A Hands-On Course in

Sensors Using the Arduino

and Raspberry Pi
Series in Sensors
Series Editors: Barry Jones and Haiying Huang

Other recent books in the series:

Resistive, Capacitive, Inductive, and Magnetic Sensor Technologies
Winncy Y. Du
Semiconductor X-Ray Detectors
B. G. Lowe and R. A. Sareen
Portable Biosensing of Food Toxicants and Environmental Pollutants
Edited by Dimitrios P. Nikolelis, Theodoros Varzakas, Arzum Erdem,
and Georgia-Paraskevi Nikoleli
Optochemical Nanosensors
Edited by Andrea Cusano, Francisco J. Arregui, Michele Giordano,
and Antonello Cutolo
Electrical Impedance: Principles, Measurement, and Applications
Luca Callegaro
Biosensors and Molecular Technologies for Cancer Diagnostics
Keith E. Herold and Avraham Rasooly
Compound Semiconductor Radiation Detectors
Alan Owens
Metal Oxide Nanostructures as Gas Sensing Devices
G. Eranna
Nanosensors: Physical, Chemical, and Biological
Vinod Kumar Khanna
Handbook of Magnetic Measurements
S. Tumanski
Structural Sensing, Health Monitoring, and Performance Evaluation
D. Huston
Chromatic Monitoring of Complex Conditions
Edited by G. R. Jones, A. G. Deakin, and J. W. Spencer
Principles of Electrical Measurement
S. Tumanski
Novel Sensors and Sensing
Roger G. Jackson
Hall Effect Devices, Second Edition
R. S. Popovic
 A Hands-On Course in
Sensors Using the Arduino
and Raspberry Pi

Volker Ziemann
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742

© 2018 by Taylor & Francis Group, LLC

CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Printed on acid-free paper

Version Date: 20180129

International Standard Book Number-13: 978-0-8153-9360-3 (Paperback)

International Standard Book Number-13: 978-0-8153-9359-7 (Hardback)

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

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

For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.

Library of Congress Cataloging-in-Publication Data

Names: Ziemann, Volker (Associate professor of physics), author.

Title: A hands-on course in sensors using the Arduino and Raspberry Pi /
Volker Ziemann.
Other titles: Sensors series.
Description: Boca Raton, FL : CRC Press, Taylor & Francis Group, [2018] |
Series: Series in sensors
Identifiers: LCCN 2017045669| ISBN 9780815393597 (hardback ; alk. paper) |
ISBN 9780815393603 (pbk. ; alk. paper) | ISBN 9781351188319 (e-book) |
ISBN 9781351188296 (e-book) | ISBN 9781351188302 (ebook) | ISBN
9781351188289 (ebook)
Subjects: LCSH: Detectors. | Raspberry Pi (Computer) | Arduino (Programmable
controller) | Microcontrollers.
Classification: LCC TK7872.D48 Z54 2018 | DDC 681/.20285464--dc23
LC record available at https://lccn.loc.gov/2017045669

Visit the eResource: https://www.crcpress.com/9780815393603

Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Contents

Preface xv

Acknowledgments xvii

Chapter 1  Introduction 1

Chapter 2  Sensors 5
2.1 ANALOG SENSORS 5
2.1.1 Resistance-based sensors 5
2.1.2 Voltage-based sensors 10
2.1.3 Current-based sensors 14
2.2 SIGNAL CONDITIONING 16
2.2.1 Voltage divider 16
2.2.2 Amplifiers 17
2.2.3 Filters 21
2.2.4 Analog-to-digital conversion 23
2.2.5 Supply voltage 26
2.3 DIGITAL SENSORS 28
2.3.1 Buttons and switches 29
2.3.2 On/off devices 30
2.3.3 I2C devices 32
2.3.4 SPI devices 35
2.3.5 RS-232 devices 36
2.3.6 Other sensors 36

Chapter 3  Actuators 39
3.1 SWITCHES 39
3.1.1 Light-emitting diodes and optocouplers 39
3.1.2 Large currents 41
3.2 MOTORS 43
3.2.1 DC motors 44
3.2.2 Servomotors and model-servos 46
3.2.3 Stepper motors 47

v
vi  Contents

3.3 ANALOG VOLTAGES 51

3.4 OTHER ACTUATORS 52

Chapter 4  Microcontroller: Arduino 55

4.1 HARDWARE 55
4.1.1 Arduino UNO 55
4.1.2 ESP8266 and NodeMCU 56
4.2 GETTING STARTED 57
4.3 HELLO WORLD, BLINK 58
4.4 INTERFACING SENSORS 60
4.4.1 Button 60
4.4.2 Analog input 61
4.4.3 I2C 65
4.4.4 SPI 77
4.4.5 Other protocols 80
4.5 INTERFACING ACTUATORS 85
4.5.1 Switching devices 85
4.5.2 DC motors 87
4.5.3 Servos 91
4.5.4 Stepper motors 92
4.5.5 Analog voltages 101
4.5.6 Human attention actuators 102
4.6 COMMUNICATION TO HOST 103
4.6.1 RS-232 and USB 103
4.6.2 Bluetooth 104
4.6.3 WiFi 105
4.6.4 Other communication 110

Chapter 5  Host Computer: Raspberry Pi 113

5.1 HARDWARE 113
5.2 GETTING STARTED 113
5.3 INSTALLING AND USING NEW SOFTWARE 117
5.4 RASPI AS A ROUTER 121
5.5 COMMUNICATION WITH THE ARDUINO 123
5.5.1 Arduino IDE 123
5.5.2 From the command line 124
5.5.3 Python 125
5.5.4 Octave 129
5.6 DATA STORAGE 132
5.6.1 Flatfile 132
5.6.2 MySQL 134
 Contents  vii

5.6.3 RRDtool 138

5.7 ONLINE PRESENTATION 141

Chapter 6  Control System: EPICS 147

6.1 INSTALLATION 147
6.2 COMMUNICATING WITH EPICS 149
6.3 ASYN AND STREAM LIBRARIES 150
6.4 WRITING AN IOC 151
6.5 STARTING THE IOC AT BOOT TIME 154

Chapter 7  Messaging System: MQTT 157

7.1 BROKER 158
7.2 NODEMCU CLIENTS 159
7.3 GATEWAY TO EPICS 161

Chapter 8  Example: Weather Station with Distributed Sensors 167

Chapter 9  Example: Geophones 177

Chapter 10  Example: Monitor for the Color of Water 185

Chapter 11  Example: Capacitance Measurement 191

Chapter 12  Example: Profile of a Laser Beam 197

Chapter 13  Example: Fire-Seeking Robot 205

Chapter 14  Presenting and Writing 223

14.1 PREPARING A PRESENTATION 223
14.2 PREPARING A REPORT 225
14.3 PRESENTING DATA 226
14.4 GOOD ENGLISH 227
14.5 POSTSCRIPTUM 228

Appendix A  Basic Circuit Theory 229

Appendix B  Least-Squares Fit and Error Propagation 231

Bibliography 235

Index 237
List of Figures

1.1 The outline of the book. 2

2.1 Image of a light-dependent resistor (LDR) on the left and how to connect it
in a plain voltage-divider configuration (center) and in a Wheatstone bridge
(right). 5
2.2 The band-level scheme and the schematics for the circuit with the LDR in a
voltage divider. The upper graph illustrates dark conditions and the lower
graph shows conditions where the LDR is exposed to light. Note that the
vertical axis by convention shows the energy of electrons. This causes the
positive pole of the battery to have the most negative energy. See the text
for a discussion. 6
2.3 NTC resistors are doped semiconductors that have donor levels just below
the conduction band. Increasing the temperature increases the kinetic energy
of the electrons that allows them to occasionally jump into the conduction
band, where they contribute to the conductivity of the material. 8
2.4 A linear potentiometer (left) and a circuit illustrating the electric connec-
tions (right). 9
2.5 A joystick is shown on the left and a fluid-level resistive sensor on the right. 10
2.6 A strain gauge. 10
2.7 Schematic of an MQ-x gas detector (left) and a sensor mounted on a small
breakout board (right). 11
2.8 Image of an LM35 temperature sensor (left) and how to connect it (right). 11
2.9 On the left we show a schematic of a thermocouple on the top and a ther-
mopile on the bottom. On the right we show an image of an MLX90614
contact-free thermometer. 12
2.10 Schematic of a Hall sensor (left) and the A1324 sensor (right). 12
2.11 The operational principle of an ADXL accelerometer. 13
2.12 An SM-24 geophone with a diameter of about 30 mm. 14
2.13 Energy-band diagram (left) and circuit (right) of a reverse-biased pin diode
in photoconductive mode. 14
2.14 Image of a BPW34 pin diode on the left and two phototransistors on the
right; an IR-sensitive BPX38 and an SFH3310, sensitive in the visible part
of the spectrum. 15
2.15 On the left we have a voltage divider to reduce the input voltage of 0-5 V
to 0-3.3 V. The right circuit shows the use of clamping diodes to protect the
input of the microcontroller to lie between ground and 5 V. 16

ix
x  LIST OF FIGURES

2.16 A bare operational amplifier shown on the left and shown wired as a line
buffer on the right. We omitted the supply wires on the latter schematic. 17
2.17 A non-inverting (left) and inverting (right) amplifier. 18
2.18 A difference amplifier (left) and the same circuit with adjustable V1 (right)
that is used to subtract the baseline. 19
2.19 Instrumentation amplifier (left) and logarithmic amplifier (right). 19
2.20 Amplifying a weak signal and shifting it to mid-range between the power
rails. 20
2.21 A simple low-pass (left) and high-pass (right) filter. 21
2.22 A simple band-pass (left) and band-stop (right) filter. 22
2.23 An active non-inverting (left) and inverting (right) low-pass filter. 22
2.24 The operating principle of a 3-bit flash ADC. 23
2.25 The operating principle of a successive approximation ADC. 24
2.26 The operating principle of a delta-sigma ADC. 24
2.27 Sampling a signal in different Nyquist zones. The dashed line displays a
frequency of 0.15fs and the solid lines show signals with frequency (1 −
0.15)fs on the left and (1+0.15)fs on the right. Note that at the times when
the signal is sampled (indicated by boxes), the signals are indistinguishable. 25
2.28 The dashed signals s1 and s2 are the images aliased into the base band of
the original signals S1 and S2 in higher Nyquist zones. 26
2.29 Schematics of very simple power supply circuits. 27
2.30 Variable-voltage (left) and fixed-voltage (right) regulator circuits. 28
2.31 Connecting a switch or button with a pull-up resistor. 29
2.32 Level shifter circuitry using a n-type MOSFET. The source of the MOSFET
is connected to the 3.3 V logic and the drain to the 5 V logic. 30
2.33 Schematic view of a PIR sensor (left) and the hardware (right). 31
2.34 HR-SR04 distance sensor. 31
2.35 Illustration of the operational principle of a barometric pressure sensor. 32
2.36 A BMP180 barometric pressure sensor and an HYT-221 humidity sensor. 33
2.37 The operational principle of one gyroscope in the MPU-6050. 34
2.38 An MPU-6050 accelerometer on a breadboard. 35
2.39 A GPS receiver on the left and a DHT11 humidity sensor on a breadboard
on the right. 36
2.40 A Shinyei PPD42NS particle sensor (left) and a GP2Y1010AU0F dust sensor
(right). 37

3.1 A close-up of a light-emitting diode is shown on the left. In the center is the
schematic of connecting an LED and the same circuit on a breadboard. 40
3.2 The physics of an LED. 41
3.3 The terminals of an NPN transistor (left), and using an NPN transistor as
switch (right). 42
3.4 Two NPN transistors connected to form a Darlington pair (left) and a
ULN2003 Darlington array (right). 42
 LIST OF FIGURES  xi

3.5 Schematic illustrating the functionality of a relay (left) and an image of a

relay (right). 43
3.6 Basic operation principle of a DC motor. The magnetic north pole of the
stator is to the left and the south pole to the right of the coil. Current is
supplied to the brushes where it enters the commutator and passes through
the rotor-coil, where the Lorentz force causes a force on the wire, moving
it upwards. After half a turn, the commutator has rotated and reverses the
polarity such that the current on the left side of the coil points in the same
direction as before. 44
3.7 Schematic illustrating the functionality of an H bridge. 45
3.8 A small model-servo. 46
3.9 The timing of the control signal for the servo. 47
3.10 Schematics of a stepper motor. Note that the coils are denoted by upper-case
letters and the corresponding terminals with lower-case letters. 48
3.11 Wiring of the coils in a bipolar stepper motor (left) where the respective
center terminals are often connected. In a unipolar motor (right) the center
terminals are exposed. 50
3.12 Connection for a unipolar stepper motor. 51
3.13 Operational principle of a digital-to-analog converter. 51
3.14 Sketch of a butterfly valve. 53

4.1 An Arduino UNO. 56

4.2 The ESP-01 on the left and a NodeMCU on the right. 56
4.3 The Arduino IDE. 57
4.4 The “Blink” sketch. 59
4.5 Interfacing a button to the Arduino. 60
4.6 Interfacing a potentiometer (left) or an LM35 temperature sensor (right) to
the Arduino. 62
4.7 Connecting the MLX90614 contact-free thermometer to an Arduino UNO. 66
4.8 Connecting the HYT-221 humidity sensor to the Arduino (left) and using
the serial plotter to display the humidity and temperature measurements
(right). 67
4.9 Connecting the MCP23017 IO-extender to the UNO. 74
4.10 Connecting the MCP3304 ADC to the NodeMCU. 77
4.11 Connecting the DS18b20 sensor to the Arduino UNO. 82
4.12 Oscilloscope trace of a pulse-width modulated signal on pin D9 of an Arduino
UNO with values of 88 (left) and 220 (right). 85
4.13 Using pulse-width modulation and a transistor to adjust the brightness of
an LED. 87
4.14 Controlling the speed of a motor with pulse width modulating the base of
a TIP-120 Darlington transistor. See the text for a discussion of the diodes. 88
4.15 The pin assignment of the L293D H-bridge driver. 88
4.16 Controlling speed and direction of a DC motor. 89
4.17 Connecting a model-servo to the Arduino UNO. 91
xii  LIST OF FIGURES

4.18 Connecting a unipolar stepper motor to the UNO with a ULN2003 Darling-
ton driver. 93
4.19 Connecting a bipolar stepper motor to the UNO with an L293D H-bridge
driver. 94
4.20 Connecting a bipolar stepper motor to the UNO with a DRV8825 stepper
motor driver with microstep capability. 98
4.21 Adjusting the maximum current on the DRV8825 breadboard. 99
4.22 Connecting an MCP4921 12-bit DAC to the Arduino UNO. 101
4.23 Connecting a speaker and a piezo buzzer to the Arduino. 103
4.24 Front and back side of the HC-06 Bluetooth dongle and the connection to
the Arduino. 104
4.25 The NodeMCU with an LM35 temperature sensor. 106
4.26 The communication with a web server showing the HTTP header. 108

5.1 A Raspberry Pi board. 114

5.2 The Accessory menu behind the button with the raspberry shows a number
of installed programs such as the Terminal (command) window. 115
5.3 The nano text-editor with available commands listed on the bottom. 118
5.4 The Synaptic Package Manager. 119
5.5 Telnet session connected wirelessly to the NodeMCU. 124
5.6 Using Python on the Raspi to communicate with the query response sketch
running on the Arduino UNO. 125
5.7 Simple ASCII graphics from querying the UNO repeatedly. 126
5.8 Plot from the temperature logger. 131
5.9 Graphics from rrdtool. 140
5.10 Web page from the Raspi after installation of the Apache2 web server. 141
5.11 The first self-made web page. 142
5.12 User pi’s home page that displays the continuously (but slowly) updated
temperature. 144

6.1 Using the EPICS commands caget and caput. 150

8.1 The weather-node circuit with the NodeMCU on the right and the LM35,
HYT221, and BMP180 sensors towards the left. 168
8.2 The weather station web page served by the Raspi. 173

9.1 The amplifier to interface an SM-24 geophone to the NodeMCU. 178

9.2 The prototype with the SM-24 geophone connected to the amplifier board
and the NodeMCU. 180
9.3 The raw time series from the geophone and the corresponding spectrum. 181

10.1 The setup to measure the color-dependent absorption (left) and the reflec-
tion from a surface (right). 185
10.2 The setup to measure the water color with an Arduino UNO. 186
 LIST OF FIGURES  xiii

10.3 The color sensor with the RGB-LED on the left and the phototransistor on
the right. 188

11.1 The setup to measure the capacitance. 192

11.2 The waveform of the voltage on the capacitor as a function of time, while a
2.2 µF Tantal capacitor discharges through a 33 kΩ resistor on a linear scale
(above) and on a logarithmic scale (below). 195

12.1 The schematic setup to measure the beam size of a laser pointer. 197
12.2 The chassis from the CD-ROM drive with the laser mounted on the top right
and the photoresistor on the bottom right. The black obstacle is mounted
on the carriage that can be moved via the spindle on the top by a small
stepper motor that is located below the laser. 198
12.3 The schematics of the circuit. The stepper motor of the frame is connected
to points labelled PA,. . . ,PD. 199
12.4 The raw sensor value as a function of the position of the obstacle and the
derived laser beam profile, which shows a moderate asymmetry. 203

13.1 The chassis of the robot with two breadboards. The smaller one can be
turned by operating a model-servo. Mounted on the larger breadboard are
a NodeMCU on the right and a second chip on the left, a bare ATmega328
that was initially tested and later replaced by an Arduino NANO. 206
13.2 Simplified setup of the remote controller on a breadboard. 207
13.3 The schematic of the remote controller. 207
13.4 The electronics circuit of the robot (color version available online). 212
13.5 The schematic of the robot electronics. 213
13.6 The operational robot from the back (left) and from the front (right). 221

A.1 Two impedances connected in series (left) and in parallel (right). 230
Preface

Some years ago three young students inquired about a moderately complex project to earn
some credits. I happily agreed to supervise them and assigned moderately difficult tasks,
namely, to build from scratch a data acquisition system for slow signals. I suggested to
connect some sensors to an Arduino microcontroller and then write a program for the
Arduino to interface the measurement values to the control system we use in our lab.
The students were very dedicated and a real joy to work with. They had the Arduino
under control within a few hours and had the first sensors reporting their measurement
values after the first day. Then they worked out a protocol that is compatible with our
EPICS-based control system, and after discussions with our control systems experts and
even more debugging, eventually the students had a prototype system working. After clean-
ing up their project, they had to give a presentation and write a report to earn their
well-deserved credits.
I soon realized that there is a clear progression of the information generated by a sensor.
The information bubbles upwards through a sequence of microcontrollers and computers
that provide data-handling, storage, and online presentation to a seminar presentation, and
eventually ends up in a report. Understanding the path the measurement data take ap-
peared like a useful concept to communicate to students. Moreover, I wanted the students
to understand the details of the signal chain and how it really works. Therefore, I used
the hands-on approach with programming the Arduino that serves as communication glue
between the sensor and the control system. This proved beneficial for the students’ under-
standing and was appreciated by them. The abstract concepts thus led to a very concrete
realization. In the final stages of the project I coached the students on how to prepare a
presentation for a seminar according to some simple guidelines, and eventually put the oral
presentation into writing for a report to hand in and receive their credits.
This book is inspired by these students and their projects, but goes a step further and
adds a number of additional topics such as signal conditioning, controlling actuators such as
switches and motors, as well as control system setup, data storage, and networking. Please
note that I cover only basic examples that are boiled down to the bare essentials in order
to illustrate the main concepts and to get started quickly. Anyway, the concepts covered
should come in handy when working with real-world data-acquisition tasks. I basically follow
Mrs. Robinson’s guideline of “help you learn to help yourself” (remember the Simon and
Garfunkel song?) and try to fill the toolbox with practical know how. This know how should
enable the reader to help herself and pick up datasheets and manuals to adapt the basics
from this book to realize far more advanced projects.

User Guide
The main theme of the book is From Sensor to Report, and that should be the guiding
principle of using it in the classroom, either in a student laboratory or as the basis for
individual projects.
For a student laboratory I suggest installing the software with some of the more arcane
instructions before starting the lab. This comprises turning the Raspi into a router (Sec-

xv
xvi  Preface

tion 5.4), installing the MySQL database (Section 5.6.2), and installing EPICS (Section 6.1).
The students should focus on the sensors and use the above systems as a background infras-
tructure. They should, on the other hand, understand the basic operation of the sensors,
learn how to interface them to a microcontroller, and move the information to the next level
on a different computer. This requires them to write network code, fill an SQL database,
prepare the protocol files for EPICS, or present data on a web server. In the lab a knowledge-
able supervisor, a “tutor”, should be available to answer questions and guide the students.
Using solderless breadboards in the lab enables the students to quickly arrive at a working
system on which to base further experiments and try out new ideas.
A suitable scope for student projects, suitable for a single or a group of two students,
is to connect a small number of sensors to an Arduino. Then they should be given a target
system where they can publish the data. This can be a database, EPICS, MQTT, or a web
page. After a prototype system is working, the students should present their system in a
seminar and prepare a report.
All code and the corresponding images of the circuits on a breadboard, prepared
with Fritzing [1], are available on this book’s web site at https://www.crcpress.com/
9780815393603.
Acknowledgments

This book only materialized because my students, Adam, Måns, and Frida, asked about
“some project” and then completed it with such enthusiasm. I gratefully acknowledge their
contributions and input.
I gratefully thank my colleagues Roger Ruber and Mattias Klintenberg. They read and
commented on parts of the manuscript and provided essential feedback. All remaining errors
are of course my responsibility.
I thank Camilla Thulin, Uppsala University, for her help with the photography.
This book is only possible thanks to the open source community that created the Arduino
and Raspberry Pi ecosystems and the large number of people who answer questions on
Internet forums.
I acknowledge the creators and maintainers of the Fritzing software. I relied on it to
prepare many drawings to illustrate the wiring of circuits.
I am indebted to my editor Francesca McGowan and to Rebecca Davies at Taylor &
Francis for competently guiding me through the intricacies of writing and publishing a book.
Last, but not least, I acknowledge my family for putting up with me during the writing
and editing period, when I was more often than not absent minded and showed a distinct
lack of response to other matters.

xvii
 CHAPTER 1

Introduction

What is the path that the electrical signal from a sensor takes to end up in a report? We
address this question because collecting sensor data, processing them, and deriving some
understanding from the data plays an important role in many circumstances. One example
is a utility company that gathers information about electricity, heating, and water in order
to prepare statements for their customers and to estimate demand for their product in the
future. Smart homes are another example; they measure temperatures or detect the presence
of beverages in the refrigerator to adjust the thermostat in the first case, or to prepare
a report for us to pick up some milk or beer on the way home in the second case. Quite
generally, many Internet of Things (IoT) technologies share a common base with the topic of
the book, but even large experimental collaborations such as the ATLAS [2] or CMS [3] that
operate the huge detectors at the Large Hadron Collider (LHC) [4] at CERN [5] get their
data from sensors that are buried deep inside the detectors. They sense currents from drift
chambers where charged particles cause a discharge between wires at different potentials, or
they cause electrical signals from semiconductor detectors, where they create electron–hole
pairs that induce a current. Other examples are Hall sensors, to measure magnetic fields, and
humidity sensors or barometric pressure sensors to detect variations of ambient conditions.
All these sensors produce electrical signals that often need to be amplified or otherwise
conditioned. This stage involves operational amplifiers and various filters to improve the
signal-to-noise ratio. Once properly processed, the analog signals are passed on to analog-
to-digital converters (ADC), where they are converted to a digital representation that is
subsequently handled by computers. Often some of the computing power is located close
to the sensor and is provided by microcontrollers that collect signals from nearby sensors
and convert them to the underlying physical quantities, formatted to have a standardized
output format. Thus they act as “communication glue” between the specific interface to
the sensor and a more generic interface to a host computer that is usually located further
away. The latter is the other end of the communication channel from the microcontroller
and provides data storage and presentation, and sometimes also shows recent data for on-
line monitoring. The host computer may run generic control-system software to provide a
further abstraction layer towards higher-level software. Examples we discuss are MQTT [6],
which is popular with IoT projects, and the EPICS control system [7], commonly found in
scientific laboratories.
In this book we will build a system that contains all the ingredients also found in
large scientific or industrial installations. In a sense it is a simplified model of a large
installation, yet containing all the hardware and logical building blocks. In particular, we
use the Arduino microcontroller [8] as the local intelligence to control switches and motors
to move the sensors around and enable reading them out and translating the signals to
a format that allows communication with a host computer using a standardized protocol.

1
2  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Analog Digital
sensors sensors

Actuators
I2C switches, motors
Signal SPI
conditioning

Digitization Reading

Turning on and moving

the sensors
Preparation to standard Arduino
output format Microcontroller

Communiating with the

host computer

RS-232, WIFI
Protocol

Communication with the

micro-controller
Raspberry Pi
Data storage Host computer

Data presentation
Publish with web server
EPICS control system
MQTT messaging system

Presenting at seminar

Writing the report

Figure 1.1 The outline of the book.

 Introduction  3

This comprises the hardware channels, which can be USB, RS-232, Bluetooth, or WiFi as
well as the logical protocol, such as a simple query-response protocol. As host computer
we use a Raspberry Pi [9] which, in its most recent incarnation, features four processing
cores and runs a standard Linux system with a huge base of available software, including
web servers, the MATLABTM -clone Octave, and even MathematicaTM , all without license
charges.
The remainder of the book is organized as laid out in Figure 1.1. We first discuss a
number of analog sensors and signal-conditioning methods, followed by a number of sensors
that already provide their measurements in digital form, and the buses and protocols used.
Next we discuss actuators. They switch things on and off, even those requiring large currents,
and we learn how to control different types of motors that are sometimes needed as part of
the measurement process. We then describe how to interface the actuators and the sensors
with the Arduino, either by digitizing the signals or by using the appropriate bus-interface.
We go on to describe a program structure that permits the Arduino to support a simple
query-response protocol in order to serve as a slave to a host computer. Next, we configure
a Raspberry Pi as a standardized host computer. It will provide data storage in databases,
and present the data in graphical form either using Octave running locally on the Raspi
or by publishing our measurement data with a web server, also running on the Raspi. We
continue the discussion by installing the EPICS control system software and turn the Raspi
into a full-blown control-system server that can join any other EPICS installation in a
transparent way. We go on to discuss the MQTT message-passing system, which plays an
important role in IoT applications. Having assembled all the parts, we consider examples
in which we build a weather station with distributed sensors, and systems to record ground
vibrations, monitor the color of water, and measure the capacitance of a capacitor. In two
more advanced examples, we build a system to measure the width of the beam of a laser
pointer, and a remote-controlled robot that also senses flames autonomously, moves to the
fire, and sounds an alarm. We conclude with presenting guidelines about how to prepare
a seminar presentation based on the examples and how to write a readable publication
describing our data acquisition system using sensors, actuators, Arduino, and Raspberry
Pi.
 CHAPTER 2

Sensors

A sensor is a device that converts a physical quantity to an electrical signal [10, 11, 12, 13],
and therefore either provides a voltage or a current, or causes a change of its resistance.
More generally, the impedance of the sensor, which also comprises capacitive and inductive
sensors, may change. We are thus faced with the task to measure either of these electrical
quantities.
Below we discuss examples of the different types of sensors. The examples only show
a selection of those available on the market, and searching the Internet for the physical
quantity one wants to measure jointly with the keyword “sensor” will give an idea of what
is available. Once the sensor is identified, careful reading of the datasheet to learn about
how to interface the sensor is mandatory.

2.1 ANALOG SENSORS

We start by considering resistance-based sensors.

2.1.1 Resistance-based sensors

The first resistance-based sensor we look at is a light-dependent resistor or LDR, shown
on the left of Figure 2.1. It changes its resistance depending on the exposure to light. The
range of variation depends on the device and typically ranges from a few 100 Ω to MΩ. The

Image of a light-dependent resistor (LDR) on the left and how to connect

Figure 2.1
it in a plain voltage-divider configuration (center) and in a Wheatstone bridge
(right).

5
6  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Few thermically Lower boundary of

Supply voltage equals exited electrons conduction band
voltage drop across
both resistors
Measured
voltage
Upper boundary
of valence band

Highest energy
for electrons
Vcc
LDR

Wire Wire Wire

Resistor

Vcc Signal

More electrons
in conduction
Measured band
voltage
Band gap
2.4 eV

Supply voltage equals

voltage drop across
both resistors Photon Upper boundary
of valence band
Hole in
valence band
Highest energy
for electrons

Conduction band
lower boundary

Vcc Photon
LDR

Wire Wire Wire

Resistor

Vcc Signal

Figure 2.2The band-level scheme and the schematics for the circuit with the LDR in
a voltage divider. The upper graph illustrates dark conditions and the lower graph
shows conditions where the LDR is exposed to light. Note that the vertical axis
by convention shows the energy of electrons. This causes the positive pole of the
battery to have the most negative energy. See the text for a discussion.
 Sensors  7

operating principle of this and many other sensors is based on the availability of electrons in
the conduction band of a material. In good (wires) or bad (resistors) conductors, electrons
partially fill the available states in the conduction band up to some energy, the Fermi level,
whereas in insulators the Fermi level is located in between the completely filled valence band
and the conduction band [14]. Therefore, no electrons are available in the conduction band.
Furthermore, the energy-difference between the upper boundary of the valence band and
the lower boundary of the conduction band, the bandgap, is large, while for semiconductors
it is on the order of electron-volt (eV).
In photoresistors the base material is often CdS, a semiconductor with a bandgap of
about 2.4 eV. This energy equals that of photons of green light with a wavelength of about
500 nm. Therefore, green photons can elevate electrons from the valence to the conduction
band and thus create electron–hole pairs. These now freely moving charge carriers conduct
electric current and therefore increase the conductivity of the material. Figure 2.2 illustrates
this in more detail. The upper figure shows a simplified band level scheme under dark
conditions. The bold lines show the lower boundary of the conduction band and the upper
boundary of the valence band. The dashed line shows the highest energy-states that electrons
occupy. For metals and resistors, this is close to the Fermi level, but in a semiconductor, like
CdS, the Fermi level lies between valence and conduction band. Yet, at room temperature,
there are a few thermally excited electrons in the conduction band of CdS. We visualize this
by the close proximity of the dashed line to the lower conduction-band boundary. In metals
there are plenty of electrons in the conduction band, and the conductivity is high. In the
resistor there are fewer electrons in the conduction band, or their mobility is impeded in
other ways such that there is a shift in the Fermi level across the resistor. In dark conditions
there are only very few thermally excited electrons in the conduction band of the LDR, and
the conductivity is very low. Consequently, there is a large voltage-drop across the LDR
and the measured voltage, which is the difference of Fermi levels between the measurement
points. The measured voltage is therefore close to the full voltage delivered by the battery.
If, on the other hand, the LDR is illuminated, the photons lift electrons from the valence
band into the conduction band. This increases the conductivity and only a small voltage is
dropped across the LDR. This consequently reduces the measured voltage, as shown in the
lower graph of Figure 2.2. Note that we do not discuss the details of the interfaces between
the different parts because it is beyond the scope of this book.
In the middle of Figure 2.1 we show how to connect a photoresistor in series with a
resistor R0 (here 10 kΩ) to create a voltage divider between the supply voltage Vcc and
ground. From the discussion of voltage dividers and a short refresher of basic circuit theory
in appendix A or [15, 16], the voltage Vs on the signal terminal is then given by Vs =
Vcc RLDR /(R0 + RLDR ). Thus, the illumination of the LDR changes its resistance RLDR
and the signal voltage varies correspondingly. Note that we have to select the resistor R0
in the middle of the range of RLDR . This causes the voltage we measure to be around one
half of the supply voltage. Therefore, we also need to use a volt meter in that voltage range.
Very small variations of the light intensity are then difficult to resolve and may need to
be amplified. Using a Wheatstone bridge, where we compare the voltages in two resistor
dividers, as shown on the right of Figure 2.1, helps to alleviate this problem. We expand on
the use of Wheatstone bridges in Section 2.2.
Other resistance-based sensors are resistance-based temperature detectors (RTD) such as
the PT100 temperature sensor. It is a calibrated platinum-based sensor with a resistance
of exactly 100 Ω at 0 o C. It is based on the fact that the resistance of a very pure metal
is determined only by scattering of electrons in the conduction band with phonons, which
are vibrations of the ions that make up the crystal lattice of the metal. Moreover, higher
8  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Energy Conduction band

lower boundary electron

Band
Occupied donor levels Partially occupied gap
donor levels

Valence band
at low upper boundary at higher
temperature temperature

Figure 2.3NTC resistors are doped semiconductors that have donor levels just below
the conduction band. Increasing the temperature increases the kinetic energy of the
electrons that allows them to occasionally jump into the conduction band, where
they contribute to the conductivity of the material.

temperatures cause stronger vibrations of the lattice, with correspondingly higher resistance.
Intuitively one might think of the crystal ions at higher temperature to oscillate with larger
amplitudes, creating a larger target for the electrons to scatter, thus impeding their motion.
Since this is an intrinsic property of the material, calibration measurements of resistance as
a function of temperature are universally valid for all sensors of the same metal, provided
the metal is very pure and free of imperfections. Commercial sensors are often made of
platinum wire wound on a ceramic support body. The PT100 sensors are connected to a
calibrated current source and the voltage drop across the sensor is measured with a volt
meter, just as any other resistance measurement.
Thermistors are resistors that have their temperature dependence deliberately made
large. In positive temperature calibration (PTC) devices, the resistance increases with tem-
perature, and in negative temperature calibration (NTC) devices, it decreases. PTCs are
mostly used as protection devices that switch the resistance from a low- to a high-resistance
state if a certain temperature is exceeded. They are based on polycrystalline materials that
change their dielectric constant at a certain temperature, the Curie temperature, by a large
amount. Above the Curie temperature, the state of the magnetic dipoles is disordered and
the dielectric constant is small. This causes the formation of large potential barriers be-
tween the crystal grains, which leads to a high resistance. Below the Curie temperature the
molecular dipoles are aligned, the dielectric constant is large, and the resistance is low. A
typical application of the PTC thermistor is a self-regulating heater, in which the heater
also warms up the thermistor, which increases the resistance and limits the current to the
heater until an equilibrium is found. PTCs can also be used to detect whether a threshold
temperature is exceeded.
The converse thermistors are NTCs, which decrease their resistance with increasing
temperature. They are often used for temperature sensing and are based on a doped semi-
conducting material that has occupied impurity donor levels below the conduction band, as
shown in Figure 2.3. Increasing the temperature thermally excites these electrons to jump
into the conduction band, thus increasing the conductivity. This effect is much larger than
the reduction of the resistance due to the ions oscillating and impeding the motion of the
electrons, which was responsible for the temperature dependence in the PT100 sensor. Both
NTC and PTC thermistors are sensed by connecting them to a constant-current source and
measuring the voltage drop across the thermistor.
 Sensors  9

A number of position sensors are based on potentiometers. A potentiometer is a variable

resistor where a slider moves up and down a resistance and shortens the distance of one end
point to the slider, thereby reducing the resistance between two terminals. The distance be-
tween the slider and the other end point lengthens, causing the resistance between the slider
and the other terminal to increase correspondingly. On the left-hand side of Figure 2.4 we
show a potentiometer with three connectors; the two end points are connected to dark wires
and the one controlled by the slider is connected to a lighter-colored wire. The schematic
view on the right of Figure 2.4 explains the functionality; the slider controls a variable mid-
point of a voltage divider and the output voltage interpolates from 0 to 5 V in this case. A
variation of the potentiometer is a joystick, which is based on two orthogonally mounted
potentiometers, controlled by a small stick. We simply need to measure the voltage on the
output with respect to ground in order to determine the position of the slider or the stick.
An image of a joystick is shown on the left of Figure 2.5.
Also, fluid levels of liquids with a small conductivity can be determined with a resistor
whose resistance is varied by reducing the resistance between conducting stripes shown in
Figure 2.5. The sensor is connected just as a potentiometer with the liquid level acting as
the slider.
Applying an external force to a material will cause a change its equilibrium shape, and
therefore strains the material. An example is a stretched wire that will get longer and
thinner if pulled. Consider the resistance R = ρL/A of the wire, given in terms of resistivity
ρ, length L, and cross section A, and consider its change. We see that increasing L and
decreasing A increases the resistance R. Thus, the simple wire converts its deformation to a
small change in the resistivity, and therefore serves as a force-sensitive resistor that is often
used as one branch in a Wheatstone bridge. The small effect of only changing the geometry
of the wire can be greatly enhanced by using heavily doped semiconductors instead. The
latter also change their resistivity ρ by a large amount when strained. Alternatively, they are
based on polymer thick-film technology. An example is the strain gauge shown in Figure 2.6.
The MQ-2 and other MQ-x are gas sensors sensitive to different types of gas, depending
on their type specification. They are based on a semiconductor substrate with a thin surface
layer of polycrystalline SnO2 tin oxide that is either sputtered or deposited by evaporation.
The specificity to various gases depends on the temperature of the active area, which is
adjusted by a heater, by the method of deposition, or by small additions of other materials,
such as palladium, gold, or platinum. The variation of the resistance depends on the grain
boundaries of the polycrystalline SnO2 and on the insulating oxide layer between the grains

Figure 2.4A linear potentiometer (left) and a circuit illustrating the electric connec-
tions (right).
10  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.5 A joystick is shown on the left and a fluid-level resistive sensor on the
right.

that is affected by the adsorbed gases. On the left of Figure 2.7 we illustrate the working
principle. The heater is located under the SnO2 active layer and powered by passing a
current through it. The resistance, which depends on the concentration of the specific gas,
can be measured between the terminals labeled 1 and 2. A device mounted on a small
breakout board is shown on the right of Figure 2.7. The sensing area is located under the
metallic hat that protects against the heated area, and potential explosive reactions on the
hot surface should a “wrong” gas ignite.
After this short selection of resistance-based sensors we progress to discuss sensors that
report a voltage directly.

2.1.2 Voltage-based sensors

An example of a sensor that directly produces a voltage at its output pin is the LM35
temperature sensor, which is a silicon-bandgap temperature sensor. The operating princi-
ple is based on passing known currents In with n = 1, 2 with current densities jn across
the base-emitter junctions of two bipolar transistors, and comparing their respective volt-
age drops VBE,n . The voltage difference is proportional to the temperature. This is easily
understandable by inverting the current–voltage curve for the diode of the base-emitter
junction h i
jn = A(T ) e(eVBE,n −Eg )/kT − 1

where Eg = 1.2 V is the bandgap energy of silicon, k is the Boltzmann constant, and
T the absolute temperature in Kelvin. A(T ) is a device-specific constant with moderate
temperature dependence. Assuming that both transistors are located on the same substrate
and have the same temperature, we solve for two current densities j1 and j2 and obtain
for the voltage difference ∆VBE = VBE,1 − VBE,2 = kT /e ln(j1 /j2 ). In the LM35, the base-

Figure 2.6 A strain gauge.

 Sensors  11

Uh

Sn2O

1 2

Heater

Schematic of an MQ-x gas detector (left) and a sensor mounted on a small

Figure 2.7
breakout board (right).

emitter diodes of the two transistors have different areas such that the ratio of the areas
determines the current densities, provided that the same macroscopic current passes through
the two transistors. There are operational amplifiers on the same substrate to provide signal
conditioning such that the LM35 produces an output voltage Vs that is related to the
temperature T by Vs = T /100. Here Vs is measured in volts and the temperature in degrees
Celsius, such that a temperature of 23 o C results in a voltage of 0.23 V. The LM35 has three
pins; one is connected to ground, one to the supply voltage, and the third one carries the
voltage Vs that is proportional to the temperature. Note the polarity for connecting the
LM35 in Figure 2.8. The flat surface is pointing to the wires on the left-hand side.
Thermocouples are temperature sensors that are based on the effects of temperature and
temperature gradient on conductors made of different materials. Directly at the junction of
the conductors, the Peltier effect causes a current that depends on the temperature. This
happens at the points labeled by their respective temperatures T1 and T2 on the top left
in Figure 2.9. On the wire segments a temperature gradient causes an additional current
to flow, the Thomson effect. And finally, joining the two junctions and the wires causes a
current to circulate, provided the loop is closed. This is called the Seebeck effect. If the loop
is open, as shown at the top left of Figure 2.9, a voltage U develops at the end terminals
as a consequence of the Peltier, Thomson, and Seebeck effects. In practice, one junction,
say at T1 , is held at known and constant temperature, for example, by immersing the

Figure 2.8 Image of an LM35 temperature sensor (left) and how to connect it (right).
12  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

material A

T1 T2
material B U material B

T1

A A A A A A
B B B B B B

T2
U

Figure 2.9On the left we show a schematic of a thermocouple on the top and a
thermopile on the bottom. On the right we show an image of an MLX90614 contact-
free thermometer.

junction in ice water. Then the voltage U is related to the temperature difference T2 − T1
of the sensing end at T2 and the reference temperature T1 . The magnitude of the voltage
generated depends on the combination of metals and is typically on the order of 50 µV/o C.
In a thermopile a number of wire segments of materials A and B are connected in series,
as shown on the bottom left in Figure 2.9. This increases the sensitivity of the device to
temperature differences. Thermopiles are often found in devices sensing heat and infrared
radiation, such as thermal imaging devices or contact-free thermometers. The sensor used
in the latter is shown on the right of Figure 2.9.
Some crystals and ceramics react to external stresses by producing a piezoelectric voltage
between opposite sides of the material, as a consequence of rearranging charges within their
crystal structure. The resulting voltages reach several kV and can be used to produce
sparks in ignition circuits or in old-fashioned vinyl record players. There, a “crystal”-stylus
is squeezed in the grooves of the record and the generated voltages are amplified and made
audible as sound. In scientific applications, piezoelectric sensors are used to measure pressure
or forces.
The speed of angular motion is easily sensed by a DC electrical motor that is operated
backwards as a generator. Instead of applying a voltage to turn the axis of the motor,
turning the axis induces an induction voltage in the motor coils that is proportional to the
angular velocity. Attached to a propeller that is turned by either a flowing liquid or a gas,
such a device can measure flow rates.

+Uh/2

I I
E
B

-Uh/2

Figure 2.10 Schematic of a Hall sensor (left) and the A1324 sensor (right).
 Sensors  13

Hall sensors produce a voltage that is proportional to the magnetic induction B. Their
mode of operation is explained in Figure 2.10 and is based on passing a known current
I through a semiconductor. In the presence of a magnetic field, the Lorentz force de-
flects the charge carriers—electrons and holes—towards perpendicularly mounted electrodes
(shaded). This creates a potential difference (a voltage) between the electrodes, which causes
a transverse electric field that counteracts the deflection from the Lorentz force such that the
following charge carriers can move towards the exit electrode undeflected. In equilibrium,
the voltage difference between the upper and lower electrodes is proportional to the mag-
netic induction B and can be measured with a voltmeter. The A1324 Hall sensor, shown on
the right in Figure 2.10, has signal-conditioning circuitry on board and only needs three pins
for ground, supply voltage, and output voltage. The latter is proportional to the magnetic
field, with a sensitivity of 50 mV/mT centered at 2.5 V when no field is present.
The ADXL335 is a three-axis integrated acceleration sensor based on micromachined
structures on a silicon substrate where an inertial mass is suspended by springs [12]. The
inertial mass is part of an assembly of capacitors driven by an AC voltage that is used to
measure the imbalance of a capacitive voltage divider. As opposed to our simplified model
with only one capacitor doublet, the real device uses a large number of interleaved capacitor
doublets in order to increase the sensitivity. Figure 2.11 illustrates the principle of operation
for a single direction. On the left in Figure 2.11 there is an AC-voltage generator that drives
the light-grey capacitor plates. The dark-grey inertial mass is placed halfway between the
driven plates, and in the absence of acceleration, the capacitances between the two light
grey plates and the inertial mass are equal. An acceleration introduces an imbalance in the
capacitances that affects the voltage level on the inertial mass. Comparing the phase and
amplitude of that signal with that of the driving AC signal yields direction and magnitude
of the acceleration. After some signal processing, it is then made available as Uacc on one
of the output pins of the ADXL335 in the range from 0 to 3 V, such that the voltage is
proportional to the acceleration in the range from -3 g to +3 g and is updated at a rate of
about 100 times per second.
The SM-24 is another type of accelerometer called a geophone, and is shown in Fig-
ure 2.12. It is based on a coil connected to the housing by springs embedded in a magnetic
field generated by permanent magnets that are attached to the housing. If the housing
moves, the coil remains stationary due to its inertia, and a voltage is induced in the coil,
which is proportional to the velocity and can be measured. The sensitivity is 28.8 V/(m/s)
and the device operates in the range of 10-240 Hz.

acc

C+dC Compare Phase

and Amplitude
C-dC Inertial
mass
Uacc
Spring

Figure 2.11 The operational principle of an ADXL accelerometer.

14  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.12 An SM-24 geophone with a diameter of about 30 mm.

Microphones convert sound to electrical signals and can be classified as sensors. Two
major classes are on the market. Dynamic microphones operate similarly to the geophones.
A coil, attached to a membrane, is excited by sound, moves in a magnetic field, and induces
an induction voltage in the coil that is amplified and measured. In electret microphones, the
membrane constitutes one electrode of a capacitor. If it moves, the capacitance changes,
and the amount of charge stored on the capacitor is pushed on and off the capacitor and
creates a current that is amplified and measured.
The electret microphone serves as a nice example with which to turn to current-based
sensors.

2.1.3 Current-based sensors

The BPW34 is a pin diode that generates a current of up to 100 nA depending on the
irradiance of up to mW/cm2 . Pin diodes are similar to conventional diodes and consist of a
semiconductor with a pn junction, which is created by doping semiconductor material, often
silicon based, with a material that has either five valence electrons, in which case it becomes
n-doped, or three valence electrons, in which case it becomes p-doped. Pin diodes, however,

E +Vb
n i p
electron raised
to cond band acceptor
levels R
Rf
nip
photon Vb _
conduction
band Ip Ip
OpAmp
donor Vo
levels
+

valence
band

Figure 2.13Energy-band diagram (left) and circuit (right) of a reverse-biased pin

diode in photoconductive mode.
 Sensors  15

Figure 2.14Image of a BPW34 pin diode on the left and two phototransistors on the
right; an IR-sensitive BPX38 and an SFH3310, sensitive in the visible part of the
spectrum.

have an additional layer of un-doped, intrinsically (‘i’) conducting silicon in order to increase
the target area for the photons and provide a chance to produce additional charge carriers.
One operational mode of the pin diode, called photoconductive, is illustrated on the left-
hand side of Figure 2.13, which shows a simplified energy-band diagram of a reverse-biased
diode. Note that by convention, the upwards energy axis corresponds to the potential energy
of electrons that is lowest at the most positive voltage and that is found on the left-hand
side. In the figure the cathode (n side) is therefore at higher voltage than the anode (p
side) and results in all charge carriers being pulled out of the intermediate zone; electrons
to the left and holes to the right. This results in the diode blocking any current flow. The
extra layer of un-doped silicon provides extra potential charge carriers that act as targets
for photons, having energy higher than the band-gap. These photons create electron–hole
pairs by lifting electrons from the valence band into the conduction band, as indicated in
Figure 2.13. The applied voltage, which is more positive on the left-hand side, causes the
electrons to move to the left and the holes move to the right. Combined, this constitutes
a current Ip . We mention in passing that ionizing radiation, such as high-energy photons
and gamma rays as well as charged particles with high energies, create electron–hole pairs.
This makes pin-diodes suitable as radiation detectors. The circuit on the right-hand side of
Figure 2.13 shows an operational amplifier that converts the current Ip flowing towards its
negative input port into a voltage Vo = −Rf Ip on its output port. We will cover operational
amplifiers in more detail in the coming sections. A BPW34 pin diode is shown on the left
of Figure 2.14.
Phototransistors such as the BPX38 or SFH3310, both shown on the right of Figure 2.14,
are similar to normal transistors, but their base-collector diode is a reverse-biased photo-
diode, similar to the one described in the previous paragraph. It causes a current to flow
as a consequence of impinging photons. The base-emitter diode is already forward biased
and will ensure that the collector–emitter connection becomes conducting. Moreover, often
there is a lens to increase the number of photons impinging onto the base terminal with the
photosensitive area. Phototransistors sensitive to special spectral ranges such as infrared
radiation, by suitably choosing their band gap, can be used as flame detectors.
The sensors in imaging applications such as cameras are charge-coupled devices, or CCDs,
which are similar to a pin diode that is attached to a small capacitor, one for each pixel
of the camera. Exposure to light transfers a small charge to the capacitor. The often large
number of pixels are read out sequentially by transferring the charge from on capacitor to
the one closer to the external readout port. A bucket chain to transfer water that is emptied
16  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.15On the left we have a voltage divider to reduce the input voltage of 0-5 V
to 0-3.3 V. The right circuit shows the use of clamping diodes to protect the input
of the microcontroller to lie between ground and 5 V.

at the end point comes to mind. In a CCD, once the charge reaches the output port, it is
passed through a resistor, where it creates a voltage drop that can be measured.
Solar cells operate in a similar fashion to photodiodes in photovoltaic mode where they
provide a voltage to a load. They are, however, optimized to absorb as large a part of the
spectrum as possible, and also to have a large absorbing area, in order to maximize the
electric power available to the load.
After this brief overview of different analog sensors, we need to address how to prepare
the signals such that they can be easily interfaced. This preparatory stage is referred to as
signal conditioning.

2.2 SIGNAL CONDITIONING

The analog signals from the sensors can be too high or too low. They can be too noisy
or otherwise inadequate to directly feed to an analog-to-digital converter (ADC) or a mi-
crocontroller. In many cases some signal conditioning is needed, and we will discuss some
common methods in the following sections.

2.2.1 Voltage divider

In case the input voltage of the sensor exceeds the input range of what the microcontroller
can handle, we have to reduce the voltage by a voltage divider, which consists of two fixed-
value resistors. A typical example is to reduce the range from 0-5 V to 0-3.3 V, which is
accomplished by a combining two resistors with a ratio of R1 /(R1 + R2 ) = 3.3/5 = 0.66.
A close example is shown in Figure 2.15, with a 12 and a 22 kΩ resistor. The ratio is
22/34 = 0.65, which is close to the desired ratio. Other combinations with larger and
smaller values will work as well, as long as the ratio is correct. The sum of both resistors
should not be too small because that will draw a larger current from the 0-5 V voltage
source, and, depending on the internal resistance of the source, may affect the measured
values.
In order to improve the measurement sensitivity, resistance-based sensors are often wired
in a Wheatstone-bridge configuration. An example was shown on the right of Figure 2.1.
Here the voltage divider on the left side of the breadboard provides half the supply voltage
 Sensors  17

Figure 2.16A bare operational amplifier shown on the left and shown wired as a line
buffer on the right. We omitted the supply wires on the latter schematic.

at its center tap (the lower-signal wire) because the two resistors are equal. The voltage
divider on the right-hand side is the same one, we encountered previously, with the upper-
signal wire connected to the point between the upper resistor and the LDR. Normally one
would choose the resistance of the right resistor to be in the middle of the range of interest
of the LDR, such that voltage-difference between the signal wires is close to zero, indicating
mid-range. In this way, depending on the light exposure, the voltage difference between the
wires varies around zero and the sign will tell us whether the exposure is lower or higher
than the expected mid-range value. Since we now deal with voltages that vary around zero,
it is easier to amplify that voltage in order to increase the sensitivity. For example, when
using a plain voltmeter, we can use a smaller voltage range.
In case we use a piezo-based sensor, the generated voltages can be much higher than is
acceptable in the following circuit such as a microcontroller. In such a situation, clamping
diodes, as shown on the right of Figure 2.15, are used. If the input voltage is between 0 and
5 V, the diodes are blocking, and the signal is passed on to the microcontroller. If, on the
other hand, the input voltage exceeds 5 V plus the forward voltage drop, the diode starts
conducting and shorts the input to the upper power rail of 5 V. If the input voltage is below
0 V, the lower diode starts conducting and shorts the input to ground. In either case, the
voltage delivered to the microcontroller is limited to 0-5 V plus or minus the forward diode
voltage drop. Many integrated circuits including microcontrollers have built-in clamping
diodes. The Raspberry Pi, however, is a notable exception.

2.2.2 Amplifiers
Very small electrical signals usually need to be amplified to reach levels adequate for further
processing. The standard device to achieve this is an operational amplifier or op-amp, shown
on the left of Figure 2.16. There are two input ports on the left, one labeled “plus”, one
labeled “minus”, and one output port. The latter delivers a voltage that depends on the
difference between the two input ports. In an ideal op-amp, the amplification factor is
infinite and we usually use some feedback mechanism to obtain a deterministic behavior,
as discussed below. Normally, op-amps require both positive and negative supply voltage,
even though sometimes it is possible to tie the negative supply rail to ground, in which case
only unipolar signals can be amplified.
Before discussing different circuits, we need to describe three basic principles that char-
acterize op-amps.
18  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.17 A non-inverting (left) and inverting (right) amplifier.

The input impedance of the input ports is “infinite”, which means that no current
flows into the op-amp and it will not load the upstream circuitry.
The op-amp tries to reduce the difference between the input ports V+ − V− to zero.
The amplification without feedback is quasi-infinite, and the output has a very low
impedance and can provide high output currents.

These simple rules will help us to design and understand the following circuits, but consult
[15] and [16] for a more extensive discussion.
We start by considering a line buffer, which is typically used as an impedance converter
that transforms the output of a sensor with a high impedance to a low-impedance signal
that is less susceptible to noise. We show the circuit on the right of Figure 2.16, where the
output of the op-amp is fed back onto the negative input port. By using the second op-amp
rule, we see that the op-amp tries to make the positive and negative input ports equal,
but negative is tied to the output, which forces the output to follow the positive input.
Moreover, by the first rule, the input impedance is high and no current (or at least very
small current) is drawn from the sensor, while the output is low impedance and can provide
a high current. Note that we omitted the connections to the supply rails on the right of
Figure 2.16, and will do so henceforth in order not to clutter the schematics.
The line buffer is essentially an amplifier with unit amplification, but if we require a
higher degree of amplification, we need to add two additional resistors to the circuit, and
arrive at the left of Figure 2.17. The two resistors R1 and R2 constitute a voltage divider
that forces the negative input voltage to be V− = Vout R2 /(R1 + R2 ). But, at the same time,
the op-amp forces V− = V+ , which, after solving for Vout , leads to Vout = V+ (R1 + R2 )/R2 ,
where (R1 + R2 )/R2 = 1 + R1 /R2 is the amplification factor. Since the output voltage Vout
has the same sign as the input voltage V+ on the positive input port, this configuration is
called a non-inverting amplifier...
...which hints at the existence of an inverting amplifier, for which we show the schemat-
ics on the right of Figure 2.17. To calculate the amplification factor, we note that the
input current Iin only flows through the input resistor R4 and the feedback resistor R5
because the input impedance of the op-amp is essentially infinite and no current flows into
the input ports. But the equality of the current in the input and the feedback resistor
implies I = (Vin − V− )/R4 = (V− − Vout )/R5 . Moreover, we observe that the positive in-
put port is grounded, which forces the negative input port to be on ground potential as
well. This implies V− = 0, and the relation between input and output voltage becomes
 Sensors  19

Figure 2.18A difference amplifier (left) and the same circuit with adjustable V1
(right) that is used to subtract the baseline.

Figure 2.19 Instrumentation amplifier (left) and logarithmic amplifier (right).

Vout = −Vin R5 /R4 , where the negative sign indicates that the amplifier is inverting. Note
that we can add several input resistors R4 in parallel with one end connected to the negative
input port. This allows us to add the currents passing through the parallel copies of R4 , and
we obtain a summing amplifier. The resistor values were simply chosen to be in a reasonable
range. They need to be determined adequately for each application.
Sometimes the signal one wants to measure changes around a non-zero baseline. Exam-
ples are the Hall sensor A1324 from the previous section, where zero magnetic field produces
2.5 V and the magnetic field added or subtracted from that value depending on its polarity.
In order to increase the resolution, we want to amplify not the signal, but the difference of
the signal to the baseline. In other words, we need a circuit to subtract the baseline and
amplify the difference. A differential amplifier as shown in Figure 2.18 accomplishes this
feat, provided that R2 = R4 and R1 = R2 . The output voltage Vout in that case is given by
Vout = (V2 − V1 )R2 /R1 . Adding a potentiometer that adjusts V1 between the positive and
negative supply rail subtracts it as baseline voltage.
20  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.20 Amplifying a weak signal and shifting it to mid-range between the power
rails.

A circuit similar to the one in Figure 2.18 but with line buffers on the inputs is called
an instrumentation amplifier, and an example is shown on the left of Figure 2.19. Normally
it is not necessary to build instrumentation amplifiers from discrete components. There are
ready-made circuits available such as the AD620 or INA131.
Often, sensors produce small voltages that vary around zero, but the analog-to-digital
converter (ADC) requires an input range of 0 to 5 V. We thus face the problem of amplifying
a bipolar signal and changing the baseline level to mid range of the ADC. We show a circuit
that amplifies by approximately a factor of 10 in Figure 2.20. The amplification is mainly
determined by the ratio of the feedback resistor R3 to the input resistors R1 and R2 . The
voltage divider of R6 and R8 provides the mid-range offset voltage. The average level of the
output voltage crucially depends on the tolerances of the resistors, and in order to place
the level safely in mid range, we use the capacitor C1 to first remove the DC level of the
output signal before adjusting it properly to mid-range, with the voltage divider consisting
of R4 and R5 .
In case we need to amplify input signals that vary over a huge range of values, a loga-
rithmic amplifier such as the one on the right of Figure 2.19 is a useful circuit. It can be
shown that the relation between input and output voltage is Vout = −Vt ln(V1 /Is R), where
Vt is the thermal voltage and Is the saturation current of the diode. Swapping the diode
and the resistor results in an exponential amplifier.
A close relative to the operational amplifier is the comparator. It can be visualized as
an op-amp with very large, even infinite, amplification, whose output port saturates at the
power rails. If the voltage at the positive input terminal of a comparator is larger than
that on the negative input terminal, the output voltage is very close to the positive supply
voltage. On the other hand, if the voltage on the positive input terminal is lower than that
on the negative, the output is close to ground potential. In this way it translates the input
voltages to a binary digital state. A comparator may therefore be considered as a 1-bit
analog-to-digital converter, and we will see in a later section how it is used to extend the
number of bits of the conversion. Some comparators have the threshold when switching
from low to high output configured to be slightly higher than the threshold switching from
the high output state back to the low one. This small hysteresis prevents the output from
 Sensors  21

Figure 2.21 A simple low-pass (left) and high-pass (right) filter.

switching back and forth uncontrollably, should the voltages on the two input terminal be
very close.
After the basics of signal amplification, we will now address the question of how to reduce
noise in the circuits and decrease the sensitivity of a circuit in an undesired frequency range.
This is the realm of filters.

2.2.3 Filters
The task of filters is to remove certain frequencies from an electrical signal such as all high
frequencies, in which case the filter is called a low-pass filter. An example is a low-pass filter
that removes “hissing” in audio-signals. The converse filter is a high-pass filter that removes
low frequencies. An example is a anti-rumble filter found in old vinyl record players. If we
know that the desired signal contains only a certain range of frequencies and we wish to
remove all others, we use a band-pass filter. An example is the IF filter found in radios that
are based on the super-heterodyne principle. And finally there are filters that remove only
frequencies in a narrow band. They are called band-stop or notch filters. An example is a
filter that removes the omnipresent 50 Hz or 60 Hz hum coming from the power grid.
We first consider a low-pass filter, which in the simplest incarnation is a frequency-
dependent voltage divider made of a resistor with resistance R and a capacitor with ca-
pacitance C, as shown on the left of Figure 2.21. For a refresher of basic concepts circuit
theory, such as impedance, please consult appendix A. In our circuit, the capacitor has an
impedance 1/iωC, which gets smaller with increasing frequency ω = 2πf. Intuitively, the
higher frequencies are shorted to ground. If we build a voltage divider as shown on the
left of Figure 2.21, the output voltage Vout is given by Vout = (Vin /iωC)/(R + 1/iωC) =
Vin /(1+iωRC) and we see that it is attenuated with increasing frequency ω and, conversely,
the low frequencies are unaffected, hence the√name low-pass filter. The frequency where the
signal amplitude is attenuated by a factor 2 is given by ωc = 1/RC, and the imaginary
unit indicates that there is a phase shift between input and output voltage that depends
on the frequency. Equivalently, a low-pass filter can be constructed with an inductor with
impedance iωL and resistor R, but in many operational situations the inductances have
values that are difficult to find; therefore filters are usually constructed from resistors and
capacitors. The frequency dependence of the filter asymptotically behaves as the first inverse
power of ω, and the filter is called a single-pole filter. Cascading two such filters results in
two-pole low-pass filters that exhibit a steeper frequency-dependence of 1/ω 2 .
Swapping the resistor and capacitor in the low-pass filter results in the high-pass filter
shown on the right of Figure 2.21, which has a frequency dependence of iωRC/(1 + iωRC)
22  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.22 A simple band-pass (left) and band-stop (right) filter.

that attenuates low frequencies due to the factor ω in the numerator and approaches unity
as ω  ωc = 1/RC.
If we want to filter out everything except a small range of frequencies, we need a band-
pass filter that we can most easily construct by a combination of a low- and a high-pass
filter, as shown on the left of Figure 2.22. Here we have to ensure that the cutoff-frequency
1/R1 C1 of the initial low-pass filter is higher than that of the high-pass filter, given by
1/R2 C2 .
In case we need to reject a certain perturbing frequency, we implement a band-stop or
notch-filter, which is shown on the right of Figure 2.22. The combination of inductance L
and capacitance C has a resonance at the frequency ωc2 = 1/LC where their series resistance
vanishes, and a signal close to this frequency is shorted to ground and not passed on to the
output.
The filters discussed so far are passive filters that only depend on resistors, capacitors,
and inductors, and will only be able to attenuate unwanted frequencies. Sometimes, how-
ever, we need to combine amplification of a signal with filtering. The simplest solution is
placing an operational amplifier immediately after the filter. Alternatively, one may include
a capacitor in the feedback branch of the amplifier, as shown in Figure 2.23.
There is a huge amount of expertise in designing filters, both passive and active, docu-
mented in the literature. For a practical overview, see [17].

Figure 2.23 An active non-inverting (left) and inverting (right) low-pass filter.
 Sensors  23

R/2 R R R R R R R R/2
Vref
Comparators - - - - - - - -
+ + + + + + + +

Vin

Sample and
Hold Circuit MSB LSB A2
8-to-3 priority encoder A1
Clock A0

Figure 2.24 The operating principle of a 3-bit flash ADC.

2.2.4 Analog-to-digital conversion

Normally we want to convert analog signals to a digital format in order to process them
on a computer. The device that achieves this feat is an analog-to-digital converter (ADC),
which can be conceptually understood as a very rapidly measuring voltmeter. It consists
of a sample-and-hold circuit that holds the voltage constant for a short time while it is
measured and digitized. Several digitization methods are employed, and we will discuss a
few different types.
In a flash ADC the voltmeter is based on large number of comparators that compare
the voltage against a sequence of voltages derived from a resistor ladder. An additional
circuit encodes the output from the comparators into a binary representation. Since all
comparators operate in parallel, flash ADCs have the highest conversion rate, in excess of
109 samples per second. The high conversion rate comes at a price, though, because for a
resolution of n bits, 2n comparators are required. We illustrate the operating principle in
Figure 2.24 for a 3-bit flash ADC. On the left there are inputs for the reference voltage
Vref , the voltage to be converted Vin , and a clock signal. First, the input voltage is held
constant for the duration of the conversion in a sample-and-hold circuit, which consists of a
switch and a small holding capacitor. The voltages to compare to are produced by a resistive
voltage divider, shown on the top. The voltages from the divider are routed to the negative
input terminal of the comparators. All of the latter then compare the sampled-and-held
input voltage, simultaneously. The outputs of the comparators are connected to the inputs
of a priority encoder that translates this into a 3-bit binary representation at the output
pins A0 , A1 , and A2 . The conversion process and synchronous operation of sampling and
encoding is coordinated with an externally supplied clock signal. Flash ADCs are normally
available with resolution of 8 bits or less, and are relatively expensive. To sample at lower
rates, we can use less expensive ADCs such as those discussed in the following paragraph.
One of the very common types of ADCs, to sample a larger number of bits at a lower
rate and lower price are successive approximation ADCs. They replace the resistor ladder
and the large number of comparators with a single comparator and a digital-to-analog
converter (DAC) to dynamically adjust the reference voltage to compare to. We discuss
the operational principle of DACs in Section 3.13, but here we just mention that they are
circuits that translate a digital word with n bits to an analog voltage, and they can do this
very quickly. They play a central role in the operation of a successive approximation ADC
whose operating mechanism is illustrated in Figure 2.25, where the input voltage is held
constant during the conversion in a sample-and-hold circuit. Then the voltage is passed to a
comparator whose negative input terminal is determined by the output voltage of the DAC.
24  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Comparator
+
Vin - One bit
Control
logic at a time
S/H n bits
DAC
Vref

Clock

Figure 2.25 The operating principle of a successive approximation ADC.

In the first comparison, the n–bit input word is set to the 0x1000. . . which results in half the
reference voltage Vref , where we prepend ‘0x’ to identify the hexadecimal representation of
a number. If the input voltage Vin is smaller, the output of the comparator is at ground
level, or logically a low state. The first, most significant bit (MSB), is therefore “0”. The
control logic then sets the DAC to 0x0100. . . and compares again on the next clock cycle. If
Vin is larger than (1/4)Vref , the comparator output will go to high-voltage level, and the
next bit to pass to the output is therefore “1,” and the DAC receives the input 0x0110. . . to
compare the next step in the bisection sequence. If this is repeated n times the ADC clocks
out all bits of the conversion from MSB to least significant bit (LSB). We see that in this
case we trade conversion speed for a simpler hardware. Most ADCs we use later in this
book are of the successive approximation type. We note in passing that often a multiplexer
switch is placed before the sample-and-hold circuitry, which allows us to select different
input voltages with a single ADC shared among the different channels. Only one channel
can be converted at the highest conversion rate.
It is possible to find a compromise between conversion speed and hardware complexity by
combining, for example, two 4-bit flash ADCs, one additional DAC, and advanced control
logic to a so-called pipelined flash ADC. They can sustain a continuous conversion-rate
determined by the 4-bit ADC. Even though each conversion requires two clock cycles, the
first ADC can already start converting the next sample while the second ADC still works

Vref R
Comparator
-Vref R
Count the fraction
_ D-Flip-flop of the clock pulses
Vin R that the signal is high
OP-amp - during a given period
D Q
+ + _ An
Q Counter
and logic
A0

Clock

Figure 2.26 The operating principle of a delta-sigma ADC.

 Sensors  25

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

−0.2 −0.2

−0.4 −0.4

−0.6 −0.6

−0.8 −0.8

−1 −1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t/tsamp t/tsamp

Figure 2.27 Sampling a signal in different Nyquist zones. The dashed line displays
a frequency of 0.15fs and the solid lines show signals with frequency (1 − 0.15)fs
on the left and (1 + 0.15)fs on the right. Note that at the times when the signal is
sampled (indicated by boxes), the signals are indistinguishable.

on the four least significant bits of the previous sample. Most high-speed ADCs for radio-
frequency applications with more than 8 bits use this method.
Instead of a very high conversion rate, it is often desirable to achieve a higher resolution;
in other words, more bits, albeit at a low conversion rate. Devices that fulfill this requirement
are delta-sigma ADCs whose simplified operating principle is illustrated in Figure 2.26. It
is based on adding quantized current pulses to the negative input terminal of the op-amp
such that its output is forced to zero voltage. Note that this terminal is a virtual ground,
because the positive input terminal is grounded, and operational amplifiers always strive to
make their input voltages equal. The op-amp is configured as an integrator and sums the
currents over time. Forcing the accumulated charge on the capacitor to zero is achieved by
the feedback, shown as a dashed line from the non-inverting output of the flip-flop to the
switch that either adds a positive or negative current of magnitude Vref /R into the inverting
input terminal of the op-amp. The purpose of the flip-flop is to produce well-defined time-
steps, equal to the clock-frequency, for the injected current pulses. All we have to do in the
end is to count the number of clock cycles when the output is high and divide by the total
number of elapsed cycles. This results in a digitized word with the number of bits given
by the time we chose to average. And that time can be quite long. Assume that we use a
clock of 10 MHz and sample for 0.1 s such that 106 clock pulses happen, which results in a
resolution of 20 bits because 106 ≈ 220 . Delta-sigma ADCs derive their name because the
small quantized difference (delta) current are summed (sigma), together with the current
from the input voltage Vin , in the integrator. The high resolution makes delta-sigma ADC a
good choice to measure the often small voltages in Wheatstone bridges. We need to keep in
mind, however, that this increased resolution comes at the price of a rather low conversion
rate; often only a few tens of conversions per second are possible.
Despite having a large number of bits, ADCs add noise to the measurements, because
they cannot measure voltage differences smaller than that corresponding to the least sig-
nificant bit. Thus, they introduce a quantization error, which could be reduced by choosing
an ADC with a larger number of bits. But even then the quantization causes a slightly
inaccurate representation of the signal amplitudes.
Sampling the signal at discrete instances in time causes a second potential mis-
26  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Base-band
s1 S1
S2
s2

fs/2 fs 3fs/2

Figure 2.28The dashed signals s1 and s2 are the images aliased into the base band
of the original signals S1 and S2 in higher Nyquist zones.

representation, because signals of frequencies f higher than half the sampling frequency
fs /2, also called the Nyquist frequency, cannot be distinguished from signals with frequency
nfs − f or nfs + f. The origin of this ambiguity is illustrated in Figure 2.27. On the left we
display a sinusoidal signal with frequency f = 0.15fs (dashed) and f = (1 − 0.15)fs (solid),
and on the right the solid line is a sine with frequency f = (1 + 0.15)fs . We see that at
the sampling instances, indicated by the squares, the curves have the same values, which
makes them indistinguishable if only sampled at rate fs . In the frequency domain, shown
in Figure 2.28, we find that a signal S1 that lies between the Nyquist boundary and fs /2
and fs is recorded by the ADC, sampling at rate fs at frequency s1 , which is S1 mirrored
at the Nyquist boundary. The signal S2 that lies above f2 is observed as signal s2 . The zone
between zero and fs /2 is commonly called the base band or first Nyquist zone, and the
appearance of higher-frequency signals in the base band is called aliasing. Since the original
frequency of the aliased signals is unknown, they are often considered to be noise. A simple
way to prevent aliasing is to use analog low-pass filters with cutoff frequencies below the
Nyquist frequency before passing the signals to the ADC. Filters such as those discussed in
Section 2.2.3 are often sufficient.
Note that ADCs are often built into microcontrollers and sensors with digital interfaces,
such as those discussed in later sections, and usually have an ADC built in. After this
treatment of the ADC, the workhorse of digital data-acquisition systems, we need to look
at the task of providing power to our circuits.

2.2.5 Supply voltage

Of course, our sensors and also the microcontroller require electric power to operate, and
that is normally supplied by a power source. A common type of power supplies use a
transformer to step down the 220 V or 120 V AC voltage from the wall to a commonly
used voltage range of around 5 to 20 V. This depends on the rating and winding ratio of
primary to secondary winding of the transformer. Since most electronics circuits require DC
voltages, we need to rectify the AC voltage from the secondary winding of the transformer.
The simplest way to do this is to use a single rectifying diode with adequate power rating,
as shown on the left of Figure 2.29. The diode only lets voltages pass in one direction, and
effectively cuts off the negative half-cycle of the AC voltage. The ratio of the voltages before
and after the diode are shown on the plot below. The output in that case is very bumpy,
and can be smoothed somewhat by adding a large electrolytic capacitor with a few 100s of
 Sensors  27

Figure 2.29 Schematics of very simple power supply circuits.

µF. It is charged during the positive half-cycle and is discharged during the negative, where
it supplies charge to the powered circuit. The more current the load draws, the bigger the
capacitor needs to be. Note also that the power in the negative half-cycle is dissipated in
the diodes, which heats them up and may make a heat sink for the diodes necessary.
A better solution is to use a bridge-rectifier circuit made of four diodes with adequate
power rating, as shown on the right side of Figure 2.29. The diodes are placed in such a
way that during the positive half-cycle, one set of diodes conducts, and during the other
half-cycle, the other set of diodes conducts. Essentially the circuit works by transforming
the sine of the AC to the absolute value of the sine, as can be seen on the plot below the
circuit where we show the ratio of the voltages before and after the bridge rectifier. We still
need the smoothing capacitor, but for the same capacitance, we obtain much less ripple on
the output voltage. Note that many of the very common wall-plug power supplies contain
a circuit similar to the one shown on the right of Figure 2.29.
But the ripple of the output voltage can be reduced further by using a linear voltage
regulator based on a temperature-stabilized voltage source with a bandgap voltage reference.
In such a circuit, the known temperature gradient of the output voltage from a configuration
with two transistors of different size is balanced by the opposite temperature gradient
of suitably chosen resistors. We discussed a similar configuration in Section 2.1.2 when
discussing the LM35 temperature sensor. The result is a very stable output voltage on
the order of 1.25 V, near the bandgap of silicon. The commercially available linear voltage
regulators usually have three pins for input IN, output OUT, and adjustment ADJ, and they
have internal circuitry that forces the voltage on the output pin to the bandgap reference
voltage of 1.25 V. It is straightforward to use an external voltage divider with suitably
chosen ratio to select any output voltage up to some maximum current specified in the
datasheet. On the left of Figure 2.30, we show how a voltage divider with two resistors is
used to set the desired output voltage. The larger capacitors have capacitances of a few µF
28  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 2.30 Variable-voltage (left) and fixed-voltage (right) regulator circuits.

and guarantee stable operation of the circuit. The smaller capacitors of typically 100 nF are
used to absorb any high frequency glitches from the load. A standard example of such a
variable voltage regulator is the LM317.
For many standard operating voltages, such as 3.3, 5, 9, and 12 V, ready-made linear
voltage regulators are available. They do not require the voltage divider, and most have the
numbers 7803, 7805, 7809, or 7812 in their name. The last two digits specify the output
voltage. The same series with 79xx exist for negative output voltages. These small circuits
are very handy for obtaining well-regulated output voltages. Also, if a power supply already
gives 5 V, but 3.3 V are required for a sub-circuit, we use a 7803 or similar. Normally the
voltage regulators require about 2 V higher input voltage than their specified output voltage,
unless we choose a special low-dropout voltage (LDO) regulator, which only requires about
0.5 V overhead. The 3.3 V regulator MCP1700 is a member of this category.
Of course, we can also power an electronic circuit from batteries or rechargeable batteries.
Of special interest are lithium-ion or lithium-polymer rechargeable, because they have the
ability to provide very high currents to a circuit, which is especially important for devices
that require large currents, such as WLAN circuits or motors. The lithium-ion cells provide
voltages in multiples of nominally 3.7 V, and require special dis- and recharging circuitry,
because their high intrinsic energy density requires special care to prevent under- and over-
charging, as well as physically damaging the battery.
For electronic circuits that are used near a computer, we can use power drawn from the
USB port. Many RS-232-to-serial converter circuits provide up to about 500 mA, which is
often sufficient for small circuits. To make circuits very portable in the field, we can also use
solar cells and so-called supercapacitors as temporary power source. In the laboratory, we
have, of course, bench power supplies with adjustable output voltage, and current-limiting
circuitry which prevents excessive currents that might damage our electronics.
After discussing analog sensors, analog-to-digital-converters, and power supplies, we now
turn to sensors that directly produce digital signals.

2.3 DIGITAL SENSORS

We now address sensors that do not require an external ADC, but report their measurement
values directly to the microcontroller in digital from. Some sensors already have the ADC
built in, and others do not need one. We start with the latter, of which the most prominent
examples are buttons and switches.
 Sensors  29

Figure 2.31 Connecting a switch or button with a pull-up resistor.

2.3.1 Buttons and switches

The simplest digital sensor is certainly a switch that is either closed or open, or a button
where the open or close state is only activated temporarily. We use these terms interchange-
ably. Our task is to sense their state in a reliable way, and this is normally done with a
pull-up resistor that is connected to the supply voltage in the way shown in Figure 2.31.
In this way the sensing pin on the microcontroller can reliably detect the supply voltage.
Only if the switch S1 is pressed does the voltage on the pin drop to zero or ground level.
Only a small current, determined by the magnitude of the resistor, flows when the switch
is closed. The actual value of the resistor is uncritical, but values around 10 to 30 kΩ are
usually reasonable. Without the pull-up resistor, the voltage potential on the pin is unde-
fined when the switch is open, and determined by stray capacitances in the system. So,
the recommendation is to always use a pull-up resistor when sensing the state of a switch.
Note that swapping the position of the resistor and the switch, the resistor functions as a
pull-down resistor, and the sensing level is zero unless the switch is closed.
We need to point out that mechanical switches have the undesirable characteristic of
bouncing. Closing the switch is often accompanied by a fast on-off sequence. Fast micro-
controllers or other computers act so rapidly that they are easily fooled by sensing multiple
switch-closures instead of a single one. Sometimes a small time-delay is introduced after the
first switch closure is detected. In this way, only a single closing event is accounted for, and
others during the short timeout period are ignored. Optionally, an analog debouncer, which
is a simple low-pass filter, as discussed in Section 2.2.3, may be used.
There are a number of other sensors that act just like a switch and can be sensed in the
same way. An example is a reed switch, which is sensitive to magnetic fields and closes a
switch if some field level is exceeded. A further example is a tilt switch, which senses the
position of a small conducting sphere that horizontally rolls back and forth, and closes a
contact when it bumps into one of the extreme positions.
A variation on the theme of switches are rotary encoders, which use two switches, A and
B, that open and close periodically as an axis is turned. The direction of the rotation can
be detected, because each switch produces a regular pattern of on and off, but the timing
of the two switches is shifted by 1/4 of the period length. Essentially, one switch produces a
cosine-like pattern and the other a sine-like pattern. From knowing both, we can determine
30  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Level shifter circuitry using a n-type MOSFET. The source of the MOS-
Figure 2.32
FET is connected to the 3.3 V logic and the drain to the 5 V logic.

the direction of rotation by observing whether switch A leads switch B or vice versa. But
the fundamental sensing process is based on the same mechanism as shown in Figure 2.31.

2.3.2 On/off devices

A number of sensors provide a voltage level to inform the microcontroller about their state
or change of state. They can be thought of as a switch with a built-in pull-up resistor and
can be sensed in the same way.
A problem can occur if the operating voltage level of the sensor and the microcontroller
do not agree. Nowadays many sensors operate on levels of 2.5 to 3.3 V, and microcontrollers
on levels from 2.5 to 5 V. Sensing higher external voltages such as those used in cars (12 V)
or industrial control applications (24 or 48 V) requires some adjustment of the voltage level,
to prevent damaging either the sensor or the microcontroller. There are level-changing
chips available, such as the 74LVC245, but in many cases a simple voltage divider with
two resistors is adequate. Yet, it only works if there is signal flowing from the high to the
low-voltage side. In case a bidirectional signal flow is necessary, such as on the data line of
the I2C bus, the solution shown in Figure 2.32, based on an n-type MOSFET transistor,
is easy to implement. In the first case, when both logic signals are high, the MOSFET is
not conducting, because the voltage difference between gate and source is close to zero. In
the second case, if the 3.3 V logic is controlling and the signal is pulled low, the difference
between gate and source is positive and the MOSFET conducts, such that even the 5 V
logic level is pulled low. In the third case, when the 5 V logic is controlling and the 5 V logic
signal is pulled low, the built-in diode (visible in the schematics) conducts, and causes the
voltage of the source to drop to about 0.7 V. At this point the gate source voltage drop is
sufficiently large to fully cause the MOSFET to conduct, which also pulls the 3.3 V logic
level low.
But let us return to the sensors. A prominent device that reports its state through a
changing voltage level is a PIR proximity sensor, shown on the right of Figure 2.33. It
senses the change in the infrared radiation level, which announces the presence of living
beings. The sensors are based on collecting the incident infrared radiation with a Fresnel
lens, which is the dome visible in Figure 2.33, on a pyroelectric sensor. The lens is often
made of polyethylene, a material chosen for its low absorption of IR radiation. One side of
the sensor is consequently warmed up and expands, which causes buckling of the piezo- or
pyroelectric material, often a polymer film. Two effects contribute to the voltage between
the upper and lower plates. First, the buckling strains the material and causes a piezoelectric
voltage. Second, heat flows from the hot to the cold side and adds a pyroelectric voltage.
 Sensors  31

Infrared
radiation

Fresnel
lens

IR-absorbing Hot
layer
Pyroelectric U
material
Cold

Figure 2.33 Schematic view of a PIR sensor (left) and the hardware (right).

The schematic setup of the device is illustrated on the left of Figure 2.33. The tiny voltage
that is generated is subsequently amplified and exposed to the surrounding electronics on
an output pin that goes from the low to the high voltage level once it is triggered by the
presence of a person, and stays there for a programmable (usually by a small potentiometer)
amount of time.
Several sensors report their measurement value encoded as a voltage pulse, with the
duration of the pulse proportional to the value. We therefore need to measure the duration
of that pulse with adequate accuracy. One device that employs this mode is an HC-SR04
distance sensor, shown in Figure 2.34, which operates like a sonar. It emits a short ultrasonic
(40 kHz) sound burst and records the duration until the echo arrives, which is the round-trip
time ∆t of the sound burst. The distance L is given by this duration, and the speed of sound
(approximately v = 340 m/s) by L = ∆t/2v, or, in convenient units, L[cm] = 0.017∆t[µs].
A pin on the device goes high when the pulse is emitted, and returns to low once the echo
arrives or some specified timeout expires. Somewhat more advanced models using the same
method, but having a larger range, are the LV-EZx sensors. They support other modes of
reporting the distance as well, such as an analog voltage proportional to the duration, or
direct reporting as RS-232 signals, but more on that below.
Another sensor, often found in modern consumer electronics, detects the infrared sig-

Figure 2.34 HR-SR04 distance sensor.

32  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

nals from the remote control. It is not strictly a scientific sensor, but still an interesting
device that works remarkably reliably, because it rejects disturbing environmental effects
and changes the channel on the TV only when a button on the remote is pressed. Typically
they operate at a wavelength of 940 nm, which matches the light-emitting diode on the re-
mote, and they have a built-in optical band-pass filter that lets only that wavelength pass.
Then the signals are modulated by 38 kHz carrier frequency, which is demodulated on the
sensor, such that only the base-band signal is reported on the output pin.
Now we turn to sensors that directly report their measurement values in digital form,
and start by discussing I2C devices.

2.3.3 I2C devices

A large number of sensors have some logic built in and support a high-level communication
protocol. An example is the I2C protocol operating on the I2C bus. The physical connection
to devices supporting I2C only needs four wires: Ground, supply voltage Vcc, clock SCL, and
data SDA. The latter two require a pull-up resistor, which is often already included in the
microcontroller that also serves as the I2C busmaster to orchestrate the communication.
It configures the sensor, initiates a measurement, and then reads data from the sensor.
Physically, the communication is based on a synchronous serial protocol, where the data
line is sampled every time the clock line changes from a high level to a low level. The
protocol is standardized and we will not go into details, but mention that the I2C devices
and also I2C sensors have a number of registers internal that can be written to in order to
configure the sensor, or read from in order to retrieve sensor data. The communication is
entirely based on exchanging digital signals, and, as mentioned before, coordinated by the
busmaster which normally is a microcontroller. Several devices can share the same SDA and
SCL lines, because each device has its own address and responds only to those messages
intended for it by specifying the device address.
The BMP085, as well as the later versions BMP180 and BMP280, are examples of baro-
metric pressure sensors. They are based on measuring the strain caused by the deformation
of a membrane that separates an evacuated test-volume and the outside air pressure [12]
with a piezoresistive strain gauge, as illustrated in Figure 2.35. The entire assembly is di-
rectly built into a silicon substrate, and the strain gauge is created by suitable doping of
a small section. A temperature sensor is included because the semiconductor-based strain
gauge is temperature dependent. Further signal conditioning and processing circuitry is in-

Air pressure
Membrane Piezoresistive
strain gauge

Temperature
Evacuated volume sensor

Silicon substrate

Figure 2.35 Illustration of the operational principle of a barometric pressure sensor.

 Sensors  33

Figure 2.36 A BMP180 barometric pressure sensor and an HYT-221 humidity sensor.

cluded in the assembly as well, such that the device communicates with a microcontroller
via an I2C-bus on address 0x76 or 0x77. The measuring range is from 300 to 1100 hPa
(mbar) with a relative accuracy of ±0.12 hPa and an absolute accuracy of about 1 hPa.
The sensor also reports the temperature with a resolution of 0.1 o C or better, depending on
which part number is selected. A sensor mounted on a small breadboard is shown left in
Figure 2.36.
The dependence of the reported pressure on the signal from the primary sensor, the strain
gauge, is rather intricate and may vary from one device to another due to manufacturing
tolerances. Each device, like many other sensors as well, therefore needs to be calibrated by
exposing it to known conditions, here the pressure, and recording device-specific constants
that allow us to accurately determine the pressure from the primary measurements. In
the case of the BMP180, a built-in ADC reports a value related to the resistance of the
strain gauge measured in a bridge circuit. The calibration constants, determined during the
manufacturing and calibration process, are stored in memory on the chip. The datasheet
describes a detailed procedure to obtain the pressure based on the value reported from the
ADC and the calibration constants. We come back to this topic in Section 4.4.3 when we
connect the BMP180 pressure sensor to a microcontroller.
The HYT221 and HYT939 are sensors that measure the relative humidity in the range
from 0 to 100 %RH with a resolution of about 0.02 %RH. The operational principle of the
measurement is based on a capacitor with a dielectric, made of a polymer as the sensing
medium. The polymer is highly hygroscopic and easily absorbs water, which changes the
relative dielectric constant εr by a large amount, because εr of the dry material is much
smaller than that of water, which is about 80. Correspondingly, the capacitance C, being
proportional to εr , changes by a large amount as well. This change of the capacitance is
determined in a Wheatstone bridge with capacitors in two branches. For the calculation
of the relative humidity, knowledge of the temperature is required, and provided on board
by a temperature sensor. On board, the raw data are compensated for a number of non-
linearities, post-processed, and made available in registers that are accessible via an I2C
interface. The I2C address of the sensor is hard-wired to 0x28 and a device is shown on the
right in Figure 2.36.
The HMC5883 is a three-axis magnetic field sensor that is based on the change of the
resistance due to the magnetoresistive effect. The active material is a long meandering strip
34  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Outer
Sensing Electro- frame
capacitor static
motor
Springs

Fc
vx

Mass

Inner
x frame
z
y

Figure 2.37 The operational principle of one gyroscope in the MPU-6050.

of a nickel-iron alloy on a nonconducting substrate. The change of resistance is due to the

change of the spin–orbit coupling of valence electrons in the material. This affects the ease
with which the conduction electrons propagate through the material and consequently the
resistance. The active material is part of a Wheatstone bridge whose output voltage is con-
ditioned and amplified before being digitized with an on-board ADC and made available on
the I2C bus. The HMC5883L measures the field along three axes with a 12-bit (4192 steps)
resolution in the range of ±8 gauss or ±800 µT. It is typically used in compass applications
for mobile phones. The noise floor of the measurements is on the order of 0.2 µT and up to
160 measurements/s can be taken.
The MPU-6050 is a three-axis motion-detection chip consisting of an accelerometer and
gyroscope to measure acceleration and angular velocities in three spatial dimensions. The
accelerometer is based on the same sensing principle as the ADXL335 discussed previously
and illustrated in Figure 2.11, where a spring-suspended inertial mass changes its position
when accelerated and varies capacitances that are measured in a bridge circuit. In the
MPU-6050, however, digitizer and post-processing digital circuitry are added to provide
the measurement data in digital form. The operational principle of the gyroscope [12] on
the MPU-6050 is illustrated in Figure 2.37, where a small mass, suspended by springs on
an inner frame, is forced to oscillate in the x direction by a micromachined electrostatic
motor. The mass therefore has a velocity component vx in the same direction. If the entire
assembly rotates around the z axis, which points into the paper, a Coriolis force, given
by the crossproduct of the velocity vector and the rotation vector, will point along the
y axis. And this force moves the inner frame in the y direction, against the force of the
springs that connect the inner to the outer frame. This motion subsequently produces an
imbalance in the sensing capacitors, very much like what happens in the accelerometer,
and is sensed in much the same way as well. Also, here the analog voltages are digitized
on the chip, further processed, and made available on an I2C bus. The performance of the
 Sensors  35

Figure 2.38 An MPU-6050 accelerometer on a breadboard.

device is rather impressive. It measures the acceleration in various ranges between ±2 g and
±16 g with 16-bit resolution and a rate of up to 1 kHz. The rotation speed can be measured
between ±250o /s and ±2000o /s with a resolution of about 10 to 100o /s and a rate of up
to 8 kHz. The I2C address of the device is 0xA0 or 0xA1, depending on the state of IO-pin
AD0.
An enhanced version of the previous device is an MPU-9250 sensor that combines the
functionality of accelerometer and gyroscope with that of a three-axis AK8963 Hall sensor,
all internally aligned. The magnetic sensor has its own I2C address 0x0C, but shares the
same I2C pins with the accelerometer and gyroscope.
The sensor chips are usually very small and difficult to work with, but luckily there are
so-called breakout boards available that route the pins of the sensors to normally spaced
(2.5 mm spacing) pins that can be attached, for example, to solderless breadboards. Fig-
ure 2.38 shows an MPU-6050 mounted on a breakout board.

2.3.4 SPI devices

The SPI interface is a synchronous serial communication bus, similar to the I2C bus, but
it can operate at much higher speed and is therefore often used for devices that require
the continuous transfer of large amounts of data, such as displays or audio equipment. SPI
communication requires one master on the bus, a role normally taken by a microcontroller.
The sensors are typically slave devices. They need at least six wires to connect: ground
and supply voltage, the clock CLK, one line to send information from the master to the
slave (MOSI, for master-out slave-in), one line for the reverse direction (MISO for master-in
slave-out), and a chip-select line CS to identify the currently active slave. CLK, MISO, and
MOSI lines can be shared among many slaves, but each slave requires its individual CS line.
Some of the devices in the I2C section support this interface as well, and the interface
can normally be selected by setting a pin on the device high or low. Details can be found
in the datasheet.
In a later project, we will connect external analog-to-digital converters to a microcon-
troller via SPI communication, in particular, the MCP3304, which has 8 single-ended 12-bit
input channels but can also be configured to use two input channels as differential inputs.
This circuit detects whether one or the other input is larger and thereby provides an addi-
tional sign bit. Thus, it provides one extra bit to obtain a 13-bit resolution. At the same time,
36  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

A GPS receiver on the left and a DHT11 humidity sensor on a breadboard

Figure 2.39
on the right.

the input range is extended to plus or minus times the supply voltage, despite operating
from a unipolar power supply. This chip or its sibling, the MCP3208, are strong candidates
to expand the number and resolution of analog input channels for many microcontrollers.

2.3.5 RS-232 devices

Several devices report their measurement values by sending them via the asynchronous RS-
232 protocol. Originally, the physical medium for the communication channel used a current
loop, but nowadays most sensors operate on 3.3 or 5 V levels. The communication happens
point-to-point between two partners who have agreed on a communication speed, which is
often 9600 baud or 115200 baud. Three wires are required at the minimum, one for ground
potential, one, labeled TX for transmitting from device A to device B, and another one,
labeled RX, for transmitting in the reverse direction. To establish communication, cables
connect the TX pin on one device to RX on the other and vice-versa. Naturally, ground
pins need to be connected as well.
One device that supports RS-232 communication is the LV-EZx distance sensor we
already discussed earlier. It can be configured to send the distance measured as an ASCII
string that can be read in a terminal program.
Sensors that query the global positioning system (GPS) use a small patch antenna on the
sensor to pick up signals from a number of satellites placed in geostationary orbits, which
broadcast their position and timing information with high precision. On-board electronics
that normally comprise a microcontroller use triangulation in order to determine the position
of the sensor with high accuracy and convert that information to an ASCII string containing
the position in a standardized format, called NMEA. The string is written, typically once
per second, to an RS-232 serial line, where it is straightforward to read and decode.

2.3.6 Other sensors

Apart from the standardized protocols, there exist a number of communication standards
that device vendors come up with. Examples of such devices are the relative humidity
sensor, DHT22, and the DHT11; the latter is shown on the right in Figure 2.39. In order
to determine the relative humidity, the temperature must be known and it is determined
with a thermistor. To measure the humidity they use a capacitive humidity sensor. The
operational principle is based on using a small capacitor with an exposed dielectric that has
a large affinity to attract water, which changes the relative dielectric constant εr and the
capacitance by a large amount. This is determined in a Wheatstone bridge with capacitors
 Sensors  37

Figure 2.40A Shinyei PPD42NS particle sensor (left) and a GP2Y1010AU0F dust
sensor (right).

in two branches. Internally, the DHT sensors have this, and further circuitry such as an
analog-to-digital converter on board to calculate the relative humidity. They provide the
data using a non-standard, though documented, digital interface that we discuss further in
Section 4.4.5, where we cover connecting a DHT11 to a microcontroller.
The DS18b20 temperature sensor, as was the LM35 discussed earlier, is based on a
bandgap temperature sensor. Here, however, ancillary digital electronics and signal pro-
cessing circuitry is added on the chip such that the measurement value is postprocessed
and made available using the so-called Dallas 1-wire bus protocol. The 1-wire protocol uses
only ground and a single additional wire to transmit power and information to and from
the device.
The air quality can be characterized by the density of microscopic particles suspended in
air. In Figure 2.40 we show two such sensors. On the left we see a PPD42NS particle sensor.
Inside this detector a resistor heats the air, which causes the air with the suspended dust
particles to rise and pass through the light emitted by an infrared diode. There the dust
particles scatter the light onto a phototransistor, which pulls an output pin to low potential.
After signal conditioning and amplification, a cleaned-up signal is available. It is low when
particles scatter light, and high otherwise. The device is calibrated such that the ratio of
time at low signal to total time can be translated to particles per liter. The GP2Y1010AU0F,
shown on the right in Figure 2.40, works in a similar way. It also detects light scattered off
of dust particles, but it periodically turns the infrared diode on and integrates the signal
from the phototransistor and one has to sample the output value 0.28 ms after the LED was
turned on. The difference between the signal with LED on versus off provides reasonable
rejection of ambient light. The performance of both dust sensors can be improved if we
place them in the airstream created by a fan, which we need to turn on and off.
And that brings us to actuators, devices such as switches or motors that cause some
change in external conditions. Sometimes they are part of the measurement process, such
as the fan mentioned in the previous paragraph, or we need to move the sensor to where
we want to measure, which typically requires motors.
38  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

QUESTIONS AND PROJECT IDEAS

1. Where is the Fermi level in Figure 2.3?
2. Discuss different ways of measuring the capacitance of a capacitor.
3. Discuss different ways of measuring the inductance.

4. Research the possibility of how to use a pin diode as a detector for ionizing radiation.
5. Like pin diodes, LEDs generate a small current when illuminated. Use this effect to
build a photometer from a number of LEDs with different colors to sense different
parts of the optical spectrum.
6. Discuss a system to measure the rotation speed of a wheel with a small magnet and
a reed switch. Can you design it such a way that you can determine both speed and
direction of the rotation?
7. Discuss methods to determine the wind direction.
8. Discuss how to measure the amount rain during a day.

9. Research how to measure the concentration of sugar (dextrose) in water by observing

the polarization change as a function of amount of sugar dissolved. Which of the
described sensors can be used?
10. Build a discrete 3-bit flash ADC from a resistor ladder and external components, such
as the LF198 sample-and-hold circuit, two LM339 quad-comparators, and a 74HC147
8-to-3 priority encoder.
 CHAPTER 3

Actuators

Even though sensors are the main topic of this book, sometimes we need to turn devices on
and off, or we need to move a sensor very accurately, with much higher precision than we can
achieve by hand. In other cases, the sensor is not at the location where we need to measure
some quantity. In such situations we need an actuator to move it in a controlled way. Here
actuator is the generic term for a device that controls external parameters. Examples are
motors, valves, or switches, and we start the discussion with the latter.

3.1 SWITCHES
Turning an electric signal on and off is very easily done by toggling an output pin of a
microcontroller, as we shall discuss in quite some detail in the next chapter. A typical
microcontroller can provide rather limited currents on the order of a few mA. This is
normally sufficient to control a single light-emitting diode, an LED, which typically draws
less than 20 mA.

3.1.1 Light-emitting diodes and optocouplers

An LED, shown on the left in Figure 3.1, is similar to a conventional diode and consists of a
semiconductor with a pn junction. If it is forward biased, with the n-terminal connected to
the lower potential, the charge carriers, electrons and holes are pulled into the space-charge
zone, provided the voltage is higher than the voltage drop of the diode. This situation is
opposite to the situation encountered in Figure 2.13, where the diode was reverse biased.
The simplified band-level scheme and the corresponding circuit for the forward-biased LED
is shown in Figure 3.2. We only show the lower boundary of the conduction band, and
the upper boundary of the valence band inside the diode. Here both electrons and holes
are pulled into the space-charge zone and have a chance to recombine. The energy that is
released by the electrons dropping from the conduction band to the valence band is emitted
as a photon. Note that the forward voltage drop of the LED is about the size of the bandgap
and the resistor is needed to limit the current. LEDs emitting different colors have different
bandgaps, and correspondingly, different forward voltage drops, where shorter wavelengths
corresponds to larger bandgaps.
As already seen in Figures 3.1 and 3.2; LEDs connect to a circuit with two wires; the
short one is the cathode and needs to be connected to ground, or the more negative potential.
The cathode is normally indicated by a flattened face of the hemispherical housing. The
other wire is the anode and is connected to the more positive potential, provided we want
the LED to light up. But simply connecting the LED to the power rails will likely destroy

39
40  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

A close-up of a light-emitting diode is shown on the left. In the center is

Figure 3.1
the schematic of connecting an LED and the same circuit on a breadboard.

the diode. We need to connect a resistor in series with the diode, as shown in the center and
right on Figure 3.1. The voltage drop Vd across most LEDs is between 1.5 and 3 V, where
the lower values apply to red and infrared LED, and the higher value to blue and ultraviolet
LEDs. Moreover, the typical operating current of the diode must be limited to some value
below I = 20 mA. The resistor thus needs to be specified according to R > (Vc − Vd )/I,
where Vc is the supply voltage. A red LED thus works nicely on Vc = 5 V with R = 180 Ω
or 220 Ω, or even larger values where I picked a larger resistor with the commonly available
values. Normally the resistance value is not very critical; if chosen too large, the LED is less
bright.
In order to dynamically vary the brightness, changing the resistor value is rather in-
convenient. The better way to achieve this is by rapidly turning the diode on and off at
a rate much faster than the human eye can resolve, typically at a kHz rate. Changing the
duty cycle of the on- versus off-time proportionally changes the brightness of the LED. The
added benefit of this method, called pulse-width modulation, is that the dissipated energy
is reduced by the same ratio.
If an LED or any other pulse-width modulated device needs to be galvanically separated
from the control electronics, we use an optocoupler. This is a small integrated circuit that
consists of an LED and a phototransistor. As discussed in Section 2.1.3, a phototransistor
behaves like a normal transistor with a high current amplification, but instead of passing
a current into the base terminal of the transistor, the LED illuminates the region with
depleted charges in the transistor and creates a large number of electron–hole pairs. Thus,
the transistor becomes conducting and switches on a device on the “other” side of the
circuit, without being electrically connected. This implies that only the on- or off-state or
digital signals are communicated across the optocoupler. Typically they are used if devices
located at different electrical potentials need to be switched on or off, or to prevent electric
perturbations from the “other” side. We note in passing that the MIDI communication
between musical instruments uses optocouplers at their respective inputs. This removes
electrical disturbances and protects the musicians from electrocution.
So far we only switched LEDs on and off, but we will look at how to control large
currents and loads.
 Actuators  41

Energy
Forward voltage
0 drop of LED

Photon

Band
Supply gap
voltage
Electrons and
holes are pulled
into the space-
charge zone and
Lower boundary recombine
of conduction band
Vcc
p n

Resistor LED

Vcc

Figure 3.2 The physics of an LED.

3.1.2 Large currents

Switching large currents and voltages requires us to “amplify” the small current that the
microcontroller provides, and transistors of various flavors do just that. Namely, they am-
plify the current flowing into their base terminal by a factor β, which is specified in the
datasheet of the transistor to a value that flows across the collector-emitter terminals. Refer
to the left side of Figure 3.3 for the naming and location of the respective terminals of an
NPN-transistor. If the amplification β is large enough, moderate base-currents cause the
output current to saturate the transistor such that it behaves like a switch that turns on
the connection from collector to emitter.
In the example shown in Figure 3.3, the 1 kΩ resistor on the base limits the input current
across the base-emitter link to 5 mA, provided the control voltage is 5 V. If the transistor
has an amplification of 100, which is typical for many small signal transistors, it can switch
up to 500 mA, which is more than is actually needed in this case. The actual current flowing
through the collector-emitter link is determined by the 12 V supply voltage to the LED and
the 680 Ω resistor limiting the current. Note that using the transistor also decouples the
voltage level of the controlling circuitry and the supply voltage, here 12 V, of the consumer,
the LED. This way of operating a transistor is called open collector, because we can think
of the collector terminal as a generic connection point for an (almost) arbitrary load that
is connected to its own supply voltage. If the emitter-collector link is in the nonconducting
state, no current flows, and the load is turned off. If a positive current flows into the base
terminal, the emitter-collector link becomes conducting, the lower terminal of the load is
shorted to ground, and a current flows from the load’s power supply through the load and
turns it on.
Since switching is very common, small switching currents are often desirable. This re-
quires a transistor with a large amplification. One way to achieve this is to connect two
transistors as a Darlington pair, as shown on the left in Figure 3.4. Such a pair has the
current amplification of approximately the product of the two individual transistors, and is
often used in switching applications, which accounts for the availability of integrated circuits
that pack seven or eight Darlington pairs into a single package. An example, the ULN2003,
is shown on the right in Figure 3.4. The seven input terminals on the left-hand side of the
42  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 3.3The terminals of an NPN transistor (left), and using an NPN transistor
as switch (right).

package are the base terminals of the Darlington pairs, and the output terminals on the
right-hand side are the corresponding open collector terminals. The ground connection is
located at the lower left, and the external positive power supply voltage is connected to
the terminal labeled COM. This specific chip can switch up to seven times 500 mA, and the
supply voltage for the load can be up to 40 V. The datasheets provide a lot more detailed
information.
In case we need to switch very high voltages of up to 1 kV or very large currents, we use
MOSFETs. They require almost no current to flow into the gate in order to switch. Only
the capacitance between the terminals needs to be charged, and this creates an electric
field that pulls charge carriers into the depletion zone of the MOSFET, which causes it to
conduct.
Using different flavors of transistors makes switching unipolar voltages very convenient,
but if we need to switch AC household voltages on or off, such as lamps or the wake-up
radio, we need a relay. Relays consist of a small electromagnet that mechanically closes or
opens a contact. This achieves a high degree of separation between the controlling circuitry
and the load circuit. The two sides only communicate via the magnetic field that toggles

Two NPN transistors connected to form a Darlington pair (left) and a

Figure 3.4
ULN2003 Darlington array (right).
 Actuators  43

Figure 3.5Schematic illustrating the functionality of a relay (left) and an image of

a relay (right).

the switch, depending on whether current flows through the coil or not. A schematic image
is shown on the left in Figure 3.5. There we see the coil with its two terminals on the left,
and the switch on the right-hand side that toggles between the two terminals. Normally
we need to prepend a transistor to switch the coil on and off because this requires a larger
current than the microcontroller provides. Moreover, we need to pay attention, because
the coil is an inductive load, and turning it off causes a large induction voltage that may
damage the transistor or other parts of the circuit. This can easily be prevented with a
flyback diode, also shown in Figure 3.5, bypassing the coil in the relay in the normally non-
conducting polarity with cathode pointing towards the positive supply voltage. Any voltage
in the reverse direction passes through the diode, rather than the transistor. Care needs
to be taken with respect to power rating of the diode, and the speed. Normally Schottky
diodes, which are particularly fast-switching, are used.
The terminals on the high-power side can handle both AC and DC voltages. The relay is
very similar to a normal switch in the way that it establishes an electrical contact between
two terminals. In Figure 3.5, it closes or opens the contact between the AC supply and a
lamp, but the lamp can be replaced by any other device that needs to be turned on and off.
Please note that the relay in the image can actually handle up to 240 V AC, but we need
to stress that this is a potentially lethal voltage. This should only be handled by suitably
trained and qualified personnel. If in doubt, do not try it yourself!
Now we have several means at our disposal to switch devices and actuators on and off.
Some of the most prominent and useful actuators are motors, and that is what we discuss
next.

3.2 MOTORS
Motors in general translate chemical or electrical energy into mechanical energy, normally
by rotating an axle in order to provide torque. Here, and in subsequent sections, we address
electrical motors only. They come in a wide variety of types and can be powered by both AC
and DC voltages. One of the most common types, and the one we will use, is a conventional
DC motor with a commutator.
44  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Part of Stator

Rotor Commutator
F

F F I Brush

I B

N S B
Commutator F
Brush
N S
I

Figure 3.6 Basic operation principle of a DC motor. The magnetic north pole of the
stator is to the left and the south pole to the right of the coil. Current is supplied
to the brushes where it enters the commutator and passes through the rotor-coil,
where the Lorentz force causes a force on the wire, moving it upwards. After half a
turn, the commutator has rotated and reverses the polarity such that the current
on the left side of the coil points in the same direction as before.

3.2.1 DC motors
The operating principle of a DC motor is most easily explained with the help of Figure 3.6. It
is based on static magnetic field that is stationary in space, the stator, and one or more coils,
electrically excited by an external voltage source. They are forced to rotate by reversing
the polarity of the exciting voltage synchronously with the rotation. The rotating part is
called the rotor and the synchronous switch is the commutator. The torque on the rotor
is provided by the Lorentz force that the electrons of the electric current experience in
the external magnetic field. After half a turn, the external magnetic field has the “wrong”
polarity to continue forcing the coil to rotate, and would brake the rotation. This is prevented
by reversing the polarity of the current flowing at just that point in time which allows the
rotation to continue. The polarity reversal is effected by the commutator, which is shown
as the two D-shaped objects that rotate between externally fixed electrodes, called brushes.
Moreover, reversing the supply voltage causes the motor to rotate in the opposite direction.
The sliding contact between the brushes and the commutator sometimes causes sparking
at the moment of polarity reversal, because the brushes briefly connect the two D-shaped
objects and thereby short circuit the supply voltage. The sparks then cause electrical distur-
bances that may be partially alleviated by connecting the brushes with a small, say 100 nF,
capacitor. This also reduces the creation of ozone with its distinct smell, which often ac-
companies the operation of DC motors. In commercially available motors these effects are
minimized, and other motor designs are used that circumvent these deficiencies. For exam-
ple, in some brushless DC motors the magnetic field of the rotor is provided by permanent
magnets whose position is continuously monitored with a Hall sensor. This information is
used in an electronic control unit to periodically excite the current through the coils of the
stator in order to maintain the rotation. The complexity of the control unit places these
motors outside the scope of this book, and we use normal DC motors for our projects despite
their shortcomings.
We can adjust the torque and the speed of the motor by varying voltage applied to the
coils. A higher voltage makes the motor turn faster, and reasonable voltages are determined
by the resistance of the coils. Often, however, it is advantageous to limit the voltages and
 Actuators  45

V+

A C

DC-motor
B D

Figure 3.7 Schematic illustrating the functionality of an H bridge.

control the currents that flow through the coils instead. Moreover, since the rotor has a
finite inertia due to its mass, we can use pulse-width modulation to control the total power
delivered to the coils in order to adjust torque and speed. In the fleeting moments when no
current flows, the inertia maintains the rotation.
In case you have a model railway, you may know that in the early days the speed of the
trains was regulated by a variable transformer, and it was very difficult to run the trains at
very slow speeds with voltages too low to reliably excite the coils. Modern model-railway
controllers use pulse-width modulation to regulate the speed. In that way, the full supply
voltage and current are always flowing, which reliably excites the coil, just not all the time.
Operating the trains at slow speed now works much more reliably.
In this context we have to keep in mind that turning the coils rapidly on and off also
rapidly changes the magnetic field, and this causes a voltage, the backwards electromagnetic
force, or back-emf that opposes the driving voltage. Thus, the faster the motor turns, the
more the back-emf reduces the voltage, and thereby also the current flowing through the
coils. This results in a reduced torque and for a given coil resistance we can either run at a
slower speed or increase the driving voltage in order to maintain the required torque.
As with the model trains, we also want to run the motor both forward and backward.
Therefore we need to control the polarity of the supply voltage. Instead of reversing the
supply cables by hand, which is not a very convenient option, we use an H bridge whose
functionality is easily explained with the help of Figure 3.7. In the center of the figure a
DC motor is located, with its supply leads extending to either side. The unipolar supply
voltage is connected to the upper terminal labeled V+ and to ground. Depending on the
position of the switches A, B, C, and D, the current flows through the motor. In case A
and D are engaged, it flows from left to right, This is indicated by the dotted lines and the
sense of rotation indicated by the small arrow in the motor. In case B and C are engaged,
while A and D are open, the current flows in the opposite direction, causing the motor to
turn in the opposite direction as well. Thus, by suitably toggling the four switches, we can
adjust the sense of rotation of the motor at will. We only have to ensure that only one of
the switches A and B is engaged at a given instance, because otherwise the supply voltage
is short circuited.
Of course, we can replace the switches with transistors and arrive at a system that is
46  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 3.8 A small model-servo.

easily controllable by a microcontroller. And the situation is made even simpler, because
ready-made integrated circuits that implement H bridges are available. The L293D is one of
them, and we will use it in later chapters to control motors. Finally, we add speed control by
pulse-width modulation. We either add a switching transistor between the upper terminal
and the supply voltage, or we directly modulate the input terminals that control the switches
A, B, C, and D.
At this point we can adjust the speed and direction of a motor, which is essential for
moving from one place to another, and the velocity at which this happens. On the other
hand, if we only want to change the position or angle by a small amount, such as the rudder
of a boat, we need means to directly adjust the position, rather than the velocity. Next we
discuss two types of systems that do this, servomotors and stepper motors.

3.2.2 Servomotors and model-servos

The term servo refers to using the motor in conjunction with a position encoder in a closed-
loop feedback, or servo, loop [18]. Here the motor speed is continuously adjusted to reduce
the difference between the desired position and the actual position, as reported by the
encoder. This type of servomotor is often used in industrial applications, and requires an
elaborate control system with a PID controller [18] and powerdrive-electronics. Servomotors
are used in large industrial robots and applications that require high accuracy of positioning.
Matching the parameters of the controller to the desired performance of the control loop in
terms of accuracy, speed, and acceptable overshoot requires expert attention, and is beyond
the scope of this book.
Instead, we use the modest cousin of the servomotor, the model-servo, often simply
referred to as a servo, shown in Figure 3.8. It is often encountered in radio-controlled cars
and planes, but can also appear in robotic applications, such as controlling the position or
angle of smaller robotic arms. Servos have a potentiometer mounted on the rotating shaft
that is used as an encoder. It provides information about the position of the shaft, which
typically has a rotation range of 0 to 180 degrees. The resistance is compared to a voltage
derived from an electric signal corresponding to the desired position. The difference signal
is amplified and used to turn a DC motor in the direction that minimizes the difference.
 Actuators  47

This constitutes a simple proportional controller. Normally a small gearbox between the
motor axis and the shaft reduces the speed, but increases the precision of control and the
torque. Note that the servo connects with three wires to the outside. Two of the wires are
ground and supply voltage, nominally 4.8 V. They are colored black and red, respectively.
The third wire, yellow or white in many cases, carries the information about the desired
position.
The information about the desired position is communicated to the servo as a pulse-
width modulated signal, as shown in Figure 3.9. The servo expects a train of pulses with a
spacing of 20 ms. Each individual pulse has a duration between 1 and 2 ms, with a 1.5 ms
duration specifying the mid-position. Producing such pulse patterns is easily done with a
microcontroller, and we will do that in later chapters.
Servomotors and servos use a closed-loop system to achieve a high accuracy of position-
ing, while stepper motors achieve this without closed-loop feedback.

3.2.3 Stepper motors

Stepper motors change the angular position of a shaft in small discrete steps, such that
counting the steps gives the position. In this way, no feedback or servomechanism is needed
to achieve high repeatability, and the motor can be operated open loop.
In Figure 3.10, we illustrate how this is accomplished in a permanent-magnet stepper
motor. The static part of the motor consists of an iron yoke and four coils, of which two each
are operated in series. Here the upper and lower coils, labeled A and B, and the right and
left coils, labeled C and D, are two such pairs. On the rotor a moderately large number of
permanent magnets are assembled. In the figure we only use six magnets, labeled 1 through
6, with the black end indicating the north pole. They are oriented radially and their polarity
alternates.
We assume that initially only coils A and B are excited in such a way that the upper
coil behaves as a magnetic north pole and the lower as a south pole. In that case, the
magnetic field lines pass through both upper and lower coils, pass through the magnets in
the rotor, and return through the surrounding yoke. With this coil excitation the rotor is
indeed oriented as shown in Figure 3.10, because the upper coil behaves as a north pole
and attracts the south pole of the upper permanent magnet, labeled 1. Consequently, they
are as close as possible. The converse is true for the bottom coil and the magnet labeled 4.
In order to turn the rotor by 30 degrees in the counterclockwise direction, we have to turn
coils A and B off and coils C and D on. Thus, coil C on the right-hand side is a north pole
that attracts the permanent magnet labeled 3, which has a south pole pointing outwards.
Again, the converse is true for coil D on the left and magnet 6. The rotor will then move
until magnet 6 faces coil D. We continue turning the rotor by turning coils C and D off and
reversing the polarity of coils A and B, such that the upper coil acts as a south pole that

V 1 - 2 ms

20 ms t

Figure 3.9 The timing of the control signal for the servo.
48  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

A yoke
D 1 C
6 c
2
5
d
4 3

Figure 3.10Schematics of a stepper motor. Note that the coils are denoted by upper-
case letters and the corresponding terminals with lower-case letters.

attracts magnet 2. Continuing in this way, we find that we can turn the motor in a number
of steps of 30 degrees each by periodically exciting the coils in the following pattern:

terminal a: 1000 1000 1...

b: 0010 0010 0...
c: 0100 0100 0...
d: 0001 0001 0...
where time runs from left to right. Here, 1 indicates that the terminal is connected to the
positive power supply voltage, and 0 indicates that the coil is connected to ground potential.
Stepping through the sequence backwards causes the motor to turn in the other direction.
Observing that time slots 1 and 3 show reversed polarity for the coils A and B implies that we
need an H bridge to implement the polarity reversal. We will return to this implementation
in Section 4.5.
It is instructive to analyze the excitation pattern of the four terminals a through d a
little more closely. The first observation is that a sequence of four time slots repeats itself
over and over again. Second, we observe that a fundamental pattern is the excitation of
terminal a with the sequence 1000. The excitation of terminal b is shifted to the right by
two time slots which we may interpret as 180 degrees out of phase if we assume that 360
degrees corresponds to the four time slots. Likewise, terminal c and d are 90 and 270 degrees
out of phase with the excitation of terminal a. This interpretation will prove useful later
when we write our own stepper motor driver and discuss other modes of operating the
motor, such as half- and microstepping modes.
Note that in the excitation pattern only one coil is excited at a time. But we can also
operate the motor with two coils excited simultaneously, one coil pulling one permanent
magnet and the other coil pushing another magnet. In this way the torque of the motor can
be increased by about 40 %, albeit at the expense of doubling the required power, because
there are two coils excited at the same time. The pattern with increased torque is based
 Actuators  49

on the fundamental sequence 1100. Applying the appropriate phasing between the four
terminals from above, we arrive at the following pattern to excite the terminals:
terminal a: 1100 1100 1...
b: 0011 0011 0...
c: 0110 0110 0...
d: 1001 1001 1...
where we see that the excitation of terminals b through d is shifted by 2, 1, and 3 time
slots with respect to that of terminal a. A special point to note is that exciting two coils
simultaneously will cause the positions towards which the permanent magnet will point to
lie in between the coils, rather than directly facing the coils. Since we now have the single-
coil excitation pattern where the magnets point at the coils and the double-coil excitation
pattern where it points in between the coils, we might consider a combination of both
patterns. . .
. . . and find that by interleaving the steps of the single-coil and double-coil excitation
patterns, we obtain a sequence where the step size is halved. This mode of operation is
called half-step mode. The pattern in which we have to excite the coils is the following:
terminal a: 11100000 11100000 1...
b: 00001110 00001110 0...
c: 00111000 00111000 0...
d: 10000011 10000011 1...
where the period length is 8 time slots with a fundamental excitation pattern of 11100000.
The excitation sequence of terminal b is the fundamental shifted by half the period length,
or four time slots, or 180 degrees. Note that here, 360 degrees, correspond to the 8 time slots.
And the sequence for terminals c and d are shifted by two and six time slots, respectively.
Note the general pattern: terminals a and b are excited by a cosine-like sequence, which
is evident by calculating the voltage applied to coils A and B, which is a-b
a - b = 1 1 1 0 -1 -1 -1 0
which admittedly is a poor approximation of a cosine. Likewise the voltage applied to coils
C and D is c-d
c - d = -1 0 1 1 1 0 -1 -1
and that is an equally poor approximation of a sine. Apparently, the signals applied to the
coil pairs roughly follow a cosine and a sine-like excitation. It is easy to see how we can
generalize this and use better approximations of the cosine- and sine-like excitations of the
coils, which is called microstepping the motors.
In microstepping modes, the permanent magnets on the rotor are moved to several
intermediate positions in between coils. This finer control comes at the expense of requiring
a more advanced driver to control the output current in finer steps than turning it on and
off. An example of such a driver is the DRV8825 circuit, which exposes very few control pins
for direction and stepping as well as selecting which microstepping mode to use. Internally
it generates the appropriate excitation pattern and the currents that are applied to the
coils.
In the above example, the stepper motor uses only six permanent magnets, resulting in
an angular increment of 30 degrees per step, or 12 steps per revolution, but it is easy to see
that we can increase the number of magnets to reduce the step size significantly. Typical
step sizes for commercially available motors are 200 or 400 steps per revolution.
50  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 3.11Wiring of the coils in a bipolar stepper motor (left) where the respec-
tive center terminals are often connected. In a unipolar motor (right) the center
terminals are exposed.

Here we need to point out that assigning one phase to coil terminals a and b and the
other to terminals c and d is only a convention, and some vendors prefer to enumerate the
coils and associated terminals sequentially when looking at the motor in Figure 3.10. This
amounts to swapping the labels of terminals b and c and implies that we may have to swap
two cables from the driver circuit to the motor in order to make it turn. It is best to check
the datasheet for the motor to find out what convention is used for that particular motor.
Furthermore, stepper motors come with different numbers of connecting cables that
expose the terminals of the coils. On the left in Figure 3.11, the situation is depicted where
all eight terminals are exposed. Sometimes the center taps are internally connected, in which
case only four terminals are exposed, and this corresponds to the situation in the discussion
earlier in this section. On the right, the center terminals are internally connected and are
exposed, in which case the total number of exposed wires is six. If the two central connections
are internally joined, five wires are exposed. Normally the exposed wires are color coded
and identified in the datasheet, but measuring the resistance between the exposed wires
allows us to determine the internal wiring of the motor experimentally.
In the previous stepper motor examples, we always reverse the polarity of the coil exci-
tation, which requires a drive circuit with an H bridge. Stepper motors, however, can also
be operated without an H bridge in unipolar mode. In Figure 3.12 we show the connection
for a unipolar stepper motor that has center taps of the coils wired together, as discussed
in the previous paragraph, and exposes only one terminal that is connected to the positive
voltage of the motor power supply. The other terminals of the coils are the collector of an
NPN or Darlington transistor and are bypassed with flyback diodes. In a ULN2003 circuit,
the diodes are already built in. Placing a positive voltage to one of the base terminals a,
b, c, or d causes the corresponding transistor to conduct and the coil to be excited.
For unipolar stepper motors, we can use the same excitation pattern we use for bipolar
motors. For example, the excitation pattern for the full-step mode with larger torque will
also turn a unipolar stepper motor in one way
terminal a: 1100 1100 1...
b: 0011 0011 0...
c: 0110 0110 0...
d: 1001 1001 1...
Reversing the order causes the motor to turn the other way around. Half- and microstepping
modes can be implemented in the same way.
All devices above were based on switching voltages or currents fully on and fully off,
 Actuators  51

Figure 3.12 Connection for a unipolar stepper motor.

but occasionally, careful adjustment of a control voltage is required; for example, to set the
control input of a power supply that determines its output current.

3.3 ANALOG VOLTAGES

A simple way to produce a constant voltage is to low-pass filter a pulse-width modulated
output voltage. The filter reduces the on–off variation and results in an average voltage level
corresponding to the peak voltage times the duty factor of the pulse-width modulation.
Varying the duty factor will produce a time-varying output voltage. There are, however,
limitations to this method, mainly due to the quality of the output voltage, which is likely
to have some ripple left over from the original modulation.
A more reliable way is to use a digital-to-analog converter (DAC), which is a device that
converts a digital word with a given length, such as 8 bits or 12 bits, to a discrete voltage
with 256 or 4096 intermediate levels, respectively. The operational principle is illustrated in
Figure 3.13 for a 4-bit DAC using an R − 2R resistor network and an operational amplifier.

V3 R V2 R V1 R V0

2R 2R 2R 2R 2R
Vref R
bit3 bit2 bit1 bit0
0 1 0 1 0 1 0 1 _

+ Vout

Figure 3.13 Operational principle of a digital-to-analog converter.

52  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

The switches labeled bit3 to bit0 represent the digital input word, and the inverting input
of the operational amplifier sums the currents that are added by toggling the respective
switches. The inverting input is a virtual ground because the op-amp forces it to be the
same potential as the noninverting input. This implies that the current flowing through
all resistors is constant, and independent of the position of the switches; if the switch is
in position 0, the current flows to the real ground, and if it is in position 1, it flows into
the virtual ground. This implies that the voltages V3 , . . . , V0 are constant and are given by
Vk−1 = Vk /2, which is easy to see for V1 and V0 . The two right-most resistors with values
2R in parallel combine to a single resistor of value R. But this means that the node with V0
is sandwiched between two resistors of value R, of which one is connected to a point with
voltage V1 , thus V0 = V1 /2. Since all the voltages Vn are fixed the current flowing into the
inverting input of the operational amplifier, provided that bit number k is set, is given by
Vk /2R, with Vk = Vref /23−k . And all currents coming from the four branches add up, and
the operational amplifier converts the current to an output voltage Vout .
Most DACs use the R − 2R ladder resistor network, with a current-adding operational
amplifier but add a digital front end with semiconductor switches that are controlled from
a parallel bus, I2C, or SPI. An example of the latter is the MCP4921, a 12-bit DAC that is
controlled via a SPI-compatible interface.
The analog voltages, switches, and motors we discussed so far are likely the most-used
actuators, but there are others, and we consider a selection of them in the following section.

3.4 OTHER ACTUATORS

In order to open doors or to move devices between well-defined end stops, solenoid magnets
are used. They are magnetic coils with a soft-iron core that are switched on and off. They
are conceptually not much different from an LED, except that the current required is much
higher, and that requires a larger power supply. Moreover, a solenoid is a large inductive
load, and turning it off causes a large back-emf that requires a large flyback protection
diode, just as we discussed in the case of relays and DC motors in Section 3.1.2.
The flow of liquids and gases can be controlled with pumps and valves. Pumps are usually
based on motors driving paddle wheels, propellers, or turbines that move the liquid or gas
from a region of low pressure to that of high pressure. The motors are controlled in the same
way as discussed before. Flow rates are controlled by valves, which constrict the aperture
through which the medium flows. An example is a butterfly valve, shown in Figure 3.14.
The aperture can be restricted by the shaded aperture, which has the same diameter as
the inner diameter of the pipe. Rotating the shaded aperture then adjusts the flow rate
from fully closed, in which case the shaded aperture covers the entire inner diameter of the
pipe, to fully open, in which case it stands perpendicular to the flow of the liquid or gas. To
change its state we can either use a servo for smaller systems, a stepper motor, or a large
servomotor, in case we need to constrict the flow in an oil pipeline.
A common means for controlling large machinery uses hydraulics, based on controlling
the flow of hydraulic oil. It uses pumps to move the oil into cylinders that cause a linear
travel, or it is pumped through a turbine or propeller-based device to cause rotary motions.
In either case the original pressure is provided by pumps, based on electric motors. The
position of valves is also controlled by the same devices we discussed earlier in this chapter.
And finally, we may consider humans as part of the larger system, and they need to be
actuated. For this purpose we may use various attention-grabbing devices, such as piezoelec-
tric buzzers, sirens, and loudspeakers, as well as optical devices comprising LEDs, neo-pixel
 Actuators  53

Figure 3.14 Sketch of a butterfly valve.

color LEDs, liquid crystal displays, or small displays such as those found on printers or
washing machines that report the current state of the apparatus.
By now we have covered quite a selection of actuators and sensors and can turn things
on, move them around, and measure a number of physical quantities. The next task is to
interface this variety of sensors and actuators with their multitude of interfaces, and provide
a standardized interface to communicate with the external world. This is the task of the
microcontroller, and the flavor we will discuss in the following chapter is the Arduino.

QUESTIONS AND PROJECT IDEAS

1. What is the purpose of a fly-back diode? When do you need it?
2. The TV images are badly disturbed as soon as you turn on a DC motor. What is the
cause and how can you alleviate the problem?
3. How do you control the speed of a DC motor if you need to run it at low speed?
4. When do you need to use an H bridge driver to control a DC motor?
5. Can you use a Darlington driver, such as the ULN2003, to control a DC motor with
pulse-width modulation?
6. When would you use a DC motor, and when is a model-servo the better option?
Discuss!
7. Use a multimeter to determine the resistance between the wires of a stepper motor
and identify the internal wiring.
8. Discuss the pros and cons of using unipolar versus bipolar stepper motors.
9. Discuss microstepping and draw the currents in the coils as a function of time.
10. We plan to control a device by low-pass filtering a pulse-width modulated signal. How
does the ripple of the control voltage depend on the input impedance of the device?
How does the capacitor enter the discussion? What happens if we plan to apply fast
changes? Discuss!
54  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

11. Investigate what a thyristor is and where it is used.

12. Investigate what a triac is and where it is used.

13. Investigate what an IGBT is and where it is used.

14. Design a small crane to lift cargo from a model boat onto a quay. Which actuators
and which sensors do you need? Motivate their use.
15. Design a mechanism that wakes you up with almost certainty. Which actuators do
you use and why?
 CHAPTER 4

Microcontroller: Arduino

The microcontrollers we select for this book come from the Arduino family, because they
are supported by an easy-to-use programming environment that allows us to quickly start
developing software. There are limitations, but it is a wonderful system with which to start
learning about microcontrollers. Here we need to point out that by “Arduino” we mainly
mean the development ecosystem, rather than the processor itself, because that has evolved
and comprises a number of different hardware incarnations. In this chapter we will discuss
a few of them.

4.1 HARDWARE
The original Arduinos are based on Atmel microcontrollers and we mostly discuss the Ar-
duino UNO. Support for a second family of controllers, based on the ESP8266 microcon-
troller, was recently integrated into the Arduino development environment. These controllers
can be programmed in much the same way as UNOs, but have native wireless support built
in. But let’s start with the UNOs.

4.1.1 Arduino UNO

An Arduino UNO is shown in Figure 4.1, where the main component is the ATmega328p
microcontroller from Atmel (now MicrochipTM ), which is the large chip with 28 legs in
the image. It is a controller with 8-bit-wide registers, and operates at a clock frequency
of 16 MHz. It has 32 kB RAM memory and 1 kB non-volatile EEPROM memory, which
can be used to store persistent data that need to survive tuning off and on the supply
voltage. There are three timers on board, which are basically counters that count clock
cycles and are programmable to perform some action, once a counter reaches some value.
The UNO interacts with its environment through 13 digital input–output (IO) pins, of
which most can be configured to be either input or output, and have software-configurable
pull-up resistors. Several of the pins are configurable to support I2C, SPI, and RS-232
communication. Moreover, there are six analog input pins. They measure voltages of up
to the supply voltage of 5 V. An alternative internal reference voltage source provides a
1.1 V reference. All digital and analog pins are routed to pin headers that are visible on
both sides of the Arduino printed circuit board (PCB) in Figure 4.1. Furthermore, the
built-in hardware RS-232 port is connected to an RS-232-to-USB converter that allows
communication and programming from a host computer. There is no WiFi, Bluetooth, or
Ethernet support on the UNO board, but extension boards, so-called shields, are available
for mounting directly onto the pin headers.

55
56  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.1 An Arduino UNO.

One can describe the Arduino UNO as having the intelligence of a washing machine. It
keeps time with the timers, it can sense voltages from, for example, temperature sensors,
and it can turn motors or pumps on or off, depending on whether some condition is met.
In this way it can also provide the glue logic to interface sensors (analog, I2C, SPI, other)
to the host computer, and that is the mode in which we will use the UNO later on. In the
next section, however, we address a second microcontroller, the ESP8266.

4.1.2 ESP8266 and NodeMCU

The ESP8266 was released by the Chinese company Espressif in 2014 and quickly became
one of the most exciting platforms for Internet of Things (IoT) projects because of its built-in
WiFi support, including a small patch antenna. The software development kit (SDK) from
Espressif was soon linked to the Arduino software development environment, and today
the ESP8266 can be programmed using the Arduino ecosystem as easily as the original
Arduinos.

Figure 4.2 The ESP-01 on the left and a NodeMCU on the right.
 Microcontroller: Arduino  57

Figure 4.3 The Arduino IDE.

But let us now look at the ESP8266 hardware, which comes in different incarnations.
We will focus on the more elaborate NodeMCU board, but also mention the basic ESP-01
in this section. Images of the ESP-01 and NodeMCU hardware are shown at the left and
right of Figure 4.2, respectively. The microcontroller on both chips is essentially the same,
only the number of internal controller pins that are routed to pins on the circuit boards is
different. Internally, the ESP8266 chips feature 32-bit RISC CPUs that normally operate
at 80 MHz or 160 MHz, and have 64 kB of RAM for instructions and 96 kB for data. There
are 16 general-purpose IO (GPIO) pins that can be configured as input or output pin, as
well as a single analog input with a 10-bit ADC. The controller supports I2C and SPI
communication as well as RS-232. All this is similar to the Arduino, except it runs faster,
and the internal registers are wider (32 instead of 8 bit in the ATmegas), but the really
cool feature is the built-in ready-to-use WiFi support. It fulfills the specifications for IEEE
802.11 b/g/n with WPA authentication, which is what most wireless networks use today.
So basically, we are able to connect an ESP8266 to any wireless network.
After having introduced the hardware, we need to move on and start programming. This
is most easily done using the Arduino IDE (Integrated Development Environment) and we
provide a quick-start guide in the following section.

4.2 GETTING STARTED

The Arduino IDE is a software bundle with a size on the order of 100 MB that can be
downloaded from
https://www.arduino.cc
58  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

for the most common computer systems, such as Linux, Mac, or Windows. Basically, we
go to the web site, choose the Download tab, select our computer type, and download the
software package. Then we install the software by following the instructions given on the
web site. Once the installation completes, we start the IDE by clicking the icon, or start it
from the command line by typing arduino followed by Enter, which should open the IDE
and show us a window similar to the one shown in Figure 4.3. If not, create a “New” sketch,
which is what Arduino programs are called, by selecting “New” from the “File” menu.
Here we already see the general structure of Arduino programs (or synonymously
“sketches”). There is a setup() function, which is executed once, immediately after power
is turned on. In this routine all initialization housekeeping is done, such as defining a pin to
be output or input, and configuring the serial line. Once the setup() function completes,
the loop() function is called repetitively, such that once it completes, it is called again and
so forth, until power is turned off. Note that the programming language supported by the
Arduino IDE is very similar to the C language. There are, however, a number of special
extensions to provide access to the specific hardware, such as the ADC.
Now we connect the Arduino UNO to any USB port on the host computer where the
Arduino IDE is running, and select Arduino UNO from the Tools→Board menu. This step
tells the IDE for which processor the compiler will generate code, as well as some hardware-
specific definitions such as the names of IO pins that we can use when programming. At
this point the IDE “knows” what the hardware is, but we still need to tell the IDE to which
USB port the UNO is connected. This we do in the Tools→Port menu, where normally
the serial port to which the UNO is connected automaticlly appears and can be selected.
On Linux this often is /dev/ttyUSB0 or /dev/ttyACM0. On a Windows computer it is COMx
where x is some number.
At this point we could start programming the UNO, even though the example program
in Figure 4.3 only contains empty functions. But before writing programs for the Arduino,
we want to install support files for ESP8266-based microcontrollers. They are easily in-
stalled by opening the File→ Preferences menu and adding http://arduino.esp8266.
com/stable/package_esp8266com_index.json to the “Additional Boards Manager URL”
text box and clicking the “OK” button. Then open the Tools→Board menu and open
the “Boards Manager,” and find the “esp8266” platform, select the newest version and
click on the “Install” button. Once installation completes, select the ESP platform from
the Tools→Board menu. For the ESP-01 the entry “Generic ESP8266 Module” is a good
choice and for the NodeMCU platform it is “NodeMCU 1.0.”
Once the correct platform (UNO, plain ESP8266, or NodeMCU) is selected, we are
ready to write programs. In the following we will tacitly assume an UNO is connected,
unless stated otherwise. So, now we are ready to write programs (sketches) and download
them to the hardware.

4.3 HELLO WORLD, BLINK

The first sketch we consider is the standard for almost any new controller, namely to
cause an LED to blink. Examples to interface hardware such as sensors can be found
in the File→Examples menu, where the example to blink an LED is located under the
01.Basics→ Blink menu item. Once selected, a new window opens with the example
code. Figure 4.4 shows the bare code, without many comment lines that are preceded by
//. It is reproduced here.
void setup() {
pinMode(LED_BUILTIN, OUTPUT);
 Microcontroller: Arduino  59

Figure 4.4 The “Blink” sketch.

}
void loop() {
digitalWrite(LED_BUILTIN, HIGH); // turn the LED on
delay(1000); // wait for a second
digitalWrite(LED_BUILTIN, LOW); // turn the LED off
delay(1000); // wait for a second
}
Let us step though this sketch one line at a time. In the setup() function, we call the
pinMode(Pin,What) function, which declares that the pin called LED BUILTIN (which is pin
13 on a UNO) will be used as output pin. That is all the initializing we do in the setup()
function. In the loop() we call digitalWrite(Pin,State), which causes the controller to
put 5 V onto the specified pin if the requested state is HIGH and 0 V if the requested state
is LOW. The latter happens two lines later. In between the changes to the output pin, we
tell the controller to wait for a specified number of milliseconds in the delay(time in ms)
function. So, all the loop function does is turn on the LED on pin 13, wait for 1000 ms or
1 s, turn it off, wait again for 1 s, and then start all over again. Note that all commands are
terminated by a semicolon, as is customary in the C-language.
Once we are satisfied with our program, we can compile it and check that the syntax is
correct and we did not forget any semicolons. Pressing the checkmark symbol just below the
File menu entry compiles the program and reports progress in the status window below
the program window. If the compilation completes without errors and the UNO board
is connected to a USB port, and it is selected in the Tools→Port menu entry, we can
download the program to the UNO by pressing the → button located to the right of the
compile button. After a few seconds, a small LED on the UNO board should start to blink
once per second. Now we can change, for example, the delay time in the sketch, compile
and download again, and observe whether the LED blinks according to the newly specified
times.
60  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.5 Interfacing a button to the Arduino.

Here we only used very few commands, such as pinMode or digitalWrite, but there
are many more, and checking the Reference section on the Arduino web site at https:
//www.arduino.cc/en/Reference/HomePage is very inspirational. Moreover, clicking on
the different commands opens a new page, with explanations and examples on how to
use the functions. A further source of wisdom is the “Learning” tab from the menu list
on the top of https://www.arduino.cc/, which offers a large number of tutorials and
examples. Keeping these resources in mind will help us to quickly locate information to
make things work. Moreover, there are a large number of books on various Arduino projects,
and compendia of tips and tricks such as [19].
After the first step to make the microcontroller follow our bidding, we move on and read
sensors.

4.4 INTERFACING SENSORS

Here, the main task of the Arduino microcontroller is to read sensors, with their different,
often idiosyncratic interfaces, and convert the measured values to a standard representation
that is communicated to a host computer. We start with the simplest sensor, a button or a
switch.

4.4.1 Button
In Figure 4.5, we show how to connect a button between ground and pin 2 of the Arduino,
such that pressing the button will cause pin 2 to read 0 V or LOW. The following listing
shows a sketch that causes the LED on pin 13 to light up, if the button is pressed.
// button_press, V. Ziemann, 161013
 Microcontroller: Arduino  61

void setup() {
pinMode(2,INPUT_PULLUP); // button, default=HIGH
pinMode(13,OUTPUT); // LED
}
void loop() {
if (digitalRead(2)==LOW) {
digitalWrite(13,HIGH);
} else {
digitalWrite(13,LOW);
}
delay(10);
}
The program follows the normal scheme of initializing the hardware in the setup() function.
We first declare that pin 2 is an input pin, and we enable the internal pull-up resistor that
causes a well-defined “high” state on the pin if no button is pressed. Pin 13 with the LED
is declared as output, so we can turn it on and off from within the sketch. In the loop()
function, we check the state of the button by calling the digitalRead(Pin) function, and
test whether pin 2 is LOW. If this is the case (note the double equal sign == in the comparison),
the LED on pin 13 is turned on by calling digitalWrite(13,HIGH). If pin 2 is found not to
be LOW, the LED on pin 13 is tuned off with digitalWrite(13,LOW);. Note the braces { and
} and their use to define blocks of code in the if (..) { } else { } construction. After
the if statement, a short delay() ensures that mechanical button bounces are ignored and
that the processor has a little time for its internal affairs. This is not absolutely necessary,
but good style.
Sensing the state of switches that indicate open doors or windows follows exactly the
same scheme, and we can sense up to 13 different switches (if we do not use the LED on
pin 13) and react in some manner.

4.4.2 Analog input

In an earlier chapter, we saw that some sensors require us to measure an analog voltage
because they change their resistance. We place them in a voltage divider and then read
the voltage change on the central tap, as shown on the left in Figure 2.1. Others, such as
the LM35 temperature sensor, directly produce a voltage and in this section we show how
to interface these sensors to the Arduino. First, we want to measure a voltage from one
of the analog input pins; here we use A0. The left of Figure 4.6 shows how to connect a
potentiometer as a variable voltage divider to the Arduino. One end of the potentiometer is
connected to ground, the other end to the supply voltage, and the wiper with the variable
tap of the potentiometer is connected to pin A0. Turning the axis of the potentiometer
causes the voltage on the center tap to vary between ground and the 5 V supply voltage.
The sketch to read the voltage and report it via the serial cable (that is connected to
the USB port) to the host computer is shown next.
// Analog and serial communication, V. Ziemann, 161130
int inp,val;
void setup() {
Serial.begin(9600); // baud rate
}
void loop() {
if (Serial.available()) {
62  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Interfacing a potentiometer (left) or an LM35 temperature sensor (right)

Figure 4.6
to the Arduino.

inp=Serial.read(); // read character from serial

val=analogRead(0); // read analog
Serial.print("Value is "); // and report back
Serial.println(val);
}
delay(50); // wait 50 ms
}
Since we want to temporarily store integer values, we have to allocate memory for the
variables, which is done in the line containing int inp,val; as it declares two integer
variables named inp and val. In the setup() function, we declare that we want to use
the built-in serial (RS-232) hardware port with a speed of 9600 baud, which is about 1000
characters per second. The receive (RX) and transmit (TX) lines of the serial port are
connected to pin 0 and pin 1, but also routed to the RS232-USB converter on the UNO
board that is connected to the host computer. In the loop() function, we first check whether
some communication from the host computer has arrived, and if that is the case, we read
the character with the Serial.read(); command, but do nothing further with it. In the
next line we read the analog pin A0 using the built-in ADC, and store the value in the
integer variable val. The value returned from the 10-bit ADC is a number between 0 and
1023 (= 210 − 1) and not the voltage. Later we will show how to convert this value to a
voltage. The result of the measurement is then sent via the serial line by Serial.print()
and Serial.println() commands. The difference between the two is that the former does
not send a carriage return character at the end of the message, and the latter does. After the
if() { } statement at the end of the loop, again a small delay is added. Now we can compile
and download the sketch to the Arduino, unless some syntax error crept in. If that is the
case, we need to check for missing semicolons or spelling errors, especially the “camelHump-
spelling” and capital letters in the middle of the commands. Care is advisable, but the IDE
 Microcontroller: Arduino  63

uses syntax highlighting, which makes such errors reasonably easy to spot. After all errors
are corrected, we download the sketch to the hardware.
Once the sketch is running on the Arduino, we want to test it, and for that we need to
have access to the other end of the serial communication channel on the host computer. The
easiest way is to open the “Serial Monitor” by pressing the icon that looks like a looking
glass at the top right in the Arduino IDE. Clicking on it opens a simple terminal program
that has an input text box at the top. Enter any character and press “Enter.” If all is well,
the Arduino should report with a line Value is nnnn in the text box below the entry box.
Now turn the potentiometer and enter a character followed by “Enter,” and see how the
returned value changes.
So far, so good, but would it not be nice if we could read out more analog ports and
even ask for which ones to read. This task is addressed in the following sketch. It is rather
similar to the previous one and we only discuss the new features.
// Analog and Serial communication, v2, V. Ziemann, 161130
int inp,val;
void setup() {
Serial.begin(9600); // baud rate
}
void loop() {
if (Serial.available()) {
inp=Serial.read(); // read serial line
if (inp==’s’) {
val=analogRead(0); // read analog
Serial.print("A0= "); // and report back
Serial.println(val);
} else if (inp==’t’) {
val=analogRead(1);
Serial.print("A1= "); Serial.println(val);
} else {
Serial.println("unknown command");
}
}
delay(50); // wait 50 ms
}
In this sketch we actually test which character was sent on the serial line, and depending
on whether it is an “s” or a “t” the value from analog pin A0 or A1 is returned. The
characters sent back to the host via the serial line contain both the pin read and the value.
If a character different from “s” or “t” is received by the Arduino, the string “unknown
command” is returned via the serial line. In this way, we build a simple query-response
protocol to make the Arduino do different things, depending on the character we send.
In the next example we assume that an LM35 temperature sensor is connected to analog
pin A0 in the way shown on the right in Figure 4.6. Note that we add a 100 nF capacitor
across the positive supply voltage and ground pins of the LM35, to increase the stability of
the circuit. Note that it is good practice to add such a decoupling or bypass capacitor close
to the power pins of each chip in a circuit. In the following, we often do not show these
capacitors on the circuit diagrams in order to make them less cluttered, but the capacitors
should be included in the hardware. The software to interface the LM35 is slightly more
elaborate than in previous examples. It allows us to send somewhat descriptive multichar-
acter commands to the Arduino and receive a response that echos the request followed by
64  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

the value. The basic idea is that we request a parameter, say A0, by sending the parameter
name with a question-mark appended, such that the request looks like this: A0?. The Ar-
duino subsequently replies with A0 <value>. In the following sketch we implement this, but
also that the Arduino first reads a full line, in the sense that it waits for a carriage-return
and a line-feed character before responding.
// query-response, V. Ziemann, 161013
int val;
float temp;
char line[30];
void setup() {
analogReference(INTERNAL); // 1.1V internal ADC reference
Serial.begin (9600);
while (!Serial) {;}
}
void loop() {
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"A0?")==line) {
val=analogRead(0);
Serial.print("A0 "); Serial.println(val);
} else if (strstr(line,"A1?")==line) {
val=analogRead(1);
Serial.print("A1 "); Serial.println(val);
} else if (strstr(line,"T?")==line) {
temp=1.1*100*analogRead(0)/1023;
Serial.print("T "); Serial.println(temp);
} else {
Serial.println("unknown");
}
}
}
In the first few lines, we declare the variables, and in particular, a 30-character buffer
line to hold the multicharacter query. In the setup() function, we use the internal 1.1 V
reference voltage for the ADC and initialize the Serial communication channel to operate at
9600 baud. The while(!Serial) statement waits for the initialization of the Serial library
to complete, and is only needed on some newer Arduinos. We include it just for safety. In
the loop() function, we wait for characters to be available before reading an entire line
until the end-of-line character \n. The readStringUntil returns a String but we want to
use a character array instead, and need to convert the type in the appended toCharArray
method. At this point the character array line contains the request that was sent from
the host computer to the Arduino; for example, A0?. The following construction of if
(strstr(line,"A0?")==line) and similar commands checks whether line starts with the
characters A0? by comparing the pointer to the first occurrence of A0? in the array line[]
to the pointer to the start of line[]. The commands that follow are executed provided
the result evaluates to true, or else the subsequent comparison in the else if sequence is
tested. In the example, we test for "A0?", "A1?", and ‘"T?" to reply with the raw values of
the analog pins 0 and 1 or with the value from analog pin 0 converted to degrees Celsius.
Consult the discussion in Section 2.1.2 regarding the conversion coefficient. If a command
arrives, and it is not in the list, the Arduino replies with unknown command. The way to
 Microcontroller: Arduino  65

specify the query with an appended question mark and the reply to echo the query with
value appended is borrowed from the SCPI command-language that is commonly used to
remotely control measurement equipment via GPIB or VXI.
As a final example, we discuss a simple temperature logger that again uses the hardware
setup as shown on the right side of Figure 4.6, but sends the temperature value continuously
to the host computer via the serial line. The sketch is shown here:
// ultra-simple LM35 temperature logger, V. Ziemann, 161201
float temp;
void setup() {
analogReference(INTERNAL); // 1.1V internal ADC reference
Serial.begin(9600);
while (!Serial) {;} // wait for serial to initialize
}
void loop() {
temp=1.1*100*analogRead(0)/1023.0;
Serial.println(temp);
delay(1000);
}
where the setup() function equals that from the previous example, but the loop() function
measures the analog pin and converts the digital word to the temperature, and sends that to
the host computer. Then it waits 1000 ms before repeating the same. In the Serial Monitor,
that can be started by clicking on the button at the top right in the IDE; we see that the
temperature value in Celsius is reported one per second in the lower text box provided the
baud rate selected in the selection box at the lower right matches the baud rate specified
in the sketch.
In the Arduino IDE, however, there is also a graphical plotting tool available. We select
it by choosing the Tools→Serial Plotter menu item. This opens a window and shows
the values received on the serial line in graphic format, provided the baud rate is selected
to match the one specified in the sketch. Note that the Serial Plotter can show several
measurement values simultaneously as traces in different colors. This functionality, built
in to the Arduino IDE, may serve as a simple logger for temperatures or other measured
values, if a quick solution to record some values as a function of time is needed.
So far, we managed to read analog voltages, and even devised a simple protocol to report
the measurement values to the host computer. In the next section, we interface I2C-based
sensors to the microcontroller, and use the same query-response protocol to communicate
with the host.

4.4.3 I2C
And we continue with another temperature sensor, but a special one that is used by nurses in
hospitals. It is based on a contact-free thermometer that takes your temperature by pointing
a gadget into your inner ear. An example is the MLX90614 infrared thermometer shown on
the right in Figure 2.9. It uses a thermopile to deduce the temperature from the picked-up
infrared radiation and it reports the temperature to our Arduino via an I2C interface. Here
we point out that the MLX90614 comes in versions for 5 V and 3.3 V supply voltage. In this
example we use the 5 V version, labeled AAA. The device has four pins—check the datasheet
to find out which pin does what—for ground, supply voltage, and the I2C lines SDA as SCL.
The latter two are connected to pin A4 and A5 on the UNO because the I2C data and clock
lines are routed to the same output pins as the analog pins A4 and A5. These pins cannot
66  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.7 Connecting the MLX90614 contact-free thermometer to an Arduino UNO.

be used for analog measurements, in case they are used for I2C connectivity. We show how
to connect the sensor in Figure 4.7, and illustrate how to talk to it in the following sketch.
// Read MLX90614 IRthermometer, V. Ziemann, 170717
#include <Wire.h>
const int MLX90614=0x5A; // I2C address
float getTemperature(uint8_t addr) { //.....getTemperature
uint16_t val;
uint8_t crc;
Wire.beginTransmission(MLX90614);
Wire.write(addr); // address
Wire.endTransmission(false);
Wire.requestFrom(MLX90614,3);
val = Wire.read();
val |= (Wire.read()<<8);
crc=Wire.read(); // not used
return -273.15+0.02*(float)val;
}
void setup() { //....................................setup
Serial.begin(9600);
Wire.begin();
}
void loop() { //.....................................loop
float Ta=getTemperature(0x06); // address for ambient temp
float To=getTemperature(0x07); // address for object1 temp
Serial.print(Ta); Serial.print("\t"); Serial.println(To);
delay(1000);
}
First we have to include support for the I2C functionality by including the Wire.h header
file, which also causes the compiler and linker to include the corresponding libraries. After we
define the I2C address 0x5A of the sensor, which we find in the datasheet, we encapsulate
the I2C communication in a separate function called getTemperature(). It receives the
 Microcontroller: Arduino  67

Figure 4.8Connecting the HYT-221 humidity sensor to the Arduino (left) and using
the serial plotter to display the humidity and temperature measurements (right).

register address of the I2C device as input parameter and returns the associate temperature,
properly scaled to degrees Celsius. The device also returns a byte that allows us to determine
transmission errors, but we do not use that feature in this simple example. In the setup()
function we only initialize Serial line and I2C communication. In the loop() function we
use the getTemperature() function to retrieve the ambient temperature from address 0x06
and the object temperature from address 0x07, as described in the datasheet, and print
both temperatures formatted with a tabulator to the Serial line. The entire process is then
repeated after waiting 1000 ms. We can use the Serial monitor or plotter built into the
Arduino IDE in order to view or display the results.
Several other sensors, often related to environmental parameters such as humidity or
barometric pressure, also report the temperature as a by-product, because it is internally
needed to calibrate the reported primary measurement value. For the next sensor, the HYT-
221, this is the humidity. Connecting the sensor to the Arduino is illustrated on the left
in Figure 4.8. From the datasheet we know that the four pins of the sensor, when looking
towards the front of the sensor, are SDA, GND, VDD, and SCL. The pins for GND and VDD are
connected to the GND and 5 V pins on the Arduino, and SDA to A4 and SCL to A5 in the
same way we did for the temperature sensor in the previous example. In the following sketch,
we read the HYT sensor, and send the measurement data of humidity and temperature via
the serial line to the host computer.
// HYT221 humidity sensor, V. Ziemann, 161203
#include <Wire.h>
const int HYT=0x28; // I2C address
void setup() { //............................setup
Serial.begin (9600); while (!Serial) {;}
Wire.begin();
}
void loop() { //..............................loop
int b1,b2,b3,b4,raw;
double humidity,temp;
Wire.beginTransmission(HYT);
Wire.requestFrom(HYT,4);
68  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

delay(50);
if (Wire.available()==4) {
b1=Wire.read(); b2=Wire.read(); // humidity rawdata
b3=Wire.read(); b4=Wire.read(); // temperature rawdata
Wire.endTransmission(true);
}
raw=(256*b1+b2) & 0x3FFF; // humidity
humidity=100.0*raw/16384.0;
raw=((256*b3+b4) & 0xFFFC)/4; // temperature
temp=165.0*raw/16384.0-40.0;
Serial.print(humidity); Serial.print("\t"); Serial.println(temp);
delay(1000);
}
First we include the I2C functionality with Wire.h, and define a variable HTY that contains
the I2C address 0x28. In the setup() function, we first initialize the serial communication
and then the I2C functionality with the call to the Wire.begin() function. In the loop()
function we first declare a number of auxiliary integer variables and double-precision float
variables to hold the measurement data before calling the Wire.beginTransmission()
function with the address HTY as argument, and tell the Arduino to fetch four bytes as
response from the sensor in the following line. After a short delay, we check whether the four
bytes have arrived, with the Wire.available() function, and in case they are available, we
read them into the variables b1,..,b4 before closing the transmission. The first two bytes
contain the raw data for the humidity, and the next two for the temperature measurement.
In the Arduino sketch, we follow the instructions from the HYT-221 datasheet to convert
the raw data to the physical quantities. For the humidity, the datasheet tells us that the
first received byte b1 is the most significant byte of a two-byte word and b2 is the least
significant byte. Therefore, we can reconstruct the word by 256*b1+b2. But we also have
to throw away (mask) the two highest bits, which we do by a logical and (&) with a binary
value that has ones everywhere except at the two highest bits, which is & 0x3FFF. Once we
have the raw value, we again follow the datasheet and scale by 100/16384 to obtain the
humidity in percent. For the temperature, we follow a similar procedure that is described
in the HYT-221 datasheet. Finally, we send the humidity and temperature values to the
host computer via the serial line and wait a while before repeating the measurement. On
the right of Figure 4.8 we show the data using the Serial-plotter in the Arduino IDE. The
upper trace shows the humidity, and the two positive excursions are due to exhaling on to
the sensor. The lower trace shows the temperature, which we briefly cause to increase by
placing a lamp close to the sensor.
The next environmental sensor—we use it later in a weather station—is the barometric
pressure sensor BMP180. Like the previous two sensors, it communicates with the Arduino
via the I2C bus, and needs four lines for SCL, SDA, GND, and VDD. Note that we can
connect this sensor in parallel to the previous sensors using the same four pins on the
Arduino because each sensor has its unique address. For this sensor, we must be careful
with the supply voltage, because higher voltages than 3.3 V can damage it. Therefore, we
use the 3.3 V supply voltage available on the Arduino to power the sensor. The I2C pins are,
however, tolerant to higher voltages. As mentioned in Section 2.3.3, each individual sensor
is factory calibrated, with calibration constants stored on an EEPROM on the sensor chip.
In order to convert raw ADC values, reported from the device, to pressure and temperature
in conventional units, we need to follow the instructions in the datasheet. The following
 Microcontroller: Arduino  69

code implements the basic functionality to retrieve the pressure and the temperature from
the sensor.
// BMP180, V. Ziemann, 170805
#include <Wire.h>
const int BMP=0x77; // I2C address
int16_t ac1,ac2,ac3,b1,b2,mb,mc,md;
uint16_t ac4,ac5,ac6;
void I2Cwrite(uint8_t addr, uint8_t reg, uint8_t val) {
Wire.beginTransmission(addr); // address
Wire.write(reg); // register
Wire.write(val); // write value
Wire.endTransmission(true);
}
uint8_t I2Cread(uint8_t addr, uint8_t reg) {
Wire.beginTransmission(addr); // address
Wire.write(reg); // register
Wire.endTransmission(false);
Wire.requestFrom(addr,1); // request 1 byte
return Wire.read(); // read and return
}
bool bmp180_init() { //......................bmp180_init
Wire.beginTransmission(BMP); // read EEPROM
Wire.write(0xAA); // start register address
Wire.endTransmission(false);
if (Wire.requestFrom(BMP,22)) {
ac1=Wire.read()<<8 | Wire.read();
ac2=Wire.read()<<8 | Wire.read();
ac3=Wire.read()<<8 | Wire.read();
ac4=Wire.read()<<8 | Wire.read();
ac5=Wire.read()<<8 | Wire.read();
ac6=Wire.read()<<8 | Wire.read();
b1=Wire.read()<<8 | Wire.read();
b2=Wire.read()<<8 | Wire.read();
mb=Wire.read()<<8 | Wire.read();
mc=Wire.read()<<8 | Wire.read();
md=Wire.read()<<8 | Wire.read();
Wire.endTransmission();
return true;
}
return false;
}
float bmp180_getdata(float *T) { //.......bmp180_getdata
float ut,x1,x2,x3,b3,b4,b5,b6,b7,tt,p;
uint16_t up;
I2Cwrite(BMP,0xF4,0x2E); // request temperature first
delay(5);
ut=(float)(I2Cread(BMP,0xF6) << 8 | I2Cread(BMP,0xF7));
x1=(ut-ac6)*ac5/32768.0;
x2=mc*2048.0/(x1+md);
70  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

b5=x1+x2;
*T=(b5+8.0)/160.0;
I2Cwrite(BMP,0xF4,0x34); // request pressure measurement
delay(5); // with oss=0, wait 5 ms, table 8
up=I2Cread(BMP,0xF6)<<8 | I2Cread(BMP,0xF7);
b6=b5-4000;
x1=(b2*(b6*b6/4096.0))/2048.0;
x2=ac2*b6/2048.0;
x3=x1+x2;
b3=((((long)ac1*4+x3))+2)/4;
x1=ac3*b6/8192;
x2=(b1*(b6*b6/4096))/65536;
x3=((x1+x2)+2)/4;
b4=ac4*(x3+32768)/32768;
b7=(up-b3)*(50000);
if (b7< 0x80000000) {p=b7*2/b4;} else {p=(b7/b4)*2;}
x1=(p/256)*(p/256);
x1=(x1*3038)/65536;
x2=(-7357*p)/65536;
p=p+(x1+x2+3791)/16;
return 0.01*p;
}
void setup() { //..................................setup
Serial.begin (9600); while (!Serial) {;}
Wire.begin();
if (!bmp180_init()) Serial.println("No sensor found!");
}
void loop() { //....................................loop
float pressure, temperature;
pressure=bmp180_getdata(&temperature);
Serial.print(temperature); Serial.print("\t");
Serial.println(pressure);
delay(2000);
}
At the top of the sketch we include the I2C functionality, define a constant with the default
I2C address of the BMP180 sensor, and declare the variables that will contain the calibra-
tion constants. Then we define two convenience functions to read and write a byte from a
device at I2C address addr, register reg, and value val. In the bmp180 setup() function,
we only read the calibration constants from the internal EEPROM. The addresses start at
0xAA and there are 11 constants with two bytes each to read. We therefore retrieve 22 bytes
and immediately assemble the constants by joining two bytes to a 16-bit word, some of them
signed and some unsigned, as specified in the datasheet. The initialization function returns
TRUE if the communication is successful and FALSE otherwise. The bpm180 getdata() func-
tion returns the pressure and the temperature. First, we declare temporary variables, before
requesting a temperature measurement by writing 0x2E to the control register 0xF4, wait
5 ms for the measurement to complete, and then retrieve the raw reading from the output
registers of the ADC, 0xF6, and 0xF7, and assemble a 16-bit word. The next four lines with
the temporary variables x1, x2, b5 are copied straight from the datasheet to convert the
raw reading to the temperature in Celsius, which we store in the variable T. Next we request
 Microcontroller: Arduino  71

a pressure measurement with low precision by writing 0x34 to the control register 0xF4,
wait a short while, and assemble the raw readings from registers 0xF6 and 0xF7 to form the
16-bit word up. The following intermediary calculations are again copied straight from the
datasheet, and at the end we return the pressure p in Pascal. Finally, we convert to mbar
by multiplying the value with 0.01, and return it. In the setup() function, we initialize the
serial line, the I2C bus, and the sensor by calling bmp180 init(), and issue a warning if
no sensor is found. In the loop() function we only call the bmp getdata() function and
display the values before waiting a short while and repeating this indefinitely.
In the previous examples with the humidity sensor and thermometer, we coded the I2C
communication by hand, but often there are libraries available that encapsulate the details
of the communication and the often-arcane prescriptions from the datasheets to convert the
raw data to physical quantities. For the BMP180 there are several libraries available, and
we will use the SFE BMP180 library. We install the library by first downloading a zip file with
library and examples and unpacking its contents in the Arduino/libraries subdirectory.
In this case the library resides in the Arduino/libraries/SFE BMP180/ directory. After
restarting the Arduino IDE, the library and examples of how to use it are ready to use. The
example equivalent to the previous one is the following.
// BMP-180 barometric pressure sensor, V. Ziemann, 161204
#include <SFE_BMP180.h>
#include <Wire.h>
SFE_BMP180 pressure;
void setup() {
Serial.begin (9600); while (!Serial) {;}
if (!pressure.begin()) {Serial.println("BMP180 failed!");}
}
void loop() {
char status;
double T,P;
status = pressure.startTemperature(); delay(status);
status = pressure.getTemperature(T);
status = pressure.startPressure(3); delay(status);
status = pressure.getPressure(P,T);
Serial.print(T,2); Serial.print("\t"); Serial.println(P,2);
delay(2000);
}
This example is much shorter than the previous one, because most of the details of the
I2C communication and data conversion from raw to physical quantities are hidden in the
library. We include the library by placing the include directive for the header files near
the top of the sketch, both for the SFE BMP180 library and the I2C library Wire.h needed
for the communication. In the next line, a variable pressure of type SFE BMP180 is de-
clared, which is the quantity (an object) that encapsulates the functionality provided by
the BMP180 sensor. In the setup() function, first the serial and then the communication
with the BMP180 sensor is initialized, and an error message is printed if the initializa-
tion fails. The loop() function is particularly simple. After declaring some variables, the
pressure sensor is requested to start a temperature measurement, which returns a status
that contains the waiting time until the measurement data are available and ready to be read
with the getTemperature function. Once the temperature is known, we start the pressure
measurement, which returns the waiting time as the status until the data can be fetched
with the getPressure function. Once the temperature and pressure are known, they are
72  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

sent via the serial line to the host computer. After a small delay, the whole procedure is
repeated. In this very basic example, we only show the basic functionality, and omit error
checking of the status. In any program used for serious data acquisition, all available status
information should of course be checked.
We can use the next sensor to determine the acceleration and speed of rotation on
carousel rides at a fair. Incidentally, the same sensor is used in smart phones to determine
whether they are held upright or sideways. This device, the MPU-6050, contains separate
sensors for acceleration in three dimensions and for angular velocity around three axes. It
uses a standard I2C with default address 0x68. In the following sketch, we explain how to
interface it to the Arduino.
// MPU6050, V. Ziemann, 170726
#include <Wire.h>
const int MPU6050=0x68;
void I2Cwrite(int addr, int reg, int value) { //...I2Cwrite
Wire.beginTransmission(addr);
Wire.write(reg);
Wire.write(value & 0xFF);
Wire.endTransmission(true);
}
uint8_t I2Cread(int addr, int reg) { //.............I2Cread
Wire.beginTransmission(addr);
Wire.write(reg);
Wire.endTransmission(false);
Wire.requestFrom(addr,1);
return Wire.read();
}
void mpu6050_init() { //.......................mpu6050_init
Wire.begin();
int whoami=I2Cread(MPU6050,0x75);
if (whoami!=0x68) {
Serial.println("No MPU6050 found!"); while (1) {yield();}
}
I2Cwrite(MPU6050,0x6B,0); // wake up device
I2Cwrite(MPU6050,0x1B,0); // Gyro config, FS=0
I2Cwrite(MPU6050,0x1C,0); // Acc config, AFS=0
}
void mpu6050_read_float(int addr, float fdata[7]) { //...read_float
int16_t intval;
Wire.beginTransmission(addr);
Wire.write(0x3B);
Wire.endTransmission(false);
Wire.requestFrom(addr,14);
intval=Wire.read()<<8 | Wire.read(); fdata[0]=intval/16.384; //ax
intval=Wire.read()<<8 | Wire.read(); fdata[1]=intval/16.384; //ay
intval=Wire.read()<<8 | Wire.read(); fdata[2]=intval/16.384; //ay
intval=Wire.read()<<8 | Wire.read(); fdata[3]=intval/340.0+36.53; //T
intval=Wire.read()<<8 | Wire.read(); fdata[4]=intval/131.0; //gx
intval=Wire.read()<<8 | Wire.read(); fdata[5]=intval/131.0; //gy
intval=Wire.read()<<8 | Wire.read(); fdata[6]=intval/131.0; //gz
 Microcontroller: Arduino  73

Wire.endTransmission();
}
void setup() { //.................................setup
Serial.begin(115200);
while (!Serial) {yield();}
mpu6050_init();
}
void loop() { //..................................loop
float fdata[7];
mpu6050_read_float(MPU6050,fdata);
Serial.print(fdata[0],0); Serial.print("\t");
Serial.print(fdata[1],0); Serial.print("\t");
Serial.print(fdata[2],0); Serial.print("\t");
Serial.print(fdata[4],0); Serial.print("\t");
Serial.print(fdata[5],0); Serial.print("\t");
Serial.print(fdata[6],0); Serial.print("\t");
Serial.println(fdata[3]);
delay(1000);
}
Here we first include I2C support by loading the Wire.h header file, and define a symbolic
name for the I2C-address of the device. Next we define the two convenience functions to
write and read a single byte from the device, which we already encountered in a previous
example in this section. In the mpu6050 init() function, we initialize the I2C library and
check for the whoami byte at address 0x75. According to the datasheet, a value of 0x68
indicates that the device is present. Then we need to write to address 0x6B in order to
wake up the device, and then configure the gyro and accelerometer to use the standard
ranges of ±250o /s and ±2g. The mpu6050 read float() function requests 14 consecutive
bytes from the device, which encodes the three gyro sensors, three accelerometers, and
temperature with two bytes each. Here 0x3B is the starting address of the range we intend
to read. We then convert the corresponding bytes to floating point values. The first three
values in the array fdata[] are the accelerations in the three spatial directions, converted
to units of 10−3 g, where g is the gravitational acceleration on the surface of the earth. The
fourth entry is the temperature that is converted from raw ADC readings to degrees Celsius
by conversion constants found in the datasheet. The last three entries contain the rate of
rotation around three axes, converted to degrees per second. Again the conversion constant
is taken from the datasheet. The setup() function only initializes the serial line and the
MPU6050. The loop() function retrieves the data from the device by using the function
we defined earlier, and displays the values in a single line. We note in passing that there are
libraries available that encapsulate the details of the communication and conversion, and
therefore simplify the code significantly, albeit at the expense of reduced transparency and
flexibility.
Occasionally the number of input or output pins on the microcontroller is insufficient for
a certain project, but luckily there are integrated circuits that help to extend the number
of IO pins. One of these is the MCP23017, a device that provides 16 IO pins, each of them
freely configurable as an input or output pin. The device has an I2C interface, such that it
is equally easy to address as the other examples in this section, and it can even share the
same I2C lines, SCL and SDA, with other devices, because it has its own address: 0x20.
Should an address-conflict with another device arise, it can be changed to any value in the
range from 0x20 to 0x27 by three address pins that need to be set to either ground or supply
74  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.9 Connecting the MCP23017 IO-extender to the UNO.

voltage. This also implies that we can place eight copies of an MCP23017 on the same bus
and can use 128 extra IO pins. In the following example, we are less ambitious, and use one
chip only. We configure it such that eight pins are output and eight pins are input. The chip
has the attractive feature that it can generate an interrupt if an input pin changes its value.
We will use that feature, despite having to use one extra pin on the microcontroller, because
it relieves the controller of continuously checking whether any input pin on the MCP23017
has changed. The controller only needs to determine which pin has changed, once a change
has occurred. In Figure 4.9 we show the circuit with an Arduino UNO interfaced to one
MCP23017. On the extender, we connect one pin to an LED that we want to turn on and
off. Another pin is connected to a button, and we want to determine whether it is pressed
or not.
In the circuit shown in Figure 4.9 we see the UNO on the left and a small solderless
breadboard with the extender on the right. Wires connect the breadboard to the 5 V and
ground terminals of the UNO, and the I2C lines for SDA and for SCL are connected to
pins A4 and A5, respectively, just as before in earlier examples. Moreover, we pull the three
address pins at the lower right of the MCP23017 to ground, which causes the device to
have the default address 0x20. We also pull the reset pin, the fourth from the bottom on
the right-hand side, to 5 V, which is required for stable operation. There are eight IO pins
on the top-right part of the chip, labeled A0 to A7, where A7 is connected to the anode of
the LED, and eight more IO pins on the top left of the chip. They are labeled B0 to B7
from top to bottom, with B0 connected to the small button on the top left. An additional
wire connects a pin called INTB on the MCP23017 to pin D2 on the UNO. We configure
the extender in such a way that INTB indicates that one of the pins B0 to B7 has changed.
INTB is normally HIGH, but will go LOW once a change occurs, and that triggers an
interrupt on the UNO, but more on that later when we discuss the inner workings of the
following sketch that runs on the UNO.
// MCP23017, V. Ziemann, 170802
#include <Wire.h>
 Microcontroller: Arduino  75

volatile uint8_t pin_has_changed=0,portB=0,portA=0;

#define MCP23017 0x20
void I2Cwrite(uint8_t addr, uint8_t reg, uint8_t val) {
Wire.beginTransmission(addr); // addr=MCP23017
Wire.write(reg); // register
Wire.write(val); // write value
Wire.endTransmission(true);
}
uint8_t I2Cread(uint8_t addr, uint8_t reg) {
Wire.beginTransmission(addr); // addr=MCP23017
Wire.write(reg); // register
Wire.endTransmission(false);
Wire.requestFrom(addr,1); // request 1 byte
return Wire.read(); // read and return
}
void action() { //..................................action
pin_has_changed=1;
}
void setup() { //....................................setup
Wire.begin(); // Init I2C bus
I2Cwrite(MCP23017,0x00,0x00); // IODIRA=all output
I2Cwrite(MCP23017,0x0D,0xFF); // GPPUB Pullups on
I2Cwrite(MCP23017,0x03,0xFF); // IPOLB reverse polarity
I2Cwrite(MCP23017,0x05,0xFF); // GPINTENB, interrupts on
pinMode(2,INPUT_PULLUP);
attachInterrupt(digitalPinToInterrupt(2),action,FALLING);
Serial.begin(9600); while (!Serial) {;}
}
void loop() { //......................................loop
if (Serial.available()) {
char line[30];
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"A7 ")==line) {
int val=(int)atof(&line[3]);
if (val==0) {portA &= B01111111;} else {portA |= B10000000;}
I2Cwrite(MCP23017,0x012,portA); // write GPIOA
}
}
if (pin_has_changed) {
portB=I2Cread(MCP23017,0x11); // 0x11=INTCAPB, 0x13=GPIOB
pin_has_changed=0;
if (portB & B00000001) {
Serial.println("Button B0 pressed");
}
}
delay(1);
}
In the sketch we first include the Wire.h header file for support of I2C communication and
a few variables. Here we use volatile to indicate that at least one of them can change in an
76  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

interrupt handler. Next we re-use the convenience functions to read and write a register on
a I2C device. The function action() is called once the interrupt is triggered, which happens
as a consequence of a pin changing on the MCP23017 that is subsequently signaled via the
INTB pin going low. In the interrupt handler we only set a variable to indicate that we
need to read the pins, which we later do in the loop() function. In the setup() function,
we initialize I2C communication by calling Wire.begin(), and configure the MCP23017.
Setting register 0x00 to 0x00 configures all pins of port A (those on the right-hand side with
the LED attached to pin A7) as output. We do not need to do that for port B, because
we use the pins in their default configuration as input. Next we enable the internal pull-up
resistors on all pins of port B, and reverse the polarity with which a change is reported.
It is convenient to invert the state, which normally is low if a button is pressed, to pull
the input pin low. We prefer, however, to have the state reported as high. Finally we write
0xFF to register 0x05, which enables interrupts to be generated whenever the state of an
input pin of port B changes. This completes the configuration of the MCP23017, but we
still need to configure the UNO to listen to a change of state on its pin 2. This pin is
configured as an input pin with pull-up enabled, and we attach the function action() to
be executed if pin 2 changes state from high to low; in other words, if it is FALLING. We
complete the configuration by initializing the serial line. In the loop() function, we first
handle commands received over the serial line, and test whether the command starts with
A7. We access the rest of the line by &line[3], because line is an array of characters, and
using the ampersand indicates the address of the array element number 3. Here we expect
an integer, rather than a float value, but it is easy to use the construction (int) fval to
cast the float value fval to an integer. This construction of parsing the rest of the line
with atof() turns out to be a compact and robust way to decipher the numbers, and we
will use it repeatedly. The result is placed in the variable val and if it is zero, we toggle
bit 7 of the variable portA off and write the variable to the output register 0x12, which is
responsible for the state of the output pins of port A, those on the right-hand side. Note that
we use the prepended letter “B” to specify a number in binary representation. In particular,
the LED is connected to the uppermost pin, labeled A7. By toggling the appropriate bit,
we switch it on and off. After handling the Serial communication, we check whether the
variable pin has changed is set, and if that is the case, we read a register with the present
state of port B. We can directly read the present state of port B by reading from the GPIOB
register 0x13. But the MCP23017 has a special feature that internally captures the state of
port B immediately after the interrupt is triggered, and we retrieve that value by reading
from register 0x11, which is called INTCAPB in the datasheet. We then check whether bit 0
is set and address that by sending some text to the serial line. Given this template and
some reading of the datasheet, it should be straightforward to adapt the software to build
a user interface with many buttons and LEDs.
To summarize this section, we point out that we can communicate with the sensors using
basic I2C communication functions such as Wire.read(). In that case, we have to convert
the retrieved raw data to physical quantities ourselves, which requires careful reading of the
datasheets. If a library for the particular sensor is available, we can include it in the Arduino
IDE by copying it to the Arduino/libraries subdirectory and use it after restarting the
IDE. In that case, we strongly advise careful inspection of the examples that are normally
included.
A second communication bus, similar to I2C, is the SPI bus. A number of sensors, but
also other devices, such as analog-to-digital converters, use it, as we shall see in the next
section.
 Microcontroller: Arduino  77

Figure 4.10 Connecting the MCP3304 ADC to the NodeMCU.

4.4.4 SPI
In this section we discuss interfacing the MCP3304 ADC. It supports an SPI interface, and
we use it to expand the number of ADC channels of a NodeMCU microcontroller, which
otherwise is equipped with a single ADC channel only. In Figure 4.10 we show how to
connect the ADC to the NodeMCU on a small solderless breadboard, with the NodeMCU
visible on the right and the ADC on the left. From left to right, the eight input pins A0
to A7 of the ADC are on the side facing downwards. A0 and A1 are used together to form
differential input channel 0. On the upper side, we connect the supply voltage to the top
left (pin 16) of the MCP3304 and ground to pins 9 and 14. The four SPI lines of the chip
are connected to pins D5, D6, D7, and D8 on the NodeMCU: SCL connects to D5, MISO
to D6, MOSI to D7, and chip-select to D8. We use a voltage divider with an 8.2 kΩ and a
68 kΩ resistor to lower the reference voltage to 0.4 V. It is connected with a wire to the Aref
pin, pin 15 on the MCP3304. This increases the resolution of the ADC to 13 bits between
±0.4 V.
In the following we discuss two ways to read the ADC values from the MCP3304. Later
we program the interface routine using the built-in SPI interface libraries, but first we use
a method called bit-banging, which means that we write and read the respective pins in the
correct order, quasi by hand. The following sketch is adapted from the example code on the
Arduino playground.
// MCP3304 bit-banged, V. Ziemann, 170726
#define CS 15 // D8, ChipSelect
#define MOSI 13 // D7, MasterOutSlaveIn
#define MISO 12 // D6, MasterInSlaveOut
#define CLK 14 // D5, Clock
void mcp3304_init_bb() { //.......................mcp3304_init
pinMode(CS,OUTPUT);
pinMode(MOSI,OUTPUT);
pinMode(MISO,INPUT);
78  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

pinMode(CLK,OUTPUT);
digitalWrite(CS,HIGH);
digitalWrite(MOSI,LOW);
digitalWrite(CLK,LOW);
}
int mcp3304_read_bb(int channel) { //...........mcp3304_read_bb
int adcvalue=0, sign=0;
byte commandbits = B10000000;
commandbits|=(channel & 0x03) << 4; // 5 config bits, MSB first
digitalWrite(CS,LOW); // chip select
for (int i=7; i>0; i--){ // clock bits to device
digitalWrite(MOSI,commandbits&1<<i);
digitalWrite(CLK,HIGH); // including two null bits
digitalWrite(CLK,LOW);
}
sign=digitalRead(MISO); // first read the sign bit
digitalWrite(CLK,HIGH);
digitalWrite(CLK,LOW);
for (int i=11; i>=0; i--){
adcvalue+=digitalRead(MISO)<<i;
digitalWrite(CLK,HIGH);
digitalWrite(CLK,LOW);
}
digitalWrite(CS, HIGH);
if (sign) {adcvalue = adcvalue-4096; }
return adcvalue;
}
void setup() { //...................................setup
mcp3304_init_bb();
Serial.begin(115200);
while (!Serial) { delay(10);}
}
void loop() { //....................................loop
int val=mcp3304_read_bb(0);
Serial.print("CH0 = "); Serial.println(val);
delay(1000);
}

In the sketch we first give meaningful names to the pins we use on the NodeMCU, and
declare the function mcp3304 init bb() where we collect the code needed to initialize the
bit-banged SPI communication. We declare the respective pins to be INPUT or OUTPUT and
initialize their state with a call to digitalWrite(). The mcp3304 read bb() function takes
the channel number as input and returns the digitized word. In this routine we assume
that we use two input pins of the MCP3304 in differential mode. Inside the routine we first
declare some variables and initialize the commandbits to start with a most significant bit 7
that has the value 1. The next bit 6 is 0, indicating differential mode, and then we add
the channel information as bits 5 and 4 and set bit 3 to zero, followed by three more zeros,
following the description from the datasheet of how to configure the ADC. Now we can pull
the CS line LOW to indicate that a transaction is about to start, and then clock the five
configuration bits plus two extra bits to allow some time for the conversion by setting the
 Microcontroller: Arduino  79

MOSI pin to the respective value and toggling the CLK pin. Once we complete this we can
read the sign bit from the MISO line and toggle the CLK pin. Then we repeat to read MISO
and toggle CLK 12 times to fill the appropriate bits of the integer adcvalue, and finally
conclude the transaction by pulling the CS pin HIGH. Before returning the adcvalue we
ensure that the sign bit is properly folded into the reading. In the setup() function we only
initialize the serial line and the MCP3304, and in the loop() function we read channel 0
from the ADC and display its value on the serial line. The latter is repeated once a second.
Instead of bit banging the pins, we can also utilize the SPI library that comes with the
Arduino IDE. The sketch, based on the <SPI.h> library, but otherwise equivalent to the
previous one, is the following.
// MCP3304, V. Ziemann, 170726
#include <SPI.h>
#define CS 15 // D8
int mcp3304_read_adc(int channel) { //...0 to 3.......read_adc
int adcvalue=0, b1=0, hi=0, lo=0, sign=0;;
digitalWrite (CS, LOW);
byte commandbits = B00001000; // Startbit+(diff=0)
commandbits |= channel & 0x03;
SPI.transfer(commandbits);
b1 = SPI.transfer(0x00); // always D0=0
sign = b1 & B00010000;
hi = b1 & B00001111;
lo = SPI.transfer(0x00); // input is don’t care
digitalWrite(CS, HIGH);
adcvalue = (hi << 8) + lo;
if (sign) {adcvalue = adcvalue-4096;}
return adcvalue;
}
void setup() {//..............................setup
pinMode(CS,OUTPUT);
SPI.begin();
SPI.setFrequency(2100000);
SPI.setBitOrder(MSBFIRST);
SPI.setDataMode(SPI_MODE0);
Serial.begin(115200);
while (!Serial) {yield();}
}
void loop() { //...........................loop
int val=mcp3304_read_adc(0);
Serial.print("CH0 = "); Serial.println(val);
delay(1000);
}
Here we first include the <SPI.h> functionality and define the CS pin before defining the
mcp3304 read adc function to read a channel from the ADC. In this function we first
declare a number of variables and pull the CS pin LOW in order to start the transaction.
Then we build the commandbits; this time they are aligned such that the start bit is bit 3
and the channel information is stored in bits 1 and 0, and we use the SPI.transfer function
to send it to the ADC. Note that the first bit recognized by the ADC is the first non-zero
bit, which is the start bit 3. We then send 0x00 in a second call to SPI.transfer and receive
80  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

the readings from the MISO pin in b1. Since we have a number of idle clock toggles, the sign
bit is bit 4 and the next four bits are the four most significant bits of the ADC reading,
which we store in the variable hi. The next call to SPI.transfer() returns the subsequent
bits, which are the lower eight bits from the MISO pin, and we store them in the variable lo.
After pulling the CS pin HIGH to conclude the transaction, we build the ADC word adcvalue
and add the sign information. Finally we return the reading to the calling program. In the
setup() function we initialize the SPI functionality by calling the SPI.begin() function,
setting the clock frequency, byte order, and MODE, and then initialize the serial line. The
loop() function is a straight copy of the previous example, where we only read channel 0
and print the value to the serial line.
This external ADC greatly expands the analog input capabilities of the NodeMCU,
which only has a single ADC channel with unipolar 10-bit resolution on board. Now we
have four extra channels at 12-bit resolution, including an additional polarity bit. We note
in passing that the MCP3304 has a close relative, the MCP3208, which has eight unipolar
channels and can be programmed in a similar way. Consult the datasheets for the details.
Sensors that support SPI communication can be controlled in a similar way to the ADCs,
either by bit banging the respective pins or by using the SPI.transfer() function. Again,
reading the detailed specifications in the datasheets is mandatory.
So far, we discussed the standard communication channels, RS-232, I2C, and SPI, but
there are several others that we briefly address in the next section.

4.4.5 Other protocols

The DHT11 and DHT22/AM2302 humidity sensors use a nonstandard communication pro-
tocol, but this is reasonably well explained in the datasheet. The protocol is based on using
a single IO-pin, which requires a pull-up resistor of about 4.7 kΩ. On breakout boards, the
resistor is likely already mounted. Initially the microcontroller configures the pin as OUTPUT
and pulls it to LOW for at least 18 ms, before releasing control to the sensor by reconfigur-
ing the pin to INPUT. The sensor then acknowledges the transaction by sending a pulse of
50 µs LOW followed by 50 µs HIGH before clocking 40 data bits to the pin using the following
convention. A zerobit is defined by a LOW state for 50 µs and a subsequent HIGH state for
about 27 µs. A onebit is defined by a 50 µs LOW state followed by a 70 µs long HIGH state.
The 40 data bits contain the information about the humidity and the temperature. In the
following code we implement this protocol.
// DHT11, V. Ziemann, 170804
#define DHT 2 // sensor pin
float read_dht11(float *T) {
bool p[500]= { 0 };
uint8_t data[5]={0},checksum;
int ic=0,goes_up=0;
pinMode(DHT,OUTPUT);
digitalWrite(DHT,LOW); // 20 ms low pulse
delay(20);
digitalWrite(DHT,HIGH);
noInterrupts(); // interrupts off for accurate timing
delayMicroseconds(20); // make sure we start at low
pinMode(DHT,INPUT);
for (int i=0;i<500;i++) { // read 5 ms worth of data
p[i]=digitalRead(DHT);
 Microcontroller: Arduino  81

delayMicroseconds(10); // one every 10 us

}
interrupts(); // interrupts on again
while (p[ic++] == 0); // next HIGH, acknowledge bit
while (p[ic++] == 1); // next LOW, first data bit
for (int i=0;i<5;i++) { // loop over the
for (int j=7;j>=0;j--) { // 40 data bits
while (p[ic++] == 0) {goes_up=ic;}
while (p[ic++] == 1) ;
(ic-goes_up > 4) ? data[i] |= (1<<j) : 0;
}
}
checksum=((data[0]+data[1]+data[2]+data[3]) & 0xFF);
if (checksum==data[4]) {
*T=(float) data[2]; // temperature
return (float) data[0]; // humidity
}
*T=-100; // only get here if bad reading
return -1;
}
void setup() { //.............................setup
Serial.begin(9600);
while (!Serial) {;}
pinMode(DHT,INPUT);
digitalWrite(DHT,HIGH);
delay(2000); // wait for DHT to wake up
}
void loop() { //...............................loop
float temperature,humidity;
humidity=read_dht11(&temperature);
Serial.print(humidity); Serial.print("\t");
Serial.println(temperature);
delay(2000);
}
In this sketch we first define the pin to which the DHT11 is connected and the function
read dht11() that encapsulates the communication with the sensor. There we first declare
a number of variables and then directly progress to configuring the pin to which the sensor
is connected as OUTPUT, and produce a 20-ms-long LOW pulse before pulling the pin HIGH
and reconfiguring it as INPUT. Since the timing of the response from the DHT11 is critical,
we briefly turn interrupts off, which disables background processing of serial or WLAN
activities. Then we read the pin every 10 µs 500 times for a total duration of 5 ms, which
is sufficient to read 40 bits, each having a maximum duration of 120 µs, into the boolean
array p[]. Once the acquisition is complete, we turn interrupts on again. Now that we have
the entire data set in memory we can decode it. The variable ic steps through the received
array p[] and searches for LOW to HIGH transitions and HIGH to LOW transitions. First we
take care of the acknowledge pulse before looping over 5 bytes of 8 bits each. For each bit
we measure the separation from the time the signal goes up until it goes down again. If
the difference of the indices is more than four, which means that the temporal separation
is longer than 40 µs, we set the corresponding bit in the respective data byte. Once 40
82  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.11 Connecting the DS18b20 sensor to the Arduino UNO.

bits are decoded we calculate the checksum, which is the sum of the first four data bytes,
truncated to eight bits. If this equals the fifth byte, the transmission is correct and the
function returns data[0] as the humidity and data[2] as the temperature, as specified in
the datasheet. If the checksum is incorrect, the function returns values that are obviously
wrong to allow checking in the calling program. Note that the DHT22/AM2302 sensors
have a slightly different interpretation of the data bytes. For these sensors, the humidity
and temperature are instead given by
humidity=(0.1*(256*data[0]+data[1]));
*T=(float) (0.1*(256*data[2]+data[3]));
In our version of the read dht11() function, we first read the entire time series of 500 points
with a spacing of 10 µs before decoding. This wastes a little memory, but is more robust
than waiting for transitions of the pin while decoding, as is done in other implementations
of the DHT libraries. Note also that sampling a pin in this way constitutes a simple logic
analyzer. This may come handy in other contexts as well.
Another data bus, similar to the one for the DHT sensors, but with additional error
handling, is the Dallas 1-wire protocol that is used by the popular DS18b20 temperature
sensor, but also by other devices, such as humidity sensors, memory circuits, or autonomous
data loggers. We will, however, focus on the DS18b20 and connect it to the Arduino using
the schematics shown in Figure 4.11 where we connect ground and supply voltage and
the single data pin to pin D2 on the UNO. We also added a 4.7 kΩ pull-up resistor from
the data pin to the supply voltage. This is needed for a bare sensor only. If we use a
breakout-board version of the DS18b20, the pull-up can likely be omitted, because it is
already mounted on the breakout board. For interfacing to the Arduino, we use ready-
made libraries that encapsulate the low-level interaction. Before using the libraries, we need
to install the OneWire and DallasTemperature libraries. An easy way is to use the library
manager of the Arduino IDE, which is accessible from the Sketch→Include Library→Manage
 Microcontroller: Arduino  83

Libraries, where we enter the keywords OneWire and DallasTemperature. This displays a
list of supported libraries that can be installed directly from the menu. After installation,
we enter the following code, which shows a simple example.
// DS18b20 1-wire temperature sensor, V. Ziemann, 170120
#include <OneWire.h>
#include <DallasTemperature.h>
OneWire oneWire(2); //use pin D2
DallasTemperature sensors(&oneWire);
void setup() {
Serial.begin(9600); while (!Serial) {;}
sensors.begin();
}
void loop() {
sensors.requestTemperatures();
float temp=sensors.getTempCByIndex(0);
Serial.println(temp);
delay(1000);
}
In the sketch, we first include support for the Onewire and DallasTemperature devices,
and initialize the OneWire bus using pin number 2, followed by the initialization of the
DallasTemperature sensors, to use the bus initialized in the previous statement. Note that
we can connect several temperature sensors in parallel on the same bus. In the setup()
function we only initialize serial communication and the sensor. The code in the loop()
function sends out the query to the connected sensors, which causes them to send back their
measurement values. In our example we only connect a single DS18b20 such that we only
can read out the temperature in Celsius from the device number zero, and print the result
to the serial line. After waiting a second we repeat the procedure.
The HC-SR04 distance sensor, shown in Figure 2.34, determines the distance to the
nearest obstacle by emitting a short high-frequency (non-audible) acoustic pulse when its
trigger pin receives a short 10-µs-long trigger pulse. Then it pulls the echo pin to low voltage
and only returns it to the high state once the echo is received. We therefore need to connect
the sensor with four wires (GND, 5V, Trig, Echo) to the Arduino, and need to measure the
duration of the echo pulse. This is accomplished in the following sketch.
// HC-SR04 distance logger, V. Ziemann, 161204
const int trig=2, echo=3;
void setup() {
Serial.begin(9600); while (!Serial) {;}
pinMode(trig,OUTPUT);
pinMode(echo,INPUT);
}
void loop() {
float duration,distance;
digitalWrite(trig,LOW); // make a 10 us trigger pulse
delayMicroseconds(2);
digitalWrite(trig,HIGH);
delayMicroseconds(10);
digitalWrite(trig,LOW);
duration=pulseIn(echo,HIGH); // wait for echo
84  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

distance=100*duration*340e-6/2; // in cm
Serial.println(distance);
delay(1000);
}
In the sketch, we first declare the constants with the pins used in this sketch. The trigger
pin of the sensor is connected to pin 2 on the Arduino and the echo pin on the sensor
to pin 3. In the setup() function, we declare the serial line and whether the respective
pins are used as input or output. In the loop() function, we first declare the variables for
distance and duration and then create a 10-µs-long trigger pulse by writing HIGH and LOW
to the trigger pin, and wait a short time in between. Once the trigger pulse is dispatched,
we use the pulseIn() function to wait for the echo pin to return to the HIGH state. The
pulseIn() function returns the elapsed time in microseconds such that 1000 µs correspond
to a distance of about 0.17 m, or 17 cm if we assume a value for speed of sound of 340 m/s
and note that the sound has to go back and forth from the sensor to the first obstacle. Then
the measured value is sent over the serial line to the host computer, and the whole process
is repeated after some delay. We note in passing that we can use the pulseIn() function
to measure the duration of short pulses in general.
Output pins of a rotary encoder carry a sine-like and a cosine-like signal, respectively.
We can determine whether the shaft is turned clockwise or counterclockwise depending on
whether the two pins are equal or not while one of them changes on a falling edge, going
from a HIGH to a LOW state. The following code implements this method using interrupts.
Using interrupts has the advantage that the state of the pins does not have to be measured
continuously and compared, but an interrupt handler is registered with a certain action,
in our case on a falling edge on the pin connected to the interrupt. The Arduino has two
pins, numbers 2 and 3, that have the interrupt functionality. But let us look at the sketch
first.
// Rotary encoder, V. Ziemann, 161205
const int pinA=2,pinB=4;
volatile int pos=0;
void setup() {
Serial.begin(9600); while (!Serial) {;}
pinMode(pinA,INPUT_PULLUP);
pinMode(pinB,INPUT_PULLUP);
attachInterrupt(0,interrupt_handler,FALLING);//0=pin2,1=pin3
}
void loop() {
Serial.println(pos);
delay(1000);
}
void interrupt_handler() {
if (digitalRead(pinA)==digitalRead(pinB)){pos++;}else{pos--;}
}
The sketch starts by declaring the used pins and a variable pos that contains the encoder
position. The variable is declared volatile, which instructs the compiler that it can change
asynchronously within the interrupt routine, and prevents the compiler from optimizing it
away, because it does not change in the main program. In the setup() function we declare
the serial line parameters and the input pins with pull-up resistors enabled. One of the pins
must be pin 2 or 3, those with interrupt capability. The attachInterrupt() function is
 Microcontroller: Arduino  85

Oscilloscope trace of a pulse-width modulated signal on pin D9 of an

Figure 4.12
Arduino UNO with values of 88 (left) and 220 (right).

used to connect a so-called callback function, here interrupt handler. It is defined at the
end of the sketch. In our sketch we chose interrupt 0 (connected to pin 2) to trigger the
callback function interrupt handler. We also specify that the callback function is called
on a FALLING edge on the interrupt pin. Instead of hard coding the value 0 as the source
of the interrupt, we can use the service function digitalPinToInterrupt() that takes the
pin number, 2 in our case, as argument, such that the line reads
attachInterrupt(digitalPinToInterrupt(2),interrupt_handler,FALLING);
instead. Using digitalPinToInterrupt() is the preferred way, because it relieves the user
of remembering which interrupt number is connected to which pin. The loop() function
is very simple and only writes the position of the encoder shaft to the serial line once a
second. Finally, we define the callback function interrupt handler, which increments the
pos variable by one step if the two encoder pins are equal, and decrements otherwise. In
this way, the sense of rotation of the shaft is taken care of. We need to point out that we
had to use separate external pull-up resistors (10 kΩ) to guarantee stable operation, and in
the sketch we may encounter a so-called race condition when the variable pos changes while
it is transmitted on the serial line.
After the discussion of interfacing the sensors, we now move on to investigate how to
work with actuators, such as switches, motors, or analog voltages.

4.5 INTERFACING ACTUATORS

In this section we discuss how to interface the actuators discussed in Chapter 3 to the
Arduino, and start with the simplest actuator, the LED.

4.5.1 Switching devices

In Section 4.3 we used the digitalWrite() function to turn an LED on and off. If we
connect a transistor, Darlington driver, or H bridge to an output pin, we can use the same
function to handle larger current than the 20 mA the Arduino can drive. Consult the circuits
shown in Section 3.1.2 for details.
Dynamically adjusting the brightness of an LED is achieved by pulse-width modulation,
and that feature is available on pins D3, D5, D6, D9, and D11 on the Arduino Uno. We use
the analogWrite(pin,value) function that takes the pin number and a value between 0
and 255 (0 to 1023 on the ESP8266) to adjust the duty cycle of the pulse-width modulated
signal from completely off to completely on. The switching frequency is between 500 and
86  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

1000 Hz, and in Figure 4.12, we show an oscilloscope trace of the result of analogWrite()
to pin D9 of a UNO with values 88 and 220, respectively. The frequency of the signal is
about 490 Hz, and we see the length of the signal differing from about 1/3 of the time at
5 V on the left to 5 V almost all the time on the right.
In order to be able to manipulate actuators in the same way as the sensors, we use the
same query-response protocol to communicate the required actions to the microcontroller.
We want to use the convention that DWx 0 turns digital pin number x off and DWx 1 turns
it on, and that sending AWx nnn sets pulse-width modulation on pin number x to value nnn.
The code that achieves this is the following:

// Switching_and_pwm, V. Ziemann, 170614

char line[30];
void setup() {
pinMode(2,OUTPUT);
Serial.begin (9600);
while (!Serial) {;}
}
void loop() {
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"DW2 ")==line) {
int val=(int)atof(&line[4]);
digitalWrite(2,val);
} else if (strstr(line,"AW9 ")==line) {
int val=(int)atof(&line[4]);
analogWrite(9,val);
} else {
Serial.println("unknown");
}
}
}
where we first declare the variable line that contains the characters received on the serial
line. In the setup() function, we only declare pin 2 as output, and initialize the serial
communication, exactly as before. In the loop() function, we use the previously seen con-
struction to parse the message received on the serial line and perform some action, depending
on the command received. If the command starts with DW2 we read the characters following
the first 4 into the variable val and pass val to the digitalWrite() function. Here we
employ the same method of using the atof() function to decode the rest of the character
array line[] that we first used in the sketch to interface the MCP23017 IO extender. We
immediately encounter the atof construction again when parsing the value received after
AW9. We then use it to set pin D9 to the desired value with the analogWrite() function. In
this way we have a simple mechanism to instruct the Arduino to turn pins on and off and
adjust those that have pulse width capability to the desired duty factor.
We note that the code above does not contain graceful handling of errors. For example,
passing any value val received from the Serial line to the digitalWrite() function when
handling the DW2 command can result in weird behavior if val is not 0 or 1. Even a
negative number could be written to the pin, which is meaningless. Replacing that line by
the following construction will only write meaningful values to the pin. Implicitly, we use
the convention that 0 turns the pin off, and anything else turns it on.
 Microcontroller: Arduino  87

Using pulse-width modulation and a transistor to adjust the brightness

Figure 4.13
of an LED.

// digitalWrite(2,(int)val);
if ((int)val == 0) {
digitalWrite(2,LOW);
} else {
digitalWrite(2,HIGH);
}
Of course, other ways of catching errors can be implemented, but that goes beyond the
scope of our presentation.
Testing the above program with a UNO is quickly done. In Figure 4.13 we show an
implementation on a small breadboard. Two cables connect the power lines to the circuit,
where the ground signal connects to the emitter of the NPN transistor. Here we use a
BC547, but any small-signal NPN transistor will work. The collector of the transistor is
directly connected to the cathode of the LED, and the anode of the LED is connected via
a 220 Ω to the positive supply voltage. The base of the transistor is connected via a 1 kΩ
resistor to the controlling pin, here pin D9, on the Arduino.
This example covers the basic functionality of turning on and off, as well as controlling
the power delivered to a load, the LED. In the next example we use very similar circuitry
to control speed and direction of motors.

4.5.2 DC motors
If we only want to control the speed of a very small motor, we can replace the LED and the
220 Ω resistor in the previous example by the motor. We also need to add flyback diodes to
prevent the back-emf from damaging the transistor. For slightly larger motors, we need to
use a transistor with a higher power rating, such as a TIP-120 Darlington power transistor.
The TIP-120 has a flyback diode from emitter to collector already built in. Note that the
performance of larger transistors normally degrades at higher frequency. Inspection of the
datasheet, however, shows that this only affects frequencies well above 10 kHz.
88  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Controlling the speed of a motor with pulse width modulating the base
Figure 4.14
of a TIP-120 Darlington transistor. See the text for a discussion of the diodes.

In Figure 4.14 we show the setup with a UNO controlling a small motor. The termi-
nals of the transistor are base, collector, and emitter, from left to right, and the motor is
connected between the collector and the supply voltage. The emitter is directly connected
to ground, and we include an external flyback diode (horizontally mounted) between the
motor leads, with the cathode pointing towards the right. The vertically mounted diode
illustrates the connection of the built-in diode. We control the speed of the motor by pulse
width modulating the base of the transistor that is connected to pin D9 on the UNO via a
1 kΩ resistor.
Controlling the speed is useful, but often we also need to control the direction of rotation;
for example, to drive a vehicle not only forward, but also backwards. The standard way to
accomplish this feature is by using an H bridge, as discussed in Chapter 3. Here we use

Figure 4.15 The pin assignment of the L293D H-bridge driver.

 Microcontroller: Arduino  89

Figure 4.16 Controlling speed and direction of a DC motor.

an L293D H bridge driver, which is called a quadruple half-bridge driver, a term that will
become clear by considering the pin layout of the 16-pin chip, shown in Figure 4.15. There
are ground pins and two supply voltages, one for logic levels and one to supply power to
the motor. Then there are four (quadruple) inputs with corresponding outputs. The circuit
works in such a way as to translate the logic level on the input pins to the corresponding
output pins, which in turn deliver the voltage level from the motor power supply to the
motor. Thus, setting input 1 to HIGH and input 2 to LOW connects one motor lead to the
motor supply voltage and the other to ground, which causes the motor to turn in one
direction. Setting input 1 to LOW and input 2 to HIGH will cause the converse, and the motor
will turn in the other direction. Two inputs and outputs therefore provide the functionality
of a single H bridge. Since there are four inputs and outputs, we can use the L293D to
implement two H bridges. Moreover, pin 1 on the L293D can be used to enable the outputs
of input 1 and 2. Applying a pulse-width modulated signal to this pin will therefore work
as speed control for the motor. Pin 9 provides the same functionality for inputs 3 and 4.
Finally, we remark that the character D in L293D indicates that flyback protection diodes
are already built into the integrated circuit.
With the theoretical background covered we are ready to build a motor controller with
an L293D on a breadboard, and control it with an Arduino UNO. We show the circuit in
Figure 4.16. The DC motor is shown on the left and the L293D is the only component on
the small breadboard. The orientation of the chip is the same as in Figure 4.15, making the
functionality easy to understand. The pin numbering goes from top left (pin 1) to bottom
left (pin 8), continuing via bottom right (pin 9) to the top right (pin 16), which is the
standard for most integrated circuits.
The upper two wires connect ground and the supply voltage to the chip. Here we assume
that the supply voltage is the same, and we connect pin 8 of the L293D to the positive logic
supply voltage. Moreover, we disable the output on the right-hand side of the chip by pulling
pin 9 to ground. Two wires connect digital outputs D2 and D3 on the Arduino to the input
of two half-bridges on pins 2 and 7 on the L293D. The corresponding outputs from pins 3
and 6 are connected to the motor cables. Finally, pin 1 on the L293D is connected to the
pulse-width modulated output pin D9 on the UNO.
We want control the motor from the Arduino by sending commands via the serial line
according to the protocol that FW nnn sets the speed in the forward direction and BW nnn
90  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

sets the speed in the reverse direction. Anything else will stop the motors. This is easy to
do with the following sketch.
// H bridge DC motor Controller, V. Ziemann, 170614
char line[30];
void setup() {
pinMode(2,OUTPUT);
pinMode(3,OUTPUT);
Serial.begin (9600);
while (!Serial) {;}
}
void loop() {
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"FW ")==line) {
digitalWrite(2,LOW);
digitalWrite(3,LOW);
digitalWrite(3,HIGH);
float val=atof(&line[3]);
analogWrite(9,(int)val);
} else if (strstr(line,"BW ")==line) {
digitalWrite(2,LOW);
digitalWrite(3,LOW);
digitalWrite(2,HIGH);
float val=atof(&line[3]);
analogWrite(9,(int)val);
} else { // STOP in all other cases
digitalWrite(2,LOW);
digitalWrite(3,LOW);
analogWrite(9,(int)0);
}
}
}
The setup() function only declares the pins used to control the H bridge as outputs and
initializes the serial line. In the loop() function we employ the standard query-response
mechanism. A command on the serial line that starts with FW first turns the motor off as
a precaution against having both outputs positive, which might short the motor supply
voltage. Then pin D3 is pulled high, resulting in the motor turning one way. Finally we
parse the rest of the command line and write the value to set the motor speed by pulse-
width modulating the enable pin of the motor-driver. Any command received, apart from
FW n and BW n, will disable both outputs and sets the speed to zero. Of course, we can
easily implement other commands, such as STOP, that will cause the motor to stop.
Even more elaborate commands can be programmed into the controller, such as turning
one way at some time for a given time, then stopping and turning backwards for another
time. The following code fragment may serve as an example.
} else if (strstr(line,"BACKANDFORTH")==line) {
digitalWrite(2,LOW); // turn all off
digitalWrite(3,LOW);
digitalWrite(2,HIGH); // chose one direction
 Microcontroller: Arduino  91

Figure 4.17 Connecting a model-servo to the Arduino UNO.

analogWrite(9,255); // full speed

delay(500);
digitalWrite(2,LOW); // stop
delay(1000);
digitalWrite(3,HIGH); // other direction
delay(500);
digtalwrite(3,LOW); // stop
analogWrite(9,0); // stop PWM
} ...
Other examples are to move for a short time, do some measurements, using code examples
from the previous section, and continue doing the same for a number of steps. Yet another
example is moving until a limit switch is engaged, which will cause the motors to turn off,
maybe do some measurement, and later return to the parking position at a second limit
switch.
Or, what about continuously reading a sensor and stopping once a certain value is
achieved? This is actually what a model-servo does. So, let us look at it more closely.

4.5.3 Servos
As we discussed in Chapter 3, servos are used to carefully change the position or orientation
of some device, such as the rudder of a boat or the steering system of a radio-controlled car.
Servos only need three wires to be connected: ground and the supply voltage as well as one
wire that carries the control information according to the timing shown in Figure 3.9. In
Figure 4.17 we show how to connect a small servo to an Arduino UNO. If the servo requires
more current or other voltages than the UNO provides, one can use a suitable external power
supply. The ground cables of the external supply and the UNO need to be connected, and
one wire from the servo, which is often red, needs to be connected to the positive terminal
of the external supply instead of the 5 V power pin on the Arduino. Controlling a servo
from the Arduino is rather simple, and we use the following code to do that.
// Servo controller, V. Ziemann, 170614
#include <Servo.h>
92  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Servo myServo;
char line[30];
void setup() {
myServo.attach(9); // pin D9
Serial.begin (9600);
while (!Serial) {;}
}
void loop() {
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"SERVO ")==line) {
float val=atof(&line[6]);
myServo.write((int) val); // 0 to 180
}
}
}
The code follows the normal template with opening serial line in the setup() function and
the query-response construction in the loop() function. Servo-specific means the inclusion
of the <Servo.h> file and the declaration of a Servo object that we call myServo. In the
setup() function, we attach the servo functionality to pin D9 on the Arduino and write new
values to the servo with the call to myServo.attach(). The latter function takes a value
between 0 and 180 as argument and moves the arm of the servo to the desired position.
Note that <Servo.h> is part of the standard Arduino distribution. Note also that we can
use any pin on the Arduino to control servos, even multiple servos, but that using the servo
functionality disables the pulse-width modulation feature of the analogWrite() function
on pin D9 and D10 because it uses the same hardware timer.
Using servos in more advanced ways, such as scanning, is also easy to implement. The
following code fragment scans back and forth through the entire range
} else if (strstr(line,"SCAN")==line) {
for (int pos=0;pos<180;pos+=1) {
myServo.write(pos); delay(10); // and do a measurement
}
for (int pos=180;pos>=0;pos-=1) {
myServo.write(pos); delay(10);
}
and it is easy to envision including some measurement inside the loops that perform the
scanning. Imagine an HR-SR04 distance sensor attached to the servo. Scanning in a semi-
circular motion and continuously measuring the distance to the closest object mimics the
functionality of radar by using a sonar-like device instead. This could serve as a collision
detection system for moving vehicles.
The last type of motors that we consider are stepper motors, and they come next.

4.5.4 Stepper motors

As we discussed before, stepper motors can be very accurately controlled by a given pulse
sequence that turns the shaft by a desired number of discrete steps. The coils inside the
motors need to be excited in certain patterns, either by a unipolar or bipolar power supply.
We start with the unipolar type.
 Microcontroller: Arduino  93

Figure 4.18Connecting a unipolar stepper motor to the UNO with a ULN2003 Dar-
lington driver.

In Figure 4.18 we show the connection of a unipolar stepper motor with the help of a
ULN2003 Darlington driver that is mounted on the small breadboard between the Arduino
UNO and the motor. The inner connection of the ULN2003 is very simple. Pins 1 to 7 on
the left-hand side are the bases of seven Darlington transistors with resistors built into the
chip. The facing pins on the right-hand side are the collectors of the respective transistors.
All emitters are wired to pin 8 on the lower left-hand side of the chip, which is also ground.
Pin 9 on the lower right-hand side is the terminal for the motor power supply, and is often
connected to an external power supply that provides adequate current and voltage to power
the motor. In the figure we assume that the stepper motor is small, and we connect pin 9 to
the 5 V pin on the Arduino with one wire connected to the left-hand side on the Arduino.
Ground is connected with the other wire connected to the left-hand side. The bases of the
Darlington transistors are directly connected to pins D2 to D5 of the Arduino and need to
be controlled by the program.
The center tap of the motor coils are connected to the positive terminal of the motor
power supply with two wires. The other four wires are hooked up to the collectors of
the upper four Darlington transistors in the ULN2003. Since the center tap of a coil is
connected to positive supply voltage, placing a positive voltage on the base of a Darlington
will cause the collector–emitter link to conduct and a current to flow through the coil. In
this way we can excite the coils in a suitable pattern to rotate the shaft either clockwise or
counterclockwise. The sketch that implements this functionality using the query-response
protocol is the following.
// Stepper controller, V. Ziemann, 170616
#include <Stepper.h>
Stepper myStepper(200, 2, 4, 3, 5);
char line[30];
void setup() {
myStepper.setSpeed(60);
Serial.begin (9600);
while (!Serial) {;}
}
void loop() {
94  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.19 Connecting a bipolar stepper motor to the UNO with an L293D H-bridge
driver.

if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"MOVE ")==line) {
int val=(int)atof(&line[5]);
myStepper.step(val);
} else {
Serial.println("unknown");
}
}
}

Here we first include the <Stepper.h> library that comes with the Arduino IDE and declare
a Stepper object called myStepper. The first argument is the number of steps per revolution
and the next four integers are the pins to which the four coils are attached. This may
differ from motor to motor, and normally this information is provided in the datasheet.
Otherwise, a little experimenting with swapping numbers may lead to a moving motor. In
the setup() function, we use the setSpeed() method to declare how fast we want the
motor to rotate, and we also initialize the serial line. In the loop() function we employ our
standard query-response protocol and make the motor respond to the command MOVE, and
read the characters behind the command as the number of steps to move. This number can
be either positive or negative, depending on the desired direction to rotate.
The advantage of using a bipolar stepper motor is increased torque for the same supply
voltage, because there is one coil pulling the permanent magnet of the rotor and the other,
which has opposite polarity, is pushing. This makes it worthwhile to describe how to drive
a bipolar stepper motor.
We connect a bipolar stepper motor in the way shown in Figure 4.19. Here we use an
L293D H-bridge driver, the same integrated circuit we use to provide direction control for
DC motors. Based on the pin-description for the circuit shown in Figure 4.15, we immedi-
ately see that two supply wires from the left-hand side of the UNO connect to the pins for
logic supply voltage and ground, and the enable pins are connected with the light-colored
wire to pin D6 on the UNO. The wires connected to pins D2, D3, D4, and D5 on the Arduino
 Microcontroller: Arduino  95

are connected to the inputs of the four half-bridge drivers on the L293D. The four leads
of the bipolar stepper motor are wired to the four output terminals of the corresponding
drivers. In this circuit the terminal for the positive motor supply voltage on pin 8 of the
L293D is connected to an external power supply. Since internally all four ground pins of the
chip are connected, we can add a wire to connect the ground terminal on the other side of
the L293D of the motor to the ground terminal of the power supply.
It is remarkable that we can use the same code we used for the unipolar stepper motor
for the bipolar motor as well. The same patterns, sent to the driver input pins, causes the
bipolar motor to operate in full step mode. We can therefore use the same Arduino sketch
as before to operate the motor, with the small reservation that we may need to swap the
assignment of pins in the declaration of myStepper.
We need to point out that at least one coil is always excited, and that causes the motor
to get warm over time, even when it is not moving. This problem can be alleviated by adding
a command to turn off all coils, albeit at the expense of losing some accuracy, because the
holding torque from the excited coil is absent. If we can accept the small loss in precision,
we can use the pin of the Arduino connected to the yellow wire to turn on and off the
outputs of the driver chip. It must, however, be pulled high with digitalWrite(6,HIGH)
for normal operation.
Using the built-in libraries to control stepper motors is the quickest way to get started,
and using well-tested libraries normally leads to robust code, but, at any rate, it is rather
instructive to code the step sequence for the stepper motors by hand so to speak. And that
is what we do in the following example, which implements the same functionality as the
previous one, but provides insight into the inner workings of the stepper library. Moreover,
the built-in library only implements full-step operation of the motor, where the permanent
magnet is directly facing a coil, such as permanent magnet 1 in Figure 3.10, which directly
faces coil A. However, by exciting coil pairs simultaneously, we can also fix the permanent
magnet halfway between the coils, and move the shaft by half a step. We previously discussed
this half-step mode of operation in Section 3.2.3.
For the following exercises, we assume that we have a bipolar stepper motor connected
to the Arduino with an L293D H bridge, as shown in Figure 4.19. We first implement the
same functionality that the stepper library provides, and later show how to easily add other
modes. The full-step mode is implemented in the following code:

// Stepper controller by hand, V. Ziemann, 170624

char line[30];
int settle_time=2; // milli-seconds
int stepcounter=0;
const int PA=2;
const int PB=3;
const int PC=4;
const int PD=5;
const int ENABLE=6;
void set_coils(int istep) { //........full-step mode
bool patA[]={1,1,0,0}; // or {1,0,0,0}
int pat_length=4;
int ii;
istep=istep % pat_length;
if (istep < 0) istep+=pat_length;
digitalWrite(PA,patA[istep]);
ii=(istep+2) % pat_length;
96  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

digitalWrite(PB,patA[ii]);
ii=(istep+3) % pat_length;
digitalWrite(PC,patA[ii]);
ii=(istep+1) % pat_length;
digitalWrite(PD,patA[ii]);
delay(settle_time);
}
void setup() { //...........................setup
Serial.begin (9600);
while (!Serial) {;}
pinMode(PA,OUTPUT);
pinMode(PB,OUTPUT);
pinMode(PC,OUTPUT);
pinMode(PD,OUTPUT);
pinMode(ENABLE,OUTPUT);
digitalWrite(ENABLE,HIGH);
}
void loop() { //...............................loop
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"MOVE ")==line) {
int steps=(int)atof(&line[5]);
if (steps > 0) {
for (int i=0;i<steps;i++) set_coils(stepcounter++);
} else {
for (int i=0;i<abs(steps);i++) set_coils(stepcounter--);
}
} else if (strstr(line,"STEPS?")==line) {
Serial.print("STEPS "); Serial.println(stepcounter);
} else if (strstr(line,"STEPS ")==line) {
stepcounter=(int)atof(&line[6]);
} else if (strstr(line,"DISABLE")==line) {
digitalWrite(ENABLE,LOW);
} else if (strstr(line,"ENABLE")==line) {
digitalWrite(ENABLE,HIGH);
}
}
}

First we declare some global variables, such as settle time, the time to wait between
changing coil excitations, and the stepcounter, which keeps track of the distance traveled.
PA through PD are the pins to which the coil terminals are connected, and ENABLE connects
to the enable pin of the H-bridge driver and can be used to de-excite all coils to prevent
them from overheating. Then the set coils() function, which will excite the coils in the
correct pattern, is declared. Inside the function, we first define the excitation pattern patA
for coil A. In this case it is 1100, the pattern in which two coils are always excited, and
and provide the larger torque. Alternatively we can also declare the pattern 1000 which
would result in the single-coil excitation pattern. See Section 3.2.3 for a discussion. The
variable pat length is the number of different steps, four in this case. Next we calculate
the remainder with respect to pat length of the input variable istep to determine what
 Microcontroller: Arduino  97

the new state of the pattern excitation is, and write the corresponding entry in the array
patA to the output that is connected to terminal PA. Next we calculate the entry in patA
that is shifted by two time slots, also modulo the pat length, and write the corresponding
entry to output pin PB. In the same fashion, we set the remaining two output pins, PC and
PD, to the entry in patA that is shifted by one or three time-slots, respectively. Finally, we
wait a short time, given by the variable settle time. This function implements a single
step, but in order to move a larger number of steps, we have to call this function repeatedly.
The remainder of the sketch consists as usual for Arduino sketches of a setup() function,
where we initialize serial communication and declare the output pins to control the motor
as OUTPUT. Finally we enable the motor driver, by setting the ENABLE pin HIGH. In the
loop() function we use the often-used construction to read from the serial line, and decode
the command with the strstr() function. If the command MOVE is received, we decode
the rest of the line as the integer steps, the number of steps we want to move the stepper
motor. This number can be positive or negative, depending on the desired direction in
which to move. If it is positive, we call the set coils() function the required number of
times while incrementing the variable stepcounter, in order to keep track of the currently
applied step as well as the accumulated number of steps. If steps is negative, we call
set coils() the necessary number of times while decrementing the variable stepcounter.
The other commands read and set the stepcounter variable with STEPS? or STEPS nnn,
respectively. The commands ENABLE and DISABLE turn the driver stage of the L293D on
and off, which may be convenient to prevent overheating of the coils during long times of
idleness. This program moves the motor back and forth. Implementing speed control by
adjusting settle time is left as an exercise.
The previous sketch implements full-step mode, but changing it to half-step mode only
requires changing the set coil() function to produce the pattern for half stepping, which,
according to Section 3.2.3, is given by 11100000. The replacement for set coil() imple-
menting half-step mode is

void set_coils(int istep) { //.............half-step mode

bool patA[]={1,1,1,0,0,0,0,0};
int pat_length=8;
int ii;
istep=istep % pat_length;
if (istep < 0) istep+=pat_length;
digitalWrite(PA,patA[istep]);
ii=(istep+4) % pat_length;
digitalWrite(PB,patA[ii]);
ii=(istep+6) % pat_length;
digitalWrite(PC,patA[ii]);
ii=(istep+2) % pat_length;
digitalWrite(PD,patA[ii]);
delay(settle_time);
}
which is very similar to the previously discussed full-step version. The only difference is the
array patA, which now contains the excitation pattern for the half-step mode. The variable
pat length is increased correspondingly to eight, and the steps to excite the other pins are
4, 2, and 6. Remember, this corresponds to 180, 90, and 270 degrees of the eight time-slots.
Depending on the sense of wiring for the coils, we may have to swap the cables, or, as an
alternative, exchange the step increment in the set coils() function. Now we can full step
98  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Connecting a bipolar stepper motor to the UNO with a DRV8825 stepper

Figure 4.20
motor driver with microstep capability.

and half step the motors, even do so by hand, but for the micro-stepping modes we need
additional hardware to control the excitation of the coils more accurately.
A driver circuit that implements these microstepping modes is the DRV8825. Interfacing
a bipolar stepper motor with an Arduino using this driver is shown in Figure 4.20. Before
connecting the motor to the circuit, we need to match the driver to the motor by adjusting
the maximum current by which the coils are excited. So we start from the configuration
shown in Figure 4.20, but without the four leads to the motor in place. The motor power
supply must provide between 8.2 and 40 V, and we add a 100 µF capacitor to stabilize the
supply voltage. We then connect the, normally black, ground lead of a multimeter to digital
ground on the motor driver and connect the other, normally red, lead of the multimeter to
the tip of a screwdriver. We use the latter to adjust the potentiometer on the top left of
the driver breadboard until the voltage shown on the multimeter is half the desired current
limit for the motor, as specified in the datasheet of the DRV8825. We illustrate this is
Figure 4.21, where we see the multimeter on the bottom left and the DRV8825 placed on
a small breadboard with the screwdriver touching the potentiometer on the driver board.
The multimeter in this case shows a voltage of 0.56 V, which implies that the current limit
for the driver is 1.12 A.
Once we have set the maximum current, we connect the motor to the terminals labeled
1A, 1B and 2A, 2B on the driver circuit. Here we chose to connect the first and third
wires on the motor to the B-pins, and the second and fourth to the A-pins, which accounts
for the crossed wires on the lower motor coils. Ground and positive voltage of the motor
power supply are connected to the two terminals on the top right, and the digital ground is
connected to the Arduino ground connector with the black wire. One wire connects UNO
pin D2 to the direction pin at the lower left of the DRV8825, and another wire connects D3
to the step input. A wire, connected to pin D4, is used to select the microstepping mode.
Several modes such as full-step, half-step, 1/4, 1/8, 1/16, and 1/32 stepping are available.
See the datasheet for the details. Here we only implement two modes. If the three mode pins
M0, M1, and M2 are pulled to ground, the driver operates in full-step mode. If the three
mode pins are pulled to the logic supply voltage, the microstepping mode with 32 microsteps
 Microcontroller: Arduino  99

Figure 4.21 Adjusting the maximum current on the DRV8825 breadboard.

is selected. Thus by toggling UNO pin D4 we can switch between full- and micro-stepping
mode. The enable pin on the top left of the driver is left unconnected, because it is internally
pulled low to enable the driver by default. Moreover, one wire, connected to the 5 V supply
of the UNO, pulls the reset and sleep pins of the driver high, thus permanently enabling
the driver.
In order to move the motor, we use a sketch for the Arduino UNO that suitably changes
the states of the direction, step, and mode pin. This is done with the following code.
// Stepper controller with DRV8825, V. Ziemann, 170626
char line[30];
int settle_time=30,stepcounter=0;
const int DIR=2; // direction pin
const int STEP=3; // step pin
const int MODE=4; // mode pin, LOW=FULLSTEP, HIGH=MICROSTEP
void setup() { //................................setup
Serial.begin (9600);
while (!Serial) {;}
Serial.println("starting");
pinMode(DIR,OUTPUT); digitalWrite(DIR,LOW);
pinMode(STEP,OUTPUT); digitalWrite(STEP,LOW);
pinMode(MODE,OUTPUT); digitalWrite(MODE,HIGH);
}
void loop() { //..................................loop
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"MOVE ")) {
int steps=(int)atof(&line[5]);
if (steps > 0) {
digitalWrite(DIR,LOW);
for (int i=0;i<steps;i++) {
100  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

stepcounter++;
digitalWrite(STEP,HIGH);
delayMicroseconds(settle_time);
digitalWrite(STEP,LOW);
delayMicroseconds(settle_time);
}
} else {
digitalWrite(DIR,HIGH);
for (int i=0;i<abs(steps);i++) {
stepcounter--;
digitalWrite(STEP,HIGH);
delayMicroseconds(settle_time);
digitalWrite(STEP,LOW);
delayMicroseconds(settle_time);
}
}
} else if (strstr(line,"STEPS?")) {
Serial.print("STEPS "); Serial.println(stepcounter);
} else if (strstr(line,"STEPS ")) {
stepcounter=(int)atof(&line[6]);
} else if (strstr(line,"WAIT?")) {
Serial.print("WAIT "); Serial.println(settle_time);
} else if (strstr(line,"WAIT ")) {
settle_time=(int)atof(&line[5]);
} else if (strstr(line,"MICROSTEP")) {
settle_time=30;
digitalWrite(MODE,HIGH);
} else if (strstr(line,"FULLSTEP")) {
settle_time=2000;
digitalWrite(MODE,LOW);
}
}
}
At the top of the sketch we declare a number of variables, such as the settle time and the
stepcounter, but also the used pins. In the setup() function we initialize serial commu-
nication, declare the control pins for the stepper driver as output, and initialize their state.
By pulling the mode pin high, we select the microstepping mode. The loop() function uses
the query-response construction in which the first command deciphers the MOVE command
and interprets the characters following it as the number of steps. Then the code branches,
depending on the sign of steps. If it is positive, we pull the direction pin HIGH and then
increment the stepcounter. Then the STEP pin is pulled high and low, with a small delay
in between. Note that we use the delayMicroseconds() function, because the rotor moves
very little between the microsteps, and we can afford to reduce the waiting time between
steps to achieve smooth motion of the motor. But in any case, the values for the waiting
time should be adapted to the actual motor connected. If steps is negative, we pull the
direction pin LOW and decrement the stepcounter before toggling the STEP pin first HIGH
and then LOW. After handling the MOVE command, we add the bookkeeping commands to
read and set both stepcounter and settle time, and finally, select the operation mode.
If the command is MICROSTEP, the settle time is set to 30 µs and the MODE pin is pulled
 Microcontroller: Arduino  101

Figure 4.22 Connecting an MCP4921 12-bit DAC to the Arduino UNO.

HIGH. If the command is FULLSTEP, we choose a larger settle time and pull the MODE pin
LOW.
After being able to control various types of motors, we move on to controlling analog
voltages.

4.5.5 Analog voltages

In this section we connect an MCP4921 12-bit digital-to-analog converter, which supports
SPI-communication, to an Arduino UNO, and discuss a sketch to set the output voltage
of the DAC using our query-response protocol. This could, for example, be used to set the
control voltage of a power supply or any other device that requires an analog voltage level
as input. The circuit that achieves this is shown in Figure 4.22, where we see the UNO on
the left and the supply and ground connections to the MCP4921. The bright wire from pin
D13 connects the respective clock pins, the wire from pin D11 connects the MOSI pins.
It carries the data from the UNO to the DAC. The cable from pin D10 connects the chip
select pins. In this case there is no information read back from the DAC to the UNO, and
we do not need a MISO wire. As a matter of fact, there is not even a MISO pin on this
DAC. On the DAC we connect the reference voltage pin to the 5 V supply voltage and pull
the LDAC pin low to transfer the new voltage immediately to the output at the end of the
SPI transaction. We also add a wire from the output pin of the DAC to analog input A0 of
the UNO, which allows us to read back and verify the output voltage. The software running
on the UNO is the following.
// MCP4921 DAC, V. Ziemann, 170801
#include <SPI.h>
#define CS 10
void setup() { //...............................setup
Serial.begin(9600);
102  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

while (!Serial){;}
pinMode(CS,OUTPUT); digitalWrite(CS,HIGH);
SPI.begin();
SPI.setBitOrder(MSBFIRST);
}
void loop() { //.................................loop
char line[30];
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"DAC ")==line) {
uint16_t val=(int)atof(&line[3]);
val|=(B0011 << 12); // 0=chA,0=unBuf,1=x1,1=ON
digitalWrite(CS,LOW);
SPI.transfer(highByte(val));
SPI.transfer(lowByte(val));
digitalWrite(CS,HIGH);
} else if (strstr(line,"A0?")==line) {
Serial.print("A0 ");
Serial.println(analogRead(0)*5.0/1023.0);
}
}
}
The sketch starts by including the header file SPI.h for the SPI functionality, and declares
the chip-select pin CS. In the setup() function, we initialize serial and SPI communication,
as well as declaring the CS pin as output, and set its value to the SPI idle-state, which
is HIGH. In the loop() function we use the query-response protocol and respond to the
command DAC n to set the DAC. First we read the desired 12-bit word from the serial line
into the variable val, and then add four configuration bits B0011 to the most significant
end, where the first bit corresponds to the channel number, which is zero for the single-
channel DAC MSP4921, and the second bit declares that we use unbuffered voltages. With
the third bit we choose the internal amplification level to be unity, and the fourth bit enables
output. Once the 16 bits in the variable val are assembled, we can pull CS low to initiate
communication and then transfer the 16 bits in two chunks of 8 bits each to the chip with
two calls to SPI.transfer() before pulling CS high again to end the communication and
internally transfer the voltage to the output buffer of the DAC. The command A0? reads
back the voltage, already converted to volts. This simple sketch will allow us to set any
device that requires analog control voltages as input, and also read back the voltage to
verify correct operation.
Now we have a number of methods to control voltages, switches, and motors, but we
will wrap up this section by discussing a few means to attract the attention of a user, which
is useful in case of malfunction or another event that requires human intervention.

4.5.6 Human attention actuators

Early in this chapter we discussed an LED as a human attention actuator. An LED obviously
addresses the eye. But we also have ears, and can also attract their attention, for instance,
in case some unexpected error happens, with an acoustic signal. The simplest example is
a small loudspeaker that we can either use as a simple buzzer, or even code with simple
 Microcontroller: Arduino  103

Figure 4.23 Connecting a speaker and a piezo buzzer to the Arduino.

acoustic patterns to identify certain things. For instance, if a moving robot gets stuck in a
corner, we may want it to sound like a siren.
We interface a speaker to the UNO by connecting one terminal to ground on the Arduino
and the other to a digital output pin, say pin 2, as shown in Figure 4.23. In order to produce
a tone of 440 Hz for 1000 ms, we use the Arduino command
tone(2,440,1000);
placed anywhere in the Arduino sketch. Initialization of the pin as digital output is done
automatically. Replacing the 2 by a 3 will sound the piezo buzzer shown in Figure 4.23.
After being able to interface a number of sensors and actuators over the serial line, we
need to take a closer look at the options to communicate to the host computer and how to
receive the data there.

4.6 COMMUNICATION TO HOST

So far, we used only the serial monitor in the Arduino IDE to communicate with the
Arduino, but normally we also want to communicate from other programs, and that is the
topic of this section.

4.6.1 RS-232 and USB

The serial monitor in the Arduino IDE uses an RS-232 protocol that is transported over
a USB line. The translation happens automatically, and plugging the USB cable into the
host computer causes the latter to automatically create a device file (/dev/ttyUSBx or
/dev/ttyACMx on Linux, COMx on Windows, and /dev/tty.usbserial-xxx on a Mac) that
represents the other end of the communication channel from the UNO to the host computer.
We can then use any terminal program on the host computer to connect to that device file,
but have to keep in mind that we use the same baud rate that we specify in the setup()
function on the Arduino. To find out the device file to which the Arduino is connected, we
can use the Arduino IDE and inspect the Tools→Port menu, and write the name of the
device file down. In my case it was /dev/ttyACM0. After we close the Arduino IDE, we can
use any terminal-emulator program, such as putty on Windows, or the very basic program
screen on Linux or Mac, to connect to the UNO. In my case the command
screen /dev/ttyACM0 9600
104  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Front and back side of the HC-06 Bluetooth dongle and the connection
Figure 4.24
to the Arduino.

does the trick, and we can send commands to the Arduino and receive the response in the
terminal window. The screen program can be exited by pressing Ctrl-a-k and confirming
the question of whether one really wants to exit with y. The number 9600 is the baud rate
used for the communication and must match the number in the Serial.begin() statement
in the Arduino sketch. Note that 9600 is the default baud rate and could be omitted in the
screen command.

4.6.2 Bluetooth
Bluetooth functionality can be added by attaching an HC-06 Bluetooth dongle to ground,
power, and the Arduino pins 0 and 1. On the top left of Figure 4.24 we see those four pins
labeled RXD, TXD, GND, and VCC. The latter two are connected to the respective power
supply connections, and RXD on the HC-06 is connected to pin 1, labeled TX on the Arduino,
which is illustrated on the bottom of Figure 4.24. On this connection the information flows
from the Arduino to the HC-06. The pin labeled TXD on the HC-06 must be connected to
pin 0, labeled RX on the Arduino, and the information flows from the HC-06 to the Arduino
on this line. This crossed connection from TXD to RX and vice-versa is equivalent to a null-
modem cable. After connecting the HC-06 in this way, all communication is sent via the
HC-06 and the USB connection in parallel. For reliable operation, the USB link should not
be used while Bluetooth is in operation.
 Microcontroller: Arduino  105

On the host computer, we have to pair a new Bluetooth device with the host computer.
On Windows, this is done in the Bluetooth administration program. On some Linuxes,
similar user interfaces exist, but we can always pair Bluetooth devices using a number of
command-line programs. First we need to find out whether the host computer has Blue-
tooth capabilities, by the command hcitool dev, which should report at least one device,
normally called hci0. Then we scan the surroundings for Bluetooth devices with the com-
mand hcitool scan. If the HC-06 is powered we should see at least one device with a line
xx:xx:xx:xx:xx:xx HC-06 where the six-byte string is the MAC address of the Bluetooth
device. To establish pairing, we have to log in as root user with the command su or sudo
and call the program bluetoothctl -a, which reports the controller and known devices. At
the [bluetooth]# prompt we initiate a search for new devices with scan on, which reports
all known devices. The following two commands establish the pairing
trust xx:xx:xx:xx:xx:xx
pair xx:xx:xx:xx:xx:xx
where xx:xx:xx:xx:xx:xx is the MAC address of the HC-06 dongle we want to pair. During
the previous actions we are prompted for a PIN number, and unless we have changed the
default on the HC-06, we use 1234. Then we exit the bluetoothctl program and create
the device file for the serial communication by issuing
rfcomm bind 0 xx:xx:xx:xx:xx:xx
which creates a device file /dev/rfcomm0 that has the same functionality as the
/dev/ttyACM0 device file we used earlier to communicate with the Arduino. Therefore
we can again use the screen command to communicate with the Arduino, but using
/dev/rfcomm0 as first argument to the screen command instead. Note that the HC-06
is configured to communicate with 9600 baud. This can be changed using AT commands,
but we will not discuss this further, and assume that all Bluetooth communication using
the HC-06 is done at 9600 baud. Once we are done using the Bluetooth serial link, we
should take it down by issuing the command rfcomm release 0, which will remove the
/dev/rfcomm0 device file, and we will no longer be able to use it. The bottom line is that
we can communicate using Bluetooth in much the same way as using native RS-232 or USB.
We only need to pair the device and the host computer once, and then create the device file
using the rfcomm bind command before using the serial line, and delete it with the rfcomm
release command once we are finished.
Apart from the channels based on serial communication, the more modern Arduino
clones such as the ESP8266 can use normal WLAN-based communication based on network
sockets. We will discuss this nifty feature in the next section.

4.6.3 WiFi
For this section, we use a NodeMCU system that is connected to the host computer, with
a USB cable for programming, and the device support package as described in Section 4.2
is installed in the Arduino IDE. Our task is to connect the NodeMCU to a local wireless
network; we assume it is called MyHomeNet and that it can be reached by other computers
connected to the same network. All these computers will be able to query the measurement
values of sensors connected to the NodeMCU. To better understand the setup, we briefly
discuss some general features of computer networks.
Everybody is probably familiar with the fact that computers on the Internet are identi-
fied by IP numbers, such as 192.168.10.200. Because the numbers are difficult to remem-
ber, there are also aliases such as www.cnn.com, and the translation is done by so-called
106  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.25 The NodeMCU with an LM35 temperature sensor.

domain name resolution or DNS servers. On our simple network, we assume that we can
keep track of the numbers, and all sensors and host computers are connected to the (class-C)
network 192.168.10.nn where nn is a number between 2 and 254. The numbers 0 and 255
are reserved for special purposes, and we assume that the router that connects our network
to the outside world (“the Internet”) has the IP number 192.168.10.1. Note that the
addresses starting with 192.168 are private numbers and can be used by anyone, provided
that there is a router that separates this network from the Internet. But we confine all our
communication to within the 192.168.10 network, and identify each computer by its IP
number. Each computer can potentially provide different services, such as running a web
server, a measurement server, a mail server, or a server to allow logging onto the computer
with standard protocols such as telnet, ftp, or ssh. The different services that a com-
puter provides are identified by a port number. As an analogy, it may help to think of the IP
number as a street address of an apartment building and the port numbers as the apartment
numbers. The communication with a server that provides a service on a computer therefore
requires the specification of the IP number and the port number.
But how does a computer know its own IP number? There are two ways to specify this:
by configuring the network setup manually and assigning the IP number explicitly. We use
the second way and dynamically acquire the IP number via the dynamic host configuration
protocol, DHCP, which is much more convenient, provided such functionality is available on
a given network. In most networks with a wireless router, the router provides this service
and one only has to tell a computer to use DHCP. In that case, the computer sends a request
for an IP number at power-up. The DHCP server responds with an IP number that is then
assigned to the newly connected computer. We assume that a DHCP server is running on
our 192.168.10 network and the NodeMCU is also configured by default to use DHCP.
Wireless networks are normally protected from unauthorized use by encrypting the com-
munication. Most networks use an encryption standard called WPA that requires entering
a password to connect to the network. We assume that the “MyHomeNet” WLAN is of that
type. Having covered the networking basics, we are ready to connect our Arduino clone, the
NodeMCU, to the WLAN.
 Microcontroller: Arduino  107

First we discuss the running of a simple web server on the NodeMCU, that uses the
standard http protocol to communicate with other computers on the network. The fol-
lowing program shows how to set up a web server that listens on the default http port
number 80. It provides access to temperature measurements and allows one to control
the brightness of an LED. In our example we query the temperature measured by the
NodeMCU, by directing a browser such as Mozilla Firefox to the address of the NodeMCU
at http://192.168.10.nn/temperature, and the NodeMCU returns a web page with
the temperature measured by an LM35 temperature sensor. The brightness of the built-
in LED is controlled by adding a value to the address after a question mark, such that
http://192.168.10.nn/led?b=1023 sets the brightness to its maximum value. The circuit
schematic is shown in Figure 4.25, and the code that brings it to life is the following.
// Web server to read temperature and set brightness, V. Ziemann, 170911
#include <ESP8266WiFi.h>
const char* ssid = "MyHomeNet";
const char* password = ".........";
WiFiServer server(80); // server listens on http port 80
void setup() {
Serial.begin(115200); delay(10);
Serial.print("Connecting to "); Serial.println(ssid);
WiFi.begin(ssid, password);
while (WiFi.status() != WL_CONNECTED) {
delay(500); Serial.print(".");
}
Serial.print("\nWiFi connected and ");
server.begin();
Serial.print("server started at ");
Serial.println(WiFi.localIP());
}
void loop() {
WiFiClient client = server.available();
while (client) {
while(!client.available()){delay(1);}
String req = client.readStringUntil(’\r’);
client.flush();
client.print("HTTP/1.1 200 OK\r\nContent-Type: text/html");
client.print("\r\n\r\n<!DOCTYPE HTML>\r\n<html>\r\n");
if (req.indexOf("/temperature") != -1) { // read sensor
float temp=100*3.3*analogRead(0)/1023;
client.print("Temperature="); client.println(temp,2);
} else if (req.indexOf("/led") != -1) { // brightness of LED
int i1=req.indexOf("?"); int i2=req.indexOf("HTTP");
String payload=req.substring(i1+1,i2-1);
if (i1>0) analogWrite(D4,1023-payload.toInt());
} else {
Serial.println("invalid request");
return;
}
client.println("</html>"); delay(1);
client.stop();
108  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 4.26 The communication with a web server showing the HTTP header.

}
}
In this example we first include the header file with the ESP8266WiFi.h library information,
define the access parameters ssid and password for the wireless network, and instantiate
the server to listen on the default http-port 80. In the setup() function, the serial line
is opened to be able to listen to debug information. Then the connection to the wireless
network is established with the WiFi.begin() function. Inside this function, the log-on
to the wireless network and the communication with the DHCP server is handled. Once
WiFi.status reports that the connection is established, we print information about the
connection, such as the acquired IP number, to the serial line, and start the server with the
call to the server.begin() function. In the loop() function a similar construction as before
receives a request from a client, stores the request in the variable req, returns a standard http
header with Content-type to the client and then checks whether the request req contains
the string /temperature. If it does, the sketch determines the temperature with a call to
the analogRead() function, and writes the value back to the client. Note that this time we
use the indexOf method of a String to determine whether a substring is present. If req
contains /led, we determine the position of the question mark and trailing characters HTTP
and extract everything in between as the payload. After conversion to an integer, we use the
payload to set the brightness of the built-in LED. We can inspect how req is constructed
and why we have to find the trailing HTTP by printing with Serial.println(req). If an
unknown request arrives, we are notified on the serial line. Finally, we add the concluding
</html> tag and close the connection to the client. We point out that in this simple example
we do not check the validity of the payload, which poses a potential security risk, if the
system is reachable from the Internet.
A few words are needed about the cryptic HTTP header that the NodeMCU returns to
the calling browser. When a browser, such as Firefox, connects to a web server on the default
port 80, it first sends the name of the requested web page, such as GET /temperature in the
above example, as an HTTP-GET request. Before returning the requested web page, the
server sends some meta-information, such as a status code, and what type of information
comes next. The status code for a properly understood request is 200, and for a missing page
it is 404, a number probably everyone has seen as a response to a typo in the specification
of a web page. The type may be an image, a media file such as a video, or HTML-formatted
text, which is what we specify in the sketch as Content-Type. After an empty line the
normal HTML header with the DOCTYPE follows, and the information embedded in <html>
 Microcontroller: Arduino  109

tags. We will discuss HTML and the structure of web pages in more detail in Section 5.7.
Browsers do not render the meta-information, which therefore remains invisible, but we
can eavesdrop on the communication with the netcat or telnet commands (more on those
commands later in Section 5.3) by pointing them to port 80 on the NodeMCU at IP address
192.68.20.184 and issuing GET /temperature by hand. All output from the server sub-
sequentially appears in the same window. Figure 4.26 illustrates the exchange.
Setting the brightness of the LED uses the same mechanism that is used to enter credit-
card numbers when shopping online. It is based on the HTTP-GET method as well, but
adds a query-string that follows the question mark to the address. The query-string has the
form name=value. This mechanism to send information to a web server is often used with
HTML forms. It allows the user to enter text or select things from a list and then click a
submit button to send it to the server, where it causes some reaction, such as sending the
desired purchase—or to set the brightness of an LED.
In a second example, we configure a NodeMCU to run a server listening for connections
on port number 1137, and once a connection is established, it waits for commands and then
replies appropriately. The hardware is the same as in the previous example and is shown in
Figure 4.25. The code that runs on the NodeMCU is the following.
// Socket-based measurement server, V. Ziemann, 161211
const char* ssid = "MyHomeNet";
const char* password = "........";
const int port = 1137;
#include <ESP8266WiFi.h>
WiFiServer server(port);
void setup() { //...........................................setup
Serial.begin(115200);
WiFi.begin(ssid, password);
while (WiFi.status() != WL_CONNECTED) {
delay(500); Serial.print(".");
}
Serial.println("");
Serial.print("WiFi connected to "); Serial.println(ssid);
Serial.print("Server IP address: "); Serial.println(WiFi.localIP());
server.begin();
Serial.print("Server started on port "); Serial.println(port);
}
void loop() { //............................................loop
char line[30];
float volt,temp;
WiFiClient client = server.available();
while (client) {
while(!client.available()) {delay(1);}
client.readStringUntil(’\n’).toCharArray(line,30);
Serial.print("Request: "); Serial.println(line);
if (strstr(line,"A0?")) {
volt=3.3*analogRead(0)/1023;
client.print("A0 "); client.println(volt);
} else if (strstr(line,"T?")) {
temp=100*3.3*analogRead(0)/1023;
client.print("T "); client.println(temp,1);
110  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

} else {
Serial.println("unknown command, disconnecting");
client.stop();
}
client.flush();
// client.stop();
}
}
In this sketch we first define two character strings, ssid and password, for the name and
passphrase of the wireless network. We also specify the port number, here 1137, in the
following line before including the functionality for WiFi support on the ESP8266 device.
Once that is included, we define a WiFiServer named server that listens on the specified
port. In the setup() function, we start up the serial communication via the USB line
that we use for logging, and then connect to the wireless network with the WiFi.begin()
function. Once the NodeMCU is connected to the wireless network, we start the server with
the server.begin() command. In the loop() function we use the same structure that
we used earlier for communication via the serial line, but this time over the socket that is
dynamically opened once a client computer connects to our server. The connection to the
requesting computer stays open until an unknown command closes the connection. But while
the connection is open, values can be requested repeatedly. If we want the connection to
close immediately after the reply is sent, we can uncomment the line with client.stop();
immediately following the line with client.flush(). Note that this server can only handle
one request at a time. If one computer is already served and a second computer tries to
establish contact, the latter is put on hold until the first computer disconnects. This program
on the NodeMCU allows us to query via the custom socket on port 1137 any parameters or
measurement values that the NodeMCU has available. How to connect to it from another
computer we defer to the next chapter.
Beyond the standard communication protocols there are several others available. We
discuss those only briefly in the following section.

4.6.4 Other communication

In the previous sections, we used standard and easy-to-use protocols (RS-232, USB, Blue-
tooth, WiFi) to communicate with the microcontroller. They have the advantage that they
are readily available on many host computers. There are, however, other communication
channels, of which we name a few in the following paragraphs. All are supported by Arduino
libraries.
One other wireless communication channel is ZigBEE, which is related to Bluetooth but
uses a simpler protocol overhead and is optimized to consume as little power as possible, to
make long-lived battery operation possible. Another option is using inexpensive 433 MHz
transceivers, which can be used to connect slave microcontrollers to their master using a
simple wireless link.
MIDI is a protocol developed to interface electronic musical instruments. The physical
interface is based on current loops that are galvanically decoupled through optocouplers at
the input of each device. Logically the communication is very similar to RS-232 at a baud
rate of 31250. Each MIDI message consists of three (or in some cases, two) bytes that have
standardized interpretation, such as channel number, tone (corresponding to a specific key
on a piano keyboard), and velocity (the intensity with which the key is struck).
It is possible to interface nearby hardware using infrared (IR) light, which is used in TV
 Microcontroller: Arduino  111

remote controls to send information from the remote to the TV to change channel or volume.
Incidentally, even the first generation LegoTM Mindstorm microcontrollers communicated
in this way. The communication is based on an IR diode using a wavelength of 940 nm
and modulating the light at a rate of 38 kHz. Sending bursts of modulated light represents
either a LOW or a HIGH signal level and is used to emulate an RS-232-like protocol at
low baud rates such as 2400 baud. Another protocol is RC-5, which is commonly used in
TV remote controls. Receivers such as the TSOP38438 have optical filters built in and are
sensitive only to a narrow band of wavelengths around 940 nm. Moreover, they demodulate
the 38 kHz carrier frequency and only deliver a 3.3 or 5 V signal on their output pin, making
them very easy to interface.
Note the generic structure: A communication channel is based on the low-level hard-
ware and on a protocol stack that often consists of several layers on top of the hardware
implementation. In our examples we use the convention to send a string, terminated with
a question mark to signify a request. The reply then consists of the same string followed
by a value. This convention defines a simple protocol for the communication. We can easily
come up with other protocols, and add features such as checksums to test the integrity of
the transmission. Also, I2C and MIDI communication are based on a standardized protocol
where a number of bytes is transmitted and each byte signifies some particular information,
such as the register to be addressed or the value that is written to the register. Of course,
it is mandatory that all participants in a transaction agree on the interpretation of the bits
and bytes that are transmitted, otherwise confusion will reign supreme.
So far we always assumed that some unspecified host computer, for example, a desktop
computer, is available and serves as the communication partner for the microcontroller.
In the next chapter we consider one specific host computer that is widely available, well
documented, and inexpensive—the Raspberry Pi.

QUESTIONS AND PROJECT IDEAS

1. Convert the hex number 0x5CA3 to binary representation.
2. The ADC that reads pin A0 on the NodeMCU is not very reliable. For example,
connecting A0 to GND will not necessarily give a zero ADC reading. Make an attempt
to calibrate it.

3. Can you connect an MCP23017 IO-extender and a BMP180 pressure sensor to the
same I2C bus?
4. Under what circumstances is it advantageous to use interrupts?
5. You need up to 128 digital input pins. Use multiple copies of the MCP23017 and
explain how to connect them.
6. Inspect the datasheet of the MCP3304 and find out how to configure it to read eight
unipolar input voltages. What do you need to change in the software?
7. If you need 128 unipolar analog input signals, how do you connect the MCP3304 to
an Arduino? Is it possible at all, and if so, how?
8. Connect a three-color LED to an Arduino and independently control the brightness
of the three colors via the serial line. Add features to directly set standard mix-colors
such as orange or magenta.
112  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

9. Explain why we use (1023-value) to set the brightness in the example with the web
server on page 107.
10. What is the difference between HTTP and HTML?

11. Learn about HTML forms and implement a web page with a user interface to set the
brightness of the LED on the NodeMCU.
12. Mount an HC-SR04 distance sensor on top of a model-servo and scan the neighborhood
for the distance to obstacles. Produce a beep, if it is closer than 20 cm.

13. Use the tone() function to make the Arduino sound like a siren.
14. Build a simple pulse generator that generates signals with millisecond precision. Dis-
cuss how to improve the temporal precision.
15. Build a magnetic field sensor with a Hall probe.

16. Build a compass using the magnetic field sensor on the MPU9250.
17. Build a device that senses the tilt angle with an MPU6050 or MPU9250.
18. Build a carousel comparator and use an MPU6050 to measure the acceleration you
experience in a merry-go-round. Use a battery-powered NodeMCU to publish a web
page with the data. Enhance the system by measuring and displaying the angular
velocity as well. Also display the maximum and minimum values during the past 5
minutes.
19. Build a contact-free wireless thermometer with an MLX90614 FIR sensors and a
NodeMCU that makes the measured values available on a web page or via a socket-
based server.
20. Measure the wind speed with a propeller that breaks the light path in a slotted optical
switch or between a discrete LED and a phototransistor.
21. Investigate materials for their phosphorescence by briefly flashing an ultraviolet LED
and record the response of the material with phototransistors or diodes sensitive to
different wavelengths.
22. Investigate how to interface the dust sensors from Section 2.3.6 to the Arduino.
23. Connect an MQ-x gas sensor to the UNO and determine the air quality.
24. Investigate how to interface the GPS sensor from Section 2.3.6 to the Arduino. Con-
sider using the SoftSerial library to add additional RS-232 ports to the microcon-
trollers.
25. Investigate the Arduino NANO. In what way does it differ from the Arduino UNO?
Discuss circumstances where you would use a NANO.

26. The ESP-01 does not have a USB interface. Investigate how to program it.
 CHAPTER 5

Host Computer: Raspberry

Pi

The Raspberry Pi [9] is a small single-board computer that first appeared in 2012, with the
intention of providing an inexpensive platform to introduce students and other interested
parties to computers in general and to programming in particular.

5.1 HARDWARE
Since its first appearance, the Raspi went through several hardware revisions, and the
present incarnation, model 3, appeared in February 2016. This version, shown in Figure 5.1,
features a quad-core ARM central-processing unit (CPU) operating at 1.2 GHz and has
1 Gbyte RAM memory on board. This hardware base is sufficiently powerful to run a full-
fledged Linux system. An external and replaceable micro-SDHC card (preferably speed class
10) serves as a hard disk to hold the operating system and user files.
But beyond the CPU, the Raspberry Pi sports a video processor that can display videos
at full-HD resolution (1920 × 1080) via the built-in HDMI-connector. Audio output is avail-
able either via the HDMI connector or via a 3.5-mm headphone connector. Moreover, there
are four USB-2 ports on board to connect peripheral components, such as USB sticks, key-
board, mice, or web cameras. Communication with the outside world is feasible via a built-in
wired Ethernet port, and since version 3, the Raspi has had built-in Bluetooth (V4.1) and
WiFi (802.11n).
The Raspis are very attractive due to their built-in low-level peripherals. There are 17
general-purpose input–output (GPIO) pins exposed on the board, some of which support
I2C, SPI, and UART (RS-232-like) communication. Moreover, a specific audio bus (I2S) is
available, as well as a high-speed CSI interface to connect the tailor-made Raspberry Pi
camera, and a DSI interface to connect LCD panels. Some of these features can be used to
implement the same functionality as on the Arduino, but we will not use that here, using
the Raspi as a standardized host computer system instead.

5.2 GETTING STARTED

The CPU and peripherals are adequate for using a Raspi as media center via the OpenElec
distribution, which runs the KODI media center software and turns a dumb TV into a
smart TV. There is the OpenWRT distribution to turn the Raspi into an Internet router
including firewall functionality. Other software packages turn it into a retro-gaming console

113
114  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.1 A Raspberry Pi board.

or into a network-addressed storage (NAS) server. Apart from these special-purpose uses,
we can also install a general-purpose Linux operating system on the SDHC card from which
we boot the Raspi. Different flavors of Linux are available, such as Ubuntu, Fedora, or
OpenSuse, but the standard system that we use in the remainder of this book is based on
the Debian distribution and is called Raspbian. In recent years, a large number of books, such
as [20, 21, 22], were published with recipes of how to use the Raspi in many circumstances.
Use those and others to complement the discussion in this book.
We download the Raspbian operating system from the www.raspberrypi.org web site
under the Download link at the top of the page. The download is a rather large, presently
1.4 GB, image file for the full system. Near the top of the download page there are installation
instructions for all common desktop operating systems. On a Linux desktop system we have
to unzip the image file, which may take some time, and then transfer the image file to the
SD card by issuing the command
dd bs=4M if=2016-11-25-raspbian-jessie.img of=/dev/sdX
as user root. We may have to prepend the dd (disk-duplicator) command by sudo depending
on the flavor of Linux on the desktop computer. In the dd command, the destination, the
output file /dev/sdX, must be the device file of the SD card. We can easily determine it
by inspecting the system log file after inserting the SD card in the writer. Just before the
end of the log file a recently inserted card is typically referenced as /dev/sdX, with X being
c or d or e. We must make sure that you use the correct drive. If we pick the wrong one
we may damage our desktop system. If unsure, we carefully follow the instructions in the
installation help, which are more extensive than those given here. Once the dd command
completes, which may take up to 15 minutes, depending on the SD card writer, we insert
the SD card into the slot on the Raspi, connect a monitor to the HDMI connector, mouse
and keyboard to USB connectors, and apply power via the micro-USB connector to boot
the Raspi.
The normal system directly boots into the desktop system on the Raspi, and automat-
ically starts the configuration program raspi-config. There we need to do some book-
keeping activities, such as expanding the file system to use the entire disk. The image file
we download only contains a moderate file system, but we can coax it to use the entire
 Host Computer: Raspberry Pi  115

Figure 5.2The Accessory menu behind the button with the raspberry shows a num-
ber of installed programs such as the Terminal (command) window.

space on the SD card, depending on the size of the card, which should be 8, 16, or even
32 GB. Next, we choose the localization option to select the time zone and especially the
keyboard layout. Then we change the user password for the default user pi from the default
password raspberrypi to something more unique, and progress to the advanced options to
select a suitable hostname for the Raspi. Make sure to remember the new password! Then
we close the raspi-config program and reboot.
Once the Raspi comes up again we are directly logged into the desktop system, without
a password being asked. To change this behavior we start a terminal program and type
sudo raspi-config followed by Enter on the command line. Under point 3 (Boot options)
we then select an option without auto-login, and close the program before rebooting.
At this point we have a running Raspbian Linux system on our Raspi, and it is time to
explore it by selecting the Menu button with the raspberry image on it. There we find all
available programs, sorted according to groups such as Programming, with links to Java,
Python, and Mathematica. The group Office contains links to the Libreoffice programs for
word processing, spreadsheet, and presentation creation. Behind the Internet menu item,
there are links to the Internet browser and mail client. Moreover, there are menu items for
Games and Accessories. The latter, shown in Figure 5.2 contains a link to the Terminal
program that provides a console to enter commands. We will use it extensively in later
sections, but here are a few basic commands that are useful to know. When we open the
terminal window, we are placed in the home directory of user pi, namely /home/pi, which
we can verify by entering the command pwd to “print the working directory.” We can list
its contents by entering the ls command, or its long form ls -l, which displays all files
116  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

and subdirectories in the present directory. In order to enter a subdirectory we use the cd
<dirnam> command to change the working directory to <dirnam>. The command cd ..
takes us back to the directory where we started. If we want to navigate to a directory for
which we only know the absolute name that starts with a slash, such as /usr/local, we use
cd /usr/local. We should play around with the programs and test them to familiarize
ourselves with the new Raspi system. Either type the commands on the command line or
click on the programs in the menus and explore. Finally, there is the Preferences menu item
with programs to customize the Raspi. In particular, the link Raspberry Pi Configuration
starts a graphical user interface with the same functionality as the raspi-config program
we used earlier.
In the menu list, next to the Menu button with the raspberry on it, there are quick-start
buttons for several programs, such as an Internet browser, file manager, and Mathemat-
ica. Programs can be added to this area by right-clicking on the menu list and selecting
Add/Remove Panel Items. In the window that appears, we highlight Application Launch
Bar and click on Preferences. The appearing window has a left side with the already
present programs in the quick-launch area. On the right is a list of installed applications
from which to select. Here we install the Terminal program from the Accessory group in
the Quick launch area because we will use it a lot later on. On the rightmost end of the
menu list, widgets for network, audio, and other features are located. We add or remove
such widgets from the Add/Remove Panel Items window that we launch by right-clicking
the menu list and pressing the Add button, whence a list of available widgets appears. From
it we select whatever we want to add. Finally, we move the entire menu-list to our preferred
location on the desktop by right-clicking on it and selecting Panel Settings where we choose
the location of the menu list. The default location is on the top, but we can also move it to
the bottom of the desktop.
In order to log on to the Raspi from our desktop computer, we need to know its IP
address, which is found by executing ifconfig in a terminal window on the Raspi and
finding the point labeled inet addr for the interface eth0. Once we know it, we log on to
the Raspi via the secure shell program ssh from our desktop computer by typing ssh -X
[email protected], where 192.168.10.nn is the IP number of the Raspi. The -X option
allows us to display additional windows on the desktop computer. If this is the first time
we connect, a warning message appears that the Raspi is unknown to your desktop and
we should answer yes that all is well. Then we are presented with a request to enter the
password for user pi on the Raspi. If we enter that correctly, we get a command prompt
from the Raspi at which you can enter commands, such as sudo raspi-config or any
other. From a Windows computer, you may use any ssh client program, such as putty, to
log on to the Raspi command line. This way of logging on can be simplified significantly by
closing the connection, and back on the desktop computer we enter
ssh-copy-id [email protected]
which presents a prompt to enter the password for user pi on the Raspi for the last time, and
henceforth we log on to the Raspi from our desktop by typing ssh -X [email protected]
and are at the command prompt of the Raspi without being asked for a password. Note that
this only works from our account on the desktop computer. The ssh-copy-id exchanges a
secret with the Raspi that is used to authenticate our account on this desktop computer and
no other. Thus, logging onto the Raspi is just as safe as logging onto our desktop computer.
We could now remove screen, mouse, and keyboard and always use the ssh command to log
onto the “naked” Raspi, but for convenience we keep the peripheral devices attached a little
longer until we have installed software to allow running a virtual screen via the network.
The default system already comes loaded with a good selection of programs, but we can
 Host Computer: Raspberry Pi  117

add more programs, and the process to do so is extremely simple. We discuss this in the
next section.

5.3 INSTALLING AND USING NEW SOFTWARE

Before installing new software we first update the current system, which is based on an
image file prepared some time ago. To do this we start a terminal program and enter the
following two commands:
sudo apt-get update
sudo apt-get upgrade
where the first command updates the database of installed programs and finds out whether
newer versions are available. The second command installs the upgrades after a question to
approve its selection. First, the newer versions are downloaded and then installed. This may
take quite a while—about 30 minutes if the time since the last update or the creation of
the initial installation media is long ago. Note that the system-administration command is
called apt-get, and we have to run it with superuser privileges by prepending sudo. After
the upgrade is complete and the program returns to the command line prompt, we have an
up-to-date system and are ready to install additional software.
Since we want to test the interface with the Arduino, we need to install the simple
terminal emulator program screen by issuing the command
sudo apt-get install screen
and after confirming that some kilo-bytes are downloaded, the program is installed in a few
seconds. We assume that we have an Arduino UNO programmed with the query-response
sketch from page 64 and connect the USB cable from the UNO to a USB port on the Raspi.
We then check the device name of the serial line by issuing dmesg in a terminal window,
and inspect the output near the end. In my case, one line contains the string ttyACM0,
but ttyUSBn with any number n can also occur, depending on the serial-to-USB converter
installed on the UNO. Since the Arduino sketch uses 9600 baud for the serial line, we connect
to it with
screen /dev/ttyACM0 9600
on the command line of the Raspi. Then we query the UNO by typing A0? followed by
Enter in the screen terminal-window. The UNO should respond with A0 nnn where nnn
is the value measured by the ADC on the UNO. Congratulations, we have established a
communication link from the Raspi to the Arduino UNO. Once we are done with getting
data from the UNO, we close the connection by pressing Ctrl-a-k in the screen window,
confirm the question of whether we really want to quit by y, and return back to the command
prompt.
The screen program and in fact most programs on a Unix or Linux system bring their
own help system, called manual pages, which can be accessed from the command line by
typing man <program name>—for example, man screen in the present case. This displays
information on how to use the program in the terminal window. In particular, all command
line switches are explained. We exit from the man program by typing “q” in the window.
Using man only works if the program name is known, but we can use the command man -k
<keyword> to search the system for manual pages that have to do with <keyword>. The
latter command works if we initialize a database once with sudo mandb. Check with man
mandb to see what it really does!
118  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.3 The nano text-editor with available commands listed on the bottom.

Later on, we will often have to edit configuration files or to write program code. The
edited files are usually plain text files, so-called ASCII files, and we create and edit them
with an editor program such as nano, vi, or emacs. The first two are already installed
on the system, and since vi has a steep learning curve, nano is a good choice to start,
unless we have another preference. We start nano from the command line by typing nano
<filename>, and after pressing Enter, the terminal window opens with the contents of
the file, if it exists, or an empty file, if it does not exist. Figure 5.3 shows nano after we
added some text. On the top are the usual menus for File and Edit operations, and at the
bottom of the window the most important commands are listed, especially Ctrl-x to exit
the program and Ctrl-g to open the built-in help system with more information about
using nano. Here, both small and capital letters work in conjunction with the Ctrl key to
execute commands, such as Ctrl-k and Ctrl-u to cut and paste the line with the cursor.
Note that once a text file is written to disk, we can also view it using the less command
or with cat. Please check the respective manual pages for additional information.
Any program, library, or other package that is available in the official Raspberry Pi
repositories can be installed in the same way by first updating the database with sudo
apt-get update, which is only necessary once per session and then installing the pack-
age with sudo apt-get install <package-name>. We use this newly won information to
install the MATLAB-clone octave [23] by issuing
sudo apt-get install octave

on the command line. After a lengthy installation that takes a few minutes, depending on
the download speed, the program is ready to use. We start it from the command line by
the command octave, followed by Enter. Octave greets the user with its version number
and some copyright information before the prompt octave:1> appears. There we can enter
MATLAB-compatible commands like A=[1,2;3,4]; B=inv(A), which defines a 2×2 matrix
A, inverts it, and stores the result in the variable B. Later we will also connect to the
Arduinos from within octave.
In order to be able to display and convert graphic files from one format to another, the
imagemagick package is invaluable. We install it by executing
sudo apt-get install imagemagick
from the command prompt. Once it is installed, we can display almost any graphics file with
display graph.png, where graph.png is just an example of any graphics file. Converting an
 Host Computer: Raspberry Pi  119

Figure 5.4 The Synaptic Package Manager.

existing file graph.png to JPEG format is done by typing convert graph.png graph.jpg
on the command line. Use man display and man convert to find out more about what the
programs can do.
Installing with apt-get is already rather convenient because dependencies to external
libraries are automatically resolved, and any missing libraries are installed in order to guar-
antee a reliably working system. There is, however, an even more convenient way, namely
by using the package manager synaptic. We install it by issuing
sudo apt-get install synaptic
on the command line. Once the installation is complete, it appears as Synaptic Package
Manager in the Preferences section on the menu, behind the raspberry logo. Starting it
from there first requests the login password before a window appears, shown in Figure 5.4.
We use the Search button on the top to find programs. Just click on it, enter a keyword
and peruse the list, install any interesting program, and try it out.
We use that feature immediately, and install programs to access computers, such as the
NodeMCU that provides server capabilities via sockets. For this we need one or both of the
netcat and telnet (client) programs, and install them via synaptic by searching for the
program name; we pick the required package from the list by right-clicking it and choosing
Mark for Installation. Once all programs are selected, we can press the Apply button near
the top left of the synaptic user interface. Upon completion of the installation, we are ready
to use the programs to connect to the NodeMCU running the socket-based server, which
we discuss in detail later. Here we briefly illustrate the capabilities of the netcat program
by opening an ad-hoc server in one terminal window. Running netcat -l 11111 starts a
server that listens (-l) on port 11111. Note that normal users can only use port numbers
120  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

above 1023. We connect to that server by running netcat localhost 11111 from a second
terminal window, where localhost is the default name for this computer and the port
number on which the server listens. Now anything written in one terminal automatically
appears on the other. In this example, we only connected two running versions of netcat
on the same computer. One is started as a server, the other as a client, but the same
functionality works between any computer with a working network connection. Find out
more about netcat by consulting its manual page with man netcat. Should we wish to
remove a previously installed program, we select it in synaptic by right-clicking the program
in the list and selecting Mark for Removal. Pressing Apply promptly removes it.
So far, we have used the Raspi as a standalone computer with screen, keyboard, and
mouse directly attached to the Raspi, but that ties up the peripheral hardware. So, now is
the time to install software to allow logging onto the Raspi via the network, and receiving
a virtual screen of the Raspi desktop on our regular desktop computer. Then we have a
Raspi-desktop in a window that behaves just like the one on a regular screen. To achieve
this feature, we install the tightvncserver on the Raspi with the command

sudo apt-get install tightvncserver

or, alternatively, via synaptic. Once the installation has finished, we run sudo vncserver
from the command line, answer the questions about passwords, and remember the password.
On the desktop computer from which we want to log onto the Raspi via the network,
we install the vncviewer software using the package manager of our Linux distribution.
During the installation, the vncpasswd program should be installed as well, and we need
to run it from the command prompt on our desktop computer. It prompts us to enter the
same password for the VNC connection that we selected on the Raspi when running the
vncserver program for the first time. Once that is done, the basic infrastructure to log
onto the Raspi using VNC is in place. To simplify the entire process, we create a file named
raspi.sh on the desktop computer with the following contents
#!/bin/bash
RASPI=192.168.10.nn
ssh -X pi@${RASPI} vncserver -geometry 1280x960 &
sleep 10
vncviewer ${RASPI}:5901 -passwd /home/ziemann/.vnc/passwd &
and place that file in the ~/bin directory. This little program remotely logs onto the Raspi
with ssh, starts the vncserver program on the Raspi with the specified desktop geometry
(we may use a smaller one such as 1024x768), then waits a few seconds and starts the
vncviewer on the desktop computer using the credentials in the .vnc/passwd file in the
home directory. After a few further seconds the desktop of the Raspi appears in a window
on the desktop computer, and we seamlessly move the mouse and keyboard focus to the
Raspi desktop, and execute commands on the Raspi. This is a very convenient way to use
the Raspi, because it only requires a network connection and no other peripheral devices.
At this point, we have a convenient setup to work on the Raspi, and are basically ready
to communicate with all the microcontrollers with attached sensors. But using the hardware
for both wired and wireless Ethernet on the Raspi is just too tempting to neglect, and in
the next section we briefly describe how to use the wireless interface on the Raspi to span
a private WLAN network to which we later connect the sensor nodes.
 Host Computer: Raspberry Pi  121

5.4 RASPI AS A ROUTER

In this setup we assume that the Raspi is connected to our normal network with its wired
Ethernet RJ45 port, and we have access to it from our desktop computer to log on using
ssh or vncviewer, as previously discussed. On the Raspi we need to install the following
packages
sudo apt-get update
sudo apt-get install hostapd dnsmasq
where we first update the repositories, and then download and install the hostapd and
dnsmasq package. The former program is responsible for spanning the private wireless net-
work, and turns the Raspi into a WLAN access point. The latter provides IP numbers on
the private network via the DHCP protocol, and translates web site names to IP numbers.
Next we need to re-configure our system. In the default configuration there is a back-
ground process, a daemon called dhcpd, that tries to obtain IP numbers for every network
interface on the Raspi, including the wireless interface called wlan0. On the other hand, in
order to operate the Raspi as an access point, we need to configure the interface ourselves
and therefore need to remove the interface wlan0 from the control of the daemon by adding
the line
denyinterfaces wlan0
at the end of the file /etc/dhcpcd.conf with any text editor, such as nano or emacs. Now
we are ready to configure the interface wlan0 manually by rewriting the relevant section in
the file /etc/network/interfaces to look like this:
auto wlan0
iface wlan0 inet static
address 192.168.20.1
netmask 255.255.255.0
up /sbin/iptables -A POSTROUTING -t nat -o eth0 -j MASQUERADE
down /sbin/iptables -D POSTROUTING -t nat -o eth0 -j MASQUERADE
which defines the IP number of the wlan0 interface to be 192.168.20.1, and config-
ures the type of network (class C) in the following line with the netmask command. The
/sbin/iptables command enables network address translation between the wireless net-
work on wlan0 and the wired network on interface eth0. In order to allow network packages
to pass between the interfaces, which is disabled by default as a security measure, we need
to enable forwarding by removing the hash # from the line
net.ipv4.ip_forward=1

in the file /etc/sysctl.conf. We omit this step if all sensor nodes should only communicate
with the Raspi, but no other computer on the local network should be able to surf the outside
Internet beyond the Raspi, which might not be needed and poses a potential security risk.
This last step completes the basic network setup, and we turn to the configuration of the
access point software.
The configuration file for the hostapd daemon is /etc/hostapd/hostapd.conf and
contains the following lines:
# /etc/hostapd/hostapd.conf
interface=wlan0
122  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

driver=nl80211
ssid=messnetz
channel=5
macaddr_acl=0
auth_algs=1
ignore_broadcast_ssid=0
wpa=2
wpa_passphrase=zxcvZXCV
wpa_key_mgmt=WPA-PSK
wpa_pairwise=TKIP
rsn_pairwise=CCMP
hw_mode=g
The file must be made readable for the owner only by executing sudo chmod 600
/etc/hostapd/hostapd.conf. It contains the definition of the network name messnetz
and the details of the encryption, such as the passphrase. The latter we should adapt to
your own secret phrase, to prevent others from misusing your wireless network. We need to
tell the system where to find the configuration file by entering the line
DAEMON_CONF=/etc/hostapd/hostapd.conf
near the top of the file /etc/init.d/hostapd and then instruct the system to start the
hostapd daemon at boot time with the command
sudo service hostapd start
After rebooting the system, we should be able to see a new WLAN named messnetz from
a computer with a wireless network interface, but before we can connect to it we need to
start the dnsmasq daemon that provides DHCP services on the 192.168.20.xx network and
distributes IP numbers. The configuration file is called /etc/dnsmasq.conf and it should
contain the following lines:
domain-needed
interface=wlan0
dhcp-range=192.168.20.100,192.168.20.200,12h
listen-address=192.168.20.1
We save the originally installed version under a new name because it contains explanations
of all parameters and is useful as a reference. For our system, however, we only need those
in the above example, and they are almost self-explanatory. Finally, we register the dnsmasq
daemon to start at boot time with the command
sudo service dnsmasq start
and reboot for good measure, to ensure that all parts of the system are synchronized.
Technically, this is not needed, but it might help, in case some parameter was changed
inadvertently during setup.
At this point we have turned the Raspi into a WLAN access point that spans the
messnetz WLAN. If the IP forwarding is enabled, we can even surf the web from any
computer on messnetz. But we are mostly interested in communicating with sensor nodes
that are connected to messnetz and will discuss how to use the Raspi as the spider in the
center of a network of sensor nodes connected by serial or WLAN links.
 Host Computer: Raspberry Pi  123

5.5 COMMUNICATION WITH THE ARDUINO

After the basic network infrastructure is set up, we connect the Arduino to the Raspi. Most
programs on the Raspi work in a similar way as on desktop computers, which is no surprise,
because a modern Raspi is about as powerful as a desktop computer from about a decade
ago. We start by communicating via the Arduino IDE first, but later we will also use the
Python programming language and octave.

5.5.1 Arduino IDE

We could install the Arduino IDE on the Raspi with sudo apt-get install arduino,
but the version in the repositories is rather old. It is better to install the most recent
version that is availabe from the https://www.arduino.cc web site, even for the Raspi.
To do so, we open a browser on the Raspi, direct it to the Arduino download web site,
choose the Linux ARM version, and download directly to the Raspi’s download directory
/home/pi/Downloads/. At the time of writing the current version was 1.8.0, and we may
need to adapt the version number to a later release.
It is customary to install large software packages under the /usr/local tree, and we
execute the following commands:
cd /usr/local
sudo tar xvJf /home/pi/Downloads/arduino-1.8.0-linuxarm.tar.xz
cd /usr/local/bin
sudo ln -s /usr/local/arduino-1.8.0/arduino arduino
where the tar command unpacks the downloaded archive into the subdirectory named
/usr/local/arduino-1.8.0/, and the ln -s creates a so-called soft link for the arduino
executable in the /usr/local/bin/ directory. In this way it is automatically found when
typing arduino at the command line in a terminal window. Optionally, we can create a
link in the Programming menu by starting the Main Menu Editor in the Preference section
behind the Menu button with the image of a raspberry. In the left panel of the Main Menu
Editor we select Programming, and then click on New Item on the right-side panel. In
the opening window, we enter the name of the new program, Arduino, and browse to the
arduino executable in /usr/local/bin directory. Once these steps are completed, there
is an Arduino entry in the Programming section of the main menu. Selecting it opens the
Arduino IDE on the Raspi, and we can start programming the Arduino directly from the
Raspi. Before using it we add support for ESP8266-based microcontrollers by following the
steps outlined at the end of Section 4.2 on page 58. On the Raspi, the procedure is just the
same.
Instead of retyping all the programs for the Arduino on the Raspi, we copy the files from
our desktop computer to the Raspi with scp, which is part of the secure-shell program suite
if it is a Unix-based system such as Linux or Mac. On a Windows system we use WinSCP.
On a Linux Desktop computer we execute
scp -r Arduino [email protected]:
which copies the Arduino subdirectory and everything in it to the Raspi, including all hard-
ware descriptions from the Arduino/hardware directory. Now we need to restart the Ar-
duino IDE, and all our previously prepared sketches are available under the File→Sketchbook
menu in the Arduino IDE on the Raspi. After we connect the Arduino UNO or NodeMCU
to an USB port on the Raspi, we are ready to program them from the Raspi.
Of course, we need to select the type of microcontroller and the port to which it is
124  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.5 Telnet session connected wirelessly to the NodeMCU.

connected the same way as on the desktop computer. In order to test that all is work-
ing properly, we download the query-response sketch to an Arduino UNO and open the
Tools→Serial Monitor in the Arduino IDE. This should allow us to communicate with the
UNO via the USB serial line. In case it does not work, ensure that the Port:’ in the Tools
menu points to the correct serial port with the Arduino UNO and that the UNO is actually
selected in the Board: menu.

5.5.2 From the command line

At this point we have established the Raspi as a development system for Arduino software,
and ensure it communicates via the USB-based serial line with the Arduino IDE. But
normally we want to access the Arduino from self-written programs rather than always
using the Arduino IDE. The first option is to use the screen program or any other terminal
program to connect to the UNO. After closing the Arduino IDE, we therefore start the
following program on the command line on the Raspi:
screen /dev/ttyACM0 9600
just as we did earlier when the UNO was connected to the Desktop computer. Bluetooth
communication works exactly the same way as on the desktop computer. Just follow the
details described in Section 4.6.2 on page 104.
In the next step, we use a NodeMCU and assume that the Router software from Sec-
tion 5.4 is installed on the Raspi. We change the variables ssid and password in the
NodeMCU sketches from Section 4.6.3 to those used on messnetz. Two lines near the top
of the sketches now need to read
const char* ssid = "messnetz";
const char* password = "zxcvZXCV";
where we need to make sure to enter the same password as the wpa passphrase in the
hostapd.conf file. We then connect the NodeMCU to a USB port on the Raspi and down-
load the updated socket-based server to the NodeMCU. We can observe progress of the
NodeMCU boot process with the Serial Monitor while it connects to the messnetz WLAN.
After a while it should report the acquired IP number that the hostapd on the Raspi dished
out. It lies in the 192.168.20.nn network, which corresponds to the IP range we specified
in Section 5.4, in the configuration files of dnsmasq. In my case the IP number received was
192.168.20.135, but yours may be different. At this point we can connect to NodeMCU
via WLAN by issuing the following command in a Raspi terminal window:
 Host Computer: Raspberry Pi  125

Figure 5.6Using Python on the Raspi to communicate with the query response
sketch running on the Arduino UNO.

telnet 192.168.20.135 1137

where 1137 is the port number that is specified in the sketch that runs on the NodeMCU.
The telnet program provides a prompt at which we can type commands to send to the
NodeMCU. So, typing A0? on the telnet prompt should result in the reply A0 nnn, also
displayed in the telnet window, as shown in Figure 5.5. We close the session by using the
command Ctrl-] and then type quit at the telnet prompt.
Instead of using telnet, we can also use netcat and type the following command in a
terminal window:
netcat -C 192.168.20.135 1137
where the -C option ensures that CR-LF characters are sent after each line, which is required
by the NodeMCU to recognize the end of a command. Otherwise the communication works
in the same way as with telnet.
The previous two paragraphs only verify that the communication works properly and
that we have our system under control. Now we proceed and use the Python language to
achieve the same feat.

5.5.3 Python
Python [24] is a modern programming language that is installed on any Raspi by default.
Actually, the Pi in Raspberry Pi stands for Python Interpreter. Unless you have some
experience with Python, I suggest you have a look at some of the tutorials at
https://wiki.python.org/moin/BeginnersGuide/Programmers
or search the web for detailed explanations of some of the simple, often even self-explanatory,
commands we use here. The purpose to access an Arduino or NodeMCU from Python is to
show how to write customized programs on the Raspi that work hand-in-hand with sketches
that run on the Arduino.
We use the following Python script to query the UNO with the query-response sketch
for a single value from analog pin 0.
# query_arduino.py
import serial, time
query="A0?\n"
ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
time.sleep(1) # wait for serial to be ready
126  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.7 Simple ASCII graphics from querying the UNO repeatedly.

ser.write(query)
time.sleep(0.1)
reply=ser.readline()
print(reply.strip())
ser.close()

Python is rather lightweight, and we have to import extra functionality such as serial com-
munication by executing
sudo apt-get install python-serial
at the command prompt and then load the new functionality with the import statement.
In the above script we import support for handling the serial line and basic time-handling.
The latter we need to implement delays in the script. In the second line we define a variable
query that contains the query string we send to the Arduino. Note that we explicitely add
the carriage return \n character. Then we open the serial port ser on /dev/ttyACM0 with
a baud rate of 9600 and a timeout of 1 second, which prevents unsatisfied read attempts
from blocking the program. We wait for one second to allow the operating system to finish
opening the serial port, and then submit the query string with the ser.write command.
Note that we use the method write on the serial device ser. Then we wait for 0.1 seconds
and read characters up to the CR-LF character with the ser.readline() function into the
variable reply. We display the reply with the print command after stripping off leading
and trailing white-space characters, such as normal spaces or CR-LF characters. Finally, we
close the serial line. We run the Python script that we give the name query arduino.py
by entering
python query_arduino.py
on the command line of the Raspi. We show this example in Figure 5.6.
The previous example script shows how to request a single measurement value. We easily
expand the script to query the UNO repeatedly, and even provide simple ASCII graphics
that show the measurement value as a function of time. This is accomplished by the Python
 Host Computer: Raspberry Pi  127

script below. Example output is shown in Figure 5.7, where we see that we call the program
by executing
python ask_arduino_repeat_plot2.py

and the program then shows the elapsed time since it started, the measurement value, and a
graphical representation of the value between the expected minimum and maximum values.
This functionality resembles a simplified version of the Serial plotter built into the Arduino
IDE. The Python script is the following:

# read arduino repeatedly from command line, V Ziemann, 161222

import serial, time, atexit
def cleanup(): # ensure serial line is closed after CTRL-c
ser.close()
atexit.register(cleanup) # register the cleanup function
query="A0?\n" # the query string
amin=0 # minimum expected value
amax=1024 # maximum value
width=70 # number of character used for plot
ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
lll=len(query)-1 #
time.sleep(1) # wait for serial to be ready
t0=time.time() # starting time
while 1: # repeat forever
ser.write(query) # send query
time.sleep(0.1) # wait a bit
reply=ser.readline() # read response
value=reply[lll:].strip() # make numeric
k=(width-1)*int(value)/1024 # where to place *
p=’%8.1f %4s |’ % (time.time()-t0,value)
for j in xrange(0,width-1):
if j==k:
p+=’*’
else:
p+=’ ’
p+=’|’
print(p)
time.sleep(1) # wait a while before next measurement

First we import support for serial communication, time, and the atexit functionality, which
allows us to register a function that does some cleaning-up activities when the program
terminates. Since we will employ an infinite loop to read the UNO, which we intend to
stop asynchronously with Ctrl-C, this ensures that the serial line is properly closed. Note
that Python uses indentation instead of brackets to define the scope of functionality, and
the function body for the cleanup() function is just ser.close(), but indented by a few
characters. After defining the query string and several variables, the serial line is opened,
and we determine the expected length lll of the characters preceding the value of the reply,
wait a short time, and record the present time in the variable t0. The while 1: statement
initiates a loop that runs forever. Inside the while loop (note the indentation to indicate the
scope), we write the query to the UNO, wait a short while, and read the response into the
string reply. In the following line, we remove the first few characters such as A0, and remove
128  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

white-space characters with the strip() method. The variable k scales the measurement
value such that it lies within the range specified in the width variable, and then we start
building the string p by first placing the seconds elapsed since t0 and the measurement
value followed by a vertical bar denoting the start of the ASCII graphic. Then we loop over
all subsequent places in the string p and place a * at the location corresponding to the
scaled measurement value k, and a space otherwise. Finally, we add another vertical bar |
and print the string p to standard output, before waiting one second for the next iteration
to start. Running this Python script results in the rudimentary ASCII graphics shown in
Figure 5.7.
The previous script works well with any serial line, including a serial port behind which
a Bluetooth link is hidden. The only change we need to implement in the previous script
is to replace the serial port /dev/ttyACM0 with the Bluetooth port that is typically called
/dev/rfcomm0 or /dev/rfcomm1.
But what about connecting to the NodeMCU running the socket server that listens on
port 1137? We only show the most basic network client that sends the query and displays
the reply on standard output. More elaborate examples such as the simple ASCII graphics
from above can be built quite easily once the basic network communication setup is under
control. We illustrate these basics in the following example:
import socket, atexit, time
def cleanup():
sock.send("quit\n")
sock.close()
atexit.register(cleanup)
sock=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
sock.connect(("192.168.20.135",1137))
sock.send("T?\n")
time.sleep(0.1)
reply=sock.recv(1000);
print(reply.strip())
The structure of the program is very similar to the one for serial communication. First,
the necessary functionality is imported, and then we define a cleanup() function that
executes before the program closes. This is the safest way to ensure that the server closes the
connection properly. Inside the cleanup() function, we instruct the python script to send
the string quit, which is unknown to the NodeMCU and causes it to close the connection,
before closing the socket on the client side. The function cleanup() is then registered in
the call to the atexit.register() function. In the next line the socket sock is created
with the specification of a normal TCP/IP socket, before connecting to the server running
on the NodeMCU at IP number 192.168.20.135 and port 1137. Then we send the query
string T? and wait a short while before we receive the reply, and then print it to standard
output after stripping white-space characters.
In this section, we cover how to interface the sensor nodes with Python, and even present
them as rudimentary ASCII graphics. This might suffice for quick-and-dirty fixes to rapidly
observe how some value changes with time, but for more professional-looking results, we
turn to octave using it to interface the sensor nodes and prepare much nicer plots for
presentations and reports.
 Host Computer: Raspberry Pi  129

5.5.4 Octave
We already installed octave in Section 5.3, but in order to add packages, we need a few
octave development tools that we install from the command line by entering
sudo apt-get update
sudo apt-get install liboctave-dev

which takes a few minutes to complete. For interfacing serial and network devices from
within octave, we need to install the instrument-control toolbox. Unfortunately, it is not
available in the normal repositories such that we can install it with apt-get. Instead, we
need to use an alternative way to make it accessible. Once the installation of liboctave-dev
finishes, we type octave on the command line and install the instrument-control toolbox
from within octave by entering
pkg install -forge instrument-control
at the octave prompt. Specifying the -forge option causes octave to download the most
recent version of the toolbox from the octave forge at https://octave.sourceforge.io
and install it on the Raspi. Note that the installation process takes up to 30 minutes and
does not report any progress. Just be patient! After completion, we add the following line:
pkg load instrument-control
to the octave startup file named .octaverc in our home directory, which in most circum-
stances on the Raspi is located in the /home/pi/ directory. This will cause the toolbox to
be loaded every time we start octave. While we edit that file, we may also add the line
graphics_toolkit(’gnuplot’)
to the .octaverc file to avoid a bug when using the octave plot command. But this may
not be needed in all circumstances.
Once the installation of the instrument-control completes, we are ready to use it to
communicate with our sensor nodes. First, we try to communicate with the UNO connected
to the USB port of the Raspi, which is accessible as /dev/ttyACM0 as before. In order to read
one measurement from the UNO, we open the serial line from the octave prompt, submit
the query, wait for the response, and display the result. The following program achieves
this.
s=serial("/dev/ttyACM0",9600); % open serial line
sleep(1) % wait for this to complete
reply=queryResponse(s,"A0?\n") % send query and receive reply
fclose(s); % close serial line
It implements the simple query-response communication protocol we used earlier. Here
we encapsulate the details of the query-response interaction in a separate function
queryResponse, because there is no native support in octave to read from the serial device
up to a termination character, and we implement that feature in the following function.
% send query and return reply up to termination character.
function out=queryResponse(dev,query,term_char)
if (nargin==2) term_char=10; end % defaults to LF=0x0A
srl_write(dev,query); % send query to device
i=1;
130  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

int_array=uint8(1);
while true % loop forever
val=srl_read(dev,1); % and read one byte
if (val==term_char) break; end % until term_char appears
int_array(i)=val; % stuff byte in output
i=i+1;
end
out=char(int_array); % convert to characters
The three arguments are the serial device dev, the query string, and an optional termination
character, which defaults the line-feed character (hex=0x0A, dec=10). In the function, we
first send the query string in the srl write() function call, and after initializing some
variables, we repeatedly read one character at a time with the srl read() function and
append the character to the int array unless it equals the termination character. Once
the termination character is received, the while true loop exits, and the int array is
converted to characters in the last line. The converted array is returned to the calling
program. Needless to say, we can also use any Bluetooth device with a serial line interface,
only the device file is /dev/rfcomm0 instead of the normal serial device files /dev/ttyACM0
or /dev/ttyUSB0.
Reading via WLAN by connecting to the socket on the NodeMCU is done in much the
same way as reading from the serial line. Instead of opening a serial line and using the
srl write and srl read functions, we open a TCP connection and use the tcp write and
tcp read functions. The basic code snippet that accomplishes this is shown here:
s=tcp("192.168.20.135",1137); % open connection
sleep(0.1) % wait, not really needed
reply=queryResponseTcp(s,"T?\n") % send query and get reply
tcp_write(s,"quit\n"); % close remote socket
tcp_close(s); % close local socket
which follows the logic from the serial communication example above. The function
queryResponseTcp() is essentially the same as the above example for the serial line, only
the srl write and srl read functions are replaced by the tcp write and tcp read func-
tions, and we therefore do not reproduce the code here. Note also that we need to explicitely
close the remote socket on the NodeMCU by sending quit.
To illustrate the usefulness of the octave interface to serial line or networked devices, we
write a simple temperature logger that connects to the NodeMCU and produces a nice plot
of the temperature as a function of time. The following octave program achieves that.
% temperature logger, V. Ziemann, 161227
clear all
s=tcp("192.168.20.135",1137);
sleep(0.01)
running=0;
while running<10
running=running+1;
reply=queryResponseTcp(s,"T?\n");
val(running)=str2double(reply(2:end));
x(running)=now;
plot(x,val,’*’)
ylim([22,28]);
 Host Computer: Raspberry Pi  131

Figure 5.8 Plot from the temperature logger.

ylabel(’Temperature [C]’)
xlabel(’Time’)
datetick(’x’,’ddd/HH:MM:SS’)
sleep(1);
end
tcp_write(s,"quit\n");
tcp_close(s);
First, we clear all variables before we establish the TCP connection to the server on the
NodeMCU, on port 1137 and IP number 192.168.20.135. Then we wait a short while and
initialize a variable running to zero. We use this variable as an iterator in the loop and
check whether the limit has been reached. In this simple example, we only iterate 10 times.
In the loop, we increment the running variable and then send the query to the NodeMCU
and receive the measurement in the string reply. Note that reply starts with a T, which
we remove in the following line by converting reply from position 2 to the end to a double
variable. The result we copy to the variable val at index position running. In this way we
use the iterator not only to count the loop iterations, but also as an index of where to put
the measurement values. In the variable x we copy the current time, which is returned by the
built-in function now(). Then we plot the value versus the time using the plot() function,
specify the vertical temperature range with the ylim() function, and specify the axis labels
for vertical and horizontal axes. Finally we use the datetick() function to specify the type
of tick marks we want. In this case we specify the day of the week and the time in hours,
minutes, and seconds. The datetick() function uses the special format in which the now()
function returns the time to extract the desired format of the tick mark. The full set of
options is explained in the help text of the datestr() function that can be accessed by
typing help datestr at the octave prompt. Before repeating the measurement, we wait a
specified time, one second in the example. Finally, once the desired number of iterations is
completed we ensure that the socket is closed on both the server and in the octave client.
We show the resulting plot in Figure 5.8, where we run for 1000 iterations instead of just
10.
132  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

By now we have managed to record data from sensor nodes via serial and WLAN and
display them online, even with reasonably attractive graphics, but for this privilege we
need to have octave running while taking data. This may be feasible for a period of a few
hours, but is hardly attractive for longer periods. In that case we need a separate repository
to store the data, and only create plots when needed. We therefore need to separate the
presentation from the data storage process, and this is the topic of the next section.

5.6 DATA STORAGE

In this section, we address several ways to store our measurement data, and the archetypical
data storage repository is a database. We introduce a number of databases, but in most cases
we use octave for the presentation of the data.

5.6.1 Flatfile
The simplest database is certainly a file containing a time stamp and one or more measure-
ment values, possibly even in human-readable form. This is called a flatfile database, and
we create one by using the following Python script:
# logger with time, V Ziemann, 161228
import serial, time, sys, atexit
def cleanup():
ser.close()
atexit.register(cleanup)
query="A0?\n"
ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
time.sleep(1) # wait for serial to be ready
while 1:
ser.write(query)
time.sleep(0.1)
reply=ser.readline()
print int(time.time()), reply[3:].strip()
sys.stdout.flush()
time.sleep(1)
and store it in a file called ask repeat.py. We recognize the same organization as before,
when we discussed Python scripts with importing functionality, registering the cleanup()
function, opening the serial line, and repeatedly sending the query string and receiving the
reply. The only difference is that we print the Unix time, which is the number of seconds
since January 1, 1970, also called the epoch, before the measurement value. In the output,
we therefore see the following text
:
1482954187 680
1482954188 678
1482954189 681
:
scrolling by. If we redirect this output into a file using the following line of code:
python ask_repeat.py > db.dat
 Host Computer: Raspberry Pi  133

that we run in a terminal window on the Raspi, we create a flatfile database db.dat with
the time stamps and the measurement values. Later we can retrieve and convert the data
to a nice plot.
Running the data acquisition program in a dedicated terminal window may be useful
for short measurement sessions, but is hardly useful if we want to log, for example, the
temperature in a building for an extended period of time, say a few months. In that case it
is much preferable to have a dedicated process that wakes up once every few minutes, reads
the temperature, and stores the value together with a timestamp in a file before going back
to sleep for a few minutes. As it turns out, Unix systems normally have a system that takes
care of these repeated tasks. It is called cron and it is a background process that wakes up
once a minute, checks whether there is a task to do, does it, and sleeps for a minute before
checking again.
Let us start by creating a Python script that we want to execute at regular inter-
vals. In this particular case we place it in the /home/pi/bin directory and call the file
single request.py. It contains the following lines:

# single_request.py, V Ziemann, 161229

import serial, time
ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
time.sleep(1) # wait for serial to be ready
ser.write("A0?\n")
time.sleep(0.1)
reply=ser.readline()
print int(time.time()), reply[3:].strip()
ser.close()
We see that it is rather similar to the previous scripts to read from the Arduino. It just
acquires one measurement and prints the value with the timestamp to standard output. In
order to make this program accessible for the cron software, it is convenient to encapsulate
it in a shell script file /home/pi/bin/readA0.sh. It contains the following lines:
#!/bin/bash
/usr/bin/python /home/pi/bin/single_request.py >> /home/pi/A0.dat

We make the file executable with the chmod program by executing chmod +x readA0.sh in
the /home/pi/bin directory. The first line instructs the operating system to interpret the
following lines using the bash shell. The next line starts the single request.py script using
the /usr/bin/python program, and redirects the output to the file /home/pi/A0.dat. Here
>> implies that new data is appended to an existing file. Note that all file names must be
given, including their absolute path. We test the readA0.sh script to ensure that it creates
the /home/pi/A0.dat file or appends a reasonable measurement value with timestamp to
the file. Once we are satisfied, we register readA0.sh with the cron software and edit the
configuration file for the cron program with the crontab program. We execute it from the
command prompt by typing
crontab -e

The first time crontab -e is called it asks for an editor. We pick our favorite or follow the
suggestion. Once the editor opens with the configuration file, we append the following line
at the end of the file:
* * * * * /home/pi/bin/readA0.sh
134  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

and save the file. This automatically registers the readA0.sh script to execute once every
minute whenever the Raspi is running. We check that the file is registered with the crontab
-l command. It lists the contents of the crontab file for user pi provided we are logged
on as user pi. The meaning of the six columns in the crontab file are minute, hour, day
of month, month, day of week, and program to execute. Placing asterisks in the first five
columns instructs cron to execute the program every time it wakes up. The output of the
crontab -l command has some basic explanations, and more is available by executing man
5 crontab at the command prompt. So, now we have a background process that records a
measurement once a minute and fills the file /home/pi/A0.dat until we remove the entry
in the crontab file.
Our next task is to produce plots, preferably with the timestamps taken into account
adequately. For this task we use octave, because it has powerful capabilities to handle
timestamps. We extract the data stored in the flatfile database and display it with the
following script.
# flatfile viewer, V. Ziemann, 161229
d=importdata(’/home/pi/A0.dat’);
TZ=1; t=719529+(d(:,1)+TZ*3600)/86400.0;
plot(t,d(:,2))
datetick(’x’,’HH:MM:SS’)
Most of the work is done by the importdata() function. It figures out the format of the
A0.dat file and loads its contents into the variable d, which is a matrix with two columns,
one with the time information and the other with the measurement values. In the next line
we convert the time into the standard format that the datetick() function expects. It will
properly format the displayed time on the horizontal axis. Note that we use the variable TZ
to denote the time zone. Since I live in Sweden, the local clock is one hour ahead of UTC
standard time, to which the epoch refers. The constant 719529 is the number of days from
January 1, 0000 until January 1, 1970, and 86400 is the number of seconds per day. The
plot looks similar to the one shown in Figure 5.8.
After storing data in a flatfile database, retrieving it, and displaying the data, we now
progress to using a more mature database, MySQL.

5.6.2 MySQL
Among several databases available for the Raspi, we choose MySQL [25] because it is widely
used and can be accessed reasonably easily from most programming languages. These in-
clude Python and, after some archaeology on the Internet, octave as well. We install the
MySQL database from the standard repositories with the following command:
sudo apt-get install mysql-server
The installation of mysql-server triggers a request to enter a password for the mysql
administrator, which is called mysql-user root. So make up your mind and invent a good
one. Note that this user root is the database administrator and not the superuser for the
computer. Just the name root happens to be the same.
Once MySQL is installed, we create a database named dataA0 that will contain roughly
the same information as the flatfile in the previous section, namely a timestamp and mea-
surement values. Creation of a new database must be done by mysql-user root, thus we first
need to log onto the MySQL administration account by executing the following command:
mysql -u root -p
 Host Computer: Raspberry Pi  135

and entering the password of the mysql-administrator root during installation, whence we
are greeted by the MySQL prompt mysql>. At the prompt, we create the new database
with the following command:

create database readA0;

where we specifically need to point out the necessity to complete every MySQL command
by a semicolon. We verify that the database readA0 exists with the show databases;
command at the MySQL prompt. Since later we will not always want to work with the
database readA0 as mysql-administrator, we create a database user. To do so, we declare
that we want to work with readA0 by writing
use readA0;
at the mysql prompt. Now we create the user me and then grant privileges to work with the
database, with the following sequence of commands:

create user ’me’@’localhost’ identified by ’pwpw’;

grant all privileges on readA0.* to ’me’@’localhost’;
where pwpw is the (too simple, invent a better one) password. Luckily the MySQL syntax
is fairly self-explanatory. Note that the granting of privileges can be tailored in a rather
detailed fashion into reading, writing, and other privileges, but for our simple example we
stick to the simple version. Once the general administration of creating a database and
assigning a mysql user is done, we exit from the mysql prompt by typing quit;.
But now we log onto MySQL as mysql-user me by typing
mysql -u me -p
at the command prompt and entering the password. Alternatively, the command mysql -u
me -ppwpw with the password pwpw immediately appended after the -p will directly log
onto the mysql-prompt, but now as the user me who only has privileges to work with the
readA0 database. We verify this by executing the show databases; command, which shows
readA0 and some other administrative database that we ignore for the time being. In order
to create a data structure, called table in our database, we select it with use readA0; and
then enter
create table dat (ts timestamp, A0 integer, A1 integer);
which creates a table dat that stores a time stamp and two values in every row. Other
options for things to store are float and BLOB, the latter being a binary large object.
The MySQL manual, available at http://dev.mysql.com/doc/, holds a wealth of detailed
information.
The administrative task to create the table is done at this point, and we could quit;,
but it is instructive to insert values into the newly created table by executing the following
command at the mysql-prompt
insert into dat (A0,A1) values (17,4);

and verify the contents of the database with the command

select * from dat;
136  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

which prints the contents of the table. We observe that the timestamp variable was auto-
matically filled with the time the values are inserted. The syntax to insert values is, again,
rather self-explanatory. We insert into a table called dat the variables (A0,A1) with
values 17 and 4. Reading all values from the table is achieved by the select command.
The way it is specified with the asterisk displays the entire table dat. If we only want
to print the timestamp ts and value A1, we can issue select ts,A1 from dat;. These
commands to create, insert, and select data from the table are standardized and are called
Structured Query Language or SQL. Please inspect tutorials and further information about
SQL on the Internet. After this exercise we exit the mysql program by executing the quit;
command.
Now we know the basic SQL commands to insert and retrieve data from tables. But
rather than typing them from the mysql prompt, our next task is to use Python to
execute the SQL commands. We add MySQL functionality to Python by installing the
python-mysqldb package with the command
sudo apt-get install python-mysqldb
at the command prompt, and are ready to access MySQL databases from within Python.
As a first attempt, we read the table dat from the database readA0 we had created earlier.
The following Python script achieves just that.
# access MySQL database, V. Ziemann, 170101
import MySQLdb
db=MySQLdb.connect("localhost","me","pwpw","readA0")
cur=db.cursor()
cur.execute("select * from dat;")
reply=cur.fetchall()
print(reply)
#for r in reply: # loop over entries
# print r # print each entry
# print str(r[0]), str(r[1]), str(r[2]) # format nicely
db.close()
First the library with MySQL support is imported, and then we connect to the database
readA0 on computer localhost, as user me with password pwpw. The MySQLdb.connect()
function returns a handle db to the database, and executing it corresponds to logging onto
the database and executing the use readA0; command at the mysql prompt. The next line
creates a cursor cur, which is equivalent to the mysql prompt and allows us to enter database
insertion or retrieval commands. In the next line we execute the select command to display
the entire database. We retrieve the output from the command in the variable reply with
the fetchall() call and print the reply before closing the database. Note the structure
of first executing a MySQL command and then retrieving the reply with the fetchall()
function. The reply, however, is a list of entries in an unfamiliar form. It is possible to
display each measurement entry on a separate line by using the lines commented out by a
hash (#). The code loops over each entry in the list, and in the first commented example,
the print command displays one entry at a time. In the second example, the print command
uses the str() function to convert each entry to a string which makes it human readable.
Finally, we close the database with the call to the db.close() function.
Our next task is to insert new entries with measurements into the database, and this is
accomplished by the following Python script.
# serial2msyql.py, V Ziemann, 161229
 Host Computer: Raspberry Pi  137

import serial, time, MySQLdb

ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
time.sleep(1) # wait for serial to be ready
ser.write("A0?\n")
reply=ser.readline()
val0=int(reply[3:].strip())
ser.write("A1?\n")
reply=ser.readline()
val1=int(reply[3:].strip())
ser.close()
db=MySQLdb.connect("localhost","me","pwpw","readA0")
cur=db.cursor()
sql="insert into dat (A0,A1) values (%d,%d);" % (val0,val1)
cur.execute(sql)
db.commit()
db.close()
The first part of the script is very similar to earlier scripts. It just queries the Arduino
UNO on the serial line and converts the measurement values from analog pin A0 and A1 to
integers, which are stored in the variables val0 and val1. In the second part we connect to
the database readA0 and obtain a cursor. We then build the sql string with the MySQL
insert command and execute it in the next command, before committing the changes to the
database and closing it. Note that we could perform several transactions in a row that might
leave the database in a bad intermediate state. To prevent this from happening, all insert
commands are buffered and all changes are committed to the database simultaneously with
the commit() function call. In order to automatically record data, we add the line
* * * * * /usr/bin/python /home/pi/python/serial2mysql.py
to the crontab file using the command crontab -e that we already discussed in the previous
section on flatfile databases.
Now that we have a process that continuously fills the database, we want to use the
stored data in octave for postprocessing and to prepare plots for presentations. For this
we need a function that allows us to access the database from within octave, which turns
out to be difficult, because MySQL is not supported by either the Raspberry Pi or oc-
tave repositories. There is, however, code to bind MySQL to MATLAB, available from
http://www.courant.nyu.edu/~almgren/mysql, and it turns out that it also works well
with octave. We download the mysql.cpp and mysql.m files from the above web site and
prepare the octave version by running
mkoctfile --mex -I/usr/include/mysql -L/usr/lib/mysql\
-lmysqlclient mysql.cpp
in a command window on the Raspi, which produces a file named mysql.mex. We then
copy both mysql.mex and mysql.m to the directory where it can be found by octave at run
time; for example, the directory with the script that we are about to write. We start octave
in that directory and type help mysql.m at the octave prompt for an explanation of the
syntax of how to use the mysql() command in octave.
In the same directory, we prepare an octave script dbread.m to read the values from the
database and display them.
% dbread.m, V. Ziemann, 170101
138  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

mysql(’open’,’localhost’,’me’,’pwpw’);
mysql(’use readA0’);
sql=’select * from dat;’
[t,a0,a1]=mysql(sql);
mysql(’close’);
tt=datenum(t,’yyyy-mm-dd HH:MM:SS’);
plot(tt,a0,tt,a1)
legend(’A0’,’A1’)
datetick(’x’,’ddd/HH:MM’)
The dbread.m script uses the mysql mex file from the previous paragraph and opens the
database before using the readA0 database. Note that there must not be a semicolon at the
end of this command, for unknown reasons. In the next line the SQL command is written
into the string sql before we execute it with the mysql() function call. It directly returns
all variables from the database into variables t, a0, and a1. Note that we do not need
the equivalent of Python’s fetchall() call; it is already built into the mysql() function.
After retrieving the requested data, we close the database connection and convert the time
stamp data that arrives in string format from the database to the datenum format, with
days since January 1, 0000. We supply the format string to aid the conversion. Finally, we
plot the data, provide a legend, and specify the tick marks on the horizontal axis to include
the day of the week, hours, and minutes.
In the previous examples we always select all available data from the database. In many
circumstances we prefer to restrict the displayed data to a smaller range. Instead of filtering
in octave, we instruct the database to only return data from the restricted time window.
To achieve this we use the following SQL query string:
sql=’select * from dat where ts > "2017-01-01 17:30"
and ts < "2017-01-01 17:55";’
instead of the command select * from dat; used earlier. Note that the command needs
to be written onto a single line. This query string explains the restricted time window in
clear text. We require the timestamp ts to be larger than some date and less than some
other date. The values returned from the command [t,a0,a1]=mysql(sql); thus only
contain data from within the requested time window.
Hopefully this short introduction to MySQL and how to access it from Python and
octave is useful to get you started in case a database is needed in a project. But now we will
turn to a second database, one that only stores values over a finite time-horizon and also
thins them out the further back in time the data originate. This database is the round-robin
database called rrdtool that we discuss in the next section.

5.6.3 RRDtool
The rrdtool [26] program was initially conceived as a tool for allowing computer-network
administrators to present network traffic over different time horizons (last hour, day, week,
month, year) in a convenient and flexible way. It generates graphical representations of
data that can be shown in a web browser by automatically generating consolidated data,
such as average, minimum, or maximum over some period of time. Rrdtool is particularly
useful to generate plots of the measured data on the fly, and we later use it to display
our measurement data, such as temperatures, on a web page. The round-robin database is
implemented as a circular buffer that is filled up to the end and then wraps around and
starts to overwrite the oldest values at the beginning of the buffer. Using rrdtool comprises
 Host Computer: Raspberry Pi  139

three steps: creating the database, filling it with data, and extracting the data. But before
delving into examples, we need to install the software with the following command:
sudo apt-get install rrdtool
after having updated the repositories with sudo apt-get update (just a friendly reminder
not to forget the update step). Now we are ready to use the software.
In what follows, we assume that all files reside in the subdirectory /home/pi/rrdtool.
The first step—creating the database—is achieved by the rrdtool create command that
we enter at the command prompt of the Raspi. In this step we define the frequency of storing
data, the type and valid range of data, and the way the stored data should be preprocessed;
more on the last point later. The simplest example, namely to create a database db1.rrd,
is shown in the following example:
rrdtool create db1.rrd --step 60 \
DS:temp:GAUGE:180:-20:100 \
RRA:AVERAGE:0.5:1:2880
where we can also omit the backslash and write the entire command on a single line.
The first part of the command creates db1.rrd and the database stores values every 60
seconds. The second line defines the data source (DS:), a variable called temp that is of
type GAUGE, which is rrdtool-speak for “measurement value.” We require a valid data point
to be uploaded to the database at least every 180 seconds before a value is marked as
invalid. The expected range of values for the data points is between -20 and 100, which is
reasonable for a temperature reading. The line starting with RRA defines the round-robin
archive that contains averaged values; the number 0.5 is used internally and should not
be changed; the number of data points to be averaged, here 1; and the total number of
(averaged) data points that the database should hold. In the example we use 2880, which
is the number of minutes in two days. Since we chose to average only one data point, we
store all values, rather than actually averaging. The above rrdtool command creates the
file db1.rrd in the directory where the command is executed. Note that we can create
several data sources; for example, for temperature, humidity, and barometric pressure, by
adding DS: statements. Moreover, note that the RRA: line creates one table in the database
db1.rrd. We can define several more tables with additional RRA: statements. For example,
adding RRA:AVERAGE:0.5:30:336 will create a table with data averaged over 30 readings,
thus one point every 30 minutes, and will store 336 values, which corresponds to one week,
because there are 336 half-hour periods in a week. Other options, instead of AVERAGE, are
MIN and MAX, which will store the minimum or maximum in the specified time period,
respectively. Please consult man rrdtool for more options. Now that we have a database,
we can start to fill it with data.
We fill the database with the rrdtool update command. The data we store are tem-
perature measurements from an LM35 temperature sensor attached to an Arduino UNO,
similar to the way we used it before. The following Python script sends T? to the UNO and
receives the temperature data as a string that is similar to T 22.5.
# read temperature, V Ziemann, 170102
import serial, time
ser=serial.Serial("/dev/ttyACM0",9600,timeout=1)
time.sleep(1) # wait for serial to be ready
ser.write("T?\n")
reply=ser.readline()
print reply[2:].strip() # just print temperature
140  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.9 Graphics from rrdtool.

ser.close()
The difference from earlier versions is that we print out the numerical value only. We call
this Python program from a shell script file with the name filldb1.sh. It contains the
following lines:
#!/bin/bash
DB=/home/pi/rrdtool/db1.rrd
TEMP=$(/usr/bin/python /home/pi/rrdtool/readtemp.py)
/usr/bin/rrdtool update $DB N:$TEMP
and we make it executable by executing chmod +x filldb1.sh. In the script, we first
define a variable DB that contains the absolute path to the database file. It resides in
/home/pi/rrdtool/ as already mentioned above. On the next line we fill the variable TEMP
with the output of the command between $( and ), but this is the temperature value that
the command /usr/bin/python /home/pi/rrdtool/readtemp.py returns. Note that the
absolute path is used for both the Python interpreter and the readtemp.py script. In the
last line we execute the rrdtool update command with the database name stored in the
variable DB, and fill it with the current time stamp, as indicated by N: and the temperature
value stored in the variable TEMP. Executing filldb1.py from the command line sends
a data point with the current time to the database, but that is rather inconvenient. A
better solution is to run filldb1.py once a minute, automatically. We therefore update
the crontab file by executing crontab -e and add the line
* * * * * /home/pi/rrdtool/filldb1.sh
to the end of the file. This will send a new data point to the database once every minute.
Finally, in the third step, we retrieve the data and generate a graph with the rrdtool
graph command. Here is an example:
rrdtool graph db1.png -s -4h \
-t "Temperature in my office" -v "T [C]" \
DEF:t0=db1.rrd:temp:AVERAGE \
LINE1:t0#FF0000:"Temperature";
Note that we also can write the entire command on a single line. It creates a graphics file
db1.png with start of the time axis 4 hours back, -s -4h in time, a title specified after the
-t option, and the vertical axis label after the -v option. In the next line the handle t0 is
defined to refer to the data coming from the db1.rrd database and as the AVERAGE of the
 Host Computer: Raspberry Pi  141

Figure 5.10 Web page from the Raspi after installation of the Apache2 web server.

variable temp. Compare this to the rrdtool create command, where temp was defined in
the DS: part of the command and AVERAGE in the RRA: part. Finally, we define a displayed
line to use the handle t0, specify the color by hexadecimal RGB values (here red: #FF0000),
and give it the label Temperature. In Figure 5.9 we show the resulting graphics produced
by the above rrdtool graph command. Note that only the last four hours are plotted, but
the first part is missing, because the cron job was not running at that time. The data are
therefore noted invalid in the database, and consequently not printed. The scale is adjusted
automatically in the previous examples, but can also be specified explicitly. The command
man rrdtool and the pointers therein provide a wealth of information on how to fine tune
the output.
After being able to measure, store, and retrieve measurement values, we want to present
them on a web page for online observation and continuous checking.

5.7 ONLINE PRESENTATION

In this section we describe how to present the measured data on a web page that is published
by a web server. It turns out that the Raspi is quite capable of running a web server on the
side while doing all the other things such as acquiring and storing data. The basic software
needed is a web server, and we chose apache2 [27], which is installed by executing
sudo apt-get install apache2
on the command line of the Raspi. After the installation is complete, we use a browser on
any computer connected to the network the Raspi is connected to and enter the IP number
of the Raspi in the address line. This also works on the browser installed on the Raspi;
142  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 5.11 The first self-made web page.

enter localhost as the web address, and you should be greeted by the web page shown in
Figure 5.10. This page instructs us to replace the file named index.html in the directory
/var/www/html/ with our own copy. Note that the apache2 web server is configured to
present a file named index.html by default, if only the address of the folder is given.
So, we need to prepare a simple web page, which is a specially formatted text file in
a format called HTML—an acronym for Hyper-Text Markup Language. There are plenty
of books on the subject, and tutorials can be found on the web in abundance. Here we
only use the most basic features. We navigate to the directory /var/www/html/ and save
index.html as a backup with mv index.html index.html.bak. Then we start our favorite
editor as superuser by prepending sudo, enter the following text, and save the contents as
index.html.
<!DOCTYPE HTML>
<HTML>
<HEAD>
<TITLE>Raspi Web Server</TITLE>
</HEAD>
<BODY>
<H1 ALIGN=CENTER>Raspi Web Server Main Page</H1>
<HR SIZE=2 WIDTH=80%>
<H3>Available Goodies:</H3>
<UL>
<LI> <A HREF="temp/">Temperature Graph</A> </LI>
<LI> <A HREF="http://www.w3schools.com/html">
HTML Tutorial</A> </LI>
<LI> <A HREF="http://www.raspberrypi.org">
Raspberry Pi web site</A> </LI>
<LI> <A HREF="http://www.arduino.cc">Arduino web site</A> </LI>
</UL>
</BODY>
</HTML>
 Host Computer: Raspberry Pi  143

If no typos crept in, by entering the address http://localhost on a browser running locally
on the Raspi we should see the page shown in Figure 5.11. This is how the web browser
renders the contents of the file we just entered and now briefly discuss. At the top of the file
we have the declaration of the document type; it is an HTML file. Then we have the opening
tag <HTML> and a matching closing tag </HTML> at the end of the file. Anything between
them describes the contents of the web page. Note that HTML tags (almost) always come
in pairs: the tag name in angle brackets and a matching closing tag with the same name,
but prepended with a slash. The next tags we encounter in the file are HEAD and TITLE,
which describe things that do not show up on the page but appear in the title bar of the
web browser. Finally, we reach the BODY tags, where we find the description of the web page
proper. The <H1> tag declares a large headline, and the ALIGN directive specifies it to be
centered on the web page. There are different levels of header tags, from H1 to H6. The next
line defines a horizontal rule with the <HR> tag that covers 80 % of the width of the page.
Then we add a smaller header with <H3> and an unnumbered list between the <UL> tags,
with each list item described by <LI> tags. The <A> tag is called an anchor. It points to
other web sites specified in the HREF directive. The first list item points to a local directory
temp/ under the directory where the file index.html resides. Since this does not yet exist,
we need to create it.
In the directory /var/www/html/, we create the temp/ subdirectory and copy a file,
also named index.html, with the following contents into the newly created subdirectory
/var/www/html/temp/:

<!DOCTYPE HTML>
<HTML>
<HEAD>
<TITLE>Raspi Web server</TITLE>
</HEAD>
<BODY>
<H1 ALIGN=CENTER>Temperature</H1>
<IMG SRC="db1.png" ALT="Temperature in my office">
</BODY>
</HTML>

The file contents follows the same general layout as before, with the DOCTYPE declared first,
followed by <HTML> tags. Next come the <HEAD> tags and then the <BODY> tags bracketing
the displayed contents of the page. Here we also find a header and a new tag <IMG> to direct
the web browser to display the image specified in the SRC directive. To make this work, we
copy the file db1.png, the one we created with rrdtool graph in the previous section, to
the directory /var/www/html/temp/. We always need to use sudo to edit or copy files to
the system areas to which the files under /var/www/ belong. This we can avoid by enabling
private web pages.
The private web pages commonly reside in a subdirectory called public html under
the user’s home directory. As user pi we therefore create it by typing mkdir/home/pi/
public_html. In order to use it we have to enable the userdir module by executing sudo
a2enmod userdir at the command prompt, and restarting apache2 with the command sudo
service apache2 restart, such that the newly enabled module is loaded. Then we copy
the files index.html and db1.png from /var/www/html/temp/ to /home/pi/public\_html
and are ready to access the same web page as before, but now under the new address
http://localhost/~pi. Any file we copy to the subdirectory /home/pi/public\_html is
then accessible from a browser at the address http://localhost/~pi/ with the filename
144  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

User pi’s home page that displays the continuously (but slowly) updated
Figure 5.12
temperature.

appended. For example, http://localhost/~pi/db1.png only displays the graphics file in

the browser window, but nothing else.
The contents on the web page in the previous example is static. We copy files to the
public html directory and then they are presented as is. If we want to update, for example,
the temperature graph, we have to run rrdtool graph again and copy the new copy of the
db1.png to public html. But this is a task that we easily automatize with the help of a
cron job. We prepare a file with the following contents:
#!/bin/bash
DB=/home/pi/rrdtool/db1.rrd
/usr/bin/rrdtool graph /home/pi/public_html/db1.png -s -4h \
-t "Temperature in my office" -v "T [C]" \
DEF:t0=$DB:temp:AVERAGE \
LINE1:t0#FF0000:"Temperature";
/usr/bin/rrdtool graph /home/pi/public_html/db2.png -s -2d \
-t "Temperature in my office" -v "T [C]" \
 Host Computer: Raspberry Pi  145

DEF:t0=$DB:temp:AVERAGE \
LINE1:t0#FF0000:"Temperature";
and name it makegraph.sh. Then we place the file in the /home/pi/rrdtool/ subdirec-
tory and make it executable with chmod +x makegraph.sh. In the file, we first define the
database to use and then have two almost identical copies of the rrdtool graph command
from the previous section. But in the first case, we create db1.png for data from the last 4
hours, and in the second case we create db2.png for data from the last two days (-s -2d).
Note that the graphics files db1.png and db2.png are specified to reside in the public html
directory, where they are accessible to the web server. Finally, we make the makegraph.sh
script execute every 10 minutes by adding the following line to our crontab file with the
command crontab -e
*/10 * * * * /home/pi/rrdtool/makegraph.sh > /dev/null
where */10 in the first column means every 10 minutes and the > /dev/null at the end
means to suppress any output from the makegraph.sh command. So now we can update
the to-be-displayed content, but we still need to coax the web browser to actually re-read
that content at some interval. For this purpose the <META http-equiv=..> tag exists, and
we include it between the HEAD tags, as shown in the following updated version of the
public html/index.html file
<!DOCTYPE HTML>
<HTML>
<HEAD>
<TITLE>Raspi Web server</TITLE>
<META http-equiv="refresh" content="300">
</HEAD>
<BODY>
<H1 ALIGN=CENTER>Temperature</H1>
<H3>The last 4 hours</H3>
<IMG SRC="db1.png" ALT="Temperature over 4 hours">
<H3>The last 2 days</H3>
<IMG SRC="db2.png" ALT="Temperature over 2 days">
</BODY>
</HTML>
which instructs the browser to reload the page every 300 seconds, as indicated by the META
tag. At the same time, we added the second image that displays the temperature over the
last two days. A screen shot of the web page is shown in Figure 5.12. We can change the
update frequency in both the crontab file and in the META tab if we are impatient. They do
not need to match.
At this point we can obtain, store, retrieve, and display measurements on an actively
updated web page. The update mechanism that we implemented using cron jobs is very
rudimentary, and there are better solutions; for example, cgi-bin or php server-side pro-
grams. They execute code on-demand at the press of a button on the web page, and update
the displayed information on the fly, but that is beyond the scope of our presentation.
So far we have used the Raspi as the hub in our sensor network to query the sensor
nodes, and store and present the data, but we can even turn the Raspi into a node of a
larger control system such as EPICS. This is the topic of the next chapter.
146  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

QUESTIONS AND PROJECT IDEAS

1. Where do you find help about programs installed on the Raspi?
2. How do you copy files on the Raspi?
3. How do you rename files on the Raspi?

4. How do you copy files between the Raspi and your desktop computer?
5. Find out what the program ping does.
6. Find out what the program touch does.
7. Describe the installation of the program nmap on the Raspi. Investigate what it does
and how it may help you in debugging your network.
8. What is the netmask?
9. Start Mathematica from the menu and play around with it. Find out what it can do.
10. Start the synaptic software installation program from Section 5.3, enter “games” as
the search keyword and explore.
11. Start python, execute help(), and follow the instructions.
12. Write a program that displays “Hello World” in Python.
13. Create a file with the name look at me in the home directory of user pi, and instruct
the cron daemon to touch it every second Thursday of the month.
14. How do you get help about commands in octave?
15. Connect a USB web cam to the Raspi and make pictures with the cheese program.
Use the imagemagick tools to cut a small portion from the image (crop) and convert
it to another format, say gif.
16. Find out how to store float values in a MySQL database.
17. Find out how to store images in a MySQL database.
18. Measure the brightness with an LDR attached to a NodeMCU and log it once per
hour in a MySQL database. If you are patient you can see see days getting longer and
shorter as the season progress.
19. Log the steps that a stepper motor takes and visualize them on a web page with
RRDtool.
20. Make a web page that shows your Arduino sketches.

21. Find out how to specify colors in HTML.

22. Use HTML forms in a web page on the Raspi to set the brightness of a LED on the
NodeMCU.
23. Prepare a web page about yourself with picture and all you care to display.
 CHAPTER 6

Control System: EPICS

The Experimental Physics and Instrumental Control System (EPICS) is used in a number
of laboratories to control large particle accelerators, such as the Advanced Photon Source
(APS) in the United States, the Swiss light source (SLS) in Zürich, or the European Spal-
lation Source (ESS) in Sweden. Other users of EPICS are the fusion reactor International
Thermonuclear Experimental Reactor (ITER) in France and the W. M. Keck astronomical
observatory on Hawaii.
EPICS is based on a number of independent computers called input-output controllers
(IOC) that announce their capabilities on the network such that other computers can in-
teract with them. Almost any type of computer can participate in an EPICS system, and
support libraries for many programming languages, such as C, Python, and MATLAB, are
available. So it is no surprise that a Raspi can also serve as an IOC and join an EPICS
control system, no matter how big it is. We illustrate this by configuring the Raspi to
communicate the measurements collected from locally attached microcontroller-based sen-
sor nodes to EPICS. This makes the measurements accessible in a larger control system
context.
The first task is to install the EPICS software on the Raspi, and we follow [28] in doing
so. The steps are somewhat arcane and the instructions resemble a cookbook. Normally
these steps are done by an experienced system administrator.

6.1 INSTALLATION
First we need to download the EPICS Base package from www.aps.anl.gov/epics. At
the time of writing, the current release is R3.15.5 and the downloaded package is called
base-3.15.5.tar.gz. In order to avoid excessive use of sudo we create a subdirectory
/home/pi/epics where all the software resides, but we also create a soft link of that di-
rectory to /usr/local/epics where it is available to all users on the Raspi. The detailed
sequence of commands is the following:
cd /home/pi
mkdir epics
sudo ln -s /home/pi/epics /usr/local
cd epics
cp /home/pi/Downloads/base-3.15.5.tar.gz .
tar xzf base-3.15.5.tar.gz
where tar is an archiving program that unpacks (x option) compressed (z option) files
(f option). Note the period ‘.’ at the end of the cp command, which is the shorthand

147
148  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

notation for the current directory, /home/pi/epics in this case. The last command creates a
subdirectory under /home/pi/epics with the name base-3.15.5. In order to avoid writing
the version number over and over again, we create a soft link with the ln -s command,
which essentially creates an alias in the same directory by executing
ln -s base-3.15.5 base
and we can henceforth refer to the directory with the EPICS base package via
/usr/local/epics/base by virtue of the soft link from /home/pi/epics to /usr/local/
epics. Now we have the source code in place, but in order to use it we still need to compile
it.
And for the compiler and later the executable programs to find the EPICS sources, we
need to copy the following lines into the file /home/pi/.bash aliases:
export EPICS_ROOT=/usr/local/epics
export EPICS_BASE=${EPICS_ROOT}/base
export EPICS_HOST_ARCH=‘${EPICS_BASE}/startup/EpicsHostArch‘
export EPICS_BASE_BIN=${EPICS_BASE}/bin/${EPICS_HOST_ARCH}
export EPICS_BASE_LIB=${EPICS_BASE}/lib/${EPICS_HOST_ARCH}
if [ "" = "${LD_LIBRARY_PATH}" ]; then
export LD_LIBRARY_PATH=${EPICS_BASE_LIB}
else
export LD_LIBRARY_PATH=${EPICS_BASE_LIB}:${LD_LIBRARY_PATH}
fi
export PATH=${PATH}:${EPICS_BASE_BIN}
If the file does not exist, just create a new copy with the above contents. Note that the
environment variable EPICS ROOT points to the subdirectory /usr/local/epics, the one
we created in an earlier step. The other variables are defined relative to that and point to
the places where the libraries and executables reside. Once we complete this step, we need
to open a new command window, because this rereads the .bash alias file and we need
these definitions before compiling the base system with
cd /home/pi/epics/base
make -j 4

which takes about 20 minutes depending on the version of Raspi. The option -j 4 causes
the compilation to start four jobs simultaneously, one for each CPU core. The terminal
window is filled with the information about what part of the base system is currently
compiled. Once the compilation completes without errors, we enter the command caget at
the command prompt. It should respond with “no pv name specified...” and this indicates
that the executables are available and located in the PATH. The “pv” is the name of a
quantity that EPICS deals with, and is called a process variable. An example of such a
name is the read-back current of power supply PSabc that might be called PSabc:current.
Reading or changing the power supply current then refers to that process variable. Reading
is done by entering caget PSabc:current on the command line. More on that topic comes
in the next section. For the next test we start the softIoc program, and at the epics>
prompt that appears, we type iocInit. This step should result in the response “iocrun: all
initialization complete” and convinces us that we have successfully installed EPICS on the
Raspi and can proceed to investigate it.
 Control System: EPICS  149

6.2 COMMUNICATING WITH EPICS

Before connecting the sensor nodes with the microcontrollers to EPICS, we explain the
basic EPICS functionality by reading and controlling virtual parameters that only exist in
the memory of the Raspi. Again, based on examples from [28], we prepare the following
EPICS database file and name it simple2.db.
# simple2.db
record(bo,"raspi:trigger") {
field(DESC,"trigger PV")
field(ZNAM,"off")
field(ONAM,"on")
}
record(stringout,"raspi:message") {
field(DESC,"message on the RPi")
field(VAL,"RPi default message")
}
record(calc, "raspi:random") {
field(SCAN,"1 second")
field(CALC,"RNDM")
}
record(ao,"raspi:A") {
field(DESC,"variable A")
field(VAL,"0")
}
record(ao,"raspi:B") {
field(DESC,"variable B")
field(VAL,"0")
}
record(calc,"raspi:C") {
field(DESC,"sum of A and B")
field(SCAN,"1 second")
field(INPA,"raspi:A")
field(INPB,"raspi:B")
field(CALC,"A+B")
}
The file contains definitions of six process variables: raspi:trigger, raspi:message,
raspi:random, raspi:A, raspi:B, and raspi:C in the record definitions. Each record
has a type, which might be bo, bi, ao, or ai, standing for binary or analog input or
output. Other types are stringout or calc. Let’s describe the records one at a time. The
first record is a binary output, bo, called raspi:trigger, which may be either 1 or 0 and
contains three defining fields, a description, a text string for state zero (ZNAM), and one for
state one (ONAM). The second record describes a string that is initialized in the VAL field.
The third record is of type calc and performs some calculation, in this case a rather trivial
one: It generates a new random number once per second as specified in the SCAN field. The
next two records define process variables raspi:A and raspi:B that are initialized to zero.
The sixth record is of type calc and calculates the sum of raspi:A and raspi:B once per
second. Once the file simple2.db is saved, we start it with the following command:
softIoc -d simple2.db
150  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 6.1 Using the EPICS commands caget and caput.

which will start an EPICS server and publishes our six process variables. The latter we can
verify by typing dbl (database list) at the epics> prompt, which lists the process variables
served from the running softIoc process.
We interact with the EPICS server from a second terminal window and enter caget
raspi:trigger in it. The response is the name of the process variable and the current
state, which initially is off. We change the state by entering caput raspi:trigger on
and subsequently verify with caget that the state has indeed changed. Note that we change
process variables with caput and retrieve their value with caget. We immediately try
this with the raspi:message process variable by entering caget raspi:message, which
displays the message from the simple2.db file. We change it with caput raspi:message
Blabla, and the next caget raspi:message should display Blabla instead. Figure 6.1
shows these transactions. When reading a process variable with caget, we can suppress the
echo of the variable name by using the -t (for terse) option and use caget -t instead. A
companion program to caget is camonitor, which monitors one or several process variables
and reports whenever one of them changes its value. We use it immediately to verify that the
raspi:random process variable indeed produces a new random number once every second,
by entering camonitor raspi:random. The last three records define process variables that
are linked in such a way that the third raspi:C calculates the sum of the other two process
variables. We test this functionality by using caput to enter the values 17 and 4 to raspi:A
and raspi:B, respectively. Subsequently reading raspi:C will report 21. In passing we
mention that a second computer on the same network, either a second Raspi or a desktop
computer that has the EPICS base package installed, can access all process variables served
by the first Raspi.
This dry run of EPICS worked without interfacing hardware. In order to communicate
with our sensor nodes, we need two additional libraries to link EPICS to the hardware.

6.3 ASYN AND STREAM LIBRARIES

The external sensor nodes communicate with the Raspi via serial RS-232 lines, Bluetooth,
Ethernet, or WLAN. EPICS requires interface libraries to be able to take over these
 Control System: EPICS  151

communication channels. For this purpose we install two additional packages, asyn and
streamdevice. We start by installing the asyn package, which is is available from https:
//www.aps.anl.gov/epics/download/modules/ as asyn4-30.tar.gz (version 4.30 was
current at time of writing). We then create a subdirectory /home/pi/epics/modules/, copy
the downloaded file to it, and unpack the file with the following sequence of commands.
cd /home/pi/epics
mkdir modules
cp /home/pi/Downloads/asyn4.30.tar.gz modules
cd modules
tar xzf asyn4.30.tar.gz
ln -s asyn4.30 asyn
Before compiling the package, we need to edit the ./asyn/configure/RELEASE file, change
the EPICS BASE variable to /usr/local/epics/base, and comment out the lines starting
with IPAC and SNCSEQ. Once this is done we compile by entering make -j 4 in the directory
/usr/local/epics/modules/asyn and wait a few minutes for completion.
The second package we install is the streamdevice package. For this pur-
pose we create a new subdirectory, /home/pi/epics/modules/stream, download the
file StreamDevice-master.zip from https://github.com/paulscherrerinstitute/
StreamDevice and unpack the zip file inside the newly created subdirectory
cd /home/pi/epics/modules/stream
unzip StreamDevice-master.zip
which creates a subdirectory StreamDevice-master. To make the build process compatible
with using a plain make system, we need to remove the GNUMakefile from that subdirectory.
Once that is done, we execute the following command in /home/pi/epics/modules/stream
makeBaseApp.pl -t support
and edit /home/pi/epics/modules/stream/configure/RELEASE. We make sure that
EPICS BASE points to /usr/local/epics/base, add the following lines to the bottom of
the file
ASYN=/usr/local/epics/modules/asyn
and finally initiate the compilation. We run make once in /home/pi/epics/modules/stream
and a second time in the StreamDevice-master subdirectory. This completes the prepara-
tion of the basic libraries, and we are ready to write IOCs that talk to our hardware, the
microcontrollers with the sensors and actuators attached.

6.4 WRITING AN IOC

As examples, we link the temperature measurement on the UNO, connected by serial line,
and on the NodeMCU, connected by WLAN, to EPICS, and publish the measurements as
process variables. For this purpose, we create a subdirectory /home/pi/epics/ioc and a
subdirectory temp in it. Inside that subdirectory we execute the following commands:
makeBaseApp.pl -t ioc temp
makeBaseApp.pl -i -t ioc temp
which creates the basic file infrastructure under /home/pi/epics/ioc/temp. We change to
that directory with the cd command and add the two lines
152  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

ASYN=/usr/local/epics/modules/asyn
STREAM=/usr/local/epics/modules/stream
to the end of the file ./configure/RELEASE.
Next, we prepare the protocol file that describes the communication protocol we used
earlier: send ‘T?’ and receive ‘T 21.2’. The file resides in the subdirectory ./tempApp/Db
under the temp base directory for this IOC. Inside it we create the following file named
temperature.proto
# ./tempApp/Db/temperature.proto
Terminator = CR LF;
get_temp {
out "T?";
in "T %f";
ExtraInput = Ignore;
}
The content is rather straightforward to understand. First we define the terminating char-
acters that denote the end of a line, and a function get temp that sends the string T? to
the device and expects T and a float (%f) number in return. Moreover, any extra characters
should be ignored. It is possible to have more than one function defined in the same protocol
file. The protocol file is the lowest level to define the communication; the next higher level
is the database file, which we already encountered in Section 6.2. Here we use the following
file
# ./tempApp/Db/temperature.db
record(ai, "$(USER):temp") {
field(DESC, "Temperature")
field(SCAN, "10 second")
field(DTYP, "stream")
field(INP, "@temperature.proto get_temp $(PORT)")
}
This file defines an analog input ai record for a process variable with the name
$(USER):temp. Here $(USER) will be defined in the calling program. The description and
update rate of 10 seconds are defined in the first two fields. The third field declares the
record to be of type stream, and as input function (INP) we use the function get temp from
the protocol file temperature.proto. We use the communication interface with the name
supplied in the variable $(PORT). We then have to edit ./tempApp/Db/Makefile and add
the line
DB += temperature.db
to it. Next we need to edit ./tempApp/src/Makefile and add the lines
temp_DBD += asyn.dbd
temp_DBD += stream.dbd
temp_DBD += drvAsynSerialPort.dbd
temp_DBD += drvAsynIPPort.dbd
after the line with temp DBD += base.dbd, which instructs the build process to include
the asyn and stream libraries as well as support for serial communication and Internet
protocol ports such as network sockets. Then, near the bottom, following temp LIBS +=
$(EPICS BASE IOC LIBS), add the lines
 Control System: EPICS  153

temp_LIBS += asyn
temp_LIBS += stream
which are needed to link against the two libraries.
In a last step, we define the startup program st.cmd for the IOC that is located in the
./iocBoot/ioctemp/ directory. In this file we add the following line
epicsEnvSet(STREAM_PROTOCOL_PATH,"../../tempApp/Db")
after the line with < envPaths. The added line describes where the protocol files are lo-
cated. Following the line with temp register RecordDeviceDriver pdbbase we add the
definition of the serial port we intend to use. For the serial line to the UNO, this looks like
drvAsynSerialPortConfigure("SERIALPORT","/dev/ttyACM0",0,0,0)
asynSetOption("SERIALPORT",-1,"baud","9600")
asynSetOption("SERIALPORT",-1,"bits","8")
asynSetOption("SERIALPORT",-1,"parity","none")
asynSetOption("SERIALPORT",-1,"stop","1")
asynSetOption("SERIALPORT",-1,"clocal","Y")
asynSetOption("SERIALPORT",-1,"crtscts","N")
and can be referred to by its symbolic name SERIALPORT. Finally, we need to load the
database records using the definition of the variables $(PORT) and $(USER)
dbLoadRecords("db/temperature.db","PORT=’SERIALPORT’,USER=’raspi’")
where SERIALPORT replaces the place holder $(PORT) in the database record file
temperature.db. Moreover, the name of the process variable is prepended by raspi, such
that we can later access the temperature with the command caget raspi:temp. Now the
software setup for the IOC is complete, and we compile it by running make in the directory
/home/epics/ioc/temp/. Once the compilation successfully completes, we make the file
st.cmd that is located in ./iocBoot/ioctemp/ executable by executing chmod +x st.cmd
and run it with
./iocBoot/ioctemp/st.cmd
This starts the EPICS server and the process variable, here only raspi:temp, is published
on the local network so that any computer with the EPICS base system installed and a
working caget program can read the temperature from our Raspi.
It remains to connect the NodeMCU microcontroller to EPICS. Since the NodeMCU
server from Section 4.6.3 listens on port 1137 at IP number 192.168.20.135 and uses the
same protocol (send ‘T?’, receive ‘T 22.1’), we just add the following two lines to the st.cmd
file:
drvAsynIPPortConfigure("SOCKET1","192.168.20.135:1137",0,0,0)
dbLoadRecords("db/temperature.db","PORT=’SOCKET1’,USER=’node’")
The first line defines a symbol SOCKET1 that points to the port on the NodeMCU, and
the second line instructs EPICS to use the same database file temperature.db as before,
and link the communication to the PORT corresponding to the one defined in the previous
line. Once we add the two lines, we need to compile the project again. This we do by
issuing make in the directory /home/epics/ioc/temp/ and restarting st.cmd by executing
./iocBoot/ioctemp/st.cmd from the same directory. The result of this exercise is that
EPICS now publishes two process variables, raspi:temp from before and node:temp from
154  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

the NodeMCU connected by WLAN, which we can verify by entering the command dbl at
the epics> prompt that st.cmd provides.
Most of the above actions we only need to do once, and adding additional sensors to the
EPICS IOC is only moderately complex. It only requires writing the appropriate protocol
file, say new.proto and database file new.db, and copying both to the ./tempApp/Db direc-
tory. Then we need to add the database file to the project by adding DB += new.db to the
Makefile in the same directory. Finally, we need to add drvAsynXConfigure to the st.cmd
file and execute dbLoadRecords to link the database file to the appropriate PORT. Finally,
we compile once again and and run the st.cmd command.

6.5 STARTING THE IOC AT BOOT TIME

In the description above we need to start the IOC st.cmd by hand and keep a terminal
window with the running program open all the time. To remedy this inconvenience, we
create a startup script named epicsioc that launches the IOC when the system boots. The
file contains the following lines:
#!/bin/sh
#/etc/init.d/epicsioc
### BEGIN INIT INFO
# Provides: empicsioc
# Required-Start: $remote_fs $syslog
# Required-Stop: $remote_fs $syslog
# Default-Start: 2 3 4 5
# Default-Stop: 0 1 6
# Short-Description: Simple script to start a program at boot
# Description: Start and stop epicsioc server
### END INIT INFO
case "$1" in
start)
echo "Starting epicsioc"
. /home/pi/.bash_aliases
cd /home/pi/epics/ioc/temp/iocBoot/ioctemp
/usr/bin/procServ -n "IOC" -L /tmp/epics.log -i ^D^C 20000 ./st.cmd
;;
stop)
echo "Stopping epicsioc"
kill $(pidof procServ)
;;
restart)
$0 stop
sleep 10
$0 start
;;
*)
echo "Usage: /etc/init.d/epicsioc {start|stop|restart}"
exit 1
;;
esac
exit 0
 Control System: EPICS  155

The file contains some INIT INFO that is required by the init program, which orchestrates
the startup of the Raspbian operating system. In the subsequent case structure, three ac-
tions are specified: start, stop, and restart. The start option reads the default EPICS
environment variables and then changes to the subdirectory with the st.cmd program.
There it executes the procServ program that in turn starts the st.cmd program. The pur-
pose of procServ is to redirect the input and output of st.cmd to the network socket 20000
to which we can connect via telnet, and interact with the program in the same way we
did earlier when we started it in a terminal window. The advantage of using procServ is
that we can close the telnet program while st.cmd keeps running independently. The other
options give the running process the name IOC, write a log file, and ignore the control se-
quences to close the process. The stop case determines the process id of procServ with
pidof and terminates the process with the kill command. The restart case executes stop
first, waits 10 seconds, and then executes start. The default case, denoted by *), displays
a brief usage note.
Before actually running the program, we need to install the procserv and telnet pack-
ages with the normal installation procedure:
sudo apt-get install procserv telnet
In case one of the programs is already present in the current system, no new software
is installed. Once all required programs are installed and the script is written, we copy it,
using sudo, to the system directory /etc/init.d, where all startup scripts for the operating
system reside. We make it executable using
sudo chmod 755 /etc/init.d/epicsioc
and start it by hand in order to verify that the script works as intended with the command

sudo /etc/init.d/epicsioc start

and if everything works we can install it permanently by executing
sudo update-rc.d epicsioc defaults
which will automatically start epicsioc at boot time. If we want to connect to the now-
running-autonomously process st.cmd, we use telnet when logged onto the Raspi
telnet localhost 20000
and see the normal output of the st.cmd program, plus some administrative info from
procServ prepended by @@@. We can enter commands such as dbl to list all process vari-
ables, and immediately receive the output back on the telnet window. The port opened by
procServ is only accessible from the Raspi itself. If we want to log from another computer,
we use ssh to first log onto the Raspi, and then execute telnet. This can be automated by
executing
ssh -t [email protected] telnet localhost 20000

from a remote desktop computer and where 192.168.1.22 is the IP-number of the Raspi.
In this way, the Raspi works unattended as an Epics IOC but we can connect to it at any
time using telnet in order to follow the performance and interact with st.cmd.
156  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

QUESTIONS AND PROJECT IDEAS

1. Why is a standardized control system advantageous to use in comparison to a home-
grown system?
2. Write protocol and database files to interface the sensors discussed in previous chap-
ters.

3. Write protocol and database files to connect an Arduino UNO running the sketch in
Section 4.5.2 to control a DC motor.
4. Write protocol and database files to connect an Arduino UNO running the sketch in
Section 4.5.3 to control a model-servo.

5. Write protocol and database files to connect an Arduino UNO running the sketch in
Section 4.5.4 to control a stepper motor.
6. Research other control systems used in industry or research institutions.
7. Research the documentation of the stream library on how to treat information that
appears on the RS-232 line unsolicited, such as a continuously appearing stream of
position data from a GPS receiver.
8. In this chapter we address the Raspi by IP number 192.168.1.22 and in Section 5.4
by 192.168.20.1. Explain, why this not a typo.
 CHAPTER 7

Messaging System: MQTT

EPICS is not the only system to integrate a large number of sensors and actuators un-
der a common interface. Another such system is the message queue telemetry transport
(MQTT) [6] protocol that was originally created by, among others, IBM, in order to collect
information from widely distributed infrastructures such as oil pipelines in a battery-saving
and energy-efficient as well as robust and secure way. Today it is a protocol commonly used
to pass messages between devices that constitute the Internet of Things (IoT) and is now
ISO-standard ISO/IEC 20922. MQTT is based on a publish/subscribe method between
clients with a message-passing broker in the middle. The broker acts as the hub passing
messages from one client that publishes data to another that subscribes to published data.
The name of the parameters that are passed around are referred to as “topic.” They are
organized hierarchically, such that the name
weatherstations/stationA/node1/temperature
refers to the temperature sensor on a weather station, called station A and connected to
node 1. When subscribing to topics it is possible to use wildcards such as the plus-sign “+”,
which is a single-level wild-card, or the hash “#,” which is a multi-level wild-card. Robustness
and reliability of transmission is guaranteed by specifying three levels of quality of service
(QoS), where the simplest level 0 implies that the sensor only publishes data without any
acknowledgment from the broker. Levels 1 and 2 implement increasingly advanced levels
of handshake signals that ensure the arrival of data. Even a last will is available, which
is transmitted to subscribing clients should a publishing client disconnect. Normally the
broker immediately transmits any received data to the subscribing clients, but it is possible
to specify that the broker retains the last good data point and transmits it to subscribers
in case the publisher is offline, and to new subscribers upon their first connection. Since
MQTT was intended to operate across public networks, encryption and security are part of
the protocol.
The core functionality of MQTT resides on the broker, such that clients can be very
simple and connect and disconnect at will. This makes it possible to use clients that only
wake up from a battery-saving deep sleep mode, perform a measurement, send it to the
broker, and go back to sleep again. This attractive feature makes MQTT very popular for
IoT applications and we therefore also discuss it as a complement to EPICS. In the following
sections we operate a broker on the Raspberry Pi and use NodeMCUs as clients that publish
and subscribe. In a final section we discuss a simple gateway that links MQTT to EPICS in
order to benefit from the best of both worlds. In this way we can access lightweight clients
connected over a public network across the gateway from our EPICS control system.

157
158  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

After this overview of MQTT, let us follow the theme of the book and describe a working
system that provides the basic functionality.

7.1 BROKER
The mosquitto broker is widely supported on many platforms, and also on the Raspberry
Pi. We install it, after bringing the system up to date with sudo apt-get update and sudo
apt-get upgrade, using the command
sudo apt-get install mosquitto mosquitto-clients
where the mosquitto package contains the broker and the mosquitto-clients package
contains the command line executables mosquitto pub and mosquitto sub. We use them
to test the base functionality on the Raspi alone, without external clients.
Immediately after the installation, mosquitto is already running and is registered as
a service that starts after every reboot. During our initial tests we inspect the logging
output with the following command, which helps to debug the problem in case something
unexpected happens:
tail -f /var/log/mosquitto/mosquitto.log

where tail -f takes the filename as argument and shows the last lines of the file, but,
instead of stopping the display at the end of the file, keeps appending newly generated
messages. Now we open a second terminal window and start publishing MQTT data using
the mosquitto pub command-line client with
mosquitto_pub -d -h localhost -i Pub1 -t dummy/value -m 42 -r

where mosquitto pub is the executable, -d enables debugging output, and -h specifies the
IP address of the broker; here it is localhost, because the broker runs on the same Raspi
where we run the mosquitto pub client. The parameter specified after -i identifies the
publisher, -t the topic, and -m specifies the message, here the value 42. Appending -r
instructs the broker to retain the message and send it in case the publisher is offline. The
debugging output can be disabled by omitting the -d option. Running mosquitto pub -h
gives a short overview of available commands, and the manual page, accessible with man
mosquitto pub, provides more information.
The mosquitto sub executable provides the receiving end of MQTT message-passing.
We run it in another terminal window where we execute
mosquitto_sub -d -h localhost -i Sub1 -t dummy/value
to subscribe to the topic dummy/value on the broker running on localhost. We identify the
subscribing client by Sub1 and enable debugging with -d. After starting the executable, we
are greeted with either an empty line or with the last published message that has the retain
flag -r enabled. The mosquitto sub keeps running, and if we publish new messages in the
second terminal with mosquitto pub, they immediately appear in the subscriber window.
At this point we have a working broker whose functionality we verified with the
command-line programs mosquitto pub and mosquitto sub running on the same Raspi
as the broker. In the next section we replace these command-line clients by external clients
running on NodeMCUs that publish temperature data and at the same time subscribe to
another topic to turn a cooling fan on or off.
 Messaging System: MQTT  159

7.2 NODEMCU CLIENTS

The hardware connected to the NodeMCU is very simple in this case. We connect ground
and supply pins of an LM35 temperature-sensor to the respective pins on the NodeMCU.
The signal pin of the LM35 is connected to the single analog input A0. Second, we connect
a transistor to an output pin as shown in previous chapters to turn the fan on and off. In
order to use MQTT functionality in the Arduino IDE, we need to install one of the many
available libraries. We select the “MQTT PubSub” library and install it from within the IDE
by going to the library manager located at Sketch→Include library→Manage Libraries, enter
“MQTT PubSub” in the search field, and install the suggestion that shows up. The following
sketches are based on the mqtt esp8266 example that is included in the installation.
It turns out that adapting the example code to write a sketch that serves as both
publisher and subscriber is straightforward, and the following code fulfills that purpose.
// MQTT client, V. Ziemann, 170816
const char* ssid = "messnetz";
const char* password = "..........";
const char* broker = "192.168.20.1";
const int fan_pin=D4;
#include <Ticker.h>
volatile uint8_t do_something=0;
Ticker tick;
void tick_action() {do_something=1;} // executed regularly
#include <ESP8266WiFi.h>
#include <PubSubClient.h>
WiFiClient espClient;
PubSubClient client(espClient);
void on_message(char* topic, byte* msg, unsigned int length) {
char ch[30]; memcpy(ch,msg,length); ch[length]=’\0’;
if (strstr(topic,"node1/fan")==topic) {
int val=(int)atof(ch);
Serial.print(" Fan="); Serial.println(val);
if (val==0) {
digitalWrite(fan_pin,HIGH);
} else {
digitalWrite(fan_pin,LOW);
}
} else if (strstr(topic,"node2/temp")==topic) {
Serial.print("Temperature on node2 = "); Serial.println(ch);
}
}
void setup() { //.................................setup
pinMode(fan_pin,OUTPUT);
Serial.begin(115200);
WiFi.begin(ssid,password);
while (WiFi.status() != WL_CONNECTED) {
delay(500); Serial.print(".");
}
Serial.print("\nWifi connected to "); Serial.println(ssid);
Serial.print("with IP address: "); Serial.println(WiFi.localIP());
client.setServer(broker, 1883); // 1883 = default MQTT port
160  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

client.setCallback(on_message); // execute when a message arrives

tick.attach(5,tick_action); // execute tick_action every 5 seconds
}
void loop() { //..................................loop
while (!client.connected()) {
if (client.connect("PubSub1")) { // identification
client.subscribe("node1/fan"); // external fan control
client.subscribe("node2/temp"); // temp on other nodemcu
} else {
delay(5000);
}
}
client.loop();
if (do_something) {
do_something=0;
char message[30];
int temperature=(int)(3.3*100*analogRead(A0)/1023.0);
sprintf(message,"%d",temperature);
client.publish("node1/temp",message);
}
}

In the first few lines of the sketch we define the usual network credentials and the IP num-
ber of the broker as well as the pin to which the fan is connected. Then we include the
header for the Ticker library, which provides a timer that is executed at regular intervals.
After declaring the variable do something, we declare the tick object and define the func-
tion tick action() to be executed in regular intervals. All this function does is to set the
variable do something to one. We declare do something volatile, because it changes asyn-
chronously from the main loop. Next we include the WiFi header, the MQTT PubSubClient
functionality, and declare both WiFiClient and a PubSubClient named client. The fol-
lowing function named on message() is executed every time a MQTT message arrives. Its
arguments are the topic, the message, and the length of the message. Within the function we
first convert the received message to a character string, because we want to handle it using
the same mechanism we used in previous chapters. Then we enter the usual construction
where we check which topic has arrived; if it is node1/fan we convert the message to an
integer value val and turn a pin on or off, depending on whether val is zero or not. If the
received topic is node2/temp we only display it on the serial line, but could easily add a
test to turn the fan on or off, depending on the temperature received.
Once variables and auxiliary functions are declared, we define the setup() func-
tion and configure the mode of the used pins, the serial line, and WiFi. The function
client.setServer() connects to the broker on the default MQTT port 1883, and the call
to the client.setCallback() function registers the function on message() to be executed
when a new message arrives. Finally, we start the ticker process to execute the function
tick action() once every 5 seconds. In the loop() function we first ensure that the con-
nection to the broker is up and running. If it is not running we connect to the broker with
the call to the client.connect() function and provide the identification of the NodeMCU
as PubSub1. In the above example using mosquitto sub, this corresponds to the parame-
ter following the -i command line switch. The subsequent calls to client.subscribe()
register the strings used as argument with the broker, which subsequently sends any up-
dated values. If the connection with the broker fails, a new attempt is made after 5 seconds.
 Messaging System: MQTT  161

Once the connection to the broker is established, we call the client.loop() function to
service any background activities related to MQTT, and finally check whether the variable
do something was set, which happens every time the ticker fires. If do something was set,
we reset it to zero and publish the parameter node1/temp.
In a terminal window on the Raspi, we can start mosquitto sub to subscribe to the topic
node1/temp and should see the updated temperature every 5 seconds. Furthermore, in order
to test multiple NodeMCU clients to communicate, we can program a second NodeMCU
with the same sketch, but swap the reference to node1 and node2. Moreover, if we only
want to publish values from a NodeMCU client, all code related to receiving messages
can be removed from the sketch, such as the on message() function and the two calls to
client.subscribe(). If we want to use a client with subscription-only functionality, we
can remove everything related to the tick function and the variable do something. In any
case, using the above sketch as a base should make it possible to serve almost any need to
connect NodeMCUs to a MQTT network.
So far, the MQTT system is unrelated to other control systems such as EPICS. So, if we
require interoperability, we need to provide a gateway that translates the message formats,
and that is what we describe in the next section.

7.3 GATEWAY TO EPICS

We want the gateway to seamlessly translate between the systems and to achieve the follow-
ing functionality: The gateway will listen on a network socket on port 51883 to communicate
with EPICS, and uses default ports for anything else. On EPICS we intend to use stream-
based protocol files such that all strings that EPICS sends to our gateway will have the
format topic value, and we configure the gateway to publish this as topic topic with the
message contents value. This is all we need to do to publish MQTT topics from EPICS. In
order to receive MQTT messages on EPICS, we need to configure the gateway to subscribe
to the messages and pass any incoming messages on to EPICS. For this we decide that any
message from EPICS to the gateway having the format SUBSCRIBE topic will instruct the
gateway to subscribe to topic. We also implement a command to UNSUBSCRIBE as a matter
of order.
We chose to implement the gateway using the Python language, because it has powerful
support for both conventional network sockets and for MQTT. The former is included in
standard installations, and for MQTT we use the paho.mqtt library that we install from
the command line using the Python installer program pip with command
sudo pip install paho-mqtt
and after installing it we are ready to write the gateway. The powerful libraries make the
following Python-program rather compact.
# Epics to MQTT gateway, V. Ziemann, 170817
import socket,atexit,paho.mqtt.client
def cleanup(): #...........................cleanup
sock.close()
atexit.register(cleanup)

def on_message(c,u,msg): #..............on_message

print "msg = ",msg.topic," ",msg.payload," ",msg.qos
epics.send(msg.topic + " " + msg.payload + "\r\n")
mqttc=paho.mqtt.client.Client()
162  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

mqttc.connect("localhost",1883)
mqttc.on_message=on_message
mqttc.loop_start()

sock=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
sock.bind((’’,51883))
sock.listen(1)
while 1:
epics,address = sock.accept()
print "Connected from ",address
while 1:
msg=epics.recv(1024)
if not msg: break
words=msg.split()
if len(words)<2: break;
if words[0].upper()=="SUBSCRIBE":
mqttc.subscribe(words[1],1)
elif words[0].upper()=="UNSUBSCRIBE":
mqttc.unsubscribe(words[1])
else:
mqttc.publish(words[0],words[1])
pass
epics.close()
print "Disconnect from ",address
At the start of the program, we import support for sockets, MQTT and atexit. The latter
is used to execute code asynchronously when terminating the program, by registering the
cleanup() function to close the network socket. Next we define the function on message,
which is called when a subscribed MQTT message arrives. After printing the message, which
is useful for debugging, we send it via the network socket named epics to EPICS. Note that
we append both a carriage return \r and a line feed \n in order to match the convention
for line terminators used in EPICS. Now we are ready to define the MQTT client mqttc,
and connect it to the broker that runs on the same computer as the gateway, namely
on localhost, and to the standard MQTT port 1883. On the following line we register
the on message function to be executed at every on message event of mqttc, and start
the MQTT event-loop with the mqttc.loop start() function. Note that the sequence of
connecting to the broker, registering a call-back function for arriving messages, and starting
the event loop mimics the code in the sketch that runs on the NodeMCU. After configuring
the MQTT connections, we define a socket sock that uses the IPv4 (AF INET) protocol and
TCP (SOCK STREAM). We bind the socket to port 51883 where it listens for requests from
EPICS. We instruct the socket to only accept one connection at a time and enter into an
infinite loop in which the socket waits for an incoming request from EPICS in the function
sock.accept(). This function blocks until a request arrives, when it returns a handle epics
to the new connection and the IP number address of the connecting computer. Once the
connection is established, the gateway enters a second loop and waits for a command on
the epics socket. If a message with an invalid format arrives, the connection closes and
otherwise splits the received message msg into words, and then branches depending on
the first word. If, after conversion to uppercase, it is SUBSCRIBE, the gateway calls the
mqttc.subscribe() function with the topic as argument. If it is UNSUBSCRIBE it calls the
mqttc.unsubscribe() function. In all other cases the two words are interpreted as topic
 Messaging System: MQTT  163

and value, and published using the mqttc.publish() function. If messages with incorrect
format are received or the calling EPICS computer disconnects, the epics socket closes
and prints a message. At this point the outer while 1: loop is still active and the gateway
reverts to the sock.accept() function and waits for new connections.
We start the gateway by executing
python epics2mqtt.py
from the command line, and writing a boot script similar to the one for EPICS from
Section 6.5 is left as an exercise. The basic functionality of the gateway can be easily tested
using netcat to emulate EPICS and send strings to the gateway. If the terminal window
with the mosquitto sub command from the end of Section 7.1 is still running, we can send
messages to it with the command
echo "dummy/value 57" | netcat -C localhost 51883
which sends the string following echo to socket 51883 on the local computer, but this is
where the gateway listens, and publishes this message on MQTT on our behalf.
If we connect to the gateway with the command netcat -C localhost 51883 executed
from the command line, we can issue the string SUBSCRIBE node1/temp to subscribe to the
topic node1/temp. Executing the following command in another terminal window
mosquitto_pub -d -h localhost -i Pub1 -t node1/temp -m 23
will cause the string node1/temp 23 to appear in the window with netcat running. Issuing
the command UNSUBSCRIBE node1/temp in netcat stops the subscription, and messages
will no longer appear in the netcat window.
Finally, we prepare EPICS protocol and database files that implement the following
behavior. Executing caput node1/fan 1 publishes the topic node1/fan with message 1,
and caput SUBSCRIBE node1/temp subscribes to the topic node1/temp such that we asyn-
chronously receive the messages in EPICS. The first turns the fan on and the second reports
the temperature measured by the LM35 on the NodeMCU. The protocol file to implement
this behavior is stored in the following file, named mqtt.proto:
# ./tempApp/Db/mqtt.proto
Terminator = CR LF;
set_fan {out "node1/fan %i";}
subscribe {out "SUBSCRIBE %s";}
unsubscribe {out "UNSUBSCRIBE %s";}
get_temp {in "node1/temp %f";}
First we define the Terminator that matches the \r\n used in the gateway code, and then
define functions that either input or output values in the same way we used in the pre-
vious chapter. Note that the functions have the MQTT names hard coded as character
strings and that the functions to subscribe and unsubscribe have a string as argument. The
corresponding database file, called mqtt.db,
# ./tempApp/Db/mqtt.db
record(ao, "node1/fan") {
field(DESC, "Fan on node1")
field(DTYP, "stream")
field(OUT, "@mqtt.proto set_fan $(PORT)")
}
164  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

record(stringout, "SUBSCRIBE") {
field(DESC, "subscribe to topic")
field(DTYP, "stream")
field(OUT, "@mqtt.proto subscribe $(PORT)")
}
record(stringout, "UNSUBSCRIBE") {
field(DESC, "unsubscribe from topic")
field(DTYP, "stream")
field(OUT, "@mqtt.proto unsubscribe $(PORT)")
}
record(ai, "node1/temp") {
field(DESC, "Temperature on node1")
field(DTYP, "stream")
field(INP, "@mqtt.proto get_temp $(PORT)")
field(SCAN,"I/O Intr")
}
links the functions defined in the protocol file to process variables where the MQTT variables
have the same name in EPICS, and we add two process variables handling subscription.
Since they register a name of a variable, their EPICS record type is stringout. All variables
that we receive from MQTT via the gateway arrive asynchronously, at the rate they are
published, and we therefore need to use the line field(SCAN,"I/O Intr") in the database
file, which handles data that arrive without explicitely being requested. Note that we only
need to subscribe to values we want to read with caget. MQTT variables we want to set,
such as turning on the fan, need no subscription.
After having defined the protocol and database file, we need to add mqtt.db to the
Makefile in the ./temp/tempApp/Db directory and add the following definition for the PORT
in st.cmd in ./iocBoot/ioctemp/.
drvAsynIPPortConfigure("SOCKET","192.168.20.1:51883",0,0,0)
dbLoadRecords("db/mqtt.db","PORT=’SOCKET’,USER=’mqtt’")
The IP address points to the IP on which both broker and gateway run, and the port on
which the gateway listens. Once this is operational, we have access to MQTT topics from
EPICS.
We’ve come a long way, starting with small sensors that are connected to microcon-
trollers, which, in turn, communicate with a more powerful computer, a Raspi in our case,
to retrieve, store, and present the measurement values. And finally, we even connected the
Raspi to full-grown control systems where all measurement values are published and are
available to all computers on that network. Since there is a large number of EPICS pro-
grams available to perform online analysis, logging, and alarm management, among other
tasks, we can say that the individual measurement values from our sensors have trickled up
the data-handling chain, all the way to a generic top layer. And that was the main task we
initially set out to illustrate.
All examples we encountered so far were deliberately chosen to be rather simple, in
order to illustrate the mechanisms, but in the coming chapters we advance to more complex
projects, and start with a weather station with distributed sensors.
 Messaging System: MQTT  165

QUESTIONS AND PROJECT IDEAS

1. What is the purpose of the broker?
2. Find out about public brokers on the Internet.
3. Connect the humidity and barometric pressure sensor to a NodeMCU and have the
data published via MQTT.

4. Control the brightness of a LED connected to a NodeMCU via MQTT.

5. Connect a DC motor with H bridge to a NodeMCU and control it via MQTT.
6. Connect a stepper motor with H bridge to a NodeMCU and control it via MQTT.

7. Connect a model-servo to a NodeMCU and control it via MQTT.

8. Prepare a boot script for the gateway following the example in Section 6.5.
9. Write gateway between MQTT and a MySQL database such that the database is
automatically filled with newly arriving data samples.

10. Write gateway that fills an RRDtool database with data from MQTT.
11. Write a gateway to interface octave to MQTT such that a plot is continuously updated
with new data.
12. Find an MQTT client for your smartphone and use it to read the temperature from
the NodeMCU.
13. Discuss the similarities and differences between EPICS and MQTT. Under what cir-
cumstances do you prefer one or the other?
14. Write an MQTT client for the NodeMCU that mimics the query-response behavior:
It receives a query, does something, and publishes a response.
 CHAPTER 8

Example: Weather Station

with Distributed Sensors

Our first project is a weather station that measures barometric pressure, humidity, and
temperature at a number of locations inside and outside a building. We use rrdtool to
prepare plots of the measurements over periods of 4 hours, 2 days, a week, and a month.
Moreover, we prepare database files to enable integration into an EPICS control system.
We select the NodeMCUs as microcontrollers for the sensor nodes because they are very
easy to program, and flexible that they may be deployed all over the place without having
to pull wires. Note that we only show how to connect a single sensor node, but multiple
copies only differ by their IP number. We can duplicate any interfacing software by simply
changing the IP number appropriately.
As sensors, we choose the I2C-based sensors BMP180 for barometric pressure and
HYT221 for humidity, both of which also provide temperature information. Since the ana-
log input is not used, we may add an LM35 temperature sensor as well, even though it is
not necessary. We show such a sensor node built on a breadboard in Figure 8.1. We see
the NodeMCU on the right, which is connected to the power rails via its 3.3 V and ground
terminals. The three sensors are placed towards the left on the breadboard and their re-
spective pins are also connected to the power rails. The I2C pins for SDA and SCL for the
barometric and the humidity sensor are directly connected to the respective pins on the
NodeMCU microcontroller, which are D1 for SCL and D2 for SDA. The pins for positive
supply voltage and ground of the LM35 are connected to the lower power rail and its analog
output pin to the analog input pin A0 of the NodeMCU. We also added 4.7 µF and 100 nF
decoupling capacitors to the power rails.
The program running on the microcontroller follows the template we used earlier, and
is shown below.
/* Simple socket server to serve barometric pressure, humidity,
* and temperature. Author: V. Ziemann, 170117
*/
#include <SFE_BMP180.h>
#include <Wire.h>
SFE_BMP180 pressure;
double P,TP,H,TH;
const int HYT=0x28; // I2C address for HYT221
const char* ssid = "messnetz";

167
168  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

The weather-node circuit with the NodeMCU on the right and the LM35,
Figure 8.1
HYT221, and BMP180 sensors towards the left.

const char* password = ".........";

const int port = 1137;
#include <ESP8266WiFi.h>
WiFiServer server(port);
void setup() { //................................................setup
pinMode(LED_BUILTIN,OUTPUT);
digitalWrite(LED_BUILTIN,LOW);
Serial.begin(115200);
WiFi.begin(ssid, password);
while (WiFi.status() != WL_CONNECTED) {
delay(500);
Serial.print(".");
}
Serial.print("\nWiFi connected to "); Serial.println(ssid);
Serial.print("Server IP address: "); Serial.println(WiFi.localIP());
server.begin();
Serial.print("Server started on port "); Serial.println(port);
digitalWrite(LED_BUILTIN,HIGH);
if(!pressure.begin()) {Serial.println("Cannot init BPM180!");}
}
void loop() { //................................................loop
char line[30];
float volt,temp;
int t0;
WiFiClient client = server.available();
 Example: Weather Station with Distributed Sensors  169

while (client) {
digitalWrite(LED_BUILTIN,LOW);
while(!client.available()) {
delay(5);
if (!client.connected()) break;
}
client.readStringUntil(’\n’).toCharArray(line,30);
Serial.print("Request: "); Serial.println(line);
if (strstr(line,"TH?")==line) {
client.print("TH "); client.println(TH,2);
} else if (strstr(line,"H?")==line) {
int b1,b2,b3,b4,raw;
Wire.beginTransmission(HYT);
Wire.requestFrom(HYT,4);
delay(200);
if (Wire.available()==4) {
b1=Wire.read(); b2=Wire.read(); // humidity rawdata
b3=Wire.read(); b4=Wire.read(); // temperature rawdata
Wire.endTransmission();
}
raw=(256*b1+b2) & 0x3FFF; // humidity
H=100.0*raw/16384.0;
raw=((256*b3+b4) & 0xFFFC)/4; // temperature
TH=165.0*raw/16384.0-40.0;
client.print("H "); client.println(H,2);
} else if (strstr(line,"TP?")==line) {
client.print("TP "); client.println(TP,2);
} else if (strstr(line,"P?")==line) {
int is=pressure.startTemperature();
if (is!=0) {
delay(is);
is=pressure.getTemperature(TP);
if (is!=0) {
is=pressure.startPressure(3); // oversampling=3
delay(is);
is=pressure.getPressure(P,TP);
if (is!=0) {
Serial.print("BMP180: T="); Serial.print(TP,2);
Serial.print(" P="); Serial.println(P,2);
client.print("P "); client.println(P,2);
} else {
Serial.println("Error: BPM180 getPressure failed");
}
} else {
Serial.println("Error: BPM180 getTemperature failed");
}
} else {
Serial.println("Error: BPM180 startTemperature failed");
}
170  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

} else if (strstr(line,"T?")==line) {
temp=100*3.3*analogRead(0)/1023;
client.print("T "); client.println(temp,1);
} else {
Serial.println("unknown command, disconnecting");
client.stop();
}
client.flush();
}
digitalWrite(LED_BUILTIN,HIGH);
yield();
}
At the top we include the I2C library Wire.h, and we use the SFE BMP180 library to interface
the BMP180 pressure because it makes the sketch more compact. Next, we instantiate
the pressure object that provides the pressure and temperature measurements from the
sensor. Then we define the I2C address for the HYT221 humidity sensor and the WLAN
information, before creating the WiFiServer server(). In the setup() function we first
establish the WLAN connection and then start the server, just as in Section 4.6.3. At the
end of the setup() function, we initialize the pressure sensor and report an error, if this
fails. In the loop() function, we first declare some variables before waiting for a client
to connect. Once it is available, we wait until a request from the client is received, and
then it is decoded in the following strstr() construction. In the request for the humidity
measurement (H?), both the humidity and the temperature from the sensor are determined,
but only the humidity returned. The temperature that is returned as a result of a TH?
request is therefore always the one determined in the previous request for humidity. The
same mechanism is used for the pressure measurement, which is started by a P? request.
It retrieves both pressure and temperature data from the sensor. Last, the temperature
from the LM35 analog sensor can be requested. If an unknown command is received, the
network connection is closed. Finally, the yield() function permits the microcontroller to
do internal bookkeeping.
On the Raspi, we use a cron job to query the microcontroller once every minute to read
the measurement values and place them into a rrdtool data base. The following Python
script reads the five values from the microcontroller via WLAN, and prepares an output
string called out with values separated by a colon “:,” which already conforms to the
syntax of the rrdtool update command.
# read_from.py reads queries from socket 1137 at IP
import socket, atexit, time, sys
def cleanup():
sock.send("quit\n")
sock.close()
if len(sys.argv) < 2:
print("Usage: read_from.py IP <list of queries>")
sys.exit(2)
else:
atexit.register(cleanup)
sock=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
sock.connect((sys.argv[1],1137)) # socket 1137
out=’’
for query in sys.argv[2:]:
 Example: Weather Station with Distributed Sensors  171

sock.send(query + "?\n")
time.sleep(0.3)
reply=sock.recv(1000);
if (len(out))>0:
out+=’:’
out+=reply[len(query):].strip()
print(out)
The above Python script requires the IP number of the sensor node and the list of parameters
to query as command-line arguments. It should be called in the following form:

python read_from.py 192.168.20.144 P TP H TH T

and it will return the five values separated by colons. In the script, we first import the
necessary functionality and define the cleanup() function to close the socket at the end of
a transaction. Then we check whether sufficient command line arguments are present, and
exit with a usage note. If sufficient command-line arguments are present, we register the
cleanup() function, open the socket to the IP number specified as the first command-line
argument, and then loop over the list of the rest of the command line arguments. Inside
the loop we build the query by appending the question mark and newline character, wait a
short time to give the sensor node some time to do the measurement, and then receive its
reply. The next two lines strip the unwanted characters from the reply and prepend a colon
unless the first entry is handled. Finally the output string out is returned.
The following shell script calls the read from.py script to obtain the measurement
values, and stuffs them into the database.
#!/bin/bash
DB=/home/pi/rrdtool/weather.rrd
TEMP=$(/usr/bin/python /home/pi/rrdtool/read_from.py \
192.168.20.144 P TP H TH T)
/usr/bin/rrdtool update $DB N:$TEMP
Note that we place the arguments of read from.py on the following line, and use the
continuation character, a backslash, to indicate that. But before we start to fill the database
with the rrdtool update command, we must create it with the following command. We
assumed it is executed in the /home/pi/rrdtool directory.
rrdtool create weather.rrd --step 60 \
DS:P:GAUGE:180:900:1100 \
DS:TP:GAUGE:180:-20:100 \
DS:H:GAUGE:180:0:100 \
DS:TH:GAUGE:180:-20:100 \
DS:T:GAUGE:180:-20:100 \
RRA:AVERAGE:0.5:1:2880 \
RRA:AVERAGE:0.5:10:2880 \
RRA:AVERAGE:0.5:60:2880

It creates the weather.rrd database file that expects values once every 60 seconds, and
contains five measurement columns (DS) for the pressure, temperature from the pressure
sensor, humidity, temperature from the humidity, sensor and the temperature from the
LM35 sensor. Then it defines the archive (RRA), which is filled by single averages for 2880
samples, which amounts to 2 days. The next two lines define averages over 10 samples or 10
172  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

minutes, for a total of 2880 samples or 20 days, and finally, hourly averages over 60 samples
for 120 days.
Now we just have to wait until some measurements make it into the database, and then
create graphs in the same way we did in Section 5.6.3 using the rrdtool graph command.
The command to create one set of graphs for temperature, pressure, and humidity, called
weathergraph.sh, is the following:
#!/bin/bash
# /home/pi/rrdtool/weathergraph.sh
S=$1
DB=/home/pi/rrdtool/weather.rrd
PICDIR=/home/pi/public_html/weather
/usr/bin/rrdtool graph $PICDIR/temperature${S}.png -s $S \
-w 800 -h 200 \
-t "Temperature" -v "T [C]" -l 16 -u 26 -r \
DEF:t0=$DB:T:AVERAGE LINE1:t0#FF0000:"LM35" \
DEF:t1=$DB:TP:AVERAGE LINE1:t1#00AF00:"T(P)" \
DEF:t2=$DB:TH:AVERAGE LINE1:t2#0000FF:"T(H)";
/usr/bin/rrdtool graph $PICDIR/pressure${S}.png -s $S \
-w 800 -h 200 \
-t "Barometric Pressure" -v "P [mbar]" -l 950 -u 1050 -r \
DEF:t0=$DB:P:AVERAGE LINE1:t0#FF0000:"Pressure";
/usr/bin/rrdtool graph $PICDIR/humidity${S}.png -s $S \
-w 800 -h 200 \
-t "Humidity" -v "Humidity [%]" \
DEF:t0=$DB:H:AVERAGE LINE1:t0#FF0000:"Humidity";
It starts by copying the first command line argument, which should contain the start time
of the display to the variable S, declares the database file to use, and the location where the
graphs will be stored. For this we choose a directory under /home/pi/public html, which
is accessible to the apache web server. The first rrdtool graph produces the temperature
plot with the temperatures from the three sensors. Note that the graph file carries the time
period appended to its name. The second produces the pressure, and the last the humidity
graph, all in the same way discussed in Section 5.6.3. In order to produce graphs with
time spans of 8 hours, 1 day, 1 week, 1 month, and 3 months, we call weathergraph.sh
with different starting dates as command line arguments, and place the five commands in
a separate script, called allweathergraph.sh. It is reproduced here.
#!/bin/bash
# /home/pi/rrdtool/allweathergraph.sh
/home/pi/rrdtool/weathergraph.sh -8h > /dev/null
/home/pi/rrdtool/weathergraph.sh -1d > /dev/null
/home/pi/rrdtool/weathergraph.sh -1w > /dev/null
/home/pi/rrdtool/weathergraph.sh -1m > /dev/null
/home/pi/rrdtool/weathergraph.sh -3m > /dev/null
The call to weathergraph.sh uses the absolute path because we want to run it from a
cron job to create new graphs every 10 minutes. We also redirect the output into the big
bit-bucket /dev/null. We execute this script regularly as a cron job with the command
crontab -e in order to add the line
*/10 * * * * /home/pi/rrdtool/allweathergraph.sh
 Example: Weather Station with Distributed Sensors  173

Figure 8.2 The weather station web page served by the Raspi.

in the crontab file.

It remains to prepare the index.html file in the directory with the graphs. The following
HTML file presents the data as the simple web page shown in Figure 8.2.
<!DOCTYPE HTML>
<HTML>
<HEAD>
<TITLE>Raspi Weather Station</TITLE>
<META http-equiv="refresh" content="120">
</HEAD>
<BODY>
<H1 ALIGN=CENTER>Raspi Weather Station (8 hours)</H1>
Display data for:
<A HREF=index.html>8 hours</A>__
174  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

<A HREF=index-1d.html>1 day</A>__

<A HREF=index-1w.html>1 week</A>__
<A HREF=index-1m.html>1 month</A>__
<A HREF=index-3m.html>3 months</A>
<HR>
<BR>
<IMG SRC="temperature-8h.png" ALT="Temperature">
<IMG SRC="pressure-8h.png" ALT="Pressure">
<IMG SRC="humidity-8h.png" ALT="Humidity">
</BODY>
</HTML>
The file contains the already-known header information with the title Raspi Weather
Station, and the automatic refresh rate of 120 seconds. Then a line with options of dif-
ferent time ranges follows. Each entry points to a different web page residing in the same
directory, but serving the graphs for the other ranges. Remember, the graphs are automat-
ically updated by the cron job, but the different index-XX.html pages are constant.
Having the weather data displayed for different time ranges is certainly adequate for
home use, but in a laboratory we want to integrate the data from the sensor nodes into
the control system. If we already have the EPICS system running, all we have to do is to
prepare the protocol and database files for our weather data. The former, written in a more
compact way, is the following.
# ./tempApp/Db/weather.proto
Terminator = CR LF;
get_T {out "T?"; in "T %f"; ExtraInput = Ignore; }
get_P {out "P?"; in "P %f"; ExtraInput = Ignore; }
get_TP {out "TP?"; in "TP %f"; ExtraInput = Ignore;}
get_H {out "H?"; in "H %f"; ExtraInput = Ignore; }
get_TH {out "TH?"; in "TH %f"; ExtraInput = Ignore; }
It follows the already-known pattern of sending a query with a question mark appended, and
then receives the query, followed by the measurement value. To accompany this file, which
orchestrates the low-level communication, we need a data base file, named weather.db,
with the following contents.
# ./tempApp/Db/weather.db
record(ai, "$(USER):T") {
field(DESC, "Temperature")
field(SCAN, "10 second")
field(DTYP, "stream")
field(INP, "@weather.proto get_T $(PORT)")
}
record(ai, "$(USER):P") {
field(DESC, "Pressure")
field(SCAN, "10 second")
field(DTYP, "stream")
field(INP, "@weather.proto get_P $(PORT)")
}
record(ai, "$(USER):TP") {
field(DESC, "TemperaturePressure")
 Example: Weather Station with Distributed Sensors  175

field(SCAN, "10 second")

field(DTYP, "stream")
field(INP, "@weather.proto get_TP $(PORT)")
}
record(ai, "$(USER):H") {
field(DESC, "Humidity")
field(SCAN, "10 second")
field(DTYP, "stream")
field(INP, "@weather.proto get_H $(PORT)")
}
record(ai, "$(USER):TH") {
field(DESC, "TemperatureHumidity")
field(SCAN, "10 second")
field(DTYP, "stream")
field(INP, "@weather.proto get_TH $(PORT)")
}
Here we find analog input ai records that define the process variables $(USER):XX, and link
them to the respective protocol functions. Moreover, we need to include the new database
file in the Makefile by adding the line
DB += weather.db
and add the following two lines to the ./temp/iocBoot/ioctemp/st.cmd EPICS command
file
drvAsynIPPortConfigure("SOCKET","192.168.20.144:1137",0,0,0)
dbLoadRecords("db/weather.db","PORT=’SOCKET’,USER=’weather’")
and finally, call make in the base directory of the temp hierarchy. Then we execute st.cmd
as before, either from the command line or automatically with an init script as discussed at
the end of Section 6.5. The process variables are now accessible from any computer on the
network. To access the barometric pressure, we issue caget weather:P, and similarly for
the other variables. In this way, they can be seamlessly included in logging, monitoring, or
other display programs that EPICS provides. Adding more sensors is as simple as adding
protocol and database files as well as including them in the Makefile in the Db directory,
and adding two lines in the st.cmd program.

QUESTIONS AND PROJECT IDEAS

1. What is the average barometric pressure at sea level, on a rainy day, in a heavy storm,
in a hurricane, in La Paz?
2. What is the typical humidity where you live, in a rain forest, in Death Valley, or at
the La Silla Observatory in Chile?
3. Add a dust sensor to turn the weather station into an environmental logging station.
4. Add an LDR or a phototransistor to the weather station to log the brightness.
5. Add a remote-controlled fan that stirs the air upon request.
6. Rewrite the client program for the weather station running on the NodeMCU and
give it an MQTT interface.
 CHAPTER 9

Example: Geophones

In the second example, we use the SM-24 geophone from Figure 2.12 to record ground
vibration spectra. Our plan is to attach the SM-24 to a battery-powered microcontroller
that samples 1024 values at a high rate and transmits the measurements via WLAN to the
host computer, our Raspi. There we use octave to postprocess and Fourier-transform the
samples, as well as present the results. Besides using octave, we also make the data available
to EPICS, such that the standard EPICS programs for display and post-processing can be
used.
For the sensor node, we use the NodeMCU microcontroller and the Ticker.h library
that allows sampling with a rate of 1000 times per second. Since the frequency range of the
SM-24 sensor is from 10 to 240 Hz, the sampling rate of 1 kHz is four times the maximum
frequency and should be adequate. The SM-24 sensor only produces a very small and bipolar
output voltage in the mV range. We therefore need to amplify the voltage in order to match
the input voltage range of 0 to 3.3 V of the ADC on the NodeMCU.
We base the amplifier on the circuit shown in Figure 2.20, but increase the amplification
to ×100 and add a few components to adapt it to our sensor and arrive at the circuit shown
in Figure 9.1. The amplification is mostly given by the ratio of R3 and R7 to R1 and R2 .
We also add a 1.5 kΩ resistor R9 across the input terminals in order to damp the resonant
peak at 10 Hz, which the SM-24 sensor exhibits according to the datasheet. If we consider
the internal coil resistance (375 Ω) of the sensor, mounting an additional 2.2 µF capacitor
across the input terminals creates a low-pass filter with a cutoff frequency of around 200 Hz.
As discussed in Section 2.2.4, this avoids aliasing of higher frequencies into the digitizing
bandwidth from 0 Hz to the Nyquist frequency of 500 Hz. The other components have the
same functionality as discussed in Section 2.2.2.
After the signal from the geophone is amplified, we use the ADC on the NodeMCU
microcontroller and collect 1024 samples: one sample every millisecond, and then pass the
digitized samples to the host computer via WLAN. The code that achieves this is the
following:
// Minimal time-series-server, V. Ziemann, 170324
const char* ssid = "messnetz";
const char* password = "zxcvZXCV";
const int port = 1137;
#include <ESP8266WiFi.h>
WiFiServer server(port);
const uint16_t npts=1024; // number of samples
const int sample_period=1; // ms

177
178  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 9.1 The amplifier to interface an SM-24 geophone to the NodeMCU.

#include <Ticker.h>
Ticker SampleFast;
uint16_t sample_buffer[npts];
volatile uint16_t isamp=0,sample_buffer_ready=0;
char line[30];
void samplefast_action() { //.............samplefast_action
sample_buffer[isamp]=analogRead(0);
isamp++;
if (npts == isamp) {
SampleFast.detach();
isamp=0;
sample_buffer_ready=1;
}
}
void setup() { //.....................................setup
pinMode(LED_BUILTIN,OUTPUT);
digitalWrite(LED_BUILTIN,LOW);
Serial.begin(115200);
WiFi.begin(ssid,password);
while (WiFi.status() != WL_CONNECTED) {
delay(500); Serial.print(".");
}
Serial.println("");
Serial.print("Wifi connected to "); Serial.println(ssid);
Serial.print("Server IP address: "); Serial.println(WiFi.localIP());
server.begin();
Serial.print("Server started on port "); Serial.println(port);
digitalWrite(LED_BUILTIN,HIGH);
}
void loop() { //......................................loop
WiFiClient client = server.available();
while (client) {
 Example: Geophones  179

while (!client.available()) {
delay(5);
if (!client.connected()) break;
}
client.readStringUntil(’\n’).toCharArray(line,30);
Serial.print("Received: "); Serial.println(line);
if (strstr(line,"WF?")==line) {
digitalWrite(LED_BUILTIN,LOW);
sample_buffer_ready=0;
SampleFast.attach_ms(sample_period,samplefast_action);
while (!sample_buffer_ready) {delay(2);} // wait until done
for (int i=0;i<npts-1;i++) {
client.print(sample_buffer[i]); client.print(", ");
}
client.println(sample_buffer[npts-1]);
digitalWrite(LED_BUILTIN,HIGH);
} else {
Serial.println("unknown command, disconnecting");
client.stop();
}
client.flush();
}
yield();
}
At the top of the code, we first enter the credentials and port number for the WLAN,
and then import support to create the server() in the next line. The constants npts
and sample period specify the number of samples to acquire and the number of millisec-
onds to wait between acquisitions, respectively. The Ticker.h library adds support for
timed and interrupt-driven functions. The variables declared next are related to filling the
sample buffer. The function samplefast action() is called automatically by the timer
and is executed once a millisecond. In this function we first do an analog conversion and
place the value into the sample buffer before incrementing the variable isamp, such that
the next sample ends up in the following slot in the array. Once the desired number npts of
samples are acquired, we disable the interrupt with the call to SampleFast.detach(), set
the sample pointer isamp to zero, and set the sample buffer ready flag to signal the main
program that the acquisition of the npts samples is complete. In the setup() function, we
first set the mode of the pin with the built-in LED and turn it on before enabling the serial
line for debugging, and connect to the WLAN. Then we start the server process with the
call to server.begin() and report everything to the serial line, before turning the LED
off.
The organization of the loop() function resembles earlier examples. We wait for a client
to connect, and then parse the request. If the query-string is WF?, we turn on the LED
and ensure the variable sample buffer ready is zero before we start the acquisition with
the call to SampleFast.attach ms(). The arguments of the function are the periodicity
in ms to call the function specified as the second argument. This launches the automatic
acquisition that continues in the background. All we need to do in the loop() function is
to monitor whether sample buffer ready is still zero, in which case we wait a bit longer.
But once it becomes nonzero, which happens in the samplefast action() function after
the desired number of samples is collected, we break the waiting loop. Finally, we are ready
180  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

The prototype with the SM-24 geophone connected to the amplifier board
Figure 9.2
and the NodeMCU.

to send the values, separated by commas, with client.print() function calls, back to the
client and turn the LED off.
We program this sketch into the NodeMCU, whence it waits for a client to connect
and to request samples. Here we use octave to request a time series of samples from the
geophone and display both the received raw time-series data and its Fourier transform, the
spectrum. The code that achieves this is the following:

% getTimeSeries.m, V. Ziemann, 170324

dev=tcp("192.168.20.144",1137);
npts=1024;
tcp_write(dev,"WF?\n");
sleep(1);
data=zeros(1,npts);
for i=1:npts
data(i)=str2double(tcp_getvalue(dev));
end
tcp_close(dev);
subplot(2,1,1)
plot(data)
xlim([0,npts])
xlabel(’Time [ms]’)
ylabel(’Amplitude [ADC bits]’)
subplot(2,1,2);
data=data-mean(data);
fftdata=2*abs(fft(data))/npts;
frequency=(0:(npts/2-1))*500/(npts/2);
plot(frequency,fftdata(1:npts/2))
xlabel(’Frequency [Hz]’)
ylabel(’Spectral density [ADC bits]’)
 Example: Geophones  181

Figure 9.3 The raw time series from the geophone and the corresponding spectrum.

print(’spectrum.png’,’-S1000,700’);
Also, the octave script follows earlier examples. We first define the socket to connect to
in the tcp() function call and then specify the number of points to acquire. This has to
match the number of samples declared on the NodeMCU. Next, we send the query string
WF? to the socket, wait a while, and then collect npts samples, which we store in the array
data(). For this we use the function tcp getvalue() that extracts one comma-separated
value from the socket. We discuss that function in more detail below. The received value is
encoded as a character string, and we therefore need to convert it to a float value with the
str2double() function. Once all samples are received and stored in data(), we close the
socket.
In the following lines, we create two subplots of which the upper contains the raw
samples. Since the sample period is a millisecond, we state that as the horizontal axis label.
The vertical axis is just the raw ADC conversion, which comes from the 10-bit ADC on
the NodeMCU, and lies between 0 and 1023. For the spectrum that we show in the lower
subplot, we first subtract the mean of the values in order to avoid a huge spectral peak at
zero. It is caused by placing the signal in mid-range of the ADC with the preamplifier. Then
we Fourier transform the samples and create an array frequency with the frequency values
from zero to the Nyquist frequency before plotting the spectrum and labeling the axes.
Finally, we use the print() function to create an image file that contains the displayed plot
with the specified size in pixels, 1000 × 700 in this case.
In the octave script, we use the function tcp getvalue to read a single sample from
the input stream. The function is similar to the queryResponse.m function we used in
Section 5.5.4. Here is the octave code.
182  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

% tcp_getvalue.m, V. Ziemann, 170324

function out=tcp_getvalue(dev)
i=1;
int_array=uint8(1);
while true
val=tcp_read(dev,1);
if ((val==’,’) || (val==0xA)) break; end
int_array(i)=val;
i=i+1;
end
out=char(int_array);
We need to provide the handle dev to the socket as input parameter, and the function
returns a character string that contains all characters until either a comma or the next end-
of-line character. In the function, we read one character at a time from the socket, and stop
reading once the comma or end-of-line character 0xA appears. Then we return all received
data after conversion to characters.
We show the sensor node with SM-24 geophone, amplifier, and NodeMCU in Figure 9.2.
Running the octave script with the NodeMCU connected to the same WLAN as the Raspi
produces a file spectrum.png, shown in Figure 9.3. Running the octave script again will
overwrite the file, and tapping with a finger on the table while recording samples results in
a much more noisy spectrum.
Recording waveforms in octave and processing them further is convenient to produce
plots of the spectra, but if we want to include the geophones in our EPICS control system,
we need to provide the database and prototype files. We start with the database record for
the time series or waveform. It is reproduced below,

# .../Db/geophone.db
record(waveform, "$(USER):wf") {
field(DESC,"Geophone waveform")
field(DTYP,"stream")
field(SCAN,"10 second")
field(NELM,"1024")
field(FTVL,"FLOAT")
field(INP,"@geophone.proto get_wf $(PORT)")
}
where we have to declare a waveform record, because we want to acquire the entire stream of
1024 samples in one step. The first three fields provide the description, the declaration as a
stream device, and the rate at which the waveforms are collected. The next two fields declare
that we record 1024 float values, before linking to the protocol file geophone.proto and the
function get wf in the last field of type INP. The referenced protocol file geophone.proto
is reproduced below.
# .../Db/geophone.proto
Terminator = CR LF;
get_wf {
ExtraInput = Ignore;
replyTimeout=2000;
out "WF?";
separator=",";
 Example: Geophones  183

in "%i";
}
Here we first define the termination string of one command as CR LF, and then define the
get wf function. For robustness, we require it to ignore any input that does not make sense,
and wait 2000 ms for the reply from the NodeMCU. The request for a new waveform consists
of sending WF? and receiving integers ("%i"), separated by commas. The number of values
to receive from the NodeMCU, 1024, is already specified in the geophone.db file.
In a final step, we need to tell EPICS where to find the NodeMCU and what socket to
connect to. This is done, as discussed in Chapter 6 on EPICS, in the st.cmd command file,
which contains the following two lines:
drvAsynIPPortConfigure("SOCKET2","192.168.20.144:1137",0,0,0)
dbLoadRecords("db/geophone.db","PORT=’SOCKET2’,USER=’geophone’")
The first defines the IP and port number of the connected NodeMCU device, and the second
line specifies the database file geophone.db that describes the protocol that is used when
communicating with the NodeMCU. After adding DB += geophone.db to the Makefile, we
recompile the EPICS IOC and then execute st.cmd from the command line, or, once we
make sure that everything works as expected, we create an init file to start the IOC at boot
time, following the procedure we discussed in Section 6.4. At this point, with a running
IOC continuously retrieving waveforms from the NodeMCU, we can obtain the most recent
waveform with caget geophone:wf, which first returns the number of samples, here 1024,
followed by the 1024 samples. Other EPICS programs, on any computer connected to the
same network, can also obtain the same waveforms, a new one every 10 seconds, as specified
in the geophone.db record file.

QUESTIONS AND PROJECT IDEAS

1. Discuss the advantage of using interrupt-driven acquisition over measuring 1000 sam-
ples in a loop with a delay(1) between acquisitions.
2. What is the purpose of the low-pass filter discussed in this chapter?

3. What is the maximum frequency we can determine uniquely when sampling at a rate
of 1000 samples per second?
4. What is the resolution (smallest detectable difference of frequencies) of the spectrum
shown in Figure 9.3? How can you increase it?
5. Discuss why the sample rate of the Ticker.h library is limited to 1000 samples per
second.
6. Use a pin diode as sensor and sample it at a rate of 1000 samples per second. Explore
different flickering light sources such as lamps, TV screens, or computer monitors.
7. Use the tone() function to generate an oscillating signal on an output pin. Then
sample that pin with the built-in ADC and observe the signal. What is the high-
est frequency you can uniquely observe? What happens when you go beyond that
frequency?
8. Instead of the geophone, use a microphone and “observe” your voice.
184  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

9. Add an MCP3304 ADC to the NodeMCU and connect two geophones, fill two
sample buffer arrays with data in the interrupt handler, and send both waveforms
to octave. If the geophones are placed on a large table far apart, try out to determine
the speed of sound by tapping on the table, and determine the shift in time between
the two waveforms.
 CHAPTER 10

Example: Monitor for the

Color of Water

In this example, we determine the color of water, or more accurately stated, the absorption
of material dissolved in water. An example is algae, which often proliferate during the
summer and turn the water green because algae absorb predominantly red light. In our
experiment, we investigate the absorption of light of different colors by turning a three-
color LED on and off, and observe the resulting modulation of the signal recorded by a
phototransistor. If red light is absorbed, the resulting modulation depth of red is reduced,
likewise for the other colors. By comparing the situation with LED on versus off, the system
removes some of the influence of ambient background light, at least to some extent. On the
left in Figure 10.1, we illustrate the absorption measurement with the RGB-LED on the
left and the phototransistor on the right, with the absorbing material in between. We note
that we can also use such a system to determine variations in the reflection from a surface,
as is shown on the right in Figure 10.1.
We construct the setup shown in Figure 10.1, using an Arduino UNO, and an RGB-LED,
which houses a red, a green, and a blue LED in the same housing with a single connector for
a common cathode or anode. In our case we use one with a common anode that is connected
to the positive supply voltage. The respective color lights up if the controlling pin on the
UNO is LOW. The phototransistor is an SFH3310, but any other model, sensitive in the
visual spectral range, should work. The very simple setup is shown in Figure 10.2.
On the left we see the Arduino UNO and on the bottom right a small breadboard

On-Off modulated Light

light shield

R
G
Common
B
Photo
transistor Three-
Three-color LED Phototransistor color LED
Absorbing
material Reflecting surface

Figure 10.1 The setup to measure the color-dependent absorption (left) and the re-
flection from a surface (right).

185
186  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 10.2 The setup to measure the water color with an Arduino UNO.

with the SFH3310 phototransistor. The emitter is connected to ground and the collector is
connected via a 68 kΩ resistor to the positive supply voltage. The resistor depends somewhat
on the phototransistor and is determined by the requirement for a clearly visible modulation
depth on the analog input pin, which is connected to the collector with the blue wire. A
little experimenting will result in a reasonable value for the resistance. We add a 10 µF
and a 100 nF capacitor to buffer the supply voltage. The RGD-LED is placed on the upper
breadboard with the common anode connected to the positive supply voltage, and the three
“color pins” via current-limiting resistors to pins D2, D3, and D4 on the UNO. We chose
220 Ω for the red LED, 180 Ω for the green, and 150 Ω for the blue in order to account for
the different voltage drops of the different-color LEDs.
In the sketch that runs on the UNO, we need to toggle the LEDs one at a time and
synchronously record the intensities recorded by the phototransistor. This is achieved by
the following code.
// Color monitor, V. Ziemann,170823
int repeat=20;
void alloff() { //.......................alloff
digitalWrite(2,HIGH); // red
digitalWrite(3,HIGH); // green
digitalWrite(4,HIGH); // blue
}
float measure(int pin, int repeat) { //..............measure
int hi,lo;
float sum=0;
alloff();
delay(10);
for (int i=0;i<repeat;i++) {
 Example: Monitor for the Color of Water  187

digitalWrite(pin,LOW);
delay(5);
lo=analogRead(0);
digitalWrite(pin,HIGH);
delay(5);
hi=analogRead(0);
sum+=(hi-lo);
}
return sum/repeat;
}
void setup() { //.........................setup
Serial.begin(9600); while (!Serial) {;}
pinMode(2,OUTPUT); // red
pinMode(3,OUTPUT); // green
pinMode(4,OUTPUT); // blue
alloff();
digitalWrite(2,LOW);
}
void loop() { //...........................loop
if (Serial.available()) {
char line[30];
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"OFF")==line) {
alloff();
} else if (strstr(line,"RED")==line) {
alloff();
digitalWrite(2,LOW);
} else if (strstr(line,"GREEN")==line) {
alloff();
digitalWrite(3,LOW);
} else if (strstr(line,"BLUE")==line) {
alloff();
digitalWrite(4,LOW);
} else if (strstr(line,"COLOR?")==line) {
float red=measure(2,repeat); // red
float green=measure(3,repeat); // green
float blue=measure(4,repeat); // blue
Serial.print("COLOR "); Serial.print(red);
Serial.print("\t");Serial.print(green);
Serial.print("\t");Serial.println(blue);
} else if (strstr(line,"REPEAT?")==line) {
Serial.print("REPEAT "); Serial.println(repeat);
} else if (strstr(line,"REPEAT ")==line) {
repeat=(int)atof(&line[6]);
}
}
delay(10);
}
First we declare a variable repeat and a function to turn all the LEDs off. The input
188  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 10.3The color sensor with the RGB-LED on the left and the phototransistor
on the right.

parameters of the function measure() are the pin to be toggled and the number of times
to repeat the measurement. We add a delay of 5 ms in order for the voltage to stabilize
after the LED is switched on or off, before measuring the voltage at the collector of the
phototransistor with the analogRead() function. The difference of the reading with LED
on and off is accumulated in the variable sum, and finally the average of the accumulated
differences is returned to the calling program. In the setup() function, we initialize the
serial communication and the pins for the RGB-LED, turn all LEDs off, and then turn the
red LED on in order to indicate that the system is up and running. The loop() function
uses the same mechanism we use throughout this book to turn all LEDs OFF or turn one of
the colors RED, GREEN, or BLUE on. If the query is COLOR? the system measures the variation
on the phototransistor when toggling the LEDs on and off, and returns three numbers for
the three color-variations. Finally, the query REPEAT nnn sets the variable repeat to nnn
and REPEAT? returns the current value.
This system measures the absorption of any material placed between the LED and the
phototransistor, which are pointing at each other as shown in Figure 10.1, and in Figure 10.3
where we show a simple prototype. Moreover, in order to measure the absorption in water,
we need to encapsulate the LED and transistor in a protective coating, for which we use
two layers of heat-shrink insulation material to cover the solder connection of wires to the
pins of the LED and phototransistor.
If we want to deploy the system at a somewhat remote location, it is rather straight-
forward to extend the MQTT client program from Chapter 7 to monitor the color and
publish the results once every minute. For this purpose, we use a NodeMCU instead of a
UNO, and connect the RGB-LED to pins D2, D3, D4 on the NodeMCU, and the collector
of the phototransistor to the analog input pin A0. The following sketch implements this
functionality.

// MQTT client water color, V. Ziemann, 170906

const char* ssid = "messnetz";
const char* password = "...........";
const char* broker = "192.168.20.1";
#include <Ticker.h>
volatile uint8_t do_something=0;
Ticker tick;
void tick_action() {do_something=1;} // executed regularly
#include <ESP8266WiFi.h>
#include <PubSubClient.h>
WiFiClient espClient;
 Example: Monitor for the Color of Water  189

PubSubClient client(espClient);
float measure(int pin, int repeat) { //..............measure
int hi,lo;
float sum=0;
digitalWrite(D2,HIGH);
digitalWrite(D3,HIGH);
digitalWrite(D4,HIGH);
delay(10);
for (int i=0;i<repeat;i++) {
digitalWrite(pin,LOW);
delay(5);
lo=analogRead(A0);
digitalWrite(pin,HIGH);
delay(5);
hi=analogRead(A0);
sum+=(hi-lo);
}
return sum/repeat;
}
void setup() { //.................................setup
pinMode(D2,OUTPUT);
pinMode(D3,OUTPUT);
pinMode(D4,OUTPUT);
Serial.begin(115200);
WiFi.begin(ssid,password);
while (WiFi.status() != WL_CONNECTED) {
delay(500); Serial.print(".");
}
Serial.print("\nWifi connected to "); Serial.println(ssid);
Serial.print("with IP address: "); Serial.println(WiFi.localIP());
client.setServer(broker, 1883); // 1883 = default MQTT port
tick.attach(60,tick_action); // execute tick_action every 60 seconds
}
void loop() { //..................................loop
while (!client.connected()) { // try to connect to broker
if (!client.connect("PubSub1")) {delay(5000);}
}
client.loop();
if (do_something) {
do_something=0;
char message[30];
float red=measure(D2,20); // red
float green=measure(D3,20); // green
float blue=measure(D4,20); // blue
sprintf(message,"%d %d %d",(int)red,(int)green,(int)blue);
client.publish("node1/color",message);
}
yield();
}
190  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

We first declare the network credentials and the IP address of the broker before in-
cluding the Ticker.h library. It orchestrates the repeated measurements by calling the
tick action() function at well-defined time-intervals. Inside the function we only set the
variable do something to one, because this variable is monitored in the loop() function
and triggers some activity, if it is set. Next, we include the WiFi libraries, and declare a
WiFiClient and the MQTT PubSubClient named client, before defining the measure()
function, which is a straight copy from the previous example. The setup() function declares
the mode of the pins, configures the serial line, connects to WLAN, configures the broker,
and starts the ticker to call tick action every 60 seconds, which triggers the measurements.
In the loop() function, we first ensure that the NodeMCU is connected to the broker be-
fore calling client.loop() to service any background activities pertaining to MQTT. Then
we test whether the variables do something is set, and execute the measurement with the
measure() function for each color, construct a string that contains the three values, and
publish the measurement.
Using this system, it is possible to submerge a sensor similar to the one shown in Fig-
ure 10.3 in a garden pond, place the NodeMCU in an enclosure next to the pond, possibly
battery powered, and make sure it is within reach of the WLAN messnetz. That is all that
is needed to monitor algae in the pond.

QUESTIONS AND PROJECT IDEAS

1. How do the reported color-variations change if you remove the capacitors from the
small breadboard? Explain why!
2. How do the reported color-variations change if the waiting time of 5 ms in the
measure() function is reduced to 1 ms? What happens if it is increased?
3. Connect the RGB-LED to output pins with pulse-width modulation capability and
write a sketch that allows you to set a number of standard colors, such as orange or
magenta.
4. Write a sketch to set the color of the LEDs via MQTT.
5. Build a reflectivity sensor similar to the one shown on the right in Figure 10.1, and
connect it to the Arduino UNO.
 CHAPTER 11

Example: Capacitance
Measurement

In this chapter, we measure the unknown capacitance of a capacitor by first charging it,
then turning off the charging supply and discharging through a parallel resistor, while
repeatedly measuring the voltage drop across the capacitor. The voltage drops exponentially
U ∝ e−t/τ with time constant τ = RC, where R is the discharging resistor and C the
unknown capacitance. Thus, from a linear least-squares fit to the logarithm of the voltage,
we can determine the time constant, and from the known resistor value, also the unknown
capacitance.
Figure 11.1 shows the experimental setup. The Arduino UNO is on the right and the
small breadboard showing the capacitor with the R = 33 kΩ resistor to discharge the capac-
itor connected in parallel; both are connected to ground and analog pin A0 on the Arduino.
The capacitor can be charged by pulling digital pin D2 high; it is connected by a wire and
a 220 Ω resistor. The latter resistor limits the initially flowing current, when the capacitor
is fully discharged.
The task of the UNO is to first charge the capacitor, by configuring the digital output
pin D2 as OUTPUT and setting its value to HIGH. Once a measurement is requested, the D2
is reconfigured as input with pinMode(2,INPUT), and setting the output nevertheless to
LOW, to ensure that the internal pull-up resistors are not used. Once the charging voltage
is disconnected, we repeatedly read analog pin A0 from an interrupt service routine until
the required number of samples is collected in memory. Once the raw data are collected, we
make a linear least-squares fit to the data and determine the capacitance from the slope.
This plan of action is realized in the following Arduino sketch.
// Capacitance measurement, V. Ziemann, 170629
const int npts=100;
volatile int isamp=0,sample_buffer_ready=0;
uint16_t sample_buffer[npts],nsamp=npts,timestep=5;
float R=33e3; // 33 kOhm
#include <MsTimer2.h>
void timer_action() { //..................timer_action
sample_buffer[isamp]=analogRead(A0);
isamp++;
if (nsamp == isamp) {
MsTimer2::stop();

191
192  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 11.1 The setup to measure the capacitance.

isamp=0;
sample_buffer_ready=1;
}
}
double linfit(int n, uint16_t y[]) { //.........linfit
double ay0=0,ay1=0;
double S0=n;
double S1=0.5*n*(n+1);
double S2=n*(n+1.0)*(2.0*n+1)/6.0;
for (int k=0;k<n;k++) {
ay0+=k*log(y[k]);
ay1+=log(y[k]);
}
return (S0*ay0-S1*ay1)/(S2*S0-S1*S1);
}
void setup() { //...............................setup
Serial.begin(9600);
while (!Serial) {delay(10);}
pinMode(2,OUTPUT);
digitalWrite(2,HIGH);
}
void loop() { //................................loop
if (Serial.available()) {
char line[30];
 Example: Capacitance Measurement  193

Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"CAP?")) {
nsamp=(int)atof(&line[5]);
nsamp=min(nsamp,npts);
if (nsamp==0) nsamp=npts;
pinMode(2,INPUT);
digitalWrite(2,LOW); // disables internal pullup
MsTimer2::set(timestep, timer_action);
MsTimer2::start();
sample_buffer[isamp]=analogRead(A0);
isamp++;
} else if (strstr(line,"WF?")) {
Serial.print("WF "); Serial.println(nsamp);
for (int i=0; i<nsamp; i++) Serial.println(sample_buffer[i]);
} else if (strstr(line,"TIMESTEP ")) {
timestep=(int)atof(&line[9]);
Serial.print("TIMESTEP "); Serial.println(timestep);
} else if (strstr(line,"TIMESTEP?")) {
Serial.print("TIMESTEP "); Serial.println(timestep);
} else if (strstr(line,"RESISTOR ")) {
R=atof(&line[9]);
Serial.print("RESISTOR "); Serial.println(R);
} else if (strstr(line,"RESISTOR?")) {
Serial.print("RESISTOR "); Serial.println(R);
}
}
if (sample_buffer_ready==1) {
sample_buffer_ready=0;
pinMode(2,OUTPUT); // start charging capacitor
digitalWrite(2,HIGH);
delay(100);
double slope=linfit(nsamp,(uint16_t)sample_buffer);
double capacitance=-1e6*timestep*1e-3/(slope*R); // in uF
Serial.print("CAP "); Serial.println(capacitance,4);
if (sample_buffer[0] < 250*sample_buffer[nsamp-1]) {
Serial.println("***Time too short, double TIMESTEP");
}
}
delay(1);
}
The sketch follows the usual format, where we first declare a number of variables in which
npts is the maximum number of points that can be acquired. The data samples are stored
in the sample buffer once every timestep milliseconds. We also declare the resistance R
of the resistor used to discharge the capacitor. Then we include the MsTimer2.h header
file and library, which provides the functionality to repeatedly call a function without user
intervention. We use MsTimer2.h because the Ticker.h library we used on the NodeMCU
in the previous chapter is unavailable for the UNO. The MsTimer2.h listens to an internal
clock and then calls the function after the specified time interval has elapsed. The function it
calls is normally referred to as the interrupt-service routine, and it is called timer action()
194  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

in our sketch. This function is called every timestep milliseconds, and first reads analog
pin A0, then stores the value in the sample buffer and increments isamp. If isamp reaches
nsamp, the timer is stopped to prevent the acquisition of further data points. Then isamp
is set to zero to prepare for the next acquisition and the variable sample buffer ready is
set to 1 to indicate that an acquisition is complete and data is ready for further processing.
Technically we can process the data within the timer action function, but normally this
routine should contain only time-critical actions, such as acquiring samples, and should
be kept as compact as possible, because it is called asynchronously to all other processing
and we should avoid excessive disturbances. The way to handle this problem is to flag
that the acquisition is ready, and in the main program, check for this flag and perform
the postprocessing. In this way, the time-critical and the slower processes are efficiently
decoupled. The linfit() function takes the acquired data as well as the number of data
points as input parameters, internally performs the least-squares fit to the logarithm of the
data points, and returns the slope of the fit. The discussion of the algorithm is somewhat
technical, and is deferred to Appendix B.
The setup() function initializes the serial line, declares digital pin D2 as output, and
sets it to HIGH to start charging the capacitor. The loop() function uses the standard
query-response protocol, and a capacitance measurement starts with the command CAP?
nnn, where nnn is the desired number of measurement points; per default the maximum
number npts is acquired. Then digital pin D2 is placed in high-impedance input mode.
We also ensure that the internal pull-up resistor is disabled, and start the timer process
with the MsTimer::set() function. It takes the time between samples and the function
to call as argument, and then we can start the timer. Before leaving this subroutine, we
take the first data sample. The command WF? can be used to retrieve the waveform data
of the last acquisition. This is useful for crosschecking the fitting in, for example, octave.
The remaining commands are used to set and read the TIMESTEP or the RESISTOR values.
Once the handling of the commands from the serial line is complete, we test whether the
variable sample buffer ready is set, which indicates that an acquisition is complete and
we can start postprocessing the raw data. First we reset the sample buffer ready to zero,
to prevent repeated calls, and then start charging the capacitor again by configuring pin D2
as output, and wait a little while. Next we calculate the slope from the samples. Since
the slope is inversely proportional to the time constant τ = RC, we can solve this for
the capacitance C with the equation defining the capacitance. The factor 1e6 causes the
displayed value to be shown in µF. Finally, we check whether the exponential decay of the
voltage actually extends over a sufficiently long time, and display a warning if that is not
the case.
We can measure capacitances by sending commands from other programs that communi-
cate over the serial line. From octave we use the following program to initiate a measurement,
and plot the data as shown in Figure 11.2.

% capacitance_plot.m, V. Ziemann, 170630

% pkg load instrument-control
close all; clear all
s=serial(’/dev/ttyACM0’,9600);
sleep(2);
srl_flush(s);
srl_write(s,"CAP? 100");
sleep(1);
reply=serialReadline(s);
capacitance=str2double(reply(4:end));
 Example: Capacitance Measurement  195

Figure 11.2 The waveform of the voltage on the capacitor as a function of time,
while a 2.2 µF Tantal capacitor discharges through a 33 kΩ resistor on a linear scale
(above) and on a logarithmic scale (below).

srl_write(s,"TIMESTEP?");
reply=serialReadline(s);
if reply(1:8)=="TIMESTEP"
timestep=str2double(reply(9:end));
else
disp(reply)
close(s);
return
end
srl_write(s,"WF?");
reply=serialReadline(s);
nsteps=str2num(reply(3:end));
xx=zeros(nsteps,1); yy=xx;
for k=1:nsteps
xx(k)=k*timestep*1e-3; % in seconds
yy(k)=str2double(serialReadline(s))*5/1023; % in Volt
end
close(s)
subplot(2,1,1); plot(xx,yy);
xlabel("Time [s]"); ylabel("Voltage [V]");
title(["Capacitance = " num2str(capacitance) " uF"]);
subplot(2,1,2); semilogy(xx,yy);
196  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

xlabel("Time [s]"); ylabel("log(Voltage)");

print(’capacitance.png’,’-S1000,700’)
We need to keep in mind that we need to load the instrument-control toolbox in order to
use the serial communication functionality and clear the workspace, before actually opening
the serial line. First we remove any remaining characters from the input queue, and initiate
a measurement with the CAP? command. After a short waiting period, we read the reply
with the serialReadline() function, which is tailored after the queryResponse() function
from Section 5.5.4. It is reproduced here.

% get response up to termination character

function out=serialReadline(dev,term_char)
if (nargin==1) term_char=10; end % defaults to LF=0x0A
i=1;
int_array=uint8(1);
while true % loop forever
val=srl_read(dev,1); % and read one byte
if (val==term_char) break; end % until term_char appears
int_array(i)=val; % stuff byte in output
i=i+1;
end
out=char(int_array); % convert to characters
This function reads characters one at a time until an end-of-line character is found, and
then returns the received characters to the calling program, where the numerical value
is extracted. For the plot, we need to know the TIMESTEP used in the measurement, and
request it from the UNO with the TIMESTEP? command. If the reply starts with TIMESTEP,
all is well, and we extract the numerical value. If we retrieve something else from the serial
input queue, the time window for the measurements was too short, and the measurement
is invalid. In that case, the program closes the serial port and returns control to the octave
command prompt. If, on the other hand, we obtain the numerical value of the time step,
we request the waveform data with the WF? command. The first returned line contains the
number of points acquired. The following lines contain the data points. Each new line is
retrieved with a call to serialReadline() until all data points are copied to the array
yy. Finally, we plot the curves with linear and logarithmic vertical scale. The latter shows
a clear linear dependence, which indicates the exponential dependence of the voltage on
time with a well-defined time constant. In the last line, we produce a png file shown in
Figure 11.2 that can be used to document the work.

QUESTIONS AND PROJECT IDEAS

1. Measure the resistance of an unknown resistor by comparing to a known resistor in a
voltage-divider.
2. Ask a friend to build a pulse generator that produces pulses with variable width;
for example, a 10 ms pulse every 200 ms. Then use a second UNO to measure the
pulse pattern, by sampling an IO-pin using the MsTimer2.h library. Display the pulse
pattern on the host computer.
 CHAPTER 12

Example: Profile of a Laser

Beam

In this example we measure the transverse profile of the laser beam from a laser pointer by
carefully moving an obstacle across the beam and observing the change of the signal from
a photoresistor. Figure 12.1 illustrates the method. We use a laser module from a sensor
kit that is mounted on a breadboard and exposes two wires, one for ground and one for
the positive supply voltage. In order to adjust the intensity of the laser, we use pulse-width
modulation of the positive supply voltage. The sensor is a voltage divider consisting of a
light sensitive resistor (LDR) and a 10 kΩ resistor. The obstacle is made of a piece of black
plastic. As the mover, we use the frame salvaged from an old CD-ROM drive, which moves
a small wagon with a stepper motor back and forth over a distance of about 40 mm, which is
approximately the width of the readable area of a CD. The stepper motor drives a spindle,
and that pulls and pushes the wagon with good precision.
Figure 12.2 shows the salvaged frame from the CD drive. The frame has plenty of holes,
which are convenient to add screws that in turn are used to attach laser, sensor, and obstacle.
The laser on its small breakout board is located at the top right with the stepper motor
beneath and hidden from view, but the spindle is clearly visible, running right to left near
the top of the frame. There it engages the white plastic part that is attached to the wagon
and moves it, when the motor turns. On the wagon there is a black piece of plastic visible
that intercepts the laser beam and prevents the laser from hitting the LDR on the lower
right on the frame. The LDR is also mounted on a small breakout board that is part of

V+

10k Move
PWM
Signal

LDR Laser

Figure 12.1 The schematic setup to measure the beam size of a laser pointer.

197
198  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 12.2The chassis from the CD-ROM drive with the laser mounted on the top
right and the photoresistor on the bottom right. The black obstacle is mounted on
the carriage that can be moved via the spindle on the top by a small stepper motor
that is located below the laser.

a sensor kit, and we use it because the mounting holes make assembly and attaching the
LDR to the screws on the frame easy. As controller we use an Arduino UNO that drives the
motor with an L293D H-bridge driver, uses pulse-width modulation on pin D9 to control
the laser power, and reads the signal from the voltage divider with the LDR on analog
input A0. Figure 12.3 shows the schematics. We use an external power supply to provide
the voltage for the stepper motor that is connected to the terminals labeled PA,. . . ,PD.
The next task is to program the UNO to control the power of the laser, move the stepper
motor, and read the sensor, all in a well-orchestrated fashion. We base the sketch on the
program to drive a stepper motor with an H bridge from Section 4.5.4 and augment the
code with sections for the laser and the sensor, as shown below.
// Laser profile measurement, V. Ziemann, 170628
char line[30];
int settle_time=2, stepcounter=0;
int laser_power=10;
const int PA=2,PB=3,PC=4,PD=5,ENABLE=6;
const int LASER=9;
void set_coils_fullstep(int istep) { //.........set_coils
bool patA[]={1,1,0,0};
int pat_length=4;
int ii;
istep=istep % pat_length;
if (istep < 0) istep+=pat_length;
digitalWrite(PA,patA[istep]);
 Example: Profile of a Laser Beam  199

Figure 12.3The schematics of the circuit. The stepper motor of the frame is con-
nected to points labelled PA,. . . ,PD.

ii=(istep+2) % pat_length;
digitalWrite(PB,patA[ii]);
ii=(istep+3) % pat_length;
digitalWrite(PC,patA[ii]);
ii=(istep+1) % pat_length;
digitalWrite(PD,patA[ii]);
delay(settle_time);
}
void setup() { //...............................setup
Serial.begin (9600);
while (!Serial) {;}
pinMode(PA,OUTPUT);
pinMode(PB,OUTPUT);
pinMode(PC,OUTPUT);
pinMode(PD,OUTPUT);
pinMode(ENABLE,OUTPUT);
digitalWrite(ENABLE,HIGH);
analogWrite(LASER,laser_power);
}
void loop() { //................................loop
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,"FMOVE ")==line) {
int steps=(int)atof(&line[6]);
digitalWrite(ENABLE,HIGH);
if (steps > 0) {
for (int i=0;i<steps;i++) set_coils_fullstep(stepcounter++);
} else {
200  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

for (int i=0;i<abs(steps);i++) set_coils_fullstep(stepcounter--);

}
} else if (strstr(line,"STEPS?")==line) {
Serial.print("STEPS "); Serial.println(stepcounter);
} else if (strstr(line,"STEPS ")==line) {
stepcounter=(int)atof(&line[6]);
} else if (strstr(line,"WAIT?")==line) {
Serial.print("WAIT "); Serial.println(settle_time);
} else if (strstr(line,"WAIT ")==line) {
settle_time=(int)atof(&line[5]);
} else if (strstr(line,"DISABLE")==line) {
digitalWrite(ENABLE,LOW);
} else if (strstr(line,"ENABLE")==line) {
digitalWrite(ENABLE,HIGH);
} else if (strstr(line,"LDR?")==line) {
Serial.print("LDR "); Serial.println(analogRead(A0));
} else if (strstr(line,"POWER?")==line) {
Serial.print("POWER "); Serial.println(laser_power);
} else if (strstr(line,"POWER ")==line) {
laser_power=(int)atof(&line[6]);
Serial.print("POWER "); Serial.println(laser_power);
analogWrite(LASER,laser_power);
} else if (strstr(line,"FSCAN")==line) {
int steps=(int)atof(&line[6]);
digitalWrite(ENABLE,HIGH);
delay(100);
for (int i=0;i<abs(steps);i++) set_coils_fullstep(stepcounter++);
delay(100);
for (int i=0;i<abs(steps);i++) {
set_coils_fullstep(stepcounter--);
delay(50);
unsigned long sum=0;
for (int k=1; k<10; k++) {sum+=analogRead(A0); delay(10);}
Serial.println(sum);
}
digitalWrite(ENABLE,LOW);
} else if (strstr(line,"CALIBRATE")==line) {
for (int power=0; power<256; power++) {
analogWrite(LASER,power);
delay(100);
unsigned long sum=0;
for (int k=1; k<10; k++) {sum+=analogRead(A0); delay(10);}
Serial.println(sum);
}
analogWrite(LASER,10);
} else {
Serial.println("unknown");
}
}
 Example: Profile of a Laser Beam  201

}
At the top of the sketch, a number of variables are declared for the time to wait between step-
per motor steps and the initial setting of the pulse-width modulation for the laser power.
Then the used pins for the stepper motor and the LASER are declared before defining the
set coils fullstep() function, which is identical to the full-step version we used earlier
in Section 4.5.4. We use full-step mode because the step resolution is sufficient, and in
half-step mode there is a small visible disturbance on the sensor from the different currents
drawn on alternate steps in half-step mode, where exciting one coil and two coils take turns.
In the setup() function, we initialize the serial communication, declare the mode of the
pins used for the motor, and initialize the laser power to be 10 (of up to 255). The loop()
function is built in much the same way as before. It expects to receive single-line commands
on the serial line. As in Section 4.5.4, the FMOVE command moves the motor by the specified
number of steps. Then there are commands to set and return the STEPS and the time to
WAIT or ENABLE the motor driver. The command LDR? reads the sensor value from analog
pin A0, and the POWER command sets and reads the laser excitation. Finally we encounter
the command FSCAN nnn where nnn is the scan range retrieved by reading the rest of line
with the atof() function we used before. Here we assume that the scanner actually blocks
the laser, and first needs to be retracted before collecting data. Inside the case block it first
reports to the serial line what it is about to do, enables the motor driver, and moves the
obstacle out of the way. Then comes a for loop, where the motor performs individual steps
and within each step, takes ten measurements of the LDR on analog pin A0 and adds them
up. At the end of each step the measurement value is written to the serial line. Once the
number of steps is complete, the motor driver is disabled. The last command implemented
is CALIBRATE, which measures the sensor response as a function of the laser power. The
latter we may later use to convert the raw sensor readings to a linear intensity scale should
the need arise.
As a first test of the system, we open the Serial Monitor in the Arduino IDE, make
sure that the baud rate is set to the one specified in the setup() function, 9600 baud in
our case, and query the LDR by sending LDR? to the UNO. The response has the format
LDR nnn, where nnn is the raw value from the call to the analogRead() function. Then we
query the laser power with POWER? and set it to a new value with POWER nnn, where nnn is
a value between 0 and 255. The brightness of the laser spot should change accordingly. Next
we move the stepper motor a few steps, for example, with FSCAN 20, and move it back to
the starting position with FSCAN -20. The wagon with the obstacle should move back and
forth. As a last point, we execute FSCAN 60, which will move the obstacle by 60 steps and
then returns to the starting position step by step, while reading the sensor and reporting
the value to the serial line. If all these initial tests complete satisfactorily, we progress to
automatizing the process, and for that we chose octave.
Basically, we hook up the Arduino UNO to the Raspi, write an octave script to open a
serial line, send commands to the UNO, and receive the response, very similar to what we
did in previous examples. Before writing the octave script, we need to calibrate the motion
of the obstacle in order to be able to show the width in mm instead of steps. In moving
the wagon a large distance, I use 250 full steps, and, taking a photo of the setup with a
ruler lying next to it before and after moving it, is easy to calculate the calibration constant
xscale in mm/step. In my case, 250 steps moved the wagon by 39.2 mm, which explains
the value for xscale near the top of the following script.
% scanplot2.m, V. Ziemann, 170628
close all;
clear all
202  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

xscale=0.161; % mm/fullstep
s=serial(’/dev/ttyACM0’,9600); % set device correctly
sleep(3);
srl_flush(s); % flush the input queue
nsteps=60;
srl_write(s,"FSCAN 60");
sleep(5);
xx=zeros(1,nsteps);
yy=xx;
for i=1:nsteps
xx(i)=i*xscale;
yy(i)=str2double(serialReadline(s));
end
close(s)
yy=smooth3(yy); % smooth data
subplot(2,1,1);
plot(xx,yy); % raw sensor data
ylabel(’arb. units’);
subplot(2,1,2);
dy=yy(2:end)-yy(1:end-1); % derivative
if (yy(1) > yy(end)) dy=-dy; end
plot(xx(1:end-1),dy);
xlabel(’x [mm]’);
ylabel(’arb. units’);
title([’FWHM = ’, num2str(xscale*fwhm(dy),"%5.2f"), ’ mm’])
print(’laser_profile.png’,’-S1000,700’)
The rest of the script should be familiar by now. After opening the serial line and flushing
whatever characters are present, we define the number of steps and write FSCAN 60 to the
UNO. Then we wait a short time and start retrieving the values from the serial line with
the serialReadline() function we already encountered in the previous chapter. It reads
characters from the serial line until the termination character is encountered, and returns
the obtained characters as a character string. But back to the main script. While retrieving
data and copying it to the array yy, we also fill the array xx with the properly scaled values
for the horizontal axis. Once the loop completes, we close the serial device.
After all data are available available within the script, we start processing it. Since we
will calculate the derivative of the raw data later, which is a process that is very sensitive
to noise, we weakly smooth the sensor values in array yy with the smooth() function. It
averages three consecutive values and replaces the central value with the average. The octave
script for smooth3() is
% smooth three consecutive points, V. Ziemann, 170628
function out=smooth3(y);
y=[y(1),y(1:end),y(end)]; % add extremities
f=ones(1,3)/3.0; % filter function
out=conv(y,f); % convolute
out=out(3:end-2); % ensure same length
It first creates a new array with the extreme points doubled up to avoid ugly artefacts at
the end points. In the next line we create the filter function f that consists of three values
1/3 and use it to convolute the data with the filter. Since the convolution creates an output
 Example: Profile of a Laser Beam  203

Figure 12.4The raw sensor value as a function of the position of the obstacle and
the derived laser beam profile, which shows a moderate asymmetry.

array that is longer than the original, we remove the extreme points before returning the
smoothed array as the value out. Once these initial preparations are complete, we plot the
smoothed raw values and add a label to the vertical axis. Then we calculate the difference
between consecutive values, which is motivated by the fact that in each step a little of the
laser beam is obscured and the change in remaining intensity is proportional to the intensity
in the small band that the obstacle transverses during that step. The result is a signal that
is proportional to the transverse intensity profile of the laser beam. For the resulting plots,
see Figure 12.4. In the title statement, we add a title to the lower plot with the full width
at half maximum (FWHM) of the laser profile, which is about 0.8 mm. Finally, we produce
an image file, which is precisely how we prepared Figure 12.4.
In this example, we use the FWHM rather than the standard deviation of the profile,
because the FWHM is a rather robust measure and works more reliably with profiles that
are non-Gaussian, asymmetric, and have tails that are moderately populated. The standard
deviation is heavily biassed by the latter. The script for the FWHM is straightforward and
is reproduced here.
% FWHM, V. Ziemann, 170628
function fwhm=fwhm(data)
N=length(data);
xmax = -1e30;
xmin=min(data);
imax=-1;
for i=1:N
if (data(i) > xmax)
xmax=data(i);
imax=i;
204  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

end
end
ileft=imax;
while (data(ileft) > (xmax+xmin)/2 && ileft>1)
ileft=ileft-1;
end
iright=imax;
while (data(iright) > (xmax+xmin)/2 && iright<N-1)
iright=iright+1;
end
fwhm=iright-ileft-1;
In the fwhm() function we first locate the minimum and maximum of the data points as
well as the location of the maximum. Then we search towards the left until the value is less
than halfway between maximum and minimum. This results in the location ileft of that
point. Then we repeat the process on the right-hand side and return the difference of the
right and left halfway points as the FWHM. The minus one in the last equation accounts
for the fact that both searches overshoot before terminating the respective while loops.
This is partially compensated by subtracting one from the final result.
In this example, we built a laser-beam profile monitor from scratch using a small stepper
motor and an LDR to illustrate the basic steps and how to combine sensors and actuators
to built an integrated system that is controlled from octave or any other language that
communicates over serial devices, such as EPICS, LabView, or C. The system is far from
perfect and many limitations and improvements come to mind. For example: Is diffraction
on the edge of the obstacle a limitation? Is the step size adequate or should we use a better,
microstepping motor-driver? Does the profile change with different laser intensities? Do we
need to correct for intensity-related nonlinearities of the sensor system? This is actually the
purpose of the CALIBRATE command to provide the base information. Does the profile show
the same asymmetry when we scan from the other side? We may also want to know the
vertical beam size and any cross-plane correlations to determine some basic aberrations. Is
the profile the same after swapping position of LDR and 10 kΩ resistor? We might want to
add a second motor moving the obstacle longitudinally in the direction of the laser beam.
Adding a lens and measuring the beam profile at several places allows us to determine the
M 2 of the laser beam, which is a measure of the laser quality and is unity for a diffraction-
limited beam. But these exercises we leave for the interested reader to do at home, while
we move on to build a robot that detects fire.
 CHAPTER 13

Example: Fire-Seeking Robot

In this final example, we discuss a small robot that detects heat radiation from a fire, moves
towards the heat source, and starts beeping once it reaches the source of the radiation. This
is, of course, a very simple prototype of an autonomous fire extinguisher. We require the
robot to either move autonomously, or to be manually controlled by a remote controller.
Despite the simple description, the project is rather ambitious and requires us to solve the
following subtasks:
1. detect the heat source and its location;
2. control the speed and direction of motors;
3. detect collisions, preferably before they happen;

4. sense joysticks and switches on the remote control;

5. send messages between remote controller and robot.
We base the hardware of the robot on a ready-made chassis with two DC motors and an on-
board battery pack with five AAA cells providing up to 7.5 V. Onto that chassis we mount
a full-size breadboard to house a NodeMCU controller, an H-bridge motor-driver, and some
linear voltage converters to provide 5 and 3.3 V. To provide additional IO-pins we add a
second Arduino that acts as a slave to the NodeMCU. On the robot chassis is a dedicated
place for a model-servo onto which we mount a very small breadboard for the BPX38 IR
phototransistors we use as flame sensors, and the HC-SR04 distance sensor. Figure 13.1
shows the chassis with the servo and the two breadboards already mounted. Mounting the
sensors on the movable breadboard makes it possible to point them in different directions
by turning the model-servo. We configure the NodeMCU as an access point that spans its
own WLAN network. We also start a server to receive commands from the remote control.
Apart from listening to the remote on the controller, we implement a simple state machine,
to which we can hand the control of the robot, which subsequently runs autonomously.
The remote control is based on a second NodeMCU. It connects to the WLAN spanned
by the first NodeMCU on the robot and to the server running on it. On the remote controller
we connect a MCP3408 8-channel ADC via SPI to the NodeMCU, and set the input voltages
of the ADC channels with two joysticks, two potentiometers, several voltage-dividers, and
two switches. The ADC readings are then interpreted as motor speed and model-servo
setting, and transmitted to the robot. The voltage on one ADC channel is set by a number
of switches and voltage dividers, which allows us to use a larger number of switches than

205
206  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 13.1 The chassis of the robot with two breadboards. The smaller one can
be turned by operating a model-servo. Mounted on the larger breadboard are a
NodeMCU on the right and a second chip on the left, a bare ATmega328 that was
initially tested and later replaced by an Arduino NANO.

available digital IO pins, albeit at the expense of being able to sense only one button at a
time.
We show a simplified version of the remote controller with less than all possible channels
set up on a breadboard in Figure 13.2, where we see the NodeMCU on the right and the
MCP3208 ADC on the left. The MCP3208 has the same pin assignment as the MCP3304
from Section 4.4.4, and is connected to the SPI port of the NodeMCU with CLK, MISO,
MOSI, and CS lines, connected to D5, D6, D7, and D8, respectively. A joystick, visible
immediately to the left of the breadboard, is connected to the power rails and the wiper to
ADC channels A0 and A1. Likewise, the potentiometer, visible on the far left, is connected
to ADC channel A5. Two buttons, visible near the bottom of Figure 13.2, connect voltage
dividers consisting of a 10 kΩ resistor to the positive supply voltage, and a 68, and 47 kΩ
resistor, respectively. By using further resistors to ground with different values, it is possible
to detect more buttons—one at a time. Pressing the buttons supplies a definite voltage level
to ADC channel A7. Finally, there is a switch connected to IO pin D4. To clarify the wiring,
we also show the schematic corresponding to the breadboard layout in Figure 13.3.
The sketch running on the remote controller that uses the hardware described in the
previous paragraph is the following. Since the sketch is fairly long, we show and describe
the preparatory sections first.
 Example: Fire-Seeking Robot  207

Figure 13.2 Simplified setup of the remote controller on a breadboard.

Figure 13.3 The schematic of the remote controller.

208  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

// RCsenderUDP, V. Ziemann, 170705

const char* ssid = "FireBot";
const char* password = "..........";
const char* host = "192.168.4.1";
const int port=1137;
#include <ESP8266WiFi.h>
#include <WiFiUdp.h>
WiFiUDP client;
int adclast[8]={0,0,0,0,0,0,0,0},D4last=HIGH;
int adccalib[8];
//...................................................ADC
#include <SPI.h>
#define CS 15
int mcp3208_read_adc(uint8_t channel) { // 8 single ended
int adcvalue=0, b1=0, hi=0, lo=0, reading;
digitalWrite (CS, LOW);
byte commandbits = B00001100; // Startbit+(single ended=1)
commandbits |= ((channel>>1) & 0x03);
SPI.transfer(commandbits);
commandbits=(channel & 0x01) << 7;
b1 = SPI.transfer(commandbits);
hi = b1 & B00011111;
lo = SPI.transfer(0x00); // input is don’t care
digitalWrite(CS, HIGH);
reading = (hi << 7) + (lo >> 1);
return reading;
}
void send_string(char line[]) { //............send_string
client.beginPacket(host,port);
client.write(line);
client.endPacket();
}
void setup() { //..................................setup
pinMode(LED_BUILTIN,OUTPUT);
digitalWrite(LED_BUILTIN,LOW);
pinMode(CS,OUTPUT);
digitalWrite(CS,HIGH);
SPI.begin();
SPI.setFrequency(100000);
SPI.setBitOrder(MSBFIRST);
SPI.setDataMode(SPI_MODE0);
pinMode(LED_BUILTIN,OUTPUT);
digitalWrite(LED_BUILTIN,LOW);
Serial.begin(115200);
pinMode(D3,INPUT_PULLUP);
pinMode(D4,INPUT_PULLUP);
delay(1000);
WiFi.mode(WIFI_STA); // needed for reliable communication
WiFi.begin(ssid,password);
 Example: Fire-Seeking Robot  209

while (WiFi.status() != WL_CONNECTED) {

Serial.print("."); delay(500);
}
Serial.print("\nConnected to "); Serial.print(ssid);
Serial.print(" with IP address: ");
Serial.println(WiFi.localIP());
client.begin(port);
for (int k=0;k<8;k++) {
adccalib[k]=mcp3208_read_adc(k);
}
digitalWrite(LED_BUILTIN,HIGH);
}
At the top of the sketch, the WLAN name and passphrase are defined, as well as the
robot’s IP number, which defaults to 192.168.4.1 for the WLAN spanned by the robot.
Then we include WiFi libraries and declare the client, which is the socket that connects
to the server. Note that we use a UDP connection instead of a TCP, because the latter uses
extensive handshaking between server and client and this makes the response of the robot to
remote commands very sluggish. UDP connections do not use handshaking, which makes the
connection less reliable, but much more responsive. Since we continuously send commands to
the robot, losing a single command is not very serious. Next we declare a number of variables
that we discuss later, and define the mcp3208 read adc() function to read a single channel
from the ADC. It is very similar to the one we discussed in Section 4.4.4, except that we
request single-ended ADC measurements, and we receive 12 data bits instead of 13. The
function receives the requested channel as input parameter and returns the 12-bit ADC
reading. The send string() function encapsulates the construction and sending of UDP
packages. In the setup() function, we configure a number of IO pins, and initialize the
SPI communication and the serial line. Then we define the WiFi.mode to be WIFI STA or a
client, rather than an access point, the latter otherwise being the default, and connect to
the WLAN. Once connected, we write the received IP number to the serial line and connect
to the server on the robot with the call to the client.begin() function. Finally, all ADC
channels are read and the values saved in the array adccalib[], which is used to calibrate
the center positions of the joysticks.
After the preparatory sections, we are ready to show and discuss the main part of the
program, the loop() function.
void loop() { //..................................loop
char line[30];
int adc[8];
for (int k=0;k<8;k++) {adc[k]=mcp3208_read_adc(k);}
if ((abs(adc[0]-adclast[0]) > 16) || (abs(adc[1]-adclast[1]) > 16)) {
adclast[0]=adc[0]; adclast[1]=adc[1];
int val0=(adc[0]-adccalib[0])*1023.0/2048.0;
int val1=(adc[1]-adccalib[1])*1023.0/2048.0;
sprintf(line,"RSPEED %d",(int)(val0+0.5*val1)); send_string(line);
sprintf(line,"LSPEED %d",(int)(val0-0.5*val1)); send_string(line);
}
if (abs(adc[5]-adclast[5]) > 16) {
adclast[5]=adc[5];
int val=adc[5]*180.0/4095;
sprintf(line,"SERVO %d",val); send_string(line);
210  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

}
if (abs(adc[7]-adclast[7]) > 5) { // check the buttons
adclast[7]=adc[7];
if (adc[7] < 1000) {
Serial.println("Red button right pressed");
sprintf(line,"FINDFIRE 1"); send_string(line);
} else if (adc[7] < 1700) {
Serial.println("Blue button right pressed");
sprintf(line,"RANGE?"); send_string(line);
} else if (adc[7] < 2250) {
Serial.println("Joystick button right pressed");
sprintf(line,"NEXTEVENT 0"); send_string(line);
} else if (adc[7] < 3580) {
Serial.println("Joystick button left pressed");
sprintf(line,"NEXTEVENT 1"); send_string(line);
} else if (adc[7] < 3660) {
Serial.println("Blue button left pressed");
sprintf(line,"BEEP 1000"); send_string(line);
Serial.println(line);
} else if (adc[7] < 3750) {
Serial.println("Red button left pressed");
sprintf(line,"NEXTEVENT 3"); send_string(line);
}
}
if (digitalRead(D4) != D4last) {
D4last=digitalRead(D4);
sprintf(line,"D0 %d",D4last); send_string(line);
}
yield();
int packetsize=client.parsePacket();
if (packetsize) {
char line[30];
int len=client.read(line,30); line[len]=’\0’;
Serial.print("Message:"); Serial.println(line);
}
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
send_string(line);
}
}
In the loop() function, we first read all ADC channels and then test whether channels A0
and A1 have changed significantly with respect to the last time; the values are saved in
the array adclast[]. This construction is useful, because it prevents continuous sending of
commands, and only does that if a value has changed. If that is the case, the difference of
the current ADC reading with respect to the calibration value is stored in variables val0
and val1. The first value is interpreted as the desired speed of the motors, and the second
value as the direction. Thus we send commands to set val0+0.5*val1 as speed to one motor
and val0-0.5*val1 to the other. Next the reading of ADC channel A5 is scaled to a value
between 0 and 180 and transmitted to set the model-servo on the robot. ADC channel A7 is
 Example: Fire-Seeking Robot  211

used to interpret the buttons pressed. In the current realization, six buttons are connected,
and pressing them sends a number of different commands to the robot. Testing for ADC
values progresses from smaller to larger values, such that the buttons are prioritized in a
natural way. If a button connected to a smaller resistor is pressed, all buttons connected to
higher-valued resistors are ignored. The details of what action the commands cause on the
robot when a button is pressed we explain later, when we discuss the software running on the
robot. The call to the yield() function ensures that all background processes for WLAN and
serial communication can complete pending tasks. The call to the client.parsePacket()
function checks whether a message from the robot has arrived. Here we only copy the
received message to the serial line. But it is easy to envision other ways to handle this; for
example, by showing them on an LCD display connected via a I2C interface to pins D1
and D2 on the NodeMCU controller. These pins are not used in the present circuit and are
available for expansions. Finally, we check whether anything has arrived on the serial line
and pass it on to the robot. The direct two-way communication with the robot is a very
convenient way to debug the system, by sending commands from the serial console if the
remote controller is connected via USB cable to the host computer.
And this brings us to the electronics on the robot. In Figure 13.4 we show the circuit
diagram to implement the functionality discussed earlier in this chapter. The central com-
ponent is the NodeMCU microcontroller visible on the right of the solderless breadboard,
with the L293D H-bridge motor driver and adjacent 7805 linear voltage regulator to its left.
The voltage regulator receives power from the battery, whose negative pole is connected to
system ground and the positive voltage to the input pin of the 7805. The positive battery
voltage is also routed to the motor and logic power pins of the L293D motor driver. The
5 V output voltage of the 7805 voltage regulator is connected to the lower power rails from
where it provides power to the Vin pin of the NodeMCU at its bottom right. The 5 V power
is further routed to the model-servo and the HC-SR04 sonar sensor visible above the large
breadboard. The 3.3 V voltage regulator on the NodeMCU provides power to the upper
power rail, which carries 3.3 V and is routed to the respective circuits. The respective power
rails have large electrolytic capacitors of 470 µF connected, to buffer intermittent voltage
requirements. Special care is needed to ensure that the three different voltages are routed
correctly. Higher than permissible voltages can destroy some of the integrated circuits. The
positive voltage from the battery only goes to the 7805 and L293D, and 5 V power the
NodeMCU via its Vin pin, the servo, and the sonar. To facilitate the correct wiring, color
images are made available on this book’s web site.
Motor 1 is connected to the pins on the lower side of the L293D H-bridge motor driver,
and the controlling input pins are connected to IO pins D3 and D4 on the NodeMCU.
Likewise, motor 2 is connected to the upper half of the L293D, and the corresponding input
pins are routed to pins D5 and D6 on the NodeMCU. The enable pins of the motor driver
are permanently wired to 3.3 V. The model-servo is connected to the 5 V power rail, and
its control wire connects to pin D7 on the NodeMCU, while the buzzer connects to pin D8.
The HC-SR04 distance sonar is powered from the 5 V rail. It is triggered by a pulse from
NodeMCU pin D1 and the echo is received on pin D2. Since the NodeMCU only operates
on 3.3 V, we need to use a voltage divider made of a 10 kΩ and a 22 kΩ resistor to step
the 5 V echo signal from the HC-SR04 to approximately 3.3 V. On the robot, the distance
sensor is actually mounted on the small breadboard visible on the top left in Figure 13.4,
which, in turn, is mounted on the movable axis of the model-servo that permits it to scan
the surroundings for obstacles. Initially only the distance sensor and the middle BPX38
phototransistor with a pull-up resistor to the 3.3 V rail are mounted on the small breadboard
to scan for both obstacles and sources of infrared radiation. The BPX38 is most sensitive
212  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Figure 13.4 The electronics circuit of the robot (color version available online).
 Example: Fire-Seeking Robot  213

Figure 13.5 The schematic of the robot electronics.

around a photon wavelength of 880 nm, which is in the infrared part of the electromagnetic
spectrum that is emitted as heat radiation by, for example, a fire or an old-fashioned light
bulb. The emitter of the BPX38 is connected to ground, and the collector connects to the
analog input pin A0 of the NodeMCU and via a 10 kΩ resistor to the 3.3 V power rail. In
this configuration we need to continuously scan the IR and distance sensor with the servo
in order to find the fire, which is a bit cumbersome. If we had two IR sensors to read out
simultaneously, we could compare excitation and have instantaneous information about the
location of the fire, without scanning with the servo. Therefore, we will add two extra IR
phototransistors to the far left and far right on the small breadboard, wired in the same
way as the middle one.
But we do not have enough analog input terminals on the NodeMCU, and have used
almost all the IO pins. This significantly limits our ability add new functionality to the
robot. We therefore add an Arduino-NANO to the large breadboard and program it to
behave as a slave to the NodeMCU, and to communicate over the serial line. The wires that
connect the respective TX and RX pins are crossed and behave similar to a null-modem
cable. The NANO behaves almost like a UNO and is also programmed in the same way,
by selecting Arduino Nano from the Tools→Board menu in the Arduino IDE. The only
obviously visible difference is that the NANO has eight instead of six analog input pins,
while there are also 13 digital IO pins. Since we power the NANO with 5 V supplied to
the pin labeled “5 V”, all IO pins are operating on 5 V logic levels. In order not to damage
the NodeMCU that operates at 3.3 V, we use a reverse-biased diode in the connection from
the TX pin of the NANO to the RX pin of the NodeMCU. The diode blocks the 5 V from
reaching the NodeMCU, but if the TX is pulled low, the signal on the input pin of the
NodeMCU is also pulled low. The prototype circuit works with a normal switching 1N4148
diode, but ideally one should use a Schottky diode, which has a smaller voltage drop. Once
the essential communication between NodeMCU and the slave-NANO works, we connect
the two additional phototransistors from the small breadboard to analog input pins A0 and
A1 on the slave NANO. This completes the description of the hardware on the robot chassis,
and we can turn to programming the NodeMCU.
214  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

The code that runs on the robot is the following, again split in two blocks because it is
rather long. First we show and discuss the preparatory sections.
// RCreceiver, V. Ziemann, 170701
#define Max(a,b) ((a)>(b)?(a):(b))
#define Min(a,b) ((a)<(b)?(a):(b))
const char *ap_ssid = "FireBot";
const char *ap_password = "..........";
const int port=1137;
#include <ESP8266WiFi.h>
#include <WiFiUdp.h>
WiFiUDP server;
#include <Servo.h>
Servo myServo;
int servo_pos=90,servo_inc=5;
int right_sensor=-1,left_sensor=-1;
long lasttime=0,sleeptime=1000,nextevent=1;
void send_string(char line[]) { //.......send_string
server.beginPacket(server.remoteIP(),port);
server.write(line);
server.endPacket();
}
int range() { //..............................range
digitalWrite(D1,LOW);
delayMicroseconds(2);
digitalWrite(D1,HIGH);
delayMicroseconds(10);
digitalWrite(D1,LOW);
int val=(int)(0.017*pulseIn(D2,HIGH));
if (val<10) tone(D8,1000,200);
return val;
}
void motor_speed(int left, int right) { //...........motor_speed
left=Max(-1023,Min(1023,left));
analogWrite(D3,0);
analogWrite(D4,0);
if (left<0) {analogWrite(D3,abs(left));}
else {analogWrite(D4,abs(left));}
right=Max(-1023,Min(1023,right));
analogWrite(D5,0);
analogWrite(D6,0);
if (right<0) {analogWrite(D5,abs(right));}
else {analogWrite(D6,abs(right));}
}
void motor_stop() { //...................motor_stop
analogWrite(D3,0);
analogWrite(D4,0);
analogWrite(D5,0);
analogWrite(D6,0);
}
 Example: Fire-Seeking Robot  215

void setup() { //..............................setup

pinMode(LED_BUILTIN,OUTPUT);
digitalWrite(LED_BUILTIN,LOW);
pinMode(D1,OUTPUT); // HCSR04-TRIG
digitalWrite(D1,LOW);
pinMode(D2,INPUT); // HCSR04-ECHO
pinMode(D3,OUTPUT);
pinMode(D4,OUTPUT);
analogWrite(D3,0);
analogWrite(D4,0);
pinMode(D5,OUTPUT);
pinMode(D6,OUTPUT);
analogWrite(D5,0);
analogWrite(D6,0);
Serial.begin(38400);
WiFi.softAP(ap_ssid,ap_password);
IPAddress myIP = WiFi.softAPIP();
server.begin(port);
Serial.print("\nAccess point and server started at address: ");
Serial.print(myIP); Serial.print(" and port: "); Serial.println(port);
Serial.print("with SSID: "); Serial.println(ap_ssid);
pinMode(D7,OUTPUT); // D7=Servo, D8=Tone
myServo.attach(D7);
digitalWrite(LED_BUILTIN,HIGH);
tone(D8,880,500);
lasttime=millis();
}
This code runs on the NodeMCU on the robot, and first defines the Max() and Min()
functions to determine the larger and smaller of two input values. Then it defines the
name and passphrase of the WLAN as well as the used port number before including WiFi
libraries for UDP and defining the server. We also include support for model-servos and
declare one instance myServo, as well as variables needed for scanning the servo, reading
the IR phototransistors, and the state machine. We then declare the convenience function
send string() to send a UDP package back to the remote controller. The range() function
controls the HC-SR04 distance sensor in the way we discussed in Section 4.4.5. It ensures
that the trigger pin D1 is LOW before pulling it HIGH for 10 µs, and then waits for the echo to
arrive on pin D2 with the pulseIn() function. The factor 0.017 converts the duration of the
echo in microseconds to distance in cm. If an obstacle is closer than 10 cm, the buzzer briefly
beeps before returning the distance as the function value. The motor speed() function
receives the speed, with sign indicating the direction, for the motors as input values. First
it clamps the values to be within ±1023, turns both motors briefly off, and then, depending
on the sign, sets the pulse-width modulation period of one or the other control pin to the
desired value with a call to analogWrite(). It first handles one motor controlled by pins D3
and D4, and then the other motor, controlled by D5 and D6. The motor stop() function
turns all relevant IO pins off.
In the setup() function we mostly configure the IO pins as input or output according to
their purpose. The LED BUILTIN pin is actually D0 and controls an LED on the NodeMCU
circuit board, which is convenient for debugging. Pulling D0 LOW lights the LED, and pulling
it HIGH turns it off again. D1 and D2 are connected to the trigger and and echo pin of the
216  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

distance sensor, and D3 to D6 to control the motors. Then we initialize the serial line to
communicate at 38400 baud to accommodate the capabilities of the slave NANO before
starting to span the WLAN with the call to the WiFi.softAP() function, which config-
ures the NodeMCU as an access point with the supplied name and passphrase. The call to
WiFi.softAPIP() returns the IP number of the access point, normally 192.168.4.1. Next
we start the server to listen on the selected port for packets to arrive from the remote con-
troller. Finally, we attach a servo controller to pin D7, turn the LED off, sound a short tone,
and remember the elapsed time in the variable lasttime, which is needed for scheduling
the state machine; but more on that topic later.
Having declared all variables and defined all preparatory functions, we can continue to
discuss the loop() function of the sketch.
void loop() { //...................................loop
char line[30];
int packetsize=server.parsePacket();
if (packetsize) {
int len=server.read(line,30);
line[len]=’\0’;
if (strstr(line,"LSPEED ")==line) {
int val=(int)atof(&line[7]);
val=Max(-1023,Min(1023,val));
analogWrite(D3,0);
analogWrite(D4,0);
if (val<0) {analogWrite(D3,abs(val));}
else {analogWrite(D4,abs(val));}
} else if (strstr(line,"RSPEED ")==line) {
int val=(int)atof(&line[7]);
val=Max(-1023,Min(1023,val));
analogWrite(D5,0);
analogWrite(D6,0);
if (val<0) {analogWrite(D5,abs(val));}
else {analogWrite(D6,abs(val));}
} else if (strstr(line,":")==line) {
line[len]=’\n’; line[len+1]=’\0’;
Serial.println(line);
} else if (strstr(line,"D0 ")==line) {
int val=(int)atof(&line[3]);
if (val==0) {
digitalWrite(LED_BUILTIN,HIGH);
} else {
digitalWrite(LED_BUILTIN,LOW);
}
} else if (strstr(line,"SERVO ")==line) {
int val=(int)atof(&line[6]);
myServo.write(val);
} else if (strstr(line,"BEEP ")) {
int val=(int)atof(&line[5]);
Serial.print("BEEP val= "); Serial.println(val);
tone(D8,440,val);
} else if (strstr(line,"A0?")) {
 Example: Fire-Seeking Robot  217

int val=analogRead(A0);
Serial.print("A0 "); Serial.println(val);
sprintf(line,"A0 %d",val); send_string(line);
} else if (strstr(line,"RANGE?")) {
int val=range();
sprintf(line,"RANGE %d",val); send_string(line);
} else if (strstr(line,"SCANRANGE ")==line) {
int val=(int)atof(&line[10]);
int minval=2000,minpos=-1;
if (val>0) {
myServo.write(10);
delay(1000);
for (int k=10;k<170;k+=5) {
myServo.write(k); delay(200);
val=range();
if (val<minval) { minval=val; minpos=k;}
sprintf(line,"SCANRANGE %d %d",k,val); send_string(line);
}
myServo.write(minpos);
sprintf(line,"MINIMUM at %d",minpos); send_string(line);
} else {
myServo.write(90);
}
} else if (strstr(line,"FINDFIRE ")==line) {
int val=(int)atof(&line[9]);
int minval=2000,minpos=-1;
if (val>0) {
myServo.write(10);
delay(1000);
for (int k=10;k<170;k+=5) {
myServo.write(k); delay(200);
val=analogRead(A0);
if (val<minval) { minval=val; minpos=k;}
sprintf(line,"FINDFIRE %d %d",k,val); send_string(line);
}
myServo.write(minpos);
sprintf(line,"MINIMUM at %d",minpos); send_string(line);
} else {
myServo.write(90);
}
} else if (strstr(line,"NEXTEVENT ")==line) {
nextevent=(int)atof(&line[10]);
} else if (strstr(line,"SLEEPTIME ")==line) {
sleeptime=(int)atof(&line[10]);
} else {
Serial.println("unknown");
}
}
yield();
218  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

if (millis()>lasttime+sleeptime) { // next scheduled event

switch (nextevent) {
case 1: // determine range
sprintf(line,"RANGE %d",range()); send_string(line);
sprintf(line,"A0 %d",analogRead(A0)); send_string(line);
nextevent=1;
break;
case 2: // scan with servo
servo_pos+=servo_inc;
if (servo_pos>170) {servo_inc=-servo_inc;}
if (servo_pos<10) {servo_inc=-servo_inc;}
myServo.write(servo_pos);
nextevent=2;
case 3: // request direction sensors
if (range()<10) {
motor_stop();
nextevent=1;
} else {
Serial.println(":A0?\n:A1?");
sleeptime=50; nextevent=4;
}
break;
case 4: // read direction sensors and take action
if ((right_sensor>0) && (left_sensor>0)) { // new data
if ((left_sensor<250) || (right_sensor<250)) {
sprintf(line,"SENSORS %d %d",left_sensor,right_sensor);
send_string(line);
int val0=(int)(600+0.2*(right_sensor-left_sensor));
val0=Max(-1023,Min(1023,val0));
int val1=(int)(600-0.2*(right_sensor-left_sensor));
val1=Max(-1023,Min(1023,val1));
motor_speed(val0,val1);
} else {
motor_stop();
}
right_sensor=-1; left_sensor=-1;
sleeptime=1000; nextevent=3;
}
default:
break;
}
lasttime=millis();
}
yield();
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
send_string(line);
if (strstr(line,".A0 ")==line) {
right_sensor=(int)atof(&line[3]);
 Example: Fire-Seeking Robot  219

} else if (strstr(line,".A1 ")==line) {

left_sensor=(int)atof(&line[3]);
}
}
}
Here we first check whether a UDP packet has arrived from the remote controller with the
call to the server.parsePacket() function. If a packet of size packetsize is available, it
is read with server.read() and its length is determined. To avoid ugly output, we ensure
that the last character is a NULL character, to indicate the end of a string and start testing
what type of command has arrived. If it is LSPEED we interpret the rest of the line as the
speed value with the call to atof(), clamp the values to the acceptable range, and set the
speed of the motor, in the same way we do in the motor speed() function. If RSPEED is
received, the speed of the other motor is adjusted. If the line starts with a colon : it is
passed on to the serial line. In this way we can send commands to the slave NANO. If
the command starts with D0 we turn the built-in LED on and off. The SERVO command
sets the angle of the model-servo, and BEEP nnn makes a sound of 440 Hz for the duration
of nnn ms. The A0? command returns the analog value read from the analog pin on the
NodeMCU to the serial line and to the remote controller, and the RANGE? command likewise
returns the distance to an obstacle as determined by the HC-SR04 sensor. The following
command, SCANRANGE 1, starts the servo to move, which causes the distance sensor to
point towards different directions, while the HC-SR04 scans the distance to an obstacle,
and simultaneously records the direction and distance to the closest object. Finally, the
servo is moved to point towards the minimum distance, and returns the direction at which
the closest object is found. Calling SCANRANGE with argument 0 causes the servo to move to
its middle position. The FINDFIRE command performs the same action, but scans the middle
IR photodiode connected to analog pin A0 on the NodeMCU to determine the direction
of a heat source. Note that the phototransistor pulls the signal line towards ground, which
causes the voltage on pin A0 to approach zero, if a heat source is detected. The last two
commands, NEXTEVENT and SLEEPTIME, allow us to set variables related to the state machine
that governs the autonomous running of the robot. They are attached to buttons on the
remote controller and can be changed by pressing these buttons.
The next section of the code, following the yield() command, implements this very
simple state machine to operate asynchronously with all other actions that the NodeMCU
performs. First we test whether the current time exceeds the value for the next scheduled
event, lasttime+sleeptime, has elapsed. If that is the case, we branch according to the
value of the nextevent variable. If it is 1 we only determine the distance from the HC-SR04
sensor with the range() function and the reading of the middle phototransistor. Then we set
nextevent=1, which will repeat the same action after sleeptime has elapsed. If nextevent
is 2, we sweep the servo position back and forth across its range, one step at a time. The
next two event codes, 3 and 4, implement the autonomous motion of the robot towards a
heat source. If nextevent is 3, we first test whether an object is closer than 10 cm, stop
the motors, and branch to event code 1 discussed above. If there is no close obstacle, the
commands :A0? and :A1? are sent on the serial line. Here we employ the convention that
a command prepended with a colon is interpreted by the slave NANO, which in this case
is requested to read its analog pins A0 and A1, to which the two outer phototransistors
on the small breadboard are connected. Since we have two transistors, the difference of the
reading will provide information about the direction of the heat source. The command is
only dispatched in event code 3, but by setting the sleeptime to 50 ms and nextevent=4, we
will execute event code 4 about 50 ms later. The response from the slave NANO is recorded
220  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

asynchronously, and we discuss the sketch running on the NANO a little later. After 50 ms,
event code 4 is executed, and there we test whether the phototransistors detected a valid
signal, and whether one of the signals is sufficiently small to be interpreted as a heat source.
In that case, the sensor readings are sent to the remote controller, and the motor speed
of one motor is set to a constant, not-too-large value, here 600, plus a small contribution
proportional to the difference between the sensor readings. This small contribution is added
to the speed of one motor and subtracted from the other. In this way we implement a
very simple proportional control loop that feeds the error signal to the motors, to make the
robot turn towards the heat source. If no heat source is detected, we stop the motors. Before
leaving this part of the program, we set the variables right sensor and left sensor to an
invalid value to indicate that no new value is present. Finally, we make the state machine
sleep for 1000 ms and restart with event code 3 in order to repeat the process of reading the
sensors and feeding the difference of their reading to the motors. Before leaving this part of
the program, we update the lasttime variable in order to know when to handle the next
event code. After the call to yield(), we test whether any characters are available on the
serial line, and if the response starts with .A0 or .A1, we interpret the numerical value as the
right sensor or the left sensor reading. We point out that handling the communication
via UDP packets and on the serial line is handled asynchronously, and interleaved with the
state machine executing the events enumerated by the variable nextevent.
The serial line serves a dual purpose: First, it displays debugging information on the
host computer if that is connected, and second, it communicates with the slave NANO.
In order to distinguish the latter, we adapted the query-response protocol used earlier by
the convention that all communication to the slave NANO starts with a colon “:,” and
all communication from the slave starts with a period “.” such that those commands can
be easily filtered out from the serial line. The sketch that runs on the slave NANO is the
following.

// Slaveduino, V. Ziemann, 170723

void setup() { //.........................setup
Serial.begin(38400);
while (!Serial) {;}
pinMode(13,OUTPUT);
digitalWrite(13,LOW);
pinMode(8,INPUT_PULLUP);
}
void loop() { //...........................loop
char line[30];
if (Serial.available()) {
Serial.readStringUntil(’\n’).toCharArray(line,30);
if (strstr(line,":A0?")==line) {
Serial.print(".A0 "); Serial.println(analogRead(A0));
} else if (strstr(line,":A1?")==line) {
Serial.print(".A1 "); Serial.println(analogRead(A1));
} else if (strstr(line,":D13 ")==line) {
int val=(int)atof(&line[4]);
if (val==0) {digitalWrite(13,LOW);} else {digitalWrite(13,HIGH);}
} else if (strstr(line,":D8?")==line) {
Serial.print(".D8 "); Serial.println(digitalRead(8));
}
delay(3);
 Example: Fire-Seeking Robot  221

Figure 13.6 The operational robot from the back (left) and from the front (right).

}
}
This sketch follows earlier examples, and in the setup() function we configure the serial
line and the used pins to be OUTPUT or INPUT, in this case even with internal pull-up resistor
enabled. In the loop() function we test whether data is available on the serial line and then
test the different requests the slave is able to handle, namely responding to :A0?, :A1?,
:D13, and :D8?, which all start with a colon, while any other query is silently ignored. Note
also that any reply back to the NodeMCU via the serial line is prepended by a period. In
the code running on the NodeMCU we only react to :A0? and :A1? and their respective
response, but extending the code running on the slave NANO is easy to implement.
In Figure 13.6 we show an early prototype of the operational robot from the back and
from the front. On the left-hand image we see the larger breadboard with the NodeMCU
on the right. In the center of the breadboard, hidden behind the wires, is the L293D, and
next to it the voltage regulator. In this prototype, instead of NANO, we use an ATmega328
that is programmed in an Arduino UNO, removed from its socket, and inserted in the
breadboard on the robot. After adding a 16 MHz crystal and two 22 pF ballast capacitors,
it works equally well on the breadboard. On the right we see the front of the robot with the
small breadboard slightly turned to the side. On it the distance sensor and below it the three
phototransistors are visible. Owing to the breadboards, the wiring has a distinct “spaghetti-
flavor,” which makes the robot rather fragile, but also very convenient for developing and
testing new functionality.
At this point we can control our simple robot with the remote controller and also make
it follow a simple algorithm to find a heat source autonomously. We have to admit that
the performance of the prototype system is less than impressive; the heat source has to be
rather close to be detectable and the motion towards it is awkward. But our main purpose
222  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

is to show how to implement all the functionality in a simple prototype system. Building a
marketable system requires significantly more effort.
The hardware with one large breadboard directly mounted on the robot chassis and a
smaller one movable on the model-servo makes the system very versatile and extendable; for
example, connecting other sensors or making the robot follow other algorithms. We use a
slave NANO as an IO extender, but also other extenders are available, such as the MCP23017
that provides up to 16 digital IO pins, or the PCA9685 that controls up to 16 pulse-width
modulated outputs or model-servos. Both extenders are controlled via the I2C-port on the
NodeMCU and require two IO-pins only. The system is open for further experiments, and
to expand the system. Other improvements comprise fine-tuning the algorithms to make
the robot more robust, and whatever else come to mind.
Now that we have discussed the hardware and software for our projects, we are ready
to use the equipment in experiments, gather measurements, and interpret them. Once we
reach firm conclusions, we want to present the results in a seminar and eventually write
a report about the experiment. These two points, related to presenting our work, are the
topic of the final chapter.

QUESTIONS AND PROJECT IDEAS

1. Discuss the pros and cons of using TCP versus UDP.

2. Add a second state-machine (thread) to the robot that periodically reads out an MQ-x
gas-sensor and sounds a special alarm, once it reports a significant presence of a gas.
3. Add an LCD display with I2C interface to the remote controller, to show status
messages from the robot.
4. Build a remote-controlled boat instead of a robot. The boat can be powered by a fan
salvaged from a computer. We can steer it by directing the air stream behind the
propeller with fins controlled by a model-servo.
5. Build a line-follower that uses LDR or phototransistors pointed towards the floor.
Program it to follow a white (or black) line made of masking tape.

6. Build a vending machine that detects the size and weight of inserted coins and that
pushes a chewing gum or other desirable item from a safe place to the bin where we
can pick it up.
7. Construct the model crane from question 13 in Chapter 3. Equip the containers with
a marker, such as a periodically flashing LED, and invent a scheme to run the crane
autonomously.
8. Contemplate how to remotely control a sailboat. What actuators do you need? In case
you want to add autonomous control, which sensors do you need?
 CHAPTER 14

Presenting and Writing

After having gone through the basic electronics and programming examples, both on Ar-
duinos and on the Raspi, and after completing several projects, we need to communicate
our activities to our colleagues and condense our activities into a well-motivated and concise
sequence of descriptions. This can be in the form of a presentation with slides, or as a report
for a thesis or a journal. As a template for the contents of either presentation or report, we
use the weather station example, just to illustrate the concepts with a specific example. We
start with the discussion of a presentation with slides.

14.1 PREPARING A PRESENTATION

The key issues when preparing a presentation are motivation and a good story line, as
well as substantial subject matter. Why is that so? The audience of a seminar usually is
rather heterogenous and we have to provide a funnel to guide them from their different
backgrounds to the subject matter of the seminar. Often a good start is stating a problem
that is easy to understand and then pointing out how our project solves or at least alleviates
the problem. The weather station came up in my home lab, because we use pressure vessels
filled with liquid helium as well as high-voltage equipment. The potential effect of barometric
pressure on the pressure vessels and the relevance of humidity for high voltage is intuitively
understandable for most people. So that is a viable motivation as to why we care about
the weather station as a device that helps us to correlate weather-based data with other
measurement data.
Once we have captured the attention of the audience with a catchy motivation, we need
to keep it alive by following a well-conceived story line that places the relevant topics—the
subject matter of the seminar—in a logically coherent sequence where one detail follows
the previous in a natural way. My suggestion is to start with easily understandable facts
and then progressively increase the complexity of the discussed material. We must avoid
situations where the audience starts wondering why we talk about a particular topic and
loses track of our story line. Since the story line addresses the overall organization of the
presentation it needs to be sorted out beforehand.
A method to organize a presentation that I found useful is the following: I start by
estimating the number of slides that fit into the allotted time. An average slide typically
requires 2 to 3 minutes for the audience to absorb. Thus, for a 20-minute presentation I
target about 8 to 10 slides as a rule of thumb. Then I prepare a title page with a catchy title
that encapsulates the essential point of the seminar, followed by one slide that motivates
the problem I intend to address. This I follow up with a brief and qualitative description
of the idea of how to solve it that I will elaborate in the remainder of the seminar. At

223
224  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

this point the audience should have a good idea about the scope of my seminar and should
be convinced that the idea I pursue has a decent chance of addressing the problem in a
meaningful way. On the subsequent slides it is useful to concentrate on the flow of some
quantity or some information from one stage to the next. In the example with the weather
station, the information flows from the sensor via the microcontroller to the host computer
where it is presented, either on a web page or accessible via control system. This flow of
information we can mimic in the organization of the slides.
For the bulk of the presentation, I recommend preparing the allotted number of empty
slides and giving each slide a title. The title should address one key issue per slide and
the sequence of slides should follow the logical flow we have identified beforehand. In the
weather-station example we have the issues: hardware with sensors, microcontroller, host
computer, then software running on the respective computing devices and protocols used to
interface the devices. Once we have the slides with their titles, we can shuffle them around
until we are satisfied with our story line.
In the next step I place one or several pictures or graphs on each slide to illustrate the
topic on the slide. Once all, or at least most, slides are equipped with a picture, I add a few
bullet-points with keywords to each slide. They remind me of what I intend to say when
presenting. They help me to understand my slides after 6 months, and those in the audience
to remember what I said after 2 months. The slides are not a substitute for a self-contained
report, but serve as illustrations for my presentation during a seminar.
Once all slides contain a picture and a few keywords, it is time to review the story line
again and see whether linking of the slides works. This addresses the flow from one slide
to the next and whether it comes in a natural and well-motivated way. If many forward
references are necessary, I consider reordering the slides to achieve a more natural flow where
information required at some stage is already discussed on a previous slide. Sometimes
forward references are difficult to avoid, but I try to minimize them.
When almost all slides are completed, I suggest preparing a final slide with a clear
synopsis of the main results and possibly some comments about how to extend the work.
After a final pass through the slides, with a check for coherence to ensure that the title and
final slides act as parentheses to enclose the subject matter, the presentation is ready.
Just for convenience I summarize the above guidelines in a presentation cookbook:
Motivation: the problem and intended way to address it.

Think of a story line and the logical flow of a concept, preferably with increasing
complexity.
Write a title on each slide and sort to follow the story line.

Add pictures to slides.

Add keywords to slides.
Add slide with conclusions and outlook.
Final check of coherence and linking. Done!

Naturally, my presentation cookbook is subjective, and should not keep you from using a
working way of preparing slides, but if you get stuck very early on in the process of preparing
a seminar, my guidelines may help you to get started.
Presenting one’s work in a seminar is the first step, and writing it up in a report or a
thesis is the next.
 Presenting and Writing  225

14.2 PREPARING A REPORT

When preparing a report or a thesis in general, I follow the same guidelines used for the
preparation of a presentation. I need to motivate what I am about to describe and then
adhere to a story line that follows the flow of a concept. As a first step I normally try to
formulate a catchy title and an abstract of 100 to 200 words to explain to myself what I
intend to write in the report. Often this is a reasonable starting point even for the final
version of the abstract.
For the main part of the report, instead of starting with empty slides and filling the title
line, I start by identifying a logical sequence of chapters, sections, or subsections, and give
each one a title before ordering them in a sequence that follows the flow of the argument.
In the next step I add pictures or other illustrative material to the respective sections. This
normally results in a valid skeleton for the report that I need to flesh out in the next step.
I start this process by adding bullet lists to the sections with topics that are relevant in
the respective section. I do this for all originally identified chapters or sections, and often
in this stage I note that I need to split sections into two or more or that I can combine
sections with similar contents into a single one. As a guideline I assume that each topic in
the bullet list will require one paragraph with about 200 words or so to discuss adequately.
I also try to roughly balance the length of sections, but this is of minor concern. Once I
know what contents should show up in the respective sections, I start writing the text and
“fill in the blanks.” Since key components—images, graphs, or tables—are already in place,
I start each section with a short description of how it fits into the flow and then describe the
material. I also try to write each section from start to end, because that helps to maintain
a logical flow. At other times, if I do not come up with a decent start, I just start at a place
where I have a good idea about how to present it and retrofit the missing parts later. In
that case, however, I need to pay special attention later to ensure the flow.
A particularly important section is the Introduction, where I need to convince the reader
that the report treats a relevant and interesting subject by stating the chosen problem and
how to solve it. But ideally the problem should also be put into the context of previous
work. What have other researchers done before on related problems? How does their work
differ from mine? This section, typically one or two paragraphs long, requires a number of
key references from the published literature. At the end of the introduction, I sometimes
give a brief outline of the report in the fashion of a commented table of contents. By this
time I have hopefully convinced the reader to continue with the remainder of my report.
And this is the prime task of the introduction: a brief statement of the problem, the context,
and a brief outline of things to come.
The main part of the report should follow the story line and the flow I initially decided
upon. I write one or two paragraphs for each topic in the bullet list and make sure that
the paragraphs are well connected in the sense that the reader knows how the following
paragraph links to the present one. I try to avoid having the reader wonder what the
lines he is presently reading have to do with the overall story line. Placing crosslinks with
references to topics addressed earlier and how they connect to the present paragraph are
helpful to make the text denser, and creates additional associations for the reader. Phrases
such as, “where we use the result from page. . . ” or “as discussed in Section. . . ” illustrate
the idea.
In the Conclusions, I give a synoptic review of the main part and the results of the
report. I try to organize the introduction and summary jointly as an executive summary.
Often, readers first read those two sections before deciding to spend time on the report in
its entirety. The most important part of the conclusions is a concise summary of the key
226  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

results. I sometimes follow this up with a number of questions that came up during the
research and that may lead to further work, either by me or some other researcher.
I like to point out that the organizational triplet of introduction, main body, and con-
clusion vaguely resembles the form of a sonata, where first the themes are introduced. In
the main body of the sonata the themes are elaborated and in the coda they often reappear
in the original form once again. In this section we discussed the structure of the report and
my recommendations are again subjective, but may help in case you are stuck.
After the discussion of the overall organization of the report, we now look at some of
the ingredients and start with the presentation of data, often using graphics and plots.

14.3 PRESENTING DATA

Many aspects of scientific work are presented with the help of figures describing the exper-
imental setup, or graphs showing the data, sometimes with comparison to a model. In [29]
E. Tufte names three guiding principles to achieve graphical excellence.
Clarity is the first principle, and it requires carefully explaining the experiment from
which the data originate, describing what quantities are plotted on a graph, and properly
labeling the axes. Moreover, using a legible font of adequate font-size and avoiding low-
contrast colors such as yellow or light green on white background, is mandatory.
Precision, the second principle, dictates presenting data honestly and truthfully. If data
points are excluded, the reason and method to select the discarded points need to be ex-
plained. The graphics data of a manuscript must be proofread in the same way as the main
text. Make sure to catch all errors, such as omitted exponents or lost or incorrect axis labels.
Choose axes adequately. A bad example is to display the average temperature on Earth in
Kelvin with zero displayed. Global warming is invisible on such a plot.
Efficiency, the third principle, aims at presenting information in a parsimonious way:
to present the largest possible insight with the least effort. In particular, display only data
that advance the main argument, but then explain all displayed features in the plot, either
in the caption or in the text. If the plot is part of a seminar, the speaker may explain it as
well. On the other hand, any unnecessary data, which Tufte calls “chart-junk,” should be
omitted from a graph. A bad example is to present all available measurement data, even
though a well-chosen subset suffices to make a point.
In The Elements of Graphing Data, W. Cleveland discusses a large number of guidelines
to produce graphical data, with many examples. He summarizes the guidelines in a concise
list of “rules” such as
– to make the data the most prominent feature and avoid cluttering the data area;

– to keep the number of tick marks limited and preferably outside the data area;
– to have compatible scales when comparing two data sets;
and many more. Consulting the book is highly recommended in case of questions about
graphing data.
Scientific journals have a strong interest that their authors produce high-quality articles,
and those often include graphical data. They often expand the general terms of Tufte
and Cleveland and provide comprehensive style guides, both for written and for graphical
material. One good example is [30], which inspired the following points.
– Ensure your audience understands four things about the data points: what quantity
they represent, their physical unit, their magnitude, and the uncertainty.
 Presenting and Writing  227

– Explain the error bars: Are they standard errors or confidence limits?
– Always label axes with quantity plotted and with units. I normally display units in
square brackets. In journals, the use of title text is discouraged in favor of text in the
caption.
– If there are several plots in the same graph, preferably label them, use a legend, or
explain them in the figure caption.
– Choose the vertical scale such that at least 80 % of the range is used.

– If an axis covers more than 2 orders of magnitude, consider using logarithmic scale,
if appropriate.
– Avoid “eye-candy” such as 3-D bar-graphs or pie-charts.
– When only a few data points need to be presented, say 10 or less, a table is often
preferable to a plot.
These basic guidelines should provide a starting point to produce presentable plots.
The last point in the above list refers to data presented in numerical form in tables or
in the main text. Under no circumstances should you state more than a sensible number of
significant figures! If a measurement generates data that is accurate to 1 %, two or maybe
three significant figures are adequate. Just because a computer displays results in double
precision with 10 or more figures, this does not mean all figures are significant. And finally,
error bars in numeric form should never be displayed with more than one or sometimes two
significant figures.
And this brings us to writing the main text. Therefore, a few words about writing good
English are in order.

14.4 GOOD ENGLISH

There are a number of style guides for the English language available, from the classic
Elements of Style [31], the MLA handbook [32], S. Pinker’s Sense of Style, [33] to S. King’s
On Writing [34]. Apart from these more general guides, several scientific journals such as
Nature [35] or Reviews of Modern Physics [36] make style guides available for their authors.
In particular, the latter style guide is very readable, and its Appendix A on “Writing a
better scientific article,” is highly recommended. In this section I highlight some of the
topics in more detail.

– In one of the style guides [35] we find the sentence “Nature journals prefer authors to
write in the active voice. . . .” The active voice usually makes the presentation clearer,
more vigorous, and often more concise. Experiments are not done by themselves, but
we, the experimenters, perform them. Moreover, often the subject matter is already
difficult to understand, so we should make the presentation as clear as possible without
obfuscating who does what. Science does not become more objective just because the
person who did the experiment hides behind the passive voice.
– Authors should strive for economy in their presentation and avoid unnecessarily com-
plicated constructions such as “owing to the fact that,” which we can usually replace
with “because.” Often replacing passive constructions with active ones helps to dis-
entangle complex sentences.
228  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

– Avoid overloading the reader with sentences containing a large number of subclauses
to explain every conceivable exception in one sentence. At the end of the sentence,
the reader may have forgotten the beginning and what the subject of the sentence is.
– Be friendly to your reader and invite her into your intellectual world by more or less
addressing her directly, using phrases such as “let us” or “we now return.” This helps
to make the report more accessible by mimicking normal colloquial speaking patterns.
The reader is not alienated by a stiff and abstract presentation but feels welcome to
mentally participate in your exposé.
– Commit yourself to what you write and avoid unnecessary hedging by conditionals.
This is easily done by trying to replace any occurrence of “might be” or “could be”
with “is.” Other hedging phrases to look out for are “may” and “could,” or anything
that hints at you being scared of writing what you really mean.
– If in doubt, use English instead of Latin phrases. In that sense, “first” is a better
choice than “initial,” use “place” instead of “location.”
– Avoid acronyms unless a long term with a commonly used acronym appears in sev-
eral places throughout the report. In that case, introduce the acronym at the first
occurrence by writing out the full term first and adding the acronym in brackets.
Henceforth use the acronym consistently!
– Avoid jargon! Someone uninitiated in the jargon specific to your particular project
might want to read—and understand—your report.
– Pay special attention to grammatical correctness and especially to agreement errors.
Ask an English-speaking colleague to proofread your report.
– Before releasing your report, make sure to run it through a spell-checker, and read it
a final time to verify that all is correct!
These guidelines are certainly incomplete, but hopefully will help you to write an interesting
and readable report. I suggest you also read one of the more comprehensive style guides
such as [36]. A further source of advice regarding writing and presenting scientific work
in general is Nature publication’s web page, English communication for scientists [37]. A
wonderful source of synonyms and antonyms is the WordNet browser [38]. You may consider
using either the online version or a version installed on your computer in order to make
your writing more lively.

14.5 POSTSCRIPTUM
And at this point we have reached the end of the book. In the process we talked about various
sensors, both analog and digital, connected them to microcontrollers, massaged the data
into a common format, and passed them on to a host computer. There we postprocessed
the measurements, stored them in databases, and prepared them for a presentation or
report. I hope that you found useful, interesting, and inspiring topics between Preface and
Postscriptum that will help you in your projects that have a data-acquisition aspect.
I tried out all circuitry and programming presented in the book, but the odd bug may
have crept in. If you find one, please do not keep the bug, but share it with me so I can
improve future versions of the book. Of course, the general disclaimer applies, namely that
the code in the book is provided as is and users are advised to use caution, but are in any
case responsible themselves for using the program code.
 APPENDIX A

Basic Circuit Theory

As a reminder, we will briefly review the basic theory of electronic circuits. We start by con-
sidering Ohm’s law, which states that the current I in a segment of a circuit is proportional
to the voltage U with the resistance R as the proportionality constant, or U = IR. This is a
consequence of the balance of the electric field accelerating electrons in the resistor and the
friction force experienced by electrons due to scattering with vibration modes, phonons, of
the ions that make up the material. The force pulling on the electrons is proportional to U
and the friction force is proportional to the electron’s drift velocity v, or by −αv where α is
a material-dependent friction coefficient. Under stationary conditions, the two forces need
to balance, and we have v = U/α. But the current I is proportional to the drift velocity v
and we find I ∝ v = U/α, which is the essence of Ohm’s law, and the friction coefficient α is
inversely proportional to the resistance R. In case the voltage is varying moderately slowly
compared to the relaxation time of the electrons, the current directly follows the applied
voltage; the current I and the voltage U are in phase.
For capacitors the situation is different, because the two plates of a capacitor are electri-
cally separated and do not allow constant transport of charges under stationary conditions.
They can, however, store a charge Q on the plates, and the capacitance C is the proportion-
ality constant between Q and the applied voltage U , given by Q = CU. Since the charge Rt Q
changes as a consequence of current I flowing onto the plates, we have U = (1/C) Idt0 ,
or by differentiation, dU/dt = (1/C)I. A sinusoidally oscillating voltage U ∝ eiωt with
frequency ω will therefore be related to the current by U = (1/iωC)I, and we can identify
ZC = 1/iωC as the generalized resistance, the impedance, of the capacitor. Here i is the
imaginary unit.
Inductors are made of coils that store energy in their magnetic field, and if we turn off the
current, a voltage develops. Expressing this behavior in a formal way, we have U ∝ dI/dt,
with the inductance L of the coil as the proportionality constant. If we again assume that
the coil is excited by a sinusoidal current I ∝ eiωt with frequency ω, we find that we have
U = (iωL)I and identify ZL = iωL as the impedance of the inductor.
Now that we have resistances and impedances of common elements found in simple
circuits, we may ask how to combine them to networks. This question is answered by
Kirchhoff ’s laws, of which the first states that under stationary conditions, all currents
flowing into a network node must add up to zero. This is a statement about the preservation
of charges, and what comes in must also come out, because under stationary conditions
piling up charges is not allowed. The second law states that the voltage differences around
a loop in the network have to add up to zero, which is the requirement that the voltages at
each node of the network with respect to a reference node are unique.

229
230  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Z2
A B C
Z1 Z2 A B

Z1
Battery

Figure A.1 Two impedances connected in series (left) and in parallel (right).

We immediately use these laws to investigate the resistance or impedances connected

in series, as shown on the left in Figure A.1. The currents in node B need to balance, and
this implies that the current I1 passing through Z1 is the same as the current I2 passing
through Z2 . At the same time, the voltage U provided by the battery is the same that is
dropped across the two impedances. We thus find U = I1 Z1 + U2 Z2 = I1 (Z1 + Z2 ), and we
find that the total impedance of the circuit equals the sum of the impedances coupled in
series.
Considering the circuit on the right in Figure A.1, we know that the currents through
the two branches have to add up: It = U/Zt = I1 + I2 = U/Z1 + U/Z2 . We can simplify this
to 1/Zt = 1/Z1 + 1/Z2 , which implies that the individual impedances add as reciprocals if
the impedances are coupled in parallel.
Note that the argument in the previous two paragraphs is valid for normal resistors
and constant (DC) voltages and currents, but works as well for AC voltages if we use the
complex frequency-dependent impedances for inductors and capacitances. We also note that
all relations among currents and voltages are linear, such that we can use the super-position
principle and analyze circuits for each voltage or current source independently and then add
the contributions in the end. A further consequence is Thevenin’s theorem, which states that
any subcircuit connected via two terminals a and b to the rest of the circuit can be replaced
by a voltage source Uth and a series resistor Rth . It turns out that Uth and Rth can be
very easily determined. Uth is given by the voltage between unconnected (open-circuit)
terminals a and b, and the resistance Rth by the current Ith = Uth /Rth that flows through
the short-circuited link between terminals a and b.
Apart from the linear circuit elements, the impedances, there are also nonlinear ele-
ments, and diodes are among them. Their
voltage-current behavior follows an exponential
dependence given by I ∝ e(eU −Eg )/kT − 1 , where Eg ≈ 1.2 V is the bandgap energy of the
semiconductor material, here silicon. We see that for negative voltages the current is very
small, while for positive currents it grows exponentially once the threshold of the bandgap
voltage is passed. Practically, diodes conduct current in one direction and block current in
the opposite direction. This makes them perfect to rectify voltages in the way we discussed
in Section 2.2.5. A diode is conducting if it is forward biased and the cathode is at a more
negative voltage than the anode.
Transistors are other nonlinear elements based on two diodes coupled antiparallel; they
come in two types, dependent on whether the central tap, called the base terminal, is an
anode or a cathode. Injecting an additional current into the base terminal enables a larger
current flow through the two outer terminals, called collector and emitter. In the main body
of the text we use transistors for switching applications, but with suitable ancillary circuitry
they can be used as amplifiers as well. That goes beyond the scope of this appendix and is
covered in books on electrical engineering, such as [15] or [16].
 APPENDIX B

Least-Squares Fit and Error

Propagation

In Chapter 11 we used the linfit() function on the Arduino to fit a straight line through
a number n of data points sampled at equidistant times. Here we briefly discuss the inner
workings of that function, but use the opportunity to expand that topic, because fitting and
error propagation analysis are essential for any experimental activity. We therefore extend
the fitting procedure to comprise generalized least-square problems, and give a rudimentary
overview of the corresponding error analysis and of error-propagation in general.
But we start with the problem of determining the slope a and intercept b of a fit to a
straight line. These parameters are determined by the requirement to minimize the sum of
squared residuals ri = yi − ati − b for a number of data points (ti , yi ). We can restate the
problem in the form of a linear equation that needs to be inverted in the least-squares sense
.. .. ..
   
. . .  
 yi  =  t i 1  a
   
    b (B.1)
.. .. ..
. . .
which we write in abbreviated form
y = Ax (B.2)
T
where A is the n×2 matrix in the previous equation and x = (a, b)P. Here the superscripted
n
T denotes the transpose, and for the squared sum of the residuals i=1 ri2 we write χ2 . The
latter we express as
n
X
χ2 = ri2 = (y T − xT AT )(y − Ax)
i=1
= y T y − xT AT y − y T Ax + xT AT Ax . (B.3)

Minimizing this expression with respect to x results in a condition for the sought solution
vector x. The condition for a minimum is the requirement that the gradient with respect
to the fit parameters xT is zero. We somewhat sloppily write it as
∂χ2
0= = −2AT y + 2AT Ax (B.4)
∂xT
where we used xT Ay = y T AT x because the expressions are scalars and any matrix of

231
232  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

the form AT A is symmetric. This allows us to combine terms and we arrive at the previous
equation. For a more detailed derivation see [39]. Left-multiplying with the inverse of (AT A),
provided the matrix is nonsingular, isolates the sought-after fit-parameter x and we obtain
 
a
−1 T
= x = AT A A y. (B.5)
b

The right-hand side is often called pseudo-inverse of the matrix A, and we need to evaluate
this expression in order to find a and b. If the t values are equidistant, we can absorb the
step size in a redefinition of a, and the matrix A obtains the form
 
1 1
 2 1 
A= 3 1  (B.6)
 
.. ..
 
. .

such that AT A becomes

 Pn Pn
k2
  
T Pk=1 Pk=1 k n(n + 1)(2n + 1)/6 n(n + 1)/2
A A= n n = (B.7)
k=1 k k=1 1 n(n + 1)/2 n

which accounts for the definition of S0,S1,S2 in the linfit routine. Inverting this 2×2 matrix
is trivial, and calculating AT y is done in the loop where ay0 and ay1 are calculated. The
linfit function then returns the parameter a, the slope, and in the main program we need
to rescale the result by multiplying the slope with the step size. It is easy to cross-check the
results with octave, and the fitting is done with the polyfit() function, but if we want to
do the calculation on the Arduino, we need methods like the one shown in this appendix.
Note also that generalizing the method to higher-order polynomials is simple. We only
need to add columns in Equation B.1 with tni and can rescale the fit parameters a, b, ...
by the appropriate power of the step size, and we are left with adding powers of positive
integers for which closed expressions exist. This means that the matrix AT A can always be
calculated in closed form for any number of data points n. Only the inversion of a matrix
with rank of the number of fit parameters remains. Being able to precompute most of the
matrices and possibly inverting them on the host computer for fixed n makes this method
rather suitable for microcontrollers.
And going beyond fitting polynomials, we realize that any problem that can be cast
into the form y = Ax is solved by the pseudo-inverse given by x = (AT A)−1 AT y. Often the
column-vector y contains the measurements that have a linear dependence on fit-parameters
x, and that dependence is characterized by a system-matrix A. Here A is the response matrix
of the measurements y on some parameters x. We observe that the pseudo-inverse provides
a linear map of values y in the “measurement-space” onto the values x in the “fit-parameter
space.” In general, there are more measurements denoted by n than fit-parameters m, thus
n > m and the system is overdetermined.
Up to now we assume that all measurements contribute equally to the fit result, but this
is often not the case, because there are measurement values y that are unreliable for some
reason, and they have large associated errors. If we denote the measurement uncertainty for
measurement yi by σi , we weigh each measurement in Equation B.2 by the inverse of σi . In
this way a large error bar σi effectively “turns off” one measurement. We formalize this by
introducing the matrix Λ = diag(1/σ1 , 1/σ2 , . . . , 1/σn ) and left-multiplying Equation B.2
by Λ with the result
Λy = ΛAx . (B.8)
 Least-Squares Fit and Error Propagation  233

Redoing the calculation that led to the pseudo-inverse now leads to

x = (AT Λ2 A)−1 AT Λ2 y (B.9)

which takes the measurement error bars into account.

The next question we may ask is how the error bars of the measurements σi determine
the error bars of the fit parameter x. In order to visualize the calculation, we assume that the
error bars σi come from repeatedly measuring yi and using the sample average hyi i as input
to the calculation, and that the root-mean squared σi2 = h(yi − hyi i)2 i determines the error
bar. Here we use the angle brackets to denote averaging over the ensemble of measurements.
If, in order to simplify the notation, we introduce the abbreviation R = (AT Λ2 A)−1 AT Λ2 ,
we obtain for the ensemble average of the fit-parameter hxi i, written for one component
* n + n
X X
hxi i = Rij yj = Rij hyj i (B.10)
j=1 j=1

because taking the average is a linear operation and allows us to extract the sum from the
average. The error bars of the fit-parameters are then given by the covariance matrix C(x).
Considering only one component of the covariance matrix C(x)ij , we find

C(x)ij = h(xi − hxi i)(xj − hxj i)i

* n ! n !+
X X
= Rik (yk − hyk i) Rjl (yl − hyl i)
k=1 l=1
n X
X n
= Rik Rjl h(yk − hyk i)(yl − hyl i)i (B.11)
k=1 l=1
Xn X n
= Rik Rjl C(y)kl
k=1 l=1

which can be written in matrix form as

C(x) = RC(y)RT . (B.12)

This equation is very powerful, because it allows us to calculate the covariance matrix of
the fit-parameters as a function of the covariance matrix of the measurements and the
pseudo-inverse that we abbreviated R. Now we realize that the covariance matrix of the
measurements is C(y) = diag(σ12 , σ22 , . . . , σn2 ) = Λ−2 and inserting this and the definition of
R leads to the result
−1
C(x) = AT Λ2 A (B.13)
which contains the squared error bars of the fit parameters on the diagonal, such that
p
σ(x)i = C(x)ii (B.14)

and the off-diagonal elements contain information of the correlations among the fit-
parameters.
We point out that Equation B.12 describes error propagation from variables y to vari-
ables x, and the derivation depends on the linearity of the map from y to x. If the un-
derlying map from some set of variables y to another set x is non-linear such that we can
write x = x(y), we can still linearize around some value ȳ such that the Jacobi matrix
234  A Hands-On Course in Sensors Using the Arduino and Raspberry Pi

Jij = ∂xi /∂yj |y=ȳ describes how much small deviations in y in the vicinity of ȳ change the
value of x. This is how error bars in y cause error bars in x. We can therefore use the Jacobi
matrix J instead of the matrix R to describe the covariance matrix in the new variables x
in Equation B.12.
To illustrate this, we consider a simple example x = y1 /y2 with covariance matrix
 2 
σ1 r12
C(y) = . (B.15)
r12 σ22

The Jacobi matrix is given by

   
∂x ∂x 1 y1
J= , = ,− 2 (B.16)
∂y1 ∂y2 y2 y2

and for the covariance matrix C(x) we find

1 y1 y2
C(x) = JC(y)J T = σ12 2 − 2r12 3 + σ22 14 (B.17)
y2 y2 y2

which agrees with the “conventional wisdom” for adding measurement errors for uncorre-
lated variables y1 and y2  2  2
2 2  ∂x  2  ∂x 
   
σ(x) = σ1  + σ2 
∂y1  ∂y2 
On the other hand, if the variables are correlated, which is expressed by a nonzero component
r12 in the covariance matrix, using Equation B.12 with the Jacobi-matrix substituted for R
yields consistent results.

QUESTIONS AND PROJECT IDEAS

1. Verify that Equation B.13 is correct by inserting the definition of R into Equation B.12.

2. Determine the covariance matrix for the linear fit to determine slope a and intercept
b. Assume that all measurement errors are equal to 5 % of the maximum measurement
value y if you have n = 10, 100, and 1000 measurements. How large are the result-
ing error bars for a and b? Discuss the relevance of the off-diagonal element in the
covariance matrix.

3. Provided the covariance matrix of two variables y1 and y2 are given by Equation B.15,
calculate the covariance matrix of the variables x1 = y1 /y2 and x2 = y2 /y1 . Are the
errors of x1 and x2 correlated?
Bibliography

[1] Fritzing. Project web site: http://fritzing.org/.

[2] ATLAS collaboration. Project web site: https://cern.ch/atlas.
[3] CMS collaboration. Project web site: https://cern.ch/cms.
[4] Large Hadron Collider. Project web site: https://cern.ch/lhc.
[5] CERN the European Center for Nuclear Research. Project web site: https://cern.ch.
[6] MQTT. Project web site: http://mqtt.org/.
[7] EPICS. Project web site: http://www.aps.anl.gov/epics.
[8] Arduino. Project web site: https://www.arduino.cc.
[9] Raspberry Pi. Project web site: https://www.raspberrypi.org.
[10] J. Fraden. Handbook of modern sensors. Springer Verlag, Berlin, third edition, 2004.
[11] J. Wilson, editor. Sensor Technology Handbook. Elsevier, Amsterdam, 2005.
[12] R. B. Northrop. Introduction to Instrumentation and Measurements. CRC Press, Boca
Raton, third edition, 2014.
[13] M. Coplan, J. Moore, and C. Davies. Building Scientific Apparatus. Cambridge Uni-
versity Press, Cambridge, UK, fourth edition, 2009.
[14] C. Kittel. Introduction to Solid State Physics. Wiley, Hoboken, NJ, eighth edition,
2005.
[15] T. Giuma and P. Peebles. Principles of Electrical Engineering. McGraw-Hill, New
York, 1991.

[16] Analog Devices: Analog Circuit Design Tutorials. Project web site: http://www.
analog.com/en/education/education-library/tutorials/ebooks.html.
[17] D. Lancaster. Active Filter Cookbook. Newnes, Oxford, second edition, 1996.

[18] G. Franklin, J. Powell, and A. Emami-Naeni. Feedback Control of Dynamic Systems.

Pearson, Boston, seventh edition, 2015.
[19] M. Margolis. Arduino Cookbook. O’Reilly, Sebastopol, CA, 2011.
[20] E. Bartmann. Die elektronische Welt mit Raspberry Pi entdecken. O’Reilly, Sebastopol,
CA, 2013.

235
236  Bibliography

[21] M. Kofler, C. Kühnast, and C. Scherbeck. Raspberry Pi, das umfassende Handbuch.
Galileo Press, Bonn, 2014.
[22] S. Monk. Raspberry Pi Cookbook. O’Reilly, Sebastopol, CA, 2014.

[23] Scientific Programming Language GNU Octave. Project web site: https://www.gnu.
org/software/octave/.
[24] Python Software Foundation. Project web site: https://www.python.org/.
[25] MySQL Database. Project web site: https://dev.mysql.com.

[26] RRDtool. Project web site: https://oss.oetiker.ch/rrdtool/.

[27] Apache web server. Project web site: https://httpd.apache.org/.
[28] Installing EPICS on the Raspberry Pi. https://prjemian.github.io/epicspi/.

[29] E. Tufte. The visual display of quantitative information. Graphic Press, Cheshire,
1983.
[30] C. Mack. How to write a good scientific paper: Figures, part 1. J. Micr/Nanolith,
12:040101–1, 2013.
[31] W. Strunk and E. White. The Elements of Style. Harcourt, 1920. Online available
from Project Gutenberg at https://www.gutenberg.org/ebooks/37134.
[32] J. Gibaldi. MLA Handbook for Writers of Research papers. Modern Language Associ-
ation of America, New York, NY, sixth edition, 2003.
[33] S. Pinker. The Sense of Style. Penguin Books, London, 2015.

[34] S. King. On writing. Hodder, London, 2012.

[35] Nature style guide. http://www.nature.com/authors/author_resources/how_
write.html.
[36] K. Friedman. Reviews of modern physics style guide, http://journals.aps.org/
files/rmpguide.pdf.
[37] Nature. English communication for scientists, http://www.nature.com/scitable/
ebooks/english-communication-for-scientists-14053993/contents.
[38] C. Fellbaum, editor. WordNet: An Electronic Lexical Database. MIT Press, Cambridge,
MA. Online version: http://wordnet.princeton.edu.

[39] W. Press et al. Numerical Recipes. Cambridge University Press, Cambridge, second
edition, 1992.
Index

7805, 28, 211 Serial communication, 61, 63

Socket communication, 109
Accelerometer, 13, 34, 35, 73 Sound, 103
Acronym, 142, 228 Stepper motor, 93, 95, 99
ADC, see Analog-to-digital Switching and PWM, 86
ADXL335, 13, 34 Weather station, 167
Aliasing, 26, 177 Web server, 107, 214
Amplifier, 17–20, 22, 177, 178
Analog-to-digital, 1, 16, 20, 23–26, 33–35, Bandgap, 7, 10, 27, 37, 39, 230
62, 68, 73, 76, 77, 79, 80, 117, 177, Barometric pressure, 1, 32, 33, 68, 165, 167,
181, 184, 205, 206, 209, 210 175, 223
Apache2, 141–144, 172 Basic Unix programs, 115
Arduino, iii, 1, 3, 55–58, 60, 63–68, 71, 74, Battery, 28, 112, 157, 177, 190, 205, 211
76, 77, 79, 82–86, 89, 91–95, Bluetooth, 104–105, 110, 113, 124, 128, 130
97–99, 101, 103–105, 110–112, 117, pairing and setup, 105
123–125, 139, 146, 185, 186, 191, BMP180, 32, 33, 70, 71, 111, 167, 170
198, 201, 205, 213, 221, 231 BPW34, 14, 15
Arduino NANO, 112, 213 BPX38, 15, 205, 211, 213
Arduino sketch Breadboard, 16, 33, 77, 89, 93, 98, 167, 185,
Analog input, 61, 63 186, 191, 197, 205–207, 211, 213,
Blink, 58 219, 221, 222
BMP180 pressure sensor, 69, 71 Brightness, 40, 85, 87, 107–109, 111, 112,
Button, 60 146, 165
Capacitance measurement, 191 Button, 32, 59–61, 65, 74, 76, 109, 115, 116,
Color monitor, 186, 188 119, 123, 145, 206, 211
DC motor with H bridge, 90
Capacitance, 3, 13, 14, 21, 22, 27, 29, 33,
DHT11 humidity sensor, 80
34, 37, 38, 42, 191, 194, 229, 230
DS18b20 temperature sensor, 83
Commutator, 43, 44
Geophone client, 177
Comparator, 20, 23, 24, 38
HC-SR04 distance sensor, 83
Compile, 59, 62, 148, 151, 153, 154
HYT221 humidity sensor, 67
Conclusions, 224, 225
Laser-beam profile, 198
Conduction band, 7, 8, 15, 39
LM35 temperature logger, 65
Control system, 3, 147–156, 161, 174
MCP23017 IO-extender, 74
Covariance matrix, 233, 234
MCP3304 ADC with SPI, 77, 79
Crane, 54, 222
MCP4921 DAC, 101
Cron process, 133, 134, 137, 140, 141, 144,
MLX90614 IR thermometer, 66
145
Model-servo, 91
MPU6050 accelerometer, 72 DAC, see Digital-to-analog
Query-response, 64 Darlington, 41, 42, 50, 53, 85, 87, 93
RC receiver, 214 Database, iv, 117, 118, 132–140, 149,
RC sender, 206 152–154, 156, 163–165, 167, 171,
Rotary encoder, 84 172, 174, 175, 182, 183

237
238  Index

DC motor, 12, 43–46, 52, 53, 89, 90, 94, HTML, 108, 109, 112, 142, 143, 173
165, 205, 210, 211, 219, 220 examples, 142, 143, 145, 173
Decoupling capacitor, 63, 167 HTML forms, 109, 112, 146
Delta-sigma ADC, 24, 25 http-equiv, 145
Desktop, 114, 116, 120, 121, 123, 124, 146, HTTP, 112
150, 156 GET, 108, 109
DHCP, 106, 108, 121, 122 header, 108
DHT11, 36, 37, 80, 81 return code, 108
Dielectric constant, 8, 33, 36 Humidity, 1, 33, 36, 67, 68, 71, 80, 82, 165,
Digital-to-analog, 24, 51, 52, 101, 102 170–172, 223
DNS, 106 Hydraulics, 52
DRV8825, 49, 98, 99 HYT-221, 33, 67, 68
DS18b20, 37, 82, 83
Dust sensor, 37, 112, 175 I2C, 30, 32–35, 52, 57, 65–68, 70–76, 111,
113, 167, 170, 211
EPICS, iv, 1, 3, 147–157, 161–165, 167, 174, Imagemagick, 118
175, 177, 182, 183 Impedance, 5, 18, 21, 53, 194, 229, 230
EPICS database file Inductance, 21, 22, 38, 229
geophone, 182 Infrared, 12, 15, 30, 40, 110, 213
MQTT interface, 163 Internet of Things, 1, 3, 56, 157
simple, 149 Interrupt, 74, 76, 81, 84, 85, 111, 179, 183,
temperature, 152 184, 191, 193
weather station, 174 IOC, 147, 153, 154, 156, 183
ESP-01, 56–58, 112 IP, 116, 121, 122, 124, 128, 131, 141, 153,
ESP8266, 55–57, 105, 110 162, 164, 167, 171, 183, 209, 216
iptables, 121
Fermi level, 7, 38
Filter, 1, 21–22, 26, 29, 32, 51, 53, 111, 177, Jacobi matrix, 233
183 Jargon, 228
Flatfile database, 132–134, 137 Joystick, 9, 10, 205, 206, 209
Flow, 12, 15, 30, 42, 45, 52, 93, 224, 225,
230 Keyword, 5, 83, 119, 224
Fluid level sensor, 9 KODI, 113
Flyback diode, 43, 50, 52, 87–89
Force sensitive resistor, see Strain gauge L293D, 46, 88, 89, 94, 95, 97, 198, 211, 221
LDO regulator, 28
Gas sensor, 9–11, 112 LDR, 5–7, 17, 146, 175, 197, 198, 201, 204,
Gateway, 157, 161–165 222
Geophone, 13, 14, 177, 178, 180–183 Least-squares, 231
GPIB, 65 LM35, 10, 11, 27, 61–63, 106, 107, 139, 159,
GPS, 36, 112 163, 167, 170
Grammar, 228 Loudspeaker, 103
Ground vibrations, 3, 177
Gyroscope, 34, 35, 73 MATLAB, 3, 118, 137, 147
MCP1700, 28
H bridge, 45, 46, 48, 50, 53, 85, 88–90, MCP23017, 74, 76, 86, 111, 222
94–96, 165, 198, 205, 211 MCP3208, 206
Hall sensor, 1, 13, 19, 35, 44, 112 MCP3304, 35, 77–80, 111, 184, 206
HC-SR04, 31, 83, 112, 205, 211, 215, 219 MCP4921, 52, 101
HDMI, 113, 114 Microcontroller, 26, 28, 30, 33, 39, 43, 53,
 Index  239

55, 57, 60, 65, 73, 74, 77, 80, 123, Pin diode, 14, 15, 38, 183
153, 167, 170, 177, 211, 224 PIR sensor, 30, 31
Microstepping, 48–50, 53, 98, 100, 204 Power supply, 26–28
MIDI, 40, 110, 111 Presentation, iii, 3, 115, 132, 223, 224,
MISO, 35, 79, 80, 101, 206 226–228
Mkoctfile, 137 Propeller, 12, 52, 112, 222
MLX90614, 12, 65, 66, 112 Pseudo-inverse, 232
MOSFET, 30, 42 PT100, 7, 8
MOSI, 35, 79, 101, 206 Pull-up, 29, 32, 55, 61, 76, 80, 82, 84, 85,
Motivation, 223, 224 191, 211, 221
MPU6050, 112 Pulse-width modulation, 40, 45–47, 51, 53,
MQ-x, 9, 11, 112 85, 86, 88–90, 92, 190, 197, 198,
MQTT, 3, 157–165, 188, 190 201, 215
Mrs. Robinson guideline, iii Python, 115, 123, 125–128, 132–134, 136,
MySQL, iv, 134–138, 146, 165 138–140, 161, 171
Python script
Nano text-editor, 118 EPICS to MQTT gateway, 161
Netcat, 109, 119, 120, 125, 163 Flatfile writer, 132, 133
Netmask, 121 Geophone reader, 180
Network address translation, 121 MySQL access, 136
Network addressed storage, 114 Query-response, 125
Network explained, 106 Read from network socket, 128
NodeMCU, 56–58, 77–80, 105–110, 119, Read from serial line, 127
123–125, 128, 130, 131, 146, 153, Weather station reader, 170
154, 159–161, 165, 167, 168, 175,
177–184, 188–190, 205–221 Query-response protocol, 3, 63, 65, 86, 90,
NPN transistor, 41, 42, 50, 87 92–94, 100–102, 117, 124, 125, 129,
165, 194, 220
Octave, 118, 128–132, 134, 137, 138, 181,
182, 196, 201, 202 R-2R resistor network, 51
Octave script Raspberry Pi, iii, v, 3, 111, 113–118,
Capacitance measurement, 194 120–125, 129, 133, 134, 137, 141,
Flatfile reader, 134 145–148, 151, 155–158, 161, 164,
FWHM, 203 170, 174, 201
Laser beam profile, 201 Raspbian, 114, 115, 155
MySQL access, 137 Reed switch, 29
Query-response, 129, 130 Relay, 42, 43
Smooth points, 202 Remote control, 32, 111, 175, 205–207, 211,
Temperature logger, 130 215, 216, 219–222
Open collector, 41, 42 Report, iii, 1, 222–226, 228
OpenELEC, 113 Resistance, 5, 7–9, 16, 21, 22, 33, 34, 40, 45,
OpenWRT, 113 50, 53, 61, 177, 193, 229, 230
Operational amplifier, 17, 18, 20, 25, 52 Robot, 3, 46, 103, 204–206, 209–215,
Optocoupler, 39, 40, 110 219–222
Rotary encoder, 29, 84
PCA9685, 222 Router, iii, 106, 113, 124
Phototransistor, 15, 37, 40, 112, 175, RRDtool, 138–141, 143–145, 170–172
185–188, 205–222 RS-232, 28, 31, 36, 55, 103, 105, 110, 151
Pictures, 146, 224, 225
Piezo buzzer, 103, 211 Sailboat, 222
240  Index

SCPI, 65 Valence band, 7, 15, 39

Screen program, 103–105, 117, 145 Valve, 39, 52
SD card, 114, 115 VNC, 120
Semiconductor, 7–9, 14, 39 Voltage regulator, 28, 221
Servomotor, 46–47, 52, 53, 91, 92, 112, 165, VXI, 65
205, 213, 216, 219
SFH3310, 15, 185, 186 Web server, iv, 3, 106–109, 112, 141–145
Slides, 223–225 Wheatstone bridge, 5, 7, 9, 16, 25, 33, 34, 37
SM-24, 13, 14, 177, 178 WiFi, 105–110, 113, 160, 167, 177, 188, 206,
Socket, 110, 128, 130, 131, 155, 161–163, 214
171, 181–183, 209 Wordnet, 228
Solar cell, 16 WPA, 57, 106, 121
Solenoid, 52
Spell-checker, 228
SPI, 35–36, 52, 55–57, 76–80, 101–102, 113,
205, 206, 209
State machine, 205, 215, 216, 219, 222
Stepper motor, 47–50, 52, 53, 92–95,
97–100, 146, 197–199, 201
Story line, 223–225
Strain gauge, 10, 32, 33
Structured query language, 136
Switch, 29–30, 39–43, 85–87
Synaptic, 119, 146

TCP, 130, 131, 162, 222

Telnet, 106, 109, 119, 124, 125, 155, 156
Temperature, 7–12, 27, 32, 33, 36, 37, 56,
61–63, 65, 67, 68, 70, 71, 73, 82,
83, 106–108, 130, 131, 133, 139,
140, 144, 145, 151, 153, 157, 158,
160, 161, 163, 165, 167, 170–172,
226
Terminal program, 36, 63, 103, 115–120,
123–125, 133, 148, 150, 154, 155,
158, 161, 163
Thermistor, 8, 36
Thermocouple, 11, 12
Thermopile, 12, 65
Tilt switch, 29
TIP-120, 87
TV, 32, 53, 111, 113

UDP, 209, 215, 219, 220

ULN2003, 50, 53, 93
Unix manual pages, 117, 119, 120, 134, 139,
141, 158
USB, 58, 59, 61, 103–105, 110, 112, 113,
117, 123, 124, 129, 146, 211
User guide, iii