategies for power electronic applications is still valid. The reader is ref-
erenced to [7] for checking the compatibility between LaunchPads and Boost-
erPacks and [31] to verify the Embedded Coder support of the adopted MCU
board from Simulink® blockset.
4
Software Installation

To the aim of programming the microcontroller mounted on the LAUNCHXL

F28069M, different software packages need to be installed. The latter are
available on the MATLAB® Add-Ons menu and on the website of Texas
Instruments™ [10]. Namely, they are:
1. Texas Instruments™ Code Composer™ Studio.
2. Texas Instruments™ ControlSUITE™ .
3. Embedded Coder® Support Package for Texas Instruments™
C2000™ Processors.
It is important to verify the installation of MinGW-64 in MATLAB® .

4.1 TI Support Packages: Code Composer™ Studio and

ControlSUITE™
An embedded engineer must answer many critical questions when select-
ing a new MCU platform, such as those reported in Figure 4.1. Texas

Figure 4.1 MCU platform selection criteria [22].

DOI: 10.1201/9781003196938-4 39
40 Software Installation

(a) (b)

Figure 4.2 Texas Instruments™ Code Composer™ Studio (a) and

ControlSUITE™ (b).

Instruments™ provides two different software aimed at MCU programming:

Code Composer™ Studio and ControlSUITE™ (see Figure 4.2).
Code Composer™ Studio is an IDE that supports Microcontroller and
Embedded Processors portfolio of TI. Code Composer™ Studio comprises a
suite of tools used to develop and debug embedded applications. It includes
an optimizing C/C++ compiler, source code editor, project build environ-
ment, debugger, profiler, and many other features. The intuitive IDE provides
a single user interface bringing the users through each step of the applica-
tion development flow. Familiar tools and interfaces allow users to get started
faster than ever before. Code Composer™ Studio combines the advantages of
the Eclipse software framework with advanced embedded debug capabilities,
resulting in a compelling feature-rich development environment for embedded
developers.
On the other hand, ControlSUITE™ for C2000™ microcontrollers is a
cohesive set of software infrastructure, tools, and documentations designed
to minimize system development time, as shown in Figure 4.3. From device-
specific drivers and support software to complete examples in sophisticated
system applications, ControlSUITE™ provides the key resources to make a fea-
sible automatic code generation procedure. Indeed, it contains all the needed

Figure 4.3 ControlSUITE™ overview: device drivers, APIs, utilities and li-
braries are used to build technology examples and system frameworks [22].
 MATLAB Support Packages 41

Figure 4.4 Embedded Coder Support Package for Texas Instruments C2000
Processors available in the Add-Ons menu of Matlab® .

settings for the MATLAB® environment (e.g. Simulink® blocks) to character-

ize the peripherals of the MCU.

4.2 MATLAB® Support Package: Embedded Coder for

Texas Instruments C2000 Processors
The Embedded Coder Support Package for Texas Instruments
C2000 Processors enables the generation of a real-time executable code
and its download to MCU boards (from 8 bit to multi-cores) through the
Simulink® environment. Such package automatically generates C code (back-
ground usage of Code Composer™ Studio) and it makes available a Simulink®
library with built-in blocks. Different blocks allows to set different I/O drivers
of the device (background usage of ControlSUITE™ ).

4.3 Installation Procedure

All the previously mentioned softwares can be installed starting from the
MATLAB® Add-Ons window following the guided procedure shown below
(reference to MATLAB® release 2019b):

• Click on the lower part of the Add-Ons icon in the home bar of

the Matlab® window and select Get Hardware Support Package.

• Search for Embedded Coder Support Package for Texas Instru-
ments C2000 Processors (see Figure 4.4) and click on it.
 42 Software Installation

• Check whether the required packages are already installed

or not. In case any of them is missing, install the required software by

clicking on its corresponding name. A guided installation procedure starts
after clicking on the Install icon .
• Once all the required packages are properly installed, click on the Install
icon back in the window of the Embedded Coder Support
Package for Texas Instruments C2000 Processors Add-On and select
install.
• Accept the license agreement. After that, click the Next button to
proceed with the installation of the following Third-Party Softwares:

– TI C2000 Code Generation Tools;

– Ostermiller Circular Buffer.

Their download and installation should start immediately.

• After that, click on the Setup Now button to go on with the
guided procedure.
• The window Select Processor Family opens. The user is asked to select
the processor family that is going to be used from a list (see Figure 4.5 (a)).
Only the TI Piccolo F2806x is needed for the solution of the exercises
proposed in this book. After this selection, click on the Next button.
• The window Install Third-Party Software opens as shown in Figure 4.5
(b). Depending on the previous selection, the user may be required to install
different softwares. In particular, the TI Piccolo F2806x family requires TI
controlSUITE and TI Code Composer Studio only. TI C2000Ware is
needed in case other boards are selected. Click on the corresponding cell in
the column Status to install their latest version.

– Texas Instruments™ ControlSUITE™ : the latest available version

of this software is automatically suggested by the guided installation
procedure. In particular, it is required to specify the ControlSUITE™
installation folder. The default path is: C:\ti\controlSUITE\. More-
over, all the needed libraries are installed by default.
– Texas Instruments™ Code Composer™ Studio: the latest avail-
able version of this software is automatically suggested by the guided
procedure. In particular, it is required to:
∗ Specify the Code Composer™ Studio installation folder. Default
path: C:\ti\ccsvx\, where x refers to the release of the software;
∗ Specify the product families to be installed:
 Installation Procedure 43

· Select C2000™ 32bit Real-time MCUs, compiler tools

1. TI C28x Compiler;
2. TI ARM Compiler;
3. GCC ARM Compiler;
∗ leave the default debug probe: TI™ XDS Debug Probe Sup-
port;
∗ Allow access for Tclsh Application and ccstudio from the settings
of Windows Defender Firewall.
It is important to note that warning messages could appear depending
on the installed Windows 10 release. Ignore them. Install any other
additional driver/software if requested during the installation process.

• The window Validate Control Suite Installation opens. In this phase,

the installation of the previously mentioned softwares is checked inserting
their installation path:

– ControlSUITE™ installation folder. The default path is:

C:\ti\controlSUITE\
– Code Composer™ Studio installation folder. The default path is:
C:\ti\ccsvx, where x refers to the installed release.
∗ Click the Validate button to check if the version of Code
Composer™ Studio is compatible with the installed MATLAB®
release.
– Specify the C2000™ Compiler installation folder. The default path is:
C:\ProgramData\MATLAB\SupportPackages\R2019b\3P.instrset\
tic2000codegentools.instrset\ti-cgt-c2000_xx.x.x.LTS, where all the x
depend on the installed release.
This compiler is downloaded during the installation of the support
package.

• At the end of the validation procedure, the Summary of third-party

installation opens. A report table showing the installed softwares appears.
• Finally, the Hardware Setup Complete window opens. The guided pro-
cedure is finished.
• In case any modification of this installation is needed, it is possible to click on
the manage button on the window of this add-on (it replaces
the Installation button). A window showing all the installed Add-Ons opens
and the guided procedure of any of them can be repeated by clicking on the
setup icon

A comprehensive summary of all these installation steps is available in [28].

44 Software Installation

(a) (b)

Figure 4.5 Example of MATLAB® windows that appear during the installa-
tion: (a) shows the processor family selection windows (only the required TI
Piccolo F2806x is selected), while (b) reports the needed third party softwares.
In this latter, all of them have been installed already.

Remark
Check if the Instrument Control Toolbox™ is installed on the adopted
MATLAB® release. This toolbox is needed to set Serial Communication be-
tween the LaunchPad™ and the host PC during the execution of the firmware.
More information on serial communication are reported in Chapter 8.

Potential Errors
For MATLAB® releases earlier than the 2019b (like the 2015b one) in
which Code Composer™ Studio v5 is used, it may be required to run
checkEnvSetup() tool to complete/adjust the installation:
To this aim, the following path should be checked by typing checkEnvSetup
(’ccsv5’,’f28069’,’setup’) in the MATLAB® prompt:
Checking CCSv5 (Code Composer Studio) version
Required version: 5.0 or later
• Required for : Code Generation
Your Version : 5.5.0
### Setting environment variable “TI_DIR” to “C:\ti\ccsv5”

Checking CGT (Texas Instruments C2000 Code Generation Tools) version

Required version: 5.2.1 to 6.0.2
Required for : Code generation
–
Your Version : 6.2.0
### Setting environment variable “C2000_CGT_INSTALLDIR” to
“C:\ti\ccsv5\tools\compiler\c2000_6.2.0”
Checking DSP/BIOS (Real Time Operating System) version
Required version: 5.33.05 to 5.41.11.38
Required for : Code Generation
–
Your Version : 5.42.01.09
### Setting environment variable “CCSV5_DSPBIOS_INSTALLDIR” to
“C:\ti\bios_5_42_01_09”
Installation Procedure 45
Checking XDC Tools (eXpress DSP Components) version
Required version: 3.16.02.32 or later
– Required for : Code Generation
√
Your Version : NOT FOUND √
The required version of XDC Tools is not found in $(XDCROOT)
Checking 2806x C/C++ Header Files version
Required version: 1.36
Required for : Code Generation
–
Your Version : 1.36
### Setting environment variable “DSP2806x_INSTALLDIR” to
“C:\ti\controlSUITE\device_support\f2806x\v136”
Checking Flash Tools (TMS320F2806x Piccolo Flash API) version
Required version: 1.00
Required for : Flash Programming
–
Your Version : 1.00
### Setting environment variable “FLASH_2806X_API_INSTALLDIR” to
“C:\ti\controlSUITE\libs\utilities\flash_api\2806x\v100”

It is important to note that this procedure could be also used to check the
correct installation of Code Composer Studio™ for other Matlab® releases up
to the 2018b one. For newer versions, this function has been removed.
 Part II

Review of Control Theory:

Closing the Loop
Introduction

The process of designing a control system usually makes many demands of

the engineer or engineering team. In general, they can be summarized in the
following step by step procedure:
1. Study of the system to be controlled and definition of the required
control objectives.
2. Modeling of the system and simplification of the model.
3. Analysis/determination of the system properties (open-loop char-
acteristics) and, if necessary, scaling of the variables.
4. Selection of the variables to be controlled (e.g., control input of a
modulator).
5. Identification of the required measurements and manipulated vari-
ables: what sensors and actuators should be used and where they
should be placed in the real system.
6. Choice of the control configuration (e.g., power electronics systems
are typically based on nested control loops).
7. Selection of the controller type to be used (e.g., PI, PID controllers).
8. Setting of the performance specifications based on the overall con-
trol objectives (closed-loop characteristics).
9. Design of the controller parameters (e.g., pole/zero cancellation in
case a PI controller and a first-order system are considered).
10. Analysis of the resulting closed-loop controlled system to see if the
specifications are satisfied. If not, modifications in the specifications
or in the controller type are needed.
11. Simulation of the closed-loop controlled system (e.g., MATLAB® -
or Simulink® -based simulations).
12. Choice of the adopted software and implementation of the regulator
on an embedded system (e.g., a microcontroller).
13. Test and validation of the resulting real-time control driving a pi-
lot/prototype of the system.
Control courses and textbooks usually focus on steps 9 and 10 in the above
procedure, that is, on controller design and control system analysis methods.

DOI: 10.1201/9781003196938-4 49
50 Introduction

Interestingly, many real control systems are designed without any considera-
tion of these two points. For example, even for complex systems with many
inputs and outputs, it may be possible to design workable control systems
using only heuristic tuning (i.e., involving steps 1, 4 5, 6, 7, 12, and 13). How-
ever, in this case a suitable control structure may not be known at the outset,
and there is a need for systematic tools and insights to assist the designer at
least with steps 4, 5, and 6, based on step 2 and 3. This book aims to provide
guidelines for steps 12 and 13, for which the control structure design has to
deal with microcontroller implementations.
Simply stated, “even the best control system cannot make a Ferrari out of
a Volkswagen.” Therefore, the process of control system design should in some
cases also include a step 0, which involves the design of the system equipment
itself. Regarding the purpose of the book, this means to deal with the power
electronics design in term of sizing passive components, consider magnetic and
saturation effects, evaluate power losses and heat exchange and many more
aspects.
Both control and power electronics engineers have to remember that the
idea of looking at hardware/system equipment and control system design as
an integrated whole is not a novel approach, as it is clear from the following
quote taken from [44]:

In the application of automatic controllers, it is important to realize

that controller and system to be controlled form a unit; credit or discredit
for results obtained are attributable to one as much as the other. A poor
controller is often able to perform acceptably on a system which is easily
controlled. The finest controller made, when applied to a miserably de-
signed system, may not deliver the desired performance. True, on badly
designed systems, advanced controllers are able to eke out better results
than older models, but on these systems, there is a definite end point which
can be approached by instrumentation and it falls short of perfection.

The aim of this chapter is to show how to design a development-oriented

PI controller for microcontroller applications. Starting from a continuous-
time model based on the physics of the system, the most important design
guidelines of a PI-based closed-loop control system are revised. In particular,
much attention is paid to how to guarantee certain steady-state and dynamic
performances, how discretize the controller, and how to cope with real-time
computation bottlenecks on physical device. Due to their digital nature, all
physical devices are designed in order to operate exclusively with discrete-
time variables. Thus, it is necessary to identify a procedure which transform
the continuous-time control problem into an equivalent discrete-time one. The
book uses simple examples to clarify the transition from “theory” to “prac-
tice”, without claiming to be exhaustive with regard to the whole detailed
aspects of control theory.
 5
Designing a Closed-Loop Control System

The objective of a closed-loop control system is to make the output y(t) behave
in a desired way by manipulating the input u(t). This “desired way” is defined
by a proper formulation of the control problem, which could refer to:
• Command-tracking: u(t) is manipulated to keep the output close to a
given reference input (set point) y ∗ (t).
• Disturbance rejection: u(t) is manipulated to counteract the effect of
disturbances;
The evolution in time of u(t) is given by the controller type and its design,
which relates to the control target. Note that the controller should be selected
according to the system to be controlled. Starting from the physics of the cir-
cuits models, i.e., based on equivalent electrical ports, the following sections
denote that both electrical machines and power conversion systems may be
described as linear or linearized models. Hence, linear controllers represent
the easiest control candidate. Furthermore, the mathematical description of
such electrical dynamics refers to multi-input multi-output (MIMO) models.
Nevertheless, this manuscript shows how the control configuration/architec-
ture may lead to translate a MIMO model into a combination of single-input
single-output (SISO) models. This is possible by taking into accounts feed-
forward terms. For SISO linear models (like those treated in this book), a
proportional–integral–derivative controller (PID) results in control loop feed-
back mechanisms widely used in industrial power electronics-based systems
and a variety of other applications requiring continuously modulated control.
A PID controller continuously calculates an error value e(t) as the difference
between the set point y ∗ (t) and the measured variable y(t). Based on this
evaluation, the controller applies a correction based on proportional, integral,
and derivative actions. Note that, PID controllers are present in the industrial
fields since the early 1920s (automatic steering systems for ships).
In general, any closed-loop control scheme for SISO models can be repre-
sented by the block diagram shown in Figure 5.1.

DOI: 10.1201/9781003196938-5 51
 52 Designing a Closed-Loop Control System

Reference Actuation Output

y  (t) e (t) u (t) y (t)

R (s) G (s)
+
−

Measurement

Figure 5.1 Example of closed-loop control system.

5.1 Dynamical Systems

From control theory, the dynamical model of a system is a set of mathematical
laws explaining in a compact form (i.e., it must be simple enough) and in a
quantitative way (i.e., it must be descriptive enough) how the system evolves
over time. The latter is fully described by the evolution of the state x(t). It is
worth noting that the main questions about dynamical models are:

• How to built the model?

(“How x(t), u(t) and y(t) influence each other?”)
• How to use such model to design a control system?
(“How to make the system behave the way I want?”)
• How to make simulations?
(“Which tool is more suitable for a given application?”)

Note that simulating a dynamical model has zero-cost compared to real

experiment and it allows to get rid of many practical complexities of the real
system. In addition, many simulations allows to reduce the order of the system
or to focus just on some parts of the whole, allowing more agile and easy runs.
To this purpose, it is important to take in mind the well-known aphorism:

“Make everything as simple as possible, but not simpler.”

“Make everything as simple as possible, butAlbert Einstein

5.1.1 Mathematical laws

Considering a continuous-time domain t ∈ R, the behavior of the state x(t)
is described by ordinary differential equations (ODEs). This book refers to
applications whose input-output electrical responses are governed by linear
ODEs with constant coefficients and, thus, linear dynamics. The continuous-
time state-space model can be formalized as:
Dynamical Systems 53

dx(t)
= Ax(t) + Bu(t) (5.1)
dt
y(t) = Cx(t) + Du(t) (5.2)

where A, B, C, and D are (time-invariant) system matrices, which sizes results

after choosing the system variables. Every time the output is entirely matching
the state x(t) via C, the term Du(t) is removed from (5.2). Given the initial
conditions x(0) = x0 ∈ Rn , the evolution of the state described in (5.1) is
equivalent to
Z t
x(t) = eAt x0 + eA(t−τ ) Bu(τ ) dτ. (5.3)
| {z } 0
natural response | {z }
(effect of x(0) ) forced response
(effect of u(t) )

Moreover, the system is asymptotically stable on condition that A < 0; this

is a key aspect.
This manuscript makes extensive use of transfer functions and of the
frequency domain, which are very useful because from simple frequency-
dependent plots it is possible to derive useful insights such as bandwidth
and peaks of closed-loop transfer functions. In particular, Laplace transforms
convert integral and differential equations into algebraic equations through
the Laplace operator s. A general time-dependent function f (t) is converted
as:
d
F (s) = L [f (t)] s= = jω (5.4)
dt
In addition, it can be demonstrated that
h i
−1 −1
y(s) = C (sI − A) x0 + C (sI − A) B + D u(s) (5.5)
| {z } | {z }
Laplace transform
of the natural response
Laplace transform
of the forced response

where I denotes the identity matrix. Assuming x0 = 0 and D = 0, the transfer

function G(s) of a continuous-time linear system defined as in (5.1) and (5.2)
is:

y(t) y(s)
G(s) = = = C(sI − A)−1 B. (5.6)
u(t) u(s)
Namely, G(s) is defined through the ratio between the Laplace transform
of the output y(s) and the Laplace transform of the input u(s). Note that,
the mapping t ← s holds. Thus, y(t)/u(t) = y(s)/u(s). This equality will
be considered in the definition of other transfer functions in the following.
Equation (5.6) directly links to Figure 5.1.
 54 Designing a Closed-Loop Control System

5.1.2 Dynamical systems in electrical applications

In general, power conversion systems and electrical machines behave as:
• RL (ohmic-inductive) or RC (ohmic-capacitive) load.
• RLC or RLE (including the motor back-electromotive force or higher-order
filter) load.
Thus, G(s) results in a 1st or a 2nd order transfer functions in the vast majority
of the cases.
A 1st order G(s) is now considered. It can be defined as:
1 1/a 1/a 1/ (a τG )
G(s) = = = = (5.7)
a + sb 1 + s (b/a) 1 + sτG s − (−1/τG )
where a and b are constant (time-invariant) coefficients. The denominator of
G(s) can be factorized such that it has the form (s − p), where p defines a
pole, e.g., p = −1/τG . When s = p, the value of G(s) becomes unbounded,
i.e., if 1 + sτG = 0 is set, it follows s = p = −1/τG , which characterizes the
natural response of the system. Moreover, if Re(p) < 0 the system is stable.
The timing required by the system output to react against transient is a
key factor either for uncontrolled, e.g., G(s), and controlled systems, e.g., L(s),
F (s) (defined later). Since different ways to define time quantities are present
in literature, it is important to specify that this book adopts the following
definition:
given a step-wise input with steady-state value Yss , the settling time Ta
is defined as the time interval required to enter in a specific tolerance
bandwidth without further violation. Namely, the steady-state value Yss
is reached with a certain degree of accuracy given by ±, i.e., entering
in a region [(1 − 0.01)Yss (1 + 0.01)Yss ]. In practice, Ta = 4 ÷ 5τ holds
(counting from the instance in which a transient happens) with  ≈ 1%,
without loss of generality.
Note that, the time constant τ is specified by the transfer function parame-
ters.1 As an example, referring to G(s), the natural settling time is defined
as Ta,G and it follows that Ta,G = 5τG , where the time constant is τG = b/a.
Note that subscripts refer to the settling time of a specific transfer function.

5.2 Design a PI Controller in Continuous-Time Domain

In power electronics-based systems, PI controllers are more common than
PIDs since the derivative action is sensitive to measurement noise (e.g., high
1 Be careful, τ must not be confused with the rise time T which is defined as the time
r
interval between the time instances related to 10% Yss and 90% Yss .
Design a PI Controller in Continuous-Time Domain 55

frequency noise entering in the system is highly amplified). However, if sudden

variations in the reference input occur, the derivative part may help to achieve
fast transients, e.g., current reference for high speed motors. Nevertheless, the
integral action could lead to wind-up effects both for PI and PID solutions,
which have to be taken into account.

5.2.1 Serial/parallel form

Using the Laplace operator, the PI controller transfer function R(s) (which is
equivalent to u(t)/e(t)) could be written both in series and parallel form:

u(s) kp (1 + sTi ) ki
R(s) = = (series) or R(s) = kp + (parallel)
e(s) sTi s
(5.8)
where kp , ki are the proportional and integral gains, respectively, while Ti
is the integration time defined as Ti = kp /ki .

5.2.2 Characterization of the closed-loop dynamics F(s)

In this subsection, an easy analytical procedure to identify the controller pa-
rameters starting from the knowledge of the natural response of G(s) and
the desired requirements for the final closed loop-system is recalled. Those
requirements are usually given in terms of:

1. Zero steady-state error (static performance)

2. Time response to transients (dynamic performances),
i.e., control bandwidth ωc
3. Robustness level against disturbances (dynamic performances),
i.e., phase margin φm
Once R(s) and G(s) are known, the open- L(s) and closed-loop F (s) transfer
functions are defined as:

y(s) y(s) L(s)

L(s) = = R(s)G(s) F (s) = = (5.9)
e(s) y (s)
∗ 1 + L(s)
Note that the definition e(s) = y ∗ (s) − y (s) holds. If G(s) shows a positive
gain and it has Re(p) < 0, it is called minimum phase system (mps). Then,
it is enough that |L(s)| cuts the 0 dB axis with a slope of −20 dB/decade to
guarantee φm ∼ = 90◦ and asymptotic stability. This means that the cut-off
frequency is imposed at ωc (i.e., at s = jωc ):
h i
|L(s)|s=jωc , 0 dB → |L(s)|s=jωc , 1. (5.10)
dB

Note that the phase margin is computed as φm = π − arg L(jωc ) = π − φc .

56 Designing a Closed-Loop Control System

Figure 5.2 Example of L (s) obtained through pole/zero cancellation and its
properties detailed through a Bode diagram.

Pole/Zero Cancellation
By considering (5.7), the open-loop transfer function can be computed as:

kp (1 + sTi ) 1/a
L(s) = (5.11)
sTi 1 + s (b/a)
If Ti = τG = b/a is imposed, the so called pole/zero cancellation is performed,
which is rather typical for first-order systems:

kp /a
L(s) = (5.12)
sTi
This transfer function behaves as a pure integrator as shown in Figure 5.2,
thus φm ∼
= 90°, and it has a cut-off frequency which is defined by the controller
parameters

(if Ti = b/a)
  
 kp /a  kp kp /b
,1 → ωc = = (5.13)
(if Ti = kp /ki )
 
 sTi 
s=jωc aTi ki /a

Note that |x + jy| = x2 + y 2 holds.

p

In practice, L(s) and F (s) are subjected to the same time constant, which
can be denoted as τF . Having an explicit form, kp and ki can be computed
as functions of ωc . For instance, by considering ωc = kp /b → kp = ωc b from
(5.13), kp may be replaced in the expression of Ti = kp /ki = ωc b/ki , which is
equivalent to Ti = b/a = ωc b/ki , which leads to ki = ωc a. Finally:

k p = ωc b ki = ωc a (5.14)
How to use the explicit forms: assume that y(t) is required to follow a
step-wise reference y ∗ (t) within a specific settling time Ta,F . For instance, the
Design a PI Controller in Continuous-Time Domain 57

requirement might be Ta,F  Ta,G , i.e., to speed up the open loop dynam-
ics by adopting a (closed) control loop. Since the final (closed-loop) settling
time is defined as Ta,F = 5τF , τF = Ta,F /5 can be computed, which repre-
sents the dominant time constant (i.e., associated to the dominant pole) of
F (s). Therefore, by choosing the cut off frequency ωc as ωc = 1/τF in rad/s,
and considering the other time-invariant parameters, all the ingredients are
available to derive kp , ki by substitution.

Generalized approach
The pole/zero cancellation is particularly suitable for 1st order linear systems.
If the system is higher-order, a different procedure should be adopted. In
general, the following formulas can be exploited:

cos(φm − π − arg(G(jωc )))

kp = AA (5.15)
|G(jωc )|
sin(φm − π − arg(G(jωc )))
ki = −ωc (5.16)
|G(jωc )|

where each parameter is function of the cut-off frequency ωc and the phase
margin φm . Of course, by using the same hypothesis on Ti for the zero pole/-
cancellation, the same results shown in (5.14) can be derived. Note that
much flexibility on the performance of the system is achieved compared to
the pole/zero tuning

Mathematical Proof

A generalized open loop transfer function cuts the 0 dB axis at ωc with phase
margin φm . In this point, its frequency response can be evaluated as:

L(jωc ) = R(jωc )G(jωc ) = 1ej(φm −π) (5.17)

Thus, the PI controller can be evaluated in the same frequency point as well:
ki ki
R(jωc ) = kp + = kp − j (5.18)
jωc ωc
Instead, a generalized G(s) is considered:

G(jωc ) = |G(jωc )| earg(G(jωc )) (5.19)

Then, (5.17) is exploited to derive the following expression:
ki 1
kp − j = ej(φm −π−arg(G(jωc ))) (5.20)
ωc |G(jωc )|
Recalling the Euler formula ejx = cos x + j sin x, it follows:
58 Designing a Closed-Loop Control System

ki 1
kp − j = [ cos (φm − π − arg (G(jωc ))) + ...
ωc |G(jωc )|
ki 1
kp − j = ... + jsin (φm − π − arg (G(jωc ))) ] (5.21)
ωc |G(jωc )|
By computing the real and imaginary parts of the previous equation, kp and ki
are obtained as reported earlier. All the angles must be considered in radians.

cos(φm − π − arg(G(jωc )))

 
ki
kp = Re kp − j = AA (5.22)
ωc |G(jωc )|
sin(φm − π − arg(G(jωc )))
 
ki
ki = −ωc Im kp − j = −ωc (5.23)
ωc |G(jωc )|

p
MATLAB® interface: pidTuner() and pidtune()
A third alternative for PI tuning is to use control design tools from Simulink® ,
e.g., pidTuner() and pidtune() from Control System Toolbox™ . In partic-
ular, pidTuner() allows to work with a user friendly tool (with a graphic
user interface as well) for simple step-response-based identification of a pro-
cess model, fast PID controller tuning and effective quality evaluation of the
control in the MATLAB® framework.
Regarding the control design, either time-domain or frequency-domain
approaches can be chosen. As an example, the MATLAB® commands
pidTuner(sys) or pidTuner(sys,PI) launch the user interface which allows
to design a desired controller for the system sys under investigation, which
in our case relates to G(s). The properties of the tuned controller can be sim-
ply and visually judged by the PID Tuner window using a simulation of the
closed-loop step response. Indeed, this tool displays the plot of the simulated
control response and the time behavior of the manipulated variable. It is easy
to compare several step responses generated with different values of kp and
ki , set points, disturbances and constraints on the manipulated variable.

Hint: By defining the Laplace operator as s=tf(’s’) and the parameters

comprised in G(s) (e.g., R=2.5e-3; L=1e-3), the system transfer function
can be defined simply as G=1/(R+s*L) in MATLAB® through the command
window and standard syntax. Then, it is enough to call pidTuner(G) to open
the interface reported in Figure 5.3.
The key commands available in this window allow to:
1. Define the system to be controlled G(s);
2. Select the type and form of the controller (e.g., PI Parallel);
Design a PI Controller in Continuous-Time Domain 59

Figure 5.3 pidTuner(G) interface window.

3. Select the design domain, i.e., frequency or time;

4. Adjust the tuning parameters (e.g., bandwidth and phase margin
for the frequency domain) and check the resulting response to a
unitary reference step;
5. Copy or export the resulting kp and ki coefficient.
Alternatively, the user can automatically obtain the required parameters us-
ing the command pidtune in a script or in the command window, which runs
the PID Tune software in background. The complete syntax is

pidtune(sys,’PI’,bandwidth,opt);

with

opt=pidtuneOptions(’PhaseMargin’,phase_margin);

The result is a structure called pid which contains the kp and ki coefficients
as field of this new variable type.
Note that booth approaches use the generalized formulas presented in the
previous section. In particular, the pidTuner() interface use pidtune() as
sub-function to compute the controller parameters.
60 Designing a Closed-Loop Control System

For a complete description of all the features implemented in PID Tuner,

the reader is referred to its documentation on the MathWorks® Website [29].

5.3 Derive a PI Controller in Discrete-Time Domain

Once the design of the controller is performed in continuous time, it is neces-
sary to define an equivalent expression that shows a similar behavior during
the real-time execution of the control algorithm. To do so, a functional equiva-
lence between the Laplace operator s for the continuous time and the complex
variable z for discrete time series must be defined. This operation goes under
the name of “root mapping” and it corresponds to the approximation of the
integral in continuous-time domain (expressed by the Laplace operator as 1/s)
with a suitable numerical integration method in discrete time, i.e., by means
of a finite set of points/samples.

5.3.1 General properties of the discretization process

The numerical integration can be carried out directly on the transfer function.
Since the starting point is in the s-plane, a continuous-time transfer function
G(s) can be mapped to the z-domain by substituting the Laplace operator as
follows:

z−1
 
G(z) = G s = (5.24)
Ts (αz + 1 − α)
where Ts is the sampling time and 0 ≤ α ≤ 1 defines the type of discretization.
Indeed, equation (5.24) refers to a generalization of a bilinear transformation
which could result in:

• Forward Euler integration α=0 → s= z−1

Ts
• Backward Euler integration α=1 → s= z−1
zTs
2 (z−1)
• Trapezoid integration α = 12 → s = Ts (z+1)

The trapezoid rule is also called Tustin’s method in digital control commu-
nity. The continuous-time transfer function G(s) shows a stable behavior when
the poles have negative real part. In discrete-time, this translates in a G(z)
with poles inside the unitary circle. The coefficient α influences such stability
because a different integration method results in a different mapping, with
different boundaries for the pole/zero locations. The three mappings are sum-
marized in Figure 5.4. However, it is possible to find a large number of other
 Derive a PI Controller in Discrete-Time Domain 61
Forward Euler Backward Euler Tustin
Im Im Im
s z

s z s z

1 Re 1 Re 1 Re

Figure 5.4 Stability regions for different discretization methods.

integration methods with different properties in terms of convergence and ac-

curacy in literature. The three methods reported in this section represent the
most common choice and they refer to single step methods in which the output
depends only on the input at the same time-step and on input/output at the
previous time instant.
Hence, they require a limited computational effort, targeting most of the
microcontroller-based platforms.2 However, keep in mind that:

• Forward Euler - if the discrete form is stable, also the continuous one is
stable too, but not vice versa;
• Backward Euler - if the continuous form is stable, also the discrete one is
stable too, but not vice versa;
• Trapezoid if the continuous form is stable, also the discrete one is stable too,
and vice-versa.

Note that the convergence region depends both on the integration method
itself and on the considered sampling time width as well. High Ts enlarges the
mapping areas. This might create a great amount of mapping distortion, even
leading to instability problem if the unitary circle is violated.
Another important property of the z-domain lies in the fact that the quan-
tity z −1 is equivalent to a unitary delay in the discrete time domain. This cor-
respondence allows to calculate the output of a generic regulator as function
of the inputs/output at the previous time steps.
In particular, Figure 5.5 shows how to interpret the described Back-
ward Euler integration method, where k denotes the discrete-time steps. This
method is adopted in the following sections.
For an exhaustive explanation of the z-transform theory, the reader is
referred to [2].
2 In multi-step methods, several points are used to evaluate the output at each time

step. These ones are characterized by a higher accuracy as the number of steps increases,
but they imply a big number of variables to be managed by the microcontroller, i.e., high
computational burden.
 62 Designing a Closed-Loop Control System

Figure 5.5 Time representation of the Backward Euler integration method.

5.3.2 Characterization of the closed-loop dynamics F(z)

The transformation reported in equation (5.24) is used to map R(s) and G(s)
in the z variable, i.e., s → z. By considering a Backward Euler integration
method s = z−1zTs for the transfer function defined in the previous subsection,
their resulting discrete counterparts are:

u(z) zTs + (z − 1)Ti y(z) zTs /a

R(z) = = kp G(z) = =
e(z) (z − 1)Ti u(z) zTs + (z − 1) (b/a)
(5.25)
The considered mapping into z-domain produces stable trajectories (i.e.,
inside the stability region) even if the corresponding continuous-time poles
show a positive real part. Then, the resulting open- and closed-loop transfer
functions are computed as for the continuous-time case:

y(z) zTs kp /a y(z) L(z)

L(z) = = F (z) = = (5.26)
e(z) (z − 1) Ti y ∗ (z) 1 + L(z)

Note that now the mapping k ← z holds, where k ∈ Z is the discrete-

time step index. For instance, this imply y(k)/e(k) = y(z)/e(z). Moreover,
the same values for kp and ki parameters computed in section 5.2 are still
valid. Hence, one may take advantage of the vast body of knowledge available
on continuous-time control theory since the discretization method does not
change the design approach, which can be summarized as:

• Perform a design based on continuous-time transfer functions. Select a cri-

terion (e.g., pole/zero cancellation) to compute kp , ki ;
• Select a discretization method (without changing kp , ki ) and an appropriate
sampling time Ts ;

• Represent the closed loop system as shown in Figure 5.6 (a) or (b).
Derive a PI Controller in Discrete-Time Domain 63

y  (k) e (k) u (k) y (k)

R (z) G (z)
+
−

(a)

y  (t) e (t) e (k) u (k) u (t) y (t)

A/D R (z) D/A G (s)
+
−

(b)

Figure 5.6 Example of discrete closed-loop control system (a) and discrete
controller (b). The scheme reported in (b) includes analog to digital (A/D)
and digital to analog (D/A) converters as interfaces between the hardware
and the real system.

Even if Figure 5.6 (a) and (b) includes the same controller R(z), they
differ on how the rest of the scheme is defined. In particular, Figure 5.6 (a)
considers a discrete transfer function G(z), which is then used to compute L(z)
and F (z) as reported in equations (5.25) and (5.26). Differently, in Figure 5.6
(b) the discretized controller is interfacing the system model in continuous-
time domain, i.e., G(s), through analog to digital (A/D) and digital to analog
(D/A) converters operating at the sampling time Ts used for the discretization.
In Simulink® , the operations of both converters can be mimicked with Rate
Transition blocks.
In any case, both methods starts from the results of the design in
continuous-time domain, i.e., based on G(s) via s.
As a reminder, note that G(z) may be directly derived by discretizing (5.1)
and (5.2), thus, deriving discrete-time differential equations such as

0 0
x(k + 1) = A x(k) + B u(k) (5.27)
0
y(k) = C x(k) (5.28)
0 0 0
The new matrices A , B and C have to be computed according to one of
the previously described discretization method.
 6
Design Example: PI-Based Current Control of
an RL Load

In order to apply the design rules presented in the previous chapter and to
show how to tailor them to a power electronics-based control problem (as
a complete exercise), a simple RL (ohmic-inductive) load circuit fed by a
controllable voltage source is considered as a case study (see Figure 6.1).
The system parameters and the desired characteristics of the controller are
reported here in the following:

• Data:
R = 25 mΩ and L = 100 mH → τG = L/R = 4 s, thus, Ta,G = 5τG = 20 s
• Objectives: reference tracking
y ∗ (t ≥ 1 s) = 3 A and Ta,F = 1 s, thus, τF = Ta,F /5 = 0.2 s

The aim of this exercise is to design a current control loop, i.e., to control the
current flowing into the winding i(t) according to a given step-wise reference
i∗ (t), by using a PI controller while achieving a final settling time of Ta,F = 1 s.
In particular, the reference changes from 0 A to 3 A (i.e., Yss = 3) at t = 1 s,
holding this value until the end of the simulation.
According to Figure 6.1, v(t) is the voltage of a variable voltage source (e.g.,
a controllable DC voltage supply), while R and L are the circuit parameters.
Applying the Kirchhoff’s voltage law, it follows that:

di(t)
v(t) − Ri(t) − L =0 (6.1)
dt

R
RL load
y  (t) e (t) Current u (t) y (t)
controller G (s) +
+ u (t) L
− −

x (t)

(a) (b)

Figure 6.1 RL-circuit under investigation.

DOI: 10.1201/9781003196938-6 65
66 Design Example: PI-Based Current Control of an RL Load

which is a differential equation describing a 1st order linear system. The volt-
age v(t) is the controllable variable, i.e., u(t) = v(t), which allows to in-
crease/decrease the current flow according to the system parameters. Hence,
such current control is based on the manipulation of the input voltage. By
considering x(t) = i(t) and y(t) = x(t), the state-space representation shown
in equations (5.1) and (5.2) is achieved considering A = −R/L, B = 1/L, and
C = 1.
Then, the transfer function of the continuous-time linear system is defined
as follows:

i(t) y(s) 1 1/R 1/L

G(s) = = = = = (6.2)
v(t) u(s) R + sL 1 + s (L/R) s − (−R/L)

where the time constant of the uncontrolled system is τG = L/R = 4 s and,

thus, the settling time is Ta,G = 5τG = 20 s. Note that τG and Ta,G charac-
terize the intrinsic response of such system (i.e., without the presence of any
controller).
A PI controller is introduced. Its design depends on the choice of its per-
formances, that is:
1. The closed-loop settling time ⇐⇒ bandwidth (cut-off fre-
quency) ωc

In this example, the controller should speed up the response of the

resulting closed-loop system against reference transients (i.e., being
faster than the uncontrolled system). According to the desired char-
acteristics reported previously, it results Ta,F = 1 s < Ta,G = 20 s
and τF = Ta,F /5 = 0.2 s < τG = 2 s. Thus, the controlled system is
going to result twenty times faster than the open-loop one.
Remember: if a desired characteristic such as Ta,F is not clearly
specified, the choice of a suitable time response must be the first
decision of the designer.
2. The robustness level ⇐⇒ phase margin φm
According to the chosen design parameters, the command-tracking perfor-
mances (i.e., how the resulting system is able to be fast and close to the given
reference) or the disturbance-rejection performances (i.e., how the resulting
system is able to cope the effects of a disturbance acting on it) may be em-
phasized. Note that, in general, high command-tracking performances results
in poor disturbance-rejection one.

Given R(s) of the same kind as that one reported in equation (5.8), the open-
loop transfer function is:

kp (1 + sTi ) 1/R Ti =L/R kp /R

L(s) = =⇒ L(s) = (6.3)
sTi 1 + s (L/R) sTi
Simulink® Simulation 67

where a pole/zero cancellation is imposed by setting Ti = τG = L/R. Since

explicit formulas for kp and ki based on ωc can be defined starting from the
requirement Ta,F = 1 s, it follows that τF = Ta,F /5 = 0.2 s, which leads to:

1 kp = ωc L = 0.50 Ω

Ta,F
Ta,F → τF = → ωc = = 5 rad/s →
5 τF ki = ωc R = 0.125 Ω/s
(6.4)
Therefore, the computed kp and ki parameters ensure the command-tracking
performances and, given L(s) as a minimum phase system (mps), the phase
margin φm is about 90° (i.e., high robustness). It must be noted that kp and
ki have precise unit values, which in this case are Ω and Ω/s, respectively. The
resulting closed-loop transfer function is:

L(s) 1 1 1
F (s) = = = = (6.5)
1 + L(s) 1 + sTi 1 + s (L/R) 1 + s (kp /ki )

At this point of the design procedure, all the characteristics of F (s) are known.
The next step is to numerically simulate the system.

6.1 Simulink® Simulation

In this book, the simulations of both uncontrolled and controlled system be-
haviors are carried out with MATLAB® and Simulink.®1 This helps power
electronics control engineers to gain an insight into the interaction of digital
control algorithms and conversion systems. Simulink® allows to quickly model
both the digital control algorithms and the analog circuits of a power elec-
tronic systems. According to the complexity and/or the requirements of the
needed control scheme, the model can just focus on the system dynamics, i.e.,
focusing on the transfer functions and the transient response, or introduce
more complex behavior at the component level, i.e., using detailed block rep-
resenting resistors, inductors, capacitors, semiconductors, . . . and their phys-
ical limitations, thus, covering a multi-domain modeling using built-in library
blocks.
Simscape™ enables a quick modeling of physical systems within the
Simulink® environment. In particular, Simscape™ Electrical™ (formerly
SimPowerSystems™ and SimElectronics® ) provides component libraries (e.g.,
semiconductors, motors, and converters) with the integration of mechanical,

1 Note that Simulink® supports both floating-point and fixed-point computations, as well

as continuous-time (analog), discrete-time (digital), hybrid (mixed-signal), and multi-rate

systems.
68 Design Example: PI-Based Current Control of an RL Load

Figure 6.2 Relationship between system dynamics and time step [32].

thermal, and other physical aspects for modeling and simulating applica-
tions such as electromechanical actuation, drives, smart grids, and renew-
able energy systems. This book mainly refers to blocks from the Simscape™
Electrical™ Specialized Power Systems library, which can be found by clicking
in the Simulink® library browser: Simscape → Electrical → Specialized Power
Systems. For further information, the reader is referred to the MathWorks®
website [30].
Using both standard blocks or Simscape™ components, analog (passive
or active) and digital components can be modeled and simulated to validate
the design of closed-loop control algorithms as well as the supervisory logic
through real-time simulations.

Trade-off between Simulation Speed and Model Complexity:

The speed at which a controller must respond to ensure stable/safe operation
of the system is a key point of Model-Based Design. Figure 6.2 shows a range
of physical systems and an order of magnitude of their corresponding time
constants; note that power electronics controllers are typically sampling in
the order of µs or faster. In general, a model described by simple dynamics
(e.g., transfer functions) will simulate really fast, though accuracy will not
be maximized. Instead, a model with more complex dynamics (e.g., built in
using elements from the Simscape™ library) will simulate slower, but will pro-
vide a more accurate description of the true power electronics behavior than
the simplified representation. In the case of the considered RL load, the vari-
able voltage source may be provided by a power converter stage. In practice,
the latter returns a squared waveform output voltage, but it can be approx-
imated by its average value. Therefore, the simulation will not capture the
effects of switching behavior, but it will be suitable for tuning the control al-
gorithms and testing the control under nominal operating conditions. The RL
branch can be modeled as a set of linear differential equations with lumped
parameters. Because it uses a simple mathematical representation of the
Simulink® Simulation 69

system behavior, the simulation will run fast. Assuming that the same system
has to be installed in a real harsh environment (e.g., the power train of an
electric vehicle) which is sensitive to temperature variations, the model should
include a detailed power electronics switching behavior, effects of temperature
on electrical efficiency, as well as the nonlinear model of the inductor, resulting
to be mathematically more complex than before. As a result, the simulation
will run really slow.
Before performing real-time simulations, the very first task is the initial-
ization of all the needed parameters. In this book, this procedure is done using
MATLAB® code by creating a script/source file (.m or m-file) that is called in
Simulink® from the Model Properties → InitFcn callback (the .m file is exe-
cuted every time the Simulink® model is running) or as PostLoadFcn callback
(the .m file is executed only once). Nevertheless, the .m file could be even ex-
ecuted manually, thus, without creating such automatic routine in Simulink® .

Be careful: The book refers to one of these procedure in every exercise,

since they allow easy initialization and edit of the simula-
tion parameters, reducing the risk of manual change inside
the blocks, which can lead to a difficult debug.
For the considered case study, the m-file which includes the code with all
the definitions of the system parameters, the computation of the controller
gains, and the transfer functions is the code reported here in the following:

%% define the Laplace operator

s = tf(’s’); % then just call s
%% system parameters and transfer function:
R = 0.025; % ohm
L = 0.1; % H
tauG = L/R; % ohm*s / ohm = s
wg = sqrt(1/(R^2)-1)/tauG; % rad/s
TaG = 5*tauG; % s
G = 1/(R + s*L); % G(s)
%% closed-loop requirements
TaF = 1; % s
tauF = TaF/5; % s
wc = 1/tauF; % rad/s (approximation)
%% PI parameters
% pole/zero cancellation procedure
kp = wc * L; % ohm
ki = wc * R; % ohm/s
Ti = kp / ki; % s - equal to L/R
Rc = kp + ki/s; % R(s) - PI controller tf
Ts = 1e-3; % sample time/discretization
%% transfer functions
Lc = Rc*G; % L(s) - open-loop tf
 70 Design Example: PI-Based Current Control of an RL Load

Fc = Lc/(1+Lc); % F(s) - closed-loop tf

%% voltage constraint
sat = 0.5; % power supply limit

Then, it is enough to recall the variable names into a specified block in

Simulink® to automatically assign the defined values, e.g., if R is used inside
a Gain block, the value 0.025 is automatically updated. Given the parameters
shown previously, the corresponding bode diagram of G(s), L(s), and F (s) are
reported in Figure 6.3, together with time constants and angular frequencies
indications. Note that, F (s) behaves exactly as a low pass filter as expected.
In order to show how to use both approaches, i.e., using standard or
Simscape™ blocks, a Simulink® scheme is created such that:

• The controller is implemented using the standard block PID Controller from
the library Continuous;
• The RL load is emulated through a transfer function (tf), i.e., a stan-
dard block, or building an equivalent circuit with blocks available in the
Simscape™ library. Thus, the user has the possibility to select which one of
the two modeling strategies should be included in simulation and then run
it.

The full scheme is shown in Figure 6.4. The possibility to switch between
the two load emulations is ensured by the manual switch Sw2 which is a stan-
dard block from the Simulink® library. Every double-click done on the block
changes its connection. Note that Figure 6.4 includes a further stage enabled
through Sw1, which can add a saturation block (standard block) to bound
the actuation signal u(t). This will be discussed in details later in Section 6.2
referring to the windup effects. Thus, it is anticipated in Figure 6.4 just for

(a) (b)

Figure 6.3 Bode diagrams for the RL-load case study.

 Simulink® Simulation 71
using Simscape

Continuous

Current Measurement
i
+
-

+
v(t) RL
y(t)
[yr] Goto
no saturation

s
-
discrete PI controller From
reference step
y*(t)
[yr]
e(t) e(k) u(k) u(t) y(t)
y*(t) PI(z)
y(t) Sw1 Sw2
saturation y(t)
m(t) [e]
e(t)
[u]
u(t)
G(s) [m]
[e] [u] [m] m(t)
using transfer function sat

> use Sw1 to include (or not) the actuation saturation

> use Sw2 to simulate the system using Simscape or transfer function G(s)

Figure 6.4 Simulink® scheme of a current control for an RL load using both
standard and Simscape™ blocks. Rate Transition blocks are inserted to mimic
the behavior of the A/D and D/A converters (see Figure 5.6 (b)).

the sake of completeness.

Remark: by using both standard or Simscape™ blocks to emulate

the RL load, the control performances and the averaged
system behavior are exactly the same. Thus, the user could
equally adopt one of the two to evaluate the controller de-
sign, while the Simscape™ approach is preferable to emu-
late a more accurate behavior of the system at the compo-
nent level.
The Simulink® file (.slx) related to Figure 6.4 can be downloaded from the
editor web page, while the file settings are further detailed in the following
sections.

6.1.1 Use of standard blocks (continuous/discrete)

Referring to the lower blue area reported in Figure 6.4, the closed-loop control
of the RL load is created by using a Transfer Fcn block, which includes the
transfer function G(s), whereas a PID Controller block inside the pink
region contains the parameters kp and ki of R(s). More specifically:

• Transfer Fcn block (Simulink® )

This block models a linear system through a transfer function in the Laplace-
domain variable s. It can model single-input single-output (SISO) and single-
input multiple-output (SIMO) systems. A generic transfer function G(s)
must be entered in the form:
 72 Design Example: PI-Based Current Control of an RL Load

num(1)sn−1 + num(2)sn−2 + · · · + num(n)

G(s) = (6.6)
den(1)sm−1 + den(2)sm−2 + · · · + den(m)

where num and den are matrices that contain the coefficients of the numera-
tor and denominator ordered in descending powers of s with order n and m,
respectively. For SISO systems (as in this example), these scalar coefficients
have to be entered as vectors: numerator coefficients: [1] → num(1) = 1
with n = 1; denominator coefficients: [L R] → den(1)s = L s, den(2) = R
with m = 2; zero initial condition.

• PID Controller block (Simulink® )

The PID Controller block can implement PID, PI, PD, P only, or I only
controllers. The block can be specified for continuous-time or discrete-time
domain operations. In both cases, the block output is, in general, a weighted
sum of the input signal (P action), its integral (I action) and its derivative
(D action). The block supports several controller types and structures. Con-
figurable options in the block include:

– Controller type (PID, PI, PD, P, I);

– Controller form (serial or parallel);
– Time domain (continuous or discrete);
– Initial conditions and reset trigger;
– Output saturation limits and built-in anti-windup mechanism;
– Signal tracking for bump-less control transfer and multi-loop control.

The internal structure of the block changes by activating different variant

subsystems depending on how those options are edited.
According to the requirements of the considered exercise, PI should be se-
lected as Controller type. Then, the parameters should be set considering
the correspondence P = kp and I = ki , which are the controller param-
eters designed in continuous-time domain. However, this exercise foresees
the implementation of a discrete-time PI controller as a final goal. Thus,
the Discrete-time option in the Time-domain settings must be selected with
Sample time Ts . All the other settings should be kept with their default
values.
• Step and Goto/From blocks (Simulink® )
The Step block provides a step between two definable levels at a specified
time. When the simulation time is less than the Step time, the block output
is the set equal to the Initial value. Otherwise, the output is the set on the
Final value.
Instead, the Goto block passes its input to its corresponding From block
without any physical connection. Thus, their usage is manly targeted to
achieve a more readable scheme without too many connection lines. The
 Simulink® Simulation 73

input can be a real- or complex-valued signal or a vector of signals of any

data type. A Goto block can pass its input signal to more than one From
block, although a From block can receive a signal from one Goto block only.
Goto/From blocks are matched through univocal tags/label.
Note that Rate Transition blocks (standard blocks) are used to interface
the continuous-time scheme with the discrete-one (i.e. A/D and D/A). They
are set either with Sample Time Ts or -1.

Case study settings:

Start time: 0(s); Stop time: 4(s); Solver type: Variable-step; Solver:
ode23t; Solver details: default. This is a first example of simulation in which
the main focus is the understanding of the closed-loop dynamics. Therefore, a
Variable-step solver is used. Neverthless, in the next chapters a Fixed-step
solver is preferred, being closer to operations in a embedded environment.

6.1.2 Use of Simscape™ (specialized power systems)

Referring to the upper blue area shown in Figure 6.4, the closed-loop control
of the RL load is now created combining the PID Controller block (i.e., us-
ing a standard block) and the Simscape™ blocks, which are used to model
the electrical system in a more physical manner. Specifically, all these new
components are described here in the following:
• Controlled Voltage Source block (Simscape™ — Specialized Power Sys-
tems)
The Controlled Voltage Source block converts a Simulink® input signal into
an equivalent voltage source. The generated voltage is driven by the input
signal of the block which is associated to a control signal (e.g., coming from a
control loop). Positive/negative terminals are associated to electrical ports.
Thus, voltage signals cannot be directly connected to standard Simulink®
block, e.g., scope, logic operators, gains. In this case, the Source type is set
on DC, the Initial amplitude on 0, whereas no measurements are foreseen.
• Series RLC Branch block (Simscape™ — Specialized Power Systems)
The Series RLC Branch block implements an impedance as a series combi-
nation of R, L, C elements. At the specified frequency, the load exhibits a
constant impedance value. By double clicking on the block, it is possible to
select which element is to be used, e.g. if the C it is not used we select the
RL only. According to the aforementioned design, the adopted settings are
Branch type: RL, then Resistance (Ohms): R and Inductance (H): L. Do
not select Set the initial inductor current checkbox and it is up to the reader
set any measurement.
• Powergui block (Simscape™ — Specialized Power Systems)
The powergui block is necessary to simulate any Simulink® model contain-
ing Simscape™ Electrical™ Specialized Power Systems blocks. It stores the
74 Design Example: PI-Based Current Control of an RL Load

(a) (b)

Figure 6.5 Performances of the current control of an RL load.

equivalent Simulink® circuit that represents the state-space equations of the

resulting scheme and it allows to choose the solver for this circuit (which
could be different from that one used for the reaming part of the scheme
based on standard blocks). The available solvers are: Continuous, which uses
a variable-step size; Discrete, which uses fixed time steps; Phasor. The pow-
ergui block also provides tools for steady-state analysis and for advanced
parameter design. For this exercise, the recommended Simulation type is
Continuous. No other settings should be edited.

In addition, a Current Measurement block must be inserted in the scheme to

measure the instantaneous current flowing in the electrical connection line. Its
output provides a data signal that can be used by other Simulink® blocks, thus,
moving from Ampere to a numerical data (e.g., 1 A is different to numerical 1).

Generally speaking, the stator windings in electrical machines behave as RL

loads plus the effects of back-emfs. Therefore, the control design proposed in
this chapter directly links to the torque regulations for electrical machines
reported in the next exercises.

6.1.3 Controller performances

Given the design and the settings of the aforementioned case study, the corre-
sponding closed-loop scheme is simulated in MATLAB® and Simulink® . The
resulting performances are shown in Figure 6.5. Figure 6.5 (a) shows the cur-
rent flowing into the winding y(t) = i(t) (blue line), which is correctly following
the step-wise reference y ∗ (t) = i∗ (t) (red line). The set point changes from
0 A to 3 A (i.e., Yss = 3) at t = 1 s. Different Ta,F could be defined according
to the value of the error . Namely, Figure 6.5 (a) shows two different Ta,F
computed as the time interval to reach: 2.79 A for  ≈ 6.67% (green dotted
Simulink® Simulation 75

line) with Ta,F ≈ 0.6 s; 2.97 A for  ≈ 1% (black dotted line) with Ta,F ≈ 1 s.
Therefore, the definition of the settling time in terms of accuracy is necessary
to classify the performances of the systems and to compare them, e.g., the
comparison among two systems must refer to the same degree of accuracy to
understand which behaves better. From now, the proposed procedure refers to
the design guidelines reported in Chapter 5, in which  ≈ 1% is suggested as
reference accuracy, thus, entering in the region [(1 − 0.01)Yss (1 + 0.01)Yss ],
where Ta = 4 ÷ 5τ holds. It results that Ta,F = 1 s < Ta,G = 20 s and
τF = Ta,F /5 = 0.2 s < τG = 2 s, so the closed-loop control system is going to
be ten times faster than the open-loop one.
The effectiveness to achieve the desired steady-state value Yss can be ver-
ified via the final value theorem. Since the Laplace transform of a step-wise
signal in time-domain is µs /s, where µs is the amplitude value (i.e., 3 A), by
multiplying F (s) = y(s)/y ∗ (s) with µs /s it results Yss = F (s)µs /s. Namely:
µs µs
lim sYss = lim sF (s) = lim = µs = 3 A (6.7)
s→0 s→0 s s→0 1 + sTi

Note that the results holds for every considered value of Ti , that is, for every
design choice.
Based on Figure 6.5 (b), the command-tracking performances are achieved
thanks to the sudden change of the input voltage u(t) = v(t) at t = 1 s. In
particular, the error e(t) and the control variable u(t) are following a decay
which is equal (but mirrored) with respect to y(t), as shown in Figure 6.5 (a).
At steady-state, e(t) = 0 while u(t) 6= 0 since the system changed its operating
point and a certain voltage is necessary to assure the required current flow.2
This is still a consequence of the Kirchhoff’s law, where at steady-state the
current derivative can be neglected while computing the voltage value,3 that
is:

di(t)
v(t) − Ri(t) − L =0 → Uss = Vss = RIss = RYss = 0.075 V (6.8)
dt
Given the aforementioned parameters and performances, the designed PI con-
troller achieves all the specifications.

After this first design, the controller parameters are modified to evaluate
the potential issues due to a wrong design. Two scenarios are considered and
the performances of the new controllers are compared to the previous one (the
correctly designed one). The plots shown in Figure 6.6 reports the behaviors
of all the design controllers.

2 Note that a voltage drop is always present among the inductor (and its internal series

resistance) and the resistance.

3 The steady state quantities are denoted as uppercase letter with subscript ss.
76 Design Example: PI-Based Current Control of an RL Load

(a) (b)

Figure 6.6 Performances of different current controllers.

More specifically, the two new controller are designed such that:
1. Integral action is removed, i.e., ki = 0.
It is worth noting that a simple P controller keeps a permanent
offset at steady-state. The kp action contributes to quickly increase
the voltage only, bringing the output close to the current reference.
The distance between y (t) and y ∗ (t) depends on the value of kp .
In case of a P controller, no pole/zero cancellation is possible. The
open-loop and closed-loop transfer functions simply result:

1 kp
L(s) = kp F (s) = (6.9)
R + sL kp + R + sL
Then, the cutoff frequency can be computed by imposing
|L(s)|s=jωc = 1. It can be easily proved that this equality leads
q
to ωc = kp2 − R2 /L. If the considered bandwidth is again ωc =
5 rad/s, the resulting kp is obviously different. Moreover, the ap-
plication of the final value theorem on the controlled system leads
to:

µs kp µs kp µs
lim sF (s) = lim = (6.10)
s→0 s s→0 kp + R + sL kp + R
where the result is kp -dependent. An increase in kp might improve
the reference tracking, while the offset persists since it could be
zeroed only if kp → ∞. This also implies that a lower voltage effort
u(t) than before might be used to keep a (wrong) steady-state value
i(t).
2. kp is kept equal to that one of the previous controller, while ki is
chosen ten time larger, i.e., 10 × ki .
Derive an Anti-Windup PI Controller Scheme 77

By increasing the integral part so significantly, the current response

shows an oscillating behavior. Figure 6.6 (a) reports that, by chang-
ing the integral action, a faster transient response might be achieved
at the cost of an higher stress on the system due to the overshoots.
As an example, if the circuit cables see their insulation designed for a
specific current peak and a certain temperature limit/heat (usually
provided as maximum values of the cable class and diameter), the
repetitive current overshoots could get them closer to those limits.
Plastic insulation and jackets that surround the wires are usually
rated to withstand up to 90 − 100 ◦C, but temperatures that ap-
proach this limit are not recommended anyway. Thus, oscillating
response of i(t) might be not acceptable for this reason.
Hence, any physical constraints must be considered at the early
design stage. Even if the transient response provided by this new
controller is faster than before, the settling time is longer due to
the oscillating behavior of i (t), thus, leading to a late entry in the
safety region  = ±1%.
Oscillations also may imply a further stress on the power supply.
Both e(t) and u(t) become negative for certain periods of time.
Therefore, the energy is also flowing back to the power supply
(which provides the actual value of DC voltage u(t)) and not only
toward the RL load. This reversed power flow is only possible if the
power supply is bidirectional, i.e., if it is able both to receive and
dissipate/store energy, otherwise, another physical constraint must
be considered during the controller design.
Even in this case, the final value theorem ensures Yss = 3 A.

6.2 Derive an Anti-Windup PI Controller Scheme

A linear controller (e.g., PI) is simple to design, but its performances are good
as long as the dynamics remain close to what is foreseen by the linear theory.
Nevertheless, real systems are affected by non-linearities and physical limi-
tations. For instance, electrical drives suffer of magnetizing phenomena, skin
effects, cable length issues, both electro-thermal and electromagnetic interfer-
ence. In addition, power supplies are limited. In particular, these latter are
among the easiest cases to treat since they represent static non-linearities. By
considering the control variable u(t) bounded by the energy flow capabilities
of a power supply, it follows that:
 78 Design Example: PI-Based Current Control of an RL Load

Saturation RL load
y  (t) e (t) Current u (t) m (t) y (t)
G (s)
+ controller
−

(a)
1.5

0.5

-0.5
-1 -0.5 0 0.5 1 1.5 2 2.5 3

(b)

Figure 6.7 Closed-loop control scheme with saturated actuation (a) and ex-
ample of saturated output (b).

if umin ≥ u(t)< umax (linear region)


 umin
m(t) = sat (u(t)) u(t) if umin < u(t) < umax (linear region) (6.11)
umax if umin <u(t) ≥ umax (linear region)


where m(t) is the actuation variable. Since the control variable is a voltage
level, umin and umax are minimum and maximum allowed voltages that the
DC power supply can output, that is, vmin and vmax . For this case study, a
(fictitious) unidirectional DC power supply which is able to provide a vari-
able voltage from umin = 0 V to umax = 0.5 V is assumed to be available. If
these limitations are introduced in the proposed model as reported in equation
(6.11), the closed-loop scheme changes as shown in Figure 6.7 (a). The satu-
ration block receives umin as input and it outputs umax . There is a Simulink®
block that can mimic this behavior among the standard ones under the name
saturation. From now on, a discrimination is introduced between:

• u(t), that is, the control variable returned by R(s) as results of the control
design (i.e., output of the PI controller);

• m(t), that is, the actuation variable seen by the RL load, where m(t) could
saturate depending on the value assumed by u(t).

From the controller perspective, note that the previous design of R(s) does
not include any information regarding the saturation of u(t).
Indeed, Figure 6.8 (a) reports the transient response of the previous control
scheme without saturation (blue dotted line), the related command-tracking
Derive an Anti-Windup PI Controller Scheme 79

4.5

3.5

2.5

1.5

0.5

0 0.5 1 1.5 2 2.5 3 3.5 4

(a) (b)

Figure 6.8 Comparison between unsaturated and saturated transient re-

sponses.

performances and the evolution of the same system including saturation effects
(blue solid line). For this latter, the output y(t) takes long to reach the steady-
state condition due to the windup effect of the integrator, which also generates
an overshoot.
Moving to Figure 6.8 (b), since R(s) operates regardless any saturation
limit, the voltage is required to be u(t) = 1.5 V at time instance t = 1 s to
bring y(t) closer to y ∗ (t) = Yss = 3 A. However, the DC power supply is able
to feed only m(t) = umax = 0.5 V. Thus, the system is fed by this maximum
voltage value. At the following time instant, the output y(t) do not increase as
expected due to a lower actuation. This limitation implies a larger error e(t)
compared to the unsaturated case, which shows an higher y(t) due to an higher
actuation signal (which has a peak value exactly equal to u(t)). The effects of
this saturation on u(t) and y(t) are more clear as long as u(t) returns below
umax , which enables the system to re-enter into the linear region. Nevertheless,
the effects due to the presence of a large error are still present. Thus, even if
m(t) = u(t), the system requires a long time to reach the steady state and it
shows an overshoot as well.
This is a well-known problem which has a really practical explanation.
More specifically, the windup effect is mainly due to the integrator part of the
PI controller, which for u(t) > umax keeps integrating the tracking error e(t)
producing m(t) 6= u(t). When m(t) = u(t), the controller re-enters into the
linear region behavior. This means to wait for the integrator discharge (also
called integral discharge time or de-saturation time), which depends on the
integration area defined by the unsaturated u(t) (yellow area in Figure 6.8
(b)).
This effect is particularly clear in Figure 6.9, which includes a further
simulation of a system with a stricter voltage limitation, i.e., umax = 0.3 V.
The larger the integrating area, the longer and the stronger the windup
effect. So, the case with saturation set at 0.3 V implies a longer windup and
 80 Design Example: PI-Based Current Control of an RL Load

(a) (b)

Figure 6.9 Comparison between different transient responses and the related
windup effects.

a quite larger settling time Ta,F , keeping  = 1%. To mitigate the windup
problem, the controller must be augmented according to one of the many anti-
windup schemes available in the literature. In this book, the back-calculation
principles are considered, which lead the compensation schemes reported in
Figure 6.10.
By exploiting the block schemes theory, it can be proven that the diagrams
reported in Figure 6.10 (a) and (b) are equivalent, creating both an anti-
windup compensation. The back-calculation method uses a feedback loop to
discharge the integrator when the controller hits specified saturation limits,
avoiding to enter in nonlinear operations. The back-calculation coefficient kaw
is computed as kaw = 1/Ti = ki /kp . This time, constant Ti determines how
quickly the integrator of the PI controller is reset by the anti-windup loop.
The regulator reported in Figure 6.10 (a) is subjected to the continuous-time
equations reported here in the following:

if

 umin umin ≥ p(t)< umax
u(t) = sat (p(t)) kp e(t) + 1
1+sTi u(t) if umin < p(t) < umax (6.12)
umax if umin <p(t) ≥ umax


where p(t) is an auxiliary variable used to keep the notation short. Therefore,
the key aspects of an anti-windup scheme can be summarized as:

• No effect when the actuator is not saturating, i.e., within the linear region
controllers with and without anti-windup scheme behave the same;
• Just one additional parameter kaw = 1/Ti is required;
• If the saturation limits are not clearly specified, a mathematical model of
the actuator should be used.
Derive an Anti-Windup PI Controller Scheme 81

Current regulator

e (t) u (t)
kp
+
+

1
1 + sTi

(a)
Current regulator

kp

+
e (t) 1 u (t)
ki
+ s +
+

− +

kaw

(b)

Figure 6.10 Example of back-calculation anti-windup scheme to update R(s).

Both schemes reported in Figure 6.10 share the idea to include the satu-
ration limits as thresholds inside a suitable saturation model. This latter may
be just implemented by looking at the data sheet of the DC power supply
or derived by using some identification techniques, especially in case where
those limits are uncertain. Note that, in this case, it is quite easy to define the
saturation limit. However, this is not always the case, e.g., accurate describe
magnetizing phenomena or skin effects are not always available.

Simulink® : to include a PI controller with anti-windup configuration in sim-

ulation, insert a PID block keeping the settings shown previously and click
on the Output Saturation tab in the block dialog. There, select Limit output
and enter the saturation limits. Then, select back-calculation from the Anti-
windup method menu and specify the Back-calculation coefficient Kb equal
to kaw . Hence, it is only necessary to change few settings into the PID block
to use an anti-windup scheme, which makes their implementation quite sim-
ple even for the implementation into MCU proposed in the following Chapters.
82 Design Example: PI-Based Current Control of an RL Load

(a) (b)

Figure 6.11 Performances of the current control of an RL load including anti-

windup compensations.

The results of simulations including anti-windup strategies are summa-

rized in Figure 6.11. In practice, the transient response reported in Figure
6.11 (a) do not present any overshoot. This is true even if stricter limitation
are introduced, like umax = 0.3 V. Looking at Figure 6.11 (b), it can be noted
that u(t) = m(t) holds for all simulations, i.e., the controller “knows” the sat-
uration limits. Nevertheless, the settling time is now longer than Ta,F = 1 s,
which was computed with the previous (unsaturated) design. This is obviously
a consequence due to the limit introduced on u(t) (limited actuation/limited
energy) which cloud not make i(t) increasing as expected from the previous
design. Therefore, the transient takes longer to reach the steady-state condi-
tion, while assuring no overshoots. In case none of them is foreseen, the Ta,F
difference is negligible. However, if the introduction of an anti-windup scheme
lead to violations in the settling time constraint, the design parameters kp , ki
might be also updated accordingly.
It must be remembered that if saturation phenomena are neglected in the
design phase, they can lead to closed-loop instability, especially if the process
is open-loop unstable.

6.3 Design Summary

The procedure to analyze a system, design a closed-loop controller, and make it
suitable to be implemented into a microcontroller can be summarized through
the following key points:
 Design Summary 83

Figure 6.12 Management of a power control topology using integrated circuits

vs discrete analog components. The term digital is used to denote that all
external signals are converted from their native analog state to a digital state
such they can be operated on by the main controller in the system [8].

• Analyze and describe the uncontrolled system dynamics/natural dynamics.

This is necessary to understand the order of G(s) as well as defining τG ;
• Select the output y(s) and input/control u(s) variables.
Given G(s), then it is possible to define a closed-loop control scheme;
• Select the controller type in continuous-time domain.
In this book, linear controllers are considered only, such as the PI ones. The
Laplace operator is exploited to derive R(s);
• Consider all the requirements to compute the controller parameters.
Once the problem requirements are set, use one of the criteria shown in
Chapter 5 to compute kp and ki ;
• Discretize the controller.
A discretization time Ts must be chosen as well as a suitable integration
method according to the location of the system poles to map R(s) ← R(z).
If necessary, the system transfer function can be discretized too;
• Check if relevant limitations are present in the system.
For instances, it should be verified possible actuation limits due to real de-
vices capabilities. Then, saturation should be included in the control scheme;
 84 Design Example: PI-Based Current Control of an RL Load

• If necessary, adjust the controller with an anti-windup scheme.

Each of the methods shown in this Chapter can be used to eliminate windup
effects which may lead to damage or aging of system components;
• Select which elements of the control scheme have to be implemented on the
MCU.

Note that the controller design has been started in continuous-time toward
discrete-time domain. This is necessary to meet the working principle of em-
bedded platforms, i.e., systems with discrete nature. Nowadays, this technol-
ogy represents the main industrial standard for power electronic-based appli-
cations (and many more) instead of using the more historical analog control
systems. Their difference is shown in Figure 6.12 which consider a closed-loop
voltage control scheme.
The proposed discrete-time domain approach led to a Simulink® -based
scheme suitable for the automatic code generation features. More details on
this topic are reported in the upcoming chapters.
7
Manipulate the Variables Format: Data Types

The main objective of this book is to empower the reader with the ability to
work and coding with standard MCU. As already discussed in the previous
sections, due the discrete/digital nature of embedded systems, tuning and sim-
ulations of power electronic setups are carried out in the discrete-time domain.
Nevertheless, real implementations of control algorithms could not be possible
without considering even the hardware structure of a MCU. Clock frequency,
maximum sampling time, memory storage, peripherals resolution, computa-
tional capabilities are just some examples of hardware-based requirements and
specifications. In particular, the computational capability of a MCU is defined
by the nature of the variable which may be involved in the computations and
the type of operations they are asked to perform. The underlying concept is
also called data-type analysis.

7.1 Fixed Point vs Floating Point Representation

The TMS320F28069M microprocessor adopted on the LAUNCHXL-F28069M
operates with 32 bit word length, for which numerical representations can be
either in fixed point or floating point format.

Fixed point: fixed point format exploits the two’s complement represen-
tation to allow the computations between rational numbers. The positive and
negative extreme values differ by one least-significant bit (LSB). Keeping a
32 bit word length, the bit format is:

SI
|{z} .f f f f f f f f f f f f f f f f f f f f f f f f f f f f f f (7.1)
| {z }
2 bit integer 30 bit quotient(fraction)

which is also called 2Q.30 (or just IQ30), where the 30 “f ” represent the
number to the right of the radix point.
Instead, the two remaining bits on the left side of the radix point represent
the integer part of the number. Namely “S” is the sign bit, where 0 is allocated
for positive numbers and 1 for negative ones, and one single bit “I”. Hence,
the maximum positive value of the integer part of such numbers is decimal

DOI: 10.1201/9781003196938-7 85
86 Manipulate the Variables Format: Data Types

Figure 7.1 General representation of a floating point format number.

“1” (i.e., integer bits S = 0, I = 1) and the minimum decimal value is “−2”
(i.e., integer bits S = 1, I = 0).
Floating point: The bit format for floating-point numbers is

S eeeeeeee
| {z } .f| f f f f f f f f f f{z
ffffffffffff (7.2)
}
8 bit exponent 23 bit quotient(fraction)

which refers to the structure reported in Figure 7.1.

Floating point format works with a sign-magnitude representation. The
sign of the number is represented the bit S, where 0 identifies positive numbers
and 1 negative numbers. The 8 “e” bits represent the exponent of the floating
point magnitude, in the range −127 . . . 126. Instead, the quotient or fraction
contains 23 “f ” bits which represent the digits at the right of the radix point.
Hence, the real value assumed by a given 32-bit data with a given biased
sign, exponent e, and a 23-bit fraction is computed as:

(−1)b31 × 2(b30 b29 ...b23 )2 −127 × (1. b22 b21 . . . b0 )2 (7.3)

which yields to the following compact form:
23
!
X
(−1) × 2
S e−127
× 1+ b23−i 2−i
(7.4)
i=1

The 32 bit floating point representation allows to reach ±2(255−127) (1−2−23 ) ≈

±3.4028 · 1038 . Thus, numbers within a very wide given range can be defined.
Note that the sign-magnitude representation adopted in the floating point
format defines a range within ± the maximum magnitude.

Example
As an example, Figure 7.2 shows how the value x = 0.15625 is represented in
floating point format. Namely, x is split in terms of bits as follows:

Figure 7.2 Representation of x = 0.15625 in floating point format.

 Fixed Point vs Floating Point Representation 87

Table 7.1 Comparison of 32 bit floating and fixed point representations: ranges
and resolutions.

Floating point Fixed point (IQ30)

±2 (255−127)
(1 − 2 −23
)ˇ (2 − 2−30 ) ˇ
Max. decimal
value 3.4028 · 1038 1.9999999991
Min. decimal ±1.1755 · 10−38 −2.000000000
value

Relative 2−24 ˇ 5.9605 · 10−8 2−32 ˇ 2.3283 · 10−10

resolution

Absolute 2−24 · 2−127 2−30 ˇ 9.3132 · 10−10

resolution

• Sign: S = b31 = 0
(−1)S = (−1)0 = +1 ∈ {−1, +1}
P7
• Exponent: e = (b30 b29 . . . b23 )2 = i=0 b23+i 2 = 124 ∈ {1, . . . , 28 − 1 −
+i

1} = {1, . . . , 254}
• 2(e−127) = 2124−127 = 2−3 ∈ {2−126 , . . . , 2127 }
P23
• Fraction: 1.b22 b21 . . . b0 = 1 + i=1 b23−i 2 = 1 + 1 × 2 = 1.25 ∈ {1, 1 +
−i −2

2 , . . . , 2 − 2 } ⊂ 1; 2 − 2
−23 −23 −23
⊂ [1; 2)

Thus:

x = (+1) × 1.25 × 2−3 = +0.15625

Note that:

• 1 + 2−23 ≈ 1.000 000 119

• 2 − 2−23 ≈ 1.999 999 881
• 2−126 ≈ 1.175 494 35 × 10−38
• 2+127 ≈ 1.701 411 83 × 10+38 .

To the aim of providing a better understanding of the floating and fixed point
representations, they are compared in Table 7.1.
The floating point format range is much wider than the fixed point one, but
this advantage comes with the penalty of a less favorable relative resolution,
for which the eight bits used for the exponent are not available for resolution.
For both floating and fixed point formats, the radix point can be moved
to any location within the 32-bit word to manage the trade off between range
and resolution.
88 Manipulate the Variables Format: Data Types

Table 7.2 Double data type.

Bits Usage
63 Sign (0 =positive, 1 =negative)
62 to 52 exponent, biased by 1023
51 to 0 Fraction f of the number 1.f

Remark
A signed 32-bit integer variable has a maximum value of 231 −1 = 2147483647,
whereas a 32-bit base-2, floating-point variable can reach numbers up to
(2 − 2−23 ) × 2127 ≈ 3.402823 × 1038 .

It is important to note that all integers with six or fewer significant digits
as well as any number that can be written as 2n (with n ∈ [−126, 127]) can
be converted into floating-point values without any loss of precision.

7.2 Single vs Double Precision

As already mentioned in the previous section, a floating point representation
can describe a wider range of numbers than a fixed point one, if the same
bit width is considered at the cost of a loss in precision. Even if the previous
section refers to 32-bit long words, when high accuracy is required, a double
precision format could be used by enlarging the length of the representa-
tion reported previously up to 64 bit. More precisely, in the IEEE 754 (1985)
standard [1], the 32-bit base-2 format is officially called single, while the
additional floating-point format of 64-bit base-2 is denoted as double.1 Note
that one of the first programming languages to provide single-and double-
precision floating-point data types was Fortran. Nevertheless, depending on
the software in use, the command name to identify those data types may
change.
Both MATLAB® and Simulink® represent numbers in floating-point no-
tation unless otherwise stated (i.e., explicitly declared). By default, the data
type is double precision, but it is possible to convert any number in single
precision (or other data types) with a simple conversion function. According
to the IEEE Standard, any value stored as a double requires 64 bits formatted
as shown in Table 7.2.
The values
 that can be+308represented with double  numbers lay in the nega-
tive range −1.79769 × 10 , −2.22507 × 10 −308
and in the positive range
2.22507 × 10 −308
, 1.79769 × 10 .
 +308


1 Note that a base-10 representations was introduced more recently.

Single vs Double Precision 89

Table 7.3 Single data type.

Bits Usage
31 Sign (0 =positive, 1 =negative)
30 to 23 exponent, biased by 127
22 to 0 Fraction f of the number 1.f

In floating-point single-precision representations, any stored value requires

32 bits, which are formatted as shown in Table 7.3. Note that, single variables
require less memory storage than double ones, since they are using 32 bits
in spite of 64. However, because they are stored with fewer bits, numbers of
type single are represented to less precision than numbers of type double.
Because the default numeric type for MATLAB® is the double one, the
value of variable x = 26.09 is spread over 64 bit. The function isfloat()
can be used to verify that x is a floating-point number. If this is the case, this
function returns logical 1 (true), logical 0 (false) otherwise. Note that, numeric
data, characters or strings, and logical data can be converted in floating-point
single- and double-precision representations using the MATLAB® functions
single() and double(), respectively.

Creating Floating-Point Data

Floating-point numbers can be used to store values greater than −3.4 ≈ ×1038
or lower than ≈ 3.4 × 1038 . For numbers that lie between these two limits,
both double- or single-precision representation can be equally used, but the
single one requires less memory. Here in the following, a simple way of defining
numbers with several data types in MATLAB® is explained.

Double Precision
Since the default numeric type for MATLAB® is double, any double-precision
number can be declared with a simple assignment statement:

x = 25.783;

The function whos shows the information associated with variable x, that is
a 1-by-1 array of type double which is stored in 8 bytes, as reported in the
following:

>> whos x
Name Size Bytes Class Attributes
x 1x1 8 double

Instead, the function isfloat can be used just to verify that x is a floating-
point number. This function returns logical 1 (true) if the input is a floating-
point number and logical 0 (false) otherwise.
90 Manipulate the Variables Format: Data Types

>> isfloat(x)
ans=
logical
1

Other numeric data, characters or strings, and logical data can be converted
to double precision using the function double(). The example reported here
in the following converts a signed integer number to double-precision floating
point one:

y=int64(-589324077574); % 64-bit integer variable

x=double(y); % y is converted to double
>> x
x=
-5.8932e+11

Single Precision
Conversion functions are needed in MATLAB® to create floating point single-
precision variables starting from the default data type:

x = single(25.783);

The whos() function can be used to store the characteristics (i.e., name, size,
bytes, class, attributes, as shown previously) of variable x in a structure. For
example, the bytes field allows to verify that single variables require 4 bytes
only:

xAttrib = whos(’x’);
>> xAttrib.bytes
ans=
4

Instead, functions double() and single() can convert other numeric data,
characters or strings, and logical data to single precision variables. The ex-
ample reported here in the following converts a signed integer number to
single-precision floating point one:

y=int64(-589324077574); % 64-bit integer variable

x=single(y); % y is converted to single
>> x
x=
-5.8932e+11
Use of Scaling in Fixed Point Representation 91

Table 7.4 Summary of some data types available in MATLAB® .

Signed 8-bit integer −27 to 27 − 1 −128 ÷ 127 int8

Signed 16-bit integer −215 to 215 − 1 −32768 ÷ 32767 int16
Signed 32-bit integer −231 to 231 − 1 − int32
Signed 64-bit integer −263 to 263 − 1 − int64

Unsigned 8-bit integer 0 to 28 − 1 0 ÷ 255 uint8

Unsigned 16-bit integer 0 to 216 − 1 0 ÷ 65535 uint16
Unsigned 32-bit integer 0 to 232 − 1 − uint32
Unsigned 64-bit integer 0 to 264 − 1 − uint64

Other Data types

MATLAB® provides several commands to directly define or convert data in a
specified base and precision. Table 7.4 summarizes the data types which are
mainly used in this book.

7.3 Use of Scaling in Fixed Point Representation

The theoretical procedure to represent real numbers using integer variables
starts by selecting a suitable base value. In digital theory, numbers are typically
computed through powers of 2. Furthermore, it is recommend to make an
intermediate step while doing the conversion, that is, moving from a real value
in SI units XSI to a per-unit (p.u.) representation xpu . Numbers like xpu can
assume values in the range [0, 1]. Namely, they are computed as follows:
XSI
xpu = (7.5)
Xb
where XSI is the real value in SI units to be translated and Xb is the base
(reference) value of the p.u. representation.
Then, xpu is translated as:
XSI Nbit
#x(Nbit ) = xpu 2Nbit = 2 (7.6)
Xb
where Nbit is number of bits of the selected base value and #x(Nbit ) is the
value of XSI translated in integer numbers. In other words, the representation
 92 Manipulate the Variables Format: Data Types

through scaled integer numbers defines a correspondence with values ranging

in one unit.

Example
Given Ib = 20 A as base value and Nbit = 6 bit, translate the current value
into scaled integers considering i(t) = 17.5 A and 10.2 A.

I Nbit 17.5 6
#i6 = 2 = 2 = 56 (7.7)
Ib 20
10.2 6
#i6 = 2 = 32.64 ≈ 32 (7.8)
20
Note that the data conversion is not exact in the second case. Thus, an ap-
proximation error is introduced. Such error is inversely proportional to the
selected number of bits for Nbit .

Regarding the value adopted for Nbit , it is possible to determine the maximum
and the minimum numbers that can be represented. Thus, for Nbit = 6 bit:

int6 : #x6 ∈ [−32, 31] uint6 : #x6 ∈ [0, 63] (7.9)

If a 6-bit base value is selected and a 8-bit data type is used for storing the
translated numbers, there is a margin not to have any overflow in case of a
wrong choice of Xb .
In addition, the choice not to saturate the bits of the processor is appro-
priate while making mathematical computations. If a processor working with
signed int16 (i.e., #x16 ∈ [−32768, 32767]), and a base values of unsigned
12-bit integers (uint12, i.e., #xu12 ∈ [0, 4095]) are considered, the following
computations might be carried out:

• Sum: in the worst case scenario, where both addends assume the maximum
allowed value, it follows that 4095 + 4095 = 8190 < 32767 → no overflow.
• Product: this operation behaves differently with respect to the sum. To
the aim of providing a better understanding to the reader, an example is
exploited starting directly from p.u. representation:

xa = 3.56790

→ xc = xa · xb = 4.54490 (7.10)
xb = 1.27383

Moving into integer numbers, it follows:

#xa,12 = 3.56790 · 212 = 14614


→ #xc,12 = 4.54490 · 212 = 18615
#xb,12 = 1.27383 · 212 = 5217
(7.11)
Converting from Decimal Representation to Single Format 93

Instead, performing the same computations using the converted values:

#xc,12 = #xa,12 · #xb,12 = 76241238 6= 18615 (7.12)

Thus, the two results are inconsistent. This ambiguity is a consequence of
the properties of exponents:

#xc,12 = xa · 212 · xb · 212 = xc · 212·2 → #xc,24 (7.13)

Namely, the base value moves from a 12 to 24 bit representation. Thus, the
right way to avoid such issue is to use a shift operator. Namely:
– Right shift (): division by a power of two, e.g., right shift of 3 bit
→ 1/23
– Left shift (): multiplication by a power of two, e.g., left shift of 3 bit
→ 23
Consequently, the previous multiplication can be performed as follows:

#xc,12 = (#xa,12 · #xb,12 )  12 = 18613 (7.14)

that is, the shift is done over a number of bit equal to that one of the base
and the approximation in the result 18613 ≈ 18615 is the consequence of
the loss of some digits in the shift operation.
Note that the choice to use uint12 as base value for processors working with
int16 allows a tolerance of 4 bit to cover model uncertainties. However, the
partial result of the multiplication #xa,12 · #xb,12 needs a storage in a 32-bit
variable (e.g., int32) since the partial result has a base value of 24 bit.

7.4 Converting from Decimal Representation to Single

Format
Another interesting representation for embedded systems is the decimal (or
base 10) one. This is a representation of non-negative real numbers r in the
form of a series, traditionally written as the sum reported here in the following:
∞
X ai
r= (7.15)
i=0
10i

where a0 is a non-negative integer, and a1 , a2 , . . . are integers satisfying 0 ≤

ai ≤ 9, which are called the digits of the decimal representation.
To convert a base 10 real number into a single (32-bit base-2) format the
outline reported here below should be used:
 94 Manipulate the Variables Format: Data Types

1. Consider a real number with an integer and a fraction part, such as

r = 12.375;
2. Convert and normalize the integer part into binary;
3. Convert the fraction part;
4. Sum the two results and adjust them to produce a proper final
conversion.
Conversion of the fractional part: consider 0.375, i.e., the fractional part
of 12.375. To convert it into a binary fraction, multiply it by 2 first, then
re-multiply the new fraction by 2 until a zero fraction is found or until the
precision limit is reached (which is 23 digits for single format). Thus:
• 0.375 × 2 = 0.750 = 0 + 0.750 → b1 = 0
The integer part is the binary fraction digit. Therefore, 0.750 must be mul-
tiplied again by 2 to proceed;
• 0.750 × 2 = 1.500 = 1 + 0.500 → b2 = 1;
• 0.500 × 2 = 1.000 = 1 + 0.000 → b3 = 1
That is, the fraction part is = 0.000.
Hence, (0.375)10 can be exactly represented in binary as (0.011)2 . Not all dec-
imal fractions can be represented in a finite digit binary fraction. For example,
0.1 does not have any exact representation, i.e., only an approximated trans-
lation can be performed. Coming back to the example under investigation, it
follows that:

(12.375)10 = (12)10 + (0.375)10 = (1100)2 + (0.011)2 = (1100.011)2

Since single data type requires real values to be represented in the format
reported in equation (7.3), 1100.011 is shifted to the right by 3 digits to become
(1100.011)2 ×2−3 . Thus, (12.375)10 = (1100.011)2 ×2−3 , from which exponent
and fraction can be determined:
• The resulting exponent is 3, that is the shift performed previously. Therefore,
e = 130 = 10000010 to obtain 130 − 127 = 3 in the biased form;
• The fraction is 100011.
From these outcomes, the resulting single representation of 12.375 is deter-
mined as:
0 − 10000010 − 10001100000000000000000 = 41460000H

Be careful: if 68.123 is to be converted into single representation using

the above procedure, 42883EF9H is expected as the final result, being the last
4 equal to 1001. However, due to the default rounding behavior of the IEEE
754 format, the result is actually 42883EFAH , whose last 4 bits are 1010.
Processing the Data: Implementation Hints 95

Example
A value of 0.375 is consider. Knowing that 0.375 = (1.1)2 × 2−2 , the exponent
is −2 and in the biased form it is 127 + (−2) = 125 = 0111 1101. The fraction
is 1 by looking at the right part of binary sequence 1.1 (1 = b22 ). Finally, the
resulting single representation of the real number 0.375 is:

0 − 01111101 − 10000000000000000000000 = 3ec00000H

7.5 Processing the Data: Implementation Hints

Floating point MCU based controllers are more expensive then fixed point
ones. Furthermore, a fixed point controller algorithm can be executed on a
floating point MCU, but not vice-versa. The software complexity of industrial
applications is related to the available hardware capability. The latter could
be dependent on, e.g., cost decisions, footprints, and harsh environment. As-
suming to implement a given control algorithm, if a powerful microprocessor
with a large memory allocation is available, the complexity of the computa-
tions is not an issue. Therefore, the best possible accuracy (i.e., floating point
variables) is achieved. On the other hand, cheap platform foresee the em-
ployment of ad hoc or underestimated memory. Then, both the architecture
and even the computations may be revisited to speed up the processing (i.e.,
exploitation of integer variables). Some typical drawbacks of the digital imple-
mentation into microcontrollers are related to the management of data types
and computations among the variables (mathematical operations). In general,
it is possible to build a firmware based either with double or single precision.
However, microcontrollers have always limited computational power. Hence,
complex variable structure such as float and double have slow computa-
tion time. It is important to remember that single variables requires less
memory than the previous ones. Finally, to speed up the processor operation
(and memory usage) further, variables can be translated into integers num-
bers. Note that, given a firmware implementation, possible loss of accuracy
while doing operations with floating-point arithmetic may be caused by the
limitations of the adopted hardware. For example, this latter could not have
enough bits to represent the result.

Summary
Computational efforts required by the design of firmware running on an em-
bedded platform is dependent on the type of required operations and by the
nature of the variable involved in such calculations. Due to the reduced avail-
ability of computational power, it is a best practice to avoid the introduction
of complex numerical structures like float or double in the code. The best
96 Manipulate the Variables Format: Data Types

solution to optimize computational speed consists in representing all the vari-

ables with integer numerical structures (int8, int16, int32) or unsigned inte-
ger (uint8, uint16, uint32) wherever it is possible. For product and/or more
complex operations (e.g., embedded functions), the complexity should be lim-
ited to single variables. It is quite handy to manage data type conversion in
Simulink® by using the data-tye conversion block (it is a standard block). Its
usage is exploited in the next chapters while studying practical cases study.
 Part III

Real-Time Control in
Power Electronics:
Peripherals Settings
Introduction

Every time a firmware foresees the employment of several peripherals, the

user must take care of their scheduling, with particular attention to their
specific settings, timing (e.g., sample time) and the synchronization between
them. For instance, ADC and ePWM peripherals, which are two of the most
important ones for any motor control/power electronic-based firmware, can
be coordinated to be executed according to a specific order/priority. This is
done by using synchronization event (e.g., one ePWM module may trigger the
Start of Conversion of several ADC blocks) and/or interrupt service routines.
Note that, even if a control scheme/firmware is well designed, its behavior is
not ensured without a proper internal timing.1
Nevertheless, several peripherals need to be investigated in order to un-
derstand how a closed-loop scheme can be implemented and which debugging
tools are possible. Simulink® Coder™ generates standalone C and C++ code
from Simulink® schemes. This part describes how to set the Simulink® envi-
ronment, the peripherals and the target hardware (i.e., LaunchXL F28069M
Piccolo™ ) in order to achieve automatic code generation. In all the examples
reported in this book, *.c files are generated and built into executable files
(which can run on the adopted board) automatically from Simulink® . Such
procedure dramatically speeds up design testing and prototyping. In the next
chapters, the peripherals reported in the following table will be presented:

Block Name I/O Firmware / Testing (.slx)

Serial Send O Testing
Serial Receive I Testing
SCI XMT O Firmware
SCI RCV I Firmware
GPIO DI I Firmware
GPIO DO O Firmware
ADC O Firmware
ePWM I Firmware
eQEP O Firmware

1 The Timing and Synchronization between Peripherals is a topic as important as the

design of the controller parameters.

DOI: 10.1201/9781003196938-7 99
 8
Basic Settings: Serial Communication COM
and Hardware Target

Before using the F28069M LaunchPad™ , the communication channel between

the board and the host PC should be set as well as the Simulink® environment.
In particular, the latter settings should cover the compiler selection as well as
hardware targeting for automatic code generation.

8.1 Virtual Serial Communication through COM port

In case of a first time connection between the F28069M LaunchPad™ and the
PC, the steps reported in the following describe how to configure the virtual
COM port of the board:
• Connect the board to the PC through the USB cable provided in the TI
package;
• Open the device manager window (see Figure 8.1 (a) and (b)):

– Open the TI XDS100 Class Auxiliary Channel (or TI XDS100

Channel B in previous Windows releases, see Figure 8.1 (b)) proper-
ties and tick Enable VCP checkbox;
– Disconnect and reconnect the board;
– A new COM port XDS100 Class USB Seral Port (COMx) should
appear in section Port (COM and LTP).

It is possible to redefine the COMx number (e.g., COMx → COM8) by click-

ing on the Advanced... button in the Port Settings pane. Note that the
baud rate 1 inside the COMx settings must not be changed. This value can be
adjusted directly in the Simulink® environment if required.

1 Baud rate refers to the number of signal or symbol changes that occur per second.

To obtain the actual Bit/Rate the formula is: BitRate = BaudRate × nBit where nBit is
the number of bit required to represent that symbol.

DOI: 10.1201/9781003196938-8 101

102 Basic Settings: Serial Communication COM and Hardware Target

(a) (b)

Figure 8.1 Device Manager window (a) and port properties window for
XDS100 Class Auxiliary Port (b).

The virtual COMx is not only used for communication (data exchange)
purposes, but as programming interface too, i.e., letting the JTAG2 Debug
Flow. Indeed, many C2000 MCU boards have a JTAG emulator implemented
on-board and they do not require an external one. Unless there is an appli-
cation requirement, TI suggests using the onboard emulator for development
purposes. The XDS100 and the XDS110 are two target emulators that are
found on TI C2000 Evaluation Modules. For completeness, note that those
emulators are physically connected to:

• XDS100v1 → FTDI FT2232

• XDS100v2 → FTDI FT2232
• XDS110 → TI MCU TM4C1294NCPDRI3R

For the JTAG emulator to communicate the PC, the driver files need to be
installed. Their installation typically occurs co-incident to the installation of

2 The use of JTAG in embedded systems aims to development, debugging, and testing.

JTAG uses a 5-pin implementation in most systems: TDI (Test Data In), TDO (Test Data
Out), TCK (Test Clock), TMS (Test Mode Select), TRSTn (Test Reset).
Virtual Serial Communication through COM port 103

Code Composer Studio. To check that the drivers are successfully installed, it
is enough to connect the PC to the JTAG emulator and power it up.

Remark
The Target Configuration File (.ccxml) contains the information necessary
to connect the PC to target device and the JTAG emulator being used. To
view the current target configurations, select Target Configurations under the
“View” tab in CCS.
9
Simulink® Configuration

Every time a blank Simulink® project is opened, the following steps must be
implemented/verified to assure a correct automatic C-code generation and
firmware deployment on the target microcontroller:

• Open the model configuration parameters menu by clicking on the

Model Settings icon in the Modeling bar1 of Simulink® window;

• Open the code generation panel:

– Select Texas Instruments™ Code Composer Studio (C2000) as

Toolchain setting as shown in Figure 9.1;
– Set the System target file as ert.tlc and C as Language.

• Open the hardware implementation pane:

– Select TI Piccolo F28069M LaunchPad as Hardware board;

Figure 9.1 Simulink® settings: Code Generation.

1 In releases which are earlier than the 2019a one, this icon is in the main menu bar of

the Simulink® window.

DOI: 10.1201/9781003196938-9 105

106 Simulink® Configuration

Figure 9.2 Simulink® settings: Hardware Implementation for TI LaunchXL

F28069M boards.

– Enlarge the Target hardware resources menu;

– Select Build options from the target hardware resources Groups:
∗ set Build action on Build, load and run, as shown in Figure
9.2;
∗ thick boot from flash2 checkbox;
∗ verify the linker command file:
$(TARGET_ROOT)\src\c28069M.cmd
∗ verify CCS hardware configuration file:
$(TARGET_ROOT)\CCS_Config\f28069.ccxml
where $(TARGET_ROOT) automatically set the directory path ac-
cording to the path of the MATLAB® installation. It must be said
that, if a different linker command file and/or CCS hardware configu-
ration file are needed, the full directory path should be specified, which
then becomes a local setting. In this case, the user must remember to
use custom paths every time new Simulink® files are created.

– Select SCI_A from target hardware resources menu (see Section

3.1.4 for the hardware arrangement that enables this channel). More-
over:
∗ define the Desired baud rate in bits/sec for the SCI_A: the
default value is set to 5.625 × 106 bit/s;
– Repeat this last step for SCI_B (see Section 3.1.4 for the hardware
arrangement that enables this channel).
2 This option allows the board to store and load the firmware on the flash memory

onboard the LaunchPad™ . Otherwise, the code is temporarily stored in the RAM memory
of the board not allowing any stand alone executions.
Simulink® Environments: Firmware vs Testing 107

9.1 Simulink® Environments: Firmware vs Testing

Automatically generated C code from Simulink® models greatly improves de-
sign, testing and prototyping speed for any project.
Simulink® Embedded Coder™ generates standalone C/C++ code from
Simulink® models for a wide variety of applications. This Section describes
how to set the Simulink® environment to achieve automatic code generation.
Basically a *.c file is generated and built into executable HEX file (alterna-
tively, binary files can be adopted) which is loaded on the Flash or RAM of
the board through the COMx port,3 and then executed on the MCU.

9.1.1 Overview
The Simulink® environment is used both to design the control system firmware
and to test it giving inputs from the host PC to the microcontroller. In order
to differentiate better the code running inside the MCU from the simulation
test code, two different .slx files are built:
1. Firmware.slx: this is the Programming Environment for the
MCU, where the internal algorithm is built. It defines the pro-
grammable logic that the input/output peripheral has to follow
(see Figure 9.3 (a)). It is mainly subejct to the samplig time Ts . By
means of automatic code generation, this Simulink® file is translated
into executable implementation through the following steps:

•C compiler from the Texas Instruments™ IDE Code

Composer™ Studio is launched along with Simulink® Embed-
ded Coder® ;
•.c and .h files are created and a report interface (Report Gen-
eration) is produced. There, the code is accessible and the com-
piler routine verified, as shown in Figure 9.4;
•The resulting file is downloaded into the MCU memory (Flash
or RAM) and, then, executed.

The Simulink® support package that allows to program the periph-

erals onboard the LaunchPad™ is the Embedded Coder Support
Package for Texas Instruments C2000 Processors. This add-
on installs a library which is entirely dedicated to the C2000-based
MCU boards.
2. Testing.slx: this is the Testing/Debugging Environment that
Simulink® executes on the host PC to send/receive data to/from
the LaunchPad™ , while the loaded firmware is in execution. This
3 UART peripheral, SCI, and COMx port refer to the same subject, i.e., the serial com-

munication between the LaunchPad® board and the Host PC.

 108 Simulink® Configuration

(a)

(b)

Figure 9.3 Firmware (a) and testing (b) environments in Simulink® .

file works as a simulation except that the signal are sent/received

from the embedded board through the COM port (see Figure 9.3
(b)). Indeed, the scheme runs at the simulation step-size Tsim , while
the speed of communication depends to the reception and transmis-
sion time TRX and TT X , respectively (see Section 10.4). Note that
the Testing environment is not always necessary, i.e., a firmware
may not require any serial COM debugging tool.

9.1.2 Execution in Simulink®

Due to this separation between the firmware and the testing code, these two
.slx files are handled in two different ways:

• To load the C-code generated from firmware.slx into the LaunchXL

F28069M, the button Build, Deploy & Start in the Hardware

bar should be clicked;

• To run the simulation file testing.slx, click on the Run button on the
Simulation bar.
MCUs and Real-Time Control with Simulink® 109

Figure 9.4 Code generation report.

Figure 9.5 Example of a real-time control system [15].

9.2 MCUs and Real-Time Control with Simulink®

Since firmware.slx and testing.slx are two separate files, the Simulink® envi-
ronment leads to build a real-time control system, where the MCU (which is
loaded with the firmware file) sends/receives data with the host PC.
Such systems have time dependence on when they need the input to remain
stable. A simple visualization of a type of real-time control system, which
is suitable for the Simulink® environment, is reported in Figure 9.5. A fully
realized controller would be capable of measuring key parameters of the system
(sensing), applying control algorithms to the incoming data (processing), and,
then, affecting the system to achieve the desired change given by the control
algorithm (actuation). Simple exercises, such as single peripheral testing, will
require only a F28069M LaunchPad™ board connected to the host PC to
110 Simulink® Configuration

program the platform through the JTAG cable. The latter will be also used
as a virtual serial port to exchange data and provide touting actions. More
complex exercises requires the effective presence of both converter and load
stages in addition to custom interface boards.
An example of complex control system is provided in Figure 9.6. Here in
the following, the required hardware will be detailed at the beginning of every
exercise.

Figure 9.6 Example of a motor control test bench based on the LaunchPad™
F28069M Piccolo™ programmed by the host PC (see Appendix B).
10
Serial Communication Interface (SCI)
Peripheral

The main board communication is based on a Universal Asynchronous

Receiver-Transmitter interface (UART, also known as SCI), which is a hard-
ware device for asynchronous serial communication1 in which the data format
and transmission speed are configurable. This is used as a serial communica-
tion protocol over the peripheral port of the PC. The electric signal levels and
methods are handled by an external driver circuit.
The UART takes bytes of data and transmits the individual bits in a sequential
fashion. At the destination, a second UART re-assembles the bits into com-
plete bytes. Each UART contains a shift register, which is the fundamental
method of conversion between serial and parallel forms.
Serial transmission of digital information (bits) through a single wire or
other medium is less expensive and complex than parallel transmission, which
requires multiple wires.
For today personal computers (PC), the serial communication interface (SCI)
can be virtually emulated over the USB link. This is the reason why the Enable
VCP (Virtual COM Port) option must be set for the USB peripheral in the
PC device manager. This operation is required for almost all the exercises
presented in this book.
The LaunchXL F28069M Piccolo™ has two SCI modules: A and B. In
particular, the receiver and transmitter of each of them can operate both in
full-duplex2 and half-duplex communication and they support the following
data types: int8, uint8, int16, uint16, int32, uint32, single. The baud-rate
that is chosen in the implementations reported in this book is 115 200 baud/s.
This value can be raised up to 5.625 Mbaud/s, although the communication
can be more prone to errors (data mismatch issues) due to an high volume
of data to be exchanged (i.e., transferred/received) by the MCU within the
sampling time Tx . For instance, if the data transmission from the MCU to the
PC cannot be realized within Tx , the PC may receive fragmented data with

1 Asynchronous serial communication is a form of serial communication in which

the interfaces of the communicating endpoints are not continuously synchronized by a com-
mon clock signal. Instead of a common synchronization signal, the data stream contains
synchronization information in form of start and stop signals before and after each unit of
transmission.
2 Full-duplex communication is adopted in point-to-point systems and it means that

the data can flow in both direction simultaneously.

DOI: 10.1201/9781003196938-10 111

 112 Serial Communication Interface (SCI) Peripheral

Figure 10.1 SCI-CPU Interface [17].

unreadable values. Hence, the transmission rate (both baud rate and sampling
time) should be set considering the computational load given by the firmware.

10.1 Hardware Details

The serial communications interface (SCI) is a two-wire asynchronous serial
port, commonly known as a UART. The SCI modules support digital com-
munications between the CPU and other asynchronous peripherals that use
the standard non-return-to-zero (NRZ) format. The SCI receiver and trans-
mitter have a 4-level deep FIFO each for reducing servicing overhead, and
they have their own separate enable and interrupt bits. Both of them can be
operated independently for half-duplex communication, or simultaneously for
full-duplex communication. As already mentioned, the communication bit rate
can be set to different speeds through a 16 bit baud-select register. Finally,
the SCI checks receive data for break detection, parity, overrun, and framing
errors to specify data integrity. A picture representing the structure of the
SCI interfaces is shown in Figure 10.1. The features of this module are listed
here in the following:

• Two external pins:

– SCITXD: SCI transmit-output pin;

– SCIRXD: SCI receive-input pin.

• Both external pins can be used as GPIO if they are not used for SCI, thanks
to a suitable arrangement of jumpers JP6 and JP7 (see Table 3.2).
• The baud rate is programmable up to 64000 different rates. However,
115 200 baud/s is the standard value adopted in all the exercises reported in
this manuscript.
 Firmware Environment: Send and Receive Data through Serial Communication 113

• Data-word format consists in:

– One start bit;
– Data-word length programmable from 1 to 8 bit;
– Optional even/odd/no parity bit;
– 1 or 2 stop bits;
– An extra bit to distinguish addresses from data (address bit mode only).
• Four error-detection flags: parity, overrun, framing, and break detection.
• Two wake-up multiprocessor modes: idle-line and address bit.
• Half- or full-duplex operation.
• Double-buffered receive and transmit functions.
• Transmitter and receiver operations can be accomplished through interrupt-
driven or polled algorithms with status flags.
• Separate enable bits for transmitter and receiver interrupts (except
BRKDT).
• NRZ3 (non-return-to-zero) format.
• Enhanced features:
– Auto-baud-detect hardware logic;
– 4-Level transmit/receive FIFO.

10.2 Firmware Environment: Send and Receive Data

through Serial Communication
The two blocks reported in the following are used in the firmware environment
to establish the serial communication between the F28069M LaunchPad™
board and the host PC to send and receive data (see Figure 10.2). They
can be found in the Embedded Coder Support Package for Texas Instruments
C2000 Processors library, inside the C2806x subset.

10.2.1 C2806x SCI receive

This block (see SCI RCV in Figure 10.2) corresponds to a code section which
allows to receive data coming from the connected PC to the LaunchPad™
through virtual COM port. In the following, the main settings of this block
are reported.
3 In telecommunication, a non-return-to-zero (NRZ) line code is a binary code in

which ones are represented by one significant condition, usually a positive voltage, while
zeros are represented by some other significant condition, usually a negative voltage, with
no other neutral or rest condition.
114 Serial Communication Interface (SCI) Peripheral
C28x C28x

Data Data

SCI RCV SCI XMT

SCI Receive SCI Transmit

Figure 10.2 SCI Receive (SCI RCV) and Transmit (SCI XMT) blocks to for
programming C2806x processors.

Block Parameters:
• SCI Module
This label sets the SCI module which is adopted for communication. For the
LaunchPad™ F28069M it can be A or B.
Remark: this setting should match with the Serial Send block channel
selected in the Testing Environment (i.e., added inside the testing.slx file,
see Section 10.3.2).
• Additional Package Header
This field specifies the data located before the received data package. It gen-
erally indicates the start of data. The additional package header must be an
ASCII value. A string or number (0–255) can be used. Single quotes around
strings must be entered in this field, but the quotes are not received or in-
cluded in the total byte count. To specify a null value (no package header),
enter two single quotes alone ”. Additional package headers specified here
must match with that one entered in the Serial Send block placed in the
testing.slx file (see Section 10.3.2).
• Additional Package Terminator
This field specifies the data located at the end of the received data package.
It generally indicates the end of data. The additional package terminator
must be an ASCII value. A string or number (0–255) can be used. Single
quotes around strings must be entered in this field, but the quotes are not
received nor included in the total byte count. To specify a null value (no
package terminator), enter two single quotes alone ”. Additional package
terminators specified here must match with that one entered in the Serial
Send block placed in the testing.slx file (see Section 10.3.2).
Firmware Environment: Send and Receive Data through Serial Communication 115

• Data Type
This option specifies the data type of the output data. Available options are
int8, uint8, int16, uint16, int32, uint32 and single.
• Data Length
This value specifies the size of data that the block should receive at each
time step. Numbers greater than 1 indicate that arrays are transferred. The
data length must match the input data length, which is sent through the
Serial Send block in the testing.slx file (see Section 10.3.2).
• Initial Output
This parameters sets the output default value from the SCI Receive block.
This value is used, for example, if a connection time-out occurs and the
Action taken when connection timeout field is set to Output the
last received value, but nothing has been received yet.
• Action taken when connection times out
This label allows to specify the output in case a connection time-out occurs.
If Output the last received value is selected, the outputs return the last
received value. If no value is received, the SCI Receive block outputs return
the Initial output value. If Output custom value is selected, the custom
value is set through the parameter Output value when connection times
out.
• Output value when connection times out
Parameter that sets the output time-out custom value.
• Sample Time
This field specifies the sample time Ts in seconds for input sampling of the
block. To inherit sample time and to execute this block asynchronously, set
it equal to −1.
• Output receiving status
By enabling this feature, a status block providing the status of the serial
communication is created.
• Enable Receive FIFO interrupts
If this option is selected, an interrupt is posted when the FIFO is filled with
a specified amount of data, allowing the subsystem to perform any action.
For example, the C28x Hardware Interrupt block can be used for triggering
the SCI Receive block to read the data as soon as it is received. If the option
is not selected, the SCI Receive block is in polling mode and it checks the
FIFO for data. If any data is present, the block reads and outputs it. If data
is not present and blocking mode is enabled, the block waits until data is
available. In non-blocking mode, the block goes on with the execution of the
algorithm without waiting for any received data.
116 Serial Communication Interface (SCI) Peripheral

Figure 10.3 Examples of Rate Transition and Data Type Conversion blocks
connected to a SCI Transmit block.

• Receive FIFO interrupt level

This parameter appears only if Enable Receive FIFO interrupts is
ticked. When it is enabled, the receive FIFO generates an interrupt when the
number of received data bytes is greater than or equal to the value selected
for this parameter.

10.2.2 C2806x SCI transmit

This block (see SCI XMT in Figure 10.2) allows the transmission of data from
the board to the host PC Through SCI. For example, this communication can
be used to visualize the values acquired by the ADC converters inside the
onboard sensors in the Simulink® environment. In the following, the main
settings of this block are reported without any further explanations. Indeed,
they are the same as those described for the C2806x SCI Receive block (see
Section 10.2.1).

Block Parameters:
The block parameters of C2806x SCI Transmit are less than C2806x SCI
Receive, but with the same meaning. In particular they are: SCI Module,
Additional Package Header, Additional Package Terminator, Enable transmit
FIFO interrupts, Transmit FIFO interrupt level. See the previous section for
further details.

Remarks
The transmission rate as well as the trasnmitted data format/length are not
internally set by this SCI blocks. These have to be (hardly) fixed through spe-
cific Simulink® blocks, such as Rate Transition and Data Type Conversion,
according to the firmware execution. This is summarized in Figure 10.3, which
shows a couple of possible arrangements related to data exchange required in
the following exercises. Note that, multiple data/variables can be sent exploit-
ing Mux blocks (data is organized in [1 n] arrays). Double-precision floating
point data is not recommended due to its large size. This is particularly critical
when more than one variable have to be transmitted/received.
Finally, note that no COM port is specified in SCI RCV or XMT, since those
blocks are translated into code which is downloaded directly on the hardware
board.
Testing Environment: Send/Receive Data through Serial Communication 117

COM5
115200 Data COM5 COM5 Data
8,none,1

Serial Configuration Serial Send Serial Receive

Figure 10.4 Serial Configuration, Send, and Receive blocks for setting serial
communications between the testing.slx file and the LaunchPad™ F28069M.

10.3 Testing Environment: Send/Receive Data through

Serial Communication
The set of blocks listed here in the following (and shown in Figure 10.4) are
used in the Testing Environment to allow interaction between the host PC and
the F28069M LaunchPad™ board through the Simulink® software. Moreover,
debugs of the generated code deployed on the board can be done in a separate
simulation file. Such blocks are available in the Instrument Control Toolbox™
library, as already mentioned in Section 4.3.

10.3.1 Serial configuration

This master block (see Figure 10.4) configures the parameters for the serial
communication, which is carried out together with the Serial Receive and
Serial Send blocks over COMx port. Its main settings are reported here in the
following.

Block Parameters:
• Communication Port
This setting specifies which available COMx port is exploited for serial com-
munication. This choice must match with the COM port number selected in
the Serial Send and Serial Receive block as well as with the VCP generated
in the PC Device Manager.
• Baud rate
This option allows to set the rate at which signals are transmitted for the
serial interface. The default value is 9600 baud/s. This number should match
the one specified in the Model Configuration Parameter window.
• Data bits
This label specifies the number of bits of a single chunk of data (i.e., baud)
that are transmitted over the serial interface. The default value is 8 and
other available values are 5, 6, and 7.
• Parity
This field allows to specify the desired check of parity bits in the data bits
118 Serial Communication Interface (SCI) Peripheral

transmitted through serial port. This is set to none by default. The other
available choices are:

– Even: parity bit is set equal to 0 if the number of ones in a data set
of bits is even;
– Odd: parity bit is set equal to 1 if the number of ones in a data set of
bits is odd;
– Mark: parity bit is always set equal to 1;
– Space: parity bit is always set equal to 0.

• Stop bits
This option specifies the number of bits used to indicate the end of the chunk
of data. The selected number of data bits impacts on this parameters since
it determines the choices available for the stop bits. If data bits is set equal
to 6, 7 or 8, then, the default value is 1 and the other choice at disposal is
2. Otherwise, the only available number is 1.5.
• Byte order
This label sets the byte order. If byte order is LittleEndian (default value),
then the instrument stores the least significant byte in the first memory
address. Conversely, if byte order is set to BigEndian, the most significant
byte is placed first in memory and the least significant one last.
• Flow control
This field specifies the process of managing the rate of data transmission on
the serial port. Flow control can be set to none or hardware, depending if
the user wants to have no flow control or to let the hardware determine it.
• Timeout
This setting specifies the amount of time that the model waits for the data
during each simulation time step. The default value is 10 s.

10.3.2 Serial send

This block (see Figure 10.4) is needed to send data from the host PC to
the LaunchPad™ F28069M board via serial communications interface (SCI)
through the COMx port. In the following, its main settings are reported.

Block Parameters:
• Communication Port
This field allows to specify an available COM port. This choice must match
with the COM port number selected in the Serial Configuration block (see
Section 10.3.1).
Testing Environment: Send/Receive Data through Serial Communication 119

• Header
This label specifies additional data placed at the beginning of the infor-
mation frame to be transmitted before sending it through serial port. By
default, <none> or no header is specified. The header entered here must
match with that one chosen in the SCI Receive block in the firmware.slx file
(see Section 10.2.1).
• Terminator
This label specifies the additional data placed at the end of the information
frame to be transmitted after sending it over the serial port. By default,
<none> or no terminator is specified. The terminator entered here must
match with that one chosen in the SCI Receive block in the firmware.slx file
(see Section 10.2.1).
• Enable blocking mode
This flag allows simulation halting while sending data. Namely, the Serial
Send do not send any other data until a signal is received. This option
is selected by default. It is cleared in case it is not desired to have write
operations that should pause the simulation.

Remark:
A Data Type Conversion block should be connected before the Serial Transmit
block to avoid the transfer of any double-precision floating point data. The
setting of this additional block should match the data type selected in the
option panel of the SCI Receive block in the Firmware Environment (see
Section 10.2.1).

10.3.3 Serial receive

This block (see Figure 10.4) is intended to receive data to the host PC via
COM port from the MCU. Its main settings are reported here in the following.
Some parameters are not explained since they are the same as those described
for the Serial Send block (see Section 10.3.2).

Block Parameters:
Some parameters of Serial Receive block are the same as those reported in the
previous Section, i.e., Communication Port, Header, Terminator. The other
parameters are:

• Data Size
Through this label, it is possible to specify the size of the transmitted data.
Both 1-D vectors and matrices are accepted, e.g., [1 1] corresponds to a
scalar value, while [1 3] corresponds to an array of size 3.
120 Serial Communication Interface (SCI) Peripheral

• Data Type
This label specifies the type of the received signal (i.e., int8, uint8, int16,
uint16, int32, uint32 and single). This setting should be coherent with the
corresponding one specified in the Data Type Conversion block connected
to the SCI XMT block in the firmware environment (see Section 10.2.2).
It is important to note that the board F28069M can handle single-precision
floating-point numbers, but this choice implies an heavier transmission and
computational load. Hence, this kind of transmission likely overloads the
communication buffer. To avoid this problem, int8/uint8 or int16/uint16
data types are preferable in case many signals have to be transferred.
• Action when data is unavailable
This option can be set choosing among one of the following actions: Output
last received value, Output custom value or Error. It is very similar
to the setting reported in Section 10.2.1, i.e., Action taken when connection
times out for firmware environment (see Section 10.2.1).
• Custom value
This parameter is enabled only if Action when data is unavailable is
set on Output custom value. It specifies a value to output when no data
is received. The default value is 0.
• Block sample time
This setting specifies the sample time of Serial Receive during simulation.
Namely, it indicates the rate at which the block is executed during simula-
tion. This value must be specified in seconds.

Remarks
Firmware and Testing Environments have common features: while both SCI
RCV/Serial Receive blocks must specify Sample Time and Data Type, SCI
Transmit and Serial Send do not have these options available and they must
be checked and possibly adapted to be synchronized with other blocks by
employing the Data Type Conversion block, the Rate Transmission block,
and Mux block in case of multiple data transmission.

10.4 Time Variable Settings (Sample Rates)

While designing both the firmware and testing environment for control sys-
tems, the reader comes across different sample rates/times options which must
be set correctly (they are worth of much attention too). At this point of the
book, a brief summary of those times is reported in the following:
Time Variable Settings (Sample Rates) 121

• CPU Clock - Tck

Typically fixed by the internal or external board oscillator; for the
LaunchPad™ F28069M it is equal to Tck = 11 ns, which means a clock
frequency of fck = 90 MHz.
• Integration Time - Ts
This value can be edited by the user in the Configuration Parameters
window in the firmware.slx file. This number represents the minimum inte-
gration time used to discretize the system and to perform the computation.
In this book, it is mainly identified with Ts . This value must be set according
to the complexity of the code running on the MCU. Given that Ts defines
the minimum, other sampling times must be multiple of Ts . Indeed, an al-
gorithm might have parts of the code running at different rate, from the
slowest (not critical) up to the fastest (highly critical).
• Sample Time Tx - TTX
This parameter represents the time rate at which data are sent to the Serial
Send block and stored in the TX Serial Buffer during the transmission of
signals to the LaunchXL F28069M board. The same occurs when data is
sent from the MCU to the host PC via the SCI XMT. In both cases, it is set
through Rate Transition blocks. This parameter must match with the SCI
Receive sample time. In this book, TTX = 0.01 s is mainly used .
• Sample Time Rx - TRX
It represents the Time Rate at which data are sampled from the RX Serial
Buffer during the reception of signals from the LaunchPad™ . This parameter
can be set in the Serial Receive block and it must match the SCI Transmit
sample time. According to the firmware requirements, the time range could
vary from TRX = 0.01 s, TRX = 0.001 s, . . . , up to TRX = Ts .
In particular, the most significant consequences of a wrong choice for
TTX /TRX (or data type) settings are mismatches such as:

– Overloading of the transmission buffer causing data spikes and/or

wrong values;
– Simulation time much slower than real time (depending on the capa-
bilities of the host PC), causing large delay between sent and received
data within a given time span.

• Fixed Step-Size - ∆t
Regarding the firmware environment, every firmware.slx relies in a discrete
framework. Thus, the book case studies will focus on Simulink® schemes
which adopt a fixed-step solver option with a discrete solver such as ODE4
(Runge-Kutta) and a fixed-step size Ts , i.e., ∆t = Ts .
In the simulation environment (i.e.,testing.slx), the simulation starts at t = 0
and ends at t = Tend . The simulation time step Tsim may be fixed or vary
122 Serial Communication Interface (SCI) Peripheral

Figure 10.5 Connection between the F28069M LaunchPad™ board and a host
PC.

during each run according to the selected solver options in Simulink® , i.e.,
discrete- or variable-step size, respectively. Since the firmware is designed
in a discrete-time domain, the same is done for the testing environment,
e.g., keeping ∆t = Tsim = Ts or ∆t = Ts /10 (to increase the visualiza-
tion resolution). Note that, every time the enabling blocking mode option
is enabled in the Serial Send block, the simulation environment waits for
the data acquisition before increasing the Simulation Time by one step (i.e.,
ti+1 = ti +∆t), resulting in simulation times which are totally different from
the real time. Namely, the more the testing environment receives data, the
longer the simulation will take to run.

10.5 Examples on Serial Communication

In order to understand the difference between firmware and testing environ-
ments, two simple exercises are proposed here in the following.

Example 1

Build a firmware aimed at generating an internal signal from the

board F28069M and send it to the PC through the serial port

A step-by-step procedure is reported here below to guide the reader through

this first exercise. The LaunchPad™ F28069M is programmed in such a way
to transmit an internally generated signal to the host PC through the virtual
 Examples on Serial Communication 123

serial port COMx. SCI module A is used. Thus, the right jumper configuration
should be realized on the board (see Table 3.2).

Firmware Environment
c28069_internalSin_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window:

– Select these two settings in the solver selection sub-menu:

∗ type = fixed step;
∗ solver = discrete (no continuous states);
– Set the fixed-step size equal to Ts = 100 µs.

• Place a SCI Transmit block in the file and double click on it to modify its
parameters:

– Select SCI module: A;

– Remove both additional package header ’S’→” and terminator
’E’→”.

• Insert a Sin Wave block to generate the internal signal. Double click on it:

– Set sine type: sample based;

– Set time: use simulation time.  
Namely, the implemented sine wave is y(k) = A sin 2π nks + b.
– Set the gain equal to A = 0.5;
– Set the bias equal to b = 0.5;
– Set the number of samples per period ns = 100;
– Set number of offset samples = 0;
– Set sample time Tsig = 1000Ts = 0.1 s.

• Insert a Pulse Generator block to have a possible alternative to the sine-

wave at disposal. Double click on it:

– Set pulse type: sample based;

– Set time: use simulation time;
– Set the gain equal to A = 0.5;
– Set the number of samples per period ns = 100;
– Set the pulse width 50 → 50 %, i.e., equally spaced square-wave;
 124 Serial Communication Interface (SCI) Peripheral
pulses
SCI A
C28x

Data

0.01 SCI XMT

sine

Figure 10.6 Simulink® scheme included in c28069_internalSin_F.slx.

– Set phase delay = 0;

– Set sample time Tsig = 0.1 s.

• Insert a Manual Switch block to change the source input.

All system parameters have to be defined and initialized by a m-script before
the firmware deploy. This can be done inside Simulink® , e.g., in Model Prop-
erties/Main/InitFcn or as a separate MATLAB® code. In this exercise, most
of the parameters are manually defined in the each block settings, while only
Ts and Tsim require a global assignment. For simplicity, both of them will be
defined directly in Simulink® for either firmware and software environment.
The first internal signal is a sample based sine-wave. Namely, the value of
k is increased every Tsig returning the actual value of y(k) at each time step,
until the reset condition is reached k = ns . Hence, the sine-wave period is equal
to T = Tsig · ns = 10 s and its corresponding frequency is f = 0.1 Hz. Note
that Tsig is a mutiple of Ts . Since the F28069M LaunchPad™ board includes
a floating-point unit, single data type could be used to send floating-point
signals through the COMx. This assignement is done via the Data Type
Conversion block. A Rate Transition block is used to fix the sample rate
of the SCI XMT to TTX = 100Ts = 0.01 s. Finally, connect the aforementioned
blocks as shown in Figure 10.6. To translate the file c28069_internalSin_F.slx
into C-code language and to run it on the board F28069M, this Simulink®
scheme should be compiled, built, deployed and run by clicking on Build,

Deploy & Start icon.

By double clicking on the manual switch icon, the input signal can be changed
to the pulse generator. This block works analogously to the sine-wave one,
thus, its explanation is not repeated here.
Be careful: the sine-wave can be edited any time by acting on the parameters
of the Sin Wave block. To make these changes effective on the board, the
translation in C-code and the deploy to hardware must be repeated every
time. The same holds for the Pulse generator block.

Testing Environment
c28069_internalSin_T.slx (solver: fixed step - ODE4, step size: Tsim = Ts )

Open a new blank Simulink® project:

 Examples on Serial Communication 125

COM14
115200 COM14 Data
8,none,1 out

Figure 10.7 Simulink® scheme included in c28069_internalSin_T.slx.

• Open the Model Configuration Parameters window:

– Select these two settings for the solver selection sub-menu:

∗ type = fixed step;
∗ solver = discrete (no continuous states);
– Set the fixed-step size equal to Ts = 100 µs;
– Set the stop time equal to inf.

• Insert a Serial Configuration block. Double click on it:

– Select the available COMx port as communication port;

– Define the baud rate equal to 115200. This value must match
with that one defined in the Model Configuration Parameters in the
c28069_internalSin_F.slx file (see Chapter 9);
– Set data bits: 8;
– Set stop bits: 1;
– Set byte order: LittleEndian;
– Set timeout: 10.

• Insert a Serial Receive block. Double click on it:

– Select the right COMx as serial port;

– No header or terminator should be specified;
– Set data size [1 1], since one signal is sent, received and visualized
only;
– Set data type: single;
– Set sample time: 0.1 s.

Note that the data type sent from the board to the PC is single. Thus, single
is set as data type in the Serial Receive parameters without any need for ad-
ditional data conversion. The resulting sine-wave can be visualized through a
Scope block. Finally, connect these blocks as shown in Figure 10.7. In particu-
lar, this scheme can also include a Dashboard Scope with signal associated
to the output of the Serial Receive. This is a simple way to dynamically vi-
sualize signals over time. This approach is suggested for the next exercises
included in the book which requires SCI as debugging tool. For simplicity rea-
sons, Dashboard Scopes will be included in the testing environments, but not
126 Serial Communication Interface (SCI) Peripheral
0.6

1
0.5

0.8
0.4

0.6 0.3

0.4 0.2

0.1
0.2

0
0

-0.1
5 10 15 20 5 10 15 20

(a) (b)

Figure 10.8 Data read from the SCI A when the internal signal is generated
by Sine Wave (a) and Pulse Generator (b) blocks.

reported in any figure. Moreover, the Testing Environment does not need to
deploy code to target. Indeed, this file should be run only as standard simu-
lations to start sending data through the serial port. In order to continuously
run the simulation, the stop time is set equal to inf (i.e., infinity). It is impor-
tant to note that the serial configuration is the “main” block which sets the
parameters of the COMx communication, possibly overwriting previous set-
tings edited in the PC device manager. The results of this exercise are shown
in Figure 10.8.

Example 2

Build a firmware aimed at processing a signal received from serial

communication by the MCU and, then, send it back to the host PC

In this example, the F28069M LaunchPad™ board must be able both to read
an external command and send it back through the virtual serial COMx port.
SCI A module is used. Thus, the right jumper configuration should be realized
on the board (see Table 3.2).

Firmware Environment
c28069_externalSin_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted in Example 1.
 Examples on Serial Communication 127
SCI A Rx SCI A Tx
C28x C28x

Data Data

SCI RCV SCI XMT

Figure 10.9 Simulink® scheme included in c28069_externalSin_F.slx.

• Place a SCI Transmit block in the file and and use the same settings
adopted in Example 1. The Rate Transition block is set with an Output
port sample time equal to 0.01 s.
• Insert a SCI Receive block in the file and double click on it to modify its
parameters as follows:

– Select SCI module: A;

– Remove both additional package header ’S’→” and terminator
’E’→”;
– Set data type: single;
– Set data length: 1;
– Set initial output: 0;
– Set sample time: 0.1 s.

• Add a gain block and set its value equal to 2. Moreover, set the output data
type as single in the Signal Attributes tab.

Finally, connect these aforementioned blocks as shown in Figure 10.9. In this

example, the board F28069M reads an external signal from the COMx port,
multiplies it through a gain and sends it back to the PC. The signal coming
from the COMx port is bounded between [0, 1] and it is in single-precision
floating point format. A data type conversion is not necessary because the
SCI RCV is already set to handle single data, while the sample time is set to
0.1 s, which is equivalent to 10 Hz communication frequency (sample & hold
effect on the processed signal).

Testing Environment
c28069_externalSin_T.slx (solver: fixed step - ODE4, step size: Tsim = Ts )

Open a new blank Simulink® project:

• Open the Model Configuration Parameters window and use the same
settings adopted in Example 1;
• Place a Serial Configuration block and use the same settings adopted in
Example 1;
• Insert a Serial Send block. Double click on it:
 128 Serial Communication Interface (SCI) Peripheral

COM14
Data COM14 115200 COM14 Data
8,none,1
sine1

Figure 10.10 Simulink® scheme included in c28069_externalSin_T.slx.

– Select the right COMx port as serial port;

– Do not insert any header or terminator.

• Add a Serial Receive block and use the same settings adopted for Exam-
ple 1;
• The external signal is generated by a Sin Wave block with the same pa-
rameters set in Example 1

In this example, the communication through COMx involves single-precision

floating point numbers and it is full-duplex. Hence, both the Serial Receive
and the SCI Receive blocks are set to work with single data type without the
need for any additional data conversion. The input and output sine-waves can
be plotted in the same Scope block through a multiplexer (MUX) block. Fi-
nally, connect these blocks as shown in Figure 10.10. The resulting waveforms
are reported in Figure 10.11.

Computational delay: it is important to note that there is a slight shift

between the two waveforms. In general, this computational delay is due to
both the finite processing time of the microcontroller (Tck ) and the sample
time of the serial communication (TRX ). In this case, the latter is much larger

1.5

0.5

10 11 12 13 14 15 16 17 18 19 20

1.2

0.8

0.6

0.4
10 10.05 10.1 10.15 10.2 10.25 10.3 10.35 10.4 10.45 10.5

Figure 10.11 Transmitted/received sine-wave signals over time with a detailed

view of the computational delay which causes the shift between them.
Examples on Serial Communication 129

than the former. Hence, this computational delay is mainly due to the serial
communication. This is detailed by a zoom view reported in the lower side of
Figure 10.11, allowing to visualize TRX .
 11
GPIO Peripheral—Digital Input/Output

This Chapter describes the general-purpose input/output (GPIO) peripheral

and how to deal with digital input/output pins. Indeed, the GPIO peripheral
has general-purpose digital state pins that can be configured as both inputs
and outputs. The peripheral has a resolution of 16 bits and its internal circuits
(and logic) work in the voltage range of 0 − 3.3 V.
When configured as an output, an internal register can be written to con-
trol the state driven on the corresponding pin. When configured as an Input,
it is possible to detect its level by reading the state of an internal register.
Both reading and writing operations for those registers can be easily done in
Simulink® through suitable blocks.
The most common settings foresees the assignment of different pins of
the LaunchPad™ board (i.e., among the 80 pins accessible on top of it) to
each GPIO channels. For instances, a GPIO peripheral may be used also for
internal purposes to control some communication features (e.g., CAN, SPI,
etc..). Moreover, some GPIO channels share the external pin of the LaunchXL
F28069M with other peripherals (e.g., ePWM, eQEP, eCAP). Hence, a mul-
tiplexing mechanism is adopted to select the right function on the shared pin.
The assignment of some pins can be reconfigured in Simulink® according to
specific necessities. The default pin-out of the adopted board is reported in
Figure 3.4 and 3.5.

11.1 Hardware Details

The GPIO pins are multiplexed by the GPxMUXn registers, which selects the
operation to be performed on the shared pins. On this board, they are labeled
with the GPIO signal to which they are connected, that is, GPIO0-GPIO58.
Moreover, the LaunchPad™ board has three I/O ports:

• Port A which includes GPIO0-GPIO31 signals;

• Port B which includes GPIO32-GPIO58 signals;
• One analog port, which includes AIO0-AIO15 signals.

DOI: 10.1201/9781003196938-11 131

132 GPIO Peripheral—Digital Input/Output

In particular, the pins of the analog port are connected to the channels of
the analog-to-digital converter peripheral ADCA0-7 and ADCB0-7. Further
details on this peripheral are reported in the next Chapter.
These pins can be operated as digital I/O (referred to as GPIO) or con-
nected to one up to three I/O peripheral signals via the GPxMUXn registers.
If digital I/O mode is selected, the pin “direction” (i.e., input or output) is
configured via the GPxDIR registers. It is also possible also to qualify the
input signals to remove unwanted noise via the GPxQSELn, GPACTRL, and
GPBCTRL registers. Note that most of the pins are re-configurable (i.e., they
are associated to many registers). However, Simulink® provides a limited ac-
cess to them. Indeed, only some of them can be edited through the dedicated
block library. For a more flexible customization of the code, the reader is
suggested to use Code Composer Studio™ .
Every time a reconfiguration of the I/O pins is needed, it is suggested to
follow these steps:
1. Organize a pin-out table.
2. Enable or disable internal pull-up resistors.
To enable or disable the internal pullup resistors, write to the re-
spective bits in the GPIO pullup disable (GPAPUD and GPBPUD)
registers. For pins that can operate as ePWM output pins, the in-
ternal pullup resistors are disabled by default. All the other GPIO-
capable pins have the pull-up resistor enabled by default. The AIOx
pins do not have internal pull-up resistors. It must be noted that
the pin register access in its full extent is available only through
Code Composer Studio™ .
3. Select input qualification.
In case a pin is used as an input, it may be needed to specify its input
qualification. This information should be set in the GPACTRL, GP-
BCTRL, GPAQSEL1, GPAQSEL2, GPBQSEL1, and GPBQSEL2
registers. By default, all the input signals are synchronized by
SYSCLKOUT.
4. Select pin function.
Configure the GPxMUXn or AIOMUXn registers such that each pin
to be re-configured is connected to one of the available peripheral.
By default, all GPIO-capable pins are configured at reset as general
purpose input pins.
5. For digital general purpose I/O, select direction of the pin.
Every time the pin is configured as a GPIO, the direction of the pin
should be specified as either input or output in the GPADIR, GPB-
DIR, or AIODIR registers. By default, all GPIO pins are inputs. To
change the pin from input to output, first load the output latch with
the value to be driven by writing the appropriate value to the GPx-
CLEAR, GPxSET, or GPxTOGGLE (or AIOCLEAR, AIOSET, or
Firmware Environment: GPIO Peripherals 133
C2806x

GPIOx

GPIO DI

Figure 11.1 C2806x GPIO Digital Input block.

AIOTOGGLE) registers. Once the output latch is loaded, change

the pin direction from input to output via the GPxDIR registers.
The output latch for all pins is cleared at reset.
6. Select low power mode wake-up sources:
Specify which pin is able to wake the device up from HALT and
STANDBY low power modes. The pins are specified in the GPI-
OLPMSEL register.
7. Select external interrupt sources.
Specify the source for the XINT1 - XINT3 interrupts. For each in-
terrupt, one of the port A signals can be specified as the source. This
is done by specifying the source in the GPIOXINTnSEL register.
The polarity of the interrupts can be configured in the XINTnCR
register.

11.2 Firmware Environment: GPIO Peripherals

In this Section, the two GPIO blocks available in the Embedded Coder Support
Package for Texas Instruments C2000 Processors library (C2806x subset) are
presented.

11.2.1 C2806x GPIO digital input (GPIO DI)

This block (see Figure 11.1) allows to configure a GPIO pin as a digital input.
Namely, it sets the GPxMUXn on digital input (DI). Once a pin is configured
as GPIO DI, it becomes unavailable for digital output or other peripheral oper-
ations. The GPIO peripheral acquires a voltage signal from the corresponding
pin Vpin and it is interpreted as high (True) or low (False), according to the
voltage thresholds reported previously.
The main settings of this block are reported here in the following.

Block Parameters:
• GPIO Group
GPIO signals are organized into groups. This setting allows the selection of
the group of pins to view or configure. Changing group, different pins can
 134 GPIO Peripheral—Digital Input/Output
C2806x

GPIOx

GPIO DO

Figure 11.2 C2806x GPIO Digital Output block.

be assessed by ticking the corresponding check-boxes. Only pins that belong

to the same GPIO group can be selected simultaneously. The reader should
refer to the pin-out reported in Figure 3.5 for the pin/GPIO correspondence.
• Sample Time
This field specifies the time interval that elapses between two consecutive
output samples. Set this parameter equal to -1 to inherit sample time from
the adopted solver.
• Data Type:
This parameter sets the data type of the input, which is read as a 16-bit
integer and, then, it is cast to the selected data type. Valid data types
are auto, double, single, int8, uint8, int16, uint16, int32, uint32 or boolean.
Regardless of the data type, high states are translated as the logic state 1
and low ones as logic state 0. Indeed, the state of the pins can be either high
or low, depending on the voltage applied on it Vpin :

– Voltage inputs between 0 and 1.3 V lead to LOW output values, i.e.,
logic 0;
– Voltage inputs over 1.5 V lead to HIGH output values, i.e., logic 1;
– Voltage inputs in the range [1.3, 1.5] V lead to random output values,
i.e., logic 0 and 1.

Remark
Since voltage inputs in the range [1.3, 1.5] V lead to random logic states, it is
recommended to keep Vpin quite far from the high/low voltage threshold.

11.2.2 C2806x digital output (GPIO DO)

This block (see Figure 11.2) allows to configure a GPIO pin as a digital out-
put. Namely, it sets the GPxMUXn on digital output (DO). Once a pin is
configured as GPIO DO, it becomes unavailable for digital input or other
peripheral operation. Digital output GPIO blocks receive an input signal in
Simulink® and it sends an output voltage on the corresponding pin:

• 0 (low/false signal): Vpin = 0 V (pin grounded);

• 1 (high/true signal): Vpin = 0 V (pin pulled high).
Examples with GPIO blocks 135

Every digital output GPIO has a toggle option that inverts the output
signal on the pin. The main settings of this block are reported here in the
following. One parameter is not explained since it is used to set the same
function described for the C2806x GPIO Digital Input block.

Block Parameters:
• GPIO Group
• Regular- and Toggle mode
For each selected pin, regular- (default) or toggle mode can be selected to
drive the GPIO signal. In regular mode, true inputs pull the GPIO pin high.
False ones ground the pin. In toggle mode, true values switch the pin output
level continuously. Thus, a true input switches the output voltage from high
to low and vice-versa. Instead, if the input is set to false, the pin output
level is unaffected. Namely, the last logic value is kept constant as long as
the the input is kept equal to 0.

Remark
The switching frequency in toggle mode is limited by the 20 MHz rate of the
pin buffers. Regardless of the input data type, every time a numerical signal
is applied to a GPIO in regular-mode, this value is interpreted as:

– Low if the signal applied to the GPIO is a 0 numerical value;

– High if the signal applied to the GPIO is any numerical value 6= 0.

11.3 Examples with GPIO blocks

Example 1

Build a firmware which is able to drive the right GPIO signal to

turn the blue led mounted onboard the LaunchPad™ board on/off

The led turn on/off is the “hello world” for micro-controller progrmming. This
simple exercise is essential to understand the electrical characteristics of the
microcontroller in use. The schematics of the LaunchXL F28069M board as
well as its pin-out can be found in the corresponding User’s guide [16]. From
this file, it can be noted that this board includes two programmable leds
which are physically connected to two digital pins: a blue (GPIO39) and red
(GPIO34) one. An initialization script can be used to load some constants,
e.g., in Model Properites/Callbacks/InitFcn.
 136 GPIO Peripheral—Digital Input/Output
3.3 V

C2806x

GPIOx

GPIO DO R
led blue GPIO39

Out pin D10

(a) (b)

Figure 11.3 Simulink® scheme included in c28069_ledB_F.slx (a) and phys-

ical connection between the diode D10 and the corresponding output pin (b).

Firmware Environment
c28069_ledB_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5;
• Insert a Digital Output (GPIO DO) block and double-click on it:

– Select GPIO Group GPIO32∼GPIO39;

– Flag GPIO39;
– It is possible to test the firmware without enabling Toggle mode and
then with it to appreciate how the output is affected. Note that once
every time a pin flag is modified, the build and deploy operation must
be repeated.

A constant block is used to produce an internal command signal which drives

the GPIO39 with 0 or 1 values. The sample time of the constant block is set
equal to Tsig = 1000Ts = 0.1 s. The led is turned on/off accordingly. Since the
resolution of GPIO is 16 bit, a conversion using a Data Type Conversion block
must be performed to obtain an uint16 format. Note that even a different data
type might be used. Finally, all the aforementioned blocks should be connected
as shown in Figure 11.3 (a). It is worth noting that Figure 11.3 (b) highlights
that the circuit works with a negative logic. Namely, the led anode is connected
to Vcc = 3.3 V through a resistor, while the cathode is in contact with GPIO39.
As a consequence, a high value for GPIO39 implies an high pin voltage, which
keeps the led off since no difference in potential (voltage) is present . On the
other hand, a zero value implies a zero pin voltage, which turns the led on.
Figure 11.4 shows a typical voltage waveform that can be measured on the
GPIO pin every time it is driven in toggle mode and its input is set equal
to 1.
 Examples with GPIO blocks 137

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Figure 11.4 Example of GPIO39 output voltage over time in toggle mode.

Remark
The input value of GPIO39 can be edited by changing the value of the constant
block. The translation in C-code and the deploy to hardware must be repeated
at each modification of this value.

Example 2

Build a firmware which is able to drive the right GPIO signals to

to make the red and the blue leds blink alternately

In this exercise, the discrete time Delay block is exploited to set a certain
delay of Nd time steps for the driving signal of one led. The time delay in
seconds is computed as Nd Tsig , where Tsig = 1000Ts and Ts is the fixed step
size of the solver.
In order to make the leds blink, the toggle mode must be enabled in the
GPIO DO block parameters window. The interleaved blinking is achieved by
selecting an odd time delay (e.g., Nd = 1 which imply one step delay)

Firmware Environment
c28069_ledBR_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5; the constant block value
must be equal to 1 in order to use the toogle mode.
• Insert a Digital Output (GPIO DO) block and double click on it:

– Select GPIO Group GPIO32∼GPIO39;

– Flag GPIO39;
 138 GPIO Peripheral—Digital Input/Output

led red
C2806x

GPIOx

GPIO DO
C2806x

Z-1 GPIOx

GPIO DO
led blue

(a) (b)

Figure 11.5 Simulink® scheme included in c28069_ledBR_F.slx (a) together

with the red (D9) and blue (D10) leds blinking accordingly (b).

– Enable Toggle mode.

• Insert a new Digital Output (GPIO DO) block and double-click on it:

– Select GPIO Group GPIO32∼GPIO39;

– Flag GPIO34;
– Enable Toggle mode.

The blinking speed depends on the sample time Tsig set in the constant
block.
• Insert a Delay block (it can be found in the Discrete library) and:

– Set Sample time equal to 0.1 s.

– Set Delay Length equal to 1, i.e., one step delay referred to Tsig .

• Insert two Data Type Conversion blocks to feed the two GPIO DO blocks
with uint16 or uint8 data.

Finally, connect these blocks as shown in Figure 11.5 (a). The blinking
effect would be directly visible on the board as reported in Figure 11.5 (b).

Example 3

Build a firmware which is able to drive the GPIO channels to turn the
leds on/off by reading an external command signal through serial port
 Examples with GPIO blocks 139

For this exercise, two Simulink® files must be created. One is needed to pro-
gram the MCU board and another one to test it by sending data through
COM port. SCI module A is used for communication purposes. Thus, a suit-
able jumper configuration should be realized on the board (see Table 3.2).

Firmware Environment
c28069_SCIledBR_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Insert two Digital Output (GPIO DO) blocks and double-click to set:

– Select GPIO Group GPIO32∼GPIO39;

– GPIODO1: flag GPIO34 (disable toggle mode);
– GPIODO2: flag GPIO39 (disable toggle mode).

• Insert a SCI Receive block and double-click on it:

– Select SCI module A;

– Remove both additional package header ’S’→” and terminator
’E’→”;
– Set data type: uint8;
– Set data length: 1;
– Set initial output: 0;
– Set sample time: Tsig = 0.1 s.

• Place an Add and a Constant block to invert the signals fed into GPIOs.
Set the signs of the add block equal to −+ and the constant block sample
time equal to 0.1 s. This arrangement is such that every time GPIO34 turns
on, then, GPIO39 turns off and vice-versa.

In this case, the input of the SCI module is uint8. Thus, an input signal
bounded between [0, 1] is converted into a integer ranging in [0, 255].
A sample time equal to 0.1 s (meaning a working frequency of 10 Hz) is set
for the serial communication to ensure good data transfer. The serial com-
munication was tested by the authors up to 5 kHz. However, this is not its
upper limit, which depends by the firmware optimization. Finally, connect the
aforementioned blocks as shown in Figure 11.6 (a).
 140 GPIO Peripheral—Digital Input/Output

led red
C28x C2806x

Data GPIOx

SCI RCV GPIO DO

SCI A Rx C2806x

GPIOx

GPIO DO
led blue

(a) (b)

Figure 11.6 Simulink® scheme included in c28069_SCIledBR_F.slx (a) and

expected outputs (condensed plots) on the corresponding GPIO pins (b).

Testing Environment
c28069_SCIledBR_T.slx (solver: fixed step - ODE4, step size: Tsim = Ts )

Open a new blank Simulink® project:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Place a Serial Configuration block and use the same settings adopted for
Example 1 in Section 10.5.
• Insert a Serial Send block and use the same settings adopted for Example
2 in Section 10.5.

This exercise consists in sending a constant command signal to drive GPIO34

and GPIO39 through the SCI A. Hence, a constant block is inserted into
the scheme. Dashboard elements can be added to the testing environment
scheme and connected to a constant block to easily perform live changes of
the command signal (0 ↔ 1). Alternatively, a manual change of this value
can be done by opening and modifying the constant block value (its sample
time can still be settled as Tsig ). Since the programming environment is set to
receive uint8 data, a data type conversion must be performed before the Serial
Send block through a Data Type Conversion block . Finally, connect these
blocks as shown in Figure 11.7. Remember to edit an initialization script, e.g.,
in Model Properites/Callbacks/InitFcn, even for the testing environment.
 Examples with GPIO blocks 141

COM10
Data COM10 115200
8,none,1

Figure 11.7 Simulink® scheme included in c28069_SCIledBR_T.slx.

Example 4

Build a firmware which is able to drive a PWM modulation of the

blue led blinking. The modulation signal is read through COM port

If the led turn on/off is the “hello world” for microcontroller programming,
the realization of a pulse-width-modulation (PWM) technique is the “hello
world” for power electronics applications. In this example, a PWM logic is
built from scratches using basic Simulink® elements. The Embedded Coder
Support Package for Texas Instruments C2000 Processors library includes
optimized PWM peripherals, but they are not used in this exercise. Indeed,
this example is aimed at providing the reader more insight into the logic
behind this modulation technique before to move in built-in peripherals.
The external command transferred thorugh COM port represents the mod-
ulation signal which is compared to a triangular carrier. The result of this op-
eration is a train of pulses which drives the blinking of the blue led (GPIO39).

Firmware Environment
c28069_SCIledBpwm_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Insert a SCI Receive block and double-click on it:

– Select SCI module A;

– Remove both additional package header ’S’→” and terminator
’E’→”;
– Set data type: single;
– Set data length: 1;
– Set initial output: 0;
– Set sample time: 0.001 s.

• Insert a SCI Transmit block and set it as reported in Section 10.5.

 142 GPIO Peripheral—Digital Input/Output

• Place a Rate Transition block and set its sample time equal to TTX =
0.001 s and a Data Type Conversion block set as single before sending
the data via SCI XMT.
• Place another Rate Transition block and set its sample time equal to
Ts = 0.001 s. Namely, this block is to be placed aftera SCI RCV and all the
computations done after it are executed every Ts .
• Add a Digital Output (GPIO DO) block and double-click on it:

– Select GPIO Group GPIO32∼GPIO39;

– Flag GPIO39 (disable toggle mode).

• Place a Saturation block [0, 1] to restrict the modulating signal in case a

value greater than one is sent over SCIA.

The simplest approach to realize a PWM logic is using a Relational Oper-

ator block. The signal coming from SCIA represents the modulating signal,
which is applied to the first input of the comparator. This is compared to a
triangular carrier at fixed frequency, which is applied to the second input of
such block. The desired behavior is such that:
• Every time the triangular waveform is greater than the modulating signal,
the output of the comparator is low (0);
• Every time the modulating signal is greater than the triangular waveform,
the output of the comparator is high (1).
This simple method produces an output square wave with a duty cycle that
varies depending on the level of the modulating signal, thus, changing the
blinking of GPIO39. The resulting switching frequency fsw is equal to the
carrier frequency fc . For this exercise, let assume fsw = fc = 20 Hz.
Nevertheless, in terms of implementation, the PWM principle can be real-
ized in many (embedded) ways. For simplicity reason, this exercise will only
discuss how implement the carrier signal. In particular, the Simulink® scheme
include two different approaches allowing to switch between them through a
Manual switch:
1. Continuous-time strategy:
The simplest approach is to realize the carrier frequency as a repet-
itive sequence of values, as shown in Figure 11.8. This scheme com-
prises:

•A Repeating Sequence used for building a symmetrical tri-

angular carrier. Double-click on it:
–Set time values: [0 1/(2fsw ), 1/fsw ]
where Tsw = 1/fsw = 1/20 = 0.05 s;
Examples with GPIO blocks 143
SCI A Tx
C28x

Data
SCI A Rx
SCI XMT
C28x
mod C2806x
Data
GPIOx
SCI RCV
GPIO DO
led blue
K Ts
z-1
carrier

discrete continuous

Figure 11.8 Simulink® scheme included in c28069_SCIledBpwm_F.slx.

–Set output values [0 1 0].

•A Rate Transition block after the carrier generation and set
its sample time equal to Ts = 100 µs, i.e., all the computations
done after this block are be executed every Ts .
•A Data Type Conversion set to single values to be con-
sistent with the modulating signal.
•A Relational Operator, which performs the comparison be-
tween modulation and carrier signals. Double-click on it:
–Set the relational operator equal to >;
–Set the output data type of this block as Boolean.
•A Not port, which allows to be consistent with the negative
logic which drives the leds, that is:
–PWM output 1 → 0
–PWM output 0 → 1
This strategy allows to obtain on the led connected to GPIO39
a slow blinking for low values of the modulating signal, i.e.,
mainly off within Tsw , while the contrary occurs for high value
of the modulating signal, i.e., mainly on within Tsw .
•Another Data Type Conversion set to uint8 is used to
bound the GPIO39 input signal.

Finally, connect these blocks as shown in Figure 11.8. The result-

ing waveforms are reported in Figure 11.9. To use continuous-time
blocks (i.e., the Repeating Sequence block), the feature for al-
lowing continuous time operation of the C2000™ family is exploited.
To enable this option, open the Simulink® Configuration Param-
eters window:
•Go in the Code Generation, Interface menu;
144 GPIO Peripheral—Digital Input/Output
1

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Figure 11.9 Resulting signals for a PWM logic which implements a carrier
built up with continuous-time blocks.

•Flag the continous time checkbox in the Software envi-

ronment/Support sub-menu.
2. Discrete-time strategy:
Alternatively, the carrier can be built using a Discrete-time In-
tegrator block. Indeed, the carrier signal can be approximated
through a staired waveform. In this case, a saw-tooth carrier is
built. Assuming to apply a Forward Euler discretization method,
the discrete-time integrator block works according to the following
equations:
(
y(k) = y0 if k = 0
y(k) = y(k − 1) + K[t(k) − t(k − 1)]u(k − 1) if k > 0

where y(k) is the output of the block, u(k) is the input of the block,
k is the time step index, K is the gain of the integrator and y0 is
the initial condition. The parameters of the block are set as follows:

•Integrator method: Integration: Forward Euler;

•Gain value: 1;
Examples with GPIO blocks 145
1

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Figure 11.10 Resulting signals for a PWM logic which implements a carrier
built up with a discrete-time integrator.

•External reset: Rising;

•Initial condition: 0;
•Sample time: Ts .

The external reset is needed to realize a saw-tooth waveform. Thus,

a constant value is integrated until the external reset trigger com-
mands the reset of the ramp. After that, the integration restarts
from its initial condition. This external reset is realized through a
Pulse Generator block set as follows:
•Pulse type: Time based;
•Time (t): Use simulation time;
•Amplitude: 1;
•Period (secs): Tsw = 0.05 s;
•Pulse Width (% of period): 50;
•Phase delay (secs): 0;
•Tick the checkbox Interpret vector parameters as 1-D.
 146 GPIO Peripheral—Digital Input/Output

COM14
Data COM14 115200 COM14 Data
8,none,1

Figure 11.11 Simulink® scheme included in c28069_SCIledBpwm_T.slx.

The Discrete-time Integrator block integrates a constant value

which is numerically equal to the desired switching frequency, which
is fsw = 20 Hz. In this case, the rising edge of the pulses created
by the Pulse Generator block are enough for enabling the reset of
the discrete integrator. The aforementioned blocks are arranged as
shown in Figure 11.8. The resulting waveforms are reported in Fig-
ure 11.10. It is important to underline that to realize a given carrier
frequency, the gain value of the integrator should be set equal to
fsw . In this case, the carrier peak value reaches 1 every Tsw = 0.05 s.
In both cases, the initialization script (which can be created in Model
Properites/Callbacks/InitFcn or in a separate m-file) now requires a global
definition of fsw . An example of such script is reported here below:

%% carrier frequency definition

fsw = 20;
Tsw = 1/fsw;
%% sampling time definition
Ts = 100e-6;

Testing Environment
c28069_SCIledBpwm_T.slx (solver: fixed step - ODE4, step size: Tsim = Ts )

Open a new blank Simulink® project:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Place a Serial Configuration block and use the same settings adopted for
Example 1 in Section 10.5.
• Insert a Serial Send block and use the same settings adopted for Example
2 in Section 10.5.
• Place a Constant block. This value defines the modulating signal which has
to be bounded between 0 and 1. A Data type conversion set to single
and a Rate Transition with sample time equal to 0.01 s are included as
well to prepare the data for transmission.
• Add a Serial Receive block and use the same settings adopted for Exam-
ple 1 in Section 10.5, apart from the fields Data type, which now should
be set to single, and Sample time, that is equal to 0.01 s.
 Examples with GPIO blocks 147

• Insert a Scope block to check if the control signal is set correctly.

Finally, connect all these blocks as shown in Figure 11.11. The waveforms
that should be visualized in the scope are the those reported in the bottom
plots of Figure 11.9 and 11.10 for the continuous and discontinuous carriers,
respectively.
12
Analog to Digital Converter Peripheral

In this Chapter, the working principles and the main configuration of ana-
log to digital converter (ADC) peripherals mounted on TI boards are briefly
outlined. ADCs are fundamental elements to sample analog external signals
and convert them into digital data to be processed by the MCU. In particular,
the ADC peripherals installed on the F28069M LaunchPad™ board has 12-bit
resolution (part SAR, part pipelined) with input pins operating in the range
0÷3.3V and it able to work up to 3.46 MSPS. Since ADC peripherals includes
many analog and digital circuits besides the analog-to-digital converter core,
it is common to refer to them as ADC modules. The analog circuits include
the front-end analog multiplexers (MUXs), sample-and-hold (S&H) circuits,
the conversion core, voltage regulators, and other analog supporting circuits.
Digital circuits include programmable conversions, interface to analog circuits
and to device peripheral bus. Moreover, ADCs work at full system clock with-
out the need of any prescaler.
The ADC core contains a single 12-bit converter fed by two sample-and-
hold circuits (A and B). The S&Hs can sample both simultaneously and se-
quentially, and they are fed by up to 16 analog input channels (8 for S&H A
and 8 for S&H B). The reader is reference to Figure 3.5 for more details on
the pinout of the board. The ADC blocks available in Simulink® allow to set
all the ADC modules options.

12.1 Operating Principle

The ADC peripheral allows the sampling and the conversion of analog external
signals to discrete ones. In general, this process involves voltage signals and
consists in two main steps:
1. Sampling and Holding;
2. Analog to Digital conversion.
In general, an analog signal changes in time continuously. Thus, its conversion
into digital values is possible only if it is kept steady for a suitably long time.
Namely, the voltage signal is held for a specific time interval and, then, it
is converted. The sample and hold (S&H) circuit keeps the value stable for

DOI: 10.1201/9781003196938-12 149

150 Analog to Digital Converter Peripheral
Sample & Hold
circuit

0
Dn−1
S u (t)
u (t) n-bit ..
Analog to Digital .
C Converter D1
D0

Figure 12.1 Operation principles of analog to digital conversions.

at least the time required for the digitizalization, which is performed by the
ADC. This process is sketched in Figure 12.1.

12.1.1 Sample & hold

The S&H circuit has a switch which is normally kept open. The sampling
process occurs every time this switch is closed, allowing the hold capacitor to
charge (holding process). Ideally, when the measurement window is opened,
the input signal charges the capacitor instantly. Actually, the charge is never
sent on the capacitor immediately. Indeed, charging the hold capacitor takes
a certain amount of time due to the dynamics of the circuit, which is typically
a RC one. Once the capacitor is fully charged, its voltage is quantized by the
analog to digital conversion stage.

12.1.2 Analog to digital converter

Once the input signal is held by the S&H, a numerical value is assigned to
the capacitor voltage. Namely, the nearest value to the peak value of the held
signal is searched. After this matching, the assigned number is encoded in
the form of a binary number, which is represented over “n” bits, that is, the
resolution of the ADC. The accuracy of this operation is strictly dependent
on the resolution of the converter, i.e., high resolutions imply accurate con-
versions. It is possible to find standalone ADC chip (e.g., the TI ADS8661),
however, the LaunchXL F28069M board integrates the 12-bit resolution ADC
core and the related circuits inside the TMS320F28069M. Note that a 12-bit
ADC translates external signals in integer numbers ranging from 0 to 4095.
Hardware Details 151

Figure 12.2 Block scheme showing the working principles of the ADC periph-
eral [19].

12.2 Hardware Details

A scheme of the 12-bit recyclic1 ADC module installed on the LaunchXL
F28069M board is shown in Figure 12.2. There, analog circuits such as the
front-end analog multiplexers MUXs, sample-and-hold (S&H) circuits and
the conversion core are represented as well as digital circuits such as pro-
grammable conversions, result registers, interface to analog circuits, interface
to device peripheral bus and interface to other on-chip modules. It is impor-
tant to note that part of this ADC module implementation is SAR 2 while
another part is pipelined.3 More specifically, the core of the ADC contains
a single converter which is fed by two S&H circuits. These two circuits can
sample signals simultaneously or sequentially up to 3.46 MSPS. The S&H cir-
cuits, in turn, are fed by 16 analog input channels by using two multiplexer.
The converter can be configured to run with an internal band-gap reference
to create true-voltage based conversions or with a pair of external voltage

1 Recyclic ADCs use the same module for both fine and coarse conversions of analog
signals.
2 SAR stands for successive approximation register. In this implementation, the signal

is temporarily stored in a register and coded so that its most significant bit is equal to 1.
Then, each bit of this code is compared with a reference voltage and converted step-by-step
in the final digital signal.
3 This implementation foresees the conversion through p stages. Each of them converts

n bit at a time. This operation leads to some delays in the output since the converted signal
is not available until all the stages complete the conversion. Then, the resulting signal is
coded over np bit.
 152 Analog to Digital Converter Peripheral

references VREFHI/LO to create full range ratiometric based conversions,

that is, the signal is translated on a 0 ÷ 3.3V voltage range.
This ADC peripheral has the ability to catch a series sequence of con-
versions, that is also called simultaneous sampling mode. This means that
each time the ADC receives a start of conversion (SOC) request, it is able to
schedule multiple conversions. The ADC can perform up to 16 sequential con-
version. Their timing is specified in his SOCx register. Namely, the peripheral
starts from the conversion specified in SOC0 followed by that one reported in
SOC1, and so on. For each conversion, any of the available 16 input channels
can be selected through the analog MUX. Input channel pairs can be simulta-
neously sampled on condition that the two coupled inputs have the same S&H
offset. Namely, the two channels must have the same number, so that the pair
ADCINA2-ADCINB2 is allowed, but ADCINA2-ADCINB3 is not. After con-
version, the digital value of the selected channel is stored in the appropriate
result register ADCRESULTn. Namely, there are 16 output registers; the first
result is stored in ADCRESULT0, the second one in ADCRESULT1, and so
on.
The basic operation principle of this peripheral is centered around the
configurations of individual conversions, triggered at the start by a SOC event
and at the end by an End Of Conversion (EOC) event. In particular, the 16
SOC registers can be configured by setting:

• Trigger source;
• Channel;
• Acquisition prescale window.

Moreover, multiple trigger sources are available:

• Software immediate start;

• ePWMx_ADCSOCA/ePWMx_ADCSOCB. Namely, they are interrupt sig-
nals generated by ePWM modules;
• XINT2_XINT2SOC. In this case, an external pin can be configured as a
trigger;
• CPUx_TINTxn, i.e., CPU timers 0/1/2 interrupts;
• ADCINT1/2. Those ADC interrupts can be fed back to generate ADC sam-
pling continuously. They can be set in the option ADCINT will trigger
SOCx.
Firmware Environment: ADC Peripheral 153

Finally, it is worth mentioning that the ADC peripheral contains 9 flexible

PIE4 interrupts, which can configure interrupt requests after any conversion.

12.2.1 Difference between acquisition window and sample

time
As an advanced feature, external drivers vary in their ability to drive an analog
signal quickly and effectively. Some circuits require longer time to properly
transfer the charge into the sampling capacitor of an ADC. To account for this
performance, the ADC peripheral supports control over the sample window
length for each SOC configuration. Each ADCSOCxCTL register has a 6 bit
field called ACQPS that determines the window size for the S&H. The value
inserted in this field is one less than the number of cycles desired for the
sampling window for that SOC. Thus, a value of 15 sets a 16 clock cycles
sample time. The minimum allowed number is 7 clock cycles (i.e., ACQPS =
6). The total sampling time is found by adding the sample window size to
the conversion time of the ADC, that is, 13 clock cycles. Hence, the overall
sample time is:

(ACQPS + 1) + 13 = ACQPS + 14 (12.1)

Instead, the Sampling Time register defines the rate of sampling from the
ADC register. It can be chosen independently from the acquisition window,
provided that it is greater than the observation interval. In this book, we are
going to change directly the sample time defined as Ts .

12.3 Firmware Environment: ADC Peripheral

In this Section, the characteristics of the ADC block available in the Embed-
ded Coder Support Package for Texas Instruments C2000 Processors library
(C2806x subset) are briefly presented.

12.3.1 C2806x ADC

This block (see Figure 12.3) allows to set the channel of the ADC peripheral for
acquiring an external analog signal. Usually, it is exploited for measurement
purposes. Its main settings are reported here in the following and organized
according to the tabs available in the Block Parameters window.

4 The Peripheral Interrupt Expansion (PIE) is used to multiplex numerous interrupt

sources into a smaller set. The PIE block can support up to 96 peripheral interrupts.
154 Analog to Digital Converter Peripheral
C2802x/03x/05x/06x C2802x/03x/05x/06x

A0 A0/B0

ADC ADC

(a) (b)

Figure 12.3 C2806x ADC block in two different configurations: (a) single sam-
pling mode and (b) simultaneous sampling mode.

Block Parameters - SOC Trigger tab

• Sampling Mode
This drop-down menu allows to select single or simultaneous sampling mode.
• SOC Trigger Number
This setting allows to identify the SOC trigger adopted for each channel.
In single sampling mode, one trigger can be selected only. In simultaneous
sampling mode, paired signals (0-1, 2-3, . . . , 14-15) are available.
• SOCx Acquisition Window
This parameter defines the length of the acquisition window in terms of ADC
clock cycles. The value of this parameter depends on the SYSCLK and on
the minimum ADC sample time. For the F28069M LaunchPad™ board, the
minimum value that can be set here is 7.
• SOCx Trigger Source
This drop-down menu allows to select the source that triggers the start
of analog to digital conversions. As already mentioned in Section 12.2, the
available entries are:

– Software;
– CPU0/1/2_XINT0/1/2n;
– XINT2_XINT2SOC.
XINT2SOC external pin should be specified inside the Configuration
Parameters window, Hardware Implementation, Target hardware re-
sources menu, ADC. This operation defines the GPIO channel con-
nected to a pin that triggers the start of conversion;
– ePWM1-8 _ADCSOCA/ePWM1-8 _ADCSOCB.

• ADCINT will Trigger SOCx

Use the ADCINT1 or ADCINT2 interrupt to trigger a SOC after an EOC.
Namely, those ADC interrupts act as feedbacks. These loops create a con-
tinuous sequence of conversions. The default selection, i.e., No ADCINT,
disables this parameter.
Example with ADC block 155

• Sample Time
This parameter specifies the time interval elapsed between two consecutive
samples, e.g., Ts .
• Data Type
This drop-down menu specifies the data type of the digital output data.
The available options are double, single, int8, uint8, int16, uint16, int32,
and uint32.

• Post Interrupt at EOC Trigger

This checkbox enables post interrupts when the ADC triggers EOC pulses.
When this option is selected, the dialog box displays the Interrupt Selec-
tion and ADCINT# Continuous Mode options.
• Interrupt Selection
This allows to select which interrupt ADCINT# the ADC posts after trig-
gering an EOC pulse. The following checkbox depends on the value specified
here.
• ADCINT# Continous Mode
Every time this checkbox is enabled, an ADCINT# is generated when the
ADC generates an EOC signal.

Block Parameters - Input Channels tab

• Conversion Channel
This drop-down menu allows to select the input channel to which the ADC
conversion applies. It is possible to choose one (for Single sample mode) or
two (for Simultaneous sample mode) channels.

12.4 Example with ADC block

In the following, the external board extPot3 is used to manipulate up to
three ADC inputs (see Appendix B for hardware details).
The extPot3 can be directly mounted on right-hand 40-pin socket of the
LaunchXL F28069M board (see Figure 12.4) and it includes three linear 20 kΩ
potentiometers connected to header J7, namely ADCB5 (pin 66), ADCA3
(pin67) and ADCB3 (pin68), which are used to create and vary input analog
signals between [0, 3.3] V. If the extPot3 is correctly mounted, a green led turns
on. Figure 12.5 (a) shows the equivalent circuit of a linear potentiometer.
 156 Analog to Digital Converter Peripheral

Figure 12.4 extPot3 mounted on right-hand 40pin socket of the F28069M

LaunchPad™ board (see Appendix B for hardware details).

Build a firmware aimed at reading an analog input from a

potentiometer which modulates the blinking of the blue led

Referring to the previous PWM logic implementation (see Example 4 in Sec-

tion 11.3), one of the external potentiometer is used now to vary the modu-
lating signal. For instance, by reading ADCA3 (pin67), the F28069M board
generates the pulse sequence depending on the modifications in the duty cycle
which are related to the variations in the voltage level applied to the selected
ADC input. Note that the digitalized measurement (from ADCA3) must be
scaled between 0 and 1 to fit the PWM logic designed in the previous Chapter.
Both the continuous- and discrete-time implementation of the carrier are still
valid.

Firmware Environment
c28069_ADCledBpwm_F.slx (solver: fixed step-ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Insert a GPIO DO block and build the same PWM logic used in Chapter
11 by following the steps reported in Example 4 in Section 11.3.
• The ADC block can be found in the C2806x subset of the library Embedded
Coder Support Package for Texas Instruments C2000 Processors. Insert it
in the scheme ans set its parameters as follows:
 Example with ADC block 157

3.3 V 4000

3500

3000

potentiometer
2500

2000
R
1500

vp (t) 1000

500

ADCA3 0
(PIN67) 0 5 10 15 20 25 30 35 40 45

(a) (b)

Figure 12.5 Equivalent circuit of a linear potentiometer (a) and example of

12-bit representation from ADCA3 before any scaling.

– Select ADCINA3 as Conversion channel in the Input Channels

tab. Then, go back in the SOC Trigger tab;
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC0;
– SOCx acqusition window: 7;
– SOCx trigger source: CPU0_TINT0n or Software;
– ADC will trigger SOCx: No ADCINT;
– Sample time: TADC = 0.001 s;
– Data type: uint16;
– Flag Post interrupt at EOC trigger (optional);
– Interrupt selection: ADCINT1 (optional);
– Flag ADCINT1 continous mode (optional).

• Insert a Data Type Conversion block set to int16 to deal with positive
signal only.
• Place a Gain block for scaling the digitalized signal within the interval [0, 1].
This is fundamental because a 12-bit reading returns a value in a range from
0 (0 V) to 212 − 1 = 4095 (3.3 V, see Figure 12.5 (b)). Hence, set the gain
as 1/(212 − 1) or, to be more conservative, as 1/212 . In this last case, the
representation range is not fully exploited.
• Insert a SCI transmit block to send the values returned by ADCA3 and the
output of the PWM logic to the host PC via SCI A. the reader is suggested
to use the same settings adopted in Chapter 11, i.e., single data type and
transmission rating equal to 0.01 s.
158 Analog to Digital Converter Peripheral
carrier
led blue
GPIO DO
ADC
GPIOx
mod
A3
C2806x
C2802x/03x/05x/06x
SCI XMT
potentiometer
Data

C28x
SCI A TX

Figure 12.6 Simulink® scheme included in c28069_ADCledBpwm_F.slx.

The remaining part of the scheme directly refers to Example 4 in Section 11.3.
Inside the ADC block, the sample time is set equal to TADC = 0.001 s 6= Ts .
In this exercise, the ADC reading does not represent a critical task in terms of
timing, i.e., such peripheral is not synchronized with any other one. Therefore,
the timing has been relaxed, i.e., TADC > Ts , to reduce the computational load
on the processor. However, this is not always the case, and most of the time
the ADC peripheral has to be synchronized with a particular signal as well as
using Ts is necessary to improve plot resolution.
Note that Rate Transition blocks are placed to run the signal comparison
every Ts . A continuous-time block is used to create the triangular carrier,
resulting in a switching frequency of fsw = fc = 20 Hz. This allow the reader
to recognize the different led blinking during operation. The resulting scheme
is reported in Figure 12.6.
The linear potentiometers mounted on the extPot3 board are connected to
the 3.3 V supply, which is directly available on the LaunchXL F28069M board.
A full rotation of a potentiometer leads to a voltage span from 0 V to 3.3 V or
vice-versa, depending on the direction of rotation. The resulting modulation
signal acquired by the ADC can be sent to the PC and visualized, as shown in
Figure 12.7. Since the ADC peripheral has a 12 bit resolution, the output range
is between 0 and 4095. This bit value can be nicely represented by unit16
data type, which allows numbers between 0 and 65535. Note that, in the ADC
settings does not exist a 12-bit data type, thus, 16-bit representation is the
right choice to avoid data cast. Regarding global definitions, an initialization
script an be created in Model Properites/Callbacks/InitFcn or in a separate
m-file. An example of such script is reported here below:

%% carrier frequency definition

fsw = 20;
Tsw = 1/fsw;
%% sampling time definition
Ts = 100e-6;
 Example with ADC block 159

0.8

0.6

0.4

0.2

5.75 5.8 5.85 5.9 5.95 6 6.05 6.1 6.15 6.2 6.25

0.8

0.6

0.4

0.2

5.75 5.8 5.85 5.9 5.95 6 6.05 6.1 6.15 6.2 6.25

5.75 5.8 5.85 5.9 5.95 6 6.05 6.1 6.15 6.2 6.25

Figure 12.7 Signal treatment from ADC reading of a linear potentiometer

input to GPIO39 output

Testing Environment
c28069_ADCledBpwm_T.slx (solver: fixed step - ODE4, step size: Tsim = Ts )

Open a new blank Simulink® project:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.
• Place a Serial Configuration block and use the same settings adopted for
Example 1 in Section 10.5.

• Add a Serial Receive block and double click on it:

– Select the right COMx as serial port;

– No header or terminator should be specified;
– Set data size [1 2], since one signal is sent, received and visualized
only;
– Set data type: single;
– Set sample time: TRX = 0.01 s (or 0.05 s).
 160 Analog to Digital Converter Peripheral
mod
COM14
115200 COM14 Data
8,none,1 out

Figure 12.8 Simulink® scheme included in c28069_ADCledBpwm_T.slx.

• Insert a Scope block to check if the control signal is set correctly.

• Insert a Mux block to split the signals coming from serial communication.

The aforementioned blocks should be connected as shown in Figure 12.8.

Buzzer extension (Optional)

Since the extPot3 has a piezoelectric buzzer available, the manipulated value
read by the ADC together with the PWM logic can be used to modulate the
sound coming out of the buzzer itself. This latter can be directly connected
to GPIO8 (pin78) or to ePWM5B (pin77) through a physical connection on
the board. Then, the buzzer will produce a sound at every edge of the PWM
output signal, with a sound which has an increasing frequency for high value of
the modulating signal. Figure 12.9 reports the block arrangement for driving
GPIO8 and a zoom on the extPot3 board showing the buzzer.

12.5 Synchronization between ADC modules

All the ADC modules developed for the family of C2000™ processors im-
plement a conversion approach based on SOC signals. Every SOC instance
allows a custom sampling for every channel, which can be edited in the Block
Parameters: ADC window. SOCx numbers determine the priority of the
chained conversion according the ascending number of the instances. Namely,
SOC1 is processed before SOC2, SOC2 before SOC3, and so on. The SOC
trigger number (0 − 15) can be chosen in the Block Parameters: ADC
window in the drop-down menu, while the SOCx trigger source option can
be set as reported in Section 12.3.1.

buzzer
C2806x

GPIOx

GPIO DO

(a) (b)

Figure 12.9 Block arrangement for driving GPIO8 (a) and detail of extPot3
board showing the buzzer (b).
 Synchronization between ADC modules 161

Moreover, it is possible to program the generation of an Interrupt signal

at every EOC event. Interrupt signals are numbered from ADCINT1 to AD-
CINT9. In particular, ADCINT1 and ADCINT2 can be chosen for triggering
other SOC instances only. To this aim, the check-box Post Interrupt at
EOC trigger should be enabled. This will generate the appearance of the
Interrupt selection frame in which an appropriate ADC interrupt signal
useful for further synchronization could be selected.

Example

Build a firmware aimed to synchronize two ADC modules

in a master-slave chain configuration

This exercise is an extension of the previous one. Two ADC modules are
exploited, ADCINA3 and ADCINB5. The main settings to create a master-
slave chain configuration between ADC modules is presented.

Firmware Environment
c28069_ADCsync_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Modify the previous example by duplicating the ADC block and updating
the settings as follows:

ADC1 settings:
• Select ADCINA3 as Conversion channel in the Input Channels tab.
Then, go back in the SOC Trigger tab.
• Sampling mode: Single sampling mode;
• SOC trigger number: SOC0;
• SOCx acqusition window: 7;
• SOCx trigger source: CPU0_TINT0n or Software;
• ADC will trigger SOCx: No ADCINT;
• Sample time: TADC = 0.001 s;
• Data type: uint16;
• Flag Post interrupt at EOC trigger;
• Interrupt selection: ADCINT1;
• Flag ADCINT1 continous mode (optional);
 162 Analog to Digital Converter Peripheral
potentiometer2
C2802x/03x/05x/06x SCI A TX
B5 C28x

ADC Data

potentiometer1 SCI XMT

C2802x/03x/05x/06x
mod C2806x
A3
GPIOx
ADC
GPIO DO
led blue
carrier

Figure 12.10 Simulink® scheme included in c28069_ADCsync_F.slx.

ADC2 settings:
• Select ADCINB5 as Conversion channel in the Input Channels tab.
Then, go back in the SOC Trigger tab.
• SOC trigger number: SOC1;
• ADC will trigger SOCx: ADCINT1;
• Unflag Post interrupt at EOC trigger;
• Keep the other settings equal to those specified for ADC1.

In this case, ADC1 (ADCINA3) is the master that creates an interrupt sig-
nal ADCINT1 at its EOC. This signal is used to trigger the SOC of ADC2
(ADCINB5), which behave as slave.
Note that if only one ADC module is used, the EOC flag is not strictly
necessary since no other modules requires such trigger; this also explain the
settings reported in Section 12.4. The new scheme is shown in Figure 12.10.

Remark

Every ADC block exploits one SOC in single sample mode or two of
them in simultaneous sample mode. It is forbidden to use the same SOC
instance for different ADCs.
13
Pulse Width Modulator Peripheral

The enhanced pulse width modulator (ePWM) peripheral is a key element in

controlling many of the power electronic systems found in both commercial
and industrial equipment. These systems include digital motor control, switch-
mode power supply control, uninterruptible power supplies (UPS), and other
forms of power conversion. Each of the available ePWM modules generates
signals through suitable switching patterns by means of a PWM logic, per-
forming a digital to analog (DAC) function.
The F28069M LaunchPad™ board provides up to 8 ePWM modules
with a total amount of 16 PWM channels, 8 of them high-resolution PWM
(HRPWM), with 16-bit independent timer in each module. In particular, from
such 16 channels, 12 of them are general purpose while 4 of them are filtered
PWM, i.e., equipped with an RC output filter. Those channels performs the
actual DAC function for the board.
Each ePWM module has two outputs, ePWMA and ePWMB. The signals
available at the corresponding output pins can drive the gates of the upper
and lower switches of a two-level converter topology, as shown in Figure 13.1.
In this Chapter, the operation principles of the ePWM peripheral are briefly
outlined as well as an explanation on how to configure the its settings in the
Simulink® environment.

comparator

1 S1
v " x
+ − 0 S1 , S4
vc
S2
x
S2 , S3

Figure 13.1 Actuation of a two-level voltage source converter (2L-VSC) with

PWM logic.

DOI: 10.1201/9781003196938-13 163

 164 Pulse Width Modulator Peripheral

13.1 Operating Principle

Since the working principle of PWM logic was already presented in Chapter 11,
a specific example is considered here in the following. Namely, the actuation
of a two-level voltage source converter (2L-VSC) which comprises two power
switches S1 and S2 (it is not important to specify their technology) with anti-
parallel diodes (see Figure 13.1). In a pulse width modulator, the reference
voltage is compared to a triangular carrier (both of them are scaled over
a [0, 1] range) and this output is used to drive the converter switches in a
complementary way. By considering a single-phase case, i.e., the drive of a
single converter leg (phase A), a sinusoidal reference voltage m(t) is compared
to the triangular carrier signal generating a switching pattern S(t) which has
to be applied on the upper switch through the gate driver, i.e., S1 (t) = S(t).
Instead, the lower switch is actuated with the logic negation of the resulting
switching pattern, that is, S2 (t) = not(S1 (t)). Therefore, S2 operates in a
complementary way with respect to S1 .
The operation of these switches given their switching patterns creates a
pulsed voltage waveform at the output of the converter. This latter can be
simply represented as vout (t) = S(t)Vdc , i.e., the product between S(t) and
the supply voltage (dc-link) Vdc . Hence, the output voltage is bounded within
two levels, i.e., 0 and Vdc . The fundamental component of this voltage is
proportional to the reference voltage. Indeed, its harmonic spectrum includes
the frequency of the modulation signal as fundamental component and higher
frequency components around the multiples of fsw , depending on the adopted
PWM logic. The carrier signal can be implemented in different ways, e.g.,
realizing increasing saw-tooth, decreasing saw-tooth or triangular waveforms.
The carrier frequency must be higher than the modulation one and, in case
of a two-level converter, the resulting switching frequency equals the carrier
one, that is fsw = fc . indeed, the comparator evaluates:
• If modulation signal < carrier signal → PWM signal high → S(t) = 1;
• If modulation signal > carrier signal → PWM signal low → S(t) = 0.
This mechanism allows to dynamically adjusts the duty cycle d(t) = Ton /Tsw
of the output square wave, i.e., by varying the control signal, and, hence, its
average value.
Assuming fsw = fc = 1 kHz, a sine wave at frequency f1 = 50 Hz
(T1 = 0.02 s) as modulation signal and Vdc = 10 V, the corresponding re-
sulting sinusoidal PWM modulation is shown in Figure 13.2. The first order
harmonic computed for vout (t) is matching with the fundamental components
f1 , which amplitude is dependent on Vdc . It is important to note that such
modulating signals are commonly used in AC drives, due to the AC nature
of many application. However, constant modulating signals are used in DC
drives, and their value represents the duty cycle m(t) = d(t) of the output
 Operating Principle 165

10

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02

Figure 13.2 Results of a sinusoidal PWM implementation in terms of switching

pattern and output voltage of the 2L-VSC.

PWM modulated signal. Therefore, this modulation technique is extremely

useful in power electronics-based application because:

• Small motors often requires low power consumption and the use of (dis-
continuous) PWM signals allows to efficiently control a power converter
by reducing the power losses in the switches (both due to conduction and
switching) compared to other kind of linear controller.1 Indeed, the switch-
ing devices foresees two states only: short and open circuit.
• The motor dynamics (which can be modeled as an RL circuit and a con-
trolled voltage source, that is, the back-emf) act as a low pass filter. Hence,
undesired high-order harmonics resulting from the PWM implementation
might be rejected by the machine itself without the need for additional filter,
being mostly sensitive to low frequency components such as the fundamental
one.

1 An old alternative was a simple circuit in which the current flow was linearly controlled

by varying a resistance, causing much energy losses and, thus, heat.

 166 Pulse Width Modulator Peripheral

It is essential to remind that the aforementioned approach has to be trans-

lated into discrete-time domain k for implementations on MCUs. This process
is represented in the top plot reported in Figure 13.2, where the discretized
modulating signal m(k) is superimposed to m(t). For instance, such discretiza-
tion is done considering Ts = Tsw /2. Thus, the value of m(k) is updated at
every peak of the carrier signal. By using Ts = Tsw , the value of m(k) is up-
dated every time the carrier equals zero. The choice of this sampling time is
quite important for embedded implementation and for the resulting voltage
spectrum as well, which may now include further harmonics.
In Chapter 11 the implementation of a PWM logic realized by using simple
Simulink® blocks was presented. This operation is conceptually fine, but it
is not an optimal solution for embedded implementation. Indeed, the PWM
processing logic can be optimized to achieve high-resolution at low latency rate
with specific routines/iterations as implemented by the ePWM peripheral.

13.2 Hardware Details

Effective PWM peripherals aim to generate complex pulse width waveforms
with minimal CPU employment or interrupts. The ePWM modules reported
here in the following address these requirements by allocating all needed tim-
ing and control resources. Cross-coupling or resource sharing are avoided. This
peripheral is built up from smaller single-channel modules with separate re-
sources that can operate together, as shown in Figure 13.3.
Each ePWM module mounted on the F28069M LaunchPad™ board has
two PWM outputs, i.e., ePWMxA and ePWMxB, where the x is the module
number. The corresponding output pins are selected through the GPIO multi-
plexer. Additionally, the ePWM module can generate an interrupt to the PIE
block as well as a SOC signal to the ADC. Inputs to the ePWM module in-
clude signals from the input cross-bar, from the ePWM cross-bar and selected
trip zone signals.
All ePWM modules installed onboard C2000™ -based devices are identical.
They can operate stand-alone, but they can be chained as well to work together
as a single system via clock synchronization scheme. Note that this operation
can be extended to other peripherals.
Here in the following, a recap list of the features supported by each ePWM
module is reported:
• Dedicated 16 bit time-base counter with both period and frequency control.
• Two PWM outputs (EPWMxA and EPWMxB) that can be used in the
following configurations:

– Two independent PWM outputs with single-edge operation;

– Two independent PWM outputs with dual-edge symmetric operation;
– One independent PWM output with dual-edge asymmetric operation.
 Hardware Details 167

Figure 13.3 Scheme of the ePWM peripheral mounted on the LAUNCHXL

F28069M board [21].

• Asynchronous over-ride control of PWM signals through software.

• Programmable phase-control support for lag or lead operation with other
ePWM modules.
• Hardware-locked (i.e., synchronized) phase relationship with other modules
on a cycle-by-cycle basis.
• Dead-band generation with independent rising and falling edge delay control.
• Programmable trip zone allocation of both cycle-by-cycle and one-shot trip
in fault conditions.

• A trip condition which can force either high, low, or high-impedance state
logic levels at PWM outputs.
• All events of the module can trigger both CPU interrupts and ADC SOC
signals.
 168 Pulse Width Modulator Peripheral

• Programmable event prescaling for minimizing CPU overhead or interrupts.

• PWM chopping by high-frequency carrier signal, useful for pulse transformer

gate drives.
Moreover, a list of the main input and output signals for each ePWM module
is reported here below:

• PWM output (EPWMxA and EPWMxB)

They are the PWM output signals which are made available external to the
device through the GPIO peripheral (see Chapter 11).

• Trip-zone (TZ1 − TZ6)

These input signals alert the ePWM module of fault conditions external to
the ePWM module. Each of them can be configured to either use or ignore
any of the trip-zone signals. The TZ1 − TZ3 trip-zone signals can be set as
asynchronous inputs through the GPIO peripheral. TZ4 is connected to an
inverted EQEP1 error signal (EQEP1ERR) from the EQEP1 module. TZ5
is connected to the system clock fail logic, whereas TZ6 is connected to the
EMUSTOP output from the CPU. This allows the configuration of a trip
action when the clock fails or the CPU halts.
• Time-base synchronization input (EPWMxSYNCI) and output
(EPWMxSYNCO)
They are the synchronization signals which are able to chain the ePWM
modules together. Each module can be configured to either use or ignore its
synchronization input. The clock synchronization input and output signals
are brought out to pins only for ePWM1 (i.e., ePWM module 1). The syn-
chronization output for ePWM1 (EPWM1SYNCO) is also connected to the
SYNCI of the first enhanced capture module (eCAP1).
• ADC start-of-conversion (EPWMxSOCA and EPWMxSOCB)
Any ePWM module can trigger a start of conversion. Indeed, each of them
is associated with two ADC start of conversion signals. Whichever event
triggers the SOC should be configured in the Event-Trigger sub-module of
the ePWM peripheral.
• Comparator output (COMPxOUT)
These are the output signals from the comparator module. In correspondence
with the trip zone signals they can generate digital compare events.
• Peripheral Bus
The peripheral bus is 32 bit wide and it allows both 16 bit and 32 bit writing
operations in the ePWM register file.
 Hardware Details 169

Figure 13.4 ePWM sub-modules [21].

13.2.1 ePWM sub-modules

Each ePWM module can communicate with other on-board peripherals
through dedicated routines which are summarized in eight sub-modules, as
shown in Figure 13.4. Namely, they are:

• Time-base; • PWM chopper;

• Counter compare; • Trip zone;
• Action qualifier; • Digital compare;
• Dead band; • Event trigger;

Almost all the working options of each sub-module can be programmed in

Simulink® through the ePWM Block Parameter window. The features of these
sub-modules are explained in details in the following:

• The time-base sub-module consists in a dedicated 16 bit counter along

with built-in synchronization logic to allow multiple ePWM modules to work
together as a single system. A clock prescaler is placed between the system
clock and the counter. A period register is used to control both frequency
and period of the generated waveform. The period register has a shadow
register, which acts like a buffer to allow the updates to be synchronized
with the counter. This avoids corruption or spurious operation due to the
asynchronous modification of the register by the software. In addition, all
the shadow registers support simultaneous global load, which avoids partial
loads from the shadow registers to the active ones. The time-base counter
operates in three modes: up count, down count, and up/down count (see
Figure 13.5).
 170 Pulse Width Modulator Peripheral

Figure 13.5 Counting modes available in the adopted MCU board. From left
to right they are: up/down, up and down count.

• The counter compare sub-module continuously compares the time-base

count value to four counter compare registers: Compare A, B, C, and D.
They generate four independent compare events. Every time the time-base
counter equals a compare register value, the compare event is fed by the
action qualifier and event trigger sub-modules. Typically, compare A and B
are used to control the duty cycle of the generated PWM waveform on a
single leg.
• The action qualifier sub-module is a key element for ePWM peripherals
because it is responsible to generate the switched PWM waveforms. It ex-
ploits events coming from the time-base and counter compare sub-modules
to perform actions on the ePWMxA and ePWMxB output pins. These ac-
tions can be chosen among the following allowed solutions:

– Do nothing;
– Clear the pin low;
– Set the pin high;
– Toggle the pin based independently on count-up and count-down time-
base match events.

Instead, the triggering events can be generated by one of these occurrences:

– The time based counter equals zero;

– The time-base counter equals Compare A;
– The time-base counter equals Compare B;
– The time-base counter equals the period register value;
– A trigger event T1 and T2 based on a comparator, trip, or synchro-
nization signal.

It is important to note that zero and period actions are fixed in time, whereas
Compare A and B actions are movable in time by programming their cor-
responding registers. These actions are configured independently for each
output using shadowed registers.
 Generation of PWM signals 171

• The dead band sub-module allows to delay the switching action of a

transistor. Since power switching devices turn on faster than they turn off,
a delay is needed to prevent any short-circuit path from the DC source to
ground. Further explanation is available in Section 13.3.
• The PWM chopper sub-module modulates a high frequency carrier sig-
nal with the PWM waveform that is generated by the action qualifier and
dead band sub-modules. Its purpose is to check the power switching elements
used with pulse transformer-based gate drivers.
• The trip zone sub-module provides a mechanism to protect the output
pins from abnormalities, such as over-voltage, over-current, and excessive
temperature rise. The trip zone signals can be generated externally from
any GPIO pin, internally from an eQEP error signal, system clock failure,
or emulation stop output from the CPU. Additionally, several trip zone
source signals can be generated from the digital compare sub-module.
• The digital compare sub-module increases the capabilities of the trip
zone sub-module by comparing signals external to the ePWM module to
directly generate PWM events or actions which are then used by the time-
base, event-trigger, and trip zone sub-modules.
• The event-trigger sub-module manages the events generated by the time-
base, counter compare, and digital compare sub-modules for generating an
interrupt to the CPU and/or a start of conversion pulse to the ADC when
selected events occur (EPWMxSOCA/EPWMxSOCB). These event triggers
can occur when the time-base counter equals:

– Zero;
– Period;
– Zero or period;
– The up- or down-count match of Compare A, B, C, or D registers.

13.3 Generation of PWM signals

The ePWM peripheral includes a complex structure capable of implementing
several features. In order to efficiently explain how to deal with all the required
settings, a short overview on how the PWM logic is implemented on the
F28069M LaunchPad™ board is reported in this Section. Particular attention
is paid on how the peripheral manages the comparison between modulation
and carrier signals by using specific counters as well as on how complementary
behaviors should be enabled.
 172 Pulse Width Modulator Peripheral

PWMfrequency PWMfrequency

up-down TBCLK
4 4
count
3 3 3 3
2 2 2 2
1 1 1 1
0 0

up up
TBCTRdir down down
TBPRD
TBCTR=
TBPRD

TBCTR=0

Figure 13.6 Example of up-down counting mode considering TBPRD = 4.

The time instants (rising edge) when TBCTR = 0 and TBCTR = TBPRD
occur are highlighted.

13.3.1 Counting modes

The ePWM modules generate PWM signals by processing discretized quanti-
ties (either modulating and carrier signals). The carrier waveform is defined in
amplitude by a discrete counter, called Time-Base Counter (TBCTR), and in
shape by the counting mode parameter. The TBCTR increases at every CPU
clock cycle TBCLK defined by the frequency fck (also called CPUfrequency or
SYSCLKOUT) as TBCLK = 1/fck = 1/CPUfrequency . All the internal sig-
nals are defined with respect to this base time. Regarding the LAUNCHXL
F28069M board, fck = CPUfrequency = 90 MHz, thus, TBCLK = 11.1 ns.
The peak value of the digital carrier is defined by the time-based period
(TBPRD), which can be also called PWMcounter_period . Based on that, every
time the peak value is reached TBCTR = TBPRD = PWMcounter_period . The
triangular carrier was defined in a [0, 1] range in Section 13.1 to the aim of pro-
viding a general introduction. In fact, at the embedded level, TBCTR varies in
the interval [0, PWMcounter period ]. Consequently, the modulation signal must
be processed to span the same interval of values. Regarding the carrier shape,
two counting modes are available for this peripheral:

• Symmetrical carrier (up-down count), i.e., a triangular carrier wave-

form. In up-down-count mode, as long as the time base counter direction
signal (TBCTRdir ) is high (1), the TBCTR counting increases till reach-
ing PWMcounter period , while as long as TBCTRdir is low (0), the TBCTR
 Generation of PWM signals 173

PWMfrequency PWMfrequency

TBCLK
up-count 4 4
3 3
2 2
1 1
0 0

Figure 13.7 Example of up-count counting mode. PWMfrequency and TBCLK

can be defined analogously for the down-count mode.

counting decreases down to 0. The value of TBCTRdir is related to two

events. When the event TBCTR = TBPRD occurs, TBCTRdir switches
from 1 to 0, while the opposite happens every time TBCTR = 0. For a
symmetrical carrier, TBCTRdir has to change value every time TBPRD is
reached, which has to be defined as an integer multiple of TBCLK. These
relationships are highlighted in Figure 13.6. In particular, the counter period
for up-down counting modes is defined as:
CPUfrequency
TBPRD = PWMcounter_period = (13.1)
2 · PWMfrequency

where fc = fsw = PWMfrequency is the frequency of the carrier. For example,

to implement a fc = 20 kHz carrier at into the F28069M LaunchPad™ board,
the counter period should be set equal to:
90 MHz
PWMcounter_period = = 2250
2 · 20 kHz

• Asymetrical carrier (up-count or down–count), i.e., sawtooth carrier

waveform. In up-count mode, TBCTR starts from zero and it increases until
TBCTR=TBPRD is reached. There, the time-base counter resets to zero
and, then, it increases again from the next time step. This kind of carrier
is shown in Figure 13.7. The down-count mode behaves similarly, with the
exception that the counter starts from TBPRD and it decrements until it
reaches 0. Therefore, the counter period for up or down countersign modes
is defined as:
CPUfrequency
TBPRD = PWMcounter_period = (13.2)
PWMfrequency

Considering again a fc = 20 kHz carrier signal, the counter period should

be set equal to:
174 Pulse Width Modulator Peripheral

Comparator
b (k) 2 [0 TBPRD]
m CMPA ePWMxA
>
ePWMxB

Carrier

ePWMx peripheral

Figure 13.8 Representation of the inner structure of an ePWM module aimed

to actuate a single-phase 2L-VSC.

90 MHz
PWMcounter_period = = 4500
20 kHz
Note that the value of TBCTRdir follows from the reasoning reported for
the symmetrical carrier.

13.3.2 ePWMxA and ePWMxB sub-modules

In terms of implementation, the comparison between the modulating signal
and the carrier is performed by setting the action qualifier (AQ) event related
to comparator CMPA and CMPB. Recalling the example reported in Section
13.1 involving the 2L-VSC, it is assumed that the upper switch is now labeled
with A while the lower one is called B. Since a single-leg is considered (i.e.,
first leg x = 1), the gate signals corresponding to the switches are driven by
ePWM1A and ePWM1B, as shown in Figure 13.8.
The amplitude of the modulating signal is defined by the (instantaneous)
value stored into the counter-compare A (CMPA) register. This value is con-
tinuously compared to TBCTR and when TBCTR = CMPA an appropriate
event is generated according to the AQ settings. In this particular case, the
AQ would be specified considering the output ePWM1A, which is the master,
while ePWM1B is computed as its complement. This is the reason why one
comparison between TBCTR and CMPA is performed only. Int is important
to underline that m(k) defined in Section 13.1 now becomes m(k),
b since it is
varying within the range [0, TBPRD]. As an example, assuming an up-down
counting mode, every time TBCTR = CMPA and the carrier slope is posi-
tive, clear is defined as AQ event on ePWM1A. Thus, ePWM1A = 0 and
ePWM1B = 1. For consistency, when TBCTR = CMPA and the carrier slope
is negative, set is defined as AQ event on ePWM1A leading to ePWM1A = 1
and ePWM1B = 0. This behavior is shown in Figure 13.9. Alternatively, the
AQ event on ePWM1A can be set on do nothing to keep it at same level
Generation of PWM signals 175

PWMfrequency PWMfrequency

4 4
3 3 3 3
CMPA=3
2 2 2 2
CMPA=2
1 1 1 1
0 0

up up
TBCTRdir down down

TBCTR=
CMPA

ePWMxA CLEAR SET CLEAR SET

ePWMxB

Figure 13.9 Counter compare operation considering CMPA = 2, 3 and with

ePWMxA behaving as master.

as currently set. Instead, it is not possible to define AQ events related to

CMPB, which is not used in this case. Note that the CMPA value is updated
at the beginning of every switching period in Figure 13.9 (i.e., at a rate of
PWMfrequency ), which correspond to the time instant in which TBCTR = 0.
This is not always the case and this behavior can be changed by modifying
the compare value reload conditions.
Even if the aforementioned switching strategy (i.e., complementary eP-
WMxA/B) is the most common one for power electronic-based applications,
ePWMxA and B can also operate independently. This is achieved using CMPA
and CMPB with specific AQ events. Since many configurations are possible,
ePWMxA and ePWMxB can be set separately. More details on the main
options available in Simulink® are reported in the following sections.

13.3.3 Setting dead bands

The ePWM1A and B switching patterns are physically applied to the switches
through specific gate driving circuits. These latter have to be designed accord-
ing to the switch technology and the application requirements. It is important
to underline that the design of gate driving circuits is out of the scope of this
176 Pulse Width Modulator Peripheral

PWMfrequency PWMfrequency

4 4
3 3 3 3
CMPA=3
2 2 2 2
CMPA=2
1 1 1 1
0 0

ePWMxA
FED RED
ePWMxA
FED

ePWMxA
RED

ePWMxA
AHC

ePWMxB
AHC

Figure 13.10 Counter compare operation for ePWMxA and ePWMxB with
dead bands in the correspondence of both RED and FED.

book. Nevertheless, it is important to remind that such circuits may cause

delay and/or particular actuation processing (e.g., bootstrap circuits) which
could lead to simultaneous application of closing command to the switches
and, thus, leg short circuit. To prevent such faulty condition, an interlock
time (also called dead-time) is set for both ePWMxA and ePWMxB. Ac-
cording to the hardware realization and the application requirements, this
time interval are in a range from 10 ns (e.g., low-power applications) up
to 20 µs (e.g., high-power applications).2 The ePWM peripheral settings of
the F28069M LaunchPad™ board allow to add programmable delays that
shift the switching patterns. Note that, the definition of interlock time or
dead time refers to a quantity expressed in second. However, for embedded
2 The gate driver circuits might add further dead-time aimed to increase the protection

of the system, i.e., a delay is always present even if dead-time equals zero at the firmware
level.
Generation of PWM signals 177

Figure 13.11 Summary of the features together with a possible application of

the ePWM peripheral [8].

implementation, the dead time should be defined in clock cycles (i.e., as mul-
tiples of TBCLK). In this case, it is easier to refer to dead band which is used
to define the time interval that separates the transition edges of two signals:
output and complemented output.
In particular, dead times can be added in the correspondence of
1. Rising edges—defining a rising edge delay (RED);
2. Falling edges—defining a falling edge delay (FED);
3. Both rising (RED) and falling edges (FED).
By default, ePWMxA is used as signal source both for RED and FED. These
latter identifies dead bands since they are defined in clock cycles. Furthermore,
the choice of the dead time, thus, the length of the RED and FED dead
bands should consider the value assumed by PWMcounter_period . For instance,
assuming PWMcounter period = 2250, the sum RED+FED must allow switching
operations inside the switching period. Namely, each dead time is limited in
the range [0 1023], so that 1023 + 1023 = 2046 < 2250.
Referring to the 2L-VSC case, since ePWM1A is operated as master, the
dead band polarity can be configured as Active High Complementary (AHC)
to compute the ePW1B starting from ePW1A. Then, the resulting switching
pattern provided to the gate driver circuits are ePW1A AHC and ePW1B
AHC. By choosing equal values for RED and FED, the pattern keeps a cen-
tered symmetry. This is shown in Figure 13.10. However, this is not always
the case in power electronics applications.
The aforementioned features and working principles of this peripheral can
be summarized as shown in Figure 13.11. The reader should refer to the Sec-
tion reported here in the following for the details on the implementation of
these functions in Simulink® .
178 Pulse Width Modulator Peripheral
C2802x/03x/05x/06x

ePWM

Figure 13.12 C2806x ePWM block.

13.4 Firmware Environment: ePWM Peripheral

The Simulink® block that allows to program the ePWM peripheral is the
ePWM one (see Figure 13.12); it can be found in the Embedded Coder Support
Package for Texas Instruments C2000 Processors library (C2806x subset).

13.4.1 C2806x ePWM

The ePWM block (see Figure 13.12) allows to set several settings and pa-
rameters which are split within ten different tabs inside the Block Parameters
window. A detailed view of the available tabs is reported here below.

In the following, given the target of this book, only the main settings useful
for power electronic-based applications are reported.

Block Parameters - General tab

This menu allows to configure the carrier signal through the settings of the
counters and registers. In particular, a key element is the TBCTR register
from which the carrier shape is defined. The editable parameters are reported
here in the following:

• Module
There are 12 different ePWM modules available: ePWM1-12. Each of them
has two outputs: ePWMxA and ePWMxB.
• Timer period units
This setting specifies whether the unit of TBPRD is specified in seconds or
directly in clock cycles (default choice). Note that, even if the timer period
unit is seconds, the software converts this value to an equivalent quantity
in clock cycles during the deploy to hardware. Therefore, it is suggested to
specify TBPRD in a clock cycles units (TBCLK) to reduce computations and
rounding errors. Regarding the F28069 LaunchPad™ board, it is important
to remind that fck = CPUfrequency = SYSCLKOUT = 90 MHz, that leads
to TBCLK = 11.1 ns.
Firmware Environment: ePWM Peripheral 179

• Specify timer period via

This drop down menu allows to specify how the timer period value should
be set. If Specify via dialog is selected, the following label becomes Timer
period. If Input port is selected, the following label switches to Timer
initial period and a timer period input port T is created on the block.
• Timer period/Timer initial period (TBPRD)
This value sets the maximum count reachable by the TBPRD counter and
it is stored in the corresponding register. This value must be compliant with
the choice specified in Time period units and the selected counting mode,
i.e., the type of the carrier waveform. Section 13.3 explains how to compute
this value according to the counting mode. It is reminded that, in this book,
TBPRD is also called PWMcounter_period and it represents the amplitude of
the triangular carrier.
• Reload for time-base period register (PRDLD)
This setting specifies the time at which the counter period is reset:

– Counter equals to zero: the counter period refreshes when the value
of the counter is 0. Namely, the shadow register contents are transferred
to the active register TBCTR = 0;
– Immediate without using shadow: the counter period refreshes
immediately.

• Counting mode
Three different counting modes are feasible: Up, Down, and Up-Down; this
directly relates to the definition of TBPRD. More details are reported in
Section 13.3.
• Synchronization action
Every time different ePWM modules should work together, the source of
a phase offset can be specified to act on the time-base synchronization in-
put signal (EPWMxSYNCI), that is, a synchronization input signal from a
previous ePWM module through the SYNC input port. Possible choices are:

– Set counter to phase value specified via dialog, which creates

the Time-Based Phase (TBPHS) offset value parameter;
– Set counter to phase value specified via input port, which cre-
ates a phase (PHS) input port on the block;
– Disable (default entry), which prevents the application of any phase
offset to the ePWM module.

• Counting direction after phase synchronization

This parameter appears when the Counting mode is set on Up-Down
180 Pulse Width Modulator Peripheral

Figure 13.13 TBPHS role while synchronizing the counter: count down on
synchronization event [21].

and the Synchronization action is Set counter to phase value spec-

ified via dialog or Input port. The timer can be configured to count up
or down after reaching the value specified in TBPHS (see next option).
This parameter corresponds to the PHSDIR field of the Time-Base Control
(TBCTL) Register.

• Phase offset value (TBPHS)

This field appears when Set counter to phase value specified via dialog
is selected in Synchronization action. The offset value is stored in the
TBPHS register and loaded in TBCTR at every Synchronization event. Each
ePWM module can be configured to use or to ignore the synchronization
input. If this parameter is set, then, the TBCTR is automatically loaded with
the content of TBPHS (see Figure 13.13) when one of these two operations
are configured:

– EPWMxSYNCI, that is, Synchronization Input Pulse: the value

of the TBPHS register is loaded into the TBCTR one when an input
synchronization pulse is detected. Each synchronization attempt occurs
at every TBCLK edge.
– Software Forced Synchronization Pulse: every time 1 is speci-
fied in TBCTL control bit, a software forced synchronization is called.
Firmware Environment: ePWM Peripheral 181

This pulse acts as a synchronization input signal, and, therefore, it is

equivalent to a pulse acting in EPWMxSYNCI.

• Synchronization output (SYNCO)

This parameter specifies the event that generates a time-base synchroniza-
tion output signal EPWMxSYNCO from the time-base sub-module and it
corresponds to the SYNCOSEL field in the TBCTL register. This signal can
be sent to next ePWM modules for synchronizing it. Indeed, the available
choices are:

– Pass through (EPWMxSYNCI or SWFSYNC): a synchroniza-

tion input pulse or software forced synchronization pulse are passed
from an ePWM module to the following chained one. This option is
used to achieve precise synchronization across multiple ePWM mod-
ules by daisy-chaining their time-base sub-modules.
Remark: ePWM modules are chained in ascending order. Namely,
ePWM1 sends synchronization signal to ePWM2, which, in turn, sends
it to ePWM3, and so on.
– Counter equals to zero (CTR=Zero): the output signal is gener-
ated every time the time-base counter is equal to zero TBCTR = 0;
– Counter equals to compare B (CTR=CMPB): the synchroniza-
tion signal is generated every time the time-base counter is equal to
counter-compare B TBCTR = CMPB;
– Disable: EPWMxSYNCO output is disabled (default choice).

• Time base clock (TBCLK) and High speed time base clock (HSP-
CLKDIV) prescaler dividers
The default prescaler value is 1. Nevertheless, a different value can be set
(i.e., additional prescaler) to reduce the size of the peripheral counters in
order to not exceed the range of values allowed by uint16 data type. This
especially happens for low carrier frequencies. As an example, considering
an up counting mode at fc = 1 Hz and CPUfrequency = 90 MHz, it follows
that:
CPUfrequency
TBPRD = PWMcounter_period = = 1 · 106 (13.3)
PWMfrequency
which cannot be realized with 16 bit data. Hence, a frequency di-
vider/prescaler is needed to adapt the representation range. The clock cycle
used as base time can be recomputed as:
1
TBCLK =
PWMfrequency · 65535
Now, it can be used to cast PWMcounter_period up to the maximum value
allowed by uint16 data type.
 182 Pulse Width Modulator Peripheral

• Enable swap module A and B

Every time this checkbox is ticked, the options set in the following two tabs
(ePWMA and ePWMB) are swapped.

Block Parameters—ePWMA tab

The settings that can be edited in this tab are used to configure the compar-
ison between the carrier and the modulating signals, the switching pattern
behavior and the reload conditions, e.g., for the upper switch of a 2L-VSC leg
(ePWMxA).

• Enable ePWMxA
This checkbox is needed to allow any change in all the options for controlling
the switch that corresponds to ePWMxA. By default, Enable ePWMxA is
ticked (whereas Enable ePWMxB is not).
• Action when counter = [· · · ]
The comparison between the input signal and the triangular carrier is de-
fined by the AQ settings. In particular, there are four compare registers:
CMPA, CMPB, CMPC, and CMPD, which are compared to TBCTR value
to generate appropriate events. The values of interest of the counter are:
ZERO, period (PRD), CMPA on up-count (CAU), CMPA on down-count
(CAD), CMPB ojn up-count (CBU) and CMPB on down-count (CBD). In
case of a 2L-VSC, CMPA is of particular interest. The reader is referred to
Section 13.3 for further details.
• Compare value reload condition
This drop-down menu determines if and when to reload the AQ operations.
The available choices are:

– Load on counter equals to zero (CTR=Zero);

– Load on counter equals to period (CTR=PRD);
– Load on either;
– Freeze.

• Add continuous software force input port

Every time this checkbox is enabled, an input port (software forced action-
SFA) is created, which can be used to select the software force logic. Three
different integer values can be sent to SFA port:

– 0, that is, forcing disable: Do nothing (default option);

– 1, that is, forcing low: Clear low;
– 2, that is, forcing high: Set high.
Firmware Environment: ePWM Peripheral 183

• Continuous software force logic

This drop-down menu appears in this tab only if the previous checkbox is
disabled. It allows to select which type of software force logic to apply. The
available choices are:

– Forcing disable: Do nothing (default setting);

– Forcing low: Clear low;
– Forcing high: Set high.

• Reload condition for software force

This drop-down menu allows to select when the reload operations occurs:

– Zero (default setting);

– Period;
– Either period or zero;
– Immediate.

Block Parameters - ePWMB tab

The settings that can be edited in this tab are used to configure the compar-
ison between the carrier and the modulation signals, the switching pattern
behavior and the reload conditions, e.g., for the lower switch of 2L-VSC leg
(ePWMxB). In this case, if a complementary switching is desired, ePWMxB
has to be the opposite of ePWMxA. This is achieved by selecting the in-
verted version of ePWMxA checkbox, without requiring other settings.
All the other parameters that can be found in this tab are the same as those
described previously.

Block Parameters - Counter Compare tab

This tab allows to set specific parameters of the counter compare sub-module.
The same options are repeated for the high and low switch. Namely, the
available settings are:

• CMPA units
This drop-down menu allows to specify the units used by the register which
is dedicated for the comparison. The available choice are:

– Clock cycles (default option);

– Percentages.

• Specify CMPA via

This second drop-down menu is used to specify in which way the control
signal is read by the ePWM module. The available choices are:
184 Pulse Width Modulator Peripheral

– Dialog (default selection). Namely, the triangular carrier wave is com-

pared to the constant value specified in the CMPA value label;
– Input Port. Namely, a new input port labeled with WA appears on
the ePWM block.

• CMPA value
This field appears when Specify CMPA via is set on dialog. A constant
control signal is specified in this option in CMPA units.
• CMPA initial value
This field appears when Specify CMPA via is set on input port. This
label is intended to assign the initial value to CMPA, which is considered
by the ePWM block every time it starts counting. Then, the input signal
through WA port drives the modulation of the high switch.
• The same settings can be provided for CMPB such as CMPB units,
Specify CMPB via Input port (which creates a new port WB), CMPB
value and CMPB initial value.

Block Parameters - Deadband unit tab

This tab allows to program dead bands on the PWM signals generated by
ePWMxA and ePWMxB modules to avoid any overlapping between the gen-
erated signals and, thus, any short-circuit in power electronic applications.
The same settings are repeated for both ePWMxA and ePWMxB.

• Use deadband for ePWMxA/ePWMxB

These two check-boxes enable/disable the use of dead bands. By default,
these two boxes are left unchecked. By ticking one or both of them, some
parameters appear.
• Deaband polarity
During the dead band time, ePWMxA and/or ePWMxB outputs have to be
settled to an inactive state. To this aim, this field configures the dead band
polarity, that is, the logic state of the inactive state. This polarity depends
on the specified settings. If only one of the two aforementioned check-boxes
are ticked, the available choices are: Positive (default setting) and Nega-
tive. Instead, every time both check-boxes are selected, the available choices
are four, that is: Active high (AH, default setting), Active low (AL),
Active high complementary (AHC) or Active low complementary
(ALC). Active high/low means that the system is active when the ePWM
output signal is set to a high/low logic state, respectively. Instead, active
high/low complementary is used every time ePWMxB is the inverse of eP-
WMxA.
Firmware Environment: ePWM Peripheral 185

• Signal source for raising edge (RED)

This entry selects the signal source on which RED has to be applied. This
field is available only if Use deadband for ePWMA is enabled. By default
ePWMxA signal is selected.
• Signal source for falling edge (FED)
This entry selects the signal source on which FED has to be applied. This
field is available only if Use deadband for ePWMB is enabled. By default
ePWMxA signal is selected.
• Deadband period source
This option specifies whether the RED and/or FED dead band period is
specified through input port or through dialog. Specify via dialog is the
default selection.
• Deadband Rising edge (RED) period (0∼1023)
This label appears only if the Deadband period source is set on Spec-
ify via dialog and Use deadband for ePWMxA is enabled. This field
specifies the rising edge dead band as multiple of TBCLK. For instance, con-
sidering CPUfrequency = 90 MHz, RED = 15 means 15 · TBCLK ≈ 167 ns.
• Deadband Falling edge (FED) period (0∼1023)
This label appears only if the Deadband period source is set on Specify
via dialog and Use deadband for ePWMxB is enabled.

Block Parameters - Event trigger tab

This tab allows to configure ADC Start of Conversion (SOC) signals through
one or both ePWMxA and ePWMxB outputs. Typically, the ePWMxA is kept
as the master module and it is used to define the ADC Start of Conversion
Event (ePWMSOCxA). PWM interrupt options are not investigated in this
book.

• Enable ADC start of conversion for module A/B

These two check-boxes enable to edit the parameters for the generation of
SOC signals in the correspondence of a specified event.
• Number of event for start of conversion A (SOCA) to be generated
If Enable ADC start of conversion for module A is selected, this
drop down menu appears and it sets the number of the event that triggers
ADC Start of Conversion for Module A (SOCA). Namely, three choices are
available:

– First event: SOC is triggered with every event (default option);

– Second event: SOC is triggered with every second event;
186 Pulse Width Modulator Peripheral

– Third event: SOC is triggered with every third event.

• Start of conversion for module A event selection

If Enable ADC start of conversion for module A is selected, this field
specifies the event (that is, a counter match), that triggers an ADC start of
conversion event. There are several choices. Among all the possibilities, it is
reported:
– Counter equals to period (CTR=PRD): a SOC signal is generated
every time the ePWM counter reaches the period value TBPRD, i.e.
the maximum carrier peak;
– Counter equals to zero or period (CTR=0 or CTR=PRD): an
SOC signal is generated both at the the maximum or minimum carrier
peaks.
These options are particularly important for what concerns current mea-
surement. Indeed, by synchronizing the SOC of an ADC reading related to
a current sensor with CTR=0 or CTR=PRD (or both), the sensed data are
coincident with the average value of the current, thus, avoiding to sense/in-
troduce high frequency contribution (ripple) in the control loop. This con-
cept is exploited in the examples reported in the following Chapters.

13.5 Example with ePWM block

Build a firmware aimed to realize a PWM logic with triangular carrier

at 3 kHz and complementary output by using an ePWM peripheral

Referring to the implementation of the PWM logic proposed in Section 12.4,

this example still uses one of the external potentiometer available on the
extPot3 to vary the modulating signal. This time, the PWM logic is embed-
ded into the ePWM block. The digitalized measurement read from ADCA3
must be scaled between 0 and 1 first and, after that, between 0 and TBPRD
to match with the range of variation of the carrier implemented by the ePWM
peripheral. This exercise is carried out aiming to realize the actuation of
a 2L-VSC (single) leg using a triangular carrier. Reminding that for the
LAUNCHXL F28069M board fck = CPUfrequency = 90 MHz, for an up-down
counting mode it follows:
CPUfrequency
TBPRD = PWMcounter_period = = 15000 (13.4)
2 · PWMfrequency
 Example with ePWM block 187

Firmware Environment
c28069_PWMbasics_F.slx (solver: fixed step-ODE4, step size: Ts = 1/3 kHz)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for Example 1 in Section 10.5.

• Insert an ePWM block, double-click on it and set the following parameters:

– General (Tab)
∗ select module ePWM1;
∗ set Timer period units to Clock cycles (TBCLK);
∗ set Timer period equal to PWMcounter_period ;
∗ verify that Counting mode is set on Up-Down;
∗ leave the other default settings.
– ePWMA (Tab)
∗ tick Enable ePWM1A check-box;
∗ Action when counter=ZERO: do nothing;
∗ Action when counter=period (PRD): do nothing;
∗ Action when counter=CMPA on up-count (CAU): clear;
∗ Action when counter=CMPA on down-count (CAD): set;
∗ Action when counter=CMPB on up-count (CBU): do noth-
ing;
∗ Action when counter=CMPB on down-count (CBD): do
nothing;
∗ set in Compare value reload condition as Load on counter
equals to zero (CTR=zero);
∗ leave the other default settings.
– ePWMB (Tab)
∗ tick Inverted version of ePWMxB check-box. This option en-
ables the implementation of the complement of ePWM1A;
∗ leave the other default settings.
– Counter Compare (Tab)
∗ select clock cycles as CMPA units;
∗ select input port in Specify CMPA via;
∗ set CMPA initial value on 0;
∗ select Counter equals to zero in Reload for compare A register
(SHDWAMODE);
188 Pulse Width Modulator Peripheral
sin50Hz

referred to the PWM_CounterPeriod

potentiometer ePWM1A: J4 - pin40

C2802x/03x/05x/06x lim C2802x/03x/05x/06x

WA
A3
ePWM
ADC

Figure 13.14 Simulink® scheme included in c28069_PWMbasics_F.slx.

∗ select clock cycles in CMPB units;

∗ select Specify via dialog in Speicfy CMPB via;
∗ set CMPB initial value 0;
∗ select Counter equals to zero in Reload for compare B register
(SHDWBMODE).
– Deadband unit (Tab)
∗ tick the Use deadband for ePWM1A check-box;
∗ tick the Use deadband for ePWM1B check-box;
∗ select Active high complementary (AHC) as Deadband polarity;
∗ select ePWMxA as Signal source for raising edge (RED);
∗ select ePWMxA as Signal source for falling edge (FED);
∗ select Specify via dialog in Deadband period source;
∗ set Deadband Raising edge (RED) period (0~1023): 15;
∗ set Deadband Falling edge (RED) period (0~1023): 15.
– Event Trigger (Tab)
∗ select Enable ADC start of conversion for module A;
∗ set Number of event for SOCA to be generated to First
event;
∗ select Counter equals to period (CTR=PRD) as Start of conver-
sion for module A event selection.
– Leave the other tabs set on their default options.

Figure 13.14 shows the complete scheme and its connections. In particular, the
ADC conversion chain from the ADCA3 reading up to the saturation block is
the same as that one reported in Section 12.4. Note that the rate transition is
used to reduce the discretization of the modulating signal, with the value of
m(k) being updated every Ts . It is common practice to set the sampling time
related to the controller, i.e., Ts , according to the chosen switching period
Ts = 1/PWM frequency or half of it Ts = 1/(2 · PWM frequency). If Ts =
1/PWM frequency is chosen, this imply a constant modulating signal over
Tc = 1/fc = 2 · TBPRD. Therefore, in the ePWMA tab, the Compare
value reload condition has to be set as Load on counter equals to zero
because a new CMPA value is evaluated at CTR=ZERO.
Example with ePWM block 189

15000

10000

5000

0 0.5 1 1.5 2 2.5 3

10-3

0 0.5 1 1.5 2 2.5 3

10-3

0 0.5 1 1.5 2 2.5 3

10-3

Figure 13.15 Measured signals at the pins 39 and 40 when the modulating
signal is a constant m(k) = 0.5.

All the system parameters have to be defined and initialized before the
firmware deploy through a separate MATLAB® .m file or included in Model
Properties/Main/InitFcn. An example of such script is reported here in the
following.

%% F28069M clock frequency

CPU_frequency = 90e6;
%% carrier frequency definition
PWM_frequency = 3e3;
Tsw = 1/PWM_frequency
%% TBPRD (counting mode) definition
PWM_counterperiod = CPU_frequency/(2*PWM_frequency)
%% sampling time definition
Ts = Tsw; or Ts = Tsw/2

The modulating signal must be scaled in the range [0, PWMcounter_period ].

Therefore, a gain block is used to compute m(k)
b = m(k) · PWMcounter_period ,
where m(k) ∈ [0, 1]. The result is converted to an int16 data type through
a Data Type Conversion block to be compatible with the ePWM peripheral
registers. The firmware allows to choose (by double-clicking on two different
190 Pulse Width Modulator Peripheral

15000

10000

5000

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02

Figure 13.16 Measured signals at pin 39 and 40 when the modulating signal
is a 50 Hz sine wave.

manual switches) between three operating modes: a constant modulating sig-

nal equal to 0.5 → 1125; a sample-based sine wave at f1 = 50 Hz; a variable
modulating signal provided by the potentiometer. In any case, ePWM1 is
used. Therefore, ePWM1A is sending its output on pin 40, whereas ePWM1B
on pin 39. The output signals connected to these pins can be visualized by
using an oscilloscope, with ground (GND) connection to, e.g., pin 20 of J2.
In case m(k) = 0.5, it is possible to highlight the behavior of the counter as
well as the sample updates for m(k),
b as shown in Figure 13.15. Regarding the
sample-based sine-wave block, the reader is referred to the settings reported
in Section 10.5, with the exception of the sample time. This latter has to be
set based on the Compare value reload condition. Thus, the sine wave is sam-
pled at Ts leading to a pulse number equal to ns = 1/(f1 · Ts ) = 60. The
resulting waveforms are shown in Figure 13.16. In particular, Figure 13.15,
13.16 and 13.17 show the voltage level at the output pins (TTL logic) related
to both ePWM1A and B. The triangular carrier is discretized every TBCLK
even if it is not visible (T BCLK ≈ 11 ns). It is important to note that the
cases m(k) = 0 (left) and m(k) = 1 (right) do not match exactly with a duty
cycle equal to 0 or 1, as shown in Figure 13.17. This is due to the potentiome-
ter, which does not provide an internally generated signal (like the constant
DAC Peripheral—Filtered PWM 191

3 3 3

2 2 2

1 1 1

0 0 0

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
10-3 10-3 10-3

3 3 3

2 2 2

1 1 1

0 0 0

0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
10-3 10-3 10-3

Figure 13.17 Measured signals at pin 39 and 40 when the modulating signal
comes from a potentiometer.

or sine-wave blocks) and, therefore, is never exactly equal to 0 or 1.3 The

ePWM peripheral settings enable the ADC SOC for module A which can be
used to synchronize the ADC reading. This is not particularity useful for this
kind of exercise, but it will be fundamental in case of ADC reading related to
measurements, e.g., phase current.

13.6 DAC Peripheral—Filtered PWM

The F28069M LaunchPad™ board has two ePWM modules which are ded-
icated to provide special Digital to Analog Converters (DACs) features and
they are simply called DACx, where x is the number of the module. The dif-
ference between standard ePWM and DAC peripherals is that these latter
are equipped with RC low pass filters right before their outputs, as shown
in Figure 13.18 (a). If the spectrum of a switching pattern generated by
a PWM logic (e.g., a chopped voltage waveform as that one reported in
Figure 13.16) is considered, the aim of filtered PWM signals is to try

3 Sometimes, it may happens that minimum RED and/or FED are introduced by default

for safety reason, i.e., to avoid short-circuits, but this is not the case in this example.
192 Pulse Width Modulator Peripheral

(a) (b)

Figure 13.18 Hardware configuration of the ePWM peripherals equipped with

the DAC: (a) schematic of the installed output filters on DAC channels and
(b) pins on the F28069M LaunchPad® board connected to those channels [16].

approximate modulating signal in a more accurate way. The cut off frequency
of the RC filter is constant because it is related to the physical components
mounted on the F28069M LaunchPad™ board. Therefore, the quality of the
filtered signal strongly depends on the choice of the switching frequency at
which the PWM logic operates. The LaunchXL F28069M board provides 4
DAC channels, which are connected to ePWM7A, ePWM7B, ePWM8A, and
ePWM8B. Their corresponding output voltages are connected to GPIO40,
GPIO41, GPIO42 and GPIO43, as shown in Figure 13.18 (b).
There are no dedicated Simulink® blocks for DAC peripheral since, at the
firmware level, the same ePWM settings still hold. Nevertheless, the correct
ePWM module must be selected.

13.7 Examples with DAC Peripherals

Example 1

Build a firmware aimed to internally generate a 10 Hz sinusoidal

waveform and visualize it through a DAC channel

Knowing how to operate with an ePWM peripheral, a DAC can be used to

visualize a an internally generated signal to debug a processing chain, e.g., for
debugging purpose.
 Examples with DAC Peripherals 193
DAC1: J4 - pin32

lim C2802x/03x/05x/06x

WA

ePWM

Figure 13.19 Simulink® scheme included in c28069_sinDAC_F.slx.

Firmware Environment
c28069_sinDAC_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted in Section 10.5, Example 1.
• Insert a Sine Wave block and double click on it:

– Set sample based as Sine type;

– Select use simulation time in Time (t);
– Set an Amplitude equal to 0.5;
– Set a Bias equal to 0.5. i.e. sine-wave ranges between 0 and 1;
– Set the number of Samples per period equal to ns = 100;
– Set the number of offset samples equal to 0;
– Set the sample time Tsig = 0.001 s. which implies

1
f1 = = 10 Hz
ns · Tsig

• A gain block is placed in the scheme. Its value is set equal to

PWMcounter_period . This multiplication allows to re-scale the modulation
signal, making it comparable with the carrier.
• The Saturation, Data type conversion (int16) and Rate trasition
blocks have the same meaning (i.e., protection and scaling) as the those
used in previous examples (see Section 13.5).
• An ePWM module is added to the scheme. Module ePWM7 is used in this
example. Its setting for the General and ePWMA tabs are the same as those
reported in Section 13.5. Sub-module ePWM7B is not used (disabled). All
the other settings are left in default conditions.
194 Pulse Width Modulator Peripheral
1

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Figure 13.20 Modulating signal and measured signal at pin 32, referred to
the ground of the board.

Connect the previously mentioned blocks as shown in Figure 13.19.

In order to have a good approximation of the 10 Hz sine wave, the car-
rier frequency is set equal to PWMfrequency = 10 kHz. Considering an
up-down counting mode, it follows that TBPRD = PWMcounter_period =
CPUfrequency /2 · PWMfrequency = 4500. The sampling time is dependent on the
modulating signal updates, i.e., Ts = 1/PWM frequency = 100 µs. The set-
tings of the ePWM block are the same as reported in Section 13.5. The
parameters are initialized by a script similar to the one reported there. The
resulting voltage signal can be checked with an oscilloscope by connecting a
probe to pin 32 of J4, which corresponds to DAC1 (the reader can use pin
20 of J2 to refer the measurement to ground). The result is shown in Figure
13.20.

Example 2

Build a firmware aimed to internally generate a 10 Hz sinusoidal

waveform, visualize it through a DAC channel, read it via an ADC and
visualize the delayed results through another (different) DAC channel

This exercise is an extension of the previous one (Example 1) to underline

how the processing delay affects both measurements and computations. The
internal signal is again sent to DAC1 which is now physically connected to
ADC via cable. The signal is processed and sent to a different DAC channel
(e.g., DAC3) for debugging purpose. The results are visualized through an
oscilloscope.
 Examples with DAC Peripherals 195
DAC1: J4 - pin32

lim1 C2802x/03x/05x/06x

WA

ePWM
ADCA7: J3 - pin23
DAC3: J8 - pin72
C2802x/03x/05x/06x
C2802x/03x/05x/06x
A7
WA
ADC Z-d
ePWM

Figure 13.21 Simulink® scheme included in c28069_sinDACinout_F.slx.

Firmware Environment
c28069_sinDACinout_F.slx (solver: fixed step - ODE4, step size: Ts = 100 µs)

Make a copy of the Simulink® file from the previous exercise (Example 1)
keeping the same settings. Then:

• Copy and paste the previous ePWM block. Double-click on the new copy
and select module ePWM8. Leave all the other settings unchanged, thus,
only ePWM8A is enabled.
• Add an ADC block and use the same configurations as those adopted in
Section 12.4. Note that ADCA7 has to be selected, with sample time equal
to Ts . Use the same approach to process the data in terms of scaling (to
obtain a [0, 1] range) and data type (single).
• Keep the same data processing in terms of saturation, PWMcounter_period
scaling, rate transition (Ts ), and data type (int16).
• Insert a manual switch and a Delay block; set Sample time as Ts and
Delay length equal to 100.

Arrange the aforementioned blocks as shown in Figure 13.21. Channel

ePWM8A corresponds to DAC3, which is connected to pin 72 from J8. Since
DAC1 (ePWM7A) corresponds to pin 32 from J4, ADCA7 is now selected
because it is located in a handy position, i.e., pin 23 from J3. A short cable is
then connecting DAC1 to ADCA7. Therefore, the sine wave approximated by
DAC1 is read, re-processed/manipulated by the LaunchXL F28069M board,
and, then, sent to DAC3. Both ePWM7A and ePWM8A operate with a
fc = 10 kHz carrier frequency, with Ts = 1/PWM frequency = 100 µs. The
oscilloscope measurements can be referred to GND connecting pin 20 from J2.
Since the sine-wave is subject to a double processing (first generated
through a DAC and, then, measured by an ADC), such workflow may be
affect the result introducing both attenuation and computational delay.
Regarding the former, Figure 13.22 shows DAC1 almost aligned with the
internally generated sine wave, while DAC3 presents a slightly different ampli-
tude. This might be due to the ADC reading which processes a noisy waveform
196 Pulse Width Modulator Peripheral
1

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25

0 0.05 0.1 0.15 0.2 0.25

Figure 13.22 Modulating signal and measured signals at pin 32 and 72, both
referred to the ground of the board.

(from DAC1), thus, returning rounded data. Moreover, also data type conver-
sion blocks are included in the processing. The cable connection from DAC1
to ADCA7 might introduce additional noise as well.
Regarding the computational delay, DAC3 is lagging DAC1, even if this
time delay is not fully recognizable from Figure 13.22. This can be somehow
expected since the data read/actuation is scheduled within TBCLK, while
the processing/computations occurs every Ts . Thus, the new data provided
by the ADC reading (i.e., board data in) is delayed with respect to the output
values (i.e., board data out). The processing effect can be emphasized by
manually adding a larger delay in the ADC conversion chain. The manual
switch inserted in the scheme (see Figure 13.21) allows to include a delay
block which is configured to add Td = 100Ts = 0.01 s to the computational
time at each step. This is evident in Figure 13.23, where DAC1 is still aligned
with the internally generated signal, while DAC3 is now delayed by Td (i.e.,
it is shifted to the right).
To make a comparison, this approach is also suitable to describe the effects
observed in Example 2 reported in Section 10.5, where the overall delay was
mainly due to serial communication sampling.

Remark
If the delay is too large, the workflow of the firmware might be corrupted in
terms of synchronization between input/output. As an example, given that
the internal sine-wave frequency is set equal to f1 = 10 Hz, if Td > 0.05 s
(i.e., greater than 1/2f1 ) the MCU starts to adapt the frequency of the sine-
wave approximated by DAC1 (which still operates at carrier frequency, fc =
10 kHz), resulting in lower values than f1 . As an example, running the previous
firmware with Td = 600Ts = 0.06 s, DAC1 returns a 5 Hz sine-wave. Hence, in
Synchronization between Multiple ePWM Modules 197
1

0.8

0.6

0.4

0.2

0 0.05 0.1 0.15 0.2 0.25

0 0.05 0.1 0.15 0.2 0.25

Figure 13.23 Modulating signal and measured signals at pin 32 and 72, both
referred to the ground of the board in presence of large latency Td = 0.01 s.

case of large delay, the PWM carrier frequency should be increased to obtain
more accurate outputs. Note that, no SOC synchronization between ADC and
ePWM/DAC has been introduced yet.

13.8 Synchronization between Multiple ePWM Modules

Synchronization events can be settled between peripherals of different kind
or between modules of the same family. In the previous Sections, all discus-
sions described the operation of a single ePWM module. Nevertheless, multiple
ePWM modules may need to work together in the same firmware. To facilitate
the understanding of this scenario, the ePWM module can be represented by
the simplified block diagram shown in Figure 13.24 . This picture shows the
key resources needed to explain how multiple ePWM modules may working
together by synchronizing their actions. Namely signals SyncIn, SyncOut, and

SyncIn
Phase reg EN
Φ=0� EPWMxA

EPWMxB
CTR = 0
CTR=CMPB
X
SyncOut

Figure 13.24 Simplified representation of the features of an ePWM module

for synchronizing peripherals [21].
198 Pulse Width Modulator Peripheral

EPWM1SYNCI
GPIO
ePWM1 MUX
EPWM1SYNCO

SYNCI
eCAP1
EPWM2SYNCI

ePWM2

EPWM2SYNCO

EPWM3SYNCI

ePWM3

EPWM3SYNCO

EPWMxSYNCI

ePWMx

EPWMxSYNCO

Figure 13.25 Time-Base Counter Synchronization Scheme [21].

Phase reg are explained to this aim. A time-base synchronization scheme con-
nects all the ePWM modules installed on a board. Each ePWM module has
a synchronization input (EPWMxSYNCI), also called SyncIn, and a synchro-
nization output (EPWMxSYNCO), also called SyncOut. The synchronization
input for the first instance (ePWM1) comes from an external pin. The synchro-
nization scheme for the remaining ePWM modules is shown in Figure 13.25.
Each ePWM module can be configured to use or ignore the synchronization
input/output. In the Simulink® environment, such feature can be configured
by acting on:
• Synchronization action (input)
This option allows to specify any possible delay between counters by setting
the reset value of ePWM modules. Actually, this quantity may be specified
through an input port with Set counter to phase value specified via
input port or via dialog with Set counter to phase value specified via
dialog (see Section 13.4.1). The latter setting generates two new parameter
fields: Counting direction after phase synchronization, which can be
set equal to Count up after sync or Count dwon after sync, and Phase
offset value (TBPHS) which has to be defined. This feature enables to
configure the selected direction of the time-base counter immediately after
a synchronization event, such as TBCTR=TBPHS.
• Synchronization output (SYNCO)
Figure 13.25 shows that the synchronization input of the second ePWM
 Synchronization between Multiple ePWM Modules 199

module (ePWM2) can be fed by the synchronization output of ePWM1.

In this case, the carrier waveform internally generated in ePWM2 is syn-
chronized with a suitable triggering signal generated inside ePWM1. More-
over, all the ePWMx modules are chained with increasing x. Therefore, a
specificSynchronization output (SYNCO) mode should be selected for
each module. Indeed, this feature define if the selected ePWM module gen-
erates its own synchronization output to be passed to the next one, i.e.,
by setting Counter equals to zero (CTR=Zero) or Counter equals
to compare B (CTR=CMPB), or if it just passes the SYNCO signal
generated by the previous module, that is, setting the option pass trough
(EPWMxSYNCI or SWFSYNC), as explained in Section 13.4.1.
Combining these two features, an ePWM module can be synchronized with
another one. Lead or lag phase control can be also added to the waveforms
generated by different ePWMs.

Example

Build a firmware aimed to synchronize three ePWM modules

in a master-slave chain configuration

This exercise is an extension of the Example reported in Section 13.5. In this

case, three consecutive ePWM modules, i.e., ePWM1, ePWM2, and ePWM3,
are considered. The main settings to create a master-slave chain configuration
between ePWM modules in Simulink® is presented here in the following.

Firmware Environment
c28069_PWMsync_F.slx (solver: fixed step - ODE4, step size: Ts = 1/3 kHz)

Modify the Simulink® file adopted for the Example proposed in Section 13.5
as follows:
• Make two copies of the ePWM block in order to have three of them in the
scheme;

ePWM1 settings:
• General (Tab)

– Set Synchronization action: to Disable;

– Set Synchronization output (SYNCO): to Counter equals to
zero (CTR=Zero);
– Leave the other parameters unchanged.
200 Pulse Width Modulator Peripheral
ePWM1
C2802x/03x/05x/06x

sin50Hz WA

referred to the PWM_CounterPeriod ePWM

ePWM2
potentiometer
lim C2802x/03x/05x/06x
C2802x/03x/05x/06x

A3 WA

ePWM
ADC
ePWM3
C2802x/03x/05x/06x

WA

ePWM

Figure 13.26 Simulink® scheme included in c28069_PWMsync_F.slx.

ePWM2 settings:
• General (Tab)

– Update the Module to ePWM2

– Set Synchronization action: to Set counter to phase value spec-
ified via dialog;
– Set Counter direction after phase synchronization: to Count
up after sync;
– Set Phase offset value (TBPHS): equal to zero;
– Set Synchronization output (SYNCO): to Pass through (EP-
WMxSYNCI or SWFSYNC);
– Leave the other parameters unchanged.

• Keep the ePWM2B as the complement of ePWM2A.

ePWM3 settings:
• General (Tab)

– Update the Module to ePWM3

– Set Synchronization action: to Set counter to phase value spec-
ified via dialog;
– Set Counter direction after phase synchronization: to Count
up after sync;
– Set Phase offset value (TBPHS): equal to zero;
– Set Syncronization output (SYNCO): to Disable;
– Leave the other parameters unchanged.

• Keep the ePWM3B as the complement of ePWM3A.

 Synchronization between Multiple ePWM Modules 201

0 1 2 3 4 5 6 7 8
10-4

Figure 13.27 Voltage waveforms measured on the pins of the LaunchXL

F28069M board corresponding to ePWM1A, 2A and 3A.

• Leave the settings in the Model Configuration Parameters windows and

those for the remaining blocks (i.e., ADC reading process) of the scheme
unchanged.

The updated Simulink® scheme is reported in Figure 13.26. These settings

make ePWM1 providing a triggering source EPWMxSYNCI to synchronize
both ePWM2 and ePWM3. In particular, the EPWMxSYNCI signal goes from
ePWM1 to ePWM3 through ePWM2 without any phase offset (TBPHS=0).
For ePWM2 and ePWM3, the Counting direction after phase synchro-
nization is set on Count up after sync to synchronize the carriers both in fre-
quency (counter period length) and shape in all three modules. The SYNCO
signal is disabled in the ePWM3 module since there are no further modules to
be triggered (i.e., Synchronization output (SYNCO): Disable). The re-
sulting outputs of the three modules can be measured through an oscilloscope.
The corresponding pins to be visualized are pin 40 for ePWM1A, pin 39 for
ePWM1B, pin 38 for ePWM2A, pin 37 for ePWM2B, pin 36 for ePWM3A and
pin 35 for ePWM3B. Figure 13.27 reports the voltage waveform corresponding
to ePWM1A, 2A and 3A.
To provide a better understanding of the adopted master-slave configura-
tion, Figure 13.28 reports a simplified block diagram.

SyncOut SyncOut SyncOut

X X X
CTR=CMPB CTR=CMPB CTR=CMPB
A
CTR = 0 CTR = 0 CTR = 0
EPWM1B EPWM2B EPWM3
B
Φ=0� EPWM1A Φ=0� EPWM2A Φ=0� EPWM3
Phase reg EN Phase reg EN Phase reg EN
1 SyncIn 2 SyncIn 3 SyncIn

Figure 13.28 Simplified block diagram explaining the master-slave synchro-

nization between three ePWM modules [21].
 202 Pulse Width Modulator Peripheral

13.9 Synchronization between ADC and ePWM Mod-

ules: Average Measurements
The synchronization between an ADC and an ePWM block is an important
subject in many applications involving switching devices. Particularly, it is
fundamental every time average values of measurements must be processed
into closed-loop control schemes. An ePWM module is able to send a syn-
chronization signal that triggers the SOC of an ADC module (or multiple
modules) when the carrier signal (i.e., TBCTR) equals a specific value (i.e.,
TBPRD). To this aim, the following settings are summarized here below:
• Event Trigger (Tab)

– Select Enable ADC start of conversion for module A;

– Set Number of event for SOCA to be generated to First event;
– Select Counter equals to period (CTR=PRD) as Start of conversion
for module A event selection.
When the carrier signal reaches the maximum value TBPRD, a SOC is sent
to the ADC. Nevertheless, this latter has to be configured as well:
• Select ADCINA0 as Conversion channel in the Input Channels tab.
Then, go back in the SOC Trigger tab;
• Sampling mode: Single sampling mode;

• SOC trigger number: SOC0;

• SOCx acqusition window: 7;
• SOCx trigger source: ePWM1_ADCSOCA;
• ADC will trigger SOCx: No ADCINT;

• Sample time: Ts ;
• Data type: uint16;
• Flag Post interrupt at EOC trigger (optional);

• Interrupt selection: ADCINT1 (optional);

• Flag ADCINT1 continous mode (optional).
The key parameter in this list of instruction is the SOCx trigger source,
which is set to follow a signal coming from the ePWM module.
This features is really useful when a current measurements (typically made
of a fundamental and several high frequency components) has to be processed
Synchronization between ADC and ePWM Modules: Average Measurements 203

shunt current
actual curent
carrier signal

Figure 13.29 Example of synchronization between ePWM and ADC modules:

measured current is compared to the actual current flowing through the low-
side shunt. Moreover, switching patterns of the considered leg is reported as
well in the bottom plot. The converter installed on the DRV8301 expansion
board is operated as a DC/DC step-down converter.

into a closed-loop control based on linear controllers. Indeed, these latter may
work poorly in presence of high ripple. Thus, averaging may be necessary.
As an example, a setup which comprise a control platform (LaunchXL
F28069M) and a converter (BOOSTXL-DRV8301) equipped with current sen-
sors is considered. In this particular case, phase current measurements are
provided by some low-side shunts (see Section 3.2). By using the previous
synchronization settings, ADCA0 starts the acquisition window when ePWM1
module equals CTR=PRD (an up-down counting mode is adopted). This SOC
event is chosen since it occurs in the correspondence of the theoretical mean
value of the current, avoiding to process any ripple inside the MCU. This result
is shown in Figure 13.29. Note that in presence of an highly distorted current,
this approach still limits the amount of noise processed in the closed-loop
scheme. Therefore, when an average measurements is required, the firmware
should be settled to follow this proposed approach (i.e., ePWM and ADC
synchronization).

Additional Consideration on Example 2

Since the sine-wave approximated by DAC1 has to be read by ADCA7 chan-
nel, the latter has to by synchronized with ePWM7A module. To this aim,
the Event Trigger tab settings are no more an optional and they must be
correctly configured. In particular:
• Event Trigger (Tab)
 204 Pulse Width Modulator Peripheral

– Select Enable ADC start of conversion for module A;

– Set Number of event for SOCA to be generated to First event;
– Select Counter equals to period (CTR=PRD) as Start of conversion
for module A event selection.

Instead, for ePWM8A module the settings can be the same as those used in
Example 1. Both ePWM7A and ePWM8A still operate with a 10 kHz carrier.
Once those module are set, the ADC block can configured for ADCA7 channel
as follows:
• Select ADCINA7 as Conversion channel in the Input Channels tab.
Then, go back to the SOC Trigger tab:
• Sampling mode: Single sampling mode;
• SOC trigger number: SOC0;

• SOCx acqusition window: 7;

• SOCx trigger source: ePWM7_ADCSOCA;
• ADC will trigger SOCx: No ADCINT;
• Sample time: Ts ;

• Data type: uint16;

• Flag Post interrupt at EOC trigger;
• Interrupt selection: ADCINT1;
• Flag ADCINT1 continous mode.

The rest of the scheme is left unchanged.

13.10 Events Execution within Sample Time

The synchronization between ADC and ePWM modules to obtain average
measurements is an example of how important is to identify what is the goal
of a closed-loop scheme (and, therefore, which kind of measurements/actuation
is going to be used/adopted) and how to choose the specific settings to avoid
issues with related to peripheral and task scheduling early during the design
stage. Many challenges are associated to integrate MCU scheduling and
peripherals in power electronics-based applications. Typical examples
are:
 Events Execution within Sample Time 205

Figure 13.30 Example of possible operations (event sequence, interrupt trigger,

and control algorithm execution time) to be performed on a target hardware
within the sample time Ts .

• ADC-ePWM synchronization to achieve current sensing in the correspon-

dence of its theoretical average value (see previous Section);
• Compensate sensor delays to achieve the desired controller response for a
closed loop system;
• Studying different modulation strategies (or ePWM settings) while designing
special algorithms.
The importance of understanding the hardware details of each peripheral (in-
cluding the identification of the effect of its parameters/registers/operating
modes) becomes more and more crucial as the complexity of a firmware in-
creases. As an example, on the F28069M LaunchPad™ Piccolo™ board the
ADC hardware contains a sample and hold circuit to sense analog inputs (see
Chapter 12). To ensure complete ADC measurement, the minimum acquisition
time must be adjusted to account for the combined effects of input circuit and
the capacitor installed in the sample and hold circuit. This means, in practice,
to update the SOCx acquisition window accordingly, i.e., choosing values
grater than 7. Another important aspect is the insight of the execution of
events. To this purpose, a firmware involving one ADC channel, one ePWM
module and a controller which use the ADC reading to compute the modu-
lating signal for the ePWM is considered. An example of possible operations
that can occur during each sample time is reported in Figure 13.30. Indeed,
 206 Pulse Width Modulator Peripheral

the controller operation or other processing/computation routines are sched-

uled to be executed within Ts = Tsw = 1/fsw . Therefore, the controller is
synchronized with the ADC interrupt. The event sequence which is scheduled
on the target hardware foresees that:

• At the beginning of Ts , when the PWM counter value equals the PWM
counter period (TBCR=TBPRD), the ePWM peripheral (which is center-
aligned in this case, that is, up-down counting mode) triggers the start-of-
conversion (SOC) event for the ADC module;
• The ADC module converts the sampled analog signal into digital counts and
triggers the end-of-conversion (EOC) event;
• The EOC triggers the ADC interrupt for the controller, which reads the
phase current value;
• Assuming that any other reference value is already provided, the controller
starts performing the required operations necessary to the computation of
the modulating signal for the ePWM module. The number as well as the
complexity of these operations define the length of this execution window;
• When the controller execution is done, the resulting modulating signal is
sent to the ePWM input and hold on until the end of the Ts .
14
Encoder Peripheral

The enhanced quadrature encoder pulse (eQEP) modules are used for direct
interface with linear or rotary encoders to get position, direction and speed
information from the rotating parts of a system. Typically, they are used for
measuring the speed or the position of rotors in electrical machines driven by
power electronics. Thus, this peripheral is used for closing speed or position
control loops. This Chapter provides some insight in the working principles of
this peripheral and of optical rotary encoders.

14.1 Operating Principle of Incremental Encoders

A rotary incremental encoder is a disk with a track of slots placed along its
periphery, as shown in Figure 14.1. These slots create an alternating pattern
of dark and light lines which can be detected, for example, by optical sensors.
A disk count is defined as the number of dark and light line pairs that occur
to complete a revolution (lines per revolution). A second track is added on the
disk to generate a signal that occurs only once per revolution. This is typically
known as index signal. Different encoder manufacturers may call this signal
using several other terms with similar meaning such as marker, home position
or zero reference. Commonly, to derive direction information, the first track on
the disk is read by two different photo-elements/optical sensors that provide
output (digital) pulses whenever a dark/light line pairs is measured. The two
sensors are placed in such a way to look at the disk pattern with a mechanical
shift of 1/4 of the pitch running between two consecutive lines. Hence, as the

Figure 14.1 Generic structure of an optical rotary encoder disk and corre-
sponding channel signals [20].

DOI: 10.1201/9781003196938-14 207

208 Encoder Peripheral

(a)

(b)

Figure 14.2 Working principle of a quadrature encoder: clockwise (a) and

counterclockwise (b) rotation.

disk rotates, the two optical sensors generate square waves signals that are
shifted by 90° out of phase from each other.
For the F28069M LaunchPad™ Piccolo™ board, these waveforms are called
QEPA (or simply A) and QEPB (or simply B) signals, while the index signal
QEPI (or simply I). QEPA and QEPB channels in quadrature (i.e., quadra-
ture encoders) means that when the incremental encoder is moving at a
constant speed, the duty cycle of each pulse is 50% (i.e., the waveform is a
square wave) realizing the 90° phase difference between A and B. Typically,
the clockwise direction is defined as the QEPA channel going positive before
the QEPB channel, as shown in Figure 14.2. Unlike absolute encoders, an
incremental encoder is not able to indicate absolute position. Indeed, it only
reports variations in position and the direction of the motion for every angular
change.
In general, direct coupled encoders (i.e., mounted on the rotor shaft be-
fore a gearbox) make one revolution for each complete rotation of the motor.
Therefore, the frequency of the digital signal coming from the QEPA and
QEPB outputs varies proportionally with the motor speed; high frequencies
indicate high speeds, whereas low frequencies mean slow speeds. The resolu-
tion of an incremental encoder refers to the precision of the produced posi-
tion information. In the case of rotary encoders, resolution is specified as the
number of pulses per revolution (ppr) or cycles/counts per revolution (cpr).
The ppr determines the number of pulses that can be counted on each signal
(QEPA and/or QEPB) in each complete revolution. Nevertheless, quadra-
ture encoders are able to increase their effective resolution by combining the
Hardware Details 209
+5V +5V

R13 1K

R29 1K

R48 1K

R49 1K

R50 1K

R51 1K
A 1 EQEP1A 1 EQEP2A
B 2 EQEP1B 2 EQEP2B
I 3 EQEP1I 3 EQEP2I

0.001u

0.001u

0.001u

0.001u

0.001u

0.001u
PWR 4 4
GND 5 5

C50

C51

C52

C53

C54

C55
QEP_A QEP_B

GND GND

Figure 14.3 Correspondence between eQEP channels and pin of the

LaunchPad™ F28069M board [16].

information carried by QEPA and QEPB to define the cpr.

cpr = 4 · ppr (14.1)

For instance, if a 1 ppr quadrature encoder is considered, its effective resolution

is increased up to 4 cpr (see the numbers reported in Figure 14.2).
Therefore, the measurement resolution of the encoder is defined as
the smallest position change that the encoder can detect. Every signal edge on
QEPA or QEPB indicates a detected position change. Since each square-wave
cycle on A (or B) encompasses four signal edges (rising A, rising B, falling A
and falling B), the measurement resolution of the encoder equals one-fourth
of the variation represented by a full QEPA (or QEPB) output cycle.

14.2 Hardware Details

The LaunchXL F28069M can be connected up to two linear and/or rotary
encoders, thanks to the presence of two eQEP channels. The interface between
board and encoder is reported in Figure 14.3. As already discussed, A, B
and index signals are called QEPxA, QEPxB and QEPxI, respectively, where
x denotes the considered module. The signals A and B are the output of
the optical sensors and they can be combined to increase the resolution of the
encoder, other than providing information on the direction of rotation. The
two channels use the index signal to assign the reference point from which the
position is encoded. It is possible to program the peripheral in such a way
that this signal resets the position counter at each revolution. In addition, it
can be used to initialize or latching the position counter every time a desired
event occurs on the index pin.
These modules are able to deal with strobe inputs eQEPxS as well. This
kind of signal can initialize or latch the position counter when a desired event
occurs on the strobe pin. This signal is typically connected to a sensor or
to a limit switch to notify that the motor reaches a certain position. Actu-
ally, the default settings of the board LaunchPad™ F28069M do not foresee
any physical pin to be connected to this signal. Indeed, it is possible to pro-
gram the correspondence between hardware pins and eQEP signals in the
210 Encoder Peripheral

Figure 14.4 LPD3806-600BM-G5-24C incremental rotary encoder.

Configuration Parameters menu, Hardware Implementation, Target hardware

resources, eQEP.

14.3 Optical Rotary Encoder LPD3806

In this book, the optical rotary encoder LPD3806-600BM-G5-24C is used.
This piece of equipment is reported in Figure 14.4. This latter provides A/B
channel quadrature-encoded with 600 ppr (thus, 2400 cpr). The index signal is
not available. The supply voltage can range from 5 up to 24 V. The maximum
measurable mechanical speed is 5000 rpm, with a frequency response in the
range [0, 20] kHz. The body height is 38 mm, whereas the size of the shaft is
6 × 13 mm.
The encoder interface is an electronic circuit (LED/photo-transistor
module-based) that processes the A/B signals, making the resulting informa-
tion available to external circuitry. The encoder interface has open collector
outputs with 20 Ω resistors for protection. The circuit draws around 30 mA.
Since it can not source current, the open-collector circuit must be connected to
positive DC voltage through a pull-up resistor. The eQEP pins of the F28069M
LaunchPad™ board have pull-up resistors1 (i.e., they are supplied with the
onboard 5 V voltage level), see Figure 14.3. Attention must be paid to limit
the cable length connecting the encoder to the board, since it is preferable
to have pull-up resistors in close proximity to the encoder interface to im-
prove noise immunity. Note that the encoder interface also has a 5 V linear
regulator onboard (LM7805), which power dissipation has to be kept within
reasonable limits. The internal LM7805 regulator in D-PAK package relies on
a big ground plane that acts as a heat-sink. Moreover, the encoder case is
connected to the chassis ground and not to circuit ground. More hardware
details are given in Appendix B.
1 If the encoder is not connected to the MCU board, the A/B signals are not visible

even with an oscilloscope; the open collector output has no voltage when there is no pull-up
resistor.
 Speed Computation 211

Figure 14.5 Encoder connected to the F28069M LaunchPad™ board through

the extPot3 custom. eQEP2 is considered. In this picture, the encoder is me-
chanically coupled to a DC motor.

In the examples reported in the following Chapters, extPot3 custom board is

considered since it provides a direct connection of the encoder terminals with
the MCU board, as shown in Figure 14.5. The wiring is:

• Red wire: voltage supply (VCC) • Green wire: A

• Black wire: ground (GND) • White wire: B

A or B wires must not be connected to VCC. Otherwise, the output stage

is shot-circuited. The correct connection between encoder and the board is
reported in Figure 14.5.

14.4 Speed Computation

Incremental encoder interfaces detect mechanical displacement without di-
rectly measuring the angular speed. Consequently, speed must be indirectly
computed either by counting or by timing the encoder output pulses (edges).
The resulting value is a frequency or a period, respectively, from which speed
can be calculated. The speed is proportional to frequency, and inversely pro-
portional to period.
In discrete time the signals are inherently quantized (e.g., interface cir-
cuits). The encoder interface samples A and B output signals frequently
212 Encoder Peripheral

enough to detect every A/B state change before the following variation occurs.
Upon detecting a state change, the position counts xp (k) is incremented or
decremented based on whether A leads or lags B. This is typically done by
storing a copy of the previous A/B state and, upon state change, compare the
current and previous A/B states to determine movement direction.
Therefore, signals QEPA and QEPB are both processed for computing
angular speeds. Nevertheless, regarding the implementation, estimating the
speed from a digital position sensor is a cost-effective strategy in motor control.
Hence, two different first order approximations for the speed computation may
be written starting from the position counter xp (k), namely:
• Standard approach:
xp (k) − xp (k − 1) ∆x(k)
ω(k) ≈ = (14.2)
Tw Tw
where ω(k) is the speed computed at time step k, xp (k) and xp (k − 1)
are the position counters at time step k and k − 1 respectively, Tw is the
acquisition time window, which is the inverse of speed calculation rate (and is
known in advance). This approach is based on the fact that the encoder count
(position) is read once at the beginning of each acquisition time period, and
the width of each pulse is a function of the motor speed and the resolution
of the sensor.
However, this method has an inherent accuracy limit directly related to the
ratio between the resolution of the position sensor and the acquisition time
period Tw that has to be chosen.
As an example, considering the LPD3806-600BM-G5-24C encoder, a 600 ppr
quadrature encoder results in 2400 cpr which implies a resolution of
1
res = 360° = 0.15° (14.3)
cpr
|{z}
1
revolution

Considering a speed calculation rate of 400 Hz, thus, Tw = 1/400 = 2.5 ms,
it follows that:
round revolution 1
ωmin = or = 60 = 10 rpm (14.4)
min min cpr · Tw
| {z }
revolution
sec

Assuming that this motor has to measure the angular frequency of a mo-
tor, which rated speed is 2000 rpm, ωmin represents the 10 rpm/2000 rpm ≈
0.42% of this quantity. While this resolution may be satisfactory from mod-
erate to high speeds, it would behave poorly at very slow rotations. Indeed,
the estimated speed would erroneously be zero much of the time for pulsa-
tions lower than 10 rpm. Note that, by varying Tw , different minimum speed
can be computed, e.g., for Tw = 1/1e3 = 1 ms it results ωmin = 25 rpm.
Firmware Environment: eQEP Peripheral 213
C28x

qposcnt

eQEP

Figure 14.6 C2806x eQEP block.

• Low speed approach:

X X
ω(k) ≈ = (14.5)
tp (k) − tp (k − 1) ∆tp

where tp (k) and tp (k − 1) are the time instant at time step k and k − 1
in which a channel A/B present a rising edge, while X is the position step,
that is, one. Equation (14.5) provides a more accurate approach at low
speed rather than (14.2). Such approach computes the speed by measuring
the elapsed time between consecutive edges. An acquisition window is not
present, while the encoder interface is required to work with a high sampling
time in order to detect rising edges even at high speed. Indeed, this approach
leads to better results at low speeds while suffering every time the motor
spins at such speed for which the sensor makes the time interval ∆tp small,
difficult to measure and greatly influenced by the timer resolution. Therefore,
considerable speed estimation errors can be introduced.
For motor control applications characterized by large speed ranges, these two
approaches can be combined by using (14.5) at low speed while switching to
(14.2) every time the motor speed overcomes a specified threshold.

14.5 Firmware Environment: eQEP Peripheral

This section shows the main characteristics of the eQEP block (see 14.6)
available in Simulink® inside the library Embedded Coder Support Package for
Texas Instrument C2000 Processors, subset C2806x.

14.5.1 C2806x eQEP

This block (see Figure 14.6) can be included into a firmware to compute both
the position and the angular speed of a rotating element. Therefore, these
measurements can be used for closing both speed or position control loops.
Its main settings are briefly explained here in the following.

Block Parameters—General tab

• Module
214 Encoder Peripheral

This drop down menu allows to select which module is in use. The adopted
board has two modules only, i.e., eQEP1 and eQEP2.
• Position counter mode
Depending on the structure of the encoder, there are four ways in which the
device and the peripheral can be interfaced:

– Quadrature count: this option allows to use the quadrature decoder,

which exploits two 90° out-of-phase signals (QEPxA and QEPxB) to
generate quadrature-clock and direction signals. Therefore, the resolu-
tion of the encoder is increased;
– Direction count: direction and clock signals are provided directly
from an external source. Some position encoders have this type of out-
put instead of quadrature output. The QEPxA pin is used for the clock
input XCLK, whereas the QEPxB pin for the direction input XDIR;
– Up-count: the position counter is exploited for measuring the fre-
quency of the QEPA signal. The direction of the counter is hard-wired
for up-count;
– Down-count: the position counter is exploited for measuring the fre-
quency of the QEPA signal. The direction of the counter is hard-wired
for down-count.

• Positive rotation
This option allows to select the convention for positive speed, i.e., clockwise
or anti-clockwise, depending on how the encoder is mechanically coupled
with a motor.
• Sample Time
This setting specifies how often the position counter is read. In case of the
standard speed computation approach, Equation (14.2), this refer to Tw
acquisition time period.

Block Parameters—Position Counter tab

In this tab, it is possible to enable (flag) the output position counter, to
set the Maximum position counter value and to choose the position
counter reset mode. Particular attention should be paid while setting the
maximum position counter value since it should be set equal to the maximum
feasible integer value depending on the chosen data type, which can be set in
the Signal data types tab.
As an example, if int32 is chosen in the Signal data types tab, the
Maximum position counter value is a number that ranges between 0 and
232 − 1. By enabling the software initialization, it is possible to choose the
Software initialization source as Set to init value at startup or via
Input port. The Initialization value is a number between 0 and 232 − 1;
 Example with eQEP block 215

0 is set by defualt. There are different options for Position counter reset
mode, e.g., Reset on the maximum position.

Block Parameters—Speed Calculation tab

The edit of the settings available in this tab is enabled through the checkbox
Enable eQEP capture. Once ticked, different timer and values can be sent
as outputs. Moreover, this tab allows to set prescaler applied on the system
clock to simplify speed computations at low speeds.
Among all the check-boxes that are reported in this tab, two of them are
briefly described here in the following:

• Output capture timer: this option enables the block to output the cap-
ture timer value in clock cycles stored inside the quadrature capture timer
(QCTMR) register, in the correspondence of the last quadrature signal. Port
qctmr is generated;
• Output capture period timer: this option enables the block to output
the capture period timer value (in clock cycles) stored in the quadrature
capture period (QCPRD) register, that is, the time elapsed between two
consecutive quadrature signals (i.e., positions). This value can be used for
speed computation purposes.

Block Parameters—Other tabs

• Compare output tab: enables synchronization signal for position compar-
ison in case of multiple blocks or triggered output signals.
• Watchdog unit tab: enables a Watchdog timer to detect misoperations.

• Interrupt tab: there, several kind of Interrupts can be enabled. In particu-

lar, eQEP module can generate Interrupt signals but it cannot be triggered
directly by any interrupts unlike the ADC and ePWM blocks.

14.6 Example with eQEP block

Build a firmware aimed to read encoder positions, compute angular

speeds and visualize pulse changes through the red led

Angular speed must be indirectly computed by counting the encoder output

pulses on QEPxA and QEPxB. When the encoder interface detect a state
variation, it increments the counter xp (k). This information is available at the
 216 Encoder Peripheral

eQEPx output and used to compute a frequency or period from which speed
can be calculated. This exercise is carried out by considering a LPD3806-
600BM-G5-24C incremental rotary encoder, aiming to implement the stan-
dard computation approach by using Equation (14.2).

Firmware Environment
c28069_encoder_F.slx (solver: fixed step-ODE4, step size: Ts = 100 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Edit an initialization script, e.g., in Model Properites/Callback-
s/InitFcn. Then:

• Insert a eQEP block, double-click on it and set the following parameters:

– General (Tab)
∗ select module eQEP2;
∗ set Position counter mode to Quadrature-count;
∗ set Positive rotation equal to Clockwise;
∗ set Sample time equal to Tw ;
∗ leave the other default settings.

– Position counter (Tab)

∗ flag Output position counter;
∗ set Maximum position counter to 232 − 1;
∗ flag Enable software initialization ;
∗ set Software initialization source equal to Set to init value at
start up;
∗ set Initialization value equal to 0;
∗ set Position counter reset mode equal to Reset on the maxi-
mum position;
∗ Leave the other default settings.

– Signal data types (Tab)

∗ set Position counter value data type equal to int32;

– Leave the other tabs with their default settings.

• Insert a Delay block with Sample time equal to Tw and Delay length
equal to 1 (i.e., one step delay).
• Insert an Add block followed by a Data type conversion block set to
single. This process the difference between xp (k) and xp (k − 1) which is
then multiplied by a Gain block set to 60/(cpr · Tw ). The result is in rpm,
which can be translated in rad/s by multiplying this quantity by π/30.
 Example with eQEP block 217
C2806x
C28x GPIOx
qposcnt GPIO DO
w
led red
eQEP Z-1
encoder

Figure 14.7 Simulink® scheme included in c28069_encoder_F.slx.

• Insert a GPIO block with the related data type as reported in Chapter 11
(red led is connected to GPIO34);

The whole Simulink® scheme is shown in Figure 14.7. Every time the position
signals QEP2A e QEP2B are sampled (discrete time signals), the pulses (or
pulse edges) are detected and counted by the encoder interface. The MCU has
read access to the interface and the eQEP module returns directly the position
counter through qposcnt. The latter is then used to compute the speed by
taking the actual position count xp (k), i.e., qposcnt(k), and the previous one
xp (k − 1), i.e., qposcnt(k-1), within a specific acquisition period Tw . This is
realized by using an add and delay block, with sample time Tw . The result is
then multiplied by 60/(cpr·Tw ). Considering a speed calculation rate of 400 Hz,
it follows Tw = 1/400 = 2.5 ms. Note that, the sample time Ts is related to the
switching period of the modulator, e.g. Ts = Tsw = 1/fsw = 1/10e3 = 100 µs.
Hence, it follows Tw = 25 · Ts , which underline that Tw is (quite) often a
multiple of Ts to keep an hierarchical scheduling of firmware executions.
A coefficient equal to 60 is used since 1/(cpr · Tw ) is expressed as counts
per unit time (e.g., counts per second), while, in practice, it may be necessary
to translate the speed in units such as revolutions per minute (rpm). Such
scaling factor takes into account the relationship between counts and desired
distance units, as well as the ratio between the sampling period and desired
time units. The other coefficient depends on the fact that LPD3806-600BM-
G5-24C encoder produces 2400 counts per revolution (cpr). The speed is finally
multiplied by another scaling factor, i.e., π/30, to move from rpm to rad/s.
Given the encoder connection on eQEP2 as shown in Figure 14.5, the firmware
can be tested by manually rotating the encoder shaft. The qposcnt value is
connected to the red led (GPIO34) to visualize the pulse frequency changes

Figure 14.8 QEP2A (pin 54) and QEP2B (pin 55) waveforms, measured
through an oscilloscope for slow rotational speeds.
218 Encoder Peripheral

Figure 14.9 Detailed QEP2A (pin 54) and QEP2B (pin 55) waveforms in case
of slow and fast rotation scenarios.

as led blinking. Nevertheless, the qposcnt behavior can be better understood

by connecting two voltage probes to QEP2A (pin54) and QEP2B (pin55)
through the available test points on the extPot3 to visualize the corresponding
waveforms. By manually increasing the rotating speed of the encoder, the pulse
frequency increase, thus, increasing qposcnt. This reflects on the behavior of
QEP2A (pin 54) and QEP2B (pin 55), which is shown in Figure 14.8. The
waveforms related to voltage signals varies in a [0, 5] V range.2 Two detailed
views of these signals are reported in Figure 14.9, in which slow and fast
rotations are compared, respectively.
The effect is that pulse frequency changes as fast as the shaft rotation, which
can be increased further by (manually) creating really quick speed transi-
tion from slow to high rotational speed, as reported in Figure 14.10 (voltages
measured at pin 54 and 55).

Figure 14.10 QEP2A (pin 54) and QEP2B (pin 55) waveforms in case a fast
transition from slow to high rotational speed is considered.
.

2 Both cables and encoder interface circuits (even considering the voltage regulator)

impact on the output with a voltage drop.

 Part IV

Real-Time Control in
Power Electronics:
Applications
 15
Open Loop Control of a Permanent Magnet
DC Motor

After a brief overview of the main peripherals and their main settings for the
F28069M LaunchPad™ Piccolo™ board, the reader is guided in this part of
the book where some experimental open- and close-loop implementations are
described step-by-step. Starting from simple examples with DC motors and
custom RL(C) loads, the reader will be shown how to solve all the proposed
control problems starting from the preliminary Simulink® simulations, then
going through the design of the PI controller and finally implementing the re-
quired control loops. In particular, this Chapter shows some examples of open-
loop control algorithms for DC Motors using the Simulink® blocks introduced
in the previous Sections coming from the Embedded Coder Support Package for
Texas Instruments C2000 Processors library. The objective is to drive a two-
pole permanent magnet DC (PMDC) motor through a BOOSTXL-DRV8301
converter connected to the LaunchXL F28069M board. The DRV8301 boost-
erpack presents three independent MOSFET-based switching legs. It is worth
noting that the actuation of a PMDC motor does not require the use of all of
them (i.e., due to its DC nature). However, the converter legs can be physi-
cally connected to create some specific topologies. In particular, two conversion
stage configurations are investigated:
• Half-Bridge (also known as Two-Quadrants converter);
• Full-Bridge (also known as H-Bridge or Four-Quadrants converter).
Both stages aim to convert a fixed input DC voltage, VDC , to a variable/con-
trolled output DC voltage, which is provided at the PMDC terminals. Such
DC-DC converters are also called choppers.

15.1 Required Hardware

• TI LaunchPad™ Piccolo™ F28069M and BOOSTXL-DRV8301 boards;
• Power supply;
• PMDC motor;

DOI: 10.1201/9781003196938-15 221

 222 Open Loop Control of a Permanent Magnet DC Motor

• extPot3 custom board (see Appendix B for technical details).

The reader is referred to Section 3.2 for all the details on BOOSTXL-
DRV8301 board. Since the three legs of the converter are independent, it
is possible to realize the two aforementioned configurations by changing the
physical connections between the motor and the power converter. Then, the
related ePWM modules must be enabled.
It is important to note that every time the converter is mounted on the
microcontroller board, it is preferable to feed both of them through the power
supply only. Indeed, jumpers JP1 and JP2 should be removed (further details
are reported in Section 3.1.3) to isolate the USB port. In this case, the USB
connection is used to deploy of the firmware and for the serial communication
between the LaunchPad™ and the PC only. Moreover, it should be remem-
bered that the BOOSTXL-DRV8301 converter must be supplied with a DC
voltage which is VDC > 6 V, to allow the proper working of the boosterpack
regardless of the state of the jumpers JP1 and JP2. The reader is referred to
the rated voltage limits of the converter reported in Section 3.2.
Now, the LaunchPad™ F28069M board connected to the BOOSTXL-
DRV8301 converter are programmed to drive the PMDC motor in open loop
by manipulating the duty cycle d(t). Both the possible converter configura-
tions are considered. This is achieved by using one of linear potentiometers
mounted on the extPot3 custom board. As shown in Section 12.4, the rotation
of the potentiometer changes an internal resistance, i.e., varying the ratio of a
voltage divider. The resulting voltage value is read by the ADC peripheral that
converts it into a digital value. By using a proper scaling (like those presented
in Chapter 13), this latter can be used as a modulating signal for the ePWM
peripheral. Depending on the chosen topology, one or two ePMW modules
are operated in up-down counting mode at fsw = PWMfrequency = 20 kHz.
Reminding that for the F28069M board fck = CPUfrequency = 90 MHz, it
follows:
CPUfrequency
TBPRD = PWMcounter−period = = 2250 (15.1)
2 · PWMfrequency

where the sampling time is Ts = Tsw = 1/PWMfrequency .

15.2 Linear Model of a PMDC Motor

A PMDC motor has a stationary set of magnets in the stator and an armature
with one or more windings of insulated wire wrapped around a soft iron core
that concentrates the magnetic field. The windings usually have multiple wire
turns around the core, with their ends connected to a commutator. The latter
allows each armature coil to be energized in turn and connects the rotating
Linear Model of a PMDC Motor 223

ia (t) La Ra

va (t) e (t)

Figure 15.1 Equivalent circuit of a PMDC motor.

coils with the external power supply through brushes. A coil of wire with a
current running through it generates an electromagnetic field aligned with
the center of the coil. Direction and magnitude of the magnetic field produced
by the coil can be changed with the direction and magnitude of the current
flowing through it.
Focusing on the system to control, the equivalent circuit of the PMDC
motor is shown in Figure 15.1 and its electrical and mechanical equations are:
dia (t)
La =va (t) − Ra ia (t) − e (t) (15.2)
dt
dΩ(t)
J =me (t) − ml (t) − βΩ(t) (15.3)
dt
where La is the armature inductance, Ra is the armature resistance, ia (t)
is the armature current (i.e., the current flowing through the winding of the
motor), va (t) is the armature voltage (i.e., the voltage applied to the terminals
of the machine), e is the back electromotive force (back-emf) induced on the
armature winding, J is the equivalent moment of inertia of the machine, β
is the rolling friction coefficient, Ω(t) is the mechanical speed of the rotor,
me (t) is the torque generated by the motor and ml (t) is load torque that may
be applied to the electrical machine. In DC machines, both electromagnetic
torque me (t) and back-emf e(t) depend on the magnetic flux linked with the
armature windings ψae and the related constants kT and ke , respectively. As
the name says, the main feature of PMDC machines is the presence of perma-
nent magnets (PM) on the stator to provide the magnetic field against which
the rotor field interacts to produce torque.1 The use of PMs is convenient
in small motors to eliminate the power consumption of the field windings,
thus, increasing their efficiency. Moreover, large PMs are costly and difficult
to assemble; this favors wound fields for large machines. To minimize overall
weight and size, small PMDC motors may use high energy magnets made by
peculiar alloys, e.g., with rare earth materials such as neodymium-iron-boron
or samarium-cobalt.
The PMs create a constant flux density B (if no field weakening tech-
nique is applied) and, accordingly, a constant excitation flux which is defined
1 Therefore, the PMs replace the excitation circuit installed in separately excited DC

motor, which is not considered here.

224 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.2 First magnetization curve of two PMs (no hysteresis is represented
here) and internal structure of the PMDC motor used in this exercise.

PM (ψPM = ψae (t)), see Figure 15.2. This is the case for all the examples
reported in this book which use PMDC motor. Moreover, kT and ke can be ap-
proximately considered equal, assuming that the energy conversion is almost
perfect. Thus, a new constant K can be defined as K = ψPM kT = ψPM ke .
Therefore, the electrical torque and the back emf can be computed as:

me (t) = Kia (t) (15.4)

e(t) = KΩ(t) (15.5)

with me (t) which is function of ia (t) only. The voltage va (t) is a controllable
variable, i.e., u(t) , which allows to increase/decrease the current flowing into
the PMDC motor (according to the system parameters). Assuming that ml (t)
is a controllable disturbance related to the characteristic of the mechanical
load, the vector of the inputs is u(t) = [va (t) ml (t)]T . As shown in Chapter
5, by considering2 x(t) = [Ω(t) ia (t)]T and y(t) = x(t), the equations (15.2)-
(15.5) can be summarized by the following state-space representation:

dx(t)
= Ax(t) + Bu(t) (15.6)
dt
y(t) = Cx(t) (15.7)

where the matrices are defined as:

0
   
−β/J K/J −1/J
A= B= C = I2 (15.8)
−K/La −Ra /La 0 −1/La

where I 2 is a two-dimensional identity matrix. To simplify the exercise, zero

load torque is assumed, i.e., ml (t) = 0, which implies u(t) = [va (t) 0]T .
Therefore, the input term can be simply interpreted as u(t) = va (t). It is
important to note that, under this assumption, even matrix B has to be

2 Bold symbols refer to both vectors of elements or matrices.

Linear Model of a PMDC Motor 225

ml (t)
−
va (t) 1 ia (t) me (t) 1
(t)
K
+ Ra + sLa +  + sJ
−
e(t)
K

PMDC motor model

Figure 15.3 Block scheme representation of a PMDC motor.

updated. Indeed, the state-space representation in matrix notation becomes:

d
      
Ω(t) −β/J K/J Ω(t) −1/J
= + va (t) (15.9)
dt ia (t) −K/La −Ra /La ia (t) 0
1 0
    
Ω(t) Ω(t)
= (15.10)
ia (t) 0 1 ia (t)
The equations reported in (15.2)–(15.5) can be also visualized by means of
a block scheme (Figure 15.3), in which the Laplace operator substitutes the
derivative one s ← d/dt. Therefore, based on Figure 15.3, the the continuous-
time linear system is constituted by two transfer functions defined as follows:
ia (t) 1 Ω(t) 1
Gi (s) = = GΩ (s) = = (15.11)
va (t) Ra + sLa me (t) β + sJ
where Gi (s) and GΩ (s) refer to the electrical and mechanical contributions
respectively. These latter are linked through the back-emf e (t) (acting as a
physical feedback) and me (t), i.e., denoting an electromechanical system. Due
to the discrete nature of the MCU, both the state-space representation and
transfer functions should be mapped into discrete-time domain with k and
z notation, that is, in terms of Gi (z) and GΩ (z). The reader is referred to
Chapters 5 and 6 for further details on this mapping. Apart from the model-
ing description, it is important to understand the operating principle of the
system. Both using the Half- or Full-Bridge configuration for the converter
stage, a modulator is necessary anyway (see Figure 15.4). The modulation
signal can be related to the rotations of a potentiometer, like in this Chapter.
However, for the considered application, the modulating signal is a (variable)
dc quantity, not sinusoidal, to be compared with a triangular carrier. There-
fore, constant modulating signal means constant duty cycle value. Hence, by
manually manipulating d(t) → d(k) it is possible to increase the average volt-
age at the PMDC motor terminals va (t), thus, the current flow ia (k). This
latter directly impacts on the torque value me (k) and on the speed Ω(k) of
the shaft as well.
 226 Open Loop Control of a Permanent Magnet DC Motor

2L-VSC
=
VDC PMDC
= motor

S(k)
d (k)
modulator

Figure 15.4 Operating principle of the open-loop control scheme in discrete-

time domain (d(k)) including the converter stage.

15.3 System Simulations

In this book, system simulations are used to analyze and design the considered
control systems. The simulation of the open-loop control of the PMDC motor
is carried out with MATLAB® and Simulink® . This is done for both converter
configurations. To this purpose, the latter and the PMDC motor are modeled
through Simscape™ Electrical elements, in particular referring to Specialized
Power Systems™ library (Simscape/Electrical/Specialized Power Systems/ ) as
shown in Section 6.1.
The Mosfet blocks (available in the subset /Fundamental Blocks/Power
Electronics) are used to represent the switches mounted into the BOOSTXL-
DRV8301 board. Each Mosfet block uses a macro model3 made by an ideal
switch controlled through an input port (i.e., a logical gate signal g(t)) and a
diode connected in parallel so that it turns on when the MOSFET is reverse
biased (vds (t) < 0). The MOSFET behavior follows the logic reported here
below:
• g(t) > 0 → ON, with the drain-source voltage vds (t) positive or negative;
• g(t) = 0 → OFF.
Referring to Chapter 13, g(t) is equivalent to S(t). In addition, the Mosfet
block allows to include more physical parameters such as the internal resis-
tance and inductance for the switch and the diode (Ron , Lon and Rd , Ld ,
respectively) as well as Rs and Cs forming the snubber circuit. The internal
diode forward voltage Vf can be inserted too. These values can be found and
computed starting from [14].
The two-level half-bridge and full-bridge configurations are obtained by
arranging these blocks. The switches are driven by a PWM modulator having
a carrier frequency of fc = fsw = 20 kHz built from standard blocks already
presented (e.g., Relational operator, Repeating Sequence, and so on).
3 This model does not take into account neither the geometry of the device or its complex

physics.
Half-Bridge Configuration 227

The PMDC motor is modeled through the DC Machine block which

can implement both a wound-field and permanent magnet DC machine by
selecting the motor model in the Field type frame. The armature circuit (A+,
A-) consists of an inductor La and resistor Ra connected in series with the
back-emf. In the permanent magnet DC machine, there is no field current
creating the excitation flux, which is established by the magnets. Thus, ke
and kT are constants as expected. It is important to underline that Torque
constant has to be selected in the Specify frame. The DC Machine block
already include a lumped (first-order) mechanical model based on the inertia
J, the viscous friction coefficient β (or Bm ), and, as a further option, the
Coulomb friction torque (which is not used in this book, i.e., = 0). The load
torque ml (t) can be provided as an external input by selecting Torque TL in
the Mechanical input frame. By selecting −1 as sample time, this block uses
the largest integration step-size allowed by the Powergui block. The latter is
essential to manage the numerical integration of the Simscape™ elements, as
explained in Section 6.1.2.

15.4 Half-Bridge Configuration

A Half-Bridge configuration foresees the use of just a single-leg of the adopted
converter, for which it is enough to use the module ePWM1 only. Namely,
the switch 1A is the master and 1B the follower (dual). The terminals of the
PMDC motor are connected between the central point of leg A and ground.
The average armature voltage is function of the duty cycle d(k) (or dA (k)) is
therefore:
Va (k) = va,avg (k) = d(k)VDC (15.12)
where the variables are already considered within discrete-time steps k, i.e.,
in view of the implementation. The voltage VDC is set by the power supply
(i.e., the voltage on the DC bus) while d (k) ∈ [0, 1]. As a consequence, it
is possible to feed the electrical machine with positive voltages only, since
va (k) ∈ [0, VDC ], which is reflected in Va (k). However, the motor can spin
in clockwise or counter-clockwise direction according to the physical connec-
tion of its terminals + and −, e.g., VDC connected to + and GND to −, or
vice-versa. So, its rotation can not be reversed once the connection is set. The
BOOSTXL-DRV8301 board adopt a two-level topology for each leg, thus, Fig-
ure 15.5 shows the equivalent circuit for the Half-Bridge topology connected
to a PMDC motor. The PMDC motor is characterized by Ra = 0.5290 Ω,
La = 0.8651 mH, K = 0.0232 Nm/A, β = 0.1803 mN m s and J = 3.559 µg m2 .
The supply voltage is VDC ≈ 10 V and the switching frequency is set equal to
fsw = 20 kHz. These parameters can be initialized in Model Properites/Call-
backs/InitFcn or in a separate m-file as follows:
228 Open Loop Control of a Permanent Magnet DC Motor

1A ia (t) La Ra
VDC +
−

va (t) e (t)
1B

Figure 15.5 Equivalent circuit of a PMDC motor driven by a Half-Bridge

topology converter.

%% power supply
Vdc = 10;
%% carrier frequency definition
fsw = 20e3;
Tsw = 1/fsw;
%% sampling time definition
Ts = 50e-6;
%% step-size definition
Tsim = Tsw/400;
%% motor parameters
Ra = 0.5290;
La = 8.6507e-4;
K = 0.0232;
B = 1.8026e-4;
J = 3.559e-6;

The equivalent circuit shown in Figure 15.5 is then realized in Simulink® as

reported in Figure 15.6. This scheme is included in the file:

HBsimDC.slx (solver: fixed step - ODE4, step size: Tsim )

HBsimDC.slx (powergui: simulation type Continuous)

The simulation time step Tsim is small enough to approximate the switching
behavior and high frequency effects that occurs in the scheme well, i.e., get-
ting the simulation closer to the practice. The reference subsystem includes
several kind of inputs to operate the system differently. Focusing on the PWM
stage, the duty cycle d(k) is compared to a triangular carrier at fc = 20 kHz,
generating a switching pattern S(k). This is done in the MCU by using the
ePWM modules. The gate signals of the 2L-VSC topology are S(k) = S1 (k)
for 1A and S2 (k) = not(S1 (k)) for 1B , which are applied to the switch through
the gate drivers. Therefore, the operation of these switches together with the
given switching pattern creates a pulsed output voltage waveform, i.e., the
Half-Bridge Configuration 229

Continuous
s1A

D
PMDC w
1A i
modA
TL m

S
modA
reference dc
m
carA s1A
A+ A-
s1B
s1B
carA

D
+
1B v va
-

S
Figure 15.6 Simulink® scheme to simulate the open-loop dynamics of the
PMDC motor driven in half-bridge configuration included in HBsimDC.slx.

armature voltage va (k). This is the reason why the average value of this latter
va,avg (k) is considered in Equation (15.12). Indeed the actual voltage va (k) is
computed as
va (k) = S(k)VDC (15.13)
which is discontinuous and bounded within two voltage levels, i.e., [0, VDC ].
Considering the ramp waveform ranging from 0 to 1 as reference signal, d(k)
reaches 1 in 0.3 ms after which such value is hold. The effects of the reference
increase are reflected into the switching pattern of 1A and 1B , which is aimed
to make the average armature voltage Va (k) increasing by enlarging the pulse
width of va (k). This is evident in Figure 15.7. This latter also shows that the
speed Ω(k) is increasing almost quadratically. In can be noted that the same
occurs for the current ia (k) and the torque me (k), but the switching effects
are more evident (see the dashed line interval reported in Figure 15.7). Indeed,
the inertia J limits these abrupt variations on Ω(k).
A longer simulation with trapezoidal duty cycle d(k) is then considered, as
shown in Figure 15.8. The carrier waveform is dropped out to improve the plot
readability. A longer simulation allows to analyze the steady-state interval for
each of the considered quantities. In particular, Figure 15.8 shows how fast
is the current ia (k) reaction to the voltage increase. Its dynamic is related to
La . When va,avg (k) reaches the first steady-state point, the current ia (k) is
at its peak value and, after that, it starts decreasing due to the presence of
e(k) (the motor is spinning) and Ra . The speed Ω(k) requires, as expected,
a longer time to reach the steady-state due to the mechanical inertia J. It is
fundamental to remind that the considered system is running in open-loop.
Thus, no forcing action to adjust the dynamics of the motor is applied.
230 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.7 Open-loop dynamics of the PMDC motor with a ramp reference
signal and Half-Bridge converter.
Half-Bridge Configuration 231
12

10

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

300

200

100

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

0.12
0.1
0.08
0.06
0.04
0.02
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

Figure 15.8 Analysis of steady-state and transient intervals for the open-loop
dynamics of the PMDC motor.

15.4.1 Control implementation

The open-loop dynamics are tested through a test-bench which includes a
TI BOOSTXL-DRV8301 converter mounted on the TI LaunchXL F28069M
Piccolo™ and connected to a PMDC motor, as shown in Figure 15.9. By
managing the cable connections from the BOOSTXL-DRV8301 to the motor
terminals, both Half-Bridge and Full-Bridge configurations can be realized.
Regarding the Half-Bridge configuration, only leg (phase) A is used, as shown
in the connection scheme reported in Figure 15.10. A step-by-step procedure
for programming the microcontroller for this task is reported here in the fol-
lowing.
 232 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.9 Test-bench including a TI BOOSTXL-DRV8301 converter

mounted on a LaunchXL F28069M Piccolo™ board and a PMDC motor (see
the possible connections in Figure 15.10 and 15.23). This setup includes the
extPot3 board connected to the LaunchXL F28069M Piccolo™ . The reader is
referenced to Appendix B for further details.

Firmware Environment
c28069_openDC_hbF.slx (solver: fixed step - ODE4, step size: Ts = 50 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:

• Open the Model Configuration Parameters window and use the same
settings adopted for the Example 1 in Section 10.5.
• Add a GPIO block. Remember that the BOOSTXL-DRV8301 is enabled
by settling GPIO50 high (see Section 3.2). Namely, the GPIO peripheral
is driven by a Constant block with value 1 and sample time Ts .
• Insert an ADC block to read the potentiometer mounted on the extPot3
custom board; double click on it:
– Select ADCINB5 as Conversion channel in the Input Channels
tab. The, go back to the SOC Trigger tab:
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC0;
– SOCx acqusition window: 7;
– SOCx trigger source: CPU0_TINT0n or Software;
– ADC will trigger SOCx: No ADCINT;
– Sample time: TADC = 0.001 s;
– Data type: uint16;
 Half-Bridge Configuration 233

GND

rotating

legA

ADCINB5

DC link

Figure 15.10 Connections of the PMDC motor with the BOOSTXL-DRV8301

converter realizing a Half-Bridge configuration.

– Flag Post interrupt at EOC trigger (optional);

– Interrupt selection: ADCINT1 (optional);
– Flag ADCINT1 continous mode (optional).
• Add a Data Type Conversion block set to int16 (since positive signal
only are used) or single to exploit 32 bit.

• Place a Gain block for scaling the digitalized signal within [0, 1]. This is
fundamental because a 12-bit reading returns a value in a range from 0
(0 V) up to 212 − 1 = 4095 (3.3 V). Hence, set the gain as 1/(212 − 1) or, to
be more conservative, as 1/212 , even if it, in this last case, the representation
range is not fully exploited.4
• Insert an ePWM block and use the same settings reported in Section 13.5:
– Select ePWM1 module with TBPRD related to fc = 20 kHz, i.e.,
PWMcounter−period = 2250;
– Remember to unflag Enable ePWM1B and tick Inverted version of
ePWMxA. This option allows to drive 1B as the complement of 1A ;
– Keep all the other settings as reported in Section 13.5.
• Add another gain block to normalize the modulation signal over the peak
value of the carrier signal PWMcounter−period . Include a data type conver-
sion set to int16 to represent the duty cycle within a 16-bit base.

4 Alternatively, this operation can be carried out by using an arithmetic shift of 12 bit,

which is the preferable choice in microcontrollers. Indeed, this kind of operation is quicker
than multiplications or divisions with this hardware. In particular, the right shift means
division (see Section 7.3).
234 Open Loop Control of a Permanent Magnet DC Motor
DRV enable
C2806x

GPIOx

potentiometer B5 GPIO DO
C2802x/03x/05x/06x
C2802x/03x/05x/06x
B5
WA
ADC
ePWM
A

pulses

Figure 15.11 Simulink® scheme included in c28069_openDC_hbF.slx.

The aforementioned blocks should be connected as shown in Figure 15.11.

Power Settings: Based on the DC motor datasheet, the power supply should
be manually set to provide VDC = 10 V and to saturate its output current to
2 A.
Once the firmware is downloaded into the MCU, the resulting behavior is
the motor spinning at a speed which can be adjusted through the potentiome-
ter. As a matter of fact, the PWM nature of va allows to adjust the average
voltage (within the PWM period) applied on the terminals of the motor by ro-
tating the potentiometer. When d(k) ≈ 1 the PMDC is fed with the full VDC .
Thus, it rotates at maximum speed. In principle, to further increase the mo-
tor speed, the voltage provided by the power supply can be, i.e. VDC > 10 V.
However, since the motor is running in open-loop, no protection or limitation
are included into the control platform, thus, potentially allowing to exceed the
rated value of the electrical machine.

15.5 Full-Bridge Configuration

A Full-Bridge configuration (or H-bridge) refers to use two legs of the adopted
boosterpack (legs A and B), for which two ePWM modules are required, e.g.,
ePWM1 and ePWM2. The terminals of the PMDC motor are connected to
the central point of legs A and B. Hence, the switches 1A , 1B and 2A , 2B
are considered. The equivalent circuit of this test case is reported in Figure
15.12. A benefit of a H-bridge configuration with respect to the Half-Bridge
one is the chance to turn the motor in both directions (clockwise or counter-
clockwise) without changing the electrical connections of the setup. The sys-
tem is simulated considering the parameters reported in Section 15.4, which
can be initialized in Model Properites/Callbacks/InitFcn or in a separate m-
file. The equivalent circuit shown in Figure 15.12 is then realized in Simulink®
as reported in Figure 15.13. This scheme is included in the file:

FBsimDC.slx (solver: fixed step - ODE4, step size: Tsim )

Full-Bridge Configuration 235

1A 2A

e (t)

+ ia (t) La Ra
VDC
−

va (t)

1B 2B

Figure 15.12 Equivalent circuit of a PMDC motor driven by a Full-Bridge

topology converter.

FBsimDC.slx (powergui: simulation type Continuous)

Similarly to Section 15.4, the simulation time step Tsim is set small. Moreover,
the reference subsystem includes different kind of inputs to operate the system
differently. A unipolar or bipolar voltage switching approaches can be selected
by acting on the manual switch, which sets the carrier phase-shift (carB)
processed by the modulator stage related to leg B.
Moreover, the actuation of the two legs (i.e., two duty cycles for two mod-
ulation stages) is done by manipulating one modulation signal d(k) only, from
which the gating signals of the four switches are generated. Such switching
pattern derivations (that is, how to compute the duty cycles) can be done by
using different strategies. Therefore, before going into the implementation de-
tails, the following sections provide the descriptions of the different converter
operations.

15.5.1 Modulation strategies

The legs A and B are driven by two different duty cycles dA (k) and dB (k).
The voltages measured from the central point of each leg and ground are:

vA (k) = dA (k)VDC (15.14)

vB (k) = dB (k)VDC (15.15)

The armature voltage va (k) results from the following difference:

va (k) = vA (k) − vB (k) = (dA (k) − dB (k)) VDC (15.16)

The two legs should be synchronously operated to control the current flow
iA (k) of the motor and regulate its rotation as well. Thus, dA (k) and dB (k)
 236 Open Loop Control of a Permanent Magnet DC Motor

Continuous
s1A s2A
modA

modB

D
reference w
s1A 1A PMDC 2A
carA i
s1B

S
TL m
carA
modB m
dc
modB
s1B A+ A- s2B
carB s2A

D
s2B

+
-
unipolar carB 1B 2B
carB

v
S

S
bipolar va
carB

Figure 15.13 Simulink® scheme included in FBsimDC.slx.

have to be linked together even it their modulation stages may be separated.

This need combined with the duty cycle manipulation, leads to define two
possible modulation strategies which are called bipolar and unipolar voltage
switching.

15.5.2 Unipolar voltage switching

In this case, va (k) is continuously switched from +VDC down to −VDC , and
vice-versa. This modulation strategy is obtained by imposing the following
condition on the duty cycles:

dB (k) = 1 − dA (k) (15.17)

Thus, by defining d(k) = dA (k), the armature voltage becomes function of

one duty-cycle only:
va (k) = (2dA (k) − 1) VDC (15.18)
The unipolar voltage switching behavior is achieved by using two modulator
stages, which keep the same frequency and phase shift for both the triangular
carriers (i.e., same settings inside ePWM1 and ePWM2). Therefore, S1 (k),
S2 (k) are considered related to 1A , 1B of leg A, respectively, whereas S3 (k),
S4 (k) to 2A , 2B of leg B, respectively. The switching pattern results from the
application of the PWM logic. It can be observed that when S1 (k), S3 (k) or
S2 (k), S4 (k) are turned on, the armature voltage va (k) is zero. The voltage
level of this latter jumps between 0 and −VDC or from 0 and +VDC . Given the
equivalence between modulating signal and duty cycle, only the computation
of dA (k) is required. Indeed:
• If dA (k) = 0, then dB (k) = 1
which imply vA (k) = 0, vB (k) = VDC and va (k) = −VDC ;
• If 0 < dA (k) < 0.5, then dB (k) > dA (k)
which imply vB (k) > vA (k) and va (k) < 0;
Full-Bridge Configuration 237

Figure 15.14 Example of unipolar voltage switching operation: associated

waveforms and dynamics of the system.
 238 Open Loop Control of a Permanent Magnet DC Motor
10

-5

-10
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10-4

10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10-4

10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10-4

Figure 15.15 Leg voltages during a unipolar voltage switching operation.

• If dA (k) = 0.5, then dA (k) = dB (k) = 0.5

which imply vA (k) = vB (k) = VDC /2 and va (k) = 0;
• If 0.5 < dA (k) < 1, then dA (k) > dB (k)
which imply vA (k) > vB (k) and va (k) > 0;
• If dA (k) = 1, then dB (k) = 0
which imply vA (k) = VDC , vB (k) = 0 and va = VDC .
Every time the potentiometer is in its centered position, it results dA (k) = 0.5.
Thus, the motor is standstill. The motor speed is assumed positive in clockwise
direction. Therefore, depending on the terminal connections realized in the
setup (see Figure 15.10), when dA (k) < 0.5 the motor speed is negative (i.e.,
counter-clockwise rotations), while it is positive for dA (k) > 0.5.
Figure 15.14 shows an example of unipolar voltage switching operations.
In particular, a positive ramp profile is applied to d(k) = dA (k) which corre-
sponds to a negative ramp profile for dB (k). At the beginning, d(k) = 0 →
dA (k) = 0, dB (k) = 1 leading to vA (k) = −VDC . Hence, the PMDC motor
starts rotating in counter-clockwise direction (Ω(k) < 0) until d(k) > 0.5. The
current ia (k) shows a similar behavior but with a different (faster) dynamic.
It can be noted that, even if va (k) can be negative, the output voltages vA (k)
and vB (k) are always positive on each leg as shown in Figure 15.15. Con-
sidering a longer simulation and a trapezoidal duty cycle d(k) (similarly to
what is reported in Section 15.4), the corresponding results are those shown
Full-Bridge Configuration 239

Figure 15.16 Analysis of a unipolar operation during a long simulation.

in Figure 15.16 (where the carrier waveforms are dropped out to improve the
plot readability). In particular, the benefits in terms of lower current ripple in
ia (k) is evident.
The unipolar voltage switching reduces the magnitude of the harmonic
components of va (k) and ia (k), which means lower ripple and harmonic dis-
tortion, while keeping the converter switching frequency fsw constant. This
goal is achieved by displacing vA (k) and vB (k), so that the effective switch-
ing frequency in va (k) is 2fsw and, consequently, in ia (k) as well. In standard
 240 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.17 Computation of the RMS voltage during an unipolar operation.

cases (i.e., without a Full-Bridge configuration), such voltage/current ripple

reduction might be achieved by actually operating the converter at twice the
switching frequency. Therefore, this operation has the benefit of doubling the
apparent fsw seen by the PMDC motor (i.e., the load stage), without increas-
ing the switching losses. This result is also a consequence of the introduction of
a zero voltage level, which allows to have shorter jumps while reaching +VDC
or −VDC (see Figure 15.15). Another important aspect regards the computa-
tion of the root mean square (RMS) value (also known as the quadratic mean)
of the armature voltage. The RMS definition is the square root of an integral
over the signal period, as shown in Figure 15.17. Namely, it is defined as:
v
u ZTs
u
u1
va,RMS = t va2 (t) dt (15.19)
Ts
0

Since the resulting va (k) is a pulse function, it can be equivalently represented

in continuous-time domain by va (t). Inside one sampling period Ts (where the
on time is defined as t1 , see Figure 15.17), the pulse function is constant and
equal to VDC for 0 ≤ t < t1 , while zero for t1 ≤ t < Ts . By replacing such
values in (15.19), it follows:
v
u Zt1
u1
r
t1
u
va,RMS = t VDC dt = VDC
2 (15.20)
Ts Ts
0

Therefore, the RMS value of va (k) is varying in time, ranging between 0 and
VDC , as shown in Figures 15.15 and 15.16.
In conclusion, this solution has the following advantages and disadvantages:

Advantages
• Better harmonic content compared both to the Half-Bridge and Full-Bridge
bipolar voltage switching configurations;
• One control signal is needed leading to three voltage levels (−VDC , 0, +VDC ).
 Full-Bridge Configuration 241

Disadvantages
• The RMS value of the armature voltage varies in time.

15.5.3 Bipolar voltage switching

The duty cycles relationship still holds

dB (k) = 1 − dA (k) (15.21)

va (k) = (2dA (k) − 1) VDC (15.22)

Even in this case, one modulating signal is still enough to fully operate the
system. Thus, the relationship between d(k) = dA (k) (i.e., potentiometer ro-
tation) and the motor speed is the same as for the unipolar voltage switching.
The main peculiarity of this solution, compared with the previous case, is
that in a bipolar voltage switching approach the output of leg A, vA (k), is
equal and opposite to the output of leg B, vB (k). This condition is carried out
by complementing the behavior of ePWM1 and ePWM2 modules. Given the
equivalent circuit reported in Figure 15.12, the diagonally opposite switches
1A , 2B and 1B , 2A are turned on or off at the same time. This logic leads to
an armature voltage va switching between two levels: −VDC and +VDC , which
is the reason beyond the name bipolar.
In terms of implementation, only one ePWM module may be used to realize
such approach theoretically speaking. For instance, if ePWM1 is settled to
provide S1 (k) and S2 (k) (i.e., driving 1A , 1B ) at pins 40 and 39 respectively,
it would be enough to physically connect S3 (k) to pin 39 and S4 (k) to pin 40
to control leg B.
On the other hand, the proposed implementation foresees to use two mod-
ulator stages (ePWM1/2). The bipolar voltage switching is achieved by using
modulators operating with different triangular carriers. The latter would be
in opposition (counter-phase) to each other. For instances, if the phase shift of
the ePWM1 carrier is set equal to 0◦ (or 90◦ ), then the one of ePWM2 should
be set equal to 180◦ (or −90◦ ), which is equivalent to have an inverted/mir-
rored carrier. Since the carrier is internally created by the ePWM blocks, the
phase shift is introduced through a specific parameter setting.
Figure 15.18 shows an example of bipolar voltage switching operation.
In particular, the same positive ramp profile adopted for the unipolar case
is applied to d(k) = dA (k) and complemented for dB (k). At the beginning,
d(k) = 0 → dA (k) = 0, dB (k) = 1 leading to vA (k) = −VDC . Nevertheless,
at the following time steps, Figure 15.18 shows how the zero voltage level is
absent from va (k), leading to a switching between −VDC and VDC to regulate
the average voltage. The resulting behavior of the output voltages vA (k) and
vB (k) (which are always positive) are shown in Figure 15.19.
Comparing Figure 15.14 and 15.18, it is evident that the bipolar voltage
switching returns an higher current and voltage harmonic distortion with re-
spect to the unipolar one. The latter benefit of the cancellation of the all the
242 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.18 Example of bipolar voltage switching operation: associated wave-

forms and dynamics of the system.
Full-Bridge Configuration 243
10

-5

-10
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
10-4

10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10-4

10

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

10-4

Figure 15.19 Leg voltages during a bipolar voltage switching operation.

odd harmonics of fsw for va (k) and, consequently, ia (k), which is not the case
for the bipolar case. This can be proved and clarified by Figure 15.20, which
compares the harmonic spectra of the armature voltages va (k) when a fixed

Figure 15.20 Spectra (truncated at the 5th harmonic) of the armature voltages
obtained in unipolar and bipolar voltage switching operator. In both cases,
the converter operates at the same fsw .
244 Open Loop Control of a Permanent Magnet DC Motor

Figure 15.21 Analysis of a bipolar operation during a long simulation.

duty cycle d(k) = 0.2 (i.e., in steady-state conditions) is applied for the unipo-
lar and bipolar operations, respectively. The spectra are computed through a
discrete fast Fourier transform (DFFT) considering 800 periods. The converter
legs are operated at the same fsw in both cases. This comparison shows how
the unipolar approach is able to operate the system at an effective switching
frequency of 2fsw , leading to a lower current distortion.
Similarly to Section 15.5.2, a long simulation and a trapezoidal duty cy-
cle d(k) are considered for the bipolar operation. The results are shown in
 Full-Bridge Configuration 245

Figure 15.22 Computation of the RMS voltage during a bipolar operation.

Figure 15.21, where the carrier waveforms are dropped out to improve the
plot readability. A larger ripple on ia (k) is particularly evident in comparison
to Figure 15.16.
Regarding the RMS value of the armature voltage, va (k) is now a bipolar
pulse function. Inside one sampling period Ts (where the on time is defined
as t1 , see Figure 15.22), va (t) = VDC for 0 ≤ t < t1 , while va (t) = −VDC
for t1 ≤ t < Ts . By replacing such values in the general equation (15.19), it
follows:
v 
Zt1 ZTs
u 
u1 
r r
t1 t1
u
2
va,RMS = t VDC + (−VDC ) dt = VDC
2  + VDC 1 − = VDC
Ts Ts Ts
0 t1
(15.23)
Therefore, the RMS value of va (k) in case of bipolar voltage switching is always
constant and equal to VDC independently from the duty cycle, as shown in
Figure 15.19 and 15.16. As an example, it can be assumed to include a lighting
device in the drive system. Light emissions are sensitive to the RMS value of
the supply voltage (this is true for LEDs, incandescent and fluorescent lights).
By connecting a lighting device in parallel to the PMDC motor terminal, the
motor control is unaffected and this device is continuously emitting light. Car
mirrors typically use this idea to regulate the position of the mirrors through
small motors, while turning on/off the integrated light indicators by directly
connecting, e.g., LEDs, to the motor terminals.5 In conclusion, the bipolar
solution shows the following advantages and disadvantages:

Advantages
• One modulation signal needed leading to two voltage levels (+VDC , −VDC );
• RMS value of the armature voltage is constant.

Disadvantages
• Higher voltage and current harmonic distortion compared to unipolar case.

5 This approach is used to reduce the number of elements aimed to the electronic circuit.
 246 Open Loop Control of a Permanent Magnet DC Motor

legB

rotating

legA

ADCINB5

DC link

Figure 15.23 Connections of the PMDC motor with the BOOSTXL-DRV8301

converter realizing a Full-Bridge configuration.

15.5.4 Control implementation

The open-loop dynamics are tested through a test-bench which includes a
TI BOOSTXL-DRV8301 converter mounted on the TI LaunchXL F28069M
Piccolo™ and connected to a PMDC motor, as shown in Figure 15.9 Regard-
ing the Full-Bridge configuration, legs A and B are used for both unipolar
and bipolar voltage switching. The cable connections from the BOOSTXL-
DRV8301 to the motor terminals are shown by the connection scheme reported
in Figure 15.23. In the following, the step-by-step procedures for programming
the microcontroller for both modulation approaches are reported.

Firmware Environment (Unipolar Approach)

c28069_openDC_fbF_uni (solver: fixed step—ODE4, step size: Ts = 50 µs)

Open a new blank Simulink® project and configure the environment as shown
in Chapter 9. Then:
• Open the Model Configuration Parameters window and use the same
settings adopted for the Example 1 in Section 10.5.
• Insert an ADC block to read the potentiometer and set it as shown in
the previous Section 15.4.1. Indeed, the potentiometer reading is the same.
Hence, add Data Type Conversion blocks to manipulate the signal (i.e.,
int16 and single). Place a Gain block set to 1/212 to scale the digitalized
signal within [0, 1].
• Insert two ePWM blocks; use the settings reported in Section 13.5:
Full-Bridge Configuration 247
B
C2802x/03x/05x/06x
db
WA

ePWM
potentiometer B5
A
C2802x/03x/05x/06x
C2802x/03x/05x/06x
B5 d da
WA
ADC
ePWM

pulses

Figure 15.24 Simulink® scheme included in c28069_openDC_fbF_uni.

– ePWM1 module is used with TBPRD corresponding to fc = 20 kHz,

i.e., PWMcounter−period = 2250;
– Remember to unflag Enable ePWM1B and tick Inverted version of
ePWMxA. This option allows to drive 1B as the complement of 1A ;
– Keep all the other settings as reported in Section 13.5;
– Set ePWM2 analogously.
Since the same settings are used for both ePWM blocks (i.e., same triangular
carrier and comparison instances), the modulating signals dA (k) and dB (k)
are built as follow:
– Add a constant and an add block to bias the input signal. Set the
value of the constant to 1 and the list of signs of the add block to +−.
Thus, the modulating signal which is fed into ePWM1 is dA (k) while
dB (k) is obtained as dB (k) = 1 − dA (k) and fed into ePWM2.
– Add gain blocks with value PWMcounter−period = 2250 to both path
for the ePWM modules to normalize the modulation signals.
∗ dbA (k) = dA (k) · PWMcounter_period
∗ dbB (k) = (1 − dA (k)) · PWMcounter_period
– Add other two data type conversion blocks (one for each ePWM
block) set to int16 to represent the duty cycle within a 16-bit base.
The aforementioned blocks should be connected as shown in Figure 15.24.
Even in this case, it is recommended to set the power supply at VDC = 10 V
with a current limitation of 2 A. After the firmware deploy on the MCU, it is
possible to manually vary the motor speed through the potentiometer as done
previously. Since the system is running in open loop, the user is recommended
to verify the maximum voltage and current which is fed into the system to
avoid any damage or premature aging of the components.

Generally, it is preferable not to fully apply d(k) = 0 (0%) or d(k) = 1 (100%)

because they could be potential causes of leg short-circuits.
 248 Open Loop Control of a Permanent Magnet DC Motor

Remark: the BOOSTXL-DRV8301 converter operates with a bootstrap cir-

cuit, which requires d(k) < 1 to keep the converter switching safely. Consid-
ering leg 1, the drain pin of 1A is connected to the positive terminal of the of
dc link while switching. It may happen that the voltage applied on the switch
is higher than the threshold needed to close the switch, i.e., the gate-source
voltage Vgs (e.g., Vgs ≈ 15 V with Vdc ≈ 48 V). Therefore, the gate circuit
is using a diode to charge a small capacitor Cgd placed between gate and
drain pins, designed to operate 1A for a given value of Vgs . In practice, when
d(k) = 1, only the switch 1A is closed. However, keeping d(k) = 1, Cgd is
slowly discharging in time until the energy is not enough to keep the switch
closed (i.e., voltage becomes lower than Vgs ). Hence, it is necessary to open
it and close 1B , which means d(k) < 1. Thus, some small interval is added to
recharge Cgd . It is important to note that for low voltage applications (as in
this case, being VSC = 10 V) it is enough to add a narrow margin on the duty
cycle, e.g., d(k) ∈ [0.01, 0.99].

Firmware Environment (Bipolar Approach)

c28069_openDC_fbF_bip (solver: fixed step - ODE4, step size: Ts = 50 µs)

Take the previous Simulink® project about the unipolar voltage switching
There, the PWM actuation must be modified. In particular, the ePWM pe-
ripherals of the TI LaunchXL F28069M Piccolo™ allow swap features be-
tween ePWMxA and ePWMxB. By selecting swap the ePWMA and eP-
WMB outputs, the computed ePWMxA output signal is redirect/provided
at the ePWMxB output terminal, and vice-versa. For the bipolar approach,
this feature allows to simplify the PWM actuation since the duty cycles
are set dA (k) = dB (k). Two modulation stages are considered, ePWM1
(which provides S1 (k), S2 (k)) and ePWM2 (which provides S3 (k), S4 (k)),
both fed with the same modulating signal. If ePWM2 is swap-enabled, then
ePWM2A=S3 (k) and ePWM2B=S3 (k). Thus, if ePWM1 has standard set-
tings, it follows that S1 (k) = S4 (k) and S2 (k) = S3 (k). This latter correspond
to the bipolar voltage switching principle and it is exactly equivalent to make
ePWM2 operating with a triangular carrier shifted by 180° with respect to
ePWM1 and having dB (k) = (1 − dA (k)).
Therefore, the previous scheme has to be updated as follows:
• Remove the dB (k) derivation, keeping one duty cycle only, i.e., dA (k) =
dB (k), by deleting the constant and the add block.
• Use the same settings adopted in the unipolar case for module ePWM1:
TBPRD corresponds to fc = 20 kHz, i.e., PWMcounterperiod = 2250.
• Keep the same settings for ePWM2 apart from the following:
– Thick enable swap module A and B at the bottom of the General tab.
• Keep the rest of the scheme set as before.
Full-Bridge Configuration 249
B
C2802x/03x/05x/06x

WA

ePWM
potentiometer B5
da=db A
C2802x/03x/05x/06x
C2802x/03x/05x/06x
B5 d
WA
ADC
ePWM

pulses

Figure 15.25 Simulink® scheme included in c28069_openDC_fbF_bip.

Finally, connect all these blocks as reported in Figure 15.25. It is recommended

to set the power supply at VDC = 10 V with a current limitation of 2 A.
After the deploy the firmware on the MCU board, it is possible to manually
vary the motor speed in absolute value and sign according to the position
of the potentiometer. Since the system is running in open loop, the user is
recommended to verify the maximum voltage and current which is fed into
the system to avoid any damage or premature aging of the components.
16
Low-Side Shunt Current Sensing

In order to close a control loop with the LaunchPad™ F28069M board, dif-
ferent measurements have to be managed and processed. Current, voltage
and speed sensors (e.g., encoders or hall effect sensors) are widely adopted
in electrical drive applications. These devices are usually placed at the power
converter terminals or mechanically coupled with the load. The resulting mea-
surements are fed into the MCU through ADC or GPIO channels. This section
focuses on how to manage current measurements.
For instance, the TI™ BOOSTXL-DRV8301 converter has both voltage
and current sensors on-board. In particular, this board includes a low-side
shunt current sensor, which is a resistor placed below the low side MOSFET
(e.g., 1B ) of each leg. A differential operational amplifier (op-amp) takes the
differential voltage across the resistor as input, which varies depending to the
current flow, to compute a filtered voltage value scaled for the peripheral in-
puts (i.e., ranging in [0, 3.3] V). The corresponding equivalent circuit is shown
in Figure 16.1 (a). It is worth noting that shunt resistors provide an indirect
current measurements since, actually, a voltage drop across them is measured.
The main advantage of low-side sensing is the simple implementation.
Moreover, being on the low side of the load, the common-mode voltage at
the shunt is approximately zero; therefore, the robustness of the amplifiers is
not an important factor. However, this solution comes at the cost of ground
variations and it foresees a resistance between in the path to ground. Indeed,
this resistance removes the ability to detect faults in the load if it is shorted
to ground. Moreover, the op-amp circuit present several gains, which can be
activated by acting on dedicated registers to adjust its output according to
the magnitude of the measurement, as shown in Figure 16.1 (b). Hereafter, to
simplify the notation, the terms op-amp or amplification stage directly refer
to the differential structure shown in Figure 16.1 (b).
In order to design a firmware able to manage a current feedbacks, an
accurate ADC conversion chain must be built taking into account the tol-
erances of the components mounted on sensoring circuits mounted on the
BOOSTXL-DRV8301 converter as well as the ADC peripheral resolution of
the LaunchPad™ F28069M board. In practice, this means to set some gains
that allow a proper calibration of the conversion chain. To this aim, the fol-
lowing sections propose and compare two approaches: a theoretical and an
experimental calibration targeted on the available hardware.

DOI: 10.1201/9781003196938-16 251

252 Low-Side Shunt Current Sensing

leg 1

1A

dc-dc
1B
vcc

v1
amplifcation
RSH vSH
stage

(a) (b)

Figure 16.1 Equivalent circuit of (a) low-side shunt resistor and op-amp circuit
for leg 1 and a (b) detailed view of the op-amp differential structure and
selectable gains through switches S1 − S4. In this last schematic vSO = vsh
[14].

16.1 Sensor Characterization: Theoretical Approach

Referring to 16.1 (b), the op-amp circuit requires a reference voltage vref =
1.65 V to process bidirectional (positive/negative) current measurements while
keeping its output bounded between [0, 3.3] V. The theoretical characteristic
of the current sensor is reported in Figure 16.2. In general, a vref = 1.65 V
is not directly available in 3.3 V systems, but it can be derived through an
internal voltage divider. This is depicted at the bottom of Figure 16.1 (b),
where a voltage follower, i.e., voltage divider and amplification stage, is used
to improve the signal stability and noise immunity. Provided that the op-amp
circuit has several selectable gains, the default configuration refer to switch
S1 closed (see Figure 16.1 (b)). Therefore, it follows:
   
R1 R1 R1 R2 R1 vcc
vSO (k) = 1+ vSP (k) − vSN (k) + 1+ (16.1)
R12 R2 R2 R12 R2 2

where R1 = 50 kΩ, R2 = 5 kΩ and R12 = R1 + R2 , while vcc /2 = vref . This

equation can be rewritten as:

vSO (k) = vref − 10(vSN (k) − vSP (k)) = vref − 10RSH i (k) (16.2)
| {z }
RSH i(t)

The BOOSTXL-DRV8301 board mounts a RSH = 0.01 Ω shunt and it is de-

signed for a maximum current peak of ±14 A. Therefore, the voltage values
Sensor Characterization: Theoretical Approach 253

3.5

2.5

1.5

0.5

-0.5

-30 -20 -10 0 10 20 30

Figure 16.2 Theoretical characteristic of the current measurement through

low-side shunt resistor.

provided at the ADC peripheral inputs are inside a [0.25, 3.05] V range. This
means that the amplification circuit has a safety margin for current manage-
ment, since the magnitude of the current can be theoretically pushed up to
±16.5 A. This last range allows to exploit the all the possible voltage values,
that is, [0, 3.3] V. Namely:

imax : i(k) = +16.5 A → vmax : vSH (k) = 3.3 V (16.3)

imin : i(k) = −16.5 A → vmin : vSH (k) = 0 V (16.4)

It is important to note that zero current corresponds to an output voltage

of vSO = vref = 1.65 Ω, as a result of the sensor characteristic. The first
conversion gain is defined to translate the measured current i (k) into voltage
vSO , including the offset vref . Considering vSO and i(k) at their maximum
values, see equation (16.2), it follows that:

vmax − vref mV
gi = 10RSH = = 0.1 Ω = 100 (16.5)
imax A
Then, vSO is processed and scaled over the 12 bit ADC peripheral through a
quantization process, which can by accounted for by the gain:
vmax 3.3 mV
gv = = ≈ 0.81 (16.6)
212 − 1 4095 bit

The output of ADC channels is a voltage value in bit, vSO bit

, for which the
interval [0, 2047] refers to negative currents, whereas [2049, 4095] to positive
ones. The value vSObit
= 2048 means to zero current. Therefore, gi and gv can
be interpreted as the current and voltage conversion steps of the quantizied
measurement chain, respectively.
254 Low-Side Shunt Current Sensing

0 − 4096 bit
i (t) ±1.65 V
gi + vSO (t) vSO (k) ADC
bit
vSO (k)
S&H + ±2048 bit i (k)
+ gv−1 gadc
vref 0 − 3.3 V −
bit
2 vref
2
ADC peripheral

Figure 16.3 ADC conversion chain for current measurement through the
BOOSTXL-DRV8301 converter board.

If the programming environment is required to work in SI units, the bit

representation has to be re-scaled to provide values in Ampere. Therefore,
combining gi and gv , a peripheral gain gadc can be defined as follows:
gv vmax − vref vmax mA
gadc = = ≈ 8.1 (16.7)
gi imax 212 − 1 bit
The conversion chain of a current measurement from a shunt resistor to
the ADC peripheral is reported in Figure 16.3, where i(k) represent a variable
ready to be processed to close a loop. This result is achieved through the
following steps:

i(t) −→ vSH (t) −→ vSO

bit
(t) → bit
vSO (k) −→ i(k) (16.8)
gi ,+vref /2 gv S&H bit ,g
−vref adc

In particular, focusing on i(t) = 16.5 A:

i(t) = 16.5 A → vSH (t) = 3.3 V → vSO

bit
(k) = 4095 → i(k) = 16.5 A (16.9)

The same process can be used to compute all the other current values.
Remark: it is convenient to synchronize the conversion procedure of the ADC
peripheral with the ePWM module as reported in Section 13.9.

16.2 Locked Rotor Test

One of the easiest way for verifying the accuracy of the theoretical charac-
terization of the current measurement is to perform a locked rotor test (also
know as short circuit test) with a PMDC motor. This test is carried out by
mechanically locking the rotor, thus, forcing the back-emf to zero. Considering
the mathematical model presented in Chapter 15, the steady state equivalent
circuit of the PMDC motor during a locked rotor test simplifies to that one
shown in Figure 16.4. At setady state, the current IA flows in the armature
resistance only since d/dt = 0, i.e., the inductance behaves like a short cir-
cuit. The rated value of the DC motor resistance Ra can be computed as
Ra = Va /Ia , or it can be measured through an ohmmeter. It follows that
Ra ≈ 0.6 Ω.
 Locked Rotor Test 255

Ra
Ia

Va e (t) = 0

Figure 16.4 Equivalent circuit of a DC motor during a locked-rotor test at

steady state.

This test can be carried out driving the PMDC motor in open-loop with
a converter in half-bridge configuration (see the equivalent circuit in Figure
15.5). The power supply is set to VDC ≈ 10 V with current limitation at 2 A.
Differently from the exercises reported in Chapter 15, the duty cycle is now
constant and internally settled into the MCU instead of using a potentimeter.
This is aimed to have a constant and stable voltage supplied at the motor
terminal, which implies a constant current absorption as well. Hence, the duty
cycle is set to d = 0.2 so that Va ≈ 2 V at armature windings. It follows that:
Va 2V
Ia = = ≈ 3.33 A (16.10)
Ra 0.6 Ω
Since the motor size is small and the applied voltage low, the rotor can be
hand locked with the help of a clamp, as reported in Figure 16.5. A simple
test is to measure the motor current at steady-state with an ADC peripheral
and visualize it through the Serial Port (e.g., SCIA). If the conversion chain is
correctly set, using gv and gadc , the result should be close to the value reported
in equation (16.10). Regarding the MCU programming, it is suggested to start
from the firmware developed in Section 15.4.1.

Firmware Environment
c28069_lockedDC_F.slx (solver: fixed step - ODE4, step size Ts = 50 µs)

Open the example of the open-loop control of a PMDC motor (half-bridge

converter configuration) presented in Section 15.4.1. Modify the scheme as
follows:
• Edit the ADC block parameters:
– Select SOCx trigger source ePWM1_ADCSOCA;
– Set sample time: Tadc = 0.001 s;
– Select Conversion channel ADCINA0 in the Input channel tab.

• Insert a data type conversion block set to int16.

 256 Low-Side Shunt Current Sensing

Figure 16.5 Example of hand locked rotor by using a clamp

• Insert a Constant equal to 2048 (as int16) and an Add block with list of
signs +− to remove the offset at the ADC output.
• Add a data type conversion block set to double (or single) for safer
mathematical computations. See see Chapter 7 for reference.
• Insert a Gain block set on gadc = 8 mA/bit to map the ADC reading over
a range of values in Ampere.
• Add a data type conversion block set to single combined with a Rate
Transition block with sample time TSCI = 0.01 s to prepare the data ex-
change.
• Insert a Mux block to send two inputs via the SCI Transmit block.
• Insert a SCI Transmit block and set it as shown in Section 10.5 (i.e.,
SCIA).
• Place a pulse generator block and double click on it:
– Set pulse type on Sample based;
– Insert an Amplitude equal to 0.2;
– Choose a period of 250 samples;
– Type a pulse width of 125 samples;
– Insert a sample time of 0.01 s.
The duty cycle is now a train of pulses with amplitude 0.2. This block
is included to limit the steady-state condition. Indeed, even if the current
is low, being related to the voltage input (which is 2 V while the rated
value for this motor is 24 V), the temperatire of the armature windings may
rise too much. The latter may lead to significant variations in the windings
resistance, thus, affecting the measurement. Hence, the pulse generator block
prevents overheating.
 Locked Rotor Test 257
ePWM1A: J4 - pin40
C2802x/03x/05x/06x

WA

ePWM
shunt
C2802x/03x/05x/06x C28x
vSO_bit

A0 Data

i SCI XMT
ADC

offset

Figure 16.6 Simulink® scheme included in c28069_lockedDC_F.slx.

• Add a data type conversion block set to single combined with a Rate
Transition block (TSCI = 0.01 s), to prepare the output data of the pulse
generator. Connect it to the Mux block
• For the ePWM settings the reader is referred to Section 15.4.1 (fc = 20 kHz)
The arrangement of these blocks is shown in Figure 16.6. Since fc =
fsw holds, the sampling time for the current is Ts = Tsw = 50 µs. A
up-down couting mode is used with a TBPRD = PWMcounter−period =
CPUfrequency /2 · PWMfrequency = 2250. An initialization script either in Model
Properites/Callbacks/InitFcn or as a separate m-file should be edited to
set the system parameters. The latter approach is used in this exercise
(lockDC_init.m) and the code is reported here in the following:

%% F28069M clock
CPU_frequency = 90e6;
%% carrier frequency definition
PWM_frequency = 20e3;
Tsw = 1/PWM_frequency
%% TBPRD (counting mode) definition
PWM_counterperiod = CPU_frequency/(2*PWM_frequency)
%% sampling time definition
Ts = Tsw;
%% ADC theoretical gain
g_adc = 33/4095;

Testing Environment
c28069_lockedDC_T (solver: fixed step - ODE4, step size Tsim = 100 µs)

Open a blank Simulink® project and set the Model Configuration Pa-
rameters as reported in Section 10.5. Then:
• insert a Serial Configuration block and set it as reported in Section 10.5.
 258 Low-Side Shunt Current Sensing

COM10
115200 COM10 Data d
8,none,1 i
iavg

Figure 16.7 Simulink® scheme included in c28069_lockedDC_T.

• Insert a Serial Receive block and double-click on it:

– Select the right COMx as serial port;
– No header or terminator should be inserted;
– Set data size: [1 2] since two signals have to be visualized;
– Set data type: single;
– Set sample time: 0.01 s.
• Insert a Demux block to split the two transmitted signals: d(k) and i(k).
• Insert a Scope block and set the number of input ports equal to two.
• Optional: a Mean block could be added to compute the mean value of i(k).
This can be used if i(k) presents high noise or the average value is not clearly
readable from the received data. Set:
– Fundamental frequency: 10 Hz or 50 Hz;
– Initial input: 0;
– Sample time: 0.
This block receives iSCI (k). A data type conversion block is added and
set on double. The resulting variable is iavg (k).
Finally, connect these blocks as reported in Figure 16.7. Since the testing
environment is used for a debug purposes only, the simulation step-size is set
equal to Tsim = Ts /2 = 100 ms.
The serial communication (here SCIA) is able to manage multiple data
at the same time. They are ordered in columns, e.g., as [1 n] vectors for any
commmunication involving n signals. In this case, the parameters of interest
are the armature voltage va (k) (which is equivalent to d(k)) and the armature
current read by the ADC (channel A0) iSCI (k). It is up to the reader decide
to further include the computation of the average current iavg (k) with a Mean
block.
In a locked rotor test, a d(k) 6= 0 implies va (k) 6= 0, which causes a positive
current ia (k) flowing in the armature windings and a positive motor torque
as well. Note that no back-emf is acting on the machine. Thus, VDC should be
much lower than the nominal voltage rating of the electrical machine not to
exceed the rated current of the motor. Given a repetitive step-wise duty cycle
at d(k) = 0.2, it follows Va = va (k) ≈ 2 V, as shown in Figure 16.8.
Locked Rotor Test 259

Figure 16.8 Comparison between the current measured through the low-side
shunt resistor and the expected steady state values. Those current levels were
evaluated thorough a current probe.

To specify that a theoretical characterization is used in this exercise, a

subscript t is added. Figure 16.8 shows that iSCIt (k) is affected by an offset
in the zero value. Moreover, the gain gadc is slightly inaccurate, since the
steady-state value Yss is not reached (dash line) as well. These inconsistencies
are due to the physical component tolerances of the measurement circuit,
which influence the gain accuracy. Figure 16.8 also shows the effect of a wrong
clamping, which can occur since the locking is done by hand.
To include a more accurate ADC conversion chain in the firmware envi-
ronments, the gain gadc and the offset must be corrected by performing an
experimental characterization.

Remark1: by blocking the rotor, the mechanical power is entirely dissipated

as Joule losses, increasing the temperature of the whole dc motor. If the tem-
perature is too high, the armature resistance increases, which means iSCI (k)
decreasing. Considering the previous example, this effect could be visible if
the duty cycle is increased, e.g., d(k) = 0.6, thus, increasing the power to be
dissipated.

Remark2: an higher Rate Transition sampling improves the visualization

of the signals. However, if the number of samples is too large (i.e., a low sam-
ple time is set) the SCIA buffer might be saturated and the communication
blocked. It can be noted that this saturation depends on how demanding is
the executed control algorithm as well. If the firmware requires a high com-
putational effort in terms of, e.g., processing time and memory usage, the
sampling time must be lowered to keep the data exchange running.
260 Low-Side Shunt Current Sensing
leg n

nA

dc-dc
nB
VS1 + vcc
−
IS2
amplifcation
RSH vSH
stage

Figure 16.9 Equivalent circuit of the test setup for the characterization of the
current sensor for the generic leg n, where n = 1, 2 or 3.

16.3 Sensor Characterization: Experimental Approach

The theoretical characterization presented in the previous section is based on
datasheet parameters only. These latter are subjected to intrinsic tolerances,
which may impact on the amplification circuit gain. In order to achieve accu-
rate ADC conversions, an experimental characterization of the measurement
channel is necessary. Such procedure is also helpful in many other cases, e.g.,
when the datasheet does not contain all the required parameters (e.g., Rin ,
RSH ) or when custom board are used. In addition, a proper offset at zero
current should be identified as well. Focusing on the half-bridge configura-
tion, a widely used procedure foresees a dc current injection into the low-side
of a converter leg. This injection is done by using an external power supply
which forces the current to flow from ground up to the output terminal of
each leg through the free-wheeling diode of the low-side switches. In this case,
the switches must be not operated, i.e., no gating signals must be provided.
However, the dc-link has to be supplied with another power supply because
the amplification stage circuit requires a voltage of 3.3 V, which is derived
from the dc-link voltage. The amplification stage can not work without it.
Figure 16.9 reports an equivalent circuit of the test setup.
Once the setup is correctly supplied, the ADC channel returns a value de-
pendent to the current injected. Different current levels may be used to map
the op-amp behavior. Since this is a forced test, the current flowing in the
free-wheeling diodes should not be too high (they are not designed for this
operation). For the sake of simplicity, only positive currents are tested, and a
symmetrical behavior of the sensors is assumed.1

1 This is not always the case and a characterization even for negative currents is recom-

mended, especially for sensors operated with ac currents.

 Sensor Characterization: Experimental Approach 261
shunt
C2802x/03x/05x/06x C28x

A0 Data

ADC SCI XMT

Figure 16.10 Simulink® scheme included in c28069_offset_F.slx.

Offset calibration: the computation of the offset value is relatively easy

to perform. The idea is to send through the serial port the bit value corre-
sponding to zero current. The settings for the firmware file are reported in the
following.

Firmware Environment
c28069_offset_F.slx (solver: fixed step - ODE4, step size: Ts = 50 µs)

Build a new firmware environment starting from the previous one:

• One ADC block is needed only. This time, no synchronzation with
any ePWM module is required. Thus, set SOCx trigger source on
CPU0_TINT0n. The input channels to be used are ADCINA0 for leg 1,
ADCINB0 for leg 2 and ADCINA1 for leg 3. Here in the following, leg 1
only is characterized only.
• Keep a data type conversion block set to single (for data exchange).
• Add a Rate Transition block with sample time 0.01 s.
• Send the output of the ADCA0 (converted in single) directly to the SCI
Transmit block.
• Delete or comment out the remaining blocks.
The connection scheme is reported in Figure 16.10.

Testing Environment
c28069_offset_T (solver: fixed step - ODE4, step size: Tsim = 100 µs)

Build a new testing environment simplifying the previous one. Indeed, only
one signal is sent trough SCI:
• Keep the Serial Configuration and the Serial Receive blocks. Change
the data size back to [1 1].
• Use a Scope or a Display block to visualize the current reading translated
in a bits.
• Delete or comment out the remaining blocks.
 262 Low-Side Shunt Current Sensing

COM10
115200 COM10 Data
8,none,1 bit

Figure 16.11 Simulink® scheme included in c28069_offset_T.slx.

• Optional: a Mean block could be added to filter the noise, which may be
present in the bit reading. Use the settings presented in the previous exercise
for this block.
Arrange these blocks as reported in Figure 16.11.
By feeding only the dc-link with a given voltage (e.g., VDC = 10 V) without
connecting the second power supply (i.e., no forced current flow), the SCIA
returns the bit value of vSH
bit
corresponding to zero current. Theoretically, this
value should be 2048 for 12 bit ADC. However, due to the components toler-
ance, the resulting value is vSH
bit
= 2067 in this case. The latter value is strictly
related to the hardware. Therefore, every reader performing such test will find
(slightly) different results. The new bit value is used to compensate the bias
highlighted in Figure 16.8.

Experimental Characterization of gadc : this second experimental char-

acterization relates the different recirculating current values (forced by the
second power supply) to the bit values read through vCOM port. The set-
tings for the Firmware Environment and the Testing Environment are
the same as for the offset characterization.
Considering again leg 1 (phase A) only, the characteristic curve of the
measurement channel is derived by injecting different currents and visualizing
the corresponding bit values. The procedure is:
1. Connect the positive terminal of the power supply to the ground
terminal of the DRV8301. Use another power supply to correctly
supply the boosterpack dc-link, thus, the op-amp circuit. If the lat-
ter is not supplied, it does not return any bit value even in presence
of a current.
2. Inject different current values if (k) to understand which are the
corresponding readings for i(t) −→ vSH (t) −→ vSObit
(k), i.e., in bits.
The Table reported in Figure 16.12 shows all these readings.
3. The values shown in Figure 16.12 are assumed to be linearly dis-
tributed. Therefore, they are linearly interpolated in MATLAB®
(e.g., through function polyval,) as shown in the same Figure. Keep-
ing the same characterization even for negative current values, the
resulting gradient of the characteristic is the experimental gain gadc :
mA mA
gadc = 8.8 6= 8.1 (16.11)
bit bit
which is slightly different with respect to that one resulting from the
Sensor Characterization: Experimental Approach 263

2200

current ADC value 2180

if (k) bit
vSO (k) 2160

0A 2067
240 mA 2097
2140

495 mA 2122 2120

800 mA 2161 2100

1A 2182
2080
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Figure 16.12 Characterization of the current measurement channel for leg 1.

theoretical characterization. It is worth noting that if more points

can be investigate, the characterization procedure results to be more
accurate.
In addition, it must be underlined that the same results can be
achieved by using a firmware which already compensates the previ-
ously computed offset, as shown in Figure 16.13.
Repeating the locked-rotor test (i.e., coming back to the full firmware set-
tings), the experimental characterization can be compared with the theoretical
one, i.e., by applying d(k) = 0.2, thus, Va = va (k) ≈ 2 V. The result is shown
in Figure 16.14. The mismatches observed in Figure 16.8 are almost fully com-
pensated. However, it must be noted that the proposed data interpolation is
not accounting for possible non-linear behaviour of the amplification stages.
Thus, the current measured through the ADC channel may overestimate the
actual value. Since an experimental approach is considered, the current is now
labeled as iSCIe (k). This latter still has some thermal influence due to current
flowing and it is sensitive to the rotor clamping, see Figure 16.14.

Validation: the experimental characterization seems to nicely approximate

the steady-state value Yss . However, the whole measurement can be validate
by directly comparing it with an exteranl measurement (a current probe

shunt
C2802x/03x/05x/06x vSO_bit
C28x
A0
Data
ADC
SCI XMT

offset exp.

Figure 16.13 Firmware for ADC channel characterization as an alternative to

c28069_offset_F.slx.
264 Low-Side Shunt Current Sensing

Figure 16.14 Locked-rotor test: comparison between theoretical and exper-

imental characterizations. The same current is evaluated by the two ap-
proaches.

connected to an oscilloscope, for example), which has better accuracy. The

results are reported in Figure 16.15. The iSCIe (k) values are now superim-
posed with iosc (k) given by the oscilloscope, validating the use of the refined
gadc .
It is important to note that the serial communication as a debugging tool
is satisfactory at steady-state only. However, due to the limited hardware
resources for communication, it is not possible to visualize fast transients,
which are typical of voltages and currents. In particular, serial communication
is affected by computational/communication delay which translate in a dis-
crepancy between the solid blue and red lines in Figure 16.15. Since i(k) and

3.5

2.5

1.5

0.5

-0.5

0.5 1 1.5 2 2.5 3 3.5

Figure 16.15 Comparison between the current waveforms measured through

ADC channel (and displayed through SCI) and a current probe connected to
an oscilloscope.
Sensor Characterization: Experimental Approach 265

d(k) are read via the SCIA, they are both affected by the same delay. This
latter can be reduced by using a faster sampling rate (in this case 0.01 s should
be reduced) until the MCU capabilities are reached. For instance, if a current
probe is able to measure a transient with a classical exponential waveform, the
serial communication tend to approximate it to a step-wise signal as a con-
sequence of the lack of samples. Figure 16.15 also shows an offset in iosc (k).
This is a consequence of the connection between the LaunchXL F28069M
board and the BOOSXL - DRV8301 converter. Indeed, the MCU is powered
by the power supply connected to the dc-link, requiring a minimum current
absorption. It can be noted that the same locked-rotor test can be performed
considering a H-bridge configuration.

Be careful: integer numbers are used to perform the sum aimed to re-
move the offset. Therefore, int16 is chosen as ADC output
data type to allow computations involving negative values.
As seen previously, it is always possible to use (directly)
single or double numbers to increase accuracy.
 17
Current Control of an RL Load

After the explanation on how to manipulate measurements with the

LaunchPad™ F28069M board, the following Sections show how to realize a
closed-loop current control of a passive RL load. The on-board shunt resistors
of the Boosterpack TI BOOSTXL-DRV8301 converter is used to measure the
load current (read via ADC peripheral), and one or two converter legs can be
involved to realize a Half-Bridge or a Full-Bridge configuration. In particular,
this Section focuses on the controller implementation (PI-based) to meet the
application requirements using a Half-Bridge configuration. Aiming to follow
a given current reference i∗ (k), the resulting PI controller regulates the dc
output voltage of the converter by manipulating the command signal(s) of
the ePWM peripheral(s). This is a first and simple example of a closed-loop
system implementation, which is equivalent (in practice) to the current control
of a DC motor in a locked rotor condition (i.e. back-emf e = 0). The refer-
ence value of the output current can be provided by an external command,
which may come from the SCIA/B (i.e., serial communication), or by the po-
tentiometers mounted on the extPot3 or internally generated (e.g., through a
pulse generator/constant block).
The whole system is simulated and analyzed first through Simulink® simu-
lations. Then, a firmware is designed and implemented based on these results.
The controller design for this application follows the methodology presented
in Chapter 6. Nevertheless, this Section focuses more on the converter stage
and how it can operate. In particular, the extRL(C) expansion board adopted
for the implementation, it is possible to vary the inductance value L. Thus,
some experimental tests are proposed to evaluate the effectiveness/robustness
of the designed controller.

17.1 Required Hardware

• TI LaunchPad™ Piccolo™ F28069M and BOOSTXL-DRV8301 boards;
• extPot3 custom board (see Appendix B for technical details);
• extRL(C) board in RL configuration (see Appendix B for technical details);
• Power supply set on 15 V, 2 A.

DOI: 10.1201/9781003196938-17 267

268 Current Control of an RL Load

Figure 17.1 Test-bench including a TI BOOSTXL-DRV8301 converter

mounted on the TI LaunchXL F28069M Piccolo™ board and an extRL(C)
bard configured in RL mode (see connection for Half-Bridge configuration
in Figure 17.10). This setup includes the extPot3 board connected to the
LaunchXL F28069M Piccolo™ . The reader is referenced to Appendix B for
further details.

This hardware should be arranged as shown in Figure 17.1. Both Half-Bridge

and Full-Bridge configurations can be used as converter stage topologies. The
corresponding equivalent circuits are shown in Figure 17.2 (a) and (b), respec-
tively. The DC link voltage is provided by a power supply set to VDC = 15 V,
with the output current limited to 2 A for safety reason. It can be noted that
the extRL(C) board is designed to sustain up to 4 A currents. Different con-
figurations can be set on this board by simply moving the on-board switches,
as shown in Appendix B. This operation physically connects or disconnects
different components on the printed circuit board. The RL configuration de-
fines a circuit with resistor R = 6.8 Ω and inductor L = 860 µH. A further
resistance might be included to model the series-resistance of the inductor,
which is R` = 150 mΩ based on the manufacturer data-sheet, resulting in an
equivalent resistive parameter equal to Req = R + R` = 6.95 Ω. The two cir-
cuits shown in Figure 17.2 differ from those reported in Chapter 15 for the
absence of the back emf. Moreover, the relationship among duty cycle d(k),
output voltage of the converter vA (k) (and vB (k)) and its averaged value are
discussed in details in this Chapter, since it is a point of paramount impor-
tance in the design of the MCU firmware. In the following, the analytical
model of the RL system is analyzed and a PI controller is designed based on
it. Then, some simulations of the controlled system are reported for the two
arrangements of the power electronics. Finally, experimental tests are carried
out once the simulations returns satisfactory results.
Linear Average Model and Controller Design 269

1A 1A 2A

L i (t)
+ L + iL (t) L R
VDC VDC
− −
vL (t) vL (t) vR (t)

vR (t) R
1B 1B 2B

(a) (b)

Figure 17.2 Equivalent circuits of the test setups involving the RL load driven
by A DC/DC converter in Half-Bridge (a) and Full-Bridge (b) configurations.

17.2 Linear Average Model and Controller Design

Chapter 6 already presented the model derivation of a simple RL load cir-
cuit fed by a controllable voltage source (i.e., 1st order linear system) and a
controller design approach. In particular, a closed-loop control regulates the
current flowing into the winding i(t) = iL (t) by manipulating the voltage v(t)
applied on the RL load, i.e., v(t) = vL (t)+vR (t). This latter is the controllable
variable u(t) = v(t). It is important to note that for a Half-Bridge configu-
ration v(t) = vA (t), while for a Full-Bridge one v(t) = vA (t) − vB (t). Besides
the description provided in Chapter 6, the circuit physics can be analyzed
more in detail during transients. In case a step-wise voltage is applied on the
circuit, the current i(t) cannot vary instantaneously, but it changes based on
an exponential law. Applying the Kirchhoff’s voltage law, it follows that:
diL (t)
v(t) − Ri(t) − L =0 (17.1)
dt
where v(t) is equal to v(t) = VDC or v(t) = 0 depending on the current
switching status of the converter. This non-linear behavior can be graphically
represented by the two simplified, equivalent circuits reported in Figure 17.3
(a) and (b), where the switch is an equivalent representation of the whole
converter. Solving the first order differential equation reported in (17.1) for
the two scenarios, the obtained solution which represents the current transient
going on in the circuit is:
t
iL (t) = (iL0 − iL∞ ) e− τ + iL∞ (17.2)

where the constant value iL0 refers to initial condition of the current for the
considered equivalent circuit (i.e., the steady state value before the operation
of the switch u(t − 1)), while iL∞ is the steady state value resulting from the
270 Current Control of an RL Load

L R iL (t) L R
iL (t)

vL (t) vL (t) vR (t)

vR (t)
+ +
VDC v (t) VDC
−
−

(a) (b)

Figure 17.3 Equivalent circuits used for studying the transients caused by a
switching operation in an RL circuit. The switches can cause (a) zero or (b)
full voltage on the load.

application of u(t) and ideally kept over an infinitely long time interval. This
latter equals Yss defined in Chapter 6. Note that, in open-loop situation, iL∞
(or Yss ) is reached after a time interval which is dependent on the natural
dynamics of the circuit, that is the time constant τ = L/R. For instance, for a
step-wise supply voltage, e.g., varying from 0 up to VDC , the evolution of the
current in time domain can be described by settling a zero initial condition
iL0 = 0 and iL∞ = VDC /R, which are the solutions of the circuits reported in
Figure 17.3. These results lead to:
VDC −t R VDC VDC  t

iL (t) = − e L + = 1 − e− τ (17.3)
R R R
Considering x(t) = i(t) and y(t) = x(t), both a state-space representation and
system transfer functions can be derived based on the solution proposed in
Chapter 6. Then, a PI controller R(s) = kp + ki /s is used to close the loop
and designed with pole/zero cancellation method. Here in the following the
system G(s) and open loop L(s) transfer functions are simply recalled:
i(t) y(s) 1/R i(t) kp /R
G(s) = = = L(s) = = R(s)G(s) = (17.4)
v(t) u(s) 1 + sτG e(t) sTi
where L(s) is computed by setting Ti = ki /kp = L/R. Alternatively, G(s) can
be computed considering the equivalent impedance Z (s) = R + sL.
As previously mentioned, the modeling can consider the parasitic resis-
tance of the inductor, which is summed up to the load resistor. An equivalent
resistance Req = 6.95 Ω is obtained. The uncontrolled system is characterized
by a small time constant τG = L/R = 124 µs, which leads to a settling time of
Ta,G = 619 µs. For a better visualization of the result, the designed controller
is slowing down the natural response by choosing Ta,F = 2.67 × Ta,G = 1.7 ms.
This choice implies τF = Ta,F /5 = 333 µs and a controller bandwidth equal
to ωc = 1/τF = 3000 rad/s. The values of kp and ki are explicitly determined
as:

kp = ωc L = 2.58 Ω (17.5)
ki = ωc R = 20 850 Ω/s (17.6)
System Simulations 271

The dc voltage VDC = 15 V provided by the power supply is considered an

upper bound for the regulator, thus, the control variable u(t) is saturated to
this value. Therefore, an anti-windup strategy must be included in the PID
controller block. Based on Chapter 6, a back-calculation approach is used with
kaw = ki /kp = 8081 s−1 .
The discrete-time PI controller R(z) is derived based on the continuous-
time one and it is integrated in the closed-loop scheme together with a Rate
Transition block in simulation to mimic the discretization process. The sam-
pling time of the controller is Ts = Tsw = 1/fsw , where the switching frequency
is fsw = 10 kHz. Backward Euler is chosen as integration method.

17.3 System Simulations

The closed-loop control of the RL load is simulated in Simulink® to analyze
its design and check the resulting performance of the regulator. This step is
carried out for both converter configurations. To this purpose, the power elec-
tronics and the RL load are modeled through Simscape™ Electrical elements,
referring to Specialized Power Systems ™ library (Simscape/Electrical/Spe-
cialized Power Systems/ ) as shown in Chapter 6.
In particular, the Mosfet blocks available in the subset /Fundamen-
tal Blocks/Power Electronics are used to represent the switch mounted on
the BOOSTXL-DRV8301 converter board. The semiconductor characteris-
tic taken from the converter datasheet, e.g., Ron and Rd , are set into the
Simulink® blocks as shown in Chapter 15. The two-level Half-Bridge configu-
ration is obtained by rearranging these block based on the equivalent circuits
reported in Figure 17.2 (a). The switches are driven by a PWM modulator
having a carrier frequency of 10 kHz, i.e., fsw = 10 kHz. The load is modeled
through a series RLC branch block set as RL branch type with L = 860 µH
and Req = R = 6.95 Ω.

17.3.1 Detailed modeling of the actuation variables

Chapter 6 details how to design a closed-loop current control for a generic RL
load, where the control variable u(t) = v(t), that is, a dc voltage, is applied
to a Transfer Fcn block which emulates the load. This operation is equivalent
to a direct application of u(t) (or, equivalently u(k)) on the RL terminals.
This approach is a high-level analysis which considers an averaged model of
the system, i.e., without including the converter stage or modeling it in a
third nested loop operating at high bandwidth. As already said, the converter
represents a non linear device, since it operates in discontinuous (switching)
mode. Since linear controllers are used, the system model has to be linear too,
thus, not including non-linear elements. Therefore, by using average models,
the controller design can be based on standard approaches from linear control
272 Current Control of an RL Load

carrier fc = fsw VDC

i∗ (t) =
+ current u (t) 1 d (t) S (t) v (t) ũ (t) 1 i (t)
controller modulator AVG
− VDC = R + sL

2L-VSC

inner loop for actuation

Figure 17.4 Closed-loop current control scheme including the inner loop for
the converter actuation and its relationship with the average modeling.

theory,1 leading to compute key parameters and figures of merit, e.g., kp and
ki . Even if the controller parameters are necessary, the firmware architecture
aimed to close a loop has to cope with the driving elements of the converter
stage (modulator, scaling and protections) and the sensing stage as well (in-
cluding measurements and sampling). Getting closer to the system physical
behavior, Chapter 13 and 15 show how to include the converter topology and
the ePWM peripheral working principle (which generates the switch positions)
into the simulation environment. In particular, the actuation variable is the
modulating signal d(t), which is directly a duty cycle when referring to dc
quantities.2 Even though the controller design should be based on a linear av-
erage model, the controller output u(t) must be related to d(t) and the switch
positions S(t). This is detailed in the scheme reported in Figure 17.4. Based
on the controller design and voltage limits (i.e., umax = VDC ), the manipu-
lated variable for a 2L-VSC is defined within u(t) ∈ [0, VDC ]. This last interval
divided by VDC becomes the duty cycle range of variation d(t) ∈ [0, 1]. This
is the input of the modulator operating at carrier frequency fc = fsw for each
converter leg, which operates according to the switching pattern S(t) ∈ [0, 1]
(see Section 13.1). Therefore, S(t) defines how long the dc-link voltage VDC
is applied to the load inside the time interval Tsw . The voltage v(t) is then a
squared waveform approximated by v(t) = S(t)VDC . The average value of v(t)
over Tsw is u b(t), which is u(t) ≈ ub(t), see Figure 17.4. Such approximation
becomes more and more accurate as fsw increases. In practice, the current i(t)
is affected by a ripple which is influenced by fsw and the resulting closed-loop
dynamic. However, the ripple should not be included (or at least limited) in the
feedback path used to compute the current error, otherwise the controller is
not able to reach a zero error at steady-state, i.e. a proper command-tracking
1 Linear control theory deals with systems governed by linear differential equations.

These systems can be analyzed by powerful frequency-domain mathematical techniques of

great generality, such as Laplace transform, Fourier transform, Z-transform, Bode plot,
root locus, and Nyquist stability criterion. These tools lead to a system description using
terms like bandwidth, frequency response, eigenvalues, gains, resonant frequencies, zeros
and poles, and to provide simple/explicit design solutions to control the system response.
An important subset of systems like these is the linear time invariant (LTI) ones.
2 This is not the case for ac quantities.
Half-Bridge Configuration 273

must be ensured.3 Indeed, the measurements sensing should be synchronized

with the ePWM peripheral as discussed in Chapter 13.
The whole approach can be easily extended to discrete-time domain and
represented by a Simulink® scheme. It can be noted that the models based
on the Simscape™ Electrical elements were already including these features in
Chapter 6. In particular, the use of the Controlled Voltage Source block was
mimicking the inner actuation loop.

17.4 Half-Bridge Configuration

As shown in Chapter 15, a Half-Bridge configuration refers to a converter
which uses a single-leg (or single-phase) of the boosterpack, requiring one
ePWM module only, e.g., ePWM1. Hence, the switch 1A is driven by the
signal S1 (k) and 1B by S2 (k), operating in a complementary way. The RL
load is connected between the central point of leg A and ground. The av-
erage output voltage of the converter leg is function of the duty cycle d(k)
(or dA (k)):
V (k) = vavg (k) = d(k)VDC (17.7)
where the variables are already considered in the discrete-time domain k, i.e.,
moving the simulation toward the implementation. The voltage VDC is set by
the power supply (i.e., it is the voltage on the DC bus) while d (k) ∈ [0, 1].
As already mentioned, the system is characterized by R = 6.95 Ω, L =
860 µH. The supply voltage is VDC = 15 V and the switching frequency is
fsw = 10 kHz. These parameters can be initialized in Model Properites/Call-
backs/InitFcn or in a separate m-file as shown here in the following:

%% power supply
Vdc = 15;
%% carrier frequency definition
fsw = 10e3;
Tsw = 1/fsw;
%% sampling time definition
Ts = Tsw;
%% step-size definition
Tsim = Tsw/400;
%% load parameters
R = 6.95;
L = 860e-6;

3 This is another fundamental principle in linear control theory.

274 Current Control of an RL Load
SOC
reload
PWM vmax
y* u s1A
reference Z-1
current controller s1B
modulator
(syncronized with PWM)
il debug
y SOC

computational delay
s1A enable

D
il

S
i
+
- +
s1B

D
+ v
v
-

Continuous

Figure 17.5 Simulink® scheme to simulate the closed-loop current control of

an RL load driven in half-bridge configuration.

%% controller design
tauG = L/R;
TaG = 5*tauG;
wc = 3000;
TaF = 5/wc;
kp = wc*L;
ki = wc*R;

The equivalent circuit shown in Figure 17.2 (a) is then created in Simulink®
shown in Figure 17.5. This scheme is included in the file:

HBsimRL_closed.slx (solver: fixed step - ODE4, step size: Tsim )

HBsimRL_closed.slx (powergui: simulation type Continuous)

The simulation time step Tsim is particularly small to well approximate the
switching behavior of current i(k) and voltage v(k). The reference subsys-
tem includes different kinds of input signals to operate the system in several
working conditions.
Considering the PWM stage, the duty cycle d(k) is compared to a tri-
angular carrier at fc = 10 kHz, generating a switching pattern S(k). This is
realized on the MCU by using the ePWM modules. Since the gate signals of
the 2L-VSC topology are S(k) = S1 (k) for 1A and S2 (k) = not(S1 (k)) for
1B , the instantaneous output voltage v(k) is a pulsed waveform that can be
computes as:
va (k) = S(k)VDC (17.8)
Half-Bridge Configuration 275

Figure 17.6 Closed-loop dynamics considering a step in the current reference

(Half-Bridge configuration).

which is discontinuous and bounded within two voltage values, i.e., 0 and VDC .
The aim of this exercise is to design a current control loop, i.e., to control the
current i(k) flowing into the load based on a step-wise reference i∗ (k) by using
a PI controller while achieving a settling time of Ta,F = 1.7 ms (see Section
17.2). To check these performances, a step-wise current reference i∗ (k) ranging
from 0 A to 1.5 A (i.e., Yss = 1.5) at 0.5 ms is applied.
Figure 17.6 shows the current flowing into the load y(k) = i(k) (blue line),
which is correctly following the step-wise reference y ∗ (k) = i∗ (k) (red dotted
line). The current dynamic follows an exponential behavior as described by
Equation (17.3). Based on the design guidelines reported in Chapter 5, a
 ≈ 1% is considered as reference accuracy. Thus, the transient is completed
once the current enters in the region [(1 − 0.01)Yss (1 + 0.01)Yss ], where
Ta ≈ 5τ holds. Indeed, it results that Ta,F = 1.7 ms, so the closed-loop control
276 Current Control of an RL Load

system is behaving as expected. The steady-state value Yss can be also be

predicted thanks to the final value theorem (see Section 6.1.3).
The effects of the step variation in the reference are reflected into the
switching patterns of 1A and 1B , which are aimed to make the average voltage
vavg (k) increase by enlarging the pulse width of v(k). Moreover, the switching
behavior implies a current ripple equal to ∆iL ≈ 359 mA at steady-state.
Based on Figure 17.6, the command-tracking performances are achieved
thanks to the sudden regulation of the converter output voltage u(k) = v(k),
where its average value is approximated by d(k). At steady-state, u(k) 6= 0
since the system works in a new operating point and a certain voltage is needed
to keep the required current flow. This value can be computed thanks to the
Kirchhoff’s voltage law, where at steady-state the current derivative can be
neglected. Namely, assuming d(k) = Dss at steady state, it follows:

di(t) RIss
v(t)−Ri(t)−L = 0 → Vss = Dss VDC = RIss → Dss = (17.9)
dt VDC
which results in Dss = 0.7. This value is reached approximately within the
settling time Ta,F .
Focusing now on power supplies, they are limited by upper and lower
bounds, which represent static non-linearity. Hence, the control variable
u(k) is bounded by the energy flow capabilities of a power supply. For this
case study, a unidirectional DC source is considered, which works between
umin = 0 V and umax = VDC = 15 V. An anti-windup PI controller based on
the back-calculation method is considered. The coefficient kaw is computed as
kaw = 1/Ti = ki /kp to define how quickly the integrator of the PI controller
is reset by the anti-windup loop. As reported in Section 6.2, the anti-windup
parameters should be defined inside the PID block. In particular, Limit out-
put must be selected and the saturation limits umin and umax have to be
specified. Then, select back-calculation in the Anti-windup method menu and
specify the Back-calculation coefficient Kb equal to kaw . The verification of
the settling time is just the first step of a closed-loop system evaluation. Simu-
lations should provide an useful insight into the system operations, including
as many implementation aspects as possible. The synchronization between
ADC and ePWM modules to obtain an average current measurement is an
example of how important it is to choose the specific sampling time to avoid
issues related to peripheral and task scheduling early during the design stage.
This directly reflects on the calculation and actuation updates that the con-
troller should provide, i.e., integrate MCU scheduling and peripherals in power
electronics-based applications, as discussed in Section 13.10. To this purpose,
an embedded closed-loop control scheme (with MCU target) involving current
reading (ADC module), 2L-PWM acutation (ePWM module) and a PI-based
controller which use the ADC reading to compute the duty cycle for the ePWM
is subjected to a sequential events execution.
 Half-Bridge Configuration 277

Figure 17.7 Example of computations to be performed within the sampling

period Ts .

The current controller is scheduled to be executed within the sampling

period Ts = Tsw = 1/fsw . Therefore, in the target hardware, the controller
routines are synchronized with the ADC interrupt every Ts . Figure 17.7 shows
the event sequence, interrupt trigger, and software execution time for the con-
trol algorithm running in the target hardware. The scheduled event sequence
is:
• At the beginning of Ts , when the value of the PWM counter equals the
PWM counter period (TBCR=TBPRD, that simply equals 1 in Simulink®
simulations), the ePWM peripheral, which is center-aligned in this case (up-
down counting mode), triggers the start-of-conversion (SOC) event for the
ADC module;
• The ADC module converts the sampled analog signal into digital counts and
triggers the end-of-conversion (EOC) event;

• The EOC triggers the ADC interrupt for the controller, which reads the
phase current value;
• Given the PI structure, the controller processes the current error to compute
the modulating signal for the ePWM module. The number as well as the
complexity of these mathematical operations define the execution window
length require by the controller to have a data ready. This time interval is
expected to be sufficiently lower than Ts ;
 278 Current Control of an RL Load
modA
Z-400
modA y*
s1A reference Z-1
1 1
carA y e u
il PI(z) Z-1
2 >0 2 u

enable carA s1B SOC

delay

3
SOC

(a) (b)
Figure 17.8 Detailed view of (a) the modulator subsystem and (b) the equiv-
alent representation of the triggered subsystem for the current controller (see
Figure 17.5).

• The resulting duty cycle is held until the end of the sampling interval and it
is sent to the adopted ePWM module. Therefore, even if the updated duty
cycle value is ready before the end of Ts , the actuation happens only at
the reload of the ePWM peripheral. During the remaining time in-between,
the MCU is assigned to other tasks (e.g., communication) or enters in sleep
mode.
Therefore, the controller adopted in the Simulink® simulation is expected to
be consistent with such event sequence. To this end, the scheme reported in
Figure 17.5 includes a modulator subsystem, in which a trigger is generated
every time the carrier equals 1. Namely, a simulated SOC signal is created.
The implementation is achieved by using a Relational Operator block which
returns one when the carrier is greater than 0.99 or zero otherwise, i.e., a
boolean squarewave is obtained. A detailed scheme of this subsystem is shown
in Figure 17.8 (a). The SOC signal is used as trigger source for a triggered
subsystem4 which realizes a current controller synchronized with the carrier
peaks.
Assuming to adopt the previously mentioned step-wise current reference
i∗ (k), both a Pulse Generator block (sample based, with gain A = 1.5, number
of samples ns = 120, pulse width 50 %, phase delay = 5, and sample time Ts )
and a square-wave of period 120Ts = 12 ms can be used. Focusing on the first
rising edge after the step shown in the top plot of Figure 17.9, the current
reference varies at the zero value of the triangular carrier. Since the ADC
SOC trigger is generated at the carrier peak, i∗ (k) is delayed by Ts /2, thus,
generating a current error at the controller input. This delay can be included
through a Delay block, where the sample time is set equal to Ts /2 and the delay
length is 1. The resulting reference current is called i∗ctrl (k). This is detailed
in the second plot of Figure 17.9 . Even if i(k) shows a superimposed ripple,
4 A triggered subsystem is a conditionally executed atomic subsystem that runs each

time the trigger signal assume one of the selected triggering condition, i.e., rising, falling,
rising-falling trigger type. For more information see https://www.mathworks.com/help/s
imulink/ug/triggered-subsystems.html.
Half-Bridge Configuration 279

Figure 17.9 Example of MCU scheduling and latency between sensing and
actuation.

the center-aligned carrier ensures an ADC SOC which allows to measure the
average value of the current, which is hold until the next SOC, i.e., for the
hwole sampling period Ts . The resulting measured current is called ictrl (k).
The subscript ctrl refers to the quantity that are processed by the controller,
which is expected to read the current error ictrl,err (k) = i∗ctrl (k) − ictrl (k) and
compute the resulting duty cycle d(k). It can be noted that, the controller
starts processing ictrl,err (k) and the related computations at the ADC EOC.
The time interval between ADC SOC and EOC is small but not zero, i.e.,
280 Current Control of an RL Load

both ictrl,err (k) and d(k) data values are not evaluated strictly simultaneously
with the carrier peak (at TBPRD). In Figure 17.9, the time interval between
SOC and EOC is not visible due to the limited figure resolution, but it is still
present. Even if the controller computations are relatively easy to perform,
providing a value for d(k) before the end of the switching period Ts = Tsw , the
controller has to wait the next reload PWM action to feed the resulting duty
cycle value at the ePWM input. This happens when CTR=zero. Therefore,
another Delay block, with sample time Ts /2 and data length 1, is included at
the controller output to synchronize the actuation with the zero value of the
carrier. This event sequence is fulfilled by the implementation of the current
controller as a triggered subsystem. To understand better its behavior, the
involved blocks are showed in Figure 17.8 (b), which is used to analyze the
controller scheduling only at simulation level.
Given the aforementioned parameters and performances, the designed PI
controller meets all the specifications.

17.4.1 Control implementation

The closed-loop dynamics are tested on a test-bench which includes a TI
BOOSTXL-DRV8301 converter mounted on the TI LaunchXL F28069M
Piccolo™ board and connected to the extRL(C) board in RL mode (see Figure
17.1). The extPot3 board is also used to manipulate analog inputs. In Half-
Bridge configuration, only leg (phase) A is used, as shown in the connection
scheme reported in Figure 17.10. Therefore, one ePMW module is used only
and it is operated in up-down counting mode at fsw = PWMfrequency = 10 kHz.
Reminding that for the F28069M board fck = CPUfrequency = 90 MHz, it fol-
lows:
CPUfrequency
TBPRD = PWMcounter−period = = 4500 (17.10)
2 · PWMfrequency

where the sampling time is Ts = Tsw = 1/PWMfrequency .

The reference current can be internally generated in the firmware or exter-
nally provided either using the available potentiometers from extPot3 board
or through COM port (e.g., SCIA). The analog solution is adopted to add ver-
satility to the control scheme, avoiding the slow communication issue through
serial communication. The needed parameters can be initialized in Model
Properites/Callbacks/InitFcn or in a separate m-file as done for simulation.
A step-by-step procedure for programming the microcontroller for this task
is reported here in the following.

Firmware Environment
c28069_RLcontrol_hbF.slx (solver: fixed step - ODE4, step size: Ts =
Tsw /2 = 50 µs)
 Half-Bridge Configuration 281

extRL(C)
used in RL mode

R=6.8 no C L=860 µH
GND
VCC
legA

enable
(ADCB3)
DC link

Figure 17.10 Connections of the extRL(C) board with the BOOSTXL-

DRV8301 converter realizing a Half-Bridge configuration.

Open a new blank Simulink® project and configure the environment as shown
in Section 9. Remember that the BOOSTXL-DRV8301 is enabled by settling
GPIO50 high (see Section 3.2).
• Insert an ePWM block and use the same settings reported in Section 13.5:
– Select ePWM1 module with TBPRD related to fc = 10 kHz, i.e.,
PWMcounter−period = 4500;
– Remember to unflag Enable ePWM1B and tick Inverted version of
ePWMxA. This option allows to drive 1B as the complement of 1A ;
– Keep all the other settings as reported in Section 13.5.
• Add a gain block to normalize the modulation signal over the peak value of
the carrier signal PWMcounter−period . Include a data type conversion set
on int16 to represent the duty cycle within a 16-bit base.
For up-down counting mode, the measurement of an average current i(k)
is feasible thanks to the ePWM module which sends a synchronization sig-
nal that triggers the SOC of an ADC module when the carrier equals
TBCTR=TBPRD, as shown previously. To this aim, the following settings
are summarized here below:
• In the Event Trigger (Tab) of ePWM1
– Select Enable ADC start of conversion for module A;
 282 Current Control of an RL Load

– Set Number of event for SOCA to be generated to First event;

– Select Counter equals to period (CTR=PRD) as Start of conversion
for module A event selection.
Every time the carrier signal reaches the maximum value TBPRD, a SOC
input (i.e., ePWM1_ADCSOCA) is sent to the ADC. This last block has
to be configured as well. According to the BOOSTXL-DRV8301 datasheet
(see Section 3.2) and Chapter 16, the shunt-resistor of leg A is connected to
ADCA0 channel:
• Add a ADC block and select ADCINA0 as Conversion channel in the
Input Channels tab. Then, go back in the SOC Trigger tab;
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC0;
– SOCx acquisition window: 7;
– SOCx trigger source: ePWM1_ADCSOCA;
– ADC will trigger SOCx: No ADCINT;
– Sample time: Ts ;
– Data type: uint16;
– Flag Post interrupt at EOC trigger (optional);
– Interrupt selection: ADCINT1 (optional);
– Flag ADCINT1 continous mode (optional).
• Add a Data Type Conversion block set to int16 (since positive signal
only are used) or single to exploit 32 bit.
• Place a Gain block for scaling the digitalized signal within [0, 1] since a
12-bit reading returns values ranging from 0 (0 V) up to 212 − 1 = 4095
(3.3 V).
• Build a proper conversion chain for the current sensing by removing the
estimated offset (e.g., 2064) as well as using the scaling factor gadc to obtain
Ampere values after moving to single data type. See the characterization
procedure in Chapter 16.
• Use the same discrete PID controller block adopted in simulation; set
the controller type as PI (Parallel) with sample time Ts ; set P=kp and
I=ki . Select an external source for the reset with rising condition. Saturate
the controller output to keep the voltage in the range [0 − VDC ]. A more
conservative choice might be to consider 0.98 · VDC as upper bound.5 Set a
back-calculation method as anti-windup strategy with Kb=kaw .
5 This choice implies to keep a sufficiently good current sensing, since the shunt-resistor

measures the current only when S2 (k) = 1 only, that occurs at every switching period when
d(k) < 1.
 Half-Bridge Configuration 283
[iref]

A
Z-1
y* C2802x/03x/05x/06x
pulses PI(z) u
[i] WA
y
ePWM
en
C2806x
enable [ref] GPIOx
>= 0.5
C2802x/03x/05x/06x
GPIO DO
B3
C28x
ADC
Data
C2802x/03x/05x/06x [ref]
SCI XMT
A0

ADC
shunt A [i]
offset exp

Figure 17.11 Simulink® scheme included in c28069_RLcontrol_hbF.slx.

• A redundant saturation block can be used to protect the system by limiting

the required voltage.
• Add a Gain block to normalize the controller output (which varies in the
[0, VDC ] range) with a scaling factor 1/VDC . Then, the signal spans a [0, 1]
range, being consistent with the definition of d(k).
• Add a Pulse Generator (sample based) block to create the reference i∗ (k).
Set the amplitude equal to A = 1.5, number of samples ns = 120, pulse
width 50 %, phase delay 5, and sample time Ts . It is equivalent to have i∗ (k)
generated by a square-wave of period 120Ts = 12 ms. It can be noted that
i∗ (k) is synchronized with the zero value of the triangular carrier
• Add a Delay block to be consistent with the ADC SOC trigger generated
at the carrier peak. Set sample time Ts /2 and delay length 1.
Arrange these aforementioned blocks as shown in Figure 17.11.

Resetting the Integrator

Every time the firmware is downloaded into the microcotnroller, the PI con-
troller (as well as the the other functions/routines) is constantly executed.
Since a closed-loop structure involves feedbacks which are processed by the
peripherals and the conversion chains, the resulting measurement values are
affected by noise, being nonzero. Therefore, even without a reference or with
its value equal to zero, during the first seconds of firmware executions the
PI controller tries to react providing a control action to change the system
state. This latter is particularly important. Indeed, solvers in Simulink® se-
quentially computes/solves states and outputs from block-to-block. When a
reference is applied but it is not synchronized with the controller action, the
closed-loop system may see a state discontinuity. This lead to a break down
of the computation process due to numerical conditions, which in this case
 284 Current Control of an RL Load

means to feed the integral part (integrator) of the controller with an infinite
value. So, even if the closed-loop system is well designed, this issue could make
the control not operating, e.g., I/O freezing or even entering in fault mode
(nFault) condition, causing the controller action to be not suitable for the
load.
To avoid this issue, the PID controller block has an external reset that
can be selected/activated in the Initialization tab, as described above. By en-
abling this feature, an additional input port is created which allows to specify
the trigger condition that causes the block to reset the integrator to initial
conditions, e.g., Source: internal - Integrator: 0. By default, the output of the
integrator is set to initial conditions when the scheme/code is initialized, while
it is not during the simulation/firmware execution. When the trigger occurs,
i.e., sending a signal to the External reset port, the integral part of the PI
controller is reset on a rising-, falling-edge, or level of such signal. The port
icon indicates the selected trigger type. For this exercise, an external reset on
the rising-edge is chosen.
In order to manually manage the activation of the PI controller, the exter-
nal reset (rising) is connected to a potentiometer (ADCINB3/pin68). Without
moving the latter, the control algorithm is forced not to provide any actuation
signal. To do that:
• Add an ADC block and select ADCINB3 as Conversion channel in the
Input Channels tab. Then, go back in the SOC Trigger tab:
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC1;
– SOCx acqusition window: 7;
– SOCx trigger source: Software;
– ADC will trigger SOCx: No ADCINT;
– Sample time: Tsig = 0.0011,s;
– Data type: uint16.
• Add a Data Type Conversion block set to int16 (since positive signal
only are used) and the 12-bit scaling to move from a range (0 V) up to
212 − 1 = 4095 (3.3 V) to [0, 1].

• Add a Compare to Constant block which is used to evaluate a greater or

equal condition referred to Constant value 0.5. If this condition is true, the
output switches from zero to one resetting the integrator of the PI controller.
The enable signal is defined as en(k).
• Additionally, en(k) can be used to add a further protection redundancy.
Indeed, by using a Product block, the enable signal is multiplied by the
actuation variable. As long as en(k) = 0 the duty cycle is forced to zero
d(k) = 0, i.e., without switching.
 Half-Bridge Configuration 285

Figure 17.12 Closed-loop dynamics measured from the F28069M board pins
through an oscilloscope (Half-Bridge configuration).

• To visually evaluate the pulsed reference i∗ (k), the Pulse Generator can be
connected to a GPIO block (e.g., GPIO7) with a proper scaling of 1/1.5
to normalize the signal in a [0, 1] range. This allow to visualize through an
oscilloscope the reference signal ranging in [0, 3.3] V at pin 79.
Optional: the firmware could be also prepared to a debug through the serial
COM. In this case:
• Add Data Type Conversion and Rate Transition (e.g., TT X = 0.01 s)
blocks to prepare the data exchange via SCI Transmit block. Figure 17.11
shows how those blocks should be connected. See Section 10.5 or Chapter
16 for further details.
• In this case, even a Testing Environment must be arranged.
Figure 17.12 reports the closed-loop behavior of the system obtained by exe-
cuting the given firmware. The system dynamics are validated through an os-
cilloscope. In particular, the current i(k) is measured by using a current probe,
while the converter output voltage v(k) and the gating signals S1 (k), S2 (k)
(considered as voltage levels at the related pins) are measured by voltage
probes. These latter are also used to check the current reference available at
pin 79.
286 Current Control of an RL Load

Figure 17.13 Effects of a processing delay on the current.

As done in simulation, the analysis of Figure 17.12 starts from the eval-
uation of the command tracking performances. The resulting i(k) (blue line)
is following the internally generated i∗ (k) (red line), which is re-scaled from
3.3 V on the actual reference value in Ampere. The settling time is slightly
larger than 1.7 ms. This is due to the presence of a delay in the converter
actuation. Indeed, both the gating signals of 1A, 1B switches and the volt-
age v(k) are operating late compared to the dynamics shown in Figure 17.6.
Moreover, this delay also implies a (small) overshoot in i(k) that modifies the
shape of the current response. The current ripple is ∆iL ≈ 540 mA. Thus, it is
slightly larger than that one computed in simulation, but it is still compara-
ble. This value is influenced by the inductor parameters as well as the probe
accuracy. For instance, the series-resistance R` might be lower than the value
reported in the datasheet. The Pulse generator which provides the internal
reference i∗ (k) is synchronized with the zero value of the triangular carrier.
Nevertheless, such synchronization is not as strict as other tasks, e.g., counter
comparison in ePWM, which is based on hard synchronization (hard-sync)
event through interrupts. Therefore, it might happen that the rising edge of
the step-wise i∗ (k) is not perfectly synchronized with the ADC SOC (i.e.,
carrier peak). This effect can be simulated by using the Simulink® scheme in
Half-Bridge Configuration 287

Figure 17.14 Closed-loop dynamics considering a delay Td = Ts /4.

Figure 17.5 which already includes the possibility to add a delay Td up to

Td = 2Ts . The results are analyzed through Figure 17.13. Theoretically, i∗ (k)
is synchronized with TBPRD=0. By adding a fixed shift of Ts /2, i∗ctrl (k) is
coincident with ADC SOC, thus, sampling a value of i∗ctrl (k) = 1.5 A. If a
delay is present (here quantified by Td ), the current sample processed at the
SOC is i∗ctrl (k) = 0 A, which implies a zero error ictrl,err (k) and no need for
any control action. It is necessary to wait until the next carrier peak to detect
the value i∗ctrl (k) = 1.5 A, after which an error is computed and d(k) 6= 0 is
applied. Therefore, in a worst case scenario, the overall interval between the
the rising edge of i∗ (k) and the starting of the system dynamics (i.e., when
i(k) starts to increase due to v(k) 6= 0) reaches 2Ts . Figure 17.14 shows the
simulation results by considering a delay of Td = Ts /4, which implies an actu-
ation delay close to 2Ts . The system dynamics and the width of the ripple are
compatible with those obtained by the experimental results shown in Figure
288 Current Control of an RL Load

17.12, validating the firmware.

Remark: if the reference i∗ (k) is particularly close to the maximum current

limit, which is related to the maximum voltage provided by the power supply,
i.e., VDC , the current sensing may be quite inaccurate because d(k) is ap-
proximately 1.6 Indeed, the low side shunt resistor is reading when S1 (k) = 0,
thus, d(k) should be sufficiently lower than 1 to allow a off-time window7 large
enough to perform the current sampling, which is centered on the off-time (see
Figure 17.6 and Section 13.9).

Remark: the same study and implementation can be repeated for a Full-
Bridge configuration of the converter, which requires two ePWM modules,
e.g., ePWM1 and ePWM2. The same procedure is suggested, that is to start
from simulation first and, then move to implementation. Moreover, the con-
siderations on the processing delay are still valid. The ripple on the current
changes depending on the adopted strategy for the converter operation, that
is, unipolar or bipolar voltage switching. For more information see Chapter
15.

17.5 Variation of Load Parameters

In order to evaluate the effectiveness/robustness of the controller, the ex-
tRL(C) board allows to change on-the-fly the load parameters, such as the
inductance L and the resistance R values. The main purpose of the designed
controller is to guarantee a stable current regulation despite load changes.
Nevertheless, even if the tracking of i∗ (k) is guaranteed, load changes imply
different settling time and current waveforms compared to the previous stan-
dard case. Such analysis is useful to understand which parameter variations
may lead to potential damage or instability in the system.
The control architecture is left equal to the previous ones as for the step-
wise current reference. The inductance can be varied from L = 860 µH up to
L = 4 · 860 µH, while the resistance from R = 6.8 Ω down to R = 6.8/2 Ω
(without considering R` ). The implementation still refers to a Half-Bridge
configuration.

17.5.1 Effects on the transient response

The same parameters adopted for the previous exercise are kept as reference,
i.e., L = 860 µH and R = 6.8 Ω + R` , as well as the resulting dynamical
responses, e.g., ∆IL = 540 mA.
An increase in L by a factor of 4 is performed while keeping the resistance
value constant. As theory predicts, a bigger inductance L implies a lower
6 Dead-time protection and signal saturation are always present.
7 Ifthe duty cycle d(k) is defined as the on-time over the switching period, the off-time
represents the remaining time interval.
Variation of Load Parameters 289

Figure 17.15 Closed-loop dynamics measured from the LaunchXL F28069M

board pins through an oscilloscope in correspondence of a variation of the load
parameters (Half-Bridge configuration).

current ripple ∆IL . At steady state, an increase of four times in the inductance,
L = 3.44 mH results in a quite relevant current ripple reduction, i.e., ∆IL =
259 mA. As a matter of fact, given the same steady-state value for the current
(i.e., average value), the continuous-time domain current waveform and, thus,
the ripple can be computed considering the integral:

1 t2
Z
i(t) = iL (t) = vL (t) dt (17.11)
L t1

which is defined within a generic switching period. For instance, t1 = 0 and

t2 = Tsw = Ts . Keeping constant fsw , the voltage applied on the inductor vL (t)
changes based on the variations of L, allowing to achieve the same average
value of the load current, but with a lower fluctuation than the reference case.
Such effects are visible in Figure 17.15, top and central plots. It can be noted
that the integral is not considering thermal variations or magnetizing effects
on the inductor, which may further influence the current ripple. Nevertheless,
since the closed-loop current controller is designed on the dynamic subject
to L = 860 µH, the settling time results to be larger than before. This is
also motivated by the fact that the controller is expecting a quicker current
 290 Current Control of an RL Load

u (t) +
L
−

x (t)

(a) (b)

Figure 17.16 Open-loop identification of an RL load: measured current (a)

and equivalent circuit (b).

reaction, which is now not possible due to a greater energy stored in four
inductors (insted of one), that implies an higher electrical inertia. Despite the
different settling time, the command tracking is still guaranteed.
On the other hand, decreasing R, e.g., moving to R = 3.48 Ω, while keeping
the reference inductance value does not have any impact on ∆Il . Due to
this low resistance, the current rapidly increase enlarging the overshoot if the
same control action applied in the reference case (i.e., designed considering a
larger R) is provided. This imply a faster dynamic, thus, a smaller settling
time, while the command tracking is still guaranteed. However, the current
overshoot might be dangerous for the system, which now have to sustain an
higher current and possible overheating.
Both cases are simple but effective test which could represent practical
scenarios that may lead to components degradation or load revamp.

Remark: the same effects caused by variations of load parameters are valid for
the Full-Bridge configuration, both for unipolar and bipolar voltage switching.

17.5.2 Parameters estimation

In case of large parameter uncertainty, different identification techniques may
be used to compute the load values. As an example, the simplest technique
consist in a open-loop identification of the system to be controlled, i.e., the
RL load (see Figure 17.16).
Based on the definition of the transfer function G(s) and  from Chapter
5 and 6, the settling time is Ta , G = 5τG . Then, the identification procedure
follows three steps:
• Drive the system in open-loop (connection scheme as shown in Figure 17.10)
evaluating the system response with an oscilloscope.
 Variation of Load Parameters 291

• By applying a constant duty cycle, the load is subject to a constant (average)

voltage Vss at steady state. It follows that i(k) = Yss = Vss /R. Since the
voltage is know, the resistance can be simply estimated as R = Vss /Yss (see
Figure 17.16 (a)).
• The constant duty cycle forces the system to react with its natural dy-
namic. The open-loop settling time can be directly read from the oscillo-
scope. Therefore, provided the resistance estimation, the inductance can be
computed as L = RτG = RTa,G /5 (see Figure 17.16 (a)).
Different operating points (i.e., different values of d(k), thus, Vss ) can be
tested in order to refine the parameters computation. This operation might
be useful when the system is sensitive to physical or environment variations.
As an example, in case the TI LaunchPad™ Piccolo™ F28069M board, the
BOOSTXL-DRV8301 converter together with the extRL(C) load are operat-
ing in a closed box where the ambient temperature is different from 25 ◦C,
the data-sheet values could be corrected for the considered working condition
with an identification procedure. It can be noted that, even in a standard
environment, a high current flowing in the RL load affects the component
temperature, impacting on the values of R and R` .
The same identification procedure can be also used to estimate the pro-
cessing delay introduced by the control algorithm.

Remark: in order to limit the components stress, it is recommend not to

apply the full dc-link voltage during the open-loop test. Set VDC as the upper
bound, suitable voltages for this test may be VDC /2 or VDC /3.
 18
Voltage Control of an RLC load

This chapter shows how to realize a closed-loop voltage control of a passive

RLC load. An extRL(C) board is adopted. The isolated output voltage sensing
stage included in the latter (i.e., it is external to the MCU board) is used to
measure the voltage across the resistor through the ADC peripheral. A Half-
Bridge converter configuration is considered, thus, using one leg of the TI
BOOSTXL-DRV8301 converter only. This aims, in practice, to realize a (non-
isolated) bidirectional step-down DC-DC circuit, i.e., two switches connected
to a LC filter. The latter is aimed to decrease the input voltage to a lower
and stable output voltage with a low harmonic content.
Even if a simple resistive load is considered, comments on the design of
the closed loop scheme proposed here in the following and on its closed-loop
dynamics are quite general. Aiming to follow a given voltage reference vo∗ (k),
the resulting PI controller regulates the dc output voltage of the converter by
manipulating the command signal of the ePWM peripheral, that is, d(k). The
reference value of the output voltage can be provided by an external command,
which may come from serial communication, or by the extPot3 potentiometers,
or internally (e.g., through a pulse generator/constant block).
The whole system, which modeling is based on the general state-space
averaging method, is simulated and analyzed first through Simulink® simu-
lations. Then, a firmware is designed and implemented on the LaunchPad™
F28069M board based on these results. Moreover, guidelines for the hardware
design of an RLC load are also provided. This Section focuses on the compo-
nents selection of the conversion stage based on given requirements, such as
power rating, current/voltage ripple and stability. Since the extRL(C) board
is a configurable load, the effectiveness/robustness of the closed-loop control
scheme is experimentally tested through online variations of the inductance
L, capacitance C and resistance R values.

18.1 Required Hardware

• TI LaunchPad™ Piccolo™ F28069M and BOOSTXL-DRV8301 boards;
• extPot3 custom board (see Appendix B for technical details);

DOI: 10.1201/9781003196938-18 293

 294 Voltage Control of an RLC load

Figure 18.1 Test-bench including a TI BOOSTXL-DRV8301 converter

mounted on the TI LaunchXL F28069M Piccolo™ board and an extRL(C)
bard configured in RLC mode. The reader is referenced to Appendix B for
further details.

• extRL(C) board in RLC configuration (see Appendix B);

• Power supply set on 15 V, 2 A.

This Chapter requires the same hardware adopted in Chapter 17, as shown
in Figure 18.1 and 17.1. Half-Bridge configuration is used as converter stage
topology. The corresponding equivalent circuit is shown in Figure 18.2. The
latter shows that the system to be controlled is a resistive load R connected to
a LC filter which smooth the discontinuous voltage supplied by the two-level
converter leg. The DC link voltage is again provided by a power supply set to
VDC = 15 V, with the output current limited to 2 A. It can be noted that the
extRL(C) board is designed to sustain up to 4 A.
There is no unique control paradigm that solves this power electronic-based
control problem [40]. Only one among the classical DC-DC converter control
approaches is presented in this book. Namely, the so called voltage-mode con-
trol is outlined together with its related embedded implementation. The ex-
tRL(C) board has a voltage sensing stage connected in parallel to the resistive
load, allowing to measure vo (k). This voltage is scaled by a voltage divider in
a [0, 3.3] V range, so that it can be directly read by an ADC peripheral.
Differently from Chapter 17, the relationship among duty cycle d(k), out-
put voltage of the converter vA (k) and its averaged value are described by
using the general state-space averaging method. In particular, the control pro-
posed in Chapter 17 is a current-mode control, which considers the output volt-
age v(k) as the controlled variable. Thus, there is no need to specify the duty
cycle in the derivation of the system transfer functions both in continuous- or
discrete-time domain. For a voltage-mode control, the duty cycle is required to
Required Hardware 295

1A
L, C and R values can
be changed on-the-fy
L iL (t)
VDC +
−
vL (t)
v (t)
C vo (t) Ro
1B
ic (t)
io (t)

Figure 18.2 Equivalent circuit of the test setup involving the RLC load, which
is connected to a bidirectional step-down DC-DC converter.

be directly considered in the transfer function. This is achieved by applying

the general state-space averaging method, which is a procedure to compute a
low frequency model of the system to be controlled.
Different configurations can be set on this extRL(C) board by simply mov-
ing the on-board switches, as shown in Appendix B. This operation physically
connects or disconnects different components on the printed circuit board.
The RLC configuration has a R = 6.8 Ω resistor, a L = 860 µH inductor
and a C = 100 µF capacitor. Given the manufacturer data-sheet, the system
modeling can be refined by including further (parasitic) resistances, such as
the series-resistance of the inductor, which is R` = 150 mΩ at 20 ◦C and the
equivalent series-resistance of the capacitor (ESR), e.g. ESR = Rc = 66 mΩ
at fsw = 10 kHz (according to the ESR-f characteristic).
As detailed in Appendix B, the extRL(C) is a configurable load that allows
to modify the L, C and R parameters through on-board manual switches. For
instances, the switch S1 can be used to put in parallel two equal 6.8 Ω resisi-
tors, thus, resulting in a Req = 3.4 Ω value seen at the output terminals (i.e.
accross the capacitor). Such manual resistance variation may be useful to test
the robustness of the control scheme. Indeed, the control algorithm should
stabilize the dc output voltage ensuring either good command-tracking perfo-
mance and sufficient disturbance rejection against load variations (i.e. output
voltage variations). In the following, the analytical model of the RLC system
is analyzed and a PI controller is designed based on it. Then, some simulations
of the controlled system are reported and considered as reference case for the
implementation on hardware. Finally, experimental tests are carried out once
the simulations returns satisfactory results.
 296 Voltage Control of an RLC load

Table 18.1 Starting data for the extRL(C) load design.

VDC Vo Ro fsw imax ipk Pmax

15 V 6V 6.8 10 kHz 2A 4A 7W

Remark: a bidirectional step-down DC-DC converter can operate in con-

tinuous conduction mode (CCM) or discontinuous conduction mode (DCM),
depending on the waveform of the inductor current. In CCM the inductor
current never reaches zero values in a switching period, whereas in DCM it
may even go towards value lower than zero for some time interval. Only CCM
operations are considered in this Chapter.

18.2 Guidelines for the Hardware Design of a RLC Load

As detailed in Chapter 5, the control design depends on a good knowledge
of the load system capabilities. In particular, this means to deal with the
power electronics design in terms of passive components sizing, magnetic and
saturation effects, evaluation of power losses and heat exchange as well as
many more aspects.1 Therefore, before entering in the details of the control,
this Section provides some guidelines to approach the LC output filter design
in a step-down DC-DC converter configuration, that is, the procedure leading
to the design of the extRL(C) board.
The starting parameters are reported in Table 18.1. Uppercase letters are
referring to steady-state quantities. The passive components have to be se-
lected aiming to make the system fulfilling a set of constraints. For example:
• The ripple of the current flowing into the inductor must be ∆iL ≤ 500 mA
• The ripple of the voltage measured across the load resistor (or equivalently
across the capacitor branch) must be ∆vo ≤ 100 mV
These two values are related to target power losses and energy quality. The
hardware design refers to the worst-case scenario, i.e., passive components are
chosen based on the most stressful condition to which the system is subjected.

Steady-State Analysis
Assuming that the capacitor C is sufficiently large to keep vo (t) = Vo , which
implies ic (t) = dvc (t)/dt = 0, the average value of the output quantities at
steady-state2 are:

1 "An engineer who deals with power electronics applications has to know the basics

of power electronics design in order to figure out system physical limits, later used as
information for the control structure" Prof. Ralph Kennel and Prof. Jose Rodriguez.
2 In order to keep a simple notation, the subscript
ss is omitted from variable definitions.
Guidelines for the Hardware Design of a RLC Load 297

Vo
Io = IL = = 882.4 mA (18.1)
Ro
Po = Ro Io2 = 5.3 W (18.2)

where the power dissipated on the load is lower than Pmax = 7 W.

A Half-Bridge configuration foresees the use of one leg of the TI
BOOSTXL-DRV8301 converter (e.g., leg A), for which switches 1A , 1B are
considered. The switch positions are defined as S1 (k) and S2 (k), respectively.
During commutations, the overall switching state is defined by the vector
S(k) = [S1 (k) S2 (k)]T . Due to the two-level converter topology, the inductor
is subjected to the following average voltages:
ON state → S(k) = [1 0]T OFF state → S(k) = [0 1]T
VL = VDC − Vo − 
R`L ≈ 9V
I VL = Vo − 
R`L ≈ 6V
I
where the dc-link voltage VDC is assumed to be constant and the inductor
voltage drop negligible, i.e., R` IL ≈ 0. The MOSFETs included in the TI
Boosterpack3 have a low gate-source voltage threshold and low drain-source
on resistance. Hence, even the MOSFET voltage drop is negligible.
From an energetic point of view, the energy stored in the inductor is
EL = (1/2)L IL2 , which increases during ON states and decreases during OFF
states. The inductor voltage can be related to the current variation within the
switching period Tsw = 1/fsw and it can be computed as:

diL (t) ∆iL

vL (t) = L → VL = L (18.3)
dt Tsw
where the inductor current variation ∆iL can be computed for the ON and
OFF states analytically as:
ON state → S(k) = [1 0]T OFF state → S(k) = [0 1]T
∆iL(on) = VDC −Vo
L DTsw ∆IL(off) = Vo
L (1 − D) Tsw
At steady-state, the amount of energy stored inside the inductor at the be-
ginning and at the end of the switching period is equal EL(on) = EL(off) .
This equality is extended to the current ripple ∆iL(on) = ∆iL(off) . Therefore,
the duty cycle at steady-state D can be computed considering the previous
equations as:
Vo
∆iL(on) = ∆iL(off) → D = = 0.4 → 40% (18.4)
VDC
As the name says, for a step-down DC-DC converter Vo ≤ VDC since the
output voltage has to be lower than the input.
3 The TI BOOSTXL-DRV8301 use CSD18533Q5A N-Channel MOSFET [22].
298 Voltage Control of an RLC load

Figure 18.3 Waveforms of the load voltage, inductor current and position of
switch 1A for a step-down DC/DC converter. The area highlighted in light blue
is the charge ∆Q that is stored (and, then, released) on the capacitor in Tsw /2.
The dotted lines refer to average values, e.g. Io = ∆iL /2 = DTsw VDC /(2L)

Reminding the aforementioned requirement for the current ripple ∆iL ≤

500 mA, ∆iL(on) = ∆iL can be assumed. Hence, the minimum inductance
value Lmin is:
VDC − Vo
Lmin = D Tsw = 720 µH → L ≥ Lmin (18.5)
∆iL(on)

Then, a first selection proposal is a commercial L = 860 µH leaded toroidal

inductor from Würth Elektronik™ Group, with ±20% tolerance, rated cur-
rent 3 A and series resistance R` = 150 mΩ at 20 ◦C. Considering R` , this
component leads to the following current ripple:
VDC − Vo − R` IL
∆iL(on) = D Tsw = 412.4 mA (18.6)
L
Vo − R` IL
∆iL(off) = (1 − D) Tsw = 409.4 mA (18.7)
L
which are both lower than 500 mA. Thus, this requirement is satisfied.
Nevertheless, the effects of the parasitic resistance R` on the load circuit
are not easy to predict with simple equations. For instance, the larger R` ,
the lower the average value of the output voltage Vo . This result is due to a
Guidelines for the Hardware Design of a RLC Load 299

larger voltage drop on the inductance. However, a large R` is useful to reduce

the current ripple ∆iL . Moreover, sufficiently large values of R` add passive
damping in the load circuit to reduce the resonating effects created by the LC
filter. This aspect is detailed in the next Section.
On the other hand, the ripple in the output voltage ∆vo strongly depends
on the capacitance C. Assuming a negligible ESR, i.e., Rc ≈ 0, then it can be
considered that vc (t) = vo (t). Similarly to the inductor voltage, the capacitor
current ic (t) depends on the voltage variation inside the switching period Tsw :

dvo (t) ∆vo

ic (t) = C → Ic = C (18.8)
dt Tsw
During CCM operations, the ripple in the output voltage ∆vo can be com-
puted considering the waveform relationships shown in Figure 18.3. Assuming
that all the ripple component of the inductor current ∆iL flows through the
capacitor only, while its average value IL flows into the load resistor Ro , the
filtering action of C provides an additional charge ∆Q that is stored and re-
leased on the capacitor layers (the light blue area reported in Figure 18.3).
Based on Figure 18.3, the voltage ripple can be computed through a geometric
approach:
Tsw

∆iL
∆Q · 2
∆vo = = 2
(18.9)
C 2C
Reminding the aforementioned voltage and current ripple requirements ∆vo ≤
100 mV, ∆iL ≤ 500 mA and considering equation (18.9), the minimum capac-
itance value Cmin is computed as:

∆iL Tsw
Cmin = = 62.5 µF → C ≥ Cmin (18.10)
8∆vo
A first selection proposal is a commercial electrolytic capacitor with C =
100 µF. Considering the specific current ripples resulting from the implementa-
tion with the selected inductor, the resulting voltage ripples can be computed
through equation (18.9), that is:

∆iL(on) Tsw
∆vo(on) = = 51.6 mV (18.11)
8C
∆iL(off) Tsw
∆vo(off) = = 51.2 mV (18.12)
8C
Nevertheless, the effects of the ESR on the voltage ripple should be verified.
Indeed, the presence of a resistance Rc implies a further voltage drop in the
capacitor branch, that leads to vo (t) = vc (t) + Rc ic (t). In terms of variations,
it becomes:
∆vo = ∆vc + Rc ∆iL (18.13)
The ripple ∆vc can be computed similarly to ∆vo from equation (18.9) through
a geometric approach. The ∆vc waveform equals ∆vo except for a different
300 Voltage Control of an RLC load

Table 18.2 Selected components for the extRL(C) load design.

L R` ∆iL C Rc ∆vo D
860 µH 150 mΩ ≈ 412 mA 100 µF 66 mΩ ≈ 52 mV 0.4

scaling. In order to keep ∆Vo ≈ ∆Vc , an easy and conservative approach4 to

estimate the maximum allowed value for the ESR Rc,max is to consider the
upper bound ∆vo /2. If the voltage drop in the capacitor branch is lower than
∆vo /2, its contribution is negligible. Given ∆vo ≤ 100 mV and ∆iL ≤ 500 mA,
it follows that:
∆vo ∆vo
= Rc,max ∆iL → Rc,max = = 100 mΩ → Rc ≤ Rc,max (18.14)
2 2∆iL
By selecting a capacitor with Rc lower than Rc,max , the parasitic resis-
tance does not have a large impact on ∆Vo . Thus, a C = 100 µF alu-
minum electrolytic capacitor with ±20% tolerance, rated voltage 100 V, and
ESR = Rc = 66 mΩ at fsw = 10 kHz from Würth Elektronik™ Group is
selected.
Many manufacturers produce different versions of the same capacitor with
equal capacitance, but different ESR values. Those with Rc < 40 mΩ are called
low-ESR type. A low Rc is useful to reduce the system power losses. However,
Rc must be large enough to damp the resonating effects together with R` .
Once components are selected, the main system parameters are summa-
rized in Table 18.2.

Remark: this chapter refers to a bidirectional step-down DC-DC converter

topology (also called synchronous buck converter). This device uses two
switches instead of a combination of one switch and one diode, which is the
classical buck converter structure. Given the application targets (e.g., power,
voltage, current ratings), a bidirectional topology may be used to reduce losses.
For instance, if the diode in a buck converter is replaced by a MOSFET switch
with low drain-source on resistance, the resulting power losses are reduced
(especially for low switching frequencies). Moreover, a bidirectional topology
allows a reverse current flow, which can be used in motor applications for
regenerative braking and/or to avoid DCM operations.

18.3 General State-Space Average Modeling Method

Middlebrook and Ćuk introduced in [36] a state-space averaging technique
aimed to derive linear models for power converters and modulation stages.
4 The method proposed here is simple, but not the optimal one. Indeed, many nonlinear

effects should be taken into account.

General State-Space Average Modeling Method 301

carrier fc = fsw VDC

veo ∗ (t) =
+ voltage de(t) S (t) v (t) veo (t)
controller modulator RLC load
− =

single leg

Gp (s) transfer function

Figure 18.4 Voltage-mode control scheme based on the state-space average

modeling method, where Gp (s) describes the whole power stage dynamic, in-
cluding the modulator, power converter and RLC load stage.

The basic idea is to focus only on small AC signals (small signal control
theory), linearizing the system around a steady-state DC operating point.
The step-down DC-DC converter topology shown in Figure 18.2 is consid-
ered. A voltage-mode control scheme based on the state-space average mod-
eling method can be represented as reported in Figure 18.4. All time domain
variables can be represented by a steady-state DC value plus a small AC per-
turbation, e.g., vo (t) = Vo + veo (t). The goal of this approach is to compute
the small signal transfer function Gp (s) = veo (s)/d(s)
e in the continuous-time
domain by using the Laplace operator. In particular, Gp (s) describes in a
condensed manner the dynamics of the modulator, power converter and RLC
load stage. The variables veo (s) and d(s)
e are the small perturbations in the
output voltage vo (t) and in the duty cycle d(t) around their steady state DC
operating values Vo and D, respectively.
Then, a state vector x(t) = [iL (t) vc (t)]T is considered, which includes
the current flowing in the inductor L and the voltage measured across the
capacitor C, as well as the switch position vector S(t) = [S1 (t) S2 (t)]T . This
latter defines the converter output voltage value v(t). The general procedure
can be outlined through the following steps:

1. Compute dynamic models for each allowed circuit config-

urations
In a 2-level topology two cases are possible: ON state S(t) =
[1, 0]T → v(t) = VDC and OFF state S(t) = [0, 1]T → v(t) = 0.
For each scenario, the state equations can be determined as
ON state → d(t)Ts OFF state → (1 − d(t)) Ts
dx(t) dx(t)
= A1 x(t) + B1 VDC = A2 x(t)
dt dt
. vo (t) = C1 x(t) . vo (t) = C2 x(t)
302 Voltage Control of an RLC load

where matrices A1 and A2 and vectors B1 , C1 and C2 are defined

based on the system parameters.
2. Average the state-variables using the duty cycle d(t)
To achieve an average description of the dynamic models in Tsw ,
the equations corresponding to the two foregoing states are merged
in a time weighted and averaged manner using d(t):
dx(t)
= [A1 d(t) + A2 (1 − d(t))] x(t) + B1 d(t)VDC (18.15)
dt
vo (t) = [C1 d(t) + C2 (1 − d(t))] x(t) (18.16)

3. Introduce small AC perturbations and split AC and DC

dynamics descriptions
Small AC perturbations, denoted by the hat m,e are superimposed to
DC steady-state quantities, which are represented by capital letters:
x(t) = X + x
e(t) vo (t) = Vo + veo (t) d(t) = D + d(t)
e (18.17)
where a constant dc-link voltage VDC is assumed. After some math-
ematical commutations using the definitions reported in equation
(18.17), (18.15) and (18.16) are rewritten as:
de
x(t)
= AX + BVDC + Ae x(t) + [A1 D + A2 (1 − D)] x
e(t)
dt
=== + [(A1 − A2 ) X + B1 VDC ] d(t)
e (18.18)
veo (t) = CX + C x e − Vo AAA
e(t) + (C1 − C2 ) X d(t) (18.19)

where terms containing products of veo (t) and d(t)e are neglected
since they are second order terms. To simplify the notation, matrices
A = A1 D +A2 (1 − D), B = B1 D and C = C1 D +C2 (1 − D) are
defined. The steady-state contributions shown in equation (18.18)
and (18.19) can be computed by setting all the AC perturbation
terms and their time derivatives equal to zero. It follows:
AX + BVDC = 0 (18.20)
CX − Vo = 0 (18.21)
This operation leads to compute the steady-state voltage transfer
function, which can be seen as a DC gain:
Vo
= −CA−1 B (18.22)
VDC
By removing the steady-state contributions from equations (18.18)
and (18.19), a small AC perturbation description is finally achieved.
dex(t)
= Aex(t) + [(A1 − A2 ) X + B1 VDC ] d(t)
e (18.23)
dt
veo (t) = C x
e(t) + [(C1 − C2 ) X] d(t)
e (18.24)
General State-Space Average Modeling Method 303

4. Rewrite AC equation in s-domain and compute the power

stage transfer function
The power stage transfer function Gp (s) can be computed using
the Laplace operator, from the small signal AC model reported in
equation (18.23) and (18.23):

Gp (s) = veo (s)/d(s)

e = (18.25)
−1
= C (sI − A) ((A1 − A2 ) X + B1 VDC ) + (C1 − C2 ) X

18.3.1 Linear average model and controller design

The state-space average modeling presented in the previous section is specific
for a 2-level converter topology and it can be extended to many different load
stages. Given x(t) = [iL (t) vc (t)]T and S(t) = [S1 (t) S2 (t)]T , the step-by-step
modeling procedure starts considering:
ON state → S(t) = [1, 0]T OFF state → S(t) = [0, 1]T
dx(t) dx(t)
= A1 x(t) + B1 VDC = A2 x(t)
dt dt
. vo (t) = C1 x(t) . vo (t) = C2 x(t)
where the matrices A = A1 = A2 , B = B1 and C = C1 = C2 are defined as
 Ro Rc + Ro R` + Rc R` Ro 
− −
L (Ro + Rc ) L (Ro + Rc ) 
A=

Ro 1 
−
C (Ro + Rc ) C (Ro + Rc )
" 1 #
B= L
0
 
Ro Rc Ro
C=
Ro + Rc Ro + Rc

The DC gain of the steady-state voltage transfer function is:

Vo Ro + Rc Vo
=D if Rc , R` ≈ 0 → =D (18.26)
VDC Ro + Rc + R` Vi
Furthermore, the power stage transfer function Gp (s) can be computed from
equation (18.25), based on the small signal AC perturbation model:
304 Voltage Control of an RLC load

veo (s) −1
Gp (s) = = C (sI − A) BVDC =
d(s)
e
1
s+
Ro Rc CRc
= VDC =
L (Ro + Rc ) C (Ro (R` + Rc ) + Rc R` ) + L Ro + R`
s +s
2
+
LC (Ro + Rc ) LC (Ro + Rc )
| {z }
oscillating response

ωo2 s + ωz
= VDC (18.27)
ωz s2 + 2ξωo s + ωo2
| {z }

The highlighted term of Gp (s) includes a 2nd order denominator which models
resonating effects. Therefore, the intrinsic behavior of a LC filter increases the
model complexity. In particular, its oscillating response is identified by the
resonance frequency ωo (which is associated to the poles), the anti-resonance
frequency ωz (which is associated to the zeros), and the damping coefficient
ξ. Based on equation (18.27), they are defined as
s
Ro + R` C (Ro (R` + Rc ) + Rc R` ) + L 1
ωo = ξ= ωz =
LC (Ro + Rc ) 2ωo (LC (Ro + Rc )) CRc
(18.28)
Figure 18.5 (a) shows the Bode plot of Gp (s) considering the numerical
values reported in Table 18.2. The transfer function presents a fixed gain
(about 23 dB) and a almost zero phase shift at frequency below ωo ≈
3431 rad/s. Beyond the resonance frequency ωo , |Gp (s)| is decreasing with
a −40 dB/dec slope, whereas the phase approaches −180°. Beyond the anti-
resonance frequency ωz ≈ 151.521 × 103 rad/s, the absolute value decreases
with a −20 dB/dec slope and the phase starts rising toward −90°. The damp-
ing coefficient is ξ ≈ 0.249, keeping the resonance peak limited below 30 dB.
From equation (18.28) it is evident how important R` and Rc are to increase
the damping factor.
As shown in Figure 18.4, a PI controller R(s) = kp + ki /s is used to close
the voltage loop. From now on the hat m e is omitted. The open-loop transfer
function is defined as:
vo (t) 1 + sTi ω2 s + ωz
L(s) = = R(s)Gp (s) = kp VDC o 2 (18.29)
e(t) sTi ωz s + 2ξωo s + ωo2
Since Gp (s) is a second order transfer function, a simple pole-zero cancellation
is not straightforward in this case. The explanation and analysis of the tuning
techniques for this kind of systems are out of the scope of this book. Never-
theless, a first set of tuning parameters can be found using the MATLAB®
General State-Space Average Modeling Method 305

(a) (b)

Figure 18.5 Bode diagrams of (a) Gp (s) and (b) comparison between Gp (s),
L(s) and F (s) transfer functions for the designed controller. The amplitude
of the overshoot strongly depends on the damping coefficient ξ.

interface pidTuner(), as shown in Chapter 5. In particular, pidTuner() al-

lows to work with a user friendly interface in which several step responses
generated with different values of kp and ki can be compared. Based on the
Bode plot shown in Figure 18.5(a), the uncontrolled system is characterized
by the angular frequency ωg ≈ 1.33 × 104 rad/s. Due to the 2nd order de-
nominator, the system produces an overshoot. Hence, the settling time is
Ta,G > 5/ωg = 376 µs. For a better visualization of the result, the designed
controller is slowing down the natural response of the system by choosing
Ta,F = 23.9 × Ta,G = 9 ms.5 This value implies τF = Ta,F /5 = 1.8 ms and a
controller bandwidth equal to ωc = 1/τF = 555.56 rad/s. By setting this last
value and a phase margin of φm = 90°, pidTuner() returns:

cos(φm − π − arg(Gp (jωc )))

kp = = 0.0047 V−1 (18.30)
|Gp (jωc )|
sin(φm − π − arg(Gp (jωc )))
ki = −ωc = 33.35 V−1 s−1 (18.31)
|Gp (jωc )|

where the units of the gains are defined to be consistent with L(s). The re-
sulting bode diagrams are shown in Figure 18.5(b).
The control variable is u(t) = d(t). However, the dc voltage VDC = 15 V
provided by the power supply is considered an upper bound. This is set as
regulator saturation by limiting the control variable u(t) in a range [0 1].

5 It is recommended to multiply the time constant T , G with a large coefficient. Indeed,

a
the estimation of Ta,G is performed from the ωg computed based on the bode plot of Gp (s).
Hence, it does not consider the magnitude of resonance. Depending on the energy introduced
in the system, the resonance may increase, implying a larger settling time than expected.
This effect is investigated through open-loop tests.
306 Voltage Control of an RLC load

Therefore, an anti-windup strategy must be included in the PID controller

block. Based on Chapter 6, a back-calculation approach is used with kaw =
ki /kp = 7096 s−1 . It can be noted that the intrinsic nature of a step-down
DC-DC converter is to achieve a steady-state output voltage much lower than
the input, which implies D  1. However, d(t) could suddenly increase during
transients to compensate for a voltage variation. Hence, this explains the need
for bounding the control variable.
The discrete-time PI controller R(z) is derived based on the continuous-
time counterpart and it is integrated in the closed-loop scheme together
with a Rate Transition block in simulation to mimic the discretization pro-
cess. The sampling time of the controller is Ts = Tsw = 1/fsw , where the
switching frequency is fsw = 10 kHz. Backward Euler is chosen as integration
method.

18.4 System Simulations

The closed-loop control of the RLC load is simulated in Simulink® to analyze
its design and check the resulting performance of the controller. This step is
carried out the Half-Bridge configuration of the converter. To this purpose,
the power electronics and the RLC load are modeled through Simscape™
Electrical elements, thanks to the Specialized Power Systems ™ library (Sim-
scape/Electrical/Specialized Power Systems/ ) as shown in Chapters 6 and 17.
In particular, the Mosfet blocks available in the subset /Fundamental
Blocks/Power Electronics are used to represent the switches mounted on
the BOOSTXL-DRV8301 converter board. The semiconductor characteris-
tics taken from the converter datasheet, e.g., Ron and Rd , are set into the
Simulink® blocks as shown in Chapters 15 and 17. The two-level Half-Bridge
configuration is obtained by rearranging these block based on the equivalent
circuits reported in Figure 18.2. The switches are driven by a PWM modula-
tor with a carrier frequency fsw = 10 kHz. The load is modeled through three
series RLC branch block set as follows: RL branch type with L = 860 µH and
R` = 150 mΩ, RC branch type with C = 100 µF and Rc = 66 mΩ, and R branch
type with Ro = 6.8 Ω. Moreover, a second R branch type with Ro = 6.8 Ω is
added in parallel to the latter component. A switch manages the parallel in-
sertion of the two Ro to check the dynamical response of the system to load
variations.

18.5 Half-Bridge Configuration

As shown in Chapter 17, a Half-Bridge configuration refers to a converter
which uses a single-leg of the boosterpack, requiring one ePWM module only,
Half-Bridge Configuration 307

SOC
reload
dmax
PWM
u s1A
y*
reference Z-1
current controller u modulator s1B
(syncronized with PWM)
vo debug SOC
y

computational delay
s1A enable

D
>= tR
S il
io
i circuit
+ i
- + +
- breaker

+
s1B

-
i
ic
g

+
+ +
v v v vo
- -

+ vc
v
S

Continuous

Figure 18.6 Simulink® scheme to simulate the closed-loop voltage control of

an RLC load driven in Half-Bridge configuration.

e.g., ePWM1. Hence, the switch 1A is driven by the signal S1 (k) and 1B
by S2 (k), operating in a complementary way. The RLC load is connected
between the central point of leg A and ground. The average output voltage of
the converter leg is function of the duty cycle d(k) (or dA (k)):

V (k) = vavg (k) = d(k)VDC (18.32)

where the variables are already considered in the discrete-time domain k, i.e.,
moving the simulation toward the implementation. The voltage VDC is set
by the power supply, while d (k) ∈ [0, 1]. Considering the PWM stage, the
duty cycle d(k) is compared to a triangular carrier at fc = 10 kHz, which
generates a switching pattern S(k). This is realized on the MCU board by
using ePWM modules. Since the gate signals of the 2-level topology are S(k) =
S1 (k) and S2 (k) = not(S1 (k)), the instantaneous output voltage v(k) is a
pulsed waveform that can be computes as:

v(k) = va (k) = S(k)VDC (18.33)

which is discontinuous and bounded within 0 and VDC . This result allows to
analyze the different states in the state-space average model.
As already mentioned, the system is characterized by the parameters re-
ported in Tables 18.1 and 18.2. The control coefficients are defined in the
previous Section. These parameters can be initialized in Model Properites/-
Callbacks/InitFcn or in a separate m-file as shown here in the following:
308 Voltage Control of an RLC load

%% power supply
Vdc = 15;
%% carrier frequency definition
fsw = 10e3;
Tsw = 1/fsw;
%% sampling time definition
Ts = Tsw;
%% step-size definition
Tsim = Tsw/400;
%% load parameters
Ro = 6.8;
L = 860e-6;
Rl = 150e-3;
C = 100e-6;
Rc = 66e-3;
imax = 2;
%% controller (pidTuner)
%% wc = 555 - phiM=90
kp = 0.0047;
ki = 33.35;

The equivalent circuit shown in Figure 18.2 is then created in Simulink® as

reported in Figure 18.6. This scheme is included in the file:

SDsimRLC_closed.slx (solver: fixed step - ODE4, step size: Tsim )

SDsimRLC_closed.slx (powergui: simulation type Continuous)

The simulation time step Tsim is particularly small to well approximate the
switching behavior of currents iL (k), ic (k), io (k) and voltages v(k), vc (t),
vo (k). The reference subsystem includes different kinds of input signals to
operate the system in several working conditions.
The time constants τG and Ta,G were estimated by analyzing the bode
diagram of Gp (s). Nevertheless, it is recommended to analyze the open-loop
behavior of the system in simulation to verify both Ta,G and the resonance ef-
fects on the variables. A constant duty cycle d(k) = D = 0.4 is directly applied
avoiding any control, as shown in Figure 18.6. Based on the previous steady-
state analysis, the output voltage is computed as vo (k) = Vo = DVDC = 6 V.
The resulting open-loop dynamics are reported in Figure 18.7. Both vo (k)
and iL (k) shows oscillating responses influenced by the forcing action pro-
vided by D 6= 0 and the resonance excitation at ωo . The peaks are attenuated
thanks due to the damping ξ, as reported in equation (18.28). In particular,
the steady state value Vo differs from the predicted one due to the voltage
drop determined across the inductor, which is due to R` . Indeed, reminding
Half-Bridge Configuration 309

Figure 18.7 Open-loop dynamics considering a load variation at tR = 8 ms.

that this is a open loop test, the duty cycle d(k) cannot be dynamically ad-
justed to compensate such voltage error, leading to Vo,open ≈ 5.718 V < 6 V.
As a consequence, at steady-state, Io,open ≈ 865 mA < Io = 882.4 mA. Even
the current ripple ∆ iL,open slightly differs from the predicted one too. The
output current io (k) presents a filtered dynamic compared to iL (k). However,
the latter has a fast dynamic which exceed the limit imax = 2 A. Similarly,
the output voltage overshoot Sv% can be analytically identified through the
coefficients reported in equation (18.28):
 
− √ πξ 2
 
Sv%
Sv% = 100e 1−ξ
≈ 44.63% → vo,pk = Vo 1 + ≈ 8.68 V
100
(18.34)
The peak value vo,pk is in-line with the results provided in Figure 18.7. The
settling time can be estimated as:
5
Ta,G,open = = 6 ms > Ta,G (18.35)
ξωo
which is larger than the one computed through the bode plot. As already
detailed, this behavior is due to the energy-related effects of the resonance,
which influences the time decay of the voltage waveform.
Moreover, the scheme shown in Figure 18.6 allows to test load variations.
At time instances tR (which is a parameter to be initialized), the circuit
breaker subsystem connects the second resistor Ro , which lead to an equiv-
alent resistance Req = Ro /2 = 3.4 Ω < Ro . Setting tR = 8 ms, Figure 18.7
shows how the system reacts to such perturbation. Since Req < Ro , both iL (k)
and io (k) are higher than before. Based on Section 18.2, it follows:
Vo
Io(Req ) = = 1.76 A > Io,open
(Req )
≈ 1.69 A > Io,open (18.36)
Req
310 Voltage Control of an RLC load

Table 18.4 Steady-state values of the currents and voltages in open loop sim-
ulations.

(R ) (R )
Vo,open Io,open ∆ iL,open ∆ vo,open eq
Vo,open eq
Io,open
5.87 V 865 mA ≈ 420 mA ≈ 52 mV 5.75 V 1.69 A

(R )
Consequently, the new steady state value of Vo,open eq
is lower than the one
before load variation due to the increase in the load current, which, in turns,
causes a larger voltage drop in the inductor and in the load resistance. The
steady-state values of the open-loop dynamics are summarized in Table 18.4.
The aim of this exercise is to design a voltage-mode control loop. Namely,
this means to control and stabilize the output voltage vo (k) measured across
the load Ro (which is effectively the output variable y(k)) based on a specific
reference vo∗ (k) (lower than VDC ) by using a PI controller while achieving a
settling time Ta,F = 9 ms (see Section 18.3.1). To check these performances,
a step-wise voltage reference vo∗ (k) jumping from 0 up to 6 V (i.e., Yss = 6) at
0.5 ms is applied.
Figure 18.8 shows the output voltage y(k) = vo (k) (blue line), which is
correctly following the step-wise reference y ∗ (k) = vo∗ (k) (red dotted line).
The current flowing in the inductor iL (k) is subjected to a more oscillating
dynamic than vo (k), but always below the upper bound imax = 2 A. The
resulting waveforms of vo (k) and iL (k) follow an exponential behavior with
oscillations superimposed on them, which is a typical case in second order
systems.
Based on the design guidelines reported in Chapter 5,  ≈ 1% is considered
as reference accuracy. Thus, the transient is completed once the voltage en-
ters in the region [(1 − 0.01)Yss (1 + 0.01)Yss ], where Ta ≈ 5τ holds. Indeed,
it results that Ta,F = 9 ms, so the closed-loop control system is behaving as
expected. The steady-state value Yss can be verified by applying the concepts
reported in Sections 18.2 and 18.3. Furthermore, the closed-loop dynamics
lead to a voltage ripple ∆vo ≈ 56 mV and current ripple ∆iL ≈ 423 mA,
which are consistent with the predicted values (see Section 18.2 to check the
requirements). The effects of the step variation in the reference are reflected
into the switching patterns of 1A and 1B , which are aimed to make the av-
erage voltage vavg (k) increase by enlarging the pulse width of v(k). This is
evident from Figure 18.9, which shows a portion of the voltage transient to
relate current and voltage dynamics to the switch positions and modulation
behavior. The command-tracking performances are achieved thanks to the
regulation of u(k) = d(k) which effectively approximates vavg (k). In particu-
lar, the load current io (k) is filtered thanks to the capacitor, which absorbs all
the high-frequency components. Indeed, ic (k) shows the same ripple of iL (k).
Note that the control variable u(k) (i.e., duty cycle dynamic) is indirectly
bounded by the energy flow capabilities of a power supply, which is a static
non-linearity. For this case study, a unidirectional DC source operating within
 Half-Bridge Configuration 311

Figure 18.8 Closed-loop dynamics considering a step in the voltage reference.

[0 15 V] is considered. Thus, referring to d(k), umin = 0 and umax = 1 holds.

An anti-windup PI controller based on the back-calculation method is consid-
ered, with the back-calculation coefficient Kb equal to kaw = 1/Ti = ki /kp .
As reported in Section 6.2, the anti-windup parameters should be defined in-
side the PID block as done for the RL load in Chapter 16. It can be noted
that, due to the voltage-mode control principle, there is no need to scale the
controller output by 1/VDC , since it is a duty cycle already.
Besides the verification of the settling time of a closed-loop system, the
choice of the sampling time to avoid issues related to peripheral and task
scheduling, e.g., controller calculation and actuation updates, is investigated,
as discussed in Section 13.10 and 17.4). To this purpose, the elements in-
volved in the closed-loop control scheme loaded into the MCU, i.e., volt-
age reading (ADC module), 2L-PWM acutation (ePWM module) and a PI-
based controller, are scheduled to be executed within the sampling period
Ts = Tsw = 1/fsw . Therefore, the target hardware is subjected to the follow-
ing sequential events execution:
• At the beginning of Ts , when the value of the PWM counter equals the PWM
counter period (TBCR=TBPRD, that simply equals 1 in Simulink® simu-
lations), the ePWM peripheral is center-aligned (up-down counting mode)
and it triggers the start-of-conversion (SOC) event for the ADC module;
 312 Voltage Control of an RLC load

Figure 18.9 Detailed view of the closed-loop dynamics, which relates the cur-
rent/voltage dynamics to the switch positions and modulation behavior.

• The ADC module converts the sampled analog signal into digital counts and
triggers the end-of-conversion (EOC) event;
• The EOC triggers the ADC interrupt for the controller, which reads the
output voltage value (sample);
• Given the PI structure, the controller processes the voltage error to compute
 Half-Bridge Configuration 313

the modulating signal for the ePWM module. The computation interval
required to obtain d(k) is expected to be sufficiently lower than Ts ;

• The resulting duty cycle is held until the end of the sampling interval and it
is sent to the adopted ePWM module. As already explained, the actuation
happens at the reload of the ePWM peripheral.
Therefore, the controller adopted in the Simulink® simulation is consistent
with such event sequence. Indeed, the scheme reported in Figure 18.6 use the
same approach adopted for the current control of an RL load (Section 17.4).
In particular, a trigger signal SOC (i.e., a boolean square-wave) is gener-
ated inside the modulator subsystem every time the carrier equals 1. The SOC
is used as trigger source for a triggered subsystem which realizes a voltage
controller synchronized with the carrier peaks. A voltage reference vo (k) in-
ternally generated by a Pulse Generator block is considered (sample based,
with gain A = 6, number of samples ns = 500, pulse width 50 %, phase delay
= 5, and sample time Ts ). This is a square-wave of period 500Ts = 25 ms. The
Pulse Generator block is synchronous with the zero value of the triangular
carrier. Since the ADC SOC trigger is generated at the carrier peak, vo∗ (k) is
delayed by Ts /2 to provide a correct voltage error at the controller input.6 The
resulting reference voltage is called vo,ctrl
∗
(k), where the subscript ctrl refers
to the quantity that are processed by the controller. Hence, the processed
voltage is called vo,ctrl (k). Even if the controller computations return a value
for d(k) before the end of Ts = Tsw , the controller has to wait the next reload
PWM action to feed d(k) to the ePWM input (this happens when CTR=zero).
Therefore, another Delay block (sample time Ts /2 and data length 1) is in-
cluded at the controller output to synchronize the actuation with the zero
value of the carrier. The whole event sequence is denoted in Figure 18.9.
As did for the open-loop test, the closed-loop dynamics are also evaluated
in presence of load variations. Figure 18.10 shows how the system reacts to
such perturbation occurring at tR = 8 ms.
Immediately after the operation of the switch, the inductor current iL (k)
cannot instantaneously increase since it is a state variable. Thus, the load
current io (k) increase (due to the lower resistance) is supplied by the capacitor,
which starts to discharge. Then, iL (k) starts to increase too. The capacitor
goes on discharging until the new steady-state value for io (k) is reached. This
current compensation is evident in Figure 18.10. In addition, it can be noted
that the output voltage quickly stabilizes around vo (k) = vo∗ (k) after the
transient.
Considering the achieved aforementioned performances, the designed PI
controller meets all the specifications.
For the sake of completeness, the close-loop dynamics can be compared
with those provided by a faster controller. Therefore, Ta,F = 3.6 ms which

6 This delay can be included through a Delay block, where the sample time is set equal

to Ts /2 and the delay length is 1.

314 Voltage Control of an RLC load

Figure 18.10 Closed-loop dynamics considering a load variation at tR = 8 ms.

implies τF = Ta,F /5 = 714 µs and a controller bandwidth equal to ωc =

1/τF = 1400 rad/s are considered. Moreover, by setting a phase margin φm =
90°, pidTuner() returns kp = 0.0133 V−1 and ki = 79.68 (Vs)−1 . Keeping the
other settings as before, the resulting plots are reported in Figure 18.11.
To achieve a faster voltage response vo,f (k), the controller feeds the system
with higher energy (i.e., df (k) > d(k)) exciting the resonance at ωo more
than before and enlarging the oscillating effects. Likewise, the dynamics of
iL,f (k) and io,f (k) may reach more stressful values, even if the current ripple
is comparable to the previous case.

18.5.1 Control implementation

The closed-loop dynamics are tested on a test-bench which includes a TI
BOOSTXL-DRV8301 converter mounted on the TI LaunchXL F28069M
Piccolo™ board and connected to the extRL(C) board in RLC mode (see
Figure 18.1). The extPot3 board is also used to manipulate analog inputs.
Since only leg (phase) A is used for this implementation, the connec-
tion scheme is reported in Figure 18.12 accordingly. Therefore, one ePMW
module is used only and it is operated in up-down counting mode at
fsw = PWMfrequency = 10 kHz. Reminding that for the F28069M board
Half-Bridge Configuration 315

Figure 18.11 Comparison of the closed-loop dynamics for two different control
design. The fastest one is denoted by subscript f .

fck = CPUfrequency = 90 MHz, it follows:

CPUfrequency
TBPRD = PWMcounter−period = = 4500 (18.37)
2 · PWMfrequency

where the sampling time is Ts = Tsw = 1/PWMfrequency .

The reference voltage can be internally generated in the firmware or exter-
nally provided both using the available potentiometers on the extPot3 board
or through COM port (e.g., SCIA). The analog solution is adopted to add
versatility to the control scheme in the proposed solution. The needed param-
eters can be initialized in Model Properites/Callbacks/InitFcn or in a separate
m-file as done for simulation.
A step-by-step procedure for programming the microcontroller for this task
is reported here in the following.

Firmware Environment
c28069_RLCcontrolSD_F.slx (solver: fixed step - ODE4, step size: Ts =
Tsw /2 = 50 µs)

Open a new blank Simulink® project and configure the environment as shown
 316 Voltage Control of an RLC load

Vo(mcu)
GND
extRL(C)
used in RLC mode

R=6.8 C=100 µF L=860 µH
GND
VCC
legA

enable
(ADCB3)
DC link

Vo(mcu)

Figure 18.12 Connections of the extRL(C) board with the BOOSTXL-

DRV8301 converter realizing a step-down DC-DC converter configuration.

in Section 9. Remember that the BOOSTXL-DRV8301 is enabled by settling

GPIO50 high (see Section 3.2).
• Insert an ePWM block and use the same settings shown in Section 13.5:
– Select ePWM1 module with TBPRD corresponding to fc = 10 kHz,
i.e., PWMcounter−period = 4500;
– Remember to unflag Enable ePWM1B and tick Inverted version of
ePWMxA. This option allows to drive 1B as the complement of 1A ;
– Keep all the other settings as reported in Section 13.5.
• Add a gain block to normalize the modulation signal over the peak value of
the carrier signal PWMcounter−period . Include a data type conversion set
on int16 to represent the duty cycle within a 16-bit base.
For up-down counting mode, the measurement of an average current iL (k) or
voltage vo (k) is feasible thanks to the ePWM module which sends a synchro-
nization signal that triggers the SOC of an ADC module when the carrier
equals TBCTR=TBPRD, as shown previously. To this aim, the following set-
tings are summarized here below:
• In the Event Trigger (Tab) of ePWM1
– Select Enable ADC start of conversion for module A;
 Half-Bridge Configuration 317

– Set Number of event for SOCA to be generated to First event;

– Select Counter equals to period (CTR=PRD) as Start of conversion
for module A event selection.
Every time the carrier signal reaches the maximum value TBPRD, a SOC
input (i.e., ePWM1_ADCSOCA) is sent to the ADC. This last block has to
be configured as well. The extRL(C) board (see Appendix B) has an external
voltage sensor, which has to be connected to one of the BOOSTXL-DRV8301
ADC channels.
• Add an ADC block and select ADCINA5 as Conversion channel in the
Input Channels tab. Then, go back in the SOC Trigger tab;
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC2;
– SOCx acquisition window: 7;
– SOCx trigger source: ePWM1_ADCSOCA;
– ADC will trigger SOCx: No ADCINT;
– Sample time: Ts ;
– Data type: uint16.
• Add a Data Type Conversion block set to int16 (since positive signals
only are used) or single to exploit 32 bit.
• Place a Gain block for scaling the digital signal over [0, 1], since a 12-bit
reading returns values ranging from 0 (0 V) up to 212 − 1 = 4095 (3.3 V).
• Build a proper conversion chain for the voltage sensing by removing the
estimated offset as well as using a scaling factor gv,adc to obtain Volt values
after moving to single data type.
The external voltage sensor is essentially a voltage divider which scales vo (k)
within a [0, 3.3] V range (suitable for MCU inputs). Since the RLC board
is designed to sustain higher voltage than the one used in this exercise, the
voltage divider uses Rhigh = 15 kΩ and Rlow = 1 kΩ to achieve:

Rlo
vo,mcu (k) = vo (k) if vo (k) = 6 V → vo,mcu (k) ≈ 375 mV
Rup + Rlo
(18.38)
Similarly, 15V → vo,mcu (k) ≈ 937.5 mV, thus, a limited portion of the
[0, 3.3] V voltage range is used. This implies that the voltage conversion
chain is characterized by the following gain:
3.3 Rup + Rlo
gv,adc = (18.39)
4095 Rlo
As done for the low-side shunt (Chapter 16), the voltage sensor should be
 318 Voltage Control of an RLC load
[ref]

A
Z-1
dmax C2802x/03x/05x/06x
pulses
PI(z) u
[vo] WA
y
en ePWM

C2806x

>= 0.5 [ref] GPIOx

enable
C2802x/03x/05x/06x GPIO DO
GPIO7 - pin79
B3

ADC C28x
C2802x/03x/05x/06x [i] [ref] Data

A0 carA SCI XMT

ADC
shunt A
C2802x/03x/05x/06x offset exp

A5
[vo]
ADC
ADCA5 pin65

offset exp

Figure 18.13 Simulink® scheme included in c28069_RLCcontrolSD_F.slx.

characterized from an experimental point of view to include the variations

given by the resistors tolerance (e.g., the setup provides some extra wiring
connections other than extra contact resistances) as well as the estimation
of the experimental offset (e.g., 40), as shown in Figure 18.13.
• Use the same discrete PID controller block adopted in simulation. Set the
controller type as PI (Parallel) with sample time Ts ; set P=kp and I=ki .
Select an external source for the reset with rising condition. Saturate the
controller output to keep the duty cycle in the range [0, 1]. Set a back-
calculation method as anti-windup strategy with Kb=kaw .
• A redundant saturation block can be used to protect the system by limiting
d(k) besides the limits provided by the PID block.
• Add a Pulse Generator (sample based) block to create the reference vo∗ (k).
Set the amplitude equal to A = 6, number of samples ns = 500, pulse width
50 %, phase delay 5, and sample time Ts . It is equivalent to have vo∗ (k)
generated by a square-wave of period 500Ts = 25 ms. It can be noted that
vo∗ (k) is synchronized with the zero value of the triangular carrier.
• Add a Delay block to be consistent with the ADC SOC trigger generated
at the carrier peak. Set sample time Ts /2 and delay length 1.

Arrange these aforementioned blocks as shown in Figure 18.13.

In order to manually manage the activation of the PI controller, the exter-
nal reset (rising) is connected to a potentiometer (ADCINB3/pin68). Without
moving it, the control algorithm is forced not to provide any actuation signal.
To do that:
 Half-Bridge Configuration 319

• Add an ADC block and select ADCINB3 as Conversion channel in the

Input Channels tab. Then, go back in the SOC Trigger tab:

– Sampling mode: Single sampling mode;

– SOC trigger number: SOC1;
– SOCx acqusition window: 7;
– SOCx trigger source: Software;
– ADC will trigger SOCx: No ADCINT;
– Sample time: Tsig = 0.001 s;
– Data type: uint16.
• Add a Data Type Conversion block set to int16 (since positive signal
only are used) and the 12-bit scaling to move from a range (0 V) up to
212 − 1 = 4095 (3.3 V) to [0, 1].
• Add a Compare to Constant block which is used to evaluate if signals
are greater or equal than a Constant value 0.5. If this condition is true, the
output switches from zero to one resetting the integrator of the PI controller.
The enable signal is defined as en(k).
• Additionally, en(k) can be used to add a further protection redundancy.
Indeed, by using a Product block, the enable signal is multiplied by the
actuation variable. As long as en(k) = 0, the duty cycle is forced to stay
equal to zero d(k) = 0, i.e., no switching is performed.
• To visually evaluate the pulsed reference vo∗ (k), the Pulse Generator can be
connected to a GPIO block (e.g., GPIO7) with a proper scaling of 1/6 to
normalize the signal in a [0, 1] range. This processing allows to visualize the
reference signal ranging in [0, 3.3] V at pin 79 through an oscilloscope.
Optional: the firmware could be also prepared to visualize the duty cycle
d(k) through a DAC peripheral. In this case:

• Copy the ePWM block, select ePWM8 (which is connected to DAC3 in J8,
pin 72), keep the other settings. Connect this block in parallel to the other
ePWM (i.e., connect d(k) just after the Data type conversion block). Hence,
d(k) can be observed by connecting a voltage probe to pin 72.
Figure 18.14 reports the closed-loop behavior of the system obtained by exe-
cuting the firmware and observing the current and voltages of interest through
an oscilloscope. In particular, voltages vo (k) and vo∗ (k) are measured by using
voltage probes (with vo∗ (k) scaled on GPIO7), while the currents iL (k) and
io (k) are measured with current probes. In case DAC3 is used to check d(k),
another voltage probe connected to pin 72 is needed.
As done in simulation, the analysis of Figure 18.14 starts from the evalu-
ation of the command tracking performances. The resulting vo (k) (blue line)
320 Voltage Control of an RLC load

Figure 18.14 Closed-loop dynamics measured from the F28069M board pins
through an oscilloscope.

is following the internally generated vo∗ (k) (red line), which is re-scaled from
3.3 V on the actual reference value in Volt. The settling time is similar to the
expected one Ta,F = 9 ms even if both voltage and current dynamics looks
smoother (i.e., reduced oscillating responses) during transient. Besides the
probe accuracy, this effect is mainly due to the parasitic resistances RL and
Rc , which differ from the theoretical ones. In particular, Rc is a great mat-
ter of concern since even a slightly increase has a significant impact on the
damping coefficient ξ, while the settling time Ta,F is almost unaffected.
Consequently, the current ripple is ∆iL ≈ 382 mA, which is lower than the
value predicted in simulation. The voltage ripple is ∆vo ≈ 100 mV, which is
larger than the value predicted in simulation. However, both ∆iL and ∆vo are
compliant with the requirements. This last discrepancy could be a consequence
of a more noisy environment. The Pulse generator which provides the internal
reference vo∗ (k) is synchronized with the zero value of the triangular carrier
(TBPRD=0). Nevertheless, such synchronization is not as strict as other tasks,
e.g., counter comparison in ePWM, which is based on hard synchronization
(hard-sync) event through interrupts. Therefore, it might happen that the
rising edge of the step-wise vo∗ (k) is not perfectly synchronized with the ADC
SOC (i.e., carrier peak). By adding a fixed shift of Ts /2, vo,ctrl
∗
(k) is coincident
with ADC SOC, thus, sampling a value of vo,ctrl (k) = 6 V. This is the same
∗
Half-Bridge Configuration 321

Figure 18.15 Closed-loop dynamics with a load variation at tR = 8 ms observed

through an oscilloscope.

processing obtained for the RL load, thus, the reader is referred to the detailed
description reported in Section 17.4.1. Based on this part, the overall interval
between the the rising edge of vo∗ (k) and the start of the system dynamics7
(i.e., when vo (k) and iL (k) start to increase due to v(k) 6= 0) reaches 2Ts in
a worst case scenario. Figure 18.15 also shows the duty cycle reading through
DAC3. The signal read at pin 72 (gray dotted line) needs to be additionally
averaged to obtain d(k) (red line). This can be easily done via software by
using an average function or a digital filter.
As did in simulation, the closed-loop dynamics are also evaluated in case of
variations in the resistance value. This is performed by manually moving the
switch S3 mounted on the extRL(C) board after the system reaches a steady-
state value. The operation changes the resistance down to Req = Ro /2. Figure
18.15 shows how the system reacts to such perturbation. The current compen-
sation behaves as explained in the previous Section. In particular, it is shown
that the inductor current iL (k) does not increase instantaneously. Thus, the
variation in the load current io (k) is compensated by the capacitor through its
branch current ic (k). The d(k) signal is also visible. Both transients and am-
plitudes are consistent with simulations, validating the stability of the voltage
loop to vo (k) = vo∗ (k) even in case of load variations.

7 Delay in the converter actuation.

322 Voltage Control of an RLC load

Figure 18.16 Closed-loop voltage and current dynamics measured through an

oscilloscope. A large inductor L = 3.4 mH is considered, keeping C = 100 µF.

Remark: as already reported in Section 18.2 and 18.3, the parasitic resis-
tances RL and Rc strongly impact on the damping coefficient ξ. The value of
this parameter might be also influenced by snubbers resistors and non-zero
impedance of the power supply. Moreover, resistance of the wires and/or the
traces of PCBs may impact on ξ too, leading to higher system losses too.

Remark: the firmware can be designed to process the inductor current by

using the low-side shunt resistor of leg A. Figure 18.13 already reports such
configuration based on the settings provided in the Chapter 17. This mea-
surement can be used in the control loop to (strictly) restrict the current
dynamic below imax = 2 A, leading to a more advanced control scheme, e.g.,
a two-degree-of-freedom (2DOF) or a model predictive control-based [40, 38].

18.6 Variations of LC Filter Parameters

To evaluate the effectiveness/robustness of the controller, the extRL(C) board
allows to change all the RLC parameters, see Appendix B. The main purpose
of the designed regulator is to guarantee a stable output voltage vo (k) =
vo∗ (k) < VDC even in case of variations in the filter parameters. Previously, it
was demonstrated that the tracking of vo∗ (k) is guaranteed in presence of a
variation of R, thanks to the sudden changes of d(k), io (k), ic (k) and iL (k).
Even the LC filter parameters can be manually varied. The inductance can
be increase from L = 860 µH up to L = 4 · 860 µH, while the capacitance may
decrease from C = 100 µF to C = 10 µF. The inductors belong to the same
Variations of LC Filter Parameters 323

Figure 18.17 Closed-loop voltage and current dynamics measured through an

oscilloscope. A small capacitor C = 10 µF is considered, keeping L = 860 µH.

family (therefore, R` is the approximately same), while the capacitors shows

different Rc (see Appendix B). Since the closed-loop voltage controller is de-
signed on L = 860 µH and C = 100 µF (i.e., the reference case), a variation in
the LC filter parameters implies different settling time and closed-loop dynam-
ics. Such analysis is useful to understand which parameter variation may lead
to potential damage or instability in the system. The firmware architecture is
left unchanged.
By manually varying the inductance up to L = 3.4 mH (keeping C =
100 µF), the inductor current iL (k) is largely filtered as well as the load current
io (k) (see Figure 18.16). A higher inductance value leads to a more stable
control loop (i.e., higher stability margin) making it easier to control. This
facilitate the controller to regulate the voltage, which is subjected to a lower
distortion too. A faster dynamic is achieved as well, that is, Ta,F ≈ 5 ms.
Moreover, a four times larger inductance implies a four time larger R` , which
is R` ≈ 450 mΩ. This variation causes an increase in the damping coefficient
ξ, reducing the oscillating effects in the voltage/current responses. Moreover,
the current ripple is lower than the reference case. Conversely, by manually
changing the capacitance value down to C = 10 µF (keeping L = 860 µH), the
capacitor voltage and the output voltage vo (k) are less filtered than before, as
shown in Figure 18.17. This indirectly affects the inductor current iL (k), which
shows a more bouncing response. Nevertheless, the ESR of the new capacitor
is Rc ≈ 470 mΩ, being larger than before. Hence, even if the voltage is less
filtered, the damping coefficient is higher, reducing the oscillating effects in
both voltage and current waveforms. This is evident comparing Figure 18.17
with 18.14.
Despite both perturbations on the LC parameters, the command tracking
is still guaranteed.
19
Cascade Speed Control of a Permanent
Magnet DC Motor

After the development of control strategies for static systems such as RL

and RLC loads, this chapter shows how to realize a cascade speed control
for PMDC motors. The same setup adopted in Chapter 15 is considered for
the implementation. In particular, two different arrangements are considered,
namely involving a single motor and two PMDC machines connected in back-
to-back (B2B) configuration.
The DecMot Hardware Kit (see Appendix B and Figure 19.1) is adopted
to test a single PMDC motor. A Full-Bridge converter configuration is con-
sidered (thus, two legs of the TI BOOSTXL-DRV8301 converter are involved)
operating in unipolar voltage switching. The DecMot Hardware Kit includes a
rotary encoder1 to measure the rotor speed. This allows to design a closed-loop
control scheme based on a speed sensor. The encoder interface is connected to
the extPot3 board to provide the speed measurement to the LaunchPad™
F28069M board through the eQEP1/eQEP2 sockets. On the contrary, a
sensorless (or encoderless) closed-loop control scheme can be also designed.
Among the many observers and estimators formulations available in literature,
this chapter presents a model reference adaptive system (MRAS) observer to
estimate the rotor speed, commenting its integration/design for a closed-loop
scheme and related effects on the closed-loop dynamics. In both cases, aim-
ing to follow a given speed reference Ω ∗ (k), the nested PI controllers lead
to regulate the dc output voltage of the converter by manipulating the com-
mand signal of the ePWM peripherals of the two legs, e.g. dA (k) and dB (k).
The reference speed value can be provided by an external command, which
may come from serial communication, or by the extPot3 potentiometers, or
internally (e.g., through a pulse generator/constant block). The B2B-PMDC
Hardware Kit (see Appendix B and Figure 19.2) is adopted to test two PMDC
motors arranged in a B2B configuration. Two Half-Bridge converters are con-
sidered to drive each motor (thus, two legs of the TI BOOSTXL-DRV8301
converter are adopted even this time), allowing one direction of rotation for
the shaft only. The PMDC motors are mechanically coupled, with one of them
behaving as a traction motor, while the other one as a braking one. Such con-
figuration is widely adopted when it is required to characterize or validate
1 This device is connected to the PMDC motor through a mechanical coupling.

DOI: 10.1201/9781003196938-19 325

326 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.1 DecMot Hardware Kit rearranged for implementations with one
PMDC motor only connected to a DC-DC converter (see connection scheme
in Figure 19.15). The reader is referenced to Appendix B for further details
on this setup.

the operating region/plane of an electrical motor. Indeed, the two motors

can be separately controlled. Assuming that one PMDC motor is tracking a
speed reference (full cascade structure) while the other is aimed to regulate
a friction/braking torque applied to the shaft (i.e., armature current control
only), the controlled B2B system is equivalent to a cascade speed control of a
PMDC (traction) motor subject to a load torque that can dynamically vary in
time. For instances, this allow to fix the speed while moving along the torque
axis of the operating region. Sensored (or encoderless) control scheme can be
designed as well.
A single motor control behaviur is simulated and analyzed first through
Simulink® simulations, being useful even to describe the B2B test case.
Then, different firmwares are designed and implemented on the LaunchPad™
F28069M board based on these results. It must be noted that no flux weak-
ening is needed for such motors, since the excitation circuit is replaced by
PMs. Moreover, due to the limited size of the PMDC motor, the presence or
absence of a back-emf compensation do not particularly impact on the system
dynamic. Hence, it is up to the reader to activate (or not) such compensation
terms in the designed algorithms.
 Required Hardware 327

Figure 19.2 B2B-PMDC Hardware Kit allowing implementations involving

two PMDC motors arranged in back-to-back configuration (see connection
scheme in Figure 19.23). The reader is referenced to Appendix B for further
details.

19.1 Required Hardware

• TI LaunchPad™ Piccolo™ F28069M and BOOSTXL-DRV8301 boards;
• extPot3 custom board (see Appendix B for technical details);
• DecMot Hardware Kit for the control of one PMDC motor (see Appendix
B and Figure 19.1);
• B2B-PMDC Hardware Kit for the control of two PMDC motors coupled in
back-to-back configuration (see Appendix B and Figure 19.2);
• Power supply set on 10 V and limited to 10 A (i.e. 100 W ).

Basically, the setups used in this chapter adopt the same hardware described in
Chapter 15. The same recommendations shown there are still valid. Namely,
jumpers JP1 and JP2 should be removed to isolate the USB/JTAG port.
Moreover, keep in mind that the BOOSTXL-DRV8301 converter requires a
dc-link voltage VDC > 6 V to properly work, otherwise the nFAULT is returned
(regardless of the state of the jumpers JP1 and JP2). The reader is referred
to the converter details reported in Section 3.2.
In the proposed control schemes, the low-side shunt-resistors are used as
current sensor to close the inner current loop (their characterization explained
in Chapter 16), whereas the rotary encoder provides the feedback for the
speed loop (see Chapter 14 for all details on this device and on the eQEP
peripherals).
328 Cascade Speed Control of a Permanent Magnet DC Motor

For the single motor configuration, the LaunchPad™ F28069M board is

connected to the BOOSTXL-DRV8301 converter and it is programmed to
follow an internally generated square-wave reference speed. This holds also
for the B2B configuration, which also includes an additional current loop to
manage the dynamic braking acting on the shaft. The application of load
torque is related to the rotation of a linear potentiometer on the extPot3
board (ADCB5). As done in Chapter 17 and 18, another potentiometer is
used as control enable (ADCA3). The reading of the corresponding ADC
channels must be suitably re-scaled. Debugging signals can be sent out of the
MCU through SCIA/B (virtual COM port). The DC link voltage is again
provided by a power supply set to VDC = 10 V, with the output current
limited to 10 A. The latter value is higher than in Chapter 15. Indeed, for the
B2B configuration, the application of a braking torque on the shaft leads to a
current rise, which has to be sustained by the power supply while fulfilling the
motor and converter data-sheets (e.g., the rated power of the PMDC motor
is 100 W).
In the following, the analytical model of the motor control system is ana-
lyzed and a cascade control scheme is designed based on it. Then, some simu-
lations of the controlled system are reported and considered as reference case
for the implementation on hardware. Finally, experimental tests are carried
out once the simulations returns satisfactory results.

19.2 Linear Model of a PMDC Motor

The model of a PMDC motor was already presented in Section 15.2. The main
equations and transfer functions characterizing such machine are reviewed
here in the following to introduce the adopted control strategies. The reader
is referred to Section 15.2 for all the details on model derivation.
Considering the equivalent circuit reported in Figure 15.1, the equations
characterizing the electrical and mechanical behavior of the system are re-
ported in (15.2) and (15.3), that are:
dia (t)
La =va (t) − Ra ia (t) − e (t)
dt
dΩ(t)
J =me (t) − ml (t) − βΩ(t)
dt
As previously mentioned, the magnetic field and the related flux flowing
inside the machine is constant, being generated through permanent magnets.
PM . As a matter of fact, a constant K is defined as K = ψPM kT = ψPM ke .
2

2 No flux weakening can be performed. Thus, the motor speed can not reach values above

the rated speed (also called base speed).

Linear Model of a PMDC Motor 329

Consequently, the electromagnetic torque and back-electromotive force can be

computed as reported in equations (15.4) and (15.5), that are:

me (t) = Kia (t)

e(t) = KΩ(t)

For the single motor configuration, the load torque refers to external causes
such as additional friction or mechanical load coupled to the PMDC motor.
If the configuration does not foresee the application of any load torque, then
ml (t) = 0. For the B2B configuration, the load torque is separately managed.
Therefore, equation (15.4) is subject to the current control loop implemented
for the braking motor. Considering this latter connected to the leg B of the
converter, it follows
ml (t) = KiB (t) (19.1)
where iB (t) is the current flowing in the armature circuit of the braking motor,
whereas ml (t) is the variable load torque.
The state-space representation is derived considering u(t) = [va (t) ml (t)]T ,
x(t) = [Ω(t) ia (t)]T and y(t) = x(t). Therefore, the resulting equations in
matrix notation are those reported in (15.9) and (15.10), that are:
d
      
Ω(t) −β/J K/J Ω(t) −1/J
= + va (t)
dt ia (t) −K/La −Ra /La ia (t) 0
1 0
    
Ω(t) Ω(t)
=
ia (t) 0 1 ia (t)
where the state-space matrices are defined as follows:
0
   
−β/J K/J −1/J
A= B= C = I2
−K/La −Ra /La 0 −1/La

It is important to note that all the parameters of the considered motors (Ra ,
La , J, β, K) must be estimated through some identification procedures such
as the locked rotor test (see Chapter 16) and by observing the open loop
mechanical dynamics. These calculations are necessary (and they must be
performed as fist task) to follow the controller design procedures presented in
the previous chapters. Since the two system configuration are different (i.e.,
three different motors and two couple of boards and converters are involved),
the results of the identification tests will be provided in the following sections
for each case.
Finally, the transfer functions Gi (s) and GΩ (s), which refer to the electrical
and mechanical behavior of the motor, respectively, are derived based on the
motor equations in the time domain, as shown in equation (15.11):
ia (t) 1 Ω(t) 1
Gi (s) = = GΩ (s) = =
va (t) Ra + sLa me (t) β + sJ
The PMDC motor in single configuration and the traction motor in B2B
330 Cascade Speed Control of a Permanent Magnet DC Motor
Ω ∗ (t)
+ speed m∗e (t) 1 i∗a (t)
+ current u∗a (t)
− controller K + va∗ (t) power PMDC
− controller
− stage motor
e(t)
2L-VSC +
modulator
K
ia (t)
Ω(t)

Figure 19.3 Block scheme for a cascade speed control of a PMDC motor with
back-emf compensation. This strategy is valid both for Half- or Full-Bridge
configurations.

configuration are controlled thanks to a current and a speed control loop

which are designed based on Gi (s) and GΩ (s), respectively. Moreover, the
braking motor in B2B configuration is controlled by a current controller only,
which is still design over Gi (s). Further details on the PI-controllers design
are reported in the following sections.

19.3 Cascade Control Architecture and Design

The speed regulation of a PMDC motor is achieved by using a so called cas-
caded control architecture, which is based on two nested loop based on a speed
and current controller, respectively. The goal of this scheme is to assure a ref-
erence speed tracking for Ω ∗ (t) by manipulating the voltage applied to the
motor terminals through the power converter. The signal Ω ∗ (t) can be spec-
ified by changing the position of a potentiometer, by internally generating it
or by sending this input via virtual COM port.
Based on to Figure 15.3, which shows the block scheme representation of
the cascade speed control for PMDC motors, the back-emf e(t) is the link
between the electrical and the mechanical contributions. From a theoretical
point of view, a cascade control architecture can be only adopted if the nested
loops can be effectively decoupled. In particular, the external loop is required
to regulate the shaft speed, as shown in Figure 19.3. The speed controller
outputs a reference torque m∗e (t), which is translated in a reference armature
current i∗a (t) by using the scaling factor 1/K. Then, the inner current controller
outputs a reference voltage u∗a (t). As a matter of fact, the back-emf e(t) can be
considered as an observable disturbance and it can be compensated as shown
in Figure 19.3. According to the control theory terminology, this is called a
feed-forward action, which is computed as:

eb(t) = KΩ
b or eb(t) = K
bΩb (19.2)
Cascade Control Architecture and Design 331

where the superscript b refers to an estimated variable. Indeed, the feed-

forward action is based on an estimation of the back-emf based on the motor
parameters (i.e., it is subjected to the accuracy of the results obtained through
the identification process) and the speed feedback, which can be returned by
a sensor (e.g., an encoder) Ω(t) or by an observer (e.g., MRAS) Ω. b
This feed-forward term allows to move from ua (t) to the reference armature
∗

voltage va∗ (t), which represents the modulating signal to be processed by the
modulator stage for generating the PWM signals. This stage is included into
the power stage block (not drawn explicitly for the sake of brevity) shown in
Figure 19.3, which includes a 2L-VSC (i.e., the DC-DC converter) and the
PWM modulator.
The back-emf compensation provides an important contribution, since it
represents and additional voltage term (on condition that the dc-link is able
to provide the required power). It should be clear that the feed-forward term
achieve an effective decoupling only at steady-state, while the cross-coupling
persists during transients. Nevertheless, the importance of such compensation
term depends on the machine design. Indeed, if the PMDC motor is of small
size and low voltage rating, e(t) can be neglected (i.e., no compensation). On
the other hand, even the load torque ml (t) can be estimated and then compen-
sated. However, for the simulation analysis, ml (t) is considered a controllable
disturbance, thus, manipulating its instantaneous value. It can be noted that
the current controller is of the same kind of that one designed in Chapter 17,
i.e., it is based on an RL circuit (see Gi (s) in (15.11)). In addition, a combined
design of both speed and current control loops is needed to ensure that the
mechanical dynamic is satisfactory.
Given the feed-forward action, the system is suitable to be controlled with
a cascade scheme. This becomes quite evident if the power stage (i.e., DC-
DC converter + modulator) and the PMDC motor shown in Figure 19.3 are
substituted with the equivalent linear average representation of the motor
dynamics shown in Figure 15.3. This require to approximate va∗ (t) with va (t)
(see Section 17.3.1).
In particular, it is important to remind that this architecture has to fulfill
cascade constraints in terms of bandwidth between the nested loops, i.e., aim-
ing to decouple the electrical and mechanical dynamics. The outer (speed)
loop must be consistently slower than the inner (current) one. This means
that, from the outer loop point of view, a current reference i∗a (t) sent as input
of the current loop is immediately provided as output ia (t) = i∗a (t), i.e., allow-
ing to approximate the inner closed-loop transfer function as a unitary gain.
This requirement translates in a constraint on the bandwidths. Indeed, this
behavior is guaranteed by imposing a cut-off frequency for the inner current
controller at least ten times smaller than the outer speed one.
ωΩ ≤ ωΩ /10 or τGΩ ≥ 10τGi (19.3)
where ωi and ωΩ are the bandwidths of the current and speed controller,
respectively. If equation (19.3) is satisfied, the two controllers can be designed
332 Cascade Speed Control of a Permanent Magnet DC Motor

separately, i.e., each one is tuned based on its specific transfer function. This
means that the control action of the speed controller is not influenced by the
dynamics of the inner loop, since the current regulator is much faster than
the outer one.
The PI-based speed controller is defined as RΩ (s) = kp,Ω + ki,Ω /s and
the current one as Ri (s) = kp,i + ki,i /s, where the integrator time constants
are Ti,Ω = kp,Ω /ki,Ω and Ti,i = kp,i /ki,i , respectively. By imposing equation
(19.3), RΩ (s) would be designed based on GΩ (s) only, while Ri (s) based on
Gi (s). It is worth noting that this simplifies the controllers design. Indeed,
since both GΩ (s) and Gi (s) are 1st order transfer functions, RΩ (s) and Ri (s)
can be explicitly designed using the pole/zero cancellation (i.e., still within
linear control theory) as shown in Chapter 5. This leads to the following open-
loop transfer functions:
i(t) kp,i /Ra
Li (s) = = Ri (s)Gi (s) = (19.4)
ei (t) sTi,i
Ω(t) kp,Ω /β
LΩ (s) = = RΩ (s)GΩ (s) = (19.5)
eΩ (t) sTi,Ω
where Ti,Ω = J/β and Ti,i = La /Ra . The open loop transfer functions behave
as a pure integrators (φm ≈ 90◦ ) and their cut off frequencies ωc,Ω , ωc,i
depends explicitly on the system parameters. As shown in Chapter 5, it follows
that:

kp,Ω = ωc,Ω J kp,i = ωc,i La (19.6)

ki,Ω = ωc,Ω β ki,i = ωc,i Ra (19.7)

Then, the closed loop transfer functions are:

Ω(t) LΩ (s) i(t) Li (s)
FΩ (s) = = Fi (s) = = (19.8)
Ω (t)
∗ 1 + LΩ (s) i (t)
∗ 1 + Li (s)
where LΩ , FΩ and Li , Fi are subject to the time constants τFΩ and τFi ,
respectively.
The cascade constraints assure that the control action of the speed
controller is not influenced by the dynamics of the inner loop, which is
approximated as a unitary gain Fi (s) ≈ 1, thus, describing steady state op-
erations only. As an example, knowing that the time constant of the elec-
trical transfer function Gi (s) is τGi = La /Ra and the corresponding set-
tling time is Ta,Gi = 5τGi , a final settling time Ta,Fi can be defined as
Ta,Fi = 1/10 × Ta,Gi in order to speed up the closed-loop system response.
Therefore, since τFi = Ta,Fi /5, it follows that ωc,i = 1/τFi . The speed con-
troller bandwidth has to satisfy equation (19.3), that is ωc,Ω = ωc,i /10.
In case the cascade constraint is not respected, the open loop transfer
function LΩ (s) becomes:

LΩ (s) = RΩ (s)GΩ (s)Fi (s) (19.9)

System Simulations 333

which is an higher order transfer function compared to the previous LΩ and,

therefore, it is not suitable for pole/zero cancellation.
Reminding that the power supply is set on VDC = 10 V, the output of both
the PI controllers must be saturated. Namely, the current controller has an
upper limit equal to 10 V, whereas the speed controller must be saturated at a
certain m̃e , which depends on the specific system parameters. Therefore, the
two controllers adopt an anti-windup structure like that one shown in Chapter
6. It is important to remind that the coefficient kaw for a back-calculation
scheme is computed as kaw = ki /kp .
Moving the attention to the B2B configuration, the braking motor is driven
by a current loop only, as shown in 19.3, where the reference torque can be
provided through an ADC channel, internally generated or sent through serial
COM port. The associated current controller is still designed with a pole/zero
cancellation as described previously. In principle, both motors mounted on
the B2B-PMDC Hardware Kit are independent, but they show quite similar
rated parameters. Therefore, the current control design for the traction motor
is be used even for the braking one.
Discrete-time PI controllers RΩ (z) and Ri (z) are derived based on the
considerations reported so far for the continuous-time counterpart. They can
be integrated in simulations using Rate Transition blocks to mimic the dis-
cretization process, as shown for the previous exercises. The sampling time
of the controller is set equal to Ts = Tsw = 1/fsw , using Forward Euler as
integration method.
It is important to note that this cascade control scheme holds for both
the setups considered in this chapter, since it can be equally used for Half-
or Full-Bridge implementations (provided that the reference voltage signal is
suitably conditioned). Thus, this approach is used to control both the single
PMDC motor and the traction motor in the B2B hardware kit. Further details
are given in the following sections about the implementations into MCU.

19.4 System Simulations

The speed control of the single PMDC motor is simulated in Simulink® to
analyze its design and check the resulting performance of the controller. This
step is carried out for the Full-Bridge configuration of the DC-DC converter
working in unipolar voltage switching. To this purpose, the power electron-
ics and the PMDC motor load are modeled through Simscape™ Electrical
elements, thanks to the Specialized Power Systems ™ library (Simscape/Elec-
trical/Specialized Power Systems/ ) as shown in Chapters 6, 17 and 18.
In particular, the Mosfet blocks available in the subset /Funda-
mental Blocks/Power Electronics are used to represent the switches
mounted on the BOOSTXL-DRV8301 converter board. The semiconductor
334 Cascade Speed Control of a Permanent Magnet DC Motor

characteristics taken from the converter data-sheet, e.g., Ron and Rd , are set
into the Simulink® blocks as shown in Chapters 15, 17 and 18. The two-level
Full-Bridge configuration is obtained by rearranging these blocks based on the
equivalent circuit and Simulink® scheme reported in Section 15.5. The switches
are driven by a PWM modulator with a carrier frequency fsw = 30 kHz. The
PMDC motor is modeled through the DC Machine block which can imple-
ment both a wound-field and permanent magnet DC machine by selecting the
motor model in the Field type frame. The armature circuit (A+, A-) consists
in an inductor La and resistor Ra connected in series with the back-emf. In the
permanent magnet DC machine, there is no field current creating the excita-
tion flux, which is established by the magnets. Thus, ke and kT are constant as
expected. It is important to underline that Torque constant has to be selected
in the Specify frame. The DC Machine block already include a lumped (first-
order) mechanical model based on the inertia J, the viscous friction coefficient
β (or Bm ), and, as a further option, the Coulomb friction torque (which is not
used here, i.e., = 0). The load torque ml (t) can be provided as an external
input by selecting Torque TL in the Mechanical input frame. By selecting −1
as sample time, this block uses the largest integration step-size allowed by the
Powergui block. This latter is essential to manage the numerical integration
of the Simscape™ elements, as explained in Section 6.1.2.

19.5 Full-Bridge Configuration

A Full-Bridge configuration (or H-bridge) foresees to use two legs of the
adopted boosterpack (legs A and B), for which two ePWM modules are re-
quired, e.g., ePWM1 and ePWM2. The terminals of the PMDC motor are
connected to the central point of legs A and B, as shown in Figure 15.12.
Hence, the switches 1A , 1B and 2A , 2B are considered. As detailed in Chapter
15, the H-bridge configuration allows to spin the motor in both directions (i.e.,
clockwise and counter-clockwise).
In the following simulations, it is assumed that some identification routines
have been already applied to the single-motor configuration to estimate the
system parameters. This procedure will be detailed in the following sections.
The PMDC motor is then characterized by: Ra = 0.529 Ω, La = 0.865 mH,
K = 0.0232 Nm/A, β = 90.13 µN m s and J = 7.12 µg m2 , which are slightly
different in comparison to the ones adopted in Chapter 15. It can be noted that
the mechanical coupling with the rotary encoder increases both equivalent
inertia and friction coefficient. The supply voltage is VDC ≈ 10 V and the
switching frequency is set equal to fsw = 30 kHz.These parameters can be
initialized in Model Properites/Callbacks/InitFcn or in a separate m-file as
shown here in the following::
Full-Bridge Configuration 335

%% power supply
Vdc = 10;
%% carrier frequency definition
fsw = 30e3;
Tsw = 1/fsw;
%% sampling time definition
Ts = Tsw;
%% step-size definition
Tsim = Tsw/400;
%% motor parameters
Ra = 0.529;
La = 8.651e-4;
K = 0.0232;
B = 9.013e-5;
J = 7.118e-6;
%% current controller
tauGc = La/Ra;
TaGc = 5*tauGc;
TaI = TaGc/10;
wcI = 5/TaI;
kpI = wcI*La;
kiI = wcI*Ra;
%% speed controller
tauGm = J/B;
TaGm = 5*tauGm;
wcW = wcI/100;
TaW = 5/wcW;
kpW = wcW*J;
kiW = wcW*B;
%% encoder settings
Tw = 5*Ts;
cpr = 2400;
minRPM = 1/(cpr*Tw)*60;
minRAD = minRPM*pi/30;
%% MRAS settings
Tsm = 5*Ts;
kpMR = kpI;
kiMR = kiI;

Based on this data, the electrical transfer function is characterized by a small

time constant τGi = L/R = 1.6 ms, which leads to a settling time of Ta,Gi =
8.2 ms. In order to improve such uncontrolled dynamic, the settling time of the
current control loop is set to Ta,Fi = Ta,Gi /10 = 818 µs. This choice implies
τFi = Ta,Fi /5 = 164 µs and a controller bandwidth equal to ωc,i = 1/τFi =
336 Cascade Speed Control of a Permanent Magnet DC Motor
SOC reload

PWM dmax closed-loop

w* Z-1 u
reference s1A

cascade controller s1B

west
Tw (syncronized with PWM)
ia modulator s2A
ia debug open-loop
s2B
SOC

west
Z-1 computational delay
w
enable

s1A s2A
load
torque
g

D
w
ia
S

S
TL m
me
dc

s1B A+ A- s2B
g

D
+
-
v
S

S
va
+
-

+
-
Continuous
v

vA vB

Figure 19.4 Simulink® scheme to simulate the closed-loop speed control of a

PMDC motor driven in Full-Bridge configuration unipolar voltage switching.

6115 rad/s. The values of kp,i and ki,i are explicitly determined as:

kp,i = ωc,i La = 5.29 Ω ki,i = ωc,i Ra = 3235 Ω/s (19.10)

The speed control dynamic is designed to fulfill the cascade constraint.

Namely, the speed controller bandwidth is set to ωc,Ω = ωc,i /100 =
61.15 rad/s. Therefore the outer loop is significantly slower than the inner
one. Therefore, the resulting settling time is Ta,FΩ = 81.8 ms, while the cor-
responding values of kp,Ω and ki,Ω are:

kp,Ω = ωc,Ω J = 4.35 × 10−4 N m s ki,Ω = ωc,Ω β = 0.0055 N m (19.11)

Discrete-time PI controllers RΩ (z) and Ri (z) are derived based on these con-
siderations. The sampling time of the controller is set equal to Ts = Tsw =
1/fsw , using Forward Euler as integration method. Getting closer to the MCU
implementation, the system variables are referred to the discrete-time domain.

The equivalent circuit shown in Figure 15.12 is then realized in Simulink® in-
cluding the control scheme discussed previously (see Figure 19.3), as reported
in Figure 19.4. This scheme is available in the file:
Full-Bridge Configuration 337

SOC
ei
reference
ew i*
iref

wref
1 m* i*
w* cut PI(z) PI(z)
noise u
e 3 1
2 uC
ww ia da
w d
mref i
e

speed estimator (MRAS)

d(k) is discontinous
d due to ctrl
i
Discrete west
1 PI(z) 2
u y va
0.0008s+1 f Adaptive ia est.
model west
wm

Figure 19.5 Detailed view of the cascade controller subsystem (see Figure
19.4). The top part of the scheme reports the two nested control loops realized
based to the scheme shown in Figure 19.3, whereas the bottom shows a possible
realization of the MRAS observer.

FBsimDC_closed.slx (solver: fixed step - ODE4, step size: Tsim )

FBsimDC_closed.slx (powergui: simulation type Continuous)

Similarly to the previous chapters, the simulation time step Tsim is set quite
small. Moreover, the reference subsystem includes different kinds of inputs
to operate the system differently.
The actuation of the two legs (i.e., two duty cycles for two modulator
stages) is done by manipulating one modulation signal d(k) only, from which
the gating signals of the four switches are generated. This duty cycle comes
from a processing done on the desired reference armature voltage va∗ (t), which
has to be divided by VDC provided by the power supply, i.e., 10 V in this case.
A detailed description of the unipolar voltage switching operation is reported
in Section 15.5.2. However, it is of great importance to recall that the PMDC
motor is subjected to the difference of the leg voltages vA and vB , which is
defined as:
va (k) = vA (k) − vB (k)
The adopted working operation foresees that the duty-cycles of the two legs
dA and dB are linked as follows:

dB (k) = 1 − dA (k)

Therefore, the armature voltage can be rewritten as:

va (k) = (2dA (k) − 1) VDC

As a result, va (k) jumps between 0 and −10 V or from 0 and 10 V depending

338 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.6 Closed-loop dynamics of the cascade speed control of a single

PMDC motor.

on the desired direction of rotation (that is, on the value of the duty cycle,
as shown in Section 15.5.2). Indeed, it must be remembered that Full-Bridge
converters allow to run DC motors in both clockwise and anti-clockwise di-
rections of rotations without modifying the physical connections of the setup.
Speed and current feedbacks are taken from the measurement port of the DC
machine block. The aim of this exercise is to design a cascade control strategy.
Full-Bridge Configuration 339

Namely, this means to control and stabilize the motor speed Ω ∗ (k) despite
the presence of a load torque, while achieving a settling time Ta,FΩ = 81.8 ms.
To check these performances, a step-wise speed reference Ω ∗ (k) jumping from
−100 rad/s up to 100 rad/s is applied. Figure 19.6 shows the closed-loop dy-
namics of the cascade speed control of a single PMDC motor for a step of
100 rad/s in the reference speed. Both speed Ω(k), torque me (k) and, conse-
quently, armature current ia (k) are correctly following their related references.
It can be noted that the settling times are respected both for the current and
the speed, being ≈ 818 µs and ≈ 82 ms, respectively. The current control loop
is quite fast compared to the speed control loop. The references m∗e (k) and
i∗a (k) are almost coincident with the actual variables me (k) and ia (k), while
the transient of Ω(k) is visible. It follows an exponential behavior. Current
and torque responses are scaled by the factor K. Due to the different system
parameters and controller settings, the plots reported in Figure 19.6 show
that the current behavior is pretty different with respect to those reported in
Chapters 15 and 17.
Since the reference Ω ∗ (k) consists in a positive speed only (i.e., clockwise
rotating direction), the voltage is varying between 0 and VDC . At the time
instance in which the speed reference changes from 0 to 100 rad/s , the current
ia (k) grows up quickly, starting the rotation of the motor shaft. Therefore, the
back-emf starts acting on the armature circuit after less than 1 ms. Then, the
current drops down to values close to 0.5 A thanks to the Ri (z) controller.
The current peak (which translates in torque peak as well) is due to a sudden
variation in the duty cycles d(k) = dA (k) and dB (k), after which they both
approach the related steady-state values. A pulse in the average and RMS
value (which is always greater than zero) of the armature voltage va (k) are
measured as well.
It can be noted that both speed and the current ripples, ∆Ω and ∆ia , are
quite small, as a consequence of the unipolar voltage switching operation.
After the steady state is reached, a load torque ml (k) = 0.02 N m is ap-
plied to the motor to test the robustness of the cascade control scheme. The
load torque is defined into the subsystem load torque (see Figure 19.4).
The command-tracking performances are quite different from the disturbance-
rejection ones, with the latter characterized by a longer time constant. Indeed,
in order to compensate the additional load torque applied to the system, the
cascade control requires a time interval which is larger than the settling time
Ta,Ω . This effect can be changed by setting a higher bandwidth ωc,Ω . It must
be noted that the control effort to compensate ml (k) is quite small, since the
duty cycles and the armature voltage are slightly higher compared to their val-
ues before the load torque transients. Besides the verification of the settling
time of the closed-loop system, the choice of the sampling time to avoid is-
sues related to peripheral and task scheduling, e.g., controller calculation and
actuation updates, is investigated as discussed in Section 13.10 and 17.4. To
this purpose, the elements involved in the closed-loop control scheme loaded
into the MCU, i.e., current reading (ADC module), encoder reading (eQEP),
340 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.7 Part I of Figure 19.8.

Full-Bridge Configuration 341

Figure 19.8 Part II of Figure 19.7. Analysis of the MCU scheduling and latency
within a sampling period. Due to the different bandwidth between the loops,
the command-tracking performances of the speed and the related speed error
are slowly changing compared to the current dynamics and the related error.
The duty cycles actuation happens at every PWM reload action.

2L-PWM actuation (ePWM module) and a PI-based controllers, are sched-

uled to be executed within the sampling period Ts = Tsw = 1/fsw . The target
hardware is subjected to a similar sequential events execution as that one ob-
served for the current control of an RL load (Section 17.4): It is summarized
in Figure 19.7 and 19.8.
Therefore, the controller adopted in the Simulink® simulation is consistent
with such event sequence.
In particular, a trigger signal SOC (i.e., a boolean square-wave) is gener-
ated inside the modulator subsystem every time the carrier equals 1. The SOC
is used as trigger source for a triggered subsystem which realizes a cascaded
controller synchronized with the carrier peaks. Inside this block, The RΩ (z),
Ri (z) controllers are modeled by using the PID controller blocks available in
the Simulink® library, as shown in Figure 19.5. A manual switch is used to
include or exclude the outer speed loop.3 Therefore, the closed-loop perfor-
mances of both traction and braking motors can be tested by using the same
simulation file (i.e., valid both for single-motor and B2B configurations). Sim-
ilarly, the back-emf can be compensated or not.
A speed reference Ω ∗ (k) is internally generated by using a Pulse Generator
block (synchronous with the zero value of the triangular carrier). Since the
ADC SOC trigger is generated at the carrier peak, Ω ∗ (k) is delayed by Ts /2
to provide a correct voltage error at the speed controller input. The resulting

3 If necessary, the subsystem cut noise can be used to bring the speed error equal to

zero every time it is below a certain threshold.

342 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.9 Closed-loop dynamics of the cascade speed control of a single

PMDC motor rotatign in clockwise and anti-clockwise direction.

reference speed is called Ωctrl

∗
(k), where the subscript ctrl refers to the quantity
that are processed by the controller. The same holds even for the other vari-
ables. The closed-loop performances for a different speed profile are shown in
Figure 19.9. The system responses are similar to those reported in Figure 19.6,
despite higher peaks during the speed transition from 100 rad/s to −100 rad/s
(a greater control effort is required). The speed measurements emulates the
Full-Bridge Configuration 343

120 120
100 100
80 80
60 60
40 40
20 20
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07

2 2

1.5 1.5

1 1

0.5 0.5

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07

(a) (b)

Figure 19.10 Comparison closed-loop dynamics (in terms of speed and arma-
ture current transients) for (a) Tw = 5Ts and (b) Tw = 200Ts
.
presence of a rotary encoder. By considering the standard approach proposed
in Section 14.4 it follows that:
xp (k) − xp (k − 1) ∆x(k)
Ω(k) ≈ = (19.12)
Tw Tw
where Ω(k) is the speed computed at time step k, xp (k) and xp (k − 1) are
the position counters at time step k and k−1 respectively, Tw is the acquisition
time window, which is the inverse of speed calculation rate. The value of Tw is
of paramount importance since it is directly related to the speed computation:
round revolution 1
Ωmin = or = 60 (19.13)
min min cpr · T
| {z w}
revolution
sec

Considering a constant cpr, a large value of Tw implies an higher speed mea-

surement accuracy, being able to measure down to small value of Ω(k). Nev-
ertheless, in order to keep the synchronized event sequence in the embedded
target, the acquisition time must be a multiple of Ts . Hence, a large value
of Ts may lead to a quite slow speed computation task adding a relevant
delay in the whole cascaded control action (e.g., the duty cycles tend to sat-
urate for long period). In the worst case scenario this may lead to system
instability.
Based on this fact, even if a sufficiently large value of Tw is necessary to
well approximate the actual speed, it is quite often to keep Tw in a [2Ts , 10Ts ]
range for switching frequencies below 30 kHz.4 In order to analyze the effect of
4 For larger switching frequencies this range is extended.
344 Cascade Speed Control of a Permanent Magnet DC Motor

the acquisition time Tw , Figure 19.10 (a) and (b) compare the closed-loop dy-
namics in terms of speed and current transients for Tw = 5Ts and Tw = 200Ts ,
respectively. It is quite evident from Figure 19.10(b) that large values of Tw
creates a discontinuous behavior even in a fast control loop like the current
one. From now on, Tw = 5Ts is set, which leads to Ωmin = 15.71 rad/s.

Remark: in case the cascade control loop is required to operate within par-
ticularly low speed (e.g., even lower than the Ωmin ) it is better to change the
speed computation approach, as described in 14.4. Moreover, by decreasing the
bandwidth ωc,Ω the acquisition time might be increased without introducing
too much delay in the control loop.

19.5.1 Model reference adaptive system (MRAS) observer

One of the evergreen research areas in the field of electrical drives is the study
of estimators and observers aimed to describe the state variables of a motor. In
order to achieve a robust cascade architecture, the closed-loop performances
strongly depends on the quality of those feedbacks.
In general, current and voltage sensors are low cost devices and simple
to measure. Therefore, it is widely common to assume that those sensors are
always present in a motor control application. Combining them with the math-
ematical description of the motor dynamics, it si possible to estimate many
variables, e.g., electromagnetic torques, speeds and fluxes. In particular, there
are different kinds of speed sensors available on the market. Their costs and
characteristics are relatively variable, and they are strongly affecting the fi-
nal cost and performances of a product, especially for small/medium size low
voltage drives. On the other hand, from medium to high voltage drives, the
problem of getting a safe and clean transmissions of the speed (or position)
measurement from the motor site to the control cabinet is of great impor-
tance and mostly solved by adopting expensive wiring harness. Due to that,
even for two different reason, the possibility to replace speed sensors with
estimators/observers could be a wise option despite the power rating of the
drive.
In the electrical drives field, control schemes which are not based on direct
measurements of speed or position in favor of estimators/observers are called
encoderless schemes or, more in general, sensorless schemes.
The aim of this Section is to built an encoderless cascade speed control
of a PMDC motor by using the so called model reference adaptive system
(MRAS) speed observer to estimate the rotor speed. The MRAS observer is
only one candidate among the large family of observers, which are not fully
detailed here.

Remark: from the terminology point of view, it is important to underline that

the estimators are based on open-loop methods which benefits of a easy im-
plementation although they are quite sensitive to parameter uncertainty and
Full-Bridge Configuration 345

u (t) Reference y (t)

+ Adaptation
model
− mechanism

Adaptive ŷ (t)
model

x̂ (t)

Figure 19.11 Block scheme showing the working principle of a generic MRAS
observer.

measurement errors. Instead, the observers are based on closed-loop methods

which provide better accuracy than estimators, at the expense of a greater
complexity of the algorithms.

The MRAS is defined as an observer because it compensates simultaneously

the armature resistance and rotor time constant variations (subjected to large
changes during operation) through a PI controller, which is used as adaptation
mechanism (i.e., in a closed-loop manner). This approach can be both applied
to the single-motor or B2B configuration.
The operating principle of the aforementioned MRAS observer is based
on an adaptive model, which is a function of the variable to be estimated.
Referring to the general scheme reported in Figure 19.11, for the simplicity of
the PMDC motor model, the implementation proposed in the following does
not foresee any reference model. The measurement y(t) = ia (t) is directly
compared to its estimation ŷ(t) = îa (t) (see Figure 19.5). A PI-based controller
is used as adaptation mechanism to provide the (output) speed estimation
x̂(t) = Ω(t).
b The armature voltage is considered as input variable u(t) = va (t).
The loop iteration refines x̂(t) until ŷ(t) ≈ y(t). To this purpose, the electrical
contribution in a PMDC motor model is reported here:

dia (t)
va (t) = La + Ra ia (t) + KΩ(t) (19.14)
dt
This expression is used to predict the deviation of the current at time
instant t (reference model). Thus, ia (t) can be computed by solving this dif-
ferential equation. In view of an implementation on MCUs, the adaptive model
is discretized considering a forward Euler method:

dia (t) ia (k + 1) − ia (k)

≈ (19.15)
dt Tsm
where Tsm is a generic discretization step. Then, the discrete-time adaptive
model follows as:
 
Tsm Tsm
ia (k + 1) = 1 − Ra ia (k) + [va (k) − KΩ(k)] (19.16)
La La
 346 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.12 Detailed view of the adaptive model subsystem (see Figure 19.5).

The corresponding Simulink® implementation of this equation is reported in

Figure 19.12, which details the adaptive model subsystem. Indeed, the MRAS
observer is included inside the triggered subsystem to process the involved
variables based on the right event sequence scheduling.
A careful sensitivity analysis of this equation shows that there is the need
to invert the signs of the sum reported in Figure 19.11. Indeed, by applying a
virtual displacement on the mechanical speed, it can be found out that:

∆ia Tsm K
=− (19.17)
∆Ω La
This equality suggests that an increase in the speed causes a drop in the
armature current. Indeed, inverting the signs before the PI regulator leads to
the following scenarios:

îa (k + 1) > ia (k + 1) → Ω (k + 1) ↑ ⇒ îa (k + 2) ↓

îa (k + 1) < ia (k + 1) → Ω (k + 1) ↓ ⇒ îa (k + 2) ↑

Besides the accuracy of the system parameters, the stability of the proposed
observer is related to Tsm . Similarly to the encoder trade-off between sensing
accuracy and latency, the MRAS observer has to be designed to well ap-
proximate the speed measurements (i.e., a small Tsm to increase the iterative
refinements). However, if the speed controller is relatively slow and the actual
speed value does not need to be update every Ts , the discretization time can be
set as Tsm = 5Ts as for the encoder. This is helpful to reduce the computations
which has to be scheduled within Ts (which may enlarge the overall execution
time), while keeping the synchronization between computations since Tsm is
a multiple of Ts . Therefore this would be the preferable choice.
Since the Adaptive Model is based on the armature current dynamic, the
kp,M R and ki,M R parameters of the the PI controller can be set equal to those
computed for the current control loop, i.e., kp,M R = kp,i and ki,M R = ki,i .
These value define the expected time of convergence of the estimated value
Ω̂(k). As previously said, this approach is suitable for both Half- and Full-
Bridge configurations (i.e., feasible for single-motor and B2B test cases). The
only difference would be in the definition of the armature voltage, which is
related to the specific adopted topology. The MRAS observer is assumed to
be used in the Full-Bridge configuration scheme adopted to drive the single
PMDC motor. The armature voltage is va (k) = (2da (k) − 1)VDC , as shown in

wm ia_est
2 1

1
Va
Single Motor Configuration 347

Figure 19.13 Comparison between command-tracking performances of the

speed loop when an encoder, a MRAS observer, and a MRAS observer with
a low pass filter are used. The control scheme is used to drive a single PMDC
motor in Full-Bridge configuration.

the implementation scheme shown in Figure 19.5. The command-tracking per-

formances of the speed loop are reported in Figure 19.13. The speed dynamics
given by using the encoder (blue line) is kept as reference case. The MRAS
observer (solid purple) is well approximating the measured speed, apart from
the introduction of a processing latency. In particular, the speed reference is
changed from −100 rad/s to 100 rad/s, requiring a sudden control action to
meet the desired settling time Ta,FΩ . The duty cycle d(k) = dA (k) is used to
compute u(t) ≈ u(k) = va (k). Therefore, a peak in d(k) is reflected in the
Ω̂(k) estimation. Based on the application requirements, if this peak is an
issue, a low pass filter can be used to smooth the speed dynamic, denoted as
Ω̂f (k), by filtering the duty cycle df (k). Figure 19.13 shows that this coun-
termeasure effectively reduces the peak in Ω̂f (k) at the cost of an additional
latency. In both cases, the use of a MRAS observer lead to a stable cascade
control scheme.

19.6 Single Motor Configuration

The implementation of the cascade speed loop for a single PMDC motor cor-
responding to the simulation proposed in Section 19.5 is reported below. First,
the procedure required for the identification of the system parameters is de-
scribed. Then, a step-by-step procedure for the realization of the firmware for
the speed control is proposed.
348 Cascade Speed Control of a Permanent Magnet DC Motor

Table 19.1 Estimated electrical parameters for PMDC motor control.

˝Gi (ms) Ra (
) La (mH)
1.635 0.529 0.865

19.6.1 Parameter identification

The parameters involved in the speed control loop (namely, Ra , La , J and β)
should be identified through suitable tests. Indeed, parameter identification
of the PMDC motor is suggested to design the PI controller properly. In
this case, they can be estimated separately through the measurement of the
dynamics of the electrical and mechanical parts of the system, respectively.
Indeed, assuming to connect one PMDC motor to a Half-Bridge converter
and an encoder (see Figure 19.1), its armature resistance and inductance can
be identified through a locked-rotor test, as reported in Section 16.2. Thus,
Ra can be evaluated measuring the voltage and the current at steady-state,
whereas the identification of La requires the study of the transient of the
current. In particular, the time constant τGi is essential to this aim. Indeed,
an exponential interpolation of the measured data should be performed in the
MATLAB® editor through functions like fittype or fminsearch. Using the
same firmware proposed in Section 16.2 (this time considering the calibrated
gain and offset for the considered ADC channel) and current/voltage probe
connected to an oscilloscope, the electrical parameters can be evaluated. The
results are reported in Table 19.1.
Actually, still one electrical parameter is missing, that is, the back-emf con-
stant K = ψPM ke which is accounting for the flux of the permanent magnets.
Moreover, this parameter is numerically equal to the torque constant ψPM kT ,
that allows to compute me (k) knowing ia (k). This K can be evaluated by
supplying the PMDC motor in open loop configuration with the same voltage
step adopted for the locked rotor test. Indeed, at steady state it holds that:

Va = Ra Ia + KΩ (19.18)

Hence, knowing the armature resistance Ra and measuring the voltage Va ,

the current Ia and the mechanical speed of the rotor Ω at steady state, it is
possible to compute K as:
Va − Ra Ia
K= (19.19)
Ω
Alternatively, this parameter can be estimated by mechanically connecting
the motor under study with another machine, building up a setup similar to
the B2B one. This second motor turns the first one, causing an open circuit
voltage drop on its terminals. Thus, K can be computed measuring this voltage
and the rotational speed of the two machines. Ω(k) can be still measured
through the rotary encoder. Moreover, its steady-state profile can be visualized
Single Motor Configuration 349

Table 19.2 Estimated mechanical parameters for PMDC control.

τGΩ (ms) K (Nm/A) β (µN m s) J (µg m2 )

79 0.0232 90.13 7.12

thanks to the virtual COM port, even if some delay is introduced. However,
it is assumed to be sufficiently low not to compromise the characterization of
mechanical transient, which is typically much slower than the electrical one.
To this purpose, a firmware exploiting the same block chain shown in Section
14.6 for speed measurement is built. The blocks presented in Section 16 for
running the motor in open loop are included as well. Indeed, this time the
rotor is not locked any more. If some delay issues arise, the DAC peripherals
can be used for data acquisition.
The mechanical dynamics is related to the following equation:

dΩ(k)
me (k) − ml (k) = βΩ(k) + J (19.20)
dt
The friction coefficient and the inertia of the motor can be computed
similarly to the electrical parameters. Indeed, β can be evaluated at steady
state without applying any load torque ml (k) on the shaft, whereas the inertia
needs a measurement of the mechanical dynamic. Therefore, the motor can
be still supplied with VDC = 10 V with a train of pulses with duty-cycle equal
to, e.g., d(k) = 0.2. Then, the electromagnetic torque can be computed as:

me (k) = Kt ia (k) = Kia (k) (19.21)

The resulting mechanical parameters are reported in Table 19.2. It is im-

portant to underline that J includes the effect of both the inertia of the motor
and of the rotary encoder.
It can be noted that the values adopted for the simulation proposed in
Section 19.5 are those obtained from such characterization procedure, which,
in principle, must be done as first task.

Firmware Environment
c28069_DCchar_hbF.slx (solver: fixed step - ODE4, step size: Ts = 50 µs)

This firmware is obtained by joining the schemes realized in Section 14.6 and
16.2 (fsw can be set equal to 20 kHz or 30 kHz). Indeed, the motor is supplied
with a d(k) = 0.2 pulse duty cycle once again and the speed measurement
through the encoder is still valid: this signal is processed so that the angular
speed in rad/s is obtained. In addition, the GPIO block is replaced with a SCI
Transmit block set as reported in Section 10.5. A Rate Transition block with
sample time equal to 0.01 s is added and the Data Type Conversion block is
set on single values. Finally, the resulting Simulink® scheme is shown in Figure
350 Cascade Speed Control of a Permanent Magnet DC Motor
encoder
eQEP
SCI XMT
qposcnt Z-1 w
Data
C28x
C28x

ePWM

WA

C2802x/03x/05x/06x
ePWM1A: J4 - pin40

Figure 19.14 Simulink® scheme included in c28069_DCchar_hbF.slx.

19.14. It is important to note that voltage and current measurements adopted

for the proposed computations are performed with voltage and current probes
connected to an oscilloscope. The testing environment for the acquisition of
the mechanical speed is the same as the one proposed in Section 16.3.

19.6.2 Control implementation

The closed-loop dynamics of a cascade speed control of a single PMDC motor
are tested through the DecMot Hardware Kit (see Figure 19.1).
Both legs (phase) A and B are used for this implementation (H-bridge):
the connection scheme is reported in Figure 19.15.
Therefore, two ePMW modules are used and they are operated in up-down
counting mode at fsw = PWMfrequency = 30 kHz. Reminding that for the
F28069M board fck = CPUfrequency = 90 MHz, it follows:

CPUfrequency
TBPRD = PWMcounter−period = = 1500 (19.22)
2 · PWMfrequency

where the sampling time is Ts = Tsw = 1/PWMfrequency .

The reference speed can be internally generated in the firmware or exter-
nally provided both using the available potentiometers on the extPot3 board
or through COM port (e.g., SCIA). The analog solution is adopted to add
versatility to the control scheme in the proposed solution (e.g. use one of the
linear potentiometer as control enable). The needed parameters can be ini-
tialized in Model Properites/Callbacks/InitFcn or in a separate m-file as done
for simulation.
A step-by-step procedure for programming the microcontroller for this task
is reported here in the following.

Firmware Environment
c28069_DCcontrol_fbF_uni (solver: fixed step - ODE4, step size: Ts =
1/2fsw = 16.7 µs)

Open a new blank Simulink® project and configure the environment as shown

ePWM1A: J4 - pin40
C2802x/03x/05x/06x

WA

ePWM

C28x
 Single Motor Configuration 351
rotating

legA

legB

encoder cable

enable
DC link (ADCB3)

Figure 19.15 Connections of the DecMot Hardware Kit with the LaunchPad
F28069M through the BOOSTXL-DRV8301 converter and the extPot3 board.
A Full-Bridge configuration is realized and the encoder wires are interfaced
with module 2 of the eQEP peripheral.

in Section 9. Remember that the BOOSTXL-DRV8301 is enabled by settling

GPIO50 high (see Section 3.2).
• Insert an ePWM block and use the same settings shown in Section 13.5:
– Select ePWM1 module with TBPRD corresponding to fc = 30 kHz,
i.e., PWMcounter−period = 1500;
– Remember to unflag Enable ePWM1B and tick Inverted version of
ePWMxA. This option allows to drive 1B as the complement of 1A ;
– Keep all the other settings as reported in Section 13.5.
• Insert a second ePWM block: select ePWM2 module and use the same
settings adopted for the previous block.
• Add two gain blocks to normalize the modulation signals over the peak
values of the carrier PWMcounter−period . Include two data type conversion
blocks set on int16 to represent the duty cycle within a 16-bit base.
These two ePWM modules are arranged to operate the converter in Full-
Bridge configuration, i.e., d(k) = dA (k) and dB (k) = 1 − dA (k). This relation-
ships can be achieved by using simple Add and Constant blocks. The reader
is referenced to Section 15.5.4 to see their connections. For up-down counting
mode, the measurement of an average current ia (k) is feasible thanks to the
ePWM module which sends a synchronization signal that triggers the SOC of
the related ADC modules when the carrier equals TBCTR=TBPRD; further
details on this topic are provided in the previous Chapters. To this aim, the
following settings are summarized here below:
 352 Cascade Speed Control of a Permanent Magnet DC Motor

• In the Event Trigger (Tab) of ePWM1:

– Select Enable ADC start of conversion for module A;

– Set Number of event for SOCA to be generated to First event;
– Select Counter equals to period (CTR=PRD) as Start of conversion
for module A event selection.
Every time the carrier signal reaches the maximum value TBPRD, a SOC
input (i.e., ePWM1_ADCSOCA) is sent to the ADC. This last block has to
be configured as well. The onboard shunt on leg 1 is exploited for the current
measurement.
• Add an ADC block and select ADCINA0 as Conversion channel in the
Input Channels tab. Then, go back in the SOC Trigger tab:
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC0;
– SOCx acquisition window: 7;
– SOCx trigger source: ePWM1_ADCSOCA;
– ADC will trigger SOCx: No ADCINT;
– Sample time: Ts ;
– Data type: uint16.
• Add a Data Type Conversion block set to int16 (since positive signals
only are used) or single to exploit 32 bit.
• Place a Gain block for scaling the digital signal over [0, 1], since a 12-bit
reading returns values ranging from 0 (0 V) up to 212 − 1 = 4095 (3.3 V).
• Build a proper conversion chain for the current sensing by removing the
estimated offset (e.g., 2064) as well as using the scaling factor gadc to obtain
Ampere values after moving to single data type. See the characterization
procedure in Chapter 16.
• Regarding the speed and current controllers, use the same discrete PID
controller blocks adopted in simulation; set the type of the controllers as
PI (Parallel) with sample time Ts . For the speed controller P=kp,Ω and
I=ki,Ω are set, with the controller output m∗e (k) saturated within the range
[−Kia,max , Kia,max ], where ia,max = 10 A. For the current controller, P=kp,i
and I=ki,i are set, with the controller output va∗ (k) saturated within the
range [0, VDC ]. A more conservative choice might be to consider 0.98 · VDC
as upper bound. In both blocks, select an external source for the reset with
rising condition, and set back-calculation method as anti-windup strategy
for both of them with Kb=kaw .
 Single Motor Configuration 353

• Redundant saturation blocks can be used to protect the system by limiting

the required voltage and torque.

• Add a Gain block to normalize the speed controller output (which varies
in the [−Kia,max , Kia,max ] range) through a scaling factor 1/K. Then, the
signal spans a [−ia,max , ia,max ] range, becoming a reference current. De-
pending on the value of kaw , dynamics close to the upper/lower bounds are
damped by the back-calculation strategy.

• Add a Gain block to normalize the current controller output (which varies
in the [0, VDC ] range) trough a scaling factor 1/VDC . Then, the signal spans
a [0, 1] range, being consistent with the definition of d(k) = dA (k) (and,
then, dB (k) is computed).
• Add a Pulse Generator (sample based) block to create the reference Ω ∗ (k).
Set the amplitude equal to A = 200, number of samples ns = 60000, pulse
width 30000 samples, phase delay 0, and sample time Ts . It is equivalent to
have a square-wave Ω ∗ (k) of period 60000Ts = 2 s with Ton = 1 s. It can be
noted that Ω∗ (k) is synchronized with the zero value of the triangular carrier
CTR=0. Then, a Add and a Constant blocks are placed to bias Ω ∗ (k) and
to generate a reference signal ranging from [−100, 100] rad/s.
• Add a Gain set on K, a constant block, a Manual Switch and an Add block
to allow the inclusion of the back-emf compensation in the control scheme.
• Regarding the speed measurement, add an eQEP block and build the related
conversion chain to process qposcnt in order to get Ω(k); the reader is
referred to Chapter 14, Section 16.3 and to the file c28069_DCchar_hbF.slx.
It can be noted that these eQEP blocks are scheduled through the sample
time Tw . To this purpose, it must be noted that Tw must be set large enough
to provide a good approximation of the measured speed. Indeed, a too low
value fo Tw may badly compute the speed value within the acquisition time,
while a too large Tw introduces further actuation delay that may lead the
controllers to saturate. As previously explained, a trade-off must be found.
IN this implementation, Tw = 5Ts is considered.
Arrange these aforementioned blocks as shown in Figure 19.16. In order to
manually manage the activation of the PI controller, the external reset (rising)
is connected to a potentiometer (ADCINB3/pin68). Without moving it, the
control algorithm is forced not to provide any actuation signal. To this aim:
• Add an ADC block and select ADCINB3 as Conversion channel in the
Input Channels tab. Then, go back in the SOC Trigger tab:
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC1;
– SOCx acqusition window: 7;
 354

C2806x
pot unipolar
[ref] GPIOx
C2802x/03x/05x/06x scaling
current test GPIO DO
A3
B
ADC C2802x/03x/05x/06x
db
speed ctrl WA

current ctrl ePWM

i*
pulses m*
[w] PI(z) A
[ref] u
PI(z) C2802x/03x/05x/06x
[ia] da
enable i
en WA
C2802x/03x/05x/06x en
ePWM
B3 >= 0.5
e DAC3: J8 - pin72
ADC C2802x/03x/05x/06x
[w]
WA
C2802x/03x/05x/06x
ePWM
A0
i
[ref]
ADC [ia]
Discrete i
shunt A 1
u y [w]
0.0009s+1 iaf

C28x
C28x w [w]
qposcnt Z-1 Data
w Discrete
1 [ia] SCI XMT
eQEP u y
0.009s+1 wf
encoder
speed estimator (MRAS)
[w]

[da]
[ia] DAC1: J4 - pin31
Continuous
1 west C2802x/03x/05x/06x
u y va
0.0008s+1 f Adaptive ia est. PI(z)
model [w] unipolar
wm WA
scaling
d(k) is discontinous
due to ctrl [en] ePWM

Figure 19.16 Simulink® scheme included in c28069_DCcontrol_fbF_uni.

Cascade Speed Control of a Permanent Magnet DC Motor
 Single Motor Configuration 355

– SOCx trigger source: Software;

– ADC will trigger SOCx: No ADCINT;
– Sample time: Tsig = 0.001 s;
– Data type: uint16.
• Add a Data Type Conversion block set to int16 (since positive signals
are used only) and the 12-bit scaling to convert the interval from 0 (0 V) to
212 − 1 = 4095 (3.3 V) in a range of variation [0, 1].
• Add a Compare to Constant block, which is used to evaluate if signals
are greater or equal than a Constant value 0.5. If this condition is true,
the output switches from zero to one and it resets the integrator of the PI
controller. The enable signal is defined as en(k).
• Additionally, en(k) can be used to add a further protection redundancy.
Indeed, by using a Product block, the enable signal is multiplied by the
actuation variable. As long as en(k) = 0, the duty cycle is forced to stay
equal to zero d(k) = 0, i.e., no switching is performed.
• To visually evaluate the step-wise reference Ω ∗ (k), the Pulse Generator can
be connected to a GPIO block (e.g., GPIO7) and read through a voltage
probe. A proper bias of 100 and a scale factor of 1/200 are used to re-
normalize the signal in a [0, 1] range. This processing allows to visualize the
reference signal ranging in [0, 3.3] V at pin 79 through an oscilloscope.
• An MRAS observer can be realized by copying the corresponding block
scheme adopted in the simulation file FBsimDC_closed.slx. Those blocks
are suitable for implementation since they were arranged for discrete-time
domain executions already.
It must be noted that some Memory blocks might be required to avoid alge-
braic loops. This is quite common when recursive loop are present. A manual
switch is then used to select if the outer speed controller is feed by the speed
measurements coming from the encoder or by the speed estimation computed
by the MRAS observer, as shown in Figure 19.16. An initialization script ei-
ther in Model Properties/Callbacks/InitFcn or as a separate m-file should be
edited to set the system parameters. The latter approach is used in this ex-
ercise (see data_init_DC.m). The script is quite similar to the one used for
simulation. The additional settings are:
 356 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.17 Closed-loop dynamics of the PMDC subjected to a step variation

in the speed reference from −100 up to 100 rad/s.

%% F28069M clock
CPU_frequency = 90e6;
%% carrier frequency definition
PWM_frequency = 30e3;
%% TBPRD (counting mode) definition
PWM_counterperiod = CPU_frequency/(2*PWM_frequency);
%% sampling time definition
Ts = 1/PWM_frequency ;
%% ADC theoretical gain
g_adc = (530e-3)/62;

Optional: similarly to the principle adopted to link Ω ∗ (k) to GPIO7, the

firmware could be also prepared to visualize the duty cycle d(k) and the rotor
speed through the DAC peripherals. In this case:
• Copy the one of the previous ePWM block, select ePWM8 (which is con-
nected to DAC3 in J8, pin 72), and keep the other settings. Connect this
block in parallel to ePWM1 (i.e., connect d(k) just after the Data type con-
version block). Hence, d(k) can be observed by connecting a voltage probe
to pin 72.
 Single Motor Configuration 357

Figure 19.18 Example of load torque application on the shaft by locking it

with a hand.

Similarly, the measured speed can be sent to an other channel of the DAC
peripheral:
• Copy the previous ePWM block, select ePWM7 (which is connected to
DAC1 in J4, pin 31), and keep the other settings. Copy and paste the Gain
block set on PWM_Counter_Period and the data type conversion set on
int16. The measured speed is first scaled through a 1/400 gain and, then, it is
biased by constant which can be defined by the reader depending on specific
needs. For instance, a 0.65 bias cause a speed variation up to 100 rad s/400+
0.65 → 0.9 × 3.3 = 3 V. This might be not satisfactory in presence of a speed
overshoot, because the DAC saturates at 3.3 V. Hence, a more conservative
choice could be to use a bias of 0.45. It must be remembered that this
value is application dependent. Finally, Ω(k) can be observed through an
oscilloscope by connecting a voltage probe to pin 31.
Figure 19.17 reports the closed-loop behavior of the cascade speed control
scheme obtained by executing the firmware and observing the speed Ω(k),
the armature current ia (k) and the duty cycle applied on leg 1 d(k) = dA (k)
through an oscilloscope. In particular, Ω(k) and d(k) were measured with
voltage probes connected to DAC1 and DAC3, respectively, while ia (k) was
acquired through a current probe.
As done in simulation, Figure 19.17 shows the command tracking perfor-
mances of the controlled system in terms of speed and current dynamics. The
resulting Ω(k) (blue line) is following the reference signal Ω ∗ (k) (red line). It
must be noted that those two signals are expressed in a 0 − 3.3 V range. In
particular, the speed lies in a shorter 0 − 3 V interval thanks to the 0.65 bias
in the DAC1 processing. This is aimed to still have voltage margin to detect
358 Cascade Speed Control of a Permanent Magnet DC Motor

Figure 19.19 Closed-loop dynamics of the PMDC subjected to a load torque

acting on the shaft.

overshoots. Therefore, it must be noted that 0 V corresponds to −100 rad/s,

whereas 3 V stands for 100 rad/s. Consequently, zero speed is equivalent to
1.5 V. DAC1 results (gray dotted line) are averaged to get a good approxi-
mation of Ω(k) dynamic. The settling time is quite similar to the expected
one (Ta,F = 82 ms), showing a response which is comparable to the simulated
one. It can be noted that the speed shows some oscillations which are a con-
sequence of the encoder measurement. Those variation in Ω(k) are reflected
in the ia (k) measurement as well, resulting in a greater current ripple ∆ia .
The Pulse generator which provides the internal reference Ω ∗ (k) is synchro-
nized with the zero value of the triangular carrier (TBPRD=0). Nevertheless,
such synchronization is not as strict as other tasks, e.g., counter comparison
in ePWM, which is based on hard synchronization (hard-sync) event through
interrupts. Therefore, it might happens that the rising edge of the step-wise
Ω ∗ (k) is not perfectly synchronized with the ADC SOC (i.e., carrier peak). As
already observed in simulation and in the previous implementations of closed-
loop controls for RL and RLC loads, an actuation delay up to Ts might be
present as a consequence of this displacement.
On the other hand, the duty cycle processed by DAC3 is fully scaled among
a [0, 3.3 V] range, where 3.3 V corresponds to d(k) = 1. The signal read at
pin 72 (gray dotted line) needs to be additionally averaged to obtain d(k) (red
Single Motor Configuration 359

Figure 19.20 Comparison between command-tracking performances of the

speed loop when an encoder and a MRAS observer are used.

line). This can be easily done via software by using an average function or a
digital filter. Despite this processing, d(k) still show spikes in the correspon-
dence of the step variations of the reference signals, being consistent with the
behavior observed in simulation. Indeed, this is a consequence of the sudden
control action provided by the PI-based current controller.
The steady-state values of Ω(k), ia (k) and d(k) are in line with the final
value theorem, as shown in Chapter 5.
Furthermore, in analogy with simulations, the robustness of the closed-
loop dynamics are evaluated in case a load torque is applied to the shaft.
This can be carefully performed by simply hand-locking the rotor, as shown
in Figure 19.18. This operation is equivalent to applying a variable an not-
a-priori known ml (k) on the rotor shaft. Figure 19.19 shows how the system
reacts to such variable perturbation. As expected, the disturbance rejection
performances differ from the command-tracking ones, with the former being
much slower. In particular, the first plot shows that Ω(k) tends to decrease
when the hand is holding the shaft, while the current ia (k) increases, trying
to compensate ml (k). Even if slow, the load disturbance is rejected by the
cascade architecture until Ω(k) tracks Ω ∗ (k) again.
Nevertheless, the rotor accelerates when ml (k) is removed, producing a
rebound effect. This behavior is due to the large rising of ia (k) during ml (k) 6=
0, which has to rapidly decrease when ml (k) = 0 producing an acceleration in
the rotor speed. The latter is, then, compensated by reducing d(k). However,
it requires some time as a consequence of the slow dynamic of the speed loop.
It must be noted that the scaling of Ω ∗ (k) and Ω(k) for the GPIO7 and
DAC1 reading is different with respect to that one adopted Figure 19.17.
Indeed, in this case 100 rad/s correspond to ≈ 1.2 V to keep a sufficiently
large voltage margin to observe the speed overshoot without saturating the
peripheral.
Finally, Figure 19.20 reports a comparison of the closed-loop dynamics for
a speed loop which processes the speed measured by the encoder LPD3806-
600BM-G5-24C (blue line) and the one obtained by using an MRAS observer.
The measured curves are in accordance with the simulated results even if the
MRAS observer shows a larger latency in comparison to simulations. More-
over, the implemented MRAS do not include a low pass filter on the duty cycle
 360 Cascade Speed Control of a Permanent Magnet DC Motor

d(k). Indeed, it is left up to the reader test possible behavioral differences due
to the inclusion of a filter.

Implementation hints: even if the realization of the firmware follows a

step-by-step procedure, some issue might arise anyway. Therefore, a hands-
on experience creates a practical know-how which is useful to solve potential
issues in a limited amount of time. Some examples are:

• Noise—if the setup shows a relevant noise of electrical nature, it must be

remembered that the MOSFET are always switching ones per period due
to the presence of a dead time. Therefore, if fsw is in an audible range, this
may create a sort of chattering sound that can be avoided by increasing the
switching frequency. In case fsw is fine, the noise might be related to a con-
tinuous computational error of the speed and/or current control. Indeed, the
feedback signals might be really noisy (and noisy speed measurements/esti-
mations are reflected in the current control action) or not accurate due to a
bad characterization of the the ADC conversion chain (i.e., the experimental
characterization should be repeated). To this end, keep in mind that discrete
low pass filters may be added in the feedback path, at the cost of additional
latency.

• Execution overload—every time many blocks or computations are in-

cluded in the firmware, execution overload effects might happen due to the
fact that all computations can not be performed within Ts . This issue is
dependent on MCU communication (e.g., serial COM) in particular, which
requires a lot of resources from the target. Therefore, if the processor has
to solve a large number of operations, the communication stage may lead
to overloads. In general, this issue depends on many key factors, e.g, value
of Ts , size of data types, etc. A simple way to verify whether the execution
is overrunning or not, is to use a free GPIO channel, implement a unitary
pulse generator with a slow period (e.g., 1 s) in fixed-point unit (i.e., reduced
computational resources and latency) and connect the corresponding out-
put pin to an oscilloscope through a voltage probe. When the signal period
starts to differ from the expected one, the MCU is close to overload without
performing all the required computations in one clock cycle.
• Limit the resources—if additional blocks are added to the firmware during
the developing stage, but they are not strictly necessary for the required task,
they should be commented out to avoid to allocate useful resources for them
and avoid overruns.
Back-to-Back (B2B) Configuration 361
IDC

1A 2A

VDC + iT iB
−

T B
1B 2B

Figure 19.21 Equivalent circuit of the test setup involving the two PMDC
motors in back-to-back configuration.

19.7 Back-to-Back (B2B) Configuration

The development of a cascaded speed control for a single PMDC motor both
using a rotary encoder or a MRAS observer is useful to understand and ana-
lyze in details all the design aspects of cascade schemes and resulting dynamic
responses for different operating points. A further point which can be investi-
gated is the dynamic regulation of a load (braking) torque to emulate specific
load conditions. This condition can be realized considering a B2B configura-
tion with two PMDC motors coupled through an elastic joint. By referring to
the B2B-PMDC hardware kit (see Appendix B), the aim of the exercise pro-
posed here in the following is to achieve a test condition for which the traction
motor is driven by a cascade speed control loop able to track a reference speed
Ω ∗ (k) (by using a rotary encoder or a MRAS observer), whereas the braking
motor is driven by a current control loop able to dynamically regulate the
value of friction torque applied to the shaft.
Each PMDC motor is connected to one leg only (i.e., Half-Bridge config-
uration) of the BOOSTXL-DRV8301 converter, thus, allowing one rotating
direction only for the whole setup. The single motor exercise already covered
all the design and implementation aspects which are useful even for this ex-
ercise. Therefore, a simple review of the most important concepts is provided
in the following.
The equivalent circuit of the B2B setup is reported in Figure 19.21. In this
representation, the electrical parameters of the motor are not explicitly shown
to allow the visualization of the mechanical coupling of the two machines. The
traction motor is connected to leg 1 (phase A), whereas the braking one to
leg 2 (phase B), as shown in Figure 15.1. The traction machine is operated as
a motor, since the absorbed power is positive (see port current and voltage
362 Cascade Speed Control of a Permanent Magnet DC Motor
ia (t) ib (t) ia (t) ib (t)

ΩT (t) ΩB (t) Ω(t)

va (t) T B vb (t) va (t) T B vb (t)
me,a (t) me,b (t) me,a (t) me,b (t)

(a) (b)

Figure 19.22 Mechanical and electrical port conventions for the test setup
involving the two PMDC motors in back-to-back configuration: (a) no me-
chanical coupling (both machines spin their shaft in the same direction when
operated as motors) and (b) with mechanical coupling (the braking motor
works as a generator).

in Figure 19.22), while the braking machine is working as a generator, being

iB < 0 and, consequently, absorbing negative power (see Figure 19.22).
Indeed, this machine is subjected to the back-emf due to the motion of the
shaft caused by the traction motor. It must be noted that it is recommended
to physically connect the two motor terminals in such a way to keep the same
sign of the back emfs for both of them. Namely, this means that positive
currents in both motors would cause the rotation of the shaft in the same
direction. This holds even for the torque direction.
Moreover, based on the equivalent circuit reported in Figure 19.21, if the
leg B connected to the braking motor is operated at dB (k) = 0, it follows
that S(k) = [0 1]T . This implies a short circuit across the motor terminals
(i.e., the armature circuit). Therefore, an open-loop control for the braking
motor is not enough to limit its armature current. In the latter case, when
no load torque is applied, the braking motor is equivalently connected to an
open circuit, since its current is forced to be 0 (which does not necessarily
mean d(k) = 0). On the other hand, when the reference (braking) current is
greater than 0 in absolute value, the PI-based controller regulates the current
flow through the motor, causing a load torque ml (k) applied on the shaft and
a net current flow toward the low-side shunt on leg B as well. This underlines
that the electrical connections of the two motors are of paramount importance.
An opposite application of the back-emf of the braking motor would create
a high loop current flowing between the two machines, causing an increase
both in the electrical and mechanical stress, other than an overheating of the
windings.
In principle, high-power motors require some equipment to dissipate losses
associated to the application of load torque on the shaft to avoid overheating
of the traction machine. This is not the case for the considered test setup, due
to the limited size of the PMDC motors and the reduced amount of energy
that is flowing in the system.
Due to the connections of both motors to the same dc link, the current
that is flowing into the dc side equals the sum of the two currents flowing in
Back-to-Back (B2B) Configuration 363

Table 19.3 Identified parameters of the BSB-PMDC Hardware Kit.

Ra (Ω) La (mH) K (Nm/A) βeq (N m s) Jeq (kg m2 )

0.578 0.920 0.0245 2.10 · 10−4 1.192 · 10−5

each motor (i.e., in each leg):

IDC = iT (k) + iB (k) = ia (k) + ib (k) (19.23)

where the labels T and B stand for traction and braking, respectively. It is
quite evident that the traction motor is related to the current ia (k) (from leg
A) and the braking motor to ib (from leg B). From now on these two currents
are called ia (k) and ib (k). It must be noted that the proposed implementation
foresees that the two involved legs are independently controlled. Thus, the
braking motor may make the shaft rotate if activated before the traction one.
The model of this system now includes two electrical and one mechanical
equations, which describe the behavior of the coupled PMDC motors (rigid
shaft and coupling elements are assumed). In the continuous-time domain, it
follows that:
dia (t)
La = va (t) − Ra ia (t) − KΩ(t) (19.24)
dt
dib (t)
Lb = vb (t) − Rb ib (t) − KΩ(t) (19.25)
dt
dΩ(t)
Jeq = me,a (t) − me,b (t) − βeq Ω(t) (19.26)
dt
where Jeq and βeq are the equivalent inertia and friction coefficient associated
to the whole system (i.e., the two PMDC motors and the mechanical joint),
whereas me,b (t) and me,a (t) correspond to me (t), ml (t) (see equation (15.3)
and Figure 19.22), respectively. Moreover, it can be noted that the subscripts
a and b now denote the two motors for all the system parameters. Given the
sample time Ts and an integration method, the B2B model and the related
variable can be discretized.

19.7.1 Parameter identification

Since a new couple of PMDC motors is installed in the B2B-PMDC Hard-
ware Kit, the characterization of the electrical and mechanical parameters
of the system should be repeated. Thus, referring to the same procedure re-
ported in Section 19.6.1, the parameters shown in Table 19.3 are identified.
The electrical parameters can be determined independently for the two mo-
tors, whereas it is convenient to keep the two electrical machine mechanically
coupled to determine the overall Jeq and βeq . It is important to note that,
due to the similarity of the motors, average electrical parameters can be con-
sidered. Therefore, the electrical identification in terms of resistance R and
364 Cascade Speed Control of a Permanent Magnet DC Motor

inductance L is performed, leading to Rb = Ra and Lb = La . In addition, the

identification of the back-emf constants can be performed by measuring the
open circuit voltage of the non-controlled motor at steady state.
Based on these values, the parameters of the speed and current controllers
are computed again using a pole-zero cancellation. Both the current PI are
identical. For what regards the current loop, starting from the time constant
τGi = La /Ra = 1.6 ms and the corresponding settling time Ta,Gi = 5τGi =
8 ms, the controlled system is set to have Ta,Fi = Ta,Gi /10 = 796 µs. Therefore,
knowing that τFi = Ta,Fi /5, it follows that ωc,i = 1/τFi = 6283 rad s. Thus,
the values of kp,i , ki,i are explicitly determined as:

kp,i = ωc,i La = 5.78 Ω ki,i = ωc,i Ra = 3631 Ω/s (19.27)

The speed control dynamic is designed to fulfill the cascade constraint.

Namely, the speed controller bandwidth is set equal to ωc,Ω = ωc,i /100 =
62.83 rad/s. Therefore, the outer loop is significantly slower than the inner
one. Therefore, the resulting settling time is Ta,FΩ = 79.6 ms, while the cor-
responding values of kp,Ω and ki,Ω are:

kp,Ω = ωc,Ω J = 0.749 mN m s ki,Ω = ωc,Ω β = 0.0132 N m (19.28)

Discrete-time PI controllers RΩ (z) and Ri (z) are derived based on these con-
siderations. The sampling time of the controller is set equal to Ts = Tsw =
1/fsw , using Forward Euler as integration method.

19.7.2 Control implementation

The closed-loop dynamics are tested for both the sensored (encoder) and sen-
sorless (MRAS observer) solution on the B2B-PMDC Hardware Kit (see Fig-
ure 19.23). Even in this implementation, legs (phase) A and B are used. How-
ever, each motor is driven through one leg only, i.e., through a Half-Bridge
converter. The connection scheme for this setup is reported in Figure 19.23.
These two ePMW modules are operated in up-down counting mode at
fsw = PWMfrequency = 30 kHz. Reminding that for the F28069M board
fck = CPUfrequency = 90 MHz, it follows that:

CPUfrequency
TBPRD = PWMcounter−period = = 1500
2 · PWMfrequency

as already seen for the control of the single PMDC motor. Moreover, the
sampling time is Ts = Tsw = 1/PWMfrequency . Once again, reference speed
can be internally generated in the firmware or externally provided both using
the available potentiometers on the extPot3 board or through COM port
(e.g., SCIA). The analog solution is adopted to add versatility to the control
scheme in the proposed solution. However, the reference braking torque is
obtain by varying a second potentiometer on the extPot3 board. The required
Back-to-Back (B2B) Configuration 365

traction motor rotating braking motor

legA

GND encoder cable

legB

load
torque

enable
DC link (ADCB3)

Figure 19.23 Connections of the B2B-PMDC Hardware Kit with the Launch-
Pad F28069M through the BOOSTXL-DRV8301 converter and the extPot3
board. A Half-Bridge configuration is realized for each motor and the encoder
wires are interfaced with module 2 of the eQEP peripheral.

parameters can be initialized in Model Properites/Callbacks/InitFcn or in a

separate m-file.
A step-by-step procedure for programming the microcontroller for this task
is reported here in the following.

Firmware Environment
c28069_B2Bcontrol_hbF (solver: fixed step - ODE4, step size: Ts = 1/fsw )

Open a new blank Simulink® project and configure the environment as shown
in Section 9. Remember that the BOOSTXL-DRV8301 is enabled by settling
GPIO50 high (see Section 3.2). This firmware can be realized starting from
the file c28069_DCcontrol_fbF_uni built previously. Indeed, this implemen-
tation requires a speed and a current loop for the traction motor, as well
as speed and current measurements. Moreover, the braking motor requires
the same current control scheme adopted previously and a MRAS observer
is needed to test sensorless runs of the code. The control (and actuation)
enable is managed again through the potentiometer connected to ADCB3,
whereas DAC modules can be used for measuring speed and duty cycle with
an oscilloscope. Therefore, it is suggested to copy and paste the whole scheme
included in c28069_DCcontrol_fbF_uni in this new blank project and, then,
to slightly modify the ePWM blocks and the processing of the modulating
signals in order to move from a Full-Bridge to an Half-Bridge configuration.
Indeed, referring to the previous file:
 366 Cascade Speed Control of a Permanent Magnet DC Motor

• Modify the generation of the speed reference. As a matter of fact, the back-
to-back setup is allowed to spin the shaft in one direction only. Therefore,
remove the Constant and the Add blocks and set the amplitude of the Pulse
Generator equal to 100. Therefore, the system tracks a speed reference rang-
ing from 0 up to 100 rad/s.
• Modify the ADC block for the current measurement by allowing the Simul-
taneous sampling of channels A0 and B0, i.e., the modules connected to the
onboard shunts on legs 1 and 2, respectively. To this aim, set the SOC trigger
number equal to SOC0-SOC1 and modify the same entry in the ADC block
for enabling the PI blocks (channel B3) setting it equal to SOC4. Indeed,
multiple SOC trigger numbers are not allowed. It can be noted that this
strategy is preferred to the use of two separated ADCs since it requires less
use of embedded resources.

• Add a Mux block to split the measurement for the currents on both legs.
Then, copy and paste the same conversion chain adopted for channel A0. It
is important to modify the offset and the gain based on the identification
process for the second leg.

• Copy and paste the current control loop from the traction motor. It must be
remembered that the two current PI controllers share the same parameters.
Then, move the gain block set equal to PWM_Counter_Period, the Data
Type Conversion block set on int16 and the ePWM2 close to this second
control loop for the braking motor. The reference braking torque is generated
through the potentiometer connected to channel B5 of the ADC peripheral.
Set the parameters of this block as follows:
– Select ADCINB5 as Conversion channel in the Input Channels
tab. Then, go back in the SOC Trigger tab;
– Sampling mode: Single sampling mode;
– SOC trigger number: SOC3;
– SOCx acquisition window: 7;
– SOCx trigger source: Software;
– ADC will trigger SOCx: No ADCINT;
– Sample time: 0.001;
– Data type: uint16.
Then, use a Data Type Conversion block set on int16 and a Gain block with
parameter 1/212 to re-scale the reading of the ADC channel into a [0, 1]
range. After that, this signal is translated into a braking torque through a
Gain block set on -maxBr, that is, −0.05 N m. A redundant saturation block
can be used to limit the load torque up to this value.
• Leave all the other blocks unchanged.
Back-to-Back (B2B) Configuration 367

Rearrange the aforementioned blocks as shown in Figure 19.24.

Remark: it must be noted that the overall firmware scheme is pretty large
and complex compared to the previous exercises. Hence, it may be possible to
saturate the computational resources of the MCU during the execution of the
firmware. To avoid this issue, some parts of the scheme which are not strictly
necessary for the execution can be commented. For example, the blocks im-
plementing the MRAS observer during the sensored control. Moreover, Data
Type Conversion blocks can be inserted and/or edited to reduce the process-
ing effort during computations as well as DAC blocks can be excluded from
the firmware.

An initialization script both in Model Properites/Callbacks/InitFcn or as a

separate m-file should be edited to set the system parameters. The latter ap-
proach is used in this exercise (data_init_B2B.m) and the code is reported
here in the following:

%% F28069M clock
CPU_frequency = 90e6;
%% carrier frequency definition
PWM_frequency = 30e3;
Tsw = 1/PWM_frequency
%% TBPRD (counting mode) definition
PWM_counterperiod = CPU_frequency/(2*PWM_frequency)
%% sampling time definition
Ts = 1/PWM_frequency;
%% ADC theoretical gain
g_adcA = 4.65/550;
g_adcB = 4.65/450;
%% maximum braking torque definition
maxBr = 0.05;
%% encoder sample time definition
Tw = Ts*5;

were the parameters of the two motors and of the controllers are not reported
for the sake of brevity. It can be noted that the encoder acquisition window
is set to Tw = 5Ts . However, the same consideration reported previously on
this parameter are still valid and a new trade-off value can be determined if
necessary. As done for all the implementations proposed in this book, Figure
19.25 shows the command tracking performances of the controlled system.
The resulting Ω(k) (blue line) is following the reference signal Ω ∗ (k) (red
line). It must be noted that those two signals are expressed in a [0, 3.3 V]
range as consequence of the measurement performed via GPIO7 and DAC1,
 368

C2802x/03x/05x/06x
[wref] C2806x
A3
speed ctrl [wref] unipolar GPIOx
ADC scaling
Traction Motor m* current ctrl GPIO DO
[w] PI(z) [da]
pulses ia* A
u
PI(z) C2802x/03x/05x/06x
[ia]
enable ia
en WA
C2802x/03x/05x/06x
en ePWM
B3 >= 0.5
e
DAC3: J8 - pin1
ADC [en]
[w] C2802x/03x/05x/06x

WA
C2802x/03x/05x/06x [ia]
[ia] ePWM
Discrete [ib]
A0/B0 1
u y
0.0009s+1 iaf [w]
ADC
shunt A ib
ib [ib] [wref] C28x
Discrete ibf
1
u y
0.009s+1 [w] Data
C28x
SCI XMT
qposcnt
w
[w]
eQEP Z-1
encoder Discrete
1 [west]
u y
0.01s+1 wf

speed estimator (MRAS)

[w]

[da]
DAC1: J4 - pin31
[ia]
Discrete C2802x/03x/05x/06x
1
u y va PI(z) unipolar
0.0008s+1 f Adaptive ia est. [w] WA
model west scaling
wm
d(k) is discontinous ePWM
due to ctrl [en]

C2802x/03x/05x/06x
current ctrl B
ib* [wref] >= 0.2
B5 C2802x/03x/05x/06x
ib*
mb*
ADC PI(z) WA
[ib]
torque is limited to avoidexcessive stress ib
Braking Motor ePWM
on the traction motor and couplings
en
en
[en]
e

[w]

Figure 19.24 Simulink® scheme included in c28069_B2Bcontrol_hbF.

Cascade Speed Control of a Permanent Magnet DC Motor
Back-to-Back (B2B) Configuration 369

Figure 19.25 Closed-loop dynamics of the back-to-back setup subjected to a

step variation in the speed reference and to the application of some braking
torque by the braking PMDC machine.

respectively. In particular, it must be noted that 0 V corresponds to 0 rad/s,

whereas 3 V stands for 100 rad/s. Actually, 3.3 V represents the highest value
which is possible to visualize in case of speed overshoots (after which the
peripheral saturates).
The settling time is pretty similar to the expected one Ta,FΩ ≈ 79.6 ms.
The same holds for the current control loop. Even in this case, the speed shows
some oscillations which are a consequence of the encoder measurement. Those
variation in Ω(k) are reflected in the current measurement as well, resulting
in an increase of the current ripple.
Figure 19.25 also shows the duty cycle measurement through DAC3. The
signal read at pin 72 (gray dotted line) needs to be additionally averaged to
obtain d(k) (red line). This can be easily done via software by using an average
function or a digital filter. Despite this processing, d(k) still shows spikes in
the correspondence of the step variation of the reference signals, similar to
the single motor case. This is a consequence of sudden control action of the
PI-based current controller.
In addition, these plots show the response of the system to the appli-
cation of a dynamical load torque on the shaft through the braking motor.
Indeed, current ib (k) is proportional to ml (k) and it is almost symmetrical
370 Cascade Speed Control of a Permanent Magnet DC Motor

with respect to ia (k). The evolution in time of ml (k) depends on the dynam-
ical variation of the potentiometer connected to ADCB5. The plots show a
rejection to disturbances which is similar to that one observed in the single
PMDC motor implementation.
 Part V

Real-Time Control in
Power Electronics: Load
Emulation
 20
Debugging Tools and Firmware Profiling

In the previous chapters, firmware design has been described in a step-by step
manner, helping the reader to understand how to make an embedded control
scheme out of system simulations, which is consistent with linear control the-
ory. Nevertheless, there are a many debugging techniques that can be adopted
both to check possible mistakes or to verify the performances of the gener-
ated code. Among them, the processor-in the-loop (PIL) testing technique and
the external mode execution are adopted in this book and described in this
Chapter. Both approaches lead to test numerical equivalence between model
computations, produced code and expected behavior (predicted from simula-
tions). As the system is running, information on block states, inputs, outputs,
and time execution can be displayed within the Simulink® Editor. This allows
the designer to investigate specific problems and/or potential improvements
in blocks, parameters, or interconnections.

20.1 Processor-in-the-loop with Simulink®

The Simulink® code generation workflow allows to perform processor-in-the-
loop (PIL) simulations of a given setup. In the previous chapters, simulation
schemes have been derived to design and analyze the closed-loop control dy-
namics. PIL simulations extends this good practice by considering the control
part running online in a dedicated processor (i.e., target specifications are re-
quired) while still emulating the system/load offline in a simulation platform
(e.g., host PC). To this aim, the control algorithm is translated/compiled into
C-code and downloaded into a development board (e.g., MCU). This latter
performs a data exchange with the host PC through a serial communication,
which has to be set (e.g., type, speed). Hence, PIL is effectively a testing
technique which should be performed after simple system simulations and be-
fore working on real firmware implementation, i.e., prior to include peripheral
blocks and their hardware constraints. This operation allows to:
• Test whether the model and generated code are numerically equivalent: PIL
tests are designed to highlight possible problems with algorithm execution
in the embedded environment. For instance, it can be used to check whether

DOI: 10.1201/9781003196938-20 373

 374 Debugging Tools and Firmware Profiling

Simulink®

control controller system test

signal model model signals

code generation

serial controller serial MCU

communication C-code communication board

Figure 20.1 Block diagram showing the main steps to be carried out during
PIL simulations.

the considered control loop fits within the execution time available on the
embedded processor.
• Observe code coverage: PIL allows an evaluation of the automatically gen-
erated code.
• Perform code execution profiling: this testing technique is useful to identify
the most demanding operations inside the control loop.
Therefore, the PIL simulations are aimed to evaluate the performances of
a controller running on a target board before its implementation on a real
system. From the hardware point of view, a control platform such as the
LaunchPad™ F28069M board is required only. Instead, the load is still emu-
lated.
As an example, by considering the closed-loop speed control of a PMDC
motor as shown in Chapter 19, PIL simulations may be performed to gain
confidence and to check that the control algorithm performs as expected once
it is physically running on the MCU, without connecting the TI BOOSTXL-
DRV8301 BoosterPack or the DC machine as well. A block diagram showing
the main steps of this procedure is shown in Figure 20.1. In particular, a sim-
ulation environment is used to model and verify the behavior of a speed-loop
control for a DC motor. Basically, starting from the simulation environment
that represents the whole controlled system, the control task which have to
be executed in the MCU are selected (e.g., PI controllers and sensing path)
and translate in C-code. This latter is then downloaded into the MCU, while
the PMDC motor model is kept in the scheme, but it is emulated only (i.e.,
code is not generated for this part of the model, see Figure 20.2).
Through a communication channel, the Simulink® environment (on the
host PC) exchanges stimulus signals (e.g., test vectors) with the MCU board
for each sample interval of the simulation. Every time the target processor
Processor-in-the-loop with Simulink® 375

Figure 20.2 PIL workflow between the LaunchPad™ F28069M and the host
PC developed in Simulink® .

receives signals from Simulink® , it executes the on-board code for one sample
step. The results, e.g., a control action, are returned to the Simulink® scheme
through a communication channel. Then, one sample cycle of the simulation
is complete, and the process keeps repeating for the next sample interval. It
can be noted that in each sample period, PIL simulations do not run in real
time. Indeed, the Simulink® environment and the object code exchange I/O
data depending on the serial communication settings. Although the workflow
reported here in the following describes how the performances of a motor
control application are assessed using PIL testing, it is possible to general-
ize this procedure for any power electronic-based application, framework and
processor/MCU supported by the Simulink® environment.

20.1.1 PMDC motor control implementation through PIL

The closed-loop speed control of a PMDC motor presented in Chapter 19 is
considered. The control architecture is based on two nested loops, i.e., the
speed and current ones, with a feedforward compensation of the back-emf.
The design rules for both controllers are already discussed in the previuous
Chapter and, thus, they are not reported here.
For the sake of simplicity, the PMDC motor equivalent circuit is created
in Simulink® based on transfer function blocks instead of using Simscape™
Electrical elements. This implies no ripple on the armature current ia (k) and
the rotor speed Ω(k), since a steady-state analysis is performed (see Chapter
6). The control scheme is included in the file:

c28069_PMDC_PIL.slx (solver: fixed step - ODE4, step size: Tsim )

376 Debugging Tools and Firmware Profiling

The system parameters reported here in the following can be initialized in

Model Properites/Callbacks/InitFcn or in a separate m-file.

%% motor data
Vdc = 10;
imax = 5;
Ra = 0.529;
La = 8.651e-4;
Kt = 0.0232;
B = 1.803e-4;
J = 3.56e-6;
%% modulation and sampling
fsw = 20e3;
Tsw = 1/fsw;
Ts = Tsw; or Ts = Tsw/2
Tsim = Ts/100;
%% current controller
tauGc = La/Ra;
TaGc = 5*tauGc;
TaI = TaGc/10;
wcI = 5/TaI;
kpI = wcI*La;
kiI = wcI*Ra;
%% speed controller
tauGm = J/B;
TaGm = 5*tauGm;
wcW = wcI/10;
TaW = 5/wcW;
kpW = wcW*J;
kiW = wcW*B;

The generation of the reference signal to operate the system in several

working conditions as well as the cascaded control loops are shown in the
schemes reported in Figures 20.3 and 20.4. Before going on with the proper
PIL simulation, it is necessary to setup the Simulink® environment. While
connecting the LaunchPad™ F28069M board to the host computer, it must
be ensured that the Serial Communication is set properly on a COM port (see
Section 8 for reference). Due to the embedded nature of the PIL simulations,
the solver type has to be necessarily fixed step (i.e., discrete-time domain).
For instance, the proposed exercise uses fixed step - ODE4 (Runge-Kutta)
solver with step size Tsim = Ts /100. These settings can be edited by open-
ing the Solver panel in the Model Configuration Parameters window.
Moreover, PIL simulations requires the enable for PIL blocks creation. To this
aim:
• Click on Hardware Implementation panel, verify that TI Piccolo
 Processor-in-the-loop with Simulink® 377
PMDC motor model ml

w*
reference w* switch to PIL mechanical
w cascade electrical ml
w w va w
speed/current loops va w
ia ia me
ia ia

va ia me

e
Place your PIL
Block here

Figure 20.3 Simulink® scheme included in c28069_PMDC_PIL.slx.

F28069M LaunchPad is set as Hardware board. Then, keep the settings

as reported in Chapter 3 and 4.
• Click on Code Generation panel and then on Verification. Open the Ad-
vanced parameters section and set the Create block frame by selecting
PIL in the drop-down menu.
As an alternative, the last setting can be found by using a shortcut. Namely,
open the Model Configuration Parameters and type CreateSILPILBlock
in the upper search box. This operation directly links to the Code Genera-
tion/Verification/Advanced parameters section.

Considering the scheme shown in Figure 20.3, the control subsystem (which
includes both speed/current controllers and back-emf compensation) is the
object of the verification through PIL testing. Only this part of the Simulink®
scheme has to be implemented and run online into the LaunchPad™ F28069M
board.
To this aim, enable the settings for running PIL over Serial by typing the
instructions reported here below in the Command Window:

setpref(’MathWorks_Embedded_IDE_Link_PIL_Preferences’,’COMPort’,’COM10’);
setpref(’MathWorks_Embedded_IDE_Link_PIL_Preferences’,’BaudRate’,115200);
setpref(’MathWorks_Embedded_IDE_Link_PIL_Preferences’,’enableserial’,true);

where the user has to set the COM port which has been defined in the PC
device manager, e.g., COM10 in this example.
Then, right-click on the control subsystem and select C/C++

speed mref iref

w* current
w*
1
w PI(z)
mref e PI(z) va
1
ia
3 va
w e
ia
2

Figure 20.4 Inner structure of the cascade speed/current loops subsystem.

378 Debugging Tools and Firmware Profiling

Figure 20.5 Subsystem parameters and the resulting PIL block.

Code/Deploy this Subsystem to Hardware from the list. This com-

mand opens a window which summarizes the control
parameters defined inside the subsystem. Select inlined for all of them as
reported in Figure 20.5 and then click on the Build button. This starts the
translation of this piece of scheme in C-code, which ends with the opening
of a new untitled.slx file containing a PIL block only. This is the trans-
lated control subsystem. Copy such block and paste it in the highlighted
area in the initial Simulink® model, as shown in Figure 20.6. Now, the file
c28069_PMDC_PIL.slx is enabled to perform PIL simulations, besides the
standard ones. Indeed, by clicking on , the PIL block of the control subsys-
tem is downloaded into the MCU board and available for executions. There-
fore, the control actions are computed online on the MCU and sent as data
arrays to the PMDC motor model input. This latter simulates the motor be-
havior and the required feedback signals are transmitted back to the MCU.
The manual switch allows to test the consistency of the results and compare
the closed-loop dynamics when the system is driven by the control scheme
running in Simulink® or implemented on-board the MCU.
In this example, a step-wise speed reference Ω ∗ (k) is applied to analyze
the command-tracking performance of speed and current loops, respectively.

PMDC motor model ml

w*
reference w* switch to PIL mechanical
w cascade electrical ml
w w va
w
speed/current loops
ia ia me w
ia ia va

va ia me
w*
e
Place your PIL
w PILhere
Block va

ia

Figure 20.6 Simulink® scheme included in c28069_PMDC_PIL.slx, which in-

volves the translated block for PIL simulations.
Processor-in-the-loop with Simulink® 379

Figure 20.7 Closed-loop dynamics resulting from standard and PIL simula-
tions.

A load torque ml (k) is also applied to test the disturbance rejection and to
verify the robustness of the proposed control scheme. Both strategies return
the same dynamics, which are reported in Figure 20.7. It is possible to note
that the PIL simulations are subjected to a slower processing, since the com-
putations of the control action are performed on-board the MCU (serial com-
munication is required),1 while in standard simulations those are computed
through the host PC CPU. The PIL framework enables another widely used

1 Note that, from MATLAB® version 2019b it might be neccessary to specify the de-

sired behavior for denormal results for arithmetic operations. Denormal numbers are any
non-zero numbers whose magnitude is smaller than the smallest normalized floating-point
number. Some hardware targets as the LaunchPad™ F28069M board flush denormal results
from arithmetic operations to zero. You can set this behavior in Model Configuration
Parameters/Math and Data Types, then change the parameter Denormal Behavior to
Flush To Zero (FTZ) (instead of Gradual Underflow which is given by default).
380 Debugging Tools and Firmware Profiling

feature which is the profiling of the task execution time and function execu-
tion time of the real-time algorithm running on the MCU board. This tool
is named Profiler or Simulink® Profiling. The Profiler captures performance
data while the software is running the simulation, e.g., it identifies which
parts of a control algorithm require most time or most iterations to be com-
puted and executed. Such information is useful to evaluate the need for code
refinements and where to focus the optimization efforts in the algorithm. To
this end, profiling is particularly useful to analyze the pipelining in real-time
platform which may benefit of multi-core processors on the target hardware.
There are two ways to enable the Profiler:
• Open the Debug tab, select Performance Advisor/Simulink Profiler.
• Select Analysis/Performance Tools/Show Profiler Report

When simulation is complete, Simulink® generates and displays the perfor-

mance data in the simulation profile report in the form of a panel. This data
shows the amount of simulation time spent executing each function in the
model.2 The block hierarchy view of the report shows the profiling information
in a nested tree form. The first row of the table, which is the top of the tree,
corresponds to the entire model. Subsequent rows at the first level of nesting
correspond to single or group blocks which can be individually assessed.

20.2 External Mode Execution with Simulink®

In the framework of firmware debugging, an alternative to PIL simulations is
the real-time external mode execution. In this mode, Simulink® is still used to
generate and download the firmware code into the MCU board, but the re-
sulting executable is able exchanges data with Simulink® via a shared memory
interface.
Compared to the PIL technique, this kind of execution is useful to debug
the firmware inner computations during the design stages. As a matter of
fact, the external mode allows to run the whole code on the microcontroller
and to visualize some variables of interest using scope and display blocks in
Simulink® . The variables are recorded from hardware through the serial com-
munication. The real-time code stores contiguous response data in memory ac-
cessible to Simulink® until a data buffer is filled (i.e., data captured within one
buffer are contiguous). Then, the firmware carries on running while Simulink®
plots data. After model computations are finished, data transfer runs at a
lower priority while the process waits for another interrupt to trigger the

2 In particular, the Profiler measures the time required to execute each invocation of

these functions.
External Mode Execution with Simulink® 381

C2806x

GPIOx

GPIO DO
led blue

Figure 20.8 Simulink® scheme included in c28069_ledB_EXT. This scheme

can be obtained from c28069_ledB_F.slx proposed in Example 1, Section
11.3. In this case, the signal source is a Pulse Generator.

next model update. Depending on the firmware complexity, such data trans-
fer could be less critical than maintaining deterministic real-time updates at
the required sample interval (i.e., not a priori).
The serial communication between PC and board is automatically settled
without any block like Serial Send and SCI Receive. Hence, differently from
what is described in Part III, this time only one Simulink® file is needed both
for programming and debugging purposes, since the external mode automat-
ically manages the data flow through the virtual COM. Scopes and displays
can be inserted directly in the firmware scheme. It is important to note that
there is a maximum number of plots that can be used depending on the per-
formances and capabilities of both the serial communication and the host PC.
Indeed, code size and memory consumption are increased when the External
Mode is enabled.

20.2.1 Simulink® setup for external mode execution

The same options set so far are to be ensured even in this case. Thus, the
proper settings for the Serial Communication as shown in Section 8 must be
verified. Moreover, the reader is referred to Chapter 9 for preparing the envi-
ronment for implementations on the MCU board (e.g., settings for jumpers,
switch positions and virtual COM port).
Considering Example 1 in Section 11.3, which consists in driving the blue
led (GPIO39) in toogle mode. A Constant block is used to produce an internal
command signal which drives the GPIO39 with 0 or 1 values. The sample time
of the constant block is set equal to Tsig = 1000Ts = 0.1 s. The led is turned
on/off accordingly. As an alternative, the constant block can be replaced by a
Pulse Generator with gain A = 1, number of samples per period 1000, pulse
width 50% (that is, 500 samples) and sample time Ts . The reader is referred
to Section 11.3 for the remaining settings. Then, place a Simulink® Scope
block connected to the Constant or Pulse Generator block output, as shown
in Figure 20.8 to evaluate the produced values. Use a Rate Transition block
to improve the accuracy of data sampling and set the sample time equal to
Ts /2. Then, once the firmware design is ready, the external mode execution
can be set and enabled as follows:
• Open the Model Configuration Parameters window and go in the
Hardware Implementation menu. Check if the TI Piccolo LaunchPad™
F28069M board is set as Hardware board. Among the list of menus in
 382 Debugging Tools and Firmware Profiling

Figure 20.9 Plot from the scope block in External Mode execution.

Target Hardware Resources, look for External mode. There, enter the se-
rial COM port set in the PC device manager, e.g., COM10 in this example.
Keep the other settings as they are.
• Coming back to the Simulink® editor, set the desired simulation time, e.g.,
20 s, and click on the monitor & tune icon to run external mode
execution.
These settings are necessary to program the MCU and to run the External
Mode execution for the supported MCU devices. Starting from MATLAB®
release R2018a, this procedure is easier due to the introduction of a specific
"TI Piccolo F28069M LaunchPad" target as Hardware board, which has de-
fault settings matching the LaunchPad™ SCI pins and oscillator frequency.3 In
previous releases, it was required to manually adjust the oscillator frequency
and the SCI pins to match the LaunchPad™ board layout. This settings al-
ready include a target buffer size which specifies how much memory should
be allocated to store signals for the external mode.4
In addition, some export settings for the variables of interest can be found
in external mode control panel, by clicking on the icon Control Panel .
Once the external model execution is launched, the model is built, compiled
and deployed in the platform. The simulation automatically starts and the
data values can be assessed through Scopes. In this case, the pulses used to
make the led blinking can be visualized as shown in Figure 20.9. An other im-
portant feature of the external mode is that constant blocks can be edited dur-
ing the execution, allowing several real-time verifications for different steady-
state working conditions.
Unfortunately, the external mode suffers from a communication bottle-
neck. It may happen that the data exchange via COM fails, causing an abrupt
stop of the Simulink® simulation. This scenario may be potentially dangerous,

3 In particular, it is possible to specify which GPIO channels are used for the External

Mode SCI connection. The corresponding pins cannot be used by other peripherals.
4 The number of words N
w required by the external mode can be computed as Nw =
Nsignals · 2 · (Nsamples + 1). In case more samples than those allowed by the memory alloca-
tion are required, Simulink® automatically truncates the scope traces every time Nsamples
points are plotted. It can be noted that cases of insufficient memory on the target result in
a build error.
External Mode Execution with Simulink® 383

since the MCU may keep executing the C-code with the last memorized state.
Based on Figure 20.9, it is particularly evident that the data samples are not
equally distributed everywhere. Extra samples may appear, implying that the
data exchange is affected by latencies. Therefore, this issue explains why this
kind of execution is suitable for debugging purposes, but it is recommended
to double check the results even with other evaluation tool.

Remark: it must be remembered that the XDS interface typically has two
serial interface channels, one for debugging and the other one for auxiliary
communication, including UART. If the External Mode connection to the
device is unsuccessful, a possible explanation is that the debug channel was
selected instead of the auxiliary communication channel during the COM port
setup.
 21
Electric Propulsion Case Studies

This last Chapter proposes two realistic cases study that deal with electric
propulsion on vehicles. No solution is reported in the following pages, since
the aim of those two exercises is to leave the reader free to organize the control
logic to check its understanding on the topics and concepts explained in the
previous Sections. Since those two exercises involve a tramway vehicle and
an electric car, no actual implementation on hardware is feasible for both
of them. However, the reader is invited to perform PIL simulations to test
the performance of the designed control strategy. The first case study deals
with a DC motor control. The reader is referred to Chapter 19 for all the
details of this control. Instead, the electric car involved in the second exercise
is controlled through an AC brushless drive. Such strategy is not explained
in the previous chapters of this book. However, this second exercise reports
some guidelines on how to deal with a synchronous motor and how to simplify
its modeling. Then, the control logic of such machines is not so different with
respect to that one adopted for PMDC motors.

21.1 Urban Tramway

Tramway vehicles “Carelli 1928” by ATM company (still in operation in Milan
and San Francisco, see Figure 21.1) are moved by four DC motors with series
excitation. Their main characteristics are reported here in the following:

• Line voltage: VDC = 600 V;

• Rated power of each motor: Pr = 21 kW;
• Maximum speed of the vehicle: vmax = 42 km/h;
• Rated speed of the motor: Ωr = Ωb = 970 rpm
• Efficiency: η = 0.9;
• Armature circuit time constant: τa = 10 ms;
• Mass of the vehicle at no load: mT = 15 t;

DOI: 10.1201/9781003196938-21 385

 386 Electric Propulsion Case Studies

Figure 21.1 A tramway vehicle “Carelli 1928” running in Milan.

• Maximum loading capacity: 130 passengers;

• Diameter of the wheel: d = 680 mm;
• Gearbox ratio (motor-to-wheels): ρ = 13/74;
• The equivalent friction force on the shaft is proportional to Ω and, at rated
speed, it is 1/10 of the traction force.
For the sake of simplicity, it is possible to assume that the four motors of
the tramway behave like one equivalent DC motor with separate excitation
having excitation rated voltage equal to Ver = 60 V, excitation rated current of
Ier = 5 A and excitation time constant τe = 0.1 s. This kind of machine differs
from the PMDC motor in the generation of the flux on the rotor. Indeed,
this flux is not constant any more and it can be regulated depending on the
working point of the machine. Namely, field weakening is regulated through a
dedicated current loop to exceed the base speed Ωb . The equation that should
be considered to tune the PI controller for the excitation is:
die (t)
ve (t) = Le + Re ie (t) (21.1)
dt
Thus, the back emf can be computed through the following equation:

e (t) = ke ie (t) Ω (t) (21.2)

and the equation for torque computations becomes:

me (t) = kT ie (t) ia (t) (21.3)

Even for a separately excited DC machine, it can be assumed that ke = kT =
k.
 Urban Tramway 387

Figure 21.2 Plots of the slope and speed profile of the track.

Based on this data:

• Design and simulate speed and current control loops in order to cover a 10 km
track considering the speed and slope profiles reported in Table 21.1 and in
Figure 21.2.1 The slope can be computed as 100 tan (ϑ), where ϑ is the angle
between the rising path and the ground, i.e., ϑ = arctan (slope %/100);
• Test the developed control logic on the microcontroller performing a PIL
simulation.

Table 21.1 Track profile for the tramway “Carrelli 1928.” The vehicle starts
from 0 speed and it stops at the end of the track.

Track (km) Slope % Speed

0−1 0 vr /2
1−3 0 vr
3−4 5 vr
4−6 0 vmax
6−8 0 vr
8−9 −5 vr
9 − 10 0 vr /2

1 In the reality, tramway vehicles go forward in coasting mode once they reach a steady

state speed. For the sake of simplicity, in this exercise it is possible to assume that the
traction torque is always applied on the motor to follow the speed profile with no error.
388 Electric Propulsion Case Studies

Example of the system parameters computations

Rated parameters of the motor
1. Overall mass of the vehicle loaded with all the passengers (standard
weight mP = 80 kg):

mtot = mT + mP NP = 25.4 t

2. Rated speed of the equivalent motor in rad/s:

2π
Ωr, rad/s = Ωr = 101.58 rad/s
60

3. Mechanical power of the equivalent motor:

Ptot = 4Pr = 84 kW

4. Electrical power absorbed by the equivalent DC motor:

Ptot
Ptot, el = = 93.33 kW
η

5. Rated speed of the vehicle in m/s:

d
vr, m/s = ρ Ωr, rad/s = 6.07 m/s
2

6. Friction power acting on the vehicle at rated speed:

1
Pfr,r = Ptot = 8.40 kW
10

7. Friction torque acting on the shaft of the equivalent motor at rated

speed:
Pfr,r
mfr,r = = 82.69 N m
Ωr, rad/s

8. Rated torque of the equivalent motor:

Ptot
mr = = 826.95 N m
Ωr, rad/s

9. Rated armature current of the equivalent motor:

Ptot, el
Ira = = 155.56 A
VDC
Urban Tramway 389

10. Rated back-emf acting on the armature circuit of the equivalent

motor:
Ptot
Er = = 540 V
Ira
11. Torque constant of the equivalent motor:
mr
K= = 1.06 Nm/A2
Ira Ire

12. Power loss associated to the equivalent armature winding:

Pl = Ptot, el − Ptot = 9.33 kW

Armature and excitation circuit parameters

1. Armature resistance of the equivalent motor:
Pl
Ra = 2
= 0.39 Ω
Ira

2. Armature inductance of the equivalent motor:

La = τa Ra = 3.86 mH

3. Excitation resistance of the equivalent motor:

Vre
Re = = 12 Ω
Ire

4. Excitation inductance of the equivalent motor:

Le = τe Re = 1.2 H

Mechanical parameters
1. Equivalent inertia of the vehicle seen on the motor shaft:
2
vr, m/s
Jeq = mtot = 90.62 kg m2
Ωr2

2. Friction coefficient2 :
mfr,r
β= = 0.81 N m s
Ωr

2 it can be assumed that the friction torque acting on the whole vehicle can be computed

as mfr (t) = βΩr (t)

 390 Electric Propulsion Case Studies

Figure 21.3 Example of an electric car during battery recharging [8].

21.2 Electric Racing Car

A full electric racing car3 like that one shown in Figure 21.3 is equipped with
a permanent-magnet synchronous motor (PMSM). It has a rated mechani-
cal power of 180 kW and it is fed by a battery stack working at 400 V. As
an example, see the car denoted in Figure 21.3. Its main characteristics are
reported here in the following:

• Efficiency η = 0.95;
• Stator resistance Rs = 0.1 Ω;
• Stator inductance Ls = 0.8 mΩ;
• Nominal speed Ωr = 400 rad/s;
• Pole pairs p = 2;
• Permanent-magnet flux φPM = 0.5 Wb.

Moreover, assume that:

• The motor control is able to keep cos ϕ = 1;

• The converter is operated through a PWM modulation with third harmonic
injection, allowing to reach a phase voltage with first harmonic peak value
2 VDC
equal to Vpeak = √ ;
3 2
• The car weights 1500 kg, whereas the driver 80 kg;
• The maximum achievable acceleration of the vehicle is amax = 4 m/s2 ;
3 The vehicle considered in this exercise is the electric counterpart of the Alfa Romeo

Giulietta Quadrifoglio Verde. Please, note that such electric car is does not exist and,
therefore, it is not available on the market.
 Electric Racing Car 391

(a)

(b)

Figure 21.4 Speed profile of the car running on the Varano circuit (also
available in Ex_E_car.mat file, variable: speed_kmh) (a) and correspond-
ing equivalent braking torque applied on the motor shaft before the gearbox
(also available in Ex_E_car.mat file, variable: T_braking) (b).

• The friction coefficient is β = 0.78 N m s, where the equivalent friction torque

acting on the motor shaft can be computed as mfr = βΩ;
• The radius of each tyre is r = 30 cm;
• The motor is connected to each axle through a gearbox with an equivalent
gear ratio equal to ρeq = 1/3;
• The car is running on the Varano circuit. Its speed profile is reported in
Figure 21.4 (a), whereas the corresponding mechanical braking torque is
shown in Figure 21.4 (b).
 392 Electric Propulsion Case Studies

Figure 21.5 Map of the track in Varano.

Knowing this data:

• Design a field-weakening strategy for the PMSM;

• Design and simulate a speed control of the motor to follow the speed profile
reported in Figure 21.4 (a) and considering the corresponding braking torque
applied on each axle reported in Figure 21.4 (b) (see the map of the track in
Figure 21.5). This latter may be due to both mechanical and electromagnetic
braking. The distribution of the braking torque on those two contributions
is left up to the reader;
• Test the developed control logic on the microcontroller performing a PIL
simulation.

Hint#1: linear model of a PMSM

Before briefly presenting the dynamical model of the PMSM, it is fundamental
to point out some assumptions to simplify this study. Namely, the machine to
be considered is perfectly symmetric (isotropic machine) with infinite iron per-
meability. In addition, the fundamental harmonic for each electrical variables
is considered only. In view of these two assumptions, it can be stated that the
electrical equations of the three-phase isotropic PMSM are the following4 :

va = Rs ia + pφa
vb = Rs ib + pφb
vc = Rs ic + pφc
φa = Lss (θm )ia + Mss (θm )ib + Mss (θm − 32 π)ic + φPM (θm )
φb = Mss (θm )ia + Lss (θm − 23 π)ib + Mss (θm + 23 π)ic + φPM (θm − 23 π)
φc = Mss (θm − 23 π)ia + Mss (θm + 23 π)ib + Lss (θm + 23 π)ic + φPM (θm + 23 π)

where p is the derivative operator, i.e., d/dt, Rs is the phase resistance, va,b,c
are the voltages applied on each phase of the motor, ia,b,c are the currents
flowing in each phase of the motor, φa,b,c stand for the magnetic flux linked
4 All the currents i
a, b, c, d, q , voltages va, b, c, d, q , fluxes φa, b, c, d, q , angles θm ,
speeds ωm and Ω, torque me and ml are in general function of time. In this paragraph,
this dependency on t is not made explicit for the sake of brevity.
Electric Racing Car 393
ea
Rs Lss

1A 2A 3A Mss ia
eb
Rs Lss
VDC + n
−
ib
ec
1B 2B 3B Rs Lss

N ic

Figure 21.6 Equivalent circuit of a three-phase PMSM connected with a three-

phase inverter.

with each winding of the motor, Lss is the phase inductance (equal for all the
windings), Mss is the mutual inductance (equal for all the magnetic couplings
between windings), φPM is the magnetic flux generated by the permanent
magnets and θm is the electrical position of the rotor. This angle differs from
the mechanical position of the rotor by a factor equal to np , that is the number
of pole pairs of the machine. Indeed, the periodicity of the electrical variables
is 1/np of the mechanical one. In addition, it is important to note that, in
general, all the phase voltages, currents, fluxes and the rotor position θm are
functions of time. The previous set of equations can be simplified reminding
that the motor adopted in this chapter is isotropic. Thus, the dependence of
the self and mutual inductances on the rotor position disappears. In addition,
the stator windings are connected in such a way that ia + ib + ic = 0 for every
considered time instant. Consequently, the new set of equations becomes:

va = Rs ia + pφa
vb = Rs ib + pφb
vc = Rs ic + pφc
φa = Ls ia + φPM (θm )
φb = Ls ib + φPM (θm − 23 π)
φc = Ls ic + φPM (θm + 23 π)

where Ls = Lss − Mss is the synchronous inductance. The corresponding

equivalent circuit of a PMSM connected to an inverter is shown in Figure
21.6. From this sketch and the previous equations, it is possible to note that
the voltage generators ea , eb and ec represent back emfs and they are due the
presence of the permanent magnets which spin together with the rotor.
Despite the fact that these new equations are far simpler than the previous
ones, they still do not allow the development of a general control strategy
in the phase variables, since they show a dependence on the rotor position.
Therefore, the Park transformation is introduced to obtain a new equivalent,
394 Electric Propulsion Case Studies

decoupled machine. This operation introduces a change in the reference frame

for the electrical equations, moving from a phase-based reference frame to a
new one with two orthogonal axis called direct and quadrature axis (d and
q) and a separate homopolar (0) axis. The typical structure of the electrical
machines is such that the homopolar components are nil for all the electrical
variables during their normal operation. Therefore, starting from a three-phase
motor, a new decoupled two-phase machine is obtained. This transformation
is defined for all the electrical variables as follows:

   
vd va
vq  = T (θm ) vb 
v0 vc
   
id ia
iq  = T (θm ) ib 
i0 ic
   
φd φa
φq  = T (θm ) φb 
φ0 φc

where the transformation matrix T (θm ) is defined as:

r  cos (θm ) cos θm − 23 π

cos θm + 23 π



2
T (θm ) = − sin√(θm ) − sin θm
√− 3π
2
− sin θm
√+ 3π
2
(21.4)
3

1/ 2 1/ 2 1/ 2

(2/3) is a coefficient that ensures the orthogonality of the transformation.

p

Then, it can be noted that this operation allows to define a number of new
reference frames, since angle θm can be chosen arbitrarily. For convenience,
a set of transformed axes which rotates synchronously with the rotor of the
machine under study is adopted commonly. In this way, all the AC electrical
variables appear as constants at steady state. Namely, the new transformed
electrical equations of a PMSM are:

vd = Rs id + pφd − np Ωφq
vq = Rs iq + pφq + np Ωφd
φd = Ls id + φPM
φq = Ls iq

where Ω is the mechanical speed of the rotor and the equations for the hom-
polar components of the voltage and the flux are not reported since they are
both nil. It can be noted that the electrical speed of the rotor and of all the
electrical variables can be defined as ωm = np Ω. Moreover, space vectors can
Electric Racing Car 395

be defined using complex numbers. The d-axis component is considered as the

real part of the space vector, whereas the projection on the q-axis becomes
its imaginary part. Namely, the electrical model can be rewritten in a more
compact form as follows:

vs = vd + jvq = Rs is + pφs + jnp Ωφs

φs = φd + jφq = Ls is + φPM

From this compact notation, the term jnp Ωφs can be highlighted. This is
a motional term which is a consequence of the adoption of a moving reference
frame and it is responsible for the electromechanical conversion of the machine.
Moreover, combining the voltage and flux equations, the following dynamical
expressions can be obtained:

vd = Rs id + Ls pid − np ΩLs iq
vq = Rs iq + Ls piq + np ΩLs id + np ΩφPM

where np ΩLs iq and np ΩLs id are cross coupling terms between the two axis
and np ΩφPM = E is the back-emf expressed in the d-q reference frame. Finally,
the torque expression and the mechanical equation are presented. The first one
is obtained by setting an energy balance between the electrical and mechanical
part of the system. For isotropic machines, it holds that:

me = np φPM iq (21.5)

where me is the electrical torque generated by the PMSM. Instead, the dy-
namical equations of the mechanical part of the system is obtained from a
torque balance as follows:

Jeq pΩ + βΩ = me − ml (21.6)

where ml is the load torque applied in the shaft of the machine, Jeq is the
equivalent inertia of the PMSM and β is the friction coefficient.

Hint#2: control strategy

The control strategy for PMSM is the vector control for AC brushless ma-
chines. As previously mentioned, it is possible to act separately on the torque
and on the flux of the machine, thanks to the Park transformation. This pecu-
liarity represents the key of the vector control, which has a similar philosophy
with respect to the strategy adopted for DC machines. Indeed, the two trans-
formed windings can be considered separately. One is used to generate torque
396 Electric Propulsion Case Studies
Ω ∗ (t)
+ speed m∗e (t) 1 i∗q (t)
+ current u∗q (t)
− controller np φPM + va∗ (t)
− controller + vq∗ (t)
+ vb∗ (t) power
+ T−1 (θm ) stage PMSM
vc∗ (t)

np φPM np Ls 2L-VSC +
modulator

i∗d (t)
+ current u∗d (t)
+ vd∗ (t)
− controller
−

np Ls
ia (t)
iq (t)
ib (t)
id (t) T (θm )
ic (t)

Ω(t)

Figure 21.7 Block diagram of a generic PI based vector control for a PMSM.

and, consequently, for regulating the speed of the rotor, whereas the other one
is exploited to provide a variation in the magnetic field. For PMSM, this last
operation is done by generating currents in the stator windings that partially
demagnetize the magnetic field due to the magnets φPM . Carefulness must be
paid while demagnetizing real PMSM since too high current may compromise
the magnets.
Focusing on the realization of the vector control, the speed regulation is
implemented in an external loop and it is cascaded with the current control
acting on iq . Indeed, this current is responsible for torque variations (see
equation 21.5). The other current control acting on id can be considered as a
loop which acts in parallel to iq . A PI-based scheme of the vector control is
reported in Figure 21.7.
For Ω lower than the base speed irefd is set equal to 0. This choice allows
the quadrature current iq to reach the maximum sustainable current of the
machine imax . Above Ωb , the value of irefd is decreased (iref
d < 0) until it
reaches the value −φPM /Ls . Once this limit is reached, the d-axis current is
kept constant. On top of that, compensation terms should be added in the
control scheme based to the model of the machine previously shown.

Speed Controller
This PI controller is tuned based on the mechanical transfer function of the
PMSM. This latter is retrieved starting from equation 21.6, that relates the
mechanical speed with the electrical torque, and moving to the Laplace do-
main. The resulting transfer functions GΩ (s) reads as:
1
GΩ (s) = (21.7)
β + Jeq s
Electric Racing Car 397

Hence, considering the classical parallel structure for PI controllers re-

ported in Section 5.2, the open loop transfer function is:

kpΩ s + kiΩ 1
LΩ (s) = RΩ (s)GΩ (s) = (21.8)
s β + Jeq s
This transfer function shows two poles in zero. Therefore, in order to guar-
antee asymptotic stability, it is crucial to properly tune kpΩ and kiΩ . From the
study of L(s), both terms kpΩ can be directly linked to the bandwidth of the
controlled system ωΩ through the following equations:

kpΩ = Jeq ωΩ (21.9)

kiΩ = βωΩ (21.10)

Current Controllers
The tuning procedure of the current controllers is based on the same electri-
cal transfer function of the PMSM. Indeed, the transformed machine in the
Laplace domain, it can be noted that the resulting transfer function has the
same structure of that one adopted for the same control in a DC machine. As
a matter of fact, the cross-coupling terms and the back-emf can be dealt with
as if they were disturbances. Namely, for both d and q axes, it holds that the
electrical transfer function is:
1
Gi (s) = (21.11)
Rs + sLs
Therefore, the open loop transfer function reads as:

kpi s + kii 1
Li (s) = Ri (s)Gi (s) = (21.12)
s Rs + sLs
As for the Speed PI controller, a relationship between the gains of the
controller and the bandwidth of the system ωi can be written as:

kpi = Ls ωi (21.13)

kii = Rs ωi (21.14)
Due to the symmetry in the transformed machine, both the current con-
trollers can be tuned with the same parameters. In addition, it must be remem-
bered to select a value at least ten time larger with respect to ωΩ to fulfill the
cascade constraint for nested loops. Indeed, the speed loop is cascaded with
the current one, as shown in Figure 21.7.
398 Electric Propulsion Case Studies

Hint#3: acceleration
The motor can start from 0 speed and accelerate up to the initial speed of the
profile shown in Figure 21.4 (a) or an initial speed can be assigned.
Set suitable torque limits and separate the mechanical torque from the elec-
tromagnetic one.
Use the block Saturation Dynamic from library Discontinuities to imple-
ment the field weakening logic.

Hint#4: computation of the simulation parameters

Rated parameters of the motor
1. Overall mass of the car and driver:

mtot = mT + mD = 1580 kg

2. Electrical power absorbed by the PMSM:

Pr
Pel = = 189.47 kW
η

3. Rated stator current (peak value):

2Pel
Irpeak = = 546.96 A
3Vpeak cos ϕ

4. Rated motor torque:

Pr
mr = = 450 N m
Ωr

Mechanical parameters
1. Equivalent inertia of the vehicle seen on the motor shaft:
2
Jeq = mtot (rρeq ) = 142.20 kg m2

2. Maximum motor torque:

mmax = mtot amax rρeq = 632 N m

3. Maximum quadrature-axis current:

mmax
iqmax = = 665.26 A
ηpφPM
Electric Racing Car 399

4. Base speed of the motor:

v
u  2
3/2
p
1 V
u
u peak
Ωb = u 2 = 192.95 rad/s
p t 
φ2PM + L2s Irpeak 3/2
p

5. Motor speed at which the direct-axis current reaches its maximum

value:
v
u  2
Vpeak 3/2
p
1
u
u
Ω∗ = u 2 = 733.21 rad/s
pt 2

−φPM + L2s Irpeak 3/2
p

6. Minimum direct-axis current:

φPM
idmax = − = −625 A
Ls
A
Appendix A: Basics of C

In this appendix, some basics of C programming which are useful to under-

stand many operations hidden by the Simulink® code generation are recalled.

A.1 Operations between numbers

A.1.1 Sum and differences
Sum and differences do not pose particular problems as regard the base system
representation, provided that the base used for the conversion has a lower value
of bits with respect to the actual base implemented in the controller (e.g., if
for the conversion it has been considered a 12-bit base, the actual type of the
variable in the C code can be chosen equal to 16-bits). In this case, the sum
of two variables represented in 12-bit base will not produce an overflow in the
result.

A.1.2 Shift operation

Consists in moving a to the left (<<) or to the right (>>) a sequence of bits
by a certain number of positions. This type of operation is much faster to be
performed for microcontrollers and it can be used to perform multiplication
between variables and constants in a faster way with respect to the same op-
erations with floating point structures: this fact holds provided that variables
are represented in integer form, with a base system multiple of 2. If a variable i
(e.g., a current) whose maximum value is 100 represented by means of uint16
numeric structure is considered, a 12-bit base is chosen (212 = 4096). Thus:
i
#I12 = · 4096
100
Assuming that the actual value of the current is 75, its representation in
the 12-bit base is:
75
#I12 = · 4096 = 3072
100
whose corresponding structure in the uint16 format is 0b0000110000000000.

DOI: 10.1201/9781003196938-A 401

402 Appendix A: Basics of C

If this variable has to be multiplied (or divided) by 2, this operation can be

obtained in a simple way by means of a left shit << by one bit (or on the
right >> for the division) of the corresponding bit sequence.
Thus, the quantity (2 · #I12 ) can be simply obtained in the microcontroller
as a shift operation 2 · #I12 = #I12 >> 2. Any multiplication (or division) by
a constant quantity can be expressed exploiting the representation in power
of 2. In case the variable #I12 has to be multiplied by π, the first step is to
express the constant coefficient as a power-two series:

#I12 · 3.1416 ∼
= #I12 · (21 + 2−1 + 2−3 + 2−6 + 2−11 )

Then, the shift properties can be exploited, making the calculation of the
product very fast. Since the shift operations have lower priority with respect
to the sum, parenthesis are required between the terms of the summation in
order to avoid errors:
#I12 · 3.1416 ∼
=(#I12 << 1) + (#I12 >> 1) + (#I12 >> 3)+
+ (#I12 >> 6) + (#I12 >> 11)

A.1.3 Multiplication
The multiplication between two variables in the same base can be written as:
#Abitbase · #Bbitbase
#Cbitbase = #(A · B)bitbase =
2bitbase
By expressing the constant division with the shift operator, it becomes:

#Cbitbase = #(A · B)bitbase = (#Abitbase · #Bbitbase ) >> bitbase

A.1.4 Division
The division of two variables follows the same method introduced for the
multiplication and it can be expressed as:
#Abitbase 1
#Cbitbase = #(A/B)bitbase = · bitbase
#Bbitbase 2
The above expression is equivalent to:
#Abitbase << bitbase
#Cbitbase = #(A/B)bitbase =
#Bbitbase

Example: Product operation in C

The base value to be consider has Nbit = 16 bit and the processor framework
is 16 bit too. Products are computation without any tolerance bandwidth.
Structure of a C program 403

The strategy to avoid overflow is to temporary move into 32 bit (long) vari-
able through left shifting, making the product computation, and then come
back to 16 bit integers. The corresponding C code is as follows:
. .
#x16 = (int) ((long)#xa,12 · #xb,12 )  4 (A.1)
. .

A.2 Structure of a C program

C-programs for micro-controller applications are generally characterized by a
constant structure. Two main types of files contribute to the definition of the
code: header files (denoted by the extension .h) and main program files (whose
extension is .c). In header files functions, constants and macros are generally
defined, while in the .c file all the aforementioned structures are called and
combined together to perform the desired operations.

The inclusion of an header file into the main program is performed by means
of the following piece of code placed at the beginning of the .c code:

#include filename.h

In order to have global visibility for both functions and variables set inside
any header file, it is necessary to define them as extern.

extern int error; // Variable

extern void function1(int,double); // Function

Alternatively, it is possible to define constant and macros with the follow-

ing syntax:

#define CONST 1.0341 // Constant

#define SUM(e1, e2) (e1 + e2) // Macro

Macros are predefined pieces of code that are are substituted to the corre-
sponding invocation before code execution. In practice, a macro is a fragment
of code which has been given a name. Whenever the name is used, it is re-
placed by the contents of the macro. Macros are useful in order to speed up
the coding by avoiding multiple writing of the same pieces of code. Macros
can be invoked in the main .c file with a syntax similar to the one adopted
for the functions:

c = SUM(a,b) // -> c = (a + b)
404 Appendix A: Basics of C

Macros are also useful for simplifying the coding of multiplications between
a variable and a constant. Recalling the properties of the shift, as defined in
Section 7.3 and previously in this appendix, it is possible to express the mul-
tiplication of any variable by π with the following macro:

#define TIMES_PI(e) ((e<<1) + (e>>1) + (e>>3) + (e>>6) + (e>>11))

A typical structure of a C file is reported here below:

//Inclusion of the compiler file, depending on the micro

processor
#include <p30fxxxx.h>
//Inclusion of the .h file with used-defined structures
#include “localfile.h”
// Macro definition
#define MUL2(in)(((int)in<<1))
// Global variable definition long int error, input, output;
//static long int
//Main function
int main () {
//do something... while(1) {
//infinite loop... do something } }
//Interrupt routine (depending on the type of the MCU)
void _ISR _T1Interrupt()
{ //do something... }

In this example, it is possible to see the difference between normal variables

and static ones: while normal variables are deleted each time the code is re-
executed (that is, at the end of any complete cycle of the control), static
ones retain their value in the local memory until a re-assigning operation is
performed.
 B
Appendix B: Custom Expansion Boards and
Hardware Kits

In this Appendix, the main characteristics of the custom expansion boards and
assembled kits targeted to the exercises proposed previously are illustrated.
As university initiative, the boards have been developed at the Laboratory
of Electrical Drives and Power Electronics, in the Department of Mechani-
cal Engineering of Politecnico di Milano, Italy, in collaboration with Würth
Elektronik™ Group and Texas Instruments™ Inc. The aim of this collabora-
tion is to provide to students ready-to-use test benches for the study of power
electronics-based applications. In particular, the developed boards and test
benches are:
• extPot3 board, which provides a direct manipulation of analog signals,
conversion chains, and buzzer;
• extRL(C) board, which is a configurable load equipped with output filter.
It is able to realize RL, RLC, R, L, LC topologies;
• RL(C) kit, which integrates the extRL(C) board with the LaunchPad™
F28069M and the BOOSTXL-DRV8301 BoosterPack boards;
• DecMot kit, which allows to test PMDC motors. An encoder sensor is
included as well as MCU interface for external power supply. Other kind of
motors can be installed and driven by the same hardware as well. Indeed,
the housing of this setup is suitable for two generic small motors;
• B2B-PMDC kit, which contains two coupled PMDC motors anchored on
an aluminum base plate, encoder sensor and MCU interface for external
power supply.
It is important to note that, according to the considered application, boards/k-
its need to be manually configured, e.g., topology settings, motor calibration
and cable connections should be edited.

Either custom expansion boards and assembled kits are open to users com-
munity. Project files can be shared and/or pre-assembled boards/kit can be
directly shipped. In case of interest, please contact:
- Mattia Rossi [email protected]
- Nicola Toscani [email protected]
- Francesco Castelli Dezza [email protected]

DOI: 10.1201/9781003196938-B 405

406 Appendix B: Custom Expansion Boards and Hardware Kits

Figure B.1 extPot3 board. Front and back views.

extPot3 Board
The extPot3 board (see Figure B.1) is a custom-made expansion board which
aims to drive three ADC channels of the LaunchPad™ F28069M board.
Namely, ADCA3, which is connected to pin 66, ADCB3 which is connected to
pin 67 and ADCB5 which is connected to pin 68. These channels are connected
to three 20 kΩ linear potentiometers (Pot1, Pot2 and Pot3), which are able to
generate analog signals in the range [0, 3.3] V at the ADC inputs. The extPot3
board can be directly mounted on top of the 40-pin plug-in module connector
(headers J5-J8) of the LaunchPad™ F28069M board. In addition, the board
includes a piezoelectric buzzer. This device can be driven by a generic GPIO
or by a ePMW, namely GPIO8 (which is connected to pin 78) and ePWM5B
(which is connected to pin 77). A bipolar transistor BJT NPN 25 V 0.5 A is
used as buzzer driver to decouple the driving signal from the actuation stage.
The extPot3 board mounts a six pins terminal header arranged to connect
hall-effect sensors through pull-up resistors. Starting from the left-hand side
of the board (front view of the board, see Figure B.1), the first three pins
corresponds to three phases for hall-effect sensor measurements, while the
other three to 3.3 V, GND, and 5 V supply, respectively. A SMD green led turns
on when the LaunchPad™ F28069M board is supplied with 3.3 V, verifying
the correct connection between boards (the 3.3V supply of the extPot3 are
taken from the MCU board).
As previously mentioned, the extPot3 board is connected to the
LaunchPad™ F28069M board through standard 40-pin plug-in module con-
nector. These pins remains accessible for debugging purpose or further connec-
tions. The maximum height of a BoosterPack is 1700 mil. This value becomes
1350 mil if there is no need for accessing header J5. Instead, there is no detailed
standard for the inner 20 pins (i.e., header J3 and J4). extPot3 board must
be attached with care to the LaunchXL F28069M one to ensure a successful
stacking with the various pins and to improve device lifetime as well.
 extPot3 Board 407

All these pins can be used as GPIOs except VCC, GND, TEST, and RE-
SET. It can be noted that most of the LaunchPad™ F28069M board pins are
multiplexed for dual functionality. Thus, attention must be paid to check if
any of the ADC pin used by the extPot3 (which is physically connected to one
of the linear potetniometers) is reconfigured. Therefore, it is recommended to
keep the default pin mapping pins unless it does not restrict the firmware
design.
The Bill of Material of the extPot3 board is reported in the following table:

Designator Description Manufacturer Part Number

Pot1, Pot2, Pot3 20 kΩ Potentiometer Alps Alpine RK11K1120A5T

J1, J2 Launchpad connectors SparkFun PRT-11376
J3, J4, J5 Launchpad connectors SparkFun PRT-11376
J6 Spring clamp terminals Würth Elektronik 691401710006B
ENABLE Enable header Würth Elektronik 61300211121
J8 Signal selector Würth Elektronik 61300311121
R1, R2, R3, R4 2.2 kΩ, SMD 0805 Vishay CRCW08052K20FKEA
R5 1 kΩ, SMD 0805 Kamaya RMC1/10K102FTP
B1 Buzzer RS PRO 617-3081
Q1 Buzzer driver Infineon BC818K40E6327HTSA1
D1 Green led Würth Elektronik 150141GS73100
TP (10x) Test point RS PRO 262-2185

Key Features
• Manipulation of analog signals through linear potentiometers;
• Piezoelectric buzzer;
• Pins of LaunchPad™ F28069M board are still accessible;
• Suitable (i.e., without conflicts) for TI BOOSTXL-DRV8301 converter board
top mouting.

Test Points
Many test points are available to debug the LaunchXL F28069M board pins:
• TP1: connected to ENC A of QUEP_A;
• TP2: connected to ENC B of QUEP_A;
• TP3: connected to ENC I of QUEP_A;
• TP4: connected to 5 V of QUEP_A;
• TP5: connected to GND of QUEP_A;
 408 Appendix B: Custom Expansion Boards and Hardware Kits

• TP6: connected to ENC A of QUEP_B;

• TP7: connected to ENC B of QUEP_B;

• TP8: connected to ENC I of QUEP_B;
• TP9: connected to 5 V of QUEP_B;
• TP10: connected to GND of QUEP_B.

Jumpers
• J8: signal selector from GPIO or ePWM connected to the buzzer;
• ENABLE: to be connected to enable QEP_B functions when using TI
BOOSTXL-DRV8323RS converter board.

Connectors
• J6: hall-effect sensor terminal block.

Hint: PCB Design Rules

For custom board, it can be noted that trace width can be estimated through
the following steps:
• Area computation:

current [A]
area mil2 =  (B.1)
 
 1c
b
k · temp.rise [◦C]

• Width computation:

area mil2
 
width [mil] = (B.2)
1.378 [mil/oz] · thickness [oz]

• For IPC-2221 internal layers: k = 0.024, b = 0.44, c = 0.725

• For IPC-2221 external layers: k = 0.048, b = 0.44, c = 0.725
where k, b, and c are numerical constants resulting from curve fitting to the
IPC-2221 curves. For further information on this topic the reader is referenced
to https://www.ipc.org and https://www.digikey.it/en/resources/on
line-conversion-calculators.
 extRL(C) Board 409

extRL(C) Board
The extRL(C) board (see Figure B.2) is an configurable load which is able
to combine different resistors R, inductors L and capacitors C, thus, allow-
ing to operate the system with several equivalent values of load resistance,
inductance and capacitance. This board is suitable for both operation in DC
or AC. The board includes dedicated current and voltage sensors as well as
an additional two-level converter leg with dedicated gate driver. The latter is
useful if no external converter board is used (e.g. the TI BOOSTXL-DRV8301
BoosterPack is not present).
The main feature of the extRL(C) load is the possibility to dynamically
change parameters values and load topology through three switches mounted
at PCB top and one at PCB bottom. Combining the extRL(C) with the TI
BOOSTXL-DRV8301 BoosterPack, the user is able to operate the system
with both Half or Full-Bridge DC-DC converter, with single-phase loads and
performing measurements comparison between the extRL(C) and BOOSTXL-
DRV8301 on-board sensors.
The details of the RLC load are reported here in the following:
• The available resistance values R are 6.8 Ω, 3.4 Ω and 0 Ω (no connected
resistors). Such value changes are made by acting on switch S1, which can:
– Put in parallel two 6.8 Ω resistors (decreasing the overall/equivalent
load resistance);
– Keep connected just one of the two 6.8 Ω resistors;
– Disconnect the passive load (switch S4 off is also required).
The adopted resistances are two Arcol HS50 Series Aluminum Housed Axial
Wire Wound Panel Mount Resistor, with nominal value 6.8 Ω±5% and rated
power 50 W.

• The available capacitance values C are 100 µF, 10 µF and 0 F (no connected
capacitor). Such value change is made by acting on switch S2, which can
connect one of the available capacitors at a time or by-pass the capacitor
stage (i.e., removing the C from the load). Electrolytic capacitors are used.
Their characteristics are reported here below:
– WCAP-ATET: Aluminum Electrolytic Capacitor, Radial, THT,
D8xH11.5mm, rated capacitance 10 µF ± 20%, rated voltage 100 V;
– WCAP-ATET: Aluminum Electrolytic Capacitor, Radial, THT,
D13xH25mm, rated capacitance 100 µF ± 20%, rated voltage 100 V.

The equivalent series-resistance of the capacitor (ESR) can be computed

from the ESR-f characteristic given in the datasheet, e.g. ESR = Rc = 66 mΩ
 410 Appendix B: Custom Expansion Boards and Hardware Kits

Figure B.2 Different views of the extRL(C) board.

at fsw = 10 kHz. By disconnecting the capacitor, the extRL(C) can operate

as single-phase RL load in AC (besides DC).
• The available inductance values L are 860 µH, 1.72 mH and 3.44 mH. Such
value change is made by acting on switch S3, which can:
– Include just one 860 µH inductor;
– Connect in series two 860 µH inductors,
– Connect in series four 860 µH inductors.
The adopted inductors are four WE-FI Leaded Toroidal Line Choke, with
rated inductance 860 µH ± 20%, rated current 3 A and series resistance R` =
150 mΩ at 20 ◦C.

The extRL(C) board allows to operate with a internal or external converter

(i.e., Half/Full-Bridge) through the switch called MODE, placed at PCB top.
In external mode, the legs of the TI BOOSTXL-DRV8301 BoosterPack are
used as the switching stage, with the extRL(C) board used as configurable
filter stage interfacing a passive resistive load. Half- or Full-Bridge converter
 extRL(C) Board 411

configurations are possible. The generation of the gating signals is performed

through a firmware operating on the LaunchPad™ F28069M board. Note that
the diode D1 must not be connected to operate in Full-Bridge configuration.
In internal mode, a dedicated switching stage, which consists in a Half-
Bridge converter, is used. This is a BTN8962TA by Infineon (aimed to motor
drive/automotive applications). Thus, no BOOSTXL-DRV8301 converter is
not needed. In particular, the BTN8962TA can be driven by:

• gating signals coming from the LaunchPad™ F28069M board, e.g., ePWM1A
and 1B. This option is called PWMext;
• gating signals provided by a timer blox monostable pulse generator inte-
grated circuit LTC6992 by Analog Devices (mounted on the PCB bottom).
This option is called PWMint.

The LTC6992 component allows to generate a PWM signal by simply control-

ling a [0, 1] V analog input (the IC is powered at 3.3 V). To this end, a 10 Ω
trimmer is used to vary the trigger values (i.e., the duty cycle) at pin 1.
The auxiliary circuits of the board are powered through the spring clamp
connectors EXT and SUPPLY. A fixed step-down regulator module and a
linear low drop voltage regulator create different voltage levels (5 V, 3.3 V and
1.65 V) to satisfy the circuit inputs.
The extRL(C) board is equipped withe a ACS711ELCTR-12AB-T current
sensor from Allegro. This sensor can operate at 3.3 V or 5 V, with a sensitivity
of 110 mV/A for the voltage output value and a bandwidth of 100 kHz. This
device consists in a linear hall-effect sensor-based circuit with a copper con-
duction path located near the surface of the die. The current flows through this
copper conduction line that generates a magnetic field which is sensed by the
integrated hall-effect sensor-based circuit and converted into a proportional
voltage. In addition, the extRL(C) board is also equipped with an isolated
voltage sensing circuit. This is a voltage divider plus a rail-to-rail CMOS op-
erational amplifiers OPA350EA operating in a [0, 3.3] V range. The circuit can
be used both in DC or AC operations. If the voltage sense is not required, e.g.,
when the capacitor is disconnected, this stage is by-passed through two test
points called BP. Both current and voltage sensing path are provided with RC
signal filters at the outputs of the related ICs. According to the application,
these filters may requires some adjustments by the user, e.g., variations in the
R and/or C values. PWMext and measurement feedbacks can be accessed by
the lateral header (PCB top) and sent to the LaunchPad™ F28069M board.
Based on the whole parameters limits, the extRL(C) board is designed
to operate at 3 A 40 V at rated conditions. Even if this values can be tempo-
rary exceeded, the board incorporates a over-current protection (OCP) circuit
based on 4 A WE fuse and an over-voltage protection circuit based on zener
diodes.
Due to the high current and voltage values, the adopted aluminum housed
resistors are separated from the PCB and placed in contact with a heat sink,
412 Appendix B: Custom Expansion Boards and Hardware Kits

Figure B.3 extRL(C) board features printed on the PCB top surface.

since those are the components heating the most in this load stage. The heat
sink ensures a sufficient temperature drop which allows to keep the extRL(C)
board on a (student) desk. The (non-scaled) voltage applied across the resis-
tance(s) is accessible from the screws placed on the top of the PCB.
A summary of the extRL(C) features is reported in the scheme printed on
the PCB top surface, which is shown in Figure B.3. The Bill of Material of
the extRL(C) Board is reported here in the following:

Designator Description Manufacturer Part Number

POT 10 kΩ PWM Trimmer Bourns 3296W-1-103LF

Q1 High Current PN (HB) Infineon BTN8962TA
EXT, SUPPLY Spring clamp terminals Würth Elektronik 691401710002B
Vo+, V- REDCUBE Pressfit Würth Elektronik 7461096
FDSM Fixed Step Down
U1 Würth Elektronik 173010535
Regulator Module
Linear low drop
U2 STmicroelectronic LD1117S33TR
voltage regulator
U3 Pulse Generator Analog Devices LTC6992CS6
U4 Current sensor 12AB Allegro ACS711ELCTR
rail-to-rail CMOS
U5 Texas Instruments OPA350EA_250
operational amplifier
J1,J2,J3 External connectors Würth Elektronik 61300211121
PWM PWM selector Würth Elektronik 61300311121
PWM_EXT external PWM header Würth Elektronik 61300211121
MODE Mode Switch C&K L202011MS02Q
SWITCH L Impedance Switch C&K L103111MS02Q
SWITCH R Resistance Switch C&K L202011MS02Q
SWITCH C Capacitor Switch C&K L103111MS02Q
S1 Switch C&K 1101M2S3CQE2
Leaded Toroidal
L1, L2, L3, L4 Würth Elektronik 7447075
Line Choke, 860 µH
C12 Elec. Capacitor 100 µF Würth Elektronik 860130878011
 extRL(C) Board 413

C13 Elec. Capacitor 10 µF Würth Elektronik 860130874006

R18, R19 6.8 Ω Resistance Arcol HS50 6R8 J
D1 Diode Comchip tech. CDBB5100-HF
D2 Diode Comchip tech. CDBB5100-HF
R1 40 kΩ, SMD 0805 Vishay CRCW080540K2
R2 309 kΩ, SMD 0805 RS PRO RS-0805-309K-1
R3 3.48 kΩ, SMD 0805 TE Connectivity CPF0805B3K48E1
R4 976 kΩ, SMD 0805 Panasonic ERJP06F9763V
R5 182 kΩ, SMD 0805 Vishay CRCW0805182K
R6, R8, R9 10 kΩ, SMD 0805 TE Connectivity CRG0805F10K
R7 33 kΩ, SMD 0805 KOA RK73H2ATTD3302
R11, R12, R13, R14 1 MΩ, SMD 0805 ROHM MCR10EZPF1004
R15, R16 11 kΩ, SMD 0805 Panasonic ERJU06F1102V
R17 68 Ω, SMD 0805 Panasonic ERJU06F68R0V
R20 15 kΩ, SMD 0805 TE Connectivity CRGH0805F15K
R21, R101 1 kΩ, SMD 0805 Kamaya RMC1/10K102FTP
C1, C2, C4, C5
100 nF, SMD 0805 Würth Elektronik 885012207128
C8, C9, C11
C3 10 µF, SMD 0805 Würth Elektronik 885012207026
C62 220 nF, SMD 0805 Würth Elektronik 885012207074
C71 47 nF, SMD 0805 Würth Elektronik 885012207126
C10 1 µF, SMD 0805 Würth Elektronik 885012207022
Fuse (3 A) RS PRO 563-542A
F1
Fuseholder Würth Elektronik 696101000002
PWM, Vext+, Vext-
Vo+, V-, Vdc, V+, GND Test point RS PRO 262-2185
3v3, 5v, GND, bp, bp
External heat sink Heat Sink Fischer Elektronik SK58-50-SA

Key Features
• Different resistance, inductance, capacitance values can be set on-the-fly;
• Choose between internal or external switching stage;

• Choose between internal or external PWM signals generation when the in-
ternal Half-Bridge converter is used;
• Over-current and over-voltage protections;
• On-board high-accuracy current sensor;

1 Varying values depending on the desired filtering action.

2 To be decided components
 414 Appendix B: Custom Expansion Boards and Hardware Kits

• On-board isolated voltage sensing;

• Both DC or AC operation;
• Compact size;
• Nice handling and pin access.

Test Points
• PWM: internal pulse width modulation;
• Vext+: positive pole of external input voltage;
• Vext-: negative pole of external input voltage;
• Vo+: positive pole of output voltage;
• V-: negative pole of converter output voltage (before L);
• Vdc: positive pole of input voltage;
• V+: positive pole of converter output voltage (before L);
• GND: ground;
• 3v3: 3.3 V connected to output pin of linear (low-drop) voltage regulator;
• 5v: 5 V connected to output pin of FDSM fixed step down regulator module;
• GND: ground;
• bp: bypass OPA350EA;
• bp: bypass OPA350EA.

Jumpers
• PWM: selection from internal or external PWM generator circuit.

Connectors
• EXT: Half-Bridge/Full-Bridge external power supply;
• SUPPLY: Half-Bridge internal power supply;
• J2: microcontroller current connector;
• J3: microcontroller voltage connector;
• PWM_EXT: external PWM genertor connector;
• Vo+: positive pole of output voltage;
• V-: negative pole of output voltage.
 RL(C) Hardware Kit 415

Figure B.4 Different views of the RL(C) kit and its connections.

RL(C) Hardware Kit

The hardware kit shown in Figure B.4 includes the following devices:
• One LaunchPad™ F28069M board;
• One TI BOOSTXL-DRV8301 BoosterPack;
• One extPot3 board;
 416 Appendix B: Custom Expansion Boards and Hardware Kits

Figure B.5 Closed-loop dynamics measured from the F28069M pins through
an oscilloscope (Full-Bridge operating with unipolar voltage switching).

• A mezzanine board to hold the MCU and manage the external power supply;
• One extRL(C) board.
This kit is a complete development test-bench for the energy management
of power topologies and passive loads. Digital power management and con-
trol capabilities provided by rapid prototyping allows the study and design of
adaptable high frequency switching power supplies, e.g. targeting both power
processing and EV on-board battery charge applications. This kit is mainly
built around the extRL(C) which was previously presented. All the afore-
mentioned features holds. The mezzanine card is a MCU housing based on a
custom made design. It can be even separately used as support for the MCU
board. The kit is designed to operate with 40 V 3 A input source at rated
conditions. Depending on the considered application target, those values can
be temporary exceeded. A usage example of this board is reported below.

Example: RL Current Control in Full-Bridge Configuration

A Full-Bridge configuration (or H-bridge) refers to use two legs of the adopted
TI BOOSTXL-DRV8301 (legs A and B), for which two ePWM modules are
required, e.g., ePWM1 and ePWM2. The design of a PI-based current control
for an RL load (i.e., extRL(C) configured in RL mode) is considered in the
Simulink® environment. The terminals of the extRL(C) board are connected to
the central point of legs A and B (i.e., the boosterpack switches 1A , 1B and 2A ,
2B are operated). Based on to Chapter 17, the actuation of the two legs (i.e.,
two duty cycles for two modulation stages) can be done by manipulating one
modulation signal d(k) only, from which the gating signals of the four switches
are generated. Since Chapter 15 already discussed the implementation and
 DecMot Hardware Kit 417

hardware settings of such strategy as well as the differences between bipolar

and unipolar voltage switching, further details are not reported here. A benefit
of a H-bridge configuration with respect to the Half-Bridge one is the chance to
regulate even negative currents without changing the electrical connections of
the setup. As an example, a bipolar current reference profile may be internally
generated in the LaunchPad™ F28069M board in addition to the closed-loop
scheme. Figure B.5 denotes the command tracking performances of a Full-
Bridge configuration driving an RL load realized through the extRL(C) board
and operating with unipolar voltage switching. The waveforms are acquired
with an oscilloscope.

DecMot Hardware Kit

The hardware kit shown in Figure B.7 includes the following devices:
• One LaunchPad™ F28069M board;
• Two TI BOOSTXL-DRV8301 BoosterPacks;
• One extPot3 board;
• A mezzanine board to hold the MCU and manage the external power supply;
• One PMDC motor;
• One brushless DC (BLDC) motor;
• Encoder LPD3806-600BM-G5-24C;
• Aluminum base plate, motor supports and a joint;
This kit is a complete motor control development test-bench aimed to test sin-
gle motor. Both PMDC and BLDC motors installed in this configuration can
be separately tested with sensored (using an encoder) or sensorless control
schemes, targeting automotive and servo-drive applications. The mezzanine
card is a MCU housing based on a custom made design. It can be even sepa-
rately used as support for the MCU board. The kit is designed to work with
24 V 10 A at rated conditions; according to the considered motor, these values
can be adjusted in the external power supply.

Encoder Connection
The encoder LPD3806-600BM-G5-24C adopted for the exercise proposed in
this book (see Chapter 14) and in the DecMot kit see an encoder interface
made by open collector outputs with 20 Ω resistors for protection. This cir-
cuit draws around 30 mA. Since it can not source current, the open-collector
 418 Appendix B: Custom Expansion Boards and Hardware Kits
+Vin +5 V +5 V
OUT

7805 4.7 kΩ 20 Ω
R4
D1 C1 C2
1 kΩ

C0
C5

Figure B.6 Internal circuit of the encoder interface, model LPD3806-600BM-

G5-24C.

circuit must be connected to positive DC voltage through a pull-up resistor.

The eQEP pins of the F28069M LaunchPad™ board have pull-up resistors,
thus, the encoder cable can be directly connected to the extPot3 pins related
to QEP_A and QEP_B. Th e extPot3 board can also supply the encoder
with the available 5 V and GND terminals. Attention must be paid to limit
the length of the cable connecting the encoder to the board, since it is prefer-
able to have pull-up resistors in close proximity to the encoder interface to
improve noise immunity. Moreover, as shown by the circuit scheme reported
in Figure B.6, the encoder interface has a LM7805 5 V linear regulator, which
power dissipation has to be kept within reasonable limits. The regulator is in
a D-PAK package and it lies on a big ground plane that acts as a heat-sink.3

B2B-PMDC Hardware Kit

This hardware kit shown in Figure B.8 includes the following devices:
• One LaunchPad™ F28069M board;
• One or two TI BOOSTXL-DRV8301 BoosterPacks;
• One extPot3 board;
• A mezzanine board to hold the MCU and manage the external power supply;
• Two equal PMDC motors;
• Encoder LPD3806-600BM-G5-24C;
• Aluminum base plate, motor supports and joints.

3 The encoder case is connected to the chassis and not to circuit ground.
B2B-PMDC Hardware Kit 419

Figure B.7 Different views of the DecMot kit and its connections.

This kit is a complete motor control development test-bench aimed at rapid

prototyping PMDC-based sensored (via encoder) or senorless back-to-back
control schemes, targeting automotive, industrial and servo-drive applications.
One PMDC motor is used as brake to manipulate the load torque. Therefore,
back-to-back operations and motor characterization are possible. The mezza-
nine card is a MCU housing based on a custom made design. It can be even
separately used as support for the MCU board. The kit is designed to operate
with a 24 V 10 A power supply at rated conditions. Depending on the appli-
cation target, those values can be temporary exceeded. The previous encoder
connection considerations still hold.
 420 Appendix B: Custom Expansion Boards and Hardware Kits

Figure B.8 Different views of the B2B-PMDC kit and its connections.

This kit can be modified by substituting a PMDC motor with a BLDC one,
resultign in the:

B2B-BLDC Hardware Kit (see Figure B.9)

Therefore, three-phase sensored (via encoder) or senorless control schemes

can be tested as well. In this case, the PMDC motor is still used as brake.
The rated voltage and current are again 24 V and 10 A, respectively. Such kit
is suitable to emulate:
• Electric power-trains; electric fuel, water and oil pumps; engine cooling fans.
• Actuators and valve controls; doors, window lift and seat control.
B2B-PMDC Hardware Kit 421

Figure B.9 Different views of the B2B-BLDC kit and its connections.

Further information can be found at the following webpage:

https://www.mecc.polimi.it/nc/us/

Politecnico di Milano
Department of Mechanical Engineering, via La Masa 1, Milan 20156, Italy
Electrical Machines, Drives, and Power Electronics Research Group
422 Appendix B: Custom Expansion Boards and Hardware Kits

Newer versions of these hardware kits would be distributed by ePEBBs Srl,

please contact [email protected] and [email protected].
Bibliography

[1] IEEE Standard for Binary Floating-Point Arithmetic. ANSI/IEEE Std

754-1985, pages 1–20, 1985.
[2] Paolo Bolzern, Riccardo Scattolini, and Nicola Schiavoni. Fondamenti di
controlli automatici. McGraw-Hill Education, 4th edition, February 2015.
[3] Bimal K. Bose. Modern Power Electronics and AC Drives. Pearson
Education, 2001.
[4] Jason Bridgmon and Carolus Andrews. Current Sensing for Inline Motor-
Control Applications. Technical Report SBOA172, Texas Instruments,
October 2016. Available at https://www.ti.com/lit/an/sboa172/sb
oa172.pdf.
[5] Bill Chou. The Joy of Generating C Code from MATLAB. https:
//it.mathworks.com/company/newsletters/articles/the-joy-of
-generating-c-code-from-matlab.html.
[6] Luca Corradini, Dragan Maksimović, Paolo Mattavelli, and Regan Zane.
Digital Control of High-Frequency Switched-Mode Power Converters.
John Wiley & Sons, 22 June 2015.
[7] Texas Instruments. BoosterPack Checker. https://dev.ti.com/bpche
cker/#/.
[8] Texas Instruments. C2000 Real-Time Control MCUs: Applications. ht
tps://www.ti.com/microcontrollers/c2000-real-time-control-
mcus/applications.
[9] Texas Instruments. Inverter & motor control. https://www.ti.com/s
olution/hev-ev-inverter-motor-control?variantid=14291&subsy
stemid=17097.
[10] Texas Instruments. Website. https://www.ti.com.
[11] Texas Instruments. Meet the TMS320F28069M - LaunchPad Develop-
ment Kit. Technical Report SPRUI02, Texas Instruments, 2014. Available
at https://www.ti.com/lit/ml/sprui02/sprui02.pdf.
[12] Texas Instruments. Lift-Off with the TI LaunchPad™ Development
Ecosystem. Technical Report SLAT152B, Texas Instruments, 2015.

423
424 Bibliography

[13] Texas Instruments. Motor Drive and Control Solutions. Technical Report
SLYB165I, Texas Instruments, 2017.

[14] Texas Instruments. DRV8301 Three-Phase Gate Driver With Dual Cur-
rent Shunt Amplifiers and Buck Regulator. Technical Report SLOS719F,
Texas Instruments, August 2011 (revised January 2016). Available at
https://www.ti.com/lit/ds/symlink/drv8301.pdf.

[15] Texas Instruments. C2000™ Key Technology Guide. Technical Re-

port SPRACN0A, Texas Instruments, August 2019 (revised March 2020).
Available at https://www.ti.com/lit/an/spracn0a/spracn0a.pdf.
[16] Texas Instruments. User’s Guide - LAUNCHXL-F28069M Overview
(Rev. B). Technical Report SPRUI11B, Texas Instruments, January 2015
(revised March 2019). Available at https://www.ti.com/lit/ug/spru
i11b/sprui11b.pdf.
[17] Texas Instruments. TMS320x280x, 2801x, 2804x Serial Communica-
tions Interface (SCI). Technical Report SPRUFK7B, Texas Instruments,
March 2009 (revised April 2011). Available at https://www.ti.com/l
it/ug/sprufk7b/sprufk7b.pdf.
[18] Texas Instruments. C2000 MCU JTAG Connectivity Debug. Technical
Report SPRACF0, Texas Instruments, May 2018. Available at https:
//www.ti.com/lit/an/spracf0/spracf0.pdf.
[19] Texas Instruments. TMS320x280x, 2801x, 2804x DSP Analog-to-Digital
Converter (ADC). Technical Report SPRU716D, Texas Instruments,
November 2004 (revised April 2010). Available at https://www.ti.c
om/lit/ug/spru716d/spru716d.pdf.
[20] Texas Instruments. TMS320x280x, 2801x, 2804x Enhanced Quadrature
Encoder Pulse (eQEP) Module. Technical Report SPRU790D, Texas
Instruments, November 2004 (revised December 2008). Available at ht
tps://www.ti.com/lit/ug/spru790d/spru790d.pdf.
[21] Texas Instruments. TMS320x280x, 2801x, 2804x Enhanced Pulse Width
Modulator (ePWM) Module. Technical Report SPRU791F, Texas In-
struments, November 2004 (revised July 2009). Available at https:
//www.ti.com/lit/ug/spru791f/spru791f.pdf.
[22] Texas Instruments. User’s Guide - BOOSTXL-DRV8301 Hardware User’s
Guide. Technical Report SLVU974, Texas Instruments, October 2013.
Available at http://www.ti.com/lit/ug/slvu974/slvu974.pdf.
[23] Paul Krause, Oleg Wasynczuk, Scott Sudhoff, and Steven Pekarek. Anal-
ysis of Electric Machinery and Drive Systems. John Wiley & Sons, 3rd
edition, 2013.
Bibliography 425

[24] Werner Leonhard. Control of Electrical Drives. Springer, 3rd edition,

2001.

[25] Renato Manigrasso, Ferdinando L. Mapelli, and Marco Mauri. Aziona-

menti elettrici vol.1: Generalità e Macchine Rotanti. Pitagora, 2007.
[26] Renato Manigrasso, Ferdinando L. Mapelli, and Marco Mauri. Aziona-
menti elettrici vol.2: Convertitori e Controllo. Pitagora, 2007.

[27] Ferdinando L. Mapelli, Davide Tarsitano, and Marco Mauri. Plug-In

Hybrid Electric Vehicle: Modeling, Prototype Realization, and Inverter
Losses Reduction Analysis. IEEE Transactions on Industrial Electronics,
57(2):598–607, 2010.
[28] MathWorks. Getting Started—Embedded Coder Support Package for
Texas Instruments C2000 Processors. Video available at https://www.
mathworks.com/videos/getting-started-with-embedded-coder-s
upport-package-for-ti-c2000-processors-1573540550102.html.
[29] MathWorks. Pid tuner documentation. https://it.mathworks.com/h
elp/control/ref/pidtuner.html.
[30] MathWorks. Specialized power system documentation. https://it.mat
hworks.com/help/physmod/sps/specialized-power-systems.html.
[31] MathWorks. TI C2000 Support from Embedded Coder. https://it.m
athworks.com/hardware-support/ti-c2000.html.
[32] MathWorks. White Paper - Hardware-in-the-Loop Testing for Power
Electronics Control Design - Featuring Simulink and Speedgoat real-
time target machines. Technical report, MathWorks, 2019. Available
at https://it.mathworks.com/campaigns/offers/power-electron
ics-control-hil-white-paper.html.

[33] Mathworks. Simulink for Developing Digital Control for Motors, Power
Converters, and Battery Systems. Technical report, MathWorks, 2020.
Available at https://www.mathworks.com/content/dam/mathworks/
brochure/speed-digital-control-motor-power-converters-syst
ems-simulink-handout.pdf.

[34] MathWorks. White Paper - Model-Based Design for Embedded Control

Systems. Technical report, MathWorks, October 2020. Available at http
s://it.mathworks.com/campaigns/offers/model-based-design-e
mbedded-control-systems.confirmation.html?elqsid=1607505922
909&potential_use=Education.
[35] Marco Mauri, Maria Stefania Carmeli, and Mattia Rossi. Macchine Elet-
triche. Modelli a regime: teoria ed esercizi. Società Editrice Esculapio,
2nd edition, 2020.
426 Bibliography

[36] Robert D. Middlebrook and Slobodan Ćuk. A general unified approach

to modelling switching-converter power stages. In 1976 IEEE Power
Electronics Specialists Conference, pages 18–34, 1976.
[37] Ned Mohan, Tore M. Undeland, and William P. Robbins. Power Elec-
tronics: Converters, Applications, and Design. John Wiley & Sons, 3rd
edition, 2002.

[38] Mattia Rossi, Francesco Castelli Dezza, Marco Mauri, and Maria Pran-
dini. Explicit Computation of Indirect Hybrid MPC for Voltage Control
in Multilevel DC-DC Converters. In 2018 20th European Conference on
Power Electronics and Applications (EPE’18 ECCE Europe), pages P.1–
P.9, 2018.
[39] Mattia Rossi, Eyke Liegmann, Petros Karamanakos, Francesco Castelli
Dezza, and Ralph Kennel. Direct Model Predictive Power Control of a
Series-Connected Modular Rectifier. In 2019 IEEE International Sym-
posium on Predictive Control of Electrical Drives and Power Electronics
(PRECEDE), pages 1–6, 2019.
[40] Mattia Rossi, Luigi Piegari, Francesco Castelli Dezza, Marco Mauri, and
Maria S. Carmeli. Voltage Control Comparison for Low-Power DC-DC
Converters in EVs: PI and Explicit MPC. In IECON 2018 - 44th Annual
Conference of the IEEE Industrial Electronics Society, pages 5005–5011,
2018.
[41] Sigurd Skogestad and Ian Postlethwaite. Multivariate Feedback Control:
Analysis and Design. John Wiley & Sons, 2nd edition, November 2005.
[42] Stephen Umans. Fitzgerald & Kingsley’s Electric Machinery. McGraw-
Hill Education, 7th edition, 16 May 2013.
[43] Peter Vas. Sensorless Vector and Direct Torque Control. Oxford Univer-
sity Press, 1998.
[44] J. G. Ziegler and N.B. Nichols. Process lags in automatic control circuits.
Transactions of the A.S.M.E., 65(5):433–444, 1943.
Index

Action qualifier, 173–179 DAC (digital-to-analog converter)

actuation variable, 78, 271–273 peripheral, 1, 33, 167,
ADC (analog-to-digital converter) 195–201, 325, 362–365
peripheral, 1, 8, 33, 39, Data type, 23–24, 87–98, 113, 117,
101, 151–163, 172, 175, 122, 136–137, 157
189–195, 257–271 DC bus, 36, 39, 233, 279
Anti-windup scheme, 77–82 DC-DC converter, 227, 275,
Automatic code generation, 1, 4, 301–306
15–25, 42, 101, 109 DecMot kit, 331–333, 411, 423–425
Discrete-time domain, 62–66, 74, 86,
B2B (back-to-back) 231
-BLDC hardware kit, Duty cycle, 5, 36, 144, 168, 228,
-PMDC hardware kit, 331, 233–237, 241–251, 279–286,
333, 411 303–320, 340–350
setup, 367–376 Dynamical systems, 54–56
Baud rate, 103, 106n2, 113–114,
119 Embedded coder for Texas
Bipolar voltage switching, 246–251, Instruments C2000
254–255 processors, 43
Bode diagram, 72, 311 Embedded platform, 11, 86, 97
Boot options, 35–36 Enhanced quadrature encoder pulse
(eQEP) peripheral, 29, 101,
Carrier signal, 175–178 211–223
Cascade control architecture, Evaluation board, 25
336–339 External mode execution, 379,
Closed-loop control scheme, 53–66 386–389
Closed-loop transfer function, extPot3 board, 157–162, 215, 228,
55, 64 273, 286, 299, 331–334,
Continuous-time domain, 56–62, 74 411–414
Counting modes, 174, 176–178, 183 extRL(C) board (RL(C) hardware
Current kit), 273–274, 287, 294,
control, 67–86, 273–298, 299–302, 322, 328, 411,
336–339, 352, 367–370, 415–423
382–383, 401–403, 422
measurement, 36, 40, 206, 208, Firmware
257–271 deployment, 22, 107
Custom board, 37, 215, 411–428 design, 1–13, 22

427
428 Index

environment, 115–119, 125–131, Open-loop transfer function, 58–59,

135–148, 155–160, 163–164, 228–231, 275–277, 309–310,
182–195, 197–200, 203–205, 334, 338, 402–403
217–222, 238–240, 252–255, Optical rotary encoder, 211–215, 331
261–265, 267, 286–294,
321–328, 355–366, 371–376 Parameter identification, 296–297,
Fixed-point data, 87–90, 93–95, 354–355, 369–370
407–409 Peripherals, 1–4, 101, 113–223
Floating-point data, 31, 87–93, PI (proportional-integral) controller,
95–97, 121–122, 407–409 21, 51–52, 56, 62, 74, 79–84,
Full-bridge configuration, 240–255, 289
275, 340–353, 422–423 current control 67–86, 273–298,
336–339, 370, 403, 422
GPIO (general purpose speed control, 336–353, 356–366,
input/output), 133–149 367–376, 402
peripheral, 133–135 tuning, 56–62, 76–79, 84–86
DI (digital input), 135–136 pole/zero cancellation 58–59,
DO (digital output), 136–137 69
generalized approach 59–60
Half-bridge configuration, 233–240, voltage control 299–330
275, 279–294, 312–328 PIL (processor-in-the-loop), 25,
Hardware kit, 29–40, 411–428 379–386, 391
Pin muxing, 32–33
Installation procedure, 43–47 Pin-out table, 32, 39, 133–134
PMDC (permanent magnet DC)
Jumper configuration, 35
motor, 227–255, 240,
LC filter, 310, 328–329 260–266, 331–376
Linear model, 228–232, 306–312, PMSM (permanent magnet
334–336, 398–401 synchronous motor),
Locked rotor test, 261–266, 396–405
269–271 PWM (pulsed-width modulation),
Low-Side Shunt Current Sensing, 21, 37, 167–172, 175–181,
39–40, 257–271 228, 236, 241–243, 248,
277–278, 281, 284–286,
MATLAB, 17–27, 41–47, 60, 90–93 291–293, 318, 346
MCU (microcontroller), 1–13, 22, Logic, 143–149, 158–162,
29–36, 41, 111–112, 167–172, 175–181, 193–195,
208–210, 285, 347, 366, 242, 278, 284
379–381 peripheral, 1, 26, 167–210, 325n1
Model-based design, 17–20, 70 Power Electronics, 1–3, 9–13, 17,
Motor control, 227–256, 331–376, 20–21, 208–210
391–405
Model reference adaptive system Rapid prototyping, 17–23
(MRAS) observer, 331, 343, Resistive-inductive-capacitive (RLC)
350–353, 365, 367, 370 load, 299–312, 328–329
Index 429

Sample rate, 122–124, 208–210 Synchronization, 101, 105n1, 172,

SCI (serial communication interface), 208, 292, 326, 352, 364
35, 108, 113–131 between channels, 162–165, 173,
receive, 101, 115–118 183–185, 201–205
transmit, 101, 118–119 between peripherals, 206–210
Sensor characterization, 258–260,
266–271 Testing environment, 109, 119–122,
Serial send, 120–121 126–128, 129–131, 142, 148,
Serial receive, 121–122 161, 263, 267
Shift operation, 95, 407 TI
Shunt current sensing, 39–40, BOOSTXL-DRV8301
257–258 BoosterPack, 36–40, 411
Simscape (specialized power C2000 MCU, 7–9, 24–26, 29, 31,
systems) 17, 20, 27n1, 43–47
69–73, 75, 232, 235, 242, Piccolo F28069M LaunchPad, 8,
277, 280, 312–313, 29–36
339, 342 Transfer function, 54–57, 64, 68, 310,
Simulink 17, 20–26, 43, 69–76, 101, 338, 402–403
107–112
Software installation, 41–47 Unipolar voltage switching, 241–247,
Speed computation, 215–217, 252–254, 422–423
219–222
State-space average modelling Virtual COM port, 103–105, 113, 115
method, 306–312 Voltage
Step-down converter, 233–237, control, 299–330
275, 279–286, 301–306, measurement, 323, 326–329
312–321 sensor, 323, 415