Control System Toolbox™ Release Notes

How to Contact MathWorks

Latest news: www.mathworks.com

Sales and services: www.mathworks.com/sales_and_services

User community: www.mathworks.com/matlabcentral

Technical support: www.mathworks.com/support/contact_us

Phone: 508-647-7000

The MathWorks, Inc.

1 Apple Hill Drive
Natick, MA 01760-2098
Control System Toolbox™ Release Notes
© COPYRIGHT 2002–2022 by The MathWorks, Inc.
The software described in this document is furnished under a license agreement. The software may be used or copied
only under the terms of the license agreement. No part of this manual may be photocopied or reproduced in any form
without prior written consent from The MathWorks, Inc.
FEDERAL ACQUISITION: This provision applies to all acquisitions of the Program and Documentation by, for, or through
the federal government of the United States. By accepting delivery of the Program or Documentation, the government
hereby agrees that this software or documentation qualifies as commercial computer software or commercial computer
software documentation as such terms are used or defined in FAR 12.212, DFARS Part 227.72, and DFARS 252.227-7014.
Accordingly, the terms and conditions of this Agreement and only those rights specified in this Agreement, shall pertain
to and govern the use, modification, reproduction, release, performance, display, and disclosure of the Program and
Documentation by the federal government (or other entity acquiring for or through the federal government) and shall
supersede any conflicting contractual terms or conditions. If this License fails to meet the government's needs or is
inconsistent in any respect with federal procurement law, the government agrees to return the Program and
Documentation, unused, to The MathWorks, Inc.
Trademarks
MATLAB and Simulink are registered trademarks of The MathWorks, Inc. See
www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be
trademarks or registered trademarks of their respective holders.
Patents
MathWorks products are protected by one or more U.S. patents. Please see www.mathworks.com/patents for
more information.
 Contents

R2022b

Create Plot Live Editor Task: Create linear analysis response plots
interactively and generate code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-2

interface: New option to obtain primal-assembly model . . . . . . . . . . . . . . 1-2

New Examples: Power electronics control design . . . . . . . . . . . . . . . . . . . . 1-2

R2022a

Extended and Unscented Kalman Filters: Support for measurements with

circular wrapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3

Renamed plot options objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-3
Default value of bdschur input argument CONDMAX changed . . . . . . . . . 2-3
Default value of canon input argument condt changed . . . . . . . . . . . . . . . 2-3

R2021b

Frequency-Domain Analysis: Support for models with complex

coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2

frd: Support for negative frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2

stepinfo and lsiminfo: Support for nonzero initial value . . . . . . . . . . . . . . 3-2

mechss and sparss: Linearize structural and thermal models to sparse

models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3

Support for opening SISO Design Tool sessions saved before release R2016a
has been removed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3
ss2ss now returns different transformation results for descriptor state-space
models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3

iii
 ss2ss: Similarity transformation is no longer supported for mechss models
...................................................... 3-4
stepinfo and lsiminfo: Response characteristics computation changes . . . . 3-4

R2021a

Model Reduction: Compute reduced-order models with small relative-

error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2

Model Reduction: Compute and plot Hankel singular values (HSV) using
balred . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2

inv and imp2exp: Implicit form returned for state-space models by

default . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3

Control System Designer: Option to disable automatic plot update during

Response Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3

State-Space Models: New Property to Support State Block Path

Management in Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3

c2dOptions: New Option to Specify Fit Order . . . . . . . . . . . . . . . . . . . . . . . 4-3

margin: Support for Sparse State-Space Models . . . . . . . . . . . . . . . . . . . . 4-4

modred, stabsep and modsep: Support for Descriptor Models . . . . . . . . . 4-4

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . . 4-4

hsvd and hsvdOptions are not recommended . . . . . . . . . . . . . . . . . . . . . . 4-4
balredOptions option StateElimMethod renamed to StateProjection . . . . . 4-4
Options to set error tolerance has changed in the balredOptions,
stabsepOptions and freqsepOptions commands . . . . . . . . . . . . . . . . . . . 4-5
Kalman Filter block: Numerical changes . . . . . . . . . . . . . . . . . . . . . . . . . . 4-5
modred, stabsep and modsep: Numerical changes . . . . . . . . . . . . . . . . . . 4-5
Support for opening SISO Design Tool sessions saved before release R2016a
will be removed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6

R2020b

sparss and mechss: Support for sparse state-space models . . . . . . . . . . . 5-2

New Example: Tune Phase-Locked Loop Using Loop-Shaping Design . . . 5-2

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2

Extended and Unscented Kalman Filter Algorithms: Numerical changes
...................................................... 5-2

iv Contents
 R2020a

Control System Designer: Perform automated loop shaping without a

Robust Control Toolbox license . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2

frdfun: Utility to apply a function across all frequencies of an frd model

.......................................................... 6-3

zgrid: Specify Sample Time to Plot True Frequencies on the Grid . . . . . . 6-3

R2019b

Live Editor Tasks: Perform model transformation and control design

tasks interactively and generate MATLAB code in a live script . . . . . . . 7-2

Residuals for Extended and Unscented Kalman Filters: Calculate

residuals and residual covariances of filter predictions . . . . . . . . . . . . . 7-2

ngrid, sgrid, and zgrid: Specify target axes as Axes and UIAxes to create
apps in App Designer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-2

Nyquist Plots: Programmatically Zoom on Critical Point . . . . . . . . . . . . . . 7-2

R2019a

getPIDLoopResponse: Obtain closed-loop and open-loop responses of

plant with PID controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2

icare and idare Commands: Solve continuous-time and discrete-time

Riccati equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2

care and gcare are not recommended . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-2
dare and gdare are not recommended . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-3

R2018b

allmargin Function: New MIMO syntax for loop-at-a-time analysis . . . . . 9-2

v
 R2018a

Particle Filter Simulink Block: Estimate states of nonlinear systems for

online tracking and control system design . . . . . . . . . . . . . . . . . . . . . . 10-2

c2d Function: Convert models to discrete time using least-squares

optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10-2

Control System Designer: Change sample time of control system . . . . . 10-2

Control System Designer: Create Simulink model for control architecture

......................................................... 10-4

R2017b

Gain Scheduling: Implement gain-scheduled controllers using a new

library of blocks configured to take scheduled parameters as inputs
......................................................... 11-2

Gain Scheduling: Achieve smooth and memory-efficient implementation

by turning gain surfaces into embedded equations . . . . . . . . . . . . . . . 11-2

Gain-Scheduled Controller Tuning: Automatically tune gain-scheduled

state observer gain, LQR gain, and other controller architectures
expressed as matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3

Gain-Scheduled Controller Tuning: Specify tuning goals that vary with

operating condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-3

Tuning Gain Surfaces: Custom normalization, lookup-table updates, and

other enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-4

Gain-Scheduled Controller Tuning: Exclude design points from tuning or

analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-4

Particle Filters: Estimate states of nonlinear systems . . . . . . . . . . . . . . . 11-4

Improved lqg Function: Compute gain matrices and optimal controller in

discrete time using current Kalman Filter estimator . . . . . . . . . . . . . 11-5

Model Reduction: balred no longer ignores MatchDC option when

specified frequency or time intervals exclude DC . . . . . . . . . . . . . . . . . 11-5

Regularization of conic-sector tuning goal in Control System Tuner . . . 11-6

Dynamic system models store Notes property as string or character

vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11-6

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . 11-7

vi Contents
 R2017a

Extended and Unscented Kalman Filter Simulink Blocks: Estimate states

of nonlinear systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12-2

New properties of generalized state-space models and matrices . . . . . . 12-2

Discrete-time frequency-dependent specifications for tuning goals . . . . 12-2

Regularization of tuning goals for improved numeric stability . . . . . . . . 12-2

Maximum Natural Frequency Option in Control System Tuner: Prevent

poles and zeros from going to infinity . . . . . . . . . . . . . . . . . . . . . . . . . . 12-3

Scaling information passed automatically to viewSpec and evalSpec . . 12-3

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . 12-4

R2016b

Conic Sector Tuning Goal: Tune control systems to enforce fixed or

frequency-dependent sector bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2

Improved Passivity Tuning Goal: Set output passivity index to a negative

value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2

MaxRadius Option for Tuning: Prevent poles and zeros from going to
infinity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2

Improved getSectorIndex and sectorplot Functions: Compute and plot

sector index for unstable systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-2

Extended and Unscented Kalman Filters: Estimate states of nonlinear

systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-3

Phase-Wrap Branch Option: Specify cutoff point for wrapping phase in

response plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13-3

R2016a

Redesigned Control System Designer App: Design SISO controllers for

feedback systems using improved interactive workflows . . . . . . . . . . . 14-2

vii
 Control System Tuner App and systune Command: Automatically tune
single-loop and multiloop control systems to meet design requirements
......................................................... 14-3

Model Reducer App: Compute and compare reduced-order models using

interactive workflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-3

Passivity and Conic Sectors: Analyze and tune control systems for
passivity and other sector bounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-3

Limited Balanced Truncation: Reduce model order according to energies

within time-domain and frequency-domain intervals . . . . . . . . . . . . . . 14-4

sampleBlock and rsampleBlock commands for sampling generalized

models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-5

Spectral factorization of LTI models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14-5

Renamed tunable control design blocks . . . . . . . . . . . . . . . . . . . . . . . . . . 14-5

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . 14-6

R2015b

pid2 and pidstd2 Model Objects: Represent, analyze, and use 2-DOF PID
controllers for control design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15-2

2-DOF PID Controller Tuning: Automatically tune the gains of 2-DOF PID
controllers with PID Tuner app and pidtune command . . . . . . . . . . . . 15-2

Save Current Controller Design as Baseline in PID Tuner . . . . . . . . . . . . 15-3

Change in LPV System block default values for model delays . . . . . . . . . 15-3

Analysis Plots Wrap Phase in Interval [0º,360º) . . . . . . . . . . . . . . . . . . . . 15-3

R2015a

Improved input disturbance rejection with the PID tuning algorithm

......................................................... 16-2

Option to specify code generation settings in LPV System block . . . . . . 16-3

connect command syntax for specifying analysis point locations . . . . . . 16-3

LTI Viewer renamed to Linear System Analyzer . . . . . . . . . . . . . . . . . . . . 16-3

viii Contents
sisotool function renamed to controlSystemDesigner . . . . . . . . . . . . . . . 16-4

getBlockValue returns all block values in structure . . . . . . . . . . . . . . . . 16-4

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . 16-5

R2014b

LPV System block for modeling and simulating linear parameter-varying

systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-2

Kalman Filter block for estimating states of linear time-invariant and

linear time-varying systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3

AnalysisPoint Control Design Block for Marking Points of Interest for

Linear Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3

pidtool function renamed to pidTuner . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17-3

getSwitches function renamed to getPoints . . . . . . . . . . . . . . . . . . . . . . . 17-4

Functionality being removed or changed . . . . . . . . . . . . . . . . . . . . . . . . . 17-4

R2014a

Redesigned PID Tuner app for improved PID tuning workflow . . . . . . . . 18-2

PID controller tuning using system identification to model the plant from
measured input-output data in the PID Tuner app (with System
Identification Toolbox) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-2

freqsep function for decomposing a linear system into fast dynamics and
slow dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18-2

damp command display includes time constant information . . . . . . . . . 18-2

R2013b

SamplingGrid property for tracking dependence of array of sampled

models on variable values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-2

ix
 Option to retain unconnected states when interconnecting models using
connect command . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-2

connect command always returns state-space or frequency response data

model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-2

updateSystem command for updating dynamic system data in a response

plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-3

getLoopID renamed to getSwitches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19-3

LoopID property of loopswitch renamed to Location . . . . . . . . . . . . . . . . 19-3

R2013a

Transient behavior slider for PID Tuner, increasing control over reference
tracking and disturbance rejection performance . . . . . . . . . . . . . . . . . 20-2

R2012b

ltiblock.pid2 and loopswitch objects for tuning two-degree-of-freedom

PID controllers and marking loop opening sites for open-loop
requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21-2

Commands for obtaining open-loop responses, closed-loop responses, and

current values of tunable components from control system models
......................................................... 21-2

Option for elementwise operation of model query commands on model

arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21-2

R2012a

Frequency Analysis Commands for Calculating Peak Gain and Finding

Gain-Crossover Frequencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22-2

Specify Target Crossover Frequency as Input to pidtune . . . . . . . . . . . . . 22-2

Rescaled Impulse Response and Impulse-Invariant Time Domain

Conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22-2

First-Order Hold Method for d2c . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22-2

x Contents
tzero Computes Invariant Zeros and Transmission Zeros . . . . . . . . . . . . 22-3

Models Created With System Identification Toolbox Can Be Used Directly

With Control System Toolbox Functions . . . . . . . . . . . . . . . . . . . . . . . . 22-3

Functionality Being Removed or Changed . . . . . . . . . . . . . . . . . . . . . . . . 22-3

R2011b

Formula-Based Specification of Summing Junctions and Vector Signal

Naming for sumblk and connect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23-2

Commands for Interacting with Control Design Blocks in Generalized LTI

Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23-2

Functionality Being Removed or Changed . . . . . . . . . . . . . . . . . . . . . . . . 23-2

R2011a

New Model Objects for Representing Tunable Parameters and Systems

with Tunable Components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24-2

New Time and Frequency Units for Models and Response Plots . . . . . . . 24-2

Discrete-Time PID Controller Objects Have Stable Derivative Filter Pole

......................................................... 24-3

New Variable q^–1 for Expressing Discrete-Time Transfer Functions . . 24-4

R2010b

New Commands and GUI for Modeling and Tuning PID Controllers . . . 25-2
PID Controller Design with the New PID Tuner GUI . . . . . . . . . . . . . . . . 25-2
PID Controller Design with the New pidtune Command . . . . . . . . . . . . . 25-2
Modeling PID Controllers in Parallel Form or Standard Form . . . . . . . . . 25-2

Improved PID Tuning Options in SISO Design Tool . . . . . . . . . . . . . . . . . 25-3

Ability to Analyze a Controller Design for Multiple Models Simultaneously

in SISO Design Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25-3

Change in Output of repsys Command . . . . . . . . . . . . . . . . . . . . . . . . . . . 25-3

xi
 R2010a

Enhanced c2d Command to Approximate Fractional Time Delays in Tustin

and Matched Discretization Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 26-2

New Commands for Specifying Options for Continuous-Discrete

Conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26-2

New FDEL Command to Remove Specified Data from Frequency Response

Data (FRD) Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26-2

R2009b

Ability to Design Compensators for New Types of Plants . . . . . . . . . . . . . 27-2

New Automated PID Tuning Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27-2

R2009a

Variable q Now Defined as the Forward Shift Operator z . . . . . . . . . . . . . 28-2

R2008b

New Design Tools for Linear-Quadratic-Gaussian (LQG) Servo Controllers

with Integral Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29-2
Current Flag Moved from lqgreg to kalman . . . . . . . . . . . . . . . . . . . . . . 29-2

New Upsampling Method for Rate Conversion in Discrete-Time Models

......................................................... 29-2

New Scaling Tools to Enhance the Accuracy of Computations with State-

Space Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29-2

New Command to Reorder the States of State-Space Models . . . . . . . . . 29-3

Enhanced Support for Customizing Response Plots . . . . . . . . . . . . . . . . . 29-3

xii Contents
 R2008a

Updated Error and Warning Message System . . . . . . . . . . . . . . . . . . . . . . 30-2

R2007b

Updated and Expanded Demos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31-2

R2007a

Analysis of Time Delay Systems Now Fully Supported . . . . . . . . . . . . . . . 32-2

New and Updated Automated Tuning Methods . . . . . . . . . . . . . . . . . . . . . 32-2

New Tustin and Prewarp Options for d2d Function . . . . . . . . . . . . . . . . . 32-2

R2006b

New Loop Configurations in the SISO Design Tool . . . . . . . . . . . . . . . . . . 33-2

New Design Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33-2

R2006a

SISO Design Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-2

Compensator Optimization Is Now Supported . . . . . . . . . . . . . . . . . . . . . 34-2
Improved Compensator Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-2
Multi-Loop Compensator Design Support . . . . . . . . . . . . . . . . . . . . . . . . 34-2
SISO Design Tool Fully Integrated with the Controls & Estimation Tools
Manager . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-2

LTI Viewer Enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-2

LTI Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-2

Descriptor and Improper State-Space Models Fully Supported . . . . . . . . 34-2
New Commands to Calculate Time Response Metrics . . . . . . . . . . . . . . . 34-3
Simplified System Interconnections Using I/O Channel Names . . . . . . . . 34-3

xiii
 Changes in the Representation of I/O Delays in State-Space Models . . . . 34-3
New Name Property for LTI Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-3
New Commands and Operations for LTI Objects . . . . . . . . . . . . . . . . . . . 34-3

Numerical Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34-3

R14SP3

No New Features or Changes

R14SP2

Command-Line API for Customizing Plots . . . . . . . . . . . . . . . . . . . . . . . . 36-2

Constraint Types for SISO Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36-2

Bode and Nichols Plots Have Additional Options . . . . . . . . . . . . . . . . . . . 36-2

Model-Approximation and Order-Reduction Commands . . . . . . . . . . . . . 36-2

xiv Contents
 1

R2022b

Version: 10.12

New Features

Bug Fixes
R2022b

Create Plot Live Editor Task: Create linear analysis response plots
interactively and generate code
You can now use the Create Plot Live Editor task to interactively create linear analysis response
plots for your dynamic system model. The task also automatically generates code that becomes part
of your live script.

For more information and an example, see the Create Plot Live Editor Task reference page.

interface: New option to obtain primal-assembly model

interface now includes a new optional argument to specify the method of physical coupling as
either the dual-assembly method or primal-assembly method. The primal-assembly method uses a
minimal number of degrees of freedom, but the system may suffer from fill-in. Use the syntax
interface( ___ ,method) to specify an assembly method of physical coupling, where method is
either 'dual' or 'primal'. If you do not specify a method, the function uses the dual-assembly
method by default. Previously, the function always returned a dual-assembly model.

For more information, see “Algorithms”.

New Examples: Power electronics control design

This release includes the following new reference examples.

• “Feedback Amplifier Design for Voltage-Mode Boost Converter” — This example illustrates tuning
the components of a power supply controller to control the output voltage of a boost converter
using loop-shape design and fixed-structure tuning methods. The workflow is demonstrated using
a type-III controller. You need a Mixed-Signal Blockset™ license to run this example.
• “Detect Replay Attacks in DC Microgrids Using Distributed Watermarking” — This example shows
how to use distributed watermarking to detect replay attacks in a cyber-physical system.

1-2
 2

R2022a

Version: 10.11.1

New Features

Bug Fixes

Compatibility Considerations
R2022a

Extended and Unscented Kalman Filters: Support for measurements

with circular wrapping
You can now use wrapped measurements with the extendedKalmanFilter and
unscentedKalmanFilter objects by enabling the HasMeasurementWrapping property, and
specifying a measurement function (through the MeasurementFcn property) with two outputs:

1 The measurement, specified as an N-element output measurement vector of the nonlinear system
at time step k, given the state vector at time step k. N is the number of measurements of the
system.
2 The measurement wrapping bounds, specified as an N-by-2 matrix where, the first column
provides the minimum measurement bound and the second column provides the maximum
measurement bound.

Enabling the HasMeasurementWrapping property wraps the measurement residuals in a defined

bound, which helps to prevent the filter from divergence due to incorrect measurement residual
values. This property is nontunable and can be set only during the object creation.

You can also enable measurement wrapping in the Extended Kalman Filter and Unscented Kalman
Filter blocks using the Has measurement wrapping check box.

2-2
For an example, see State Estimation with Wrapped Measurements Using Extended Kalman Filter.

Functionality being removed or changed

Renamed plot options objects
Behavior change

The following plot options objects have been renamed for R2022a:

Creation Function Old Name New Name

timeoptions TimePlotOptions object TimeOptions object
bodeoptions BodePlotOptions object BodeOptions object
pzoptions PZMapOptions object PZOptions object
nyquistoptions NyquistPlotOptions object NyquistOptions object
hsvoptions HSVPlotOptions object HSVOptions object
sigmaoptions SigmaPlotOptions object SigmaOptions object
nicholsoptions NicholsPlotOptions object NicholsOptions object

Default value of bdschur input argument CONDMAX changed

Behavior change

The default value of bdschur input argument CONDMAX is now 1e4. Previously, the default value was
1/sqrt(eps).

CONDMAX specifies an upper bound on the condition number of the transformation matrix T.
Increasing CONDMAX reduces the size of the eigenvalue clusters in the transformed model but also
decreases the accuracy of the transformation. The new default value balances the conditioning and
transformation accuracy. If you have code that relies on the default value of CONDMAX being 1/
sqrt(eps), update your code to explicitly set this input argument.

Default value of canon input argument condt changed

Behavior change

The default value of canon input argument condt is now 1e4. Previously, the default value was 1e8.

condt specifies an upper bound on the condition number of the transformation matrix T. Increasing
condt reduces the size of the eigenvalue clusters in the transformed model but also decreases the
accuracy of the transformation. The new default value balances the conditioning and transformation
accuracy. If you have code that relies on the default value of condt being 1e8, update your code to
explicitly set this input argument.

2-3
 3

R2021b

Version: 10.11

New Features

Bug Fixes

Compatibility Considerations
R2021b

Frequency-Domain Analysis: Support for models with complex

coefficients
You can now visualize and analyze frequency-domain responses of complex-coefficient systems using
bode, margin, nyquist, nichols, sigma, passiveplot, and sectorplot functions. You can also
now specify negative frequencies in the frequency vector w when using the syntax bode(sys,w) or
similar.

For bode, margin, sigma, passiveplot, and sectorplot, when you plot complex-coefficient
systems in:

• Log frequency scale, the plots show two branches, one for positive frequencies and one for
negative frequencies. The arrows indicate the direction of increasing frequency values for each
branch.
• Linear frequency scale, the plots show a single branch with a symmetric frequency range centered
at a frequency value of zero. If you specify a frequency range of [wmin,wmax] for your plot, the
frequency limits are set to [–wmax,wmax].

For an example, see Bode Plot of Model with Complex Coefficients .

The nyquist plot now shows a contour comprised of both positive and negative frequencies. The
arrows indicate the direction of increasing frequency for each branch. For complex-coefficient
systems, the two branches are not symmetric. For real-coefficient systems, the negative branch is
obtained by symmetry. For an example, see Nyquist Plot of Model with Complex Coefficients

For complex-coefficient models, a nichols plot now shows a contour comprised of both positive and
negative frequencies. For real-coefficient models, the plot shows only positive frequencies, even when
complex-coefficient models are present. For an example, see Nichols Plot of Model with Complex
Coefficients.

frd: Support for negative frequencies

You can now specify negative frequency points in a frequency-response data (frd) model object. frd
models with negative frequencies are used to represent models with complex coefficients. Previously,
the negative frequencies were not supported.

stepinfo and lsiminfo: Support for nonzero initial value

The stepinfo and lsiminfo commands now let you specify an initial value for your response data.
Previously, the commands always assumed a default initial value to be zero.

To compute step-response characteristics when you have a nonzero initial value, use the following
syntax.

S = stepinfo(y,t,yfinal,yinit)

This command computes step-response characteristics from an array of step-response data y,

corresponding time vector t, steady-state value yfinal, and initial value yinit. For more
information, see stepinfo.

To compute linear response characteristics when you have a nonzero initial value, use the following
syntax.

3-2
S = lsiminfo(y,t,yfinal,yinit)

This command computes response characteristics from an array of linear response data y,
corresponding time vector t, steady-state value yfinal, and initial value yinit. For more
information, see lsiminfo.

Additionally, the output structures of stepinfo and lsiminfo now contain a TransientTime field.
Use this characteristic to measure how quickly the transients die off. This characteristic also applies
to any response, including responses with zero final values (impulse and initial).

Compatibility Considerations
As a result of these changes to stepinfo and lsiminfo, the computation of some response
characteristics has changed. For more information, see “stepinfo and lsiminfo: Response
characteristics computation changes” on page 3-4.

mechss and sparss: Linearize structural and thermal models to sparse

models
You can now obtain mechss and sparss model objects from structural and thermal models using the
linearize (Partial Differential Equation Toolbox) function. This function also lets you extract finite
element matrices used to obtain the mechss and sparss models. For examples, see Linear Analysis
of Cantilever Beam and Linear Analysis of Tuning Fork.

Functionality being removed or changed

Support for opening SISO Design Tool sessions saved before release R2016a has been
removed
Errors

Support for opening SISO Design Tool sessions saved before release R2016a will be removed in
release R2021b.

If you have sessions saved before release R2016a, open and resave the session files using Control
System Designer in any release from R2016a through R2021a.

ss2ss now returns different transformation results for descriptor state-space models
Behavior change

ss2ss performs the similarity transformation x = Tx on the state vector x and produces the
equivalent state-space model sysT. For descriptor state-space models ss2ss now returns a different
transformation result.

For a descriptor state-space model

Eẋ = Ax + Bu
y = Cx + Du,

ss2ss now returns

ET −1ẋ = AT −1x + Bu
y = CT −1x + Du .

3-3
R2021b

Previously, the function returned the following transformation.

TET −1ẋ = T AT −1x + TBu

y = CT −1x + Du

For more information, see ss2ss.

ss2ss: Similarity transformation is no longer supported for mechss models

Errors

ss2ss no longer supports sparse second-order (mechss) models. Performing similarity

transformations on mechss models destroys symmetry and has no obvious general form.

stepinfo and lsiminfo: Response characteristics computation changes

Behavior change

Because of changes to stepinfo and lsiminfo, the computation of some response characteristics
has changed. Additionally, the settling time calculation is now based on how quickly the response gets
within a specified threshold of the final value.

The following table summarizes the changes to the fields of the structure returned by stepinfo.

Before R2021b R2021b

RiseTime — Time it takes for the response to RiseTime — Time it takes to go from 10% to
rise from 10% to 90% of the way from y(1) to 90% of the way from yinit to yfinal.
yfinal.
SettlingTime — The first time T such that the SettlingTime — The first time T such that the
error |y(t) – yfinal| ≤ SettlingTimeThreshold × emax error |y(t) – yfinal| ≤ SettlingTimeThreshold × |yfinal
for t ≥ T, where emax is the maximum error |y(t) – – yinit| for t ≥ T.
yfinal| for t ≥ 0.
By default, SettlingTime measures the time it
By default, SettlingTimeThreshold = 0.02 (2% of takes for the error to stay below 2% of |yfinal –
the peak error). SettlingTime measures the yinit|.
time for the error to fall below 2% of the peak
value of the error.
Peak — Peak absolute value of y(t). Peak — Peak absolute value of y(t) – yinit.

The following table summarizes the changes to the fields of the structure returned by lsiminfo.

Before R2021b R2021b

SettlingTime — The first time T such that the SettlingTime — The first time T such that the
error |y(t) – yfinal| ≤ SettlingTimeThreshold × emax error |y(t) – yfinal| ≤ SettlingTimeThreshold × |yfinal
for t ≥ T, where emax is the maximum error |y(t) – – yinit| for t ≥ T.
yfinal| for t ≥ 0.
By default, SettlingTime measures the time it
By default, SettlingTimeThreshold = 0.02 (2% of takes for the error to stay below 2% of |yfinal –
the peak error). SettlingTime measures the yinit|.
time for the error to fall below 2% of the peak
value of the error.

3-4
Additionally, the output structures of stepinfo and lsiminfo now contain a TransientTime field.
This characteristic contains the peak-error-based settling time calculation used in releases before
R2021b. Transient time is used to measure how quickly the transient dynamics die off.

Here:

• y(t) is the system response.

• yinit is the initial value of y(t) before the response occurs. By default, yinit = 0.
• yfinal is the final value of y(t). By default, yfinal = last sample value of y(t).
• SettlingTimeThreshold is the threshold for defining settling time. By default,
SettlingTimeThreshold = 0.02.

These changes also apply to the characteristics of step, impulse, and initial plots. Additionally:

• For step plots, yinit is always assumed to be zero and yfinal is the steady-state value.
• For the step response, transient time and settling time tend to differ for models with feedthrough,
zeros at the origin, unstable zeros (undershoot), or large overshoot. They match for models with
no undershoot or feedthrough, and with less than 100% overshoot.
• For the step response of models with feedthrough, the new RiseTime value can differ because
y(1) is nonzero whereas yinit is zero by default. Before R2021b, the rise time computed was the
time it takes to go from 10% to 90% of the way from y(1) to yfinal, instead of yinit to yfinal now.

3-5
 4

R2021a

Version: 10.10

New Features

Bug Fixes

Compatibility Considerations
R2021a

Model Reduction: Compute reduced-order models with small relative-

error
You can now use the Balanced Stochastic Truncation method in balred. This method lets you control
the relative error of the approximation and preserves roll-off characteristics of the original model.
Use the new ErrorBound option in balredOptions to choose between absolute or relative error
control. The Regularization option ensures a well-defined relative error at all frequencies.

• ErrorBound — You can choose between absolute and relative error approximation. Previously,
you could only use absolute error approximation when performing model reduction. Relative error
approximation is useful when your original model has a very high or low gain in the region that is
important to your application. In such regions, absolute error approximations are not effective.
• Regularization — Use this option to choose the regularization level that ensures a well-defined
relative error at all frequencies.

For more information, see balredOptions.

Relative-error approximation is also available in the Model Reducer app. To use it, in the Balanced
Truncation tab, in the Error Bound drop-down list, select relative. You can also specify the
regularization value in the Regularization field inside the Options menu in the Balanced
Truncation tab.

For more information, see Balanced Truncation Model Reduction.

Compatibility Considerations
balred and balredOptions replaces the hsvd and hsvdOptions commands to compute and plot
Hankel singular values and includes options that preserves roll-off characteristics from the original
model. The hsvd and hsvdOptions commands are not recommended. For more information, see the
Functionality being removed or changed section.

Model Reduction: Compute and plot Hankel singular values (HSV)

using balred
You can now use balred to compute and plot the Hankel singular values (HSVs) and reduce the
model with this information based on your desired fidelity. Previously, you had to use hsvd to
compute the HSVs and then extract the reduced model-order approximation using balred.
Additionally, balred can also return the error bounds, regularization level and the Cholesky factors
of the gramians.

For more information, see balred.

Compatibility Considerations
balred replaces hsvd to compute and plot Hankel singular values. The hsvd command is not
recommended. For more information, see the Functionality being removed or changed section.

4-2
inv and imp2exp: Implicit form returned for state-space models by
default
The inv and imp2exp functions now return the implicit form for ss, genss and uss (Robust Control
Toolbox) model objects, by default. When the inverse is proper, you can use inv(sys,'min') and B
= imp2exp(A,yidx,uidx,'min') to eliminate the extra states and obtain a model with as many
states as sys or A respectively.

Use isproper or ss(sys,'explicit') to extract an explicit model if desired.

For more information, see inv and imp2exp.

Compatibility Considerations
If your code uses the inv and imp2exp commands with state-space models, then it may produce
results that are different from the results you obtained using previous versions.

Control System Designer: Option to disable automatic plot update

during Response Optimization
You can now disable automatic plot updates during response optimization in the Control System
Designer. Use the Update plots during optimization checkbox in the Response Optimization
dialog under Tuning Methods to disable automatic plot updates during optimization.

You can also use the Update plots in real-time checkbox in the Control System Designer
Preferences dialog to enable or disable synchronous updates of the response plots while using
graphical tuning methods.

Disabling this option improves performance especially for multimodel control design.

State-Space Models: New Property to Support State Block Path

Management in Linearization
The state-space model object, ss, now has a new StatePath property to facilitate state block path
management in linearization.

For more information, see the ss reference page.

c2dOptions: New Option to Specify Fit Order

You can now use the new FitOrder option to specify the fit order when using the least-squares
method for continuous-discrete conversion. FitOrder specifies the order of the discrete-time model
to be fitted to the continuous-time frequency response in the options set obtained using c2dOptions.
Reducing the order helps with unstable poles or pole/zero cancellations at z = -1. Previously, some
discretized models had pole-zero cancellations at z = 1 and z = -1 or unstable poles at |z| > 1.

For more information, see the c2dOptions reference page.

4-3
R2021a

margin: Support for Sparse State-Space Models

You can now specify a frequency vector or frequency range to plot the Bode response of sparss and
mechss model objects using margin. Previously, margin did not support sparse models.

For more information, see the margin reference page.

modred, stabsep and modsep: Support for Descriptor Models

modred, stabsep and modsep now supports descriptor models. Previously, the commands would
first convert descriptor models to an explicit form before model simplification or decomposition,
respectively.

Compatibility Considerations
Numerical improvements to the algorithms used by modred, stabsep and modsep commands might
produce results that are different from the results you obtained using previous versions for descriptor
models.

Functionality being removed or changed

hsvd and hsvdOptions are not recommended
Still runs

The hsvd and hsvdOptions commands are not recommended. Use balred to obtain the Hankel
singular values (HSV) of dynamic system and also compute reduced-order model approximations
using the same command, and balredOptions to create the required HSV option set.
balredOptions also includes new options that preserve roll-off characteristics.

The following table shows some typical uses of hsvd and hsvdOptions, and how to update your
code to use balred and balredOptions instead.

Not Recommended Recommended

hsv = hsvd(sys) [~,info] = balred(sys) computes the
Hankel singular values and the error
approximations of sys. For more information, see
balred.
hsv = hsvd(sys,opts) [~,info] = balred(sys,opts) computes the
Hankel singular values with options specified in
the option set opts. For more information, see
balred and balredOptions.
hsvd(sys) balred(sys) displays the Hankel singular
values and approximation error on a plot. For
more information, see balred.
opts = hsvdOptions(Name,Value) opts = balredOptions(Name,Value) creates
the option set with the specified options. For
more information, see balredOptions.

balredOptions option StateElimMethod renamed to StateProjection

Still runs

4-4
The StateElimMethod option of the balredOptions command has been renamed to
StateProjection. This option specifies the method of eliminating weakly coupled states for model
reduction. For more information, see the 'StateProjection' option in the balredOptions reference
page.

Compatibility Considerations
If your code uses this option, consider modifying it to use the new option name.

Options to set error tolerance has changed in the balredOptions, stabsepOptions and
freqsepOptions commands
Still runs

The AbsTol and RelTol options for the balredOptions, freqsepOptions and stabsepOptions
commands have been replaced by SepTol. Use this option to specify the accuracy loss factor and
increasing SepTol helps separate modes straddling the stable/unstable or slow/fast boundary at the
expense of accuracy. Previously, you had to use two options to achieve the same result. For more
information, see the balredOptions, freqsepOptions and stabsepOptions function reference
pages.

Compatibility Considerations
If your code uses the AbsTol and RelTol options, consider modifying it to use the new SepTol
option.

The following table shows how to update your code to use the SepTol option instead.

Not Recommended Recommended

opts = opts = balredOptions(...,'SepTol',1e3)
balredOptions(...,'AbsTol',0.01,'RelTo creates the option set with the accuracy loss
l',0.02) factor. For more information, see
balredOptions.
opts = opts =
stabsepOptions(...,'AbsTol',0.01,'RelT stabsepdOptions(...,'SepTol',1e3)
ol',0.02) creates the option set with the accuracy loss
factor. For more information, see
stabsepOptions.
opts = opts =
freqsepOptions(...,'AbsTol',0.01,'RelT freqsepdOptions(...,'SepTol',1e3)
ol',0.02) creates the option set with the accuracy loss
factor. For more information, see
freqsepOptions.

Kalman Filter block: Numerical changes

Behavior change

Numerical improvements in the algorithms used by the Kalman Filter block might produce results
that are different from the results you obtained using previous versions.

modred, stabsep and modsep: Numerical changes

Behavior change

4-5
R2021a

Numerical improvements to the algorithms used by modred, stabsep and modsep commands might
produce results that are different from the results you obtained using previous versions for descriptor
models.

Support for opening SISO Design Tool sessions saved before release R2016a will be
removed
Warns

Support for opening SISO Design Tool sessions saved before release R2016a will be removed in
release R2021b.

If you have sessions saved before release R2016a, open and resave the session files using Control
System Designer in any release from R2016a through R2021a.

4-6
 5

R2020b

Version: 10.9

New Features

Bug Fixes

Compatibility Considerations
R2020b

sparss and mechss: Support for sparse state-space models

Use the new sparss and mechss objects to represent first-order and second-order sparse state-
space models, respectively. Using the sparse representation for state-space models increases
computational efficiency while reducing the amount of memory required for data storage. For more
information, see the sparss and mechss reference pages.

You can perform time-domain and frequency-domain analysis on sparss and mechss model objects
with functions listed in Time and Frequency Domain Analysis. Signal-based interconnections are also
supported for sparse model objects. Use the functions listed in Model Interconnection to interface
signals between sparse models and with other LTI models. For more information, see Sparse Model
Basics. For an example, see Transient Modeling and Linear Analysis of a Cantilever Beam.

Additionally, you can:

• Access the sparse matrices and sample time of the sparse model objects using sparssdata and
mechssdata.
• Obtain a summary of how the state vector is partitioned into components, interfaces, and signals
in your sparss and mechss model objects using showStateInfo.
• Sort states based on state partition using xsort.
• Specify the physical interface between two sparse models or between subcomponents of a mechss
object using the interface command. For an example, see Rigid Assembly of Model
Components.
• Use the Sparse Second Order block to represent your second-order sparse state-space model in
Simulink®. For an example, see Linearize Simulink Model to a Sparse Second-Order Model Object.
• Linearize and obtain a sparse model from a Simulink model when a Descriptor State-Space or a
Sparse Second Order block is present. For more information, see Sparse Model Basics. For an
example, see Linearize Simulink Model to a Sparse Second-Order Model Object.
• Convert between continuous-time and discrete-time and resample sparse models using c2d, d2c
and d2d. For available methods, see Sparse Model Basics.

New Example: Tune Phase-Locked Loop Using Loop-Shaping Design

The new example, Tune Phase-Locked Loop Using Loop-Shaping Design, shows how to tune the
components of a loop filter to improve the loop bandwidth of a phase-locked loop (PLL) system, using
systune to achieve the specified loop shape. The example uses the Mixed-Signal Blockset library to
model the PLL system.

Functionality being removed or changed

Extended and Unscented Kalman Filter Algorithms: Numerical changes
Behavior change

Numerical improvements in the algorithms used by the Extended Kalman Filter and Unscented
Kalman Filter blocks and the extendedKalmanFilter and unscentedKalmanFilter functions
might produce results that are different from the results you obtained using previous versions.

5-2
 6

R2020a

Version: 10.8

New Features

Bug Fixes
R2020a

Control System Designer: Perform automated loop shaping without a

Robust Control Toolbox license
You can now perform automated loop shaping in the Control System Designer without the
requirement of a license for Robust Control Toolbox™. Control System Designer uses looptune
instead of loopsyn in the absence of a Robust Control Toolbox license.

If you already have a Robust Control Toolbox license, you can choose one of the following options
from the Compensator Structure dropdown menu:

• Free-form Structure — Uses loopsyn to perform the automated loop shaping with the option
to specify your controller order preference using the Desired controller order slider.
• Fixed Structure — Uses looptune to perform the automated loop shaping without the option
to specify your controller order preference. When you select Fixed Structure, Control System
Designer will use the supplied controller order to perform the automated loop shaping.

For more information, see Design Compensator Using Automated Tuning Methods.

6-2
frdfun: Utility to apply a function across all frequencies of an frd
model
Use frdfun to apply a function across all frequencies of a frequency-response data model frd. The
function to be used on the frd model must take a single matrix and return a scalar, vector, or matrix
of fixed size across the frequencies.

For more information, see frdfun.

zgrid: Specify Sample Time to Plot True Frequencies on the Grid

zgrid now includes a new optional argument to specify the sample time T. You can use the syntax
zgrid(...,T) to plot true frequencies on the z-plane grid.

For more information, see zgrid.

6-3
 7

R2019b

Version: 10.7

New Features

Bug Fixes
R2019b

Live Editor Tasks: Perform model transformation and control design

tasks interactively and generate MATLAB code in a live script
Use new Live Editor tasks to perform model reduction, model rate conversion, and PID controller
tuning without writing code. The tasks generate plots that let you interactively explore the effects of
changing parameter values and options. The tasks also automatically generate code that becomes
part of your live script.

In R2019b, Control System Toolbox includes three tasks:

• Reduce Model Order — Reduce complexity of linear time-invariant (LTI) models

• Convert Model Rate — Convert LTI models between continuous time and discrete time;
resample discrete-time models
• Tune PID Controller — Tune PID controller gains for LTI plants

To use tasks in the Live Editor, on the Live Editor tab, in the Task menu, select a task. Alternatively,
in a code block in a live script, begin typing the task name and select the task from the suggested
command completions. For more information about Live Editor tasks generally, see Add Interactive
Tasks to a Live Script (MATLAB).

Residuals for Extended and Unscented Kalman Filters: Calculate

residuals and residual covariances of filter predictions
The new function residual computes residuals for extended Kalman filter (EKF) and unscented
Kalman filter (UKF) objects. These residuals represent the error between the filter predictions and
the measurements. Use residual to help validate filter performance.

For more information, see residual.

For more information on extended and unscented Kalman filter objects, see
extendedKalmanFilter and unscentedKalmanFilter.

ngrid, sgrid, and zgrid: Specify target axes as Axes and UIAxes to
create apps in App Designer
You can now use an Axes or UIAxes object with ngrid, sgrid, and zgrid to specify target axes
when creating apps in App Designer.

For more information, see ngrid, sgrid and zgrid.

Nyquist Plots: Programmatically Zoom on Critical Point

You can now use the zoomcp command to zoom programmatically on the critical point of a Nyquist
plot that you create with nyquistplot. Previously, you could zoom the plot to the critical point by
right-clicking and selecting Zoom on (-1,0). Now, for a plot with handle h, entering the command
zoomcp(h) achieves the same result as the right-click option. For an example, see the nyquistplot
reference page.

7-2
 8

R2019a

Version: 10.6

New Features

Bug Fixes

Compatibility Considerations
R2019a

getPIDLoopResponse: Obtain closed-loop and open-loop responses of

plant with PID controller
The new getPIDLoopResponse command generates responses for a control system formed by a PID
controller and plant model. The function returns the closed-loop, open-loop, controller action, or
disturbance response that you specify. Previously, you had to construct each response transfer
function manually, using interconnection commands such as feedback(G*C,1). For more
information, see getPIDLoopResponse.

icare and idare Commands: Solve continuous-time and discrete-time

Riccati equations
Use the new icare and idare commands to solve continuous-time and discrete-time Riccati
equations. icare replaces care and gcare for solving continuous-time Riccati equations while,
idare replaces dare and gdare for solving discrete-time Riccati equations. The new commands
have improved accuracy through better scaling and the computation of gain K is more accurate when
R is ill-conditioned.

For more information, see icare and idare.

Compatibility Considerations
• icare replaces care and gcare for solving continuous-time Riccati equations. The care and
gcare commands are not recommended. For more information, see “care and gcare are not
recommended” on page 8-2.
• idare replaces dare and gdare for solving discrete-time Riccati equations. The dare and gdare
commands are not recommended. For more information, see “dare and gdare are not
recommended” on page 8-3.

Functionality being removed or changed

care and gcare are not recommended
Still runs

The care and gcare commands are not recommended. Use icare to solve continuous-time implicit
Riccati equations instead. This approach has improved accuracy through better scaling and the
computation of K is more accurate when R is ill-conditioned relative to care and gcare.
Furthermore, icare includes an optional info structure to gather the implicit solution data of the
Riccati equation.

The following table shows some typical uses of care and gcare, and how to update your code to use
icare instead.

Not Recommended Recommended

• [X,L,G] = care(A,B,Q,R,S,E) [X,K,L] = icare(A,B,Q,R,S,E,G) computes
• [X,L] = gcare(H,J,NS) the stabilizing solution X, the state-feedback gain
K and the closed-loop eigenvalues L of the
continuous-time algebraic Riccati equation. For
more information, see icare.

8-2
Not Recommended Recommended
• [X,L,G,report] = care(A,B,Q,R,S,E) [X,K,L,info] = icare(A,B,Q,R,S,E,G)
• [X,L,report] = gcare(H,J,NS) computes the stabilizing solution X, the state-
feedback gain K, the closed-loop eigenvalues L of
the continuous-time algebraic Riccati equation.
The info structure contains the implicit solution
data. For more information, see icare.

There are no plans to remove care and gcare at this time.

dare and gdare are not recommended

Still runs

The dare and gdare commands are not recommended. Use idare to solve discrete-time implicit
Riccati equations instead. This approach has improved accuracy through better scaling and the
computation of K is more accurate when R is ill-conditioned relative to dare and gdare.
Furthermore, idare includes an optional info structure to gather the implicit solution data of the
Riccati equation.

The following table shows some typical uses of dare and gdare, and how to update your code to use
idare instead.

Not Recommended Recommended

• [X,L,G] = dare(A,B,Q,R,S,E) [X,K,L] = idare(A,B,Q,R,S,E) computes
• [X,L] = gdare(H,J,NS) the stabilizing solution X, the state-feedback gain
K and the closed-loop eigenvalues L of the
discrete-time algebraic Riccati equation. For
more information, see idare.
• [X,L,G,report] = dare(A,B,Q,R,S,E) [X,K,L,info] = idare(A,B,Q,R,S,E)
• [X,L,report] = gdare(H,J,NS) computes the stabilizing solution X, the state-
feedback gain K, the closed-loop eigenvalues L of
the discrete-time algebraic Riccati equation. The
info structure contains the implicit solution
data. For more information, see idare.

There are no plans to remove dare and gdare at this time.

8-3
 9

R2018b

Version: 10.5

New Features

Bug Fixes
R2018b

allmargin Function: New MIMO syntax for loop-at-a-time analysis

You can now use allmargin to compute loop-at-a-time classical stability margins of MIMO systems.
Previously, allmargin computed stability margins for SISO systems only.

For more information, see the allmargin reference page.

9-2
 10

R2018a

Version: 10.4

New Features

Bug Fixes
R2018a

Particle Filter Simulink Block: Estimate states of nonlinear systems for

online tracking and control system design
Perform state estimation for arbitrary nonlinear models using the new Particle Filter block in
Simulink. Particle filters are flexible in comparison to Kalman filters, that is, they can also perform
state estimation for nonlinear systems with non-Gaussian distributions.

Particle Filter block uses particles and sensor data to estimate the posterior distribution of the
current state. The filter predicts the states using the nonlinear state transition function. Then, it
corrects the estimate based on sensor data and measurement likelihood model. You can specify a
fixed number of particles to use, a fixed number of state variables to estimate, and your state
estimation method.

You can find the Particle Filter block in the Control System Toolbox > State Estimation block
library in Simulink.

You can use Simulink Coder™ to deploy particle filters with multiple measurement models and fixed-
size arrays for your application.

For more information on the Particle Filter block, see Particle Filter. For more information on the
detailed workflow, see Parameter and State Estimation in Simulink Using Particle Filter Block.

c2d Function: Convert models to discrete time using least-squares

optimization
You can now convert continuous-time dynamic system models to discrete time using a new least-
squares optimization method. This algorithm minimizes the error between the frequency responses of
the continuous-time and discrete-time systems up to the Nyquist frequency. This method is useful
when you want to capture fast system dynamics but must use a larger sample time, for example,
when computational resources are limited.

To convert a model using this approach, specify the discretization method as 'least-squares'.

discreteModel = c2d(contModel,Ts,'least-squares');

Alternatively, you can create a c2dOptions option set, and set the Method property to 'least-
squares'. You can then use this option set with the c2d function.

options = c2dOptions('Method','least-squares');
discreteModel = c2d(contModel,Ts,options);

This conversion method supports only SISO models.

For more information, see c2d and c2dOptions.

Control System Designer: Change sample time of control system

You can now modify the sample time of your control system in Control System Designer. To do so,
on the Control System tab, under Edit Architecture, click Sample Time Conversion.

10-2
In the Sample Time Conversion dialog box, specify the Sample time, and select a Conversion
method for each block in the control system. For more information on the available conversion
methods, see Continuous-Discrete Conversion Methods. Control System Designer does not support
the new least-squares vector fitting approach.

Click OK.

The app converts the dynamic model of each block to discrete time using the specified sample time
and conversion method.

10-3
R2018a

If your model is already in discrete time, you can choose to convert it to continuous time or to
resample the system using a different sample time.

Control System Designer: Create Simulink model for control

architecture
You can now generate a Simulink model for your tuned control system architecture in Control
System Designer.

For more information, see Generate Simulink Model for Control Architecture.

10-4
 11

R2017b

Version: 10.3

New Features

Bug Fixes

Compatibility Considerations
R2017b

Gain Scheduling: Implement gain-scheduled controllers using a new

library of blocks configured to take scheduled parameters as inputs
A new library of Simulink blocks lets you implement common control-system elements with variable
parameters. The new blocks in the Linear Parameter Varying library take parameter values as inputs,
letting you compute those values at run-time in your model.

These blocks are useful for implementing gain-scheduled controller elements in which the parameter
values vary as a function of scheduling variables. For instance, the new Varying Notch Filter block
accepts as inputs the notch frequency, the gain at the notch frequency, and the damping ratio of the
filter poles. In your model, you connect each of these inputs to an element such as a lookup table or
MATLAB function block that computes the parameter value from the scheduling variables.

The available variable-parameter control-system elements include:

• Varying Transfer Function, Discrete Varying Transfer Function

• Varying Lowpass Filter, Discrete Varying Lowpass
• Varying Notch Filter, Discrete Varying Notch
• Varying State Space, Discrete Varying State Space
• Varying Observer Form, Discrete Varying Observer Form

The library also includes versions of the PID Controller and PID Controller (2DOF) blocks that are
preconfigured to accept PID coefficients as inputs.

To access the new library, in the Simulink Library Browser, select Control System Toolbox > Linear
Parameter Varying. For more information about using the blocks to implement gain-scheduled
control systems, see Model Gain-Scheduled Control Systems in Simulink.

Gain Scheduling: Achieve smooth and memory-efficient

implementation by turning gain surfaces into embedded equations
In a gain-scheduled control system, controller gains vary as a function of one or more scheduling
variables. In a Simulink model, one way to implement a scheduled gain is to use a MATLAB function
block to represent the equations relating controller gains to scheduling-variable values. You can now
use systune to tune those equations automatically and write the resulting relationship back to the
MATLAB function block. Previously, you had to convert the tuned relationship to a discrete lookup
table relating scheduling variables to gain values. While you can still use this approach, the new
functionality can achieve smoother variation of scheduled gains by evaluating the equation that
relates gains to scheduling variables.

To tune MATLAB function blocks representing scheduled gain values, you parameterize each
MATLAB function block with a tunableSurface object that represents the equation relating gain to
scheduling parameters. When you write the tuned parameters back to your model, the MATLAB
function is automatically updated with MATLAB® code for the tuned gain surface. Use the new
codegen command to examine the MATLAB code for the gain schedule.

If you have a code-generation product such as Simulink Coder, you can implement the tuned gain
schedule in hardware. The new functionality can result in more memory-efficient implementation,
storing only the gain surface coefficients rather than a potentially long list of lookup-table values.

For more information about modeling and tuning gain-scheduled control systems in Simulink, see:

11-2
• Model Gain-Scheduled Control Systems in Simulink
• Tune Gain Schedules in Simulink

Tuning control systems modeled in Simulink requires Simulink Control Design™ software.

Gain-Scheduled Controller Tuning: Automatically tune gain-scheduled

state observer gain, LQR gain, and other controller architectures
expressed as matrices
You can now use systune for automatic tuning of matrix-valued gain schedules and implement them
in MATLAB function blocks or Matrix Interpolation blocks. For instance, suppose that you want to
implement a time-varying LQG controller of the form:

dxe = Axe + Bu + L t y − Cxe − Du

u = − K t xe,

where the state-feedback matrix K and the observer-gain matrix L vary with time. In your Simulink
model, use the new Varying Observer Form block to represent the LQG controller. Then use MATLAB
function blocks to implement the time-varying matrices as inputs to the Varying Observer Form block.
When you set up the model for tuning, parameterize the MATLAB function block using a
tunableSurface that computes a matrix value as a function of time. When you tune the control
system with systune and write the tuned controller parameters back to the model, the MATLAB
function block is automatically updated with MATLAB code for the tuned gain surface.

For more information about:

• The Varying Observer Form block and other new variable-parameter blocks for gain scheduling,
see “Gain Scheduling: Implement gain-scheduled controllers using a new library of blocks
configured to take scheduled parameters as inputs” on page 11-2.
• Modeling gain-scheduled control systems in Simulink, see Model Gain-Scheduled Control Systems
in Simulink.
• Creating gain surfaces for tuning gain schedules, see Parameterize Gain Schedules.

Tuning control systems modeled in Simulink requires Simulink Control Design software.

Gain-Scheduled Controller Tuning: Specify tuning goals that vary with

operating condition
When tuning fixed or gain-scheduled controllers at multiple design points representing different
operating conditions, it is sometimes useful to adjust the design objectives as a function of operating
condition. For example, you might want to relax design requirements in some regions of the
operating range. The new varyingGoal function lets you construct tuning goals that depend
implicitly or explicitly on the design point.

For more information about configuring varying requirements for gain scheduling, see:

• The varyingGoal reference page

• Change Requirements with Operating Condition

11-3
R2017b

Tuning Gain Surfaces: Custom normalization, lookup-table updates,

and other enhancements
This release includes enhancements to tunable gain surfaces that improve the workflow for tuning
gain schedules.

• Custom normalization of scheduling variables in tunable surface — By default, the

tunableSurface representation of a gain surface normalizes the scheduling variables so that the
design range of each variable falls within the interval [–1,1]. You can change this normalization
using the new Normalization property of tunableSurface. Changing the normalization is
useful, for example, when you have a known gain value for a design point, or want to restrict a
scheduling variable to positive values. For more information, see tunableSurface.
• Update portion of lookup table — You can now update a single point or a portion of a lookup table
in a Simulink model that you tuned using slTuner and systune. To perform the update, use the
new writeLookupTableData command. This command is useful when you are retuning a single
design point or a subset of design points covered by the lookup table. Previously, you could only
use writeBlockValue to update the entire lookup table.
• Name scheduling variables in basis functions — New syntaxes for polyBasis and fourierBasis
let you assign names to the input variables of the basis functions generated by these commands.
The names are preserved when you combine basis functions using ndBasis. Naming basis-
function variables can improve readability of the generated basis functions and of code generated
from a tunableSurface object that you create with the basis functions. For more information,
see the reference pages for these commands.
• Tunable surface with constant gain — A new syntax of tunableSurface lets you create a flat
gain surface with constant, tunable gain. For more information, see tunableSurface.

Gain-Scheduled Controller Tuning: Exclude design points from tuning

or analysis
When you have created a design grid of tunable, linearized models for gain-scheduled controller
tuning, you can now exclude one or more design points from tuning without removing the
corresponding model from the array. Doing so can be useful, for example, to identify problematic
design points when tuning over the entire design grid fails to meet your design requirements. It can
also be useful when there are design points that you want to exclude from a particular tuning run, but
preserve for performance analysis or further tuning. To exclude design points from tuning, use the
new SkipModels option of systuneOptions, which lets you specify models in the design grid to
exclude from tuning.

As an alternative, you can eliminate design points from the model grid entirely, so that they do not
contribute to any stage of tuning or analysis. To do so, use the new voidModel command, which
replaces specified models in a model array with NaN. This approach is useful when your sampling grid
includes points that represent irrelevant or unphysical design points. Using voidModel lets you
design over a grid of design points that is almost regular.

For more information about controlling how different design points contribute to tuning, see Change
Requirements with Operating Condition.

Particle Filters: Estimate states of nonlinear systems

Perform state estimation for arbitrary nonlinear system models using the new particleFilter
command. Particle filters are flexible, that is, they can also perform state estimation for nonlinear

11-4
systems with non-Gaussian distributions. Previously, you could perform state estimation only for
systems with unimodal distributions using extended or unscented Kalman filters.

particleFilter uses particles and sensor data to estimate the posterior distribution of the current
state. The filter predicts the states using the nonlinear state transition function. Then, it corrects the
estimate based on sensor data and measurement likelihood model. You can specify a fixed number of
particles to use, a fixed number of state variables to estimate, and your state estimation method
based on the particle weights.

To use a particle filter for state estimation:

1 Create a particle filter, and set the transition and measurement likelihood functions.
2 Initialize the particle filter by specifying the number of particles to be used and your initial state
guess. Also specify state bounds or covariance of the initial particle distribution.
3 Specify the state estimation and resampling method.
4 Perform state estimation.

You can use MATLAB Compiler™ or MATLAB Coder software to deploy the particle filter for your
application.

For more information and examples, see the particleFilter reference page.

Improved lqg Function: Compute gain matrices and optimal controller

in discrete time using current Kalman Filter estimator
When designing an LQG controller for a discrete-time plant using the lqg function, you can now use
the current type of Kalman estimator, which uses x[n|n] as the state estimate. Previously, the lqg
function supported using only the delayed type of Kalman estimator; that is, using x[n|n-1] as the
state estimate. For more information about the types of Kalman estimators, see kalman.

Also, you can now return the controller and estimator gain matrices when using the lqg function. You
can use the controller and estimator gains to, for example, implement the controller in observer form.

For more information, see lqg.

Model Reduction: balred no longer ignores MatchDC option when

specified frequency or time intervals exclude DC
When you use balred for model reduction, you can use balredOptions to restrict the computation
to specified frequency or time intervals. If the StateElimMethod option of balredOptions is set to
'MatchDC' (the default value), then balred attempts to match the DC gain of the original and
reduced models, even if the specified intervals exclude DC (frequency = 0 or time = Inf). This
behavior might reduce the quality of the match in the specified intervals. To improve the match
within intervals that exclude DC, set StateElimMethod = 'Truncate'. For more information, see
balredOptions.

In the Model Reducer app, there is no change in behavior.

Compatibility Considerations
Previously, if time or frequency intervals excluded DC, balred did not attempt to match the DC gain
of the original and reduced models, even if StateElimMethod = 'MatchDC'. If you have scripts or

11-5
R2017b

functions that use balred with restricted time or frequency intervals that exclude DC, consider
updating them to set StateElimMethod = 'Truncate'. The balred command now issues a
warning when StateElimMethod = 'MatchDC' and the specified time or frequency intervals
exclude DC.

Regularization of conic-sector tuning goal in Control System Tuner

The Conic Sector Goal in the Control System Tuner app has a new Regularization option that
allows you to specify a nonzero regularization parameter. This option is useful when other tuning
goals tend to make the sector bound ill-conditioned at some frequencies. For more information, see
Conic Sector Goal.

This new option is equivalent to the Regularization property of TuningGoal.ConicSector,

introduced in R2017a for command-line tuning.

Dynamic system models store Notes property as string or character

vector
The Notes property of a dynamic system model stores any text that you want to associate with the
model. This property now accepts either character-vector or string values, and stores whichever
type you provide. For instance, if sys1 and sys2 are dynamic system models, you can set their
Notes properties as follows:

sys1.Notes = "sys1 has a string.";

sys2.Notes = 'sys2 has a character vector.';
sys1.Notes
sys2.Notes

ans =

"sys1 has a string."

ans =

1×1 cell array

{'sys2 has a character vector.'}

When you create a new model, the default value of Notes is now [0×1 string]. Previously, you
could only specify the Notes property as a character vector or cell array of character vectors, and
the default value was {}.

Some other dynamic system model properties accept strings as inputs, but store the values as
character vectors or a cell array of character vectors.

11-6
 Functionality being removed or changed
Functionality Resu Use This Instead Compatibility Considerations
lt
balred with Warn StateElimMethod = See “Model Reduction: balred no
StateElimMethod = s 'Truncate' longer ignores MatchDC option
'MatchDC' and restricted time when specified frequency or time
or frequency range that intervals exclude DC” on page 11-5.
excludes DC
viewSpec, evalSpec Still viewGoal, evalGoal If you have functions or scripts that
work use viewSpec or evalSpec,
s consider updating them to use
viewGoal and evalGoal instead.

11-7
 12

R2017a

Version: 10.2

New Features

Bug Fixes

Compatibility Considerations
R2017a

Extended and Unscented Kalman Filter Simulink Blocks: Estimate

states of nonlinear systems
You can now use the Extended Kalman Filter and Unscented Kalman Filter blocks to estimate the
states of a discrete-time nonlinear system in Simulink. The blocks use first-order extended and
unscented Kalman filter algorithms to estimate states as new data becomes available during the
operation of the system. Previously, nonlinear state estimation using these algorithms was available
at the command line only. You can use the state estimates for state-feedback controllers and for
applications such as condition monitoring and fault detection. You can also generate C/C++ code for
these blocks using Simulink Coder software.

For information about how to use these blocks, see the Extended Kalman Filter and Unscented
Kalman Filter block reference pages. For examples, see Estimate States of Nonlinear System with
Multiple, Multirate Sensors and Nonlinear State Estimation of a Degrading Battery System.

New properties of generalized state-space models and matrices

The generalized state-space model object, genss, has the following new properties:

• StateName and StateUnit — Track the state names and state units of the fixed LTI components
and control design blocks that make up the model. When you build a genss model from fixed and
tunable LTI components, it inherits the StateName and StateUnit values from these
components
• A, B, C, D, and E — Access the state-space matrices of a genss model. These properties model the
dependency of the state-space matrices on static control design blocks, such as realp, ureal,
ucomplex, or ucomplexm.

For more information and examples, see the genss reference page.

Additionally, the generalized matrix object, genmat, now has a Name property. Use the property to
assign a name to the generalized matrix. When you convert a static control design block such as
tunableSurface to a generalized matrix using genmat(blk), the Name property of the block is
preserved.

Discrete-time frequency-dependent specifications for tuning goals

You can now use discrete-time LTI models to specify frequency-dependent gain profiles for tuning in
discrete time. Tuning goals that you can now specify in discrete time include:

• Frequency-dependent minimum gains, maximum gains, and loop gains.

• Frequency-dependent rejection, sensitivity, or error profiles.
• Frequency-dependent weighting functions.

If you specify the gain profile in continuous time for tuning in discrete time, the tuning software
discretizes the profile. Specifying the gain profile in discrete time gives you more control over the
gain profile near the Nyquist frequency. For more information, see the documentation for the
individual tuning goals.

Regularization of tuning goals for improved numeric stability

When you use a tuning goal with a frequency-dependent specification, the tuning algorithm uses a
frequency-weighting function to compute the normalized value of the tuning goal. This weighting

12-2
function is derived from the gain profile that you specify. For numeric stability and tractability, the
software now adjusts the specified gain profile when necessary to eliminate undesirable low-
frequency or high-frequency dynamics or asymptotes. This adjustment process is called
regularization.

The regularized gain profile is displayed on tuning-goal plots generated with viewSpec or in Control
System Tuner. For affected tuning goals, the getWeight or getWeights command extracts the
regularized frequency-weighting functions. For more information about regularization, see Visualize
Tuning Goals and the documentation for the individual tuning goals.

Also, the conic sector goal has a new Regularization property that allows you to specify a nonzero
regularization parameter. This property is useful when other tuning goals tend to make the sector
bound ill-conditioned at some frequencies. For more information, see TuningGoal.ConicSector.

Maximum Natural Frequency Option in Control System Tuner: Prevent

poles and zeros from going to infinity
Most tuning goals in the Control System Tuner app include implicit stability or minimum-phase
constraints. The new Maximum natural frequency tuning option constrains the maximum natural
frequency of the corresponding stabilized poles and zeros. This option is useful to prevent poles and
zeros from going to infinity as a result of algebraic loops becoming singular or control effort growing
unbounded. To access the option in the app, on the Tuning tab, click Tuning Options.

For more information about stabilized poles and zeros, see the documentation for each tuning goal,
listed on the Tuning Goals page.

The new option is equivalent to the MaxRadius option of systuneOptions, introduced in R2016b
for command-line tuning.

Scaling information passed automatically to viewSpec and evalSpec

When you use systune to tune a MIMO feedback loop, some tuning goals are sensitive to the relative
scaling of each SISO loop. systune tries to balance the overall loop-transfer matrix while enforcing
such goals. The optimal loop scaling is now stored in the tuned closed-loop model returned by
systune. When you pass the tuned model to viewSpec or evalSpec to examine tuning results,
these functions take this scaling into account. Previously, you had to pass the info output of
systune to these functions to ensure consistent scaling.

For more information, see viewSpec or evalSpec.

Compatibility Considerations
You no longer need to use the syntaxes viewSpec(Req,CL,info) and evalSpec(Req,CL,info)
to ensure consistent scaling. The syntaxes viewSpec(Req,CL) and evalSpec(Req,CL) obtain any
necessary scaling information from the closed-loop model CL returned by systune. To force the
functions to disregard scaling information, use viewSpec(Req,CL,[]) or evalSpec(Req,CL,[]).

12-3
R2017a

Functionality being removed or changed

Functionality Resu Use This Instead Compatibility Considerations
lt
Q property of Still SectorMatrix If you have scripts or functions that
TuningGoal.ConicSector work refer to the Q property of a
s TuningGoal.ConicSector
object, update them to use the new
property name SectorMatrix.
• viewSpec(Req,CL,info) Still • viewSpec(Req,CL) If you have scripts or functions that
• evalSpec(Req,CL,info) work • evalSpec(Req,CL) use the info argument, consider
s updating them to remove the
argument. See “Scaling information
passed automatically to viewSpec
and evalSpec” on page 12-3.

12-4
 13

R2016b

Version: 10.1

New Features

Bug Fixes
R2016b

Conic Sector Tuning Goal: Tune control systems to enforce fixed or

frequency-dependent sector bounds
A conic system is a system whose trajectories lie in a given conic sector of I/O space. In some control
applications, it is useful to restrict system trajectories to a particular sector. Such bounds arise, for
example, in robust control of feedback loops with static nonlinearities. New goals for control system
tuning let you impose conic sector bounds on responses of the tuned system.

• Use TuningGoal.ConicSector for tuning at the command line with systune.

• Use Conic Sector Goal for tuning in Control System Tuner.

For more information about conic systems, see About Sector Bounds and Sector Indices.

Improved Passivity Tuning Goal: Set output passivity index to a

negative value
You can now specify a negative output passivity index when tuning a control system using systune
or Control System Tuner. Specifying a negative output passivity index for a particular response lets
the tuned response have a shortage of passivity at the output. Negative output passivity index values
are permitted in:

• TuningGoal.Passivity and TuningGoal.WeightedPassivity, for tuning at the command

line with systune.
• Passivity Goal and Weighted Passivity Goal, for tuning with Control System Tuner.

Previously, you could specify a negative or positive input passivity index with these tuning goals, but
only a positive output passivity index.

For more information about passivity indices, see About Passivity and Passivity Indices.

MaxRadius Option for Tuning: Prevent poles and zeros from going to
infinity
Most tuning goals used for control system tuning with systune include implicit stability or minimum-
phase constraints. The new MaxRadius option of systuneOptions constrains the maximum natural
frequency of the corresponding stabilized poles and zeros. This option is useful to prevent poles and
zeros from going to infinity as a result of algebraic loops becoming singular or control effort growing
unbounded. For more information, see the systuneOptions reference page and the reference pages
for each tuning goal.

Improved getSectorIndex and sectorplot Functions: Compute and plot

sector index for unstable systems
You can now use getSectorIndex and sectorplot to analyze the sector index for both stable and
unstable systems. Previously, providing an unstable system as input to either of these functions
generated an error.

13-2
Extended and Unscented Kalman Filters: Estimate states of nonlinear
systems
You can now estimate the states of discrete-time nonlinear systems at the command line using first-
order extended Kalman filter algorithms and unscented Kalman filter algorithms. The new state
estimation commands extendedKalmanFilter and unscentedKalmanFilter are useful for
estimation of states when new data is available during the operation of the system. You can use the
state estimates for state-feedback controllers and for applications such as condition monitoring and
fault detection. You can use MATLAB Compiler or MATLAB Coder software to deploy the estimators
in your application.

For an example of online state estimation, see Nonlinear State Estimation Using Unscented Kalman
Filter.

Phase-Wrap Branch Option: Specify cutoff point for wrapping phase in

response plots
By default, response plots that show phase response, such as Bode and Nichols plots, display the
exact phase. You can make these plots wrap the phase into the interval [–180º,180º) by checking
Wrap Phase in the plot Property Editor, the Linear System Analyzer Preferences Editor, or the
Toolbox Preferences Editor.

In R2016b, checking Wrap Phase enables a new Branch field that lets you specify the value at
which accumulated phase wraps in the response plot. For example, entering 0 causes the plot to wrap
the phase into the interval [0º,360º).

At the command line, turn on phase wrapping by setting the PhaseWrapping option of
bodeoptions or nicholsoptions to 'on'. Specify the phase-wrap cutoff point using the new
PhaseWrappingBranch option.

In R2015b and R2016a, phase-wrapped plots used the interval [0º,360º). Before R2015b, phase-
wrapped plots used the interval [–180º,180º).

13-3
 14

R2016a

Version: 10.0

New Features

Bug Fixes

Compatibility Considerations
R2016a

Redesigned Control System Designer App: Design SISO controllers for

feedback systems using improved interactive workflows
The redesigned Control System Designer app streamlines workflows for designing SISO controllers
for feedback control systems using graphical and automated tuning methods.

For more information on using Control System Designer, see:

• Control System Designer

• Control System Designer Tuning Methods
• Bode Diagram Design
• Design Compensator Using Automated Tuning Methods
• Analyze Designs Using Response Plots

14-2
Control System Tuner App and systune Command: Automatically tune
single-loop and multiloop control systems to meet design
requirements
Control System Toolbox now includes automatic tuning tools that previously required a Robust
Control Toolbox license. Control System Tuner and the systune command automatically tune control
systems from high-level design goals you specify, such as reference tracking, disturbance rejection,
and stability margins.

To tune a control system, you specify the tunable elements of your control system. You then capture
your design requirements using the library of tuning goals. The software jointly tunes all the free
parameters of your control system regardless of architecture, number of feedback loops, or whether
it is modeled in MATLAB or Simulink. (Tuning control systems modeled in Simulink requires a
Simulink Control Design license.)

For information about using these tools, see:

• Tuning with Control System Tuner

• Programmatic Tuning

You can also use the systune command to tune gain-scheduled controllers for control systems in
which plant dynamics change with operating conditions or time. For more information, see Gain
Scheduling.

Model Reducer App: Compute and compare reduced-order models

using interactive workflows
The new Model Reducer app is an interactive tool for computing reduced-order approximations of
high-order models. Working with lower-order models can simplify analysis and control design.
Simpler models are also easier to understand and manipulate. You can reduce a plant model to focus
on relevant dynamics before designing a controller for the plant. Or, you can use model reduction to
simplify a full-order controller. Using any of the following model-reduction methods, Model Reducer
helps you reduce model order while preserving model characteristics that are important to your
application:

• Balanced Truncation — Remove states with low energy contributions.

• Pole/Zero Simplification — Eliminate canceling or near-canceling pole-zero pairs.
• Mode Selection — Select modes by specifying a region of interest in the complex plane.

Model Reducer also provides response plots and error plots to help ensure that the reduced-order
model preserves important dynamics.

For an example showing how to use Model Reducer, see Reduce Model Order Using the Model
Reducer App. For more information about model reduction generally, see Model Reduction Basics.

Passivity and Conic Sectors: Analyze and tune control systems for
passivity and other sector bounds
A linear system is passive when it cannot produce energy on its own and can only dissipate the
energy initially stored in it. More generally, an I/O map is passive if, on average, increasing the output

14-3
R2016a

requires increasing the input. Passive control is often part of the safety requirements in process
control, tele-operation, human-machine interfaces, and system networks.

Use the following new commands to analyze the passivity of linear systems:

• isPassive — Check passivity of linear system.

• getPassiveIndex — Compute various measures of the excess or shortage of passivity of a linear
system.
• passiveplot — Calculate and plot passivity index as a function of frequency.

Mathematically, a system is passive when all its I/O trajectories are restricted to a particular sector of
the I/O space. More generally, a conic system is a system whose trajectories lie in a given conic
sector. Conic sector bounds arise, for example, in robust control of feedback loops with static
nonlinearities. The following new commands let you analyze how well a linear system lies within any
conic sector.

• getSectorIndex — Check whether the output trajectories of a linear system lie in a particular
conic sector. Compute the relative sector index, a measure of how tightly the trajectories fit within
the sector.
• getSectorCrossover — Compute the frequencies at which the range of trajectories crosses the
sector boundary.
• sectorplot — Calculate and plot sector index as a function of frequency.

New tuning goals let you enforce passivity when tuning a control system using Control System Tuner
or the systune command.

Constraint Command Line Control System Tuner

Enforce passivity on specified TuningGoal.Passivity Passivity Goal
I/Os in the control system
Enforce passivity on specified TuningGoal.WeightedPassi Weighted Passivity Goal
I/Os with frequency-dependent vity
weighting

For more background and details about the notions of passivity and sector bounds, see:

• About Passivity and Passivity Indices

• About Sector Bounds and Sector Indices

Limited Balanced Truncation: Reduce model order according to

energies within time-domain and frequency-domain intervals
You can now perform balanced-truncation model reduction and compute Hankel singular values
based on state energies calculated within specified time and frequency intervals.

• To perform frequency-limited or time-limited balanced truncation, use balredOptions to set the

TimeIntervals or FreqIntervals options for the balred command. When you use these
options, balred determines which states to truncate based on their energy contributions within
the specified interval only. For more details, see the reference pages for balredOptions and
balred.

14-4
• To compute or plot Hankel singular values for specific time or frequency, use hsvdOptions or
hsvoptions to set the TimeIntervals and FreqIntervals options for the hsvd and hsvplot
commands, respectively. For more details, see the reference page for hsvdOptions.

These operations use the new functionality in the gram command that computes time-limited and
frequency-limited controllability and observability Gramians. For details, see the reference pages for
gram and the new gramOptions command.

sampleBlock and rsampleBlock commands for sampling generalized

models
The new sampleBlock and rsampleBlock commands sample Control Design blocks in a
generalized model such as a genss or uss model. You can sample tunable blocks, uncertain blocks,
or both. The output is the model array obtained by replacing the sampled blocks by values you specify
(sampleBlock) or randomized values (rsampleBlock). Some uses of these functions include:

• Perform sensitivity analysis by varying tunable parameters randomly or across a grid.

• Study robustness by varying parameters over an uncertainty range.

For more information, see the sampleBlock and rsampleBlock reference pages.

Spectral factorization of LTI models

The new spectralfact command computes the spectral factorization of LTI models. The spectral
factorization of a model H is:

H = G'*S*G,

where S is a symmetric matrix and G is a square, stable, and minimum-phase system with unit
(identity) feedthrough. H must satisfy H = H'. For more information, see the spectralfact
reference page.

Renamed tunable control design blocks

The tunable control design blocks have been renamed. Starting in R2016a, use the following block
names:

Control Design Block New Name

Tunable gain block tunableGain
Fixed-order state-space model with tunable tunableSS
coefficients
SISO fixed-order transfer function with tunable tunableTF
coefficients
One-degree-of-freedom PID controller with tunablePID
tunable coefficients
Two-degree-of-freedom PID controller with tunablePID2
tunable coefficients

Also, several properties of tunable state-space and tunable transfer function blocks have changed.
For more information, see “Functionality being removed or changed” on page 14-6.

14-5
R2016a

The remaining block functionality and properties remain unchanged.

Compatibility Considerations
If your code uses tunable control design blocks, modify your code to use the new block names. For
more information, see “Renamed tunable control design blocks” on page 14-5.

Functionality being removed or changed

Functionality Resu Use This Instead Compatibility Considerations
lt
ltiblock.gain, Still tunableGain, tunableSS, For more information, see
ltiblock.ss, ltiblock.tf, work tunableTF, tunablePID, and “Renamed tunable control design
ltiblock.pid, and s tunablePID2 respectively. blocks” on page 14-5.
ltiblock.pid2
a, b, c, d, and e properties of ss Still A, B, C, D, and E respectively. If your code uses any of these
and tunableSS models. work properties, consider modifying your
s code to use the new property
names. For more information, see
ss.
num, den, and ioDelay Still Numerator, Denominator, and If your code uses any of these
properties of tf and work IODelay respectively. properties, consider modifying your
tunableTF models. s code to use the new property
names. For more information, see
tf.
z, p, k, and ioDelay properties Still Z, P, K, and IODelay If your code uses any of these
of zero-pole-gain models work respectively. properties, consider modifying your
s code to use the new property
names. For more information, see
zpk.

14-6
 15

R2015b

Version: 9.10

New Features

Bug Fixes
R2015b

pid2 and pidstd2 Model Objects: Represent, analyze, and use 2-DOF
PID controllers for control design
The new numeric LTI models pid2 and pidstd2 are specialized for modeling two-degree-of-freedom
(2-DOF) PID controllers. These are analogous to the 1-DOF PID models pid and pidstd.

Use pid2 and pidstd2 to represent a 2-DOF PID controller directly with the PID parameters,
expressed in parallel (pid2) or standard (pidstd2) form. For example, C2 =
pid2(Kp,Ki,Kd,Tf,b,c) creates a 2-DOF PID controller in parallel form with proportional,
integral, and derivative gains Kp, Ki, and Kd, derivative filter time constant Tf, and setpoint weights
b and c. In previous releases, to model a 2-DOF PID controller, you had to derive the controller's
equivalent transfer function (or other model), and could not directly store the 2-DOF PID parameters.

The pid2 and pidstd2 commands can also convert to PID form any type of LTI object that
represents a 2-DOF PID controller.

This release also includes new functions to help you work with 2-DOF PID controllers. These
functions include:

• getComponents — Extract two SISO control components from a 2-DOF pid2 or pidstd2
controller.
• make1DOF and make2DOF — Convert 1-DOF pid and pidstd controllers to 2-DOF pid2 and
pidstd2 controllers, and vice versa.
• piddata2 and pidstddata2 — Access parameters of 2-DOF PID controllers.

For more information about working with 2-DOF PID controller objects, see:

• Two-Degree-of-Freedom PID Controllers

• pid2 and pidstd2 function reference pages

2-DOF PID Controller Tuning: Automatically tune the gains of 2-DOF

PID controllers with PID Tuner app and pidtune command
You can now use pidtune and PID Tuner to tune all parameters of a two-degree-of-freedom (2–DOF)
PID controller, including the setpoint weights b and c. When you call pidtune or open the PID Tuner
app with a plant, the software automatically tunes all parameters of the block to achieve a balance
between performance and robustness. When you use the Response Time and Transient Behavior
sliders to adjust that balance, PID Tuner adjusts all parameters, including b and c if necessary.

PID Tuner and pidtune also include options for tuning 2-DOF PID controllers with fixed setpoint
weights, such as I-PD (b = 0, c = 0) and P-ID (b = 1, c = 0).

For more information, see:

• Tune 2-DOF PID Controller (PID Tuner)

• Tune 2-DOF PID Controller (Command Line)
• PID Controller Types for Tuning

15-2
Save Current Controller Design as Baseline in PID Tuner
In PID Tuner, you can now make the current controller design the baseline controller at any time.
This feature allows to you compare the performance of a PID Tuner controller design to the
performance of controllers obtained by further adjustment of the design.

To make the current PID Tuner design the baseline controller, by click the Export arrow and select
Save as Baseline.

When you do so, the current Tuned response becomes the Baseline response. Further adjustment of
the current design creates a new Tuned response line.

Previously, you could only designate a baseline controller when you opened the PID Tuner using the
syntax pidTuner(sys,C0).

For more information about analyzing controller performance in PID Tuner, see Analyze Design in PID
Tuner.

Change in LPV System block default values for model delays

The default value is now false for Input delay, Output delay, and Internal delay, in the Fixed
Entries tab of LPV System Block Parameters dialog box. A false value means that model delays are
treated as free during simulation.

For information about changing the default values, see the LPV System block reference page.

Analysis Plots Wrap Phase in Interval [0º,360º)

By default, response plots that show phase response, such as Bode and Nichols plots, display the
exact phase. You can make these plots wrap the phase into the interval [0º,360º) by unchecking

15-3
R2015b

Unwrap Phase in the plot Property Editor. Previously, unchecking this option caused the plots to
wrap the phase into [–180º,180º). The change makes it easier to visualize 180º phase crossings on
analysis plots.

The default plot behavior (exact or unwrapped phase) is unchanged. The behavior of commands that
return numerical phase response data, such as bode, is also unchanged. These commands always
return the unwrapped phase.

15-4
 16

R2015a

Version: 9.9

New Features

Bug Fixes

Compatibility Considerations
R2015a

Improved input disturbance rejection with the PID tuning algorithm

Controller tuning with the PID Tuner app or the pidtune command now yields better disturbance
rejection by default. For a given target phase margin, the tuning algorithm selects PID coefficients
that achieve a balance between reference tracking and input disturbance rejection. If you require
more disturbance rejection or better reference tracking than the default algorithm provides, PID
Tuner and pidtune have a new Design Focus option. Use this option to alter the balance that the
tuning algorithm sets between reference tracking and input disturbance rejection. For instance,
setting the design focus to reference tracking improves the reference tracking performance of the
tuned controller, with some cost to disturbance rejection. Similarly, setting the design focus to input
disturbance rejection improves rejection with some cost to reference tracking. Changing design focus
is most effective when tuning PID and PIDF controllers, rather than controllers with fewer free
parameters, such as PI.

To use the Design Focus option in PID Tuner, click Options and select a design focus from the Focus
menu.

You can still use the Response Time and Transient Behavior sliders to further adjust the balance
between reference tracking and input disturbance rejection.

To specify a design focus for command-line tuning with pidtune, use pidtuneOptions to set the
DesignFocus option. For example, the following commands design a PIDF controller for the plant G
with a crossover frequency of 10 rad/s, specifying reference tracking as the design focus.

opt = pidtuneOptions('DesignFocus','reference-tracking');
C = pidtune(G,'pidf',10,opt);

For more information about using the design focus option, see:

• Tune PID Controller to Favor Reference Tracking or Disturbance Rejection (PID Tuner)
• Tune PID Controller to Favor Reference Tracking or Disturbance Rejection at Command Line

16-2
For more information about using PID Tuner, see Designing PID Controllers with the PID Tuner. For
more information about command-line PID tuning, see PID Controller Design at the Command Line.

Option to specify code generation settings in LPV System block

You can now specify code generation settings in the LPV System block. You specify these settings in
the Code Generation tab of the block parameters dialog box.

For more information on Linear Parameter-Varying models, see Linear Parameter-Varying Models.

connect command syntax for specifying analysis point locations

When you interconnect dynamic system models using the connect command, you can now specify
analysis point locations as an input argument to the command. The following syntax creates a
dynamic system model with analysis points by interconnecting multiple models
sys1,sys2,...,sysN:
sys = connect(sys1,sys2,...,sysN,inputs,outputs,APs);

inputs and outputs are string vectors that specify the names of the inputs and outputs of the
interconnected model. APs is a string vector that lists the signal locations at which to insert analysis
points. The software automatically inserts an AnalysisPoint block with channels corresponding to
these locations. Previously, you had to create AnalysisPoint blocks separately and include them in
the list of models to connect.

For example, consider the following control system.

Use connect to build this system with an analysis point at the plant input, u.
C.InputName = 'e'; C.OutputName = 'u';
G.InputName = 'u'; G.OutputName = 'y';
Sum = sumblk('e = r-y');
T = connect(G,C,Sum,'r','y','u');

These commands return a generalized state-space (genss) model with one AnalysisPoint block.
You can use the analysis point, for example, to extract the open-loop response of the system at u:
L = getLoopTransfer(T,'u',-1);

For a more detailed example, see Mark Analysis Points in Closed-Loop Models. For more information
about using analysis points in dynamic system models, see the AnalysisPoint reference page.

LTI Viewer renamed to Linear System Analyzer

The LTI Viewer app is now called Linear System Analyzer. The functionality of the app is unchanged.

16-3
R2015a

You can access Linear System Analyzer in two ways:

• From the MATLAB desktop, in the Apps tab, in the Control System Design and Analysis
section, click Linear System Analyzer.
• From the MATLAB command line, use the linearSystemAnalyzer function. Previously, this
function was called ltiview. Using ltiview to open Linear System Analyzer does not generate
an error in this release, but the function might be removed in a future release.

Compatibility Considerations
If you have scripts or functions that use ltiview, consider replacing those calls with
linearSystemAnalyzer.

sisotool function renamed to controlSystemDesigner

The sisotool function is now called controlSystemDesigner. The controlSystemDesigner
opens the SISO Design Tool. You can also access SISO Design Tool from MATLAB desktop. In the
Apps tab, in the Control System Design and Analysis section, click Control System Designer.

Using sisotool to open SISO Design Tool does not generate an error in this release, but the
function might be removed in a future release.

Compatibility Considerations
If you have scripts or functions that use sisotool, consider replacing those calls with
controlSystemDesigner.

getBlockValue returns all block values in structure

A new syntax of the getBlockValue command now returns the current values of all Control Design
Blocks of a generalized model (genss) in a structure. The following syntax returns a structure, S,
whose field names are the names of the blocks in the genss model M. The values of the fields are
numerical LTI models or numerical values equal to the current values of the corresponding Control
Design Blocks.

S = getBlockValue(M)

This syntax lets you transfer the block values from one generalized model to another model that uses
the same Control Design Blocks, as follows:

S = getBlockValue(M1);
setBlockValue(M2,S);

For more information, see the getBlockValue reference page.

Compatibility Considerations
Previously, the syntax getBlockValue(M) returned the current values of the blocks of M as a vector
list:

[Val1,Val2,...,ValN] = getBlockValue(M)

16-4
 Now, using this syntax causes an error. You can still obtain block values in a list by specifying the
block names as input arguments, as follows.

[Val1,Val2,...,ValN] = getBlockValue(M,Blkname1,Blkname2,...,BlknameN)

Functionality being removed or changed

Functionality What Use This Instead Compatibility Considerations
Happens
When You
Use This
Functionality
?
[Val1,Val2,...] = Error S = getBlockValue(M) getBlockValue(M) now returns
getBlockValue(M) a structure containing current
values of all blocks. Update scripts
and functions that use
getBlockValue(M) to use output
structure.
ltiview function Still works linearSystemAnalyzer Consider replacing ltiview with
linearSystemAnalyzer in
scripts and functions.
sisotool function Still works controlSystemDesigner Consider replacing sisotool
with controlSystemDesigner

16-5
 17

R2014b

Version: 9.8

New Features

Bug Fixes

Compatibility Considerations
R2014b

LPV System block for modeling and simulating linear parameter-

varying systems
This release introduces the LPV System block. You use this block to represent Linear Parameter
Varying (LPV) systems in Simulink.

An LPV system is a linear state-space system whose dynamics vary as a function of certain time-
varying parameters called the scheduling parameters. Mathematically, an LPV system is represented
as:

dx t = A p x t + B p u t
y t =C p x t +D p u t
x 0 = x0

where

• u(t) are the inputs

• y(t) the outputs
• x(t) are the model states with initial value x0
• dx t is the state derivative vector ẋ for continuous-time systems and the state update vector
x t + ΔT for discrete-time systems. ΔT is the sample time.
• A(p), B(p), C(p) and D(p) are the state-space matrices parameterized by the scheduling
parameter vector p.
• The parameters p = p(t) are measurable functions of the inputs and the states of the model.
They can be a scalar quantity or a vector of several parameters. The set of scheduling parameters
define the scheduling space over which the LPV model is defined.

The linear system can be extended to contain offsets in the system’s states, input, and output signals.
Mathematically, the LPV system is represented by the following equations:

dx t = A p x t + B p u t + dx p − A p x p − B(p)u(p)
y t = C p x t + D p u t + y p − C p x p − D(p)u(p)
x 0 = x0

dx p ,     x p ,    u p ,    y p are the offsets in the values of dx(t), x(t), u(t) and y(t) at a given
parameter value p = p(t).

LPV system can be thought of as a first-order approximation of a nonlinear system over a grid of
scheduling parameter values. For example, you can linearize a Simulink model between a given input
and output ports over a grid of equilibrium operating conditions. The values of the model inputs,
outputs and state values at each operating point define the offsets, while the linear state-space
models obtained by linearization define the state-space data. The LPV system thus generated can
work as a proxy for the original model for facilitating faster simulations and control system design.
For more information, see Linear Parameter-Varying Models.

The LPV System block accepts the state-space matrices and offsets over a grid of scheduling
parameter values. The state-space matrices must be specified as an array of model objects. The
SamplingGrid property of the array defines the scheduling parameters for the LPV system. For
examples of using this block, see:

17-2
• Using LTI Arrays for Simulating Multi-Mode Dynamics
• Approximating Nonlinear Behavior using an Array of LTI Systems
• LPV Approximation of a Boost Converter Model

Kalman Filter block for estimating states of linear time-invariant and

linear time-varying systems
Use the Kalman Filter block to estimate the states of linear time-invariant and linear time-varying
systems online. The states are estimated as new data becomes available during the operation of the
system. The system can be continuous-time or discrete-time. You can generate code for this block
using code generation products such as Simulink Coder.

You can access this block from the Control System Toolbox library. For an example of using this block,
see State Estimation Using Time-Varying Kalman Filter.

AnalysisPoint Control Design Block for Marking Points of Interest for

Linear Analysis
The new AnalysisPoint block is a unit-gain Control Design Block that you can insert anywhere in a
control system model to mark points of interest for linear analysis and tuning. Incorporate
AnalysisPoint blocks into generalized state-space (genss) control system models by
interconnecting them with numeric LTI models and other Control Design Blocks. When you mark a
location in a control system model with an AnalysisPoint block, you can use that location for linear
analysis tasks, such as extracting responses using getIOTransfer or getLoopTransfer. You can
also use such locations to specify design requirements for control system tuning using systune or
Control System Tuner (requires Robust Control Toolbox software).

For more information about using AnalysisPoint blocks, see:

• AnalysisPoint reference page

• Control System with Multi-Channel Analysis Points
• Managing Signals in Control System Analysis and Design

Compatibility Considerations
AnalysisPoint replaces the loopswitch Control Design Block.

Models that contain loopswitch blocks continue to work, for backward compatibility. However, it is
recommended that you use AnalysisPoint blocks in new models. If you have scripts or functions
that use loopswitch blocks, consider updating them to use AnalysisPoint instead.

For documentation of loopswitch, see loopswitch in the R2014a documentation.

pidtool function renamed to pidTuner

The pidtool function is now called pidTuner. To open PID Tuner, use the pidTuner command or,
in the MATLAB desktop Apps tab, click PID Tuner.

Using pidtool does not generate an error in this release, but the function may be removed in a
future release.

17-3
R2014b

Compatibility Considerations
If you have scripts that use pidtool, consider replacing those calls with pidTuner.

getSwitches function renamed to getPoints

The getSwitches function is now called getPoints to match the renaming of loopswitch to
AnalysisPoint. Using getSwitches does not generate an error in this release, but the function
may be removed in a future release.

Compatibility Considerations
If you have scripts or functions that use getSwitches, consider replacing those calls with
getPoints.

Functionality being removed or changed

Functionality What Happens When Use This Instead Compatibility
You Use This Considerations
Functionality?
loopswitch Control Still works AnalysisPoint Consider replacing
Design Block loopswitch with
AnalysisPoint in
scripts and functions.
getSwitches function Returns loopswitch getPoints Consider replacing
and AnalysisPoint getSwitches with
blocks in model getPoints in scripts
and functions.
pidtool function Still works pidTuner Consider replacing
pidtool with
pidTuner in scripts.

17-4
 18

R2014a

Version: 9.7

New Features

Bug Fixes

Compatibility Considerations
R2014a

Redesigned PID Tuner app for improved PID tuning workflow

The redesigned PID Tuner streamlines workflows for interactively tuning PID controllers for
reference tracking and disturbance rejection.

To access the PID Tuner, use the pidtool command. For example, to tune a PI controller for an LTI
model, G:

pidtool(G,'PI')

For more information about the PID Tuner, see Designing PID Controllers with the PID Tuner.

PID controller tuning using system identification to model the plant

from measured input-output data in the PID Tuner app (with System
Identification Toolbox)
If you have System Identification Toolbox™ software, you can use PID Tuner to fit a linear model to
the measured SISO response data from your system and tune a PID controller for the resulting model.
For example, if you want to design a PID controller for a manufacturing process, you can start with
response data from a bump test on your system.

PID Tuner uses system identification to estimate an LTI model from the response data. You can
interactively adjust the identified parameters to obtain an LTI model with a response that fits your
response data. PID Tuner automatically tunes a PID controller for the estimated model. You can then
interactively adjust the performance of the tuned control system, and save the estimated plant and
tuned controller.

For an example, see Interactively Estimate Plant Parameters from Response Data.

freqsep function for decomposing a linear system into fast dynamics

and slow dynamics
Use the new freqsep command for separating numeric LTI models into fast and slow components.
freqsep allows you to specify the cutoff frequency about which the model is decomposed. The slow
component contains poles with natural frequency below the cutoff frequency. The fast component
contains poles at or above the cutoff.

For more information, see the freqsep reference page.

damp command display includes time constant information

When you call the damp command with no output arguments, the display now includes the time
constant for each pole. The time constant is calculated as follows:

1
τ= .
ωnζ

ωn is the natural frequency of the pole, and ζ is its damping ratio.

18-2
Compatibility Considerations
For a discrete-time system with unspecified sample time (Ts = -1), damp now calculates the natural
frequency and damping ratio by assuming Ts = 1. Previously, the software returned [] for the
natural frequency and damping ratio of such systems.

damp returns outputs in order of increasing natural frequency. Therefore, this change can result in
reordered poles for systems with unspecified sample times.

For more information on the outputs, see the damp reference page.

18-3
 19

R2013b

Version: 9.6

New Features

Bug Fixes

Compatibility Considerations
R2013b

SamplingGrid property for tracking dependence of array of sampled

models on variable values
In Control System Toolbox, you can derive arrays of numeric or generalized LTI models by sampling
one or more independent variables. The new SamplingGrid property of LTI models tracks the
variable values associated with each model in such an array.

Set this property to a structure whose fields are the names of the sampling variables and contain the
sampled variable values associated with each model. All sampling variables should be numeric and
scalar valued, and all arrays of sampled values should match the dimensions of the model array.

For example, suppose you create a 11-by-1 array of linear models, sysarr, by taking snapshots of a
linear time-varying system at times t = 0:10. The following code stores the time samples with the
linear models.

sys.SamplingGrid = struct('time',0:10)

For an additional examples, see:

• Array With Variations in Two Parameters

• Sample a Tunable (Parametric) Model for Parameter Studies

Option to retain unconnected states when interconnecting models

using connect command
By default, the connect command discards states that do not contribute to the dynamics in the path
between the inputs and outputs of the interconnected system. You can now optionally retain such
unconnected states. This option can be useful, for example, when you want to compute the
interconnected system response from known initial state values of the components.

To instruct connect to retain unconnected states, use the new connectOptions command with the
existing connect command.

For more information, see the connectOptions reference page.

connect command always returns state-space or frequency response

data model
The connect command now always returns a state-space model, such as an ss, genss, or uss
model, unless one or more of the input models is a frequency response data model. In that case,
connect returns a frequency response data model, such as an frd or genfrd model.

For more information, see the connect reference page.

Compatibility Considerations
In previous releases, connect returned a tf or zpk model when all input models were tf or zpk
models. Therefore, connect might now return state-space models in cases where it previously
returned tf or zpk models.

19-2
updateSystem command for updating dynamic system data in a
response plot
The new updateSystem command replaces the system data used to compute a response plot with
data derived from a different dynamic system, and updates the plot. updateSystem is useful, for
example, to cause a plot in a GUI to update in response to interactive input.

For more information, see:

• updateSystem reference page

• Build GUI With Interactive Plot Updates

getLoopID renamed to getSwitches

The getLoopID function is now called getSwitches to more clearly reflect the purpose of the
function. Using getLoopID does not generate an error in this release, but the function may be
removed in a future release.

Compatibility Considerations
If you have scripts or functions that use getLoopID, consider replacing those calls with
getSwitches.

LoopID property of loopswitch renamed to Location

The LoopID property of the loopswitch model component is now called Location to more clearly
reflect the purpose of the property. Using LoopID does not generate an error in this release, but the
name may be removed in a future release.

Compatibility Considerations
If you have scripts or functions that use the LoopID property, consider updating your code to use
Location instead.

19-3
 20

R2013a

Version: 9.5

New Features

Bug Fixes
R2013a

Transient behavior slider for PID Tuner, increasing control over

reference tracking and disturbance rejection performance
The PID Tuner now has a Transient behavior slider for emphasizing either reference tracking or
disturbance rejection. When you open the PID Tuner, the tool starts in the Time domain design
mode, displaying a step plot of the reference tracking response. The new Transient behavior slider
is beneath the Response time slider.

You can use the Transient behavior slider when:

20-2
• The tuned system’s disturbance rejection response is too sluggish for your requirements. In this
case, try moving the Transient behavior slider to the left to make the controller more aggressive
at disturbance rejection.
• The tuned system’s reference tracking response has too much overshoot for your requirements. In
this case, try moving the Transient behavior slider to the right to increase controller robustness
and reduce overshoot.

In Frequency domain design mode, the PID Tuner has Bandwidth and Phase margin sliders.
These sliders are the frequency-domain equivalents of the Response time and Transient behavior
sliders, respectively.

20-3
 21

R2012b

Version: 9.4

New Features

Bug Fixes

Compatibility Considerations
R2012b

ltiblock.pid2 and loopswitch objects for tuning two-degree-of-freedom

PID controllers and marking loop opening sites for open-loop
requirements
New Control Design Blocks allow you to specify more control structures and more types of
constraints for fixed-structure control system tuning in MATLAB:

• ltiblock.pid2 — Tunable two-degree-of-freedom PID controller

• loopswitch — Control Design Block for specifying feedback loop opening locations in a tunable
genss model of a control system

You can use these Control Design Blocks to build control systems for tuning with Robust Control
Toolbox tuning commands such as systune and looptune. For more information, see the
ltiblock.pid2 and loopswitch reference pages.

Commands for obtaining open-loop responses, closed-loop responses,

and current values of tunable components from control system
models
New commands allow you to compute open-loop and closed-loop responses from a Generalized LTI
model representing a control system.

• getLoopTransfer — Compute point-to-point open-loop response of a Generalized LTI model of a

control system, at a loop-opening site defined by a loopswitch block. The new command
getLoopID returns a list of such loop-opening sites.
• getIOTransfer — Extract the closed-loop response from a specified input to a specified output
of a control system.

These commands are particularly useful for validating the response functions of control systems
tuned using Robust Control Toolbox tuning commands such as systune.

Additionally, the new showTunable command displays the current value of tunable components in a
generalized LTI model of a control system. This command is useful for querying tuned parameter
values of control systems tuned using Robust Control Toolbox tuning commands such as systune.

For more information, see the reference pages for these new commands and the following topics:

• Generalized Models
• Models with Tunable Coefficients

Option for elementwise operation of model query commands on model

arrays
The new 'elem' flag causes elementwise operation on model arrays of the model query commands:

• hasInternalDelay
• hasdelay
• isstatic
• isreal

21-2
• isfinite
• isproper
• isstable

For example, for an array, sysarray, of dynamic system models,

B = hasdelay(sysarray,'elem');

returns a logical array. B of the same size as sysarray indicating whether the corresponding model
in sysarray contains a time delay. Without the 'elem' flag,

B = hasdelay(sysarray);

returns a scalar logical value that is equal to 1 if any entry in sysarray contains a time delay.

Compatibility Considerations
isfinite and isstable now return a scalar logical value when invoked without the 'elem' flag.
Previously, isfinite and isstable returned a logical array by default.

If you have scripts or functions that use isfinite(sysarray) or isstable(sysarray), replace

those calls with isfinite(sysarray,'elem') or isstable(sysarray,'elem') to perform an
elementwise query and obtain a logical array.

21-3
 22

R2012a

Version: 9.3

New Features

Compatibility Considerations
R2012a

Frequency Analysis Commands for Calculating Peak Gain and Finding

Gain-Crossover Frequencies
Control System Toolbox software includes two new frequency analysis commands:

• getPeakGain — Peak gain of frequency response of a dynamic system model

• getGainCrossover — Frequencies at which system gain crosses a specified gain level

For more information, see the getPeakGain and getGainCrossover reference pages.

These functions use the SLICOT library of numerical algorithms. For more information about the
SLICOT library, see http://slicot.org.

Specify Target Crossover Frequency as Input to pidtune

A new syntax for pidtune lets you specify a target crossover frequency directly as an input
argument. For example, the following command designs a PI controller, C, for a plant model sys. The
command also specifies a target value wc for the 0 dB gain crossover frequency of the open-loop
response L = sys*C.

C = pidtune(sys,'pi',wc);

Previously, you had to use pidtuneOptions to specify a target crossover frequency.

For more information, see the pidtune reference page.

Rescaled Impulse Response and Impulse-Invariant Time Domain

Conversion
For discrete-time dynamic system models, the input signal applied by impulse is now a unit area
pulse of length Ts and height 1/Ts. Ts is the sampling time of the discrete-time system. Previously,
impulse applied a pulse of length Ts and unit height.

Compatibility Considerations
Results of this change include:

• The amplitude of the impulse response calculated by impulse and impulseplot is scaled by
1/Ts relative to previous versions.
• Discretization using the impulse-invariant ('impulse') method of c2d returns a model that is
scaled by Ts compared to previous releases. This scaling ensures a close match between the
frequency responses of the continuous-time model and the impulse-invariant discretization as Ts
approaches zero (for strictly proper models). In previous releases, the frequency responses
differed by a factor of Ts.

First-Order Hold Method for d2c

The d2c command now supports the first-order hold (FOH) method for converting a discrete-time
dynamic system model to continuous time. The FOH method converts by performing linear
interpolation of the inputs, assuming the control inputs are piecewise linear over the sampling period.

22-2
 For more information about using this method, see the d2c reference page and Continuous-Discrete
Conversion Methods.

tzero Computes Invariant Zeros and Transmission Zeros

The tzero command computes the invariant zeros of SISO and MIMO dynamic system models. For
minimal realizations, tzero computes transmission zeros. tzero also returns the normal rank of the
transfer function of the system. For more information, see the tzero reference page.

Models Created With System Identification Toolbox Can Be Used

Directly With Control System Toolbox Functions
Identified linear models that you create using System Identification Toolbox software can now be
used directly with Control System Toolbox analysis and compensator design commands. In prior
releases, doing so required conversion to Control System Toolbox LTI model types.

Identified linear models include idfrd, idss, idproc, idtf, idgrey and idpoly models.

Identified linear models can be used directly with:

• Any Control System Toolbox or Robust Control Toolbox functions that operate on dynamic
systems, including:

• Response plots — nichols, margin, and rlocus

• Model simplification — pade, balred and minreal
• System interconnections — series, parallel, feedback and connect

For a complete list of these functions, enter:

methods('DynamicSystem')
• Analysis and design tools such as ltiview, sisotool and pidtool.
• The LTI System block in Simulink models.

Functionality Being Removed or Changed

Functionality What Happens Use This Instead Compatibility
When You Use This Considerations
Functionality?
impulse(sys) and Still works. N/A Amplitude of response is
impulseplot(sys), for scaled by 1/Ts compared
discrete-time sys to previous versions. Ts is
sampling time of sys.
c2d(sys,Ts,'impulse') Still works. N/A Resulting discretized
model is scaled by Ts
compared to previous
releases.

22-3
R2012a

Functionality What Happens Use This Instead Compatibility

When You Use This Considerations
Functionality?
[y,t] = For discrete-time N/A Expect the number of
impulse(sys,Tfinal) sys with undefined simulation data points to
[y,t] = step(sys,Tfinal) sample time (Ts=-1), be Tfinal + 1 instead of
[y,t,x] = Tfinal is Tfinal.
initial(sys,Tfinal) interpreted as the
number of sampling
periods to simulate.

22-4
 23

R2011b

Version: 9.2

New Features

Compatibility Considerations
R2011b

Formula-Based Specification of Summing Junctions and Vector Signal

Naming for sumblk and connect
You can now use formula strings to specify the behavior of summing junctions with sumblk. For
example, to create a summing junction, S, that takes the difference between signals r and y to
produce signal e, enter the following command:

S = sumblk('e = r-y');

Additionally, both sumblk and connect now support vector-based signal naming for interconnecting
multi-input, multi-output (MIMO) models. For more information, see the sumblk and connect
reference pages.

Commands for Interacting with Control Design Blocks in Generalized

LTI Models
The following new commands allow you to examine and set the values of Control Design Blocks in
Generalized LTI Models:

• getValue — Get nominal value of Generalized Model (replaces getNominal)

• setValue — Modify value of Control Design Block
• getBlockValue — Get nominal value of Control Design Block in Generalized Model
• setBlockValue — Set value of Control Design Block in Generalized Model
• showBlockValue — Display nominal values of Control Design Blocks in Generalized Model

For more information about these commands, see the reference pages for each command.

Functionality Being Removed or Changed

Functionality What Happens Use This Instead Compatibility
When You Use This Considerations
Functionality?
delay2z Errors absorbDelay Replace delay2z with
absorbDelay.
getNominal Errors getValue Replace getNominal with
getValue.
Scale and Info properties of Errors None None
realp parameter
sumblk('a','b','c','+-') Still works sumblk('a=b-c') Use new formula-based
syntax for sumblk.

23-2
 24

R2011a

Version: 9.1

New Features

Compatibility Considerations
R2011a

New Model Objects for Representing Tunable Parameters and Systems

with Tunable Components
Control System Toolbox includes new model objects that you can use to represent systems with
tunable components. You can use these models for parameter studies or controller synthesis using
hinfstruct. The new model types include:

• Control Design Blocks—Parametric components that are the building blocks for constructing
tunable models of control systems. Control Design Blocks include:

• realp—Tunable real parameter

• ltiblock.gain—Tunable static gain block
• ltiblock.tf—Fixed-order SISO transfer function with tunable coefficients
• ltiblock.ss—Fixed-order state-space model with tunable coefficients
• ltiblock.pid—One-degree-of-freedom PID controller with tunable coefficients
• Generalized Matrices—Matrices that include parametric (tunable) values. Generalized matrices
are genmat models.
• Generalized and Uncertain LTI Models—Models representing systems that have both fixed and
tunable coefficients. Generalized LTI models include:

• genss—Generalized state-space model

• genfrd—Generalized frequency response data model

These models arise from interconnections between numeric LTI models (such as tf , ss, or frd)
and Control Design Blocks. You can also create genss models by using the tf or ss commands
with one or more realp or genmat inputs.

This release also adds new functions for working with generalized models:

• getNominal—Nominal value of generalized model

• replaceBlock—Replace Control Design Blocks in generalized model
• nblocks—Number of blocks in generalized model
• isParametric — Determine if model has tunable blocks
• getLFTModel—Decompose generalized model

For more information about the new model types and about modeling systems that contain tunable
coefficients, see the following in the Control System Toolbox User's Guide:

• Types of Model Objects

• Models with Tunable Coefficients

New Time and Frequency Units for Models and Response Plots
All linear model objects now have a TimeUnit property for specifying unit of the time variable, time
delays in continuous-time models, and sampling time in discrete-time models. The default time units
is seconds. You can specify the time units, for example, as hours. See Specify Model Time Units for
examples.

Frequency-response data ( frd and genfrd) models also have a new FrequencyUnit property for
specifying units of the frequency vector. The default frequency units is rad/TimeUnit, where

24-2
TimeUnit is the system time units. You can specify the units, for example as kHz, independently of
the system time units. See Specify Frequency Units of Frequency-Response Data Model for examples.
If your code uses the Units property of frequency-response data models, it continues to work as
before.

See the model reference pages for available time and frequency units options.

Changing the TimeUnit and FrequencyUnit properties changes the overall system behavior. If you
want to simply change the time and frequency units without modifying system behavior, use
chgTimeUnit and chgFreqUnit, respectively.

The time and frequency units of the model appear on the response plots by default. For multiple
systems, the units of the first system are used. You can change the units of the time and frequency
axes:

• Graphically, using the following editors:

• Toolbox Preferences Editor

• LTI Viewer Preferences Editor
• Graphical Tuning Window Preferences Editor
• Property Editor of individual plots
• Programmatically, by setting the following properties of plots:

• TimeUnits for time-domain plots using timeoptions

• FreqUnits for frequency-domain plots using, for example, bodeoptions

Discrete-Time PID Controller Objects Have Stable Derivative Filter

Pole
New requirements for creating pid and pidstd controller objects ensure that the derivative filter
pole is always stable.

• For a discrete-time pid controller with a derivative filter (Tf≠ 0) and Dformula set to
'ForwardEuler', the sampling time Ts must be less than 2*Tf.
• For a discrete-time pidstd controller with a derivative filter (N≠ Inf) and Dformula set to
'ForwardEuler', the sampling time Ts must be less than 2*Td/N.
• The Trapezoidal value for DFormula is not available for a discrete-time pid or pidstd
controller with no derivative filter (Tf = 0 or N = Inf).

Compatibility Considerations
On loading pid or pidstd controllers saved under previous versions, the software changes certain
properties of controllers that do not have stable derivative filter poles.

• For a discrete-time pid controller with a derivative filter (Tf≠ 0), Dformula set to
'ForwardEuler', and sampling time Ts ≥ 2*Tf, the derivative filter time is reset to Tf = Ts.
• For a discrete-time pidstd controller with a derivative filter (N≠ Inf), Dformula set to
'ForwardEuler', the sampling time Ts ≥ 2*Td/N, the derivative filter constant is reset to N =
Td/Ts.

24-3
R2011a

• For a discrete-time pid or pidstd controller with no derivative filter and DFormula =
'Trapezoidal', the derivative filter integrator formula is reset to DFormula =
'ForwardEuler'.

The software issues a warning when it changes any of these values. If you receive such a warning,
validate your controller to ensure that the new values achieve the desired performance.

New Variable q^–1 for Expressing Discrete-Time Transfer Functions

You can now express discrete-time tf and zpk models in terms of the inverse shift operator q^-1.
The variable q^-1 is equivalent to z^-1.

Note This new definition is consistent with the System Identification Toolbox definition of q^-1.

Use the new variable by setting the Variable property of a tf or zpk model to q^-1. For example,
entering:

H = tf([1 2 3],[5 6 7],0.1,'Variable','q^-1')

creates the following discrete-time transfer function:

Transfer function:
1 + 2 q^-1 + 3 q^-2
-------------------
5 + 6 q^-1 + 7 q^-2

Sampling time (seconds): 0.1

When you set Variable to q^-1, tf interprets the numerator and denominator vectors as ascending
powers of q^-1.

For more information, see the tf and zpk reference pages.

24-4
 25

R2010b

Version: 9.0

New Features

Compatibility Considerations
R2010b

New Commands and GUI for Modeling and Tuning PID Controllers
This release introduces specialized tools for modeling and designing PID controllers.

PID Controller Design with the New PID Tuner GUI

The new PID Tuner GUI lets you interactively tune a PID controller for your required response
characteristics. Using the GUI, you can adjust and analyze your controller's performance with
response plots, such as reference tracking, load disturbance rejection, and controller effort, in both
time and frequency domains.

The PID Tuner supports all types of SISO plant models, including:

• Continuous- or discrete-time plant models

• Stable, unstable, or integrating plant models
• Plant models that include I/O time delays or internal time delay

For more information about using PID Tuner, see:

• Designing PID Controllers in the Control System Toolbox Getting Started Guide
• The new demo Designing PID for Disturbance Rejection with PID Tuner

PID Controller Design with the New pidtune Command

The new pidtune command lets you tune PID controller gains at the command line.

pidtune automatically tunes the PID gains to balance performance (response time) and robustness
(stability margins). You can specify your own response time and phase margin targets using the new
pidtuneOptions command.

pidtune supports all types of SISO plant models, including:

• Continuous- or discrete-time plant models.

• Stable, unstable, or integrating plant models.
• Plant models that include I/O time delays or internal time delays.
• Arrays of plant models. If sys is an array, pidtune designs a separate controller for each plant in
the array.

For additional information, see:

• The pidtune and pidtuneOptions reference pages

• The new Control System Toolbox demo Designing Cascade Control System with PI
Controllers

Modeling PID Controllers in Parallel Form or Standard Form

The new LTI model objects pid and pidstd are specialized for modeling PID controllers.

With pid and pidstd you can model a PID controller directly with the PID parameters, expressed in
parallel (pid) or standard (pidstd) form. The pid and pidstd commands can also convert to PID
form any type of LTI object that represents a PID controller.

25-2
Previously, to model a PID controller, you had to derive the controller's equivalent transfer function
(or other model), and could not directly store the PID parameters.

For additional information, see the pid and pidstd reference pages

Improved PID Tuning Options in SISO Design Tool

This release includes improvements to the PID Tuning options in the Automated Tuning pane of SISO
Design Tool.

In addition to the Robust Response Time tuning algorithm, SISO Design Tool offers a collection of
classical design formulas, including the following:

• Approximate M-Constrained Integral Gain Optimization (MIGO) Frequency Response

• Approximate MIGO Step Response
• Chien-Hrones-Reswick
• Skogestad Internal Model Control (IMC)
• Ziegler-Nichols Frequency Response
• Ziegler-Nichols Step Response

For information about using SISO Design Tool, see SISO Design Tool in the Control System Toolbox
User's Guide. For specific information about the automatic PID Tuning options in SISO Design Tool,
see PID Tuning in the Control System Toolbox User's Guide.

Ability to Analyze a Controller Design for Multiple Models

Simultaneously in SISO Design Tool
You can now analyze a controller design for multiple models simultaneously using the SISO Design
Tool. This feature helps you analyze whether the controller satisfies design requirements on a system
whose exact dynamics are not known and may vary.

System dynamics can vary because of parameter variations or different operating conditions. You
represent variations in system dynamics of the plant (G), sensor (H), or both in a feedback structure
using arrays of LTI models. Then, design a controller for a nominal model in the array and analyze
that the controller satisfies the design requirements on the remaining models using the design and
analysis plots. For more information, see:

• Control Design Analysis of Multiple Models.

• Compensator Design for a Set of Plant Models demo.
• Reference Tracking of a DC Motor with Parameter Variations demo in Simulink
Control Design software.

Change in Output of repsys Command

The output of the repsys command when called with a single dimension argument has changed.

In prior versions, the output of repsys(sys,N) was the same as that of append(sys,...,sys).

Now, repsys(sys,N) returns the same result as repsys(sys,[N N]).

25-3
R2010b

The results of other syntaxes for repsys have not changed.

See the repsys and append reference pages for more information.

Compatibility Considerations
Code that depends upon the previous result of repsys(sys,N) no longer returns that result. To
obtain the previous result, replace repsys(sys,N) with sys*eye(N).

25-4
 26

R2010a

Version: 8.5

New Features

Compatibility Considerations
R2010a

Enhanced c2d Command to Approximate Fractional Time Delays in

Tustin and Matched Discretization Methods
The c2d command can now approximate fractional time delays when discretizing linear models with
the tustin or matched methods. The new c2dOptions command lets you specify an optional
Thiran all-pass filter. The Thiran filter approximates fractional delays for improved phase matching
between continuous and discretized models. Previously, c2d rounded fractional time delays to the
nearest multiple of the sampling time when using the tustin or matched methods. For more
information, see the c2d and c2dOptions reference pages and Continuous-Discrete Conversion
Methods in the Control System Toolbox User Guide.

New Commands for Specifying Options for Continuous-Discrete

Conversions
New commands c2dOptions, d2dOptions, and d2cOptions make it easier to specify options for

• Discretization using c2d

• Resampling using d2d.
• Conversion from discrete to continuous time using d2c.

Compatibility Considerations
This release deprecates the prewarp method for c2d, d2d, and d2c. Instead, use c2dOptions,
d2dOptions, or d2cOptions to specify the tustin method and a prewarp frequency. For more
information, see Continuous-Discrete Conversion Methods and the c2d, d2d, and d2c reference
pages.

New FDEL Command to Remove Specified Data from Frequency

Response Data (FRD) Models
You can now remove selected data from frd models using the new fdel command. For example, use
fdel to:

• Remove spurious or unneeded data from frd models you create from measured frequency
response data.
• Remove data at intersecting frequencies from frd models before merging them into a single frd
model with fcat, which can only merge frd models containing no common frequencies.

For more information, see fdel reference page.

26-2
 27

R2009b

Version: 8.4

New Features
R2009b

Ability to Design Compensators for New Types of Plants

In the SISO Design Tool, you can now design compensators for plants models that:

• Contain time delays

Previously, you had to approximate delays before designing compensators.

• You specify as frequency-response data (FRD)

For more information on designing compensators using the SISO Design Tool, see SISO Design Tool.

New Automated PID Tuning Method

You can now tune compensators using a new automated PID tuning algorithm called Robust Response
Time, which is available in the SISO Design Tool. You specify the open-loop bandwidth and phase
margin, and the software computes PID parameters to robustly stabilize your system.

For information on tuning compensators using automated tuning methods, see Automated Tuning.

27-2
 28

R2009a

Version: 8.3

New Features

Compatibility Considerations
R2009a

Variable q Now Defined as the Forward Shift Operator z

The variable q is now defined in the standard way as the forward shift operator z. Previously, q was
defined as z-1.

Note This new definition is consistent with the System Identification Toolbox definition of q.

Compatibility Considerations
If you use the q variable, you may receive different results than in previous releases when you:

• Create a transfer function

• Modify the num or den properties of an existing transfer function

The resulting transfer function differs from previous releases when both the

• Variable property is set to q

• num and den properties have different lengths

For example, the following code:

H = tf([1,2],[1 3 8],0.1,'Variable','q')

now returns the transfer function

q+2 z+2
≡
q2 + 3q + 8 2
z + 3z + 8

Previously, the code returned the transfer function

1 + 2q 1 + 2z−1 z2 + 2z
2
≡ −1 −2
≡ 2
1 + 3q + 8q 1 + 3z + 8z z + 3z + 8

The two transfer functions have different numerators.

28-2
 29

R2008b

Version: 8.2

New Features

Compatibility Considerations
R2008b

New Design Tools for Linear-Quadratic-Gaussian (LQG) Servo

Controllers with Integral Action
You can now design a Linear-Quadratic-Gaussian (LQG) servo controller for set-point tracking using
the new lqi and lqgtrack commands. This compensator ensures that the system output tracks the
reference command and rejects process disturbances and measurement noise.

For more information on forming LQG servo controllers, see Linear-Quadratic-Gaussian (LQG)
Design, the lqi reference page, and the lqgtrack reference page.

Current Flag Moved from lqgreg to kalman

The 'current' flag was moved from the lqgreg function to the kalman function.

Compatibility Considerations
The following code:

kest = kalman(sys,Qn,Rn)
c = lqgreg(kest,k)

now returns the current regulator u n = − Kx n n instead of the delayed regulator

u n = − Kx n n − 1 .

To update your code to return the same results as in previous releases, use the following code with
the added string 'delayed' in the kalman command:

kest = kalman(sys,Qn,Rn,'delayed')
c = lqgreg(kest,k)

For information on using these functions with the current flag in the kalman function, see the
kalman and lqgreg reference pages.

New Upsampling Method for Rate Conversion in Discrete-Time Models

You can now upsample a discrete-time system to an integer multiple of the original sampling rate
without any distortion in the time or frequency domain using the upsample command.

For more information on upsampling, see the upsample reference page and Upsample a Discrete-
Time System in the Control System Toolbox User's Guide.

New Scaling Tools to Enhance the Accuracy of Computations with

State-Space Models
You can now scale state-space models to maximize accuracy over the frequency band of interest using
the prescale command and associated GUI. Use this functionality when you cannot achieve good
accuracy at all frequencies and some tradeoff is necessary. A warning alerts you when accuracy may
be poor and using prescaling is recommended.

For more information on setting the frequency band for scaling state-space realizations, see Scaling
State-Space Models and the prescale reference page.

29-2
New Command to Reorder the States of State-Space Models
You can now reorder the states of state-space models according to a specified permutation using the
xperm command.

For more information on reordering states, see the xperm reference page.

Enhanced Support for Customizing Response Plots

You can now make the following changes to your Control System Toolbox response plots using the
figure plotting tools:

• System name
• Line color
• Line style
• Line width
• Marker type

For more information on customizing the appearance of response plots using plot tools, see
Customizing Response Plots Using Plot Tools in the Control System Toolbox User's Guide.

29-3
 30

R2008a

Version: 8.1

New Features
R2008a

Updated Error and Warning Message System

The Control System Toolbox error and warning IDs and messages have been updated. If you use error
and warning IDs in your code, you must update your code to reflect the new IDs.

30-2
 31

R2007b

Version: 8.0.1

New Features
R2007b

Updated and Expanded Demos

The Control System Toolbox demos have been reformatted and expanded to include more examples
and content. Demos in the following categories now have new and improved content:

• Getting Started with LTI Models

• Discretization and Sampling Rate Conversions
• How to Get Accurate Results

To open the Control System Toolbox demos, type

demo toolbox control

at the MATLAB prompt.

31-2
 32

R2007a

Version: 8.0

New Features
R2007a

Analysis of Time Delay Systems Now Fully Supported

Control System Toolbox software now lets you:

• Model, simulate, and analyze any interconnection of linear systems with delays, such as systems
containing feedback loops with delays.
• Exactly analyze and simulate control systems with long delays. You can evaluate control
strategies, such as Smith Predictor and PID control for first-order-plus-dead-time plants.
• Use new commands for modeling state-space models with delays including: delayss,
getDelayModel, and setDelayModel.

For more information, see the section on Models with Time Delays.

New and Updated Automated Tuning Methods

Control System Toolbox software now provides the following new and updated automated tuning
methods:

• New Singular Frequency Based Tuning lets you design PID compensators for both stable and
unstable plants.
• New H-infinity Loop Shaping lets you find compensators based on a desired open-loop bandwidth
or loop shape. This feature requires Robust Control Toolbox software.
• Updated Internal Model Control (IMC) Tuning now supports unstable plants.

For more information, see the section on automated tuning in the Control System Toolbox
documentation.

New Tustin and Prewarp Options for d2d Function

The d2d function now includes the following new options for the resampling method:

• 'tustin'—Performs Bilinear (Tustin) approximation

• 'prewarp'—Performs Tustin approximation with frequency prewarping

For more information, see the d2d reference pages.

32-2
 33

R2006b

Version: 7.1

New Features
R2006b

New Loop Configurations in the SISO Design Tool

Two new loop configurations are available from the SISO Design Tool. See Modifying Block Diagram
Structure for more information.

New Design Requirements

The LTI Viewer now supports step response and upper/lower time bound design requirements. See
Adding Design Requirements to the LTI Viewer for more information.

33-2
 34

R2006a

Version: 7.0

New Features
R2006a

SISO Design Tool

The SISO Design Tool now provides one-click automated tuning using systematic algorithms such as
Ziegler-Nichols PID tuning, IMC design, and LQG design. In addition, you can calculate low-order
approximations of the IMC/LQG compensators to keep the control system complexity low.

Compensator Optimization Is Now Supported

If you have installed Simulink Response Optimization™ software, you can now optimize the
compensator parameters inside the SISO Design Tool GUI. You can specify time- and frequency-
domain requirements on SISO Design Tool plots such as bode and step, and use numerical
optimization algorithms to automatically tune your compensator to meet your requirements. See the
Simulink Response Optimization documentation for more details.

Improved Compensator Editor

The Compensator Editor used to edit the numerical values of poles and zeros has been upgraded to
better handle common control components such as lead/lag and notch filters.

Multi-Loop Compensator Design Support

Many control systems involve multiple feedback loops, some of which are coupled and need joint
tuning. The SISO Design Tool now lets you analyze and tune multi-loop configurations. You can focus
on a specific loop by opening signals to remove the effects of other loops, gain insight into loop
interactions, and jointly tune several SISO loops.

SISO Design Tool Fully Integrated with the Controls & Estimation Tools Manager

To improve workflow and better leverage other tools, such as Simulink Control Design software and
Simulink Response Optimization software, the SISO Design Tool is now fully integrated with the
Controls & Estimation Tools Manager (CETM). This provides a signal environment for the design and
tuning of compensators.

When you open the SISO Design Tool, the CETM also opens with a SISO Design Task. Many SISO
Design Tool features, such as importing models, changing loop configurations, etc., have been moved
to the SISO Design Task in CETM. In addition, related tasks such as Simulink based Tuning and
Compensator Optimization are seamlessly integrated with the SISO Design Task. See the Control
System Toolbox Getting Started Guide for details on the new work flow.

LTI Viewer Enhancements

The LTI Viewer now lets you plot the response of a system to user-defined input signals (lsim) and
initial conditions (initial). A new GUI lets you select input signals from a signal generator library,
or import signal data from a variety of file formats.

LTI Objects
Descriptor and Improper State-Space Models Fully Supported

There is now full support for descriptor state-space models with a singular E matrix. This now lets
you build state-space representations, such as PID, and manipulate improper models with the
superior accuracy of state-space computations. In previous versions, only descriptor models with a
nonsingular E matrix were supported.

34-2
New Commands to Calculate Time Response Metrics

The new stepinfo and lsiminfo commands compute time-domain performance metrics, such as
rise time, settling time, and overshoot. You can use these commands to write scripts that
automatically verify or optimize such performance requirements. Previously, these metrics were
available only from response plots.

Simplified System Interconnections Using I/O Channel Names

The commands connect, feedback, series, parallel, and lft now let you connect systems by
matching names of I/O channels. A helper function, sumblk, has also been added to simplify the
specification of summing junctions. Altogether this considerably simplifies the task of deriving models
for complicated block diagrams. In previous releases, only index-based system connection was
supported.

Changes in the Representation of I/O Delays in State-Space Models

The ioDelay property is deprecated from state-space models. Instead, these models have a new
property called InternalDelay for logging all delays that cannot be pushed to the inputs or
outputs. Driving this change is the switch to a representation of delays in terms of delayed
differential equations rather than frequency response. See Models with Time Delays for more details
on internal delays, and ss/getdelaymodel for details on the new internal representation of state-
space models with delays.

New Name Property for LTI Objects

This new property lets you attach a name (string) to a given LTI model. The specified name is
reflected in response plots.

New Commands and Operations for LTI Objects

The new exp command simplifies the creations of continuous-time transfer functions with delays. For
more information, type help lti/exp at the MATLAB prompt.

The frd object has the following new methods:

• fcat — Concatenates one or more FRD models along the frequency dimension (data merge).
• fselect — Selects frequency points or range in frd model.
• fnorm — Calculates pointwise peak gain of frd model.

The .* operation is supported for transfer functions and zero-pole-gain objects. This allows you to
perform element-by-element multiplication of MIMO models.

Numerical Algorithms
There have been several major improvements in the Control System Toolbox numerical algorithms,
many of which benefit the upgraded SISO Design Tool:

• New scaling algorithm that maximizes accuracy for badly scaled state-space models
• Performance improvement in time and frequency response computations through MEX-files
• More accurate computations of the zero-pole-gain and transfer function representations of a state-
space model

34-3
R2006a

• More accurate state-space representations of zero-pole-gain models

• Better handling of nonminimal modes in model reduction commands (balred, balreal)
• canon now computes a block modal form for A matrices that are not diagonizable or are nearly
defective
• Exact phase computation for zero-pole-gain models in bode and nichols
• Accurate handling of improper models using the descriptor state-space representation

34-4
 35

R14SP3

Version: 6.2.1

No New Features or Changes

 36

R14SP2

Version: 6.2

New Features
R14SP2

Command-Line API for Customizing Plots

The Control System Toolbox software now provides a command-line API for customizing units, labels,
limits, and other plot options. You can now change default plot options before generating a plot, or
modify plot properties after creation.

For a detailed description of the commands, see the Control System Toolbox documentation.

Constraint Types for SISO Design

You can now create

• Single piecewise linear constraints for root-locus and Bode plots

• Gain/phase exclusion regions for Nichols plots

Design constraints are displayed as shaded regions.

Bode and Nichols Plots Have Additional Options

When editing Bode and Nichols plots, you can now

• Set the lower limit of the magnitude manually.

• Adjust the phase offsets by multiples of 360 degrees to facilitate comparing multiple responses.

Model-Approximation and Order-Reduction Commands

New commands have been added for model approximation and order reduction:

• hsvd computes and plots the Hankel singular values.

• balred computes low-order approximations using a numerically stable, balancing-free algorithm.
You can perform multiple order reductions with a single command.

36-2