ELEC2146

Electrical Engineering Modelling and Simulation

Simulation Programming and MATLAB

Dr Julien Epps
S2, 2011

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 1

 Overview

Programming vs. simulation
programming

Programming principles

How MATLAB works

Getting to know MATLAB

C : A refresher

Debugging

Good programming practise

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 2

 Simulation Programming

Most programming:
– Uses a variety of data structures
– Need to declare all variables (‘strongly typed’)
– Often need to worry about memory allocation of
variables
– Visual component = GUI

Often need libraries to implement this easily
– Few pre-defined functions

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 3

 Simulation Programming

Simulation programming:
– Often uses just numbers, arrays and matrices

Often not very focused on data structures
– Algorithm-focused
– Visual component = plots (GUI possible)

Aim is to visualise data easily
– MATLAB

No need to worry about
– declaring variables
– memory allocation
– Nearly every function is predefined

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 4

 Simulation Programming

Comparing programming languages
C
– Readability
Fair
– Writability OK
– Reliability

Everything except pointers OK

Pointers (buffer overflow) Bad
– Cost

Writing/debugging code Fair

Execution time Good
– Portability OK
– Generality OK

Adapted from Reus, B. “Principles of Programming Languages” http://www.cogs.susx.ac.uk/users/bernhard/ppl2007/Slides/ppl2s.pdf

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 5

 Simulation Programming

Comparing programming languages
MATLAB
– Readability
OK
– Writability Good
– Reliability

Everything except memory management Good

Memory management Bad
– Cost

Writing/debugging code Good

Execution time Not great
– Portability Bad
– Generality Bad

Adapted from Reus, B. “Principles of Programming Languages” http://www.cogs.susx.ac.uk/users/bernhard/ppl2007/Slides/ppl2s.pdf

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 6

 Programming Principles

How computers are organised
Central Processing Random Access
Unit Caches Memory
Instruction
Data
I/O Storage Data
Keyboard HDD Data
Mouse DVD Instruction
Ports etc USB etc Instruction . . . .

Programs are a set of instructions on

data, executed by the CPU

Data exists in storage (slow), RAM (fast)
or in a CPU cache (fastest)
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 7
 Programming Principles

Top-down design
– In procedural languages, code development is
based on a step-wise refinement of the abstract
function that the code is required to perform
– Breaks a problem into sub-problems

And from there into sub-sub-problems

Deals with each sub-problem separately
– Example:
% Problem: Get to UNSW

CheckTimetable(date,timetable);
DoINeedToGo(date,timetable); CheckSocialSchedule(date);

CheckGearNeeded(date,timetable);
WhatDoINeed(date,timetable); Raining(windowCam);
HotCold(temperature);
DrivingHassle(date,trafficCond);
GetThere(homeGPS,UNSWGPS);
NextBus(date,time);
NextTrain(date,time);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 8

 Programming Principles

Top-down design
– Start solution to sub-problems by writing pseudo-
code

Comments containing code-like statements or even just text
descriptions of the function being performed
– Wherever possible, find generic, atomic pieces of the
problem and write them as functions

Should be as independent and self-contained as possible

Typically have one entry and one exit point

Usually short

Makes code easier to read; improves underlying logic

Makes code easier to test

Large projects: can split up work

MATLAB: can start new file for function or include it as part of a

script file function b = sqrt(a)

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 9

 Programming Principles

Data structures: MATLAB
– Most common type of variable: double

64-bit double-precision floating-point

Can hold real, imaginary and complex numbers
– Arrays

The fundamental data type of MATLAB

1-D, e.g. array = [1 3 2 4];

2-D (matrix), e.g. A = [1 3; 2 4];

3-D and beyond, e.g. B(2,1,4) = 16;
– Characters (not used much in this course)
– Cell arrays (not used much in this course)

Generic containers – very handy

ExampleCell{1} = [1 3 2 4]; ExampleCell{2} = 56;
ExampleCell{3} = ‘Hello’;

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 10

 Programming Principles

MATLAB detaches developer from
memory management
– Dynamic variable size
A = 1;

A is a scalar: double A;
A(2) = 4;

A is now a 1x2 array: double A[2];
A(3,2) = -1;

A is now a 3x2 matrix: double A[3][2];
A(3,2,2) = 6;

A is now 3x2x2: double A[3][2][2];
– Growing an array within a loop
B=0; for k=1:100, B(k)=100-k; end
– “Weakly dynamically typed”
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 11
 How MATLAB works

High level language

Huge amount of abstraction from the machine-level instructions

Programming languages don’t come much higher level than this

Procedural language

Complex problems can be decomposed into a hierarchy of
functions

Interpreted language
– Each line of code is decoded as the compiler
reaches it

Slower than compiled languages. MATLAB now has a compiler

Based on fast numerical methods

LINPACK, EISPACK

Like an advanced calculator

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 12
 How MATLAB works

Command window
Workspace browser Note that variables defined here
can be seen inside MATLAB
scripts

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 13

 How MATLAB works

Where data is stored in memory

Used during command MATLAB workspace

execution command window
Not accessible to programmer workspace browser

Central Processing Random Access

Unit Caches Memory
Instruction
Data
I/O Storage Data
Keyboard file
HDD load file
Data
Mouse DVD
Ports etc
system
USB etc
Instruction
Instruction . . . .

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 14

 How MATLAB works

Scripts
– A list of commands, stored in exampleScript.m
– Execute by running exampleScript in the
command window
– Think of scripts as identical to the command window

Functions
– Should always be used in preference to scripts,
where possible to produce generic, reusable code
– Take inputs, produce outputs
– Execute in command window by running e.g.
c = dist(a,b);
function d = dist(x,y)
d = sqrt(sum((x-y).^2));

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 15

 How MATLAB works

Built-in functions
– Lots of them
– Very useful
– Functions are “overloaded”

Same function can take different inputs or produce different
outputs depending on how it is called
>> x = rand(10,10,10);
>> mean(x)
ans(:,:,1) =
0.6239 0.6602 0.5873 0.3978 0.5614 0.4500
0.5154 0.4969 0.5980 0.3891 etc . . . .
>> x = rand(10,10);
>> mean(x)
ans =
0.4441 0.5751 0.3666 0.6025 0.4448 0.6426
0.3992 0.6124 0.5794 0.4805
>> x = rand(1,10);
>> mean(x)
ans =
0.5878
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 16
 How MATLAB works

Some MATLAB tips
– If your code is slow, avoid for loops

Try to use vectors and matrices – MATLAB is fast for these

Other languages: for loops are fine
– If your code is slow, work out where the problem is

Use the profiler: help profile
– If your code is slow, reduce your memory usage

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 17

 Getting to know MATLAB

Plots
I = [1 2 3 4 5]; % shorthand: I = 1:5;
R = 4.7;
V = I*R;
plot(I,V);
xlabel(‘Current (A)’); % ALWAYS lable axes
ylabel(‘Voltage (V)’);
title(‘Resistor voltage-current relationship’);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 18

 Getting to know MATLAB

Plots
Resistor voltage-current relationship
24

22

20

18

16
Voltage (V)

14

12

10

4
1 1.5 2 2.5 3 3.5 4 4.5 5
Current (A)

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 19

 Getting to know MATLAB

Plots
I = 1:0.1:5; % now in increments of 0.1 A
R1 = 4.7;
V1 = I*R1;
R2 = 10;
V2 = I*R2;
subplot(2,1,1); plot(I,V1);
subplot(2,1,2); plot(I,V2);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 20

 Getting to know MATLAB

Plots
25

20

15
Note:
10
axes
5
have 0
1 1.5 2 2.5 3 3.5 4 4.5 5
different
scaling: 50

difficult to 40

interpret 30

20

10
1 1.5 2 2.5 3 3.5 4 4.5 5

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 21

 Getting to know MATLAB

Plots
I = 1:0.1:5; % now in increments of 0.1 A
R1 = 4.7;
V1 = I*R1;
R2 = 10;
V2 = I*R2;
subplot(2,1,1); plot(I,V1); axis([1 5 0 50]);
ylabel(‘Voltage (V1)’);
title(‘V-I relationship for two resistors’);
subplot(2,1,2); plot(I,V2); axis([1 5 0 50]);
ylabel(‘Voltage (V2)’);
xlabel(‘Current (A)’);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 22

 Getting to know MATLAB

Plots
V-I relationship for two resistors
50

40
Voltage (V1)

30

20

10

0
1 1.5 2 2.5 3 3.5 4 4.5 5

50

40
Voltage (V2)

30

20

10

0
1 1.5 2 2.5 3 3.5 4 4.5 5
Current (A)
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 23
 Getting to know MATLAB

Plots
I = 5:-0.1:1; % now in decreasing steps of 0.1 A
R1 = 4.7;
V1 = I*R1;
R2 = 10;
V2 = I*R2;
plot(I,V1,I,V2); % plots on same axes
legend(‘R1’,’R2’);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 24

 Getting to know MATLAB

Plots
50

45 R1
R2
40

35

30
Voltage (V)

25

20

15

10

0
1 1.5 2 2.5 3 3.5 4 4.5 5
Current (A)
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 25
 Getting to know MATLAB

Plots
I = 5:-0.1:1; % now in decreasing steps of 0.1 A
R1 = 4.7;
V1 = I*R1;
R2 = 10;
V2 = I*R2;
plot(I,V1);
hold;
plot(I,V2); % plots on same axes (alternative)

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 26

 Getting to know MATLAB

Plots
– See also:

– text – good way to put text where you want it in

the figure
– ginput – returns x-y coordinates of wherever you
click in the axes
– plot(x,y,’r’) – plots in red
– plot(x,y,’:’) – plots dotted line

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 27

 Getting to know MATLAB

Matrix/vector definition
x = [4 7 2 1]; % row vector
x = zeros(1,3);
x = ones(1,7);

x = [4; 7; 2; 1]; % column vector

x = zeros(3,1);
x = ones(7,1);

x = [1 3; 4 6; 7 9]; % 3x2 matrix

x = [ones(1,150); zeros(1,150)]; % 2x150 matrix

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 28

 Getting to know MATLAB

Matrix/vector indexing
x = [4 7 2 1];
x(3) % 3rd element of x (=2)

x = [4; 7; 2; 1];
x(3) % 3rd element of x (=2)

x = [1 3; 4 6; 7 9];
x(2,1) % 2nd row, 1st column (=4)
x(2,:) % all of 2nd row (= [4 6])
x(:,1) % all of 1st column (= [1; 4; 7])

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 29

 Getting to know MATLAB

Matrix/vector manipulation
x = [4 7 2 1];
x’ % matrix/vector transpose
% (= [4; 7; 2; 1])
% careful if x is complex (-> conj.)

A = [1 3; 4 6; 7 9; 2 5];
x*A % matrix multiplication
% dimensions must agree
% (= [48 77])

A’*x’ = [48; 77]

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 30

 Getting to know MATLAB

Matrix/vector manipulation: “.”
x = [4 7 2 1]; y = [1 2 3 4];
x.*y % element by element multiplication
% = [4 14 6 4]
2.^y % element by element “^”
% = [2 4 8 16]
1./y % element by element “1/”
% = [1 0.5 0.33 0.25]
(x.^2+y.^2)./(x-2*y)
% element by element operations
% require careful use of “.”

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 31

 Getting to know MATLAB

Loops
x = [4 7 2 1]; y = [1 2 3 4];
for k = 1:4,
z(k) = (x(k)^2+y(k)^2)/(x(k)-2*y(k));
end

% slower than (x.^2+y.^2)./(x-2*y)

% see also: while

% see also: control flow

% if / then / else, case / switch

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 32

 C : A Refresher

Array (vector) operations
#include math.h
void main {
float x[4] = {4,7,2,1},
y[4] = {1,2,3,4}, z[4];

for (k=0; k<4; k++) {

z[k] = (pow(x[k],2)+pow(y[k],2))
/(x[k]-2*y[k]);
printf(“%f ”,z[k]);
} // can use cout for printf in C++
}
// MATLAB: (x.^2+y.^2)./(x-2*y)
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 33
 C : A Refresher

Pointers
void main {
int* x;
int* y;

x = malloc(sizeof(int));
// Allocate an int pointee,
// and set x to point to it
// can use “new” (C++) instead

source: http://cslibrary.stanford.edu/106/
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 34
 C : A Refresher

Pointers
*x = 42; // Dereference x to store
42 in its pointee

*y = 13; // CRASH -- y does not

have a pointee yet

source: http://cslibrary.stanford.edu/106/
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 35
 C : A Refresher

Pointers
y = x; // Pointer assignment
sets y to point to x's pointee

*y = 13; // Dereference y to store

13 in its (shared) pointee

source: http://cslibrary.stanford.edu/106/
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 36
 C : A Refresher

Functions
– Don’t forget to

Declare prototype

Give types for all arguments and outputs
– e.g.
float dist(float x, float y, int d);
float a[5], b[5]; int k = 5;
void main {
dist(a,b,k);
}
float dist(float x, float y, int d) {
// code implementing ‘dist’ function
}

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 37

 Debugging

Writing code is the easy part
– Figuring out why it doesn’t work is the hard part
Some advice:

Understand the big objective of
debugging: Systematically isolating the
point of error
– Try to eliminate parts of the code that could not
have produced the error

Work out what should have happened
where the bug occurred
– Try the code using a simpler input
Adapted from White, A., “Teaching debugging – giving novices expert knowledge”, Artificial Intelligence in
Education, Boulay and Mizoguchi (Eds) , IOS Press, 1997
Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 38
 Debugging

MATLAB will usually locate the problem
line of code

Try executing the code in the command
window (if using a script or function)

Try leaving the “;” off the end of the
command

Prints output to command window

Check the dimensions of g and r

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 39

 MATLAB Debugging Tools

echo
– Displays every command before it is executed

size(x)
– Returns the dimensions of array x

keyboard
– Stops execution, allows input from keyboard
– Afterwards, use return to return to execution

plot – code runs OK, results wrong

pause – need to step through code

Use the debug menu
– Set / clear breakpoints
– Step in and out
– Mainly good practise for similar debuggers in C etc

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 40

 Debugging

Typical problem:
– “I built it and it doesn’t work”

Understand exactly what “working”
means

Break “it” up into smaller subsections

Test each subsection
– With the simplest possible test you can construct

Better still:
– Do this as you create the code
– Test as you go

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 41

 Good programming practise

Why ?
– Good programming is about communication
– Well-written code is more likely to have fewer errors
– Well-written code is easier to debug
– Well-written code is more extensible and re-usable
– Well-written code is easier for someone else to
understand

Very important in industry

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 42

 Good programming practise

Code formatting
– A new level of indentation should be used for every
nested statement
for k = 1:N,
if array[k] > threshold,
disp(‘above’);
else
disp(‘below’);
end
end

– MATLAB has “smart indent” – use it

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 43

 Good programming practise

Variables
– Newly declared variables should have a comment
explaining their use
VC = zeros(1,100); % Array of capacitor voltages
– Use meaningful, descriptive variable names

E.g. VC or CapVolt or CapacitorVoltage for a capacitor voltage
– Don’t change the value of loop variables within a
loop
– Avoid use of global variables

Easy if you use functions
– Pre-allocate arrays to some fixed length

Not critical for MATLAB (no need to declare arrays), but faster
and good practise
– Avoid using i, j – MATLAB complex numbers

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 44

 Good programming practise

Function headers
– All useful functions/subroutines/scripts/methods
have headers
– Should give:

The name of the function

Its purpose

Description of all inputs

Description of all outputs

Author, date

Version number (if you create more than one version)
% function C = lbg(data,csize)
%
% LBG algorithm for codebook design
% data : matrix with training vectors in rows
% csize : desired codebook size (must be a power of 2)
% C : codebook with csize rows
%
% Author: Julien Epps Date: A few years ago
% Reference: Linde, Y., Buzo, A., and Gray, R. M. (1980). "An algorithm for vector quantiser
% design", IEEE Trans. Commun., vol. COM-28, no. 1, pp. 84-95, January.

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 45

 Good programming practise

Use pseudocode
% linfilt – plots various quantities for a digital filter

% Define filter
NumCoeff = [1 -2 1] % numerator coefficients
DenCoeff = 1; % denominator coefficients

% Plot frequency response

freqz(NumCoeff,DenCoeff);

% Plot pole-zero diagram

figure;
zplane(NumCoeff,DenCoeff);

% Plot impulse response

figure;
stem(NumCoeff); % filter is FIR

% Plot step response

figure;
Step = [zeros(1,100) ones(1,100)];
y = filter(NumCoeff,DenCoeff,Step);
plot(y);

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 46

 Good programming practise

Keep your code modular
– Use functions

Bad:
for k = 1:N,
if sqrt(sum((x(k,:)-y).^2)) < min,
d(k) = sqrt(sum((x(k,:)-z).^2));
end
end

Good:
for k = 1:N,
if dist(x(k,:),y) < min,
d(k) = dist(x(k,:),z);
end
end New function dist created by the programmer

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 47

 Good programming practise

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 48

 Good programming practise

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 49

 Good programming practise

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 50

 Good programming practise

In this course:
– 10% of lab, project and assignment marks allocated
to good code formatting and programming practise

Plenty of examples exist, see
http://www.datatool.com/downloads/matlab_style_guidelines.pdf
– 5% of lab and assignment marks allocated to
correct and effective use of plotting

Correctly labelled axes

Axis ranges selected to show interesting part of plot

Good use of MATLAB plotting functions to visualise simulation
results
– e.g. Plotting two curves one same axis for easy
comparison

Only the key essential information shown (not plot after plot of
similar results)

Dr Julien Epps ELEC2146 Electrical Engineering Modelling and Simulation 51

 Object-Oriented Programming

see e.g.
http://www.cs.ubc.ca/~mdunham/tutorial/
Dr Julien Epps objectOriented.html
ELEC2146 Electrical Engineering Modelling and Simulation 52