Matlab DSP

A1
Introduction to MATLAB
and Simulink
Appendix A
This appendix provides a quick reference for using MATLAB with its toolboxes
and Simulink with its blocksets for DSP applications. These tools are used exten-
sively in the experiments and examples in this book. This appendix covers useful
topics related to DSP in MATLAB, the Signal Processing Toolbox, the Filter Design
Toolbox, Simulink, the DSP Blockset, and the Fixed-Point Blockset. More detailed
descriptions are documented in the MATLAB, toolbox, and Simulink users guides
listed in references at the end of this appendix.
A.1 USING MATLAB
MATLAB stands for MATrix LABoratory and is a technical-computing language
that allows the user to perform numerical computation, simulation, acquisition, and
visualization of data, algorithm design, analysis, development, and implementation.
Unlike other high-level programming languages such as C, MATLAB provides a com-
prehensive suite of mathematical, statistical, and engineering functions. The functional-
ity is extended with interactive graphical capabilities for creating plots. Furthermore,
extensive toolboxes are available for working under the MATLAB environment. Tool-
boxes are collections of algorithms, written by experts in their fields, that provide appli-
cation-specific capabilities. These toolboxes enhance MATLABs functionality in
KuoAppAv3.qxd 2/16/04 3:07 PM Page 1
2005 Pearson Education, Inc., Upper Saddle River, NJ. All rights reserved.
For exclusive distribution with the book Digital Signal Processors: Architectures, Implementations, and
Applications by Sen M. Kuo and Woon-Seng Gan, ISBN 0-13-035214-4
A2 Appendix A Introduction to MATLAB and Simulink

signal and image processing, data analysis and statistics, mathematical modeling, con-
trol system design, etc.
In addition to MATLAB and its toolboxes, there is another software package
called Simulink for modeling, simulating, and analyzing dynamic systems. Simulink
is integrated closely with the MATLAB environment. Variables and results derived
from Simulink can be put in the MATLAB workspace for postprocessing and visu-
alization. Like the toolboxes that extend MATLAB functions, many blocksets that
add additional blocks to the Simulink environment are available. These blocksets
include the DSP Blockset, Communication Blockset, and Fixed-Point Blockset.
A.1.1 Startup
To start MATLAB, double-click on the icon on the desktop. A MATLAB win-
dow is displayed, as shown in Fig. A.1. This window provides an integrated environ-
ment for developing MATLAB code. The command window is the main window in
which the user keys MATLAB commands after the prompt Some MATLAB
commands are listed as follows:
1. The help command is used to show a list of programs installed in the MAT-
LAB environment.
2. The help topic command is used to display the usage of a particular MAT-
LAB topic or syntax.
3. The demo command is used to bring up a demo window.
7 7.
Figure A.1 MATLAB environment (version 6.5) using a five-panel desktop layout
Section A.1 Using MATLAB A3

4. The whos command is used to display a list of variables currently loaded into
the workspace.
5. The what dir command is used to display the files in the directory dir.
6. The clear all command is used to clear all of the variables.
7. The diary ('record.out') command is used create a file called record.out
that contains all of the functions entered in the command window. The com-
mand diary('off') turns off the record mode.
8. The quit command is used to exit MATLAB.
In addition to the Command Window shown in the right side of Fig. A.1, there
are Command History and Current Directory windows located in the bottom left of
the screen. The Command History window records all of the executed commands, as
well as the date and time when these commands were executed. This feature is use-
ful in recalling the commands that have been executed previously. The Current
Directory window keeps track of the files inside the current directory. In the upper-
left screen are two default windows: Launch Pad and Workspace. The Launch Pad
window contains all of the installed toolboxes that can be easily launched by dou-
ble-clicking on the desired toolbox. The Workspace window is used to organize the
loaded variables and also displays the size of variables, bytes, and class.
A set of toolbars at the top of the screen performs the following functions:
New file, open file
Cut, copy, paste
Undo last action, redo last action
Simulink library browser (to be discussed in Section A.3)
Open help window
The MATLAB environment also includes the following features under the
File pull-down menu:
1. Import data from a file directory into the MATLAB workspace.
2. Save the MATLAB workspace to a file.
3. Set path, which allows commonly used files in the set directory to be searched.
4. Preference, which allows the user to specify the windows color and font size,
number display formats, editor used, print options, figure copy options, etc.
A.1.2 Useful Syntax
MATLAB uses very simple programming notations to execute mathematical
statements. MATLAB syntax is expressed in a matrix-style operation. The user

TABLE A.1 Some Useful MATLAB Syntaxes and Examples
Input Comment
g is a row vector
p is a column vector
G is a matrix
Performs an inner-product operation
Performs element-by-element multiplication
Performs element-by-element addition
Performs matrix-vector multiplication
r is a six-element
is used to separate the start and end
G(1,:) Selects the first row of G
G(:,1) Selects the first column of G
G(1:2,1:2) Selects the upper submatrix of G
Performs (square of the matrix)
Performs an element-by-element square
[g, g] Concatenates the row vector
[g; g] Places the next vector in the next row
ones (1,3) Creates a vector of all
zeros (2,3) Creates a matrix of all
save result Saves all variables to file result.mat
save result1 x,y,z Saves variables x, y, and z to the file result1.mat
load result.mat Loads all variables from the file result.mat
clear all Clears all of the workspace variables
elements = 0 2 * 3
elements = 1 1 * 3
G.
2
G
2
G
2
2 * 2
vector = [0 2 4 6 8 10] r = 0:2:10
G * g
g+g
g. * g
g * g'
3 * 3 G = [1 2 3; 4 5 6; 7 8 9]
13 * 12 p = [1 2 3]'
11 * 32 g = [1 2 3]
can represent and manipulate scalars, vectors, and matrices in the MATLAB
workspace, as shown by the examples summarized in Table A.1.
A.1.3 Plots
MATLAB is known for its outstanding graphics capabilities. Both 2-D and 3-D
graphics can be plotted and manipulated using some commonly used commands
summarized in Table A.2.
MATLAB version 6 (or higher) provides a set of powerful editing tools for
editing graphics. It supports a point-and-click editing mode to modify the plots.

TABLE A.2 MATLAB Commands and Examples for 2-D and 3-D Graphics
Input Comment
Sets up the time axis
Creates a vector x
plot(t,x); Line plot of the x-axis (t) vs. the y-axis (x)
plot(t,x,'ro') Line plot using a red circle
figure Creates a new figure plot
stem(t,x) Discrete sequence plot terminated with an
stem(t,x, 'filled') empty circle or full circle ('filled')
subplot(1,2,1); Display window divided into one row
plot(t,x); and two columns
subplot(1,2,2); Note: of rows, number
plot(x,t); of columns, index)
close all Closes all of the figures
Time index from 0 to 50,
Generates two sinewaves
plot (x,y1,x,y2) Plots two sinewaves in the same plot
plot (x,y1, 'r'); hold on; Plots two sinewaves, one at a time in the
plot (x,y2, 'g'); hold off; same plot
Sets up the x-axis and y-axis limit
grid on Shows grid lines
xlabel('time') Specifies the name of the x-label
ylabel('amplitude') Specifies the name of the y-label
title('Two sine waves') Specifies the title of the plot
Prepares the coordinates
Performs a 3-D plot whose coordinates are
elements of x, y, and z
axis square; grid on Changes to a square axis plot with a grid
Transforms the vector into an array X, Y
Generates another array Z
mesh (X, Y, Z); Mesh surface plot
surf (X, Y, Z); Surface plot
meshc (X, Y, Z); Surface plot with a contour beneath
meshz (X, Y, Z); Surface plot with a curtain
pcolor (X, Y, Z); Pseudo-color plot
surfl (X, Y, Z); 3-D-shaded surface with lighting
Z = X. * exp1-X.
2 - Y.
22;
[X,Y] = meshgrid 1[-2:0.1:2]2;
plot31sin1t2, cos1t2,t2
t = 0:pi/50:5 * pi
axis1[0 10 -1 1]2
y2 = cos12 * pi * x/82;
y1 = sin12 * pi * x/82;
step = 1 x = 0:1:50;
1x,y,z2 = 1number
x = exp1-10. * t2;
t = 0:0.001:1;
Graphical objects can be selected by clicking on the object, and multiple objects can
be selected by shift-clicking. Several editing features are highlighted in Fig. A.2.
Besides selecting, moving, resizing, changing color, cutting, copying, and past-
ing graphic objects, MATLAB also provides tools for zoom-in, zoom-out, camera

Insert line
Insert arrow
Insert text
Select object in graph
Figure A.2 Example of editing graphics
settings, basic data fitting, and displaying data statistics. These tools can be executed
from the Tools pull-down menu. Once the diagram has been edited, the user can
export the diagram into different graphical file types such as .bmp, .jpeg, etc., or
save the diagram as a .fig file. A more detailed explanation can be found in the
manual Using MATLAB Graphics [2].
A.1.4 Programming
We can enter and execute MATLAB commands directly in the command window.
Since we may need to reuse these MATLAB commands and use several com-
mands in sequence for a specific task, a better way is to use the MATLAB M-file
editor (or another text editor) to write the MATLAB code (consisting of a
sequence of commands) and save it into an M-file with an extension .m. The M-file
editor can be activated by clicking on either the New M-file icon or the Open
File icon .
There are two types of M-file: scripts and functions. Comments that start with
a percent sign (%) can be inserted anywhere inside a function or script file. A script

file has no input and output arguments and contains a sequence of MATLAB
statements that make use of workspace data globally. To execute the M-file, simply
type the name of the M-file in the command window and press Enter or Return. The
user can interrupt a running program by pressing or at any time.
A function file must contain the word function in the first line of the program.
For example,
function [mean,stdev] = stat(x)
% STAT statistics of mean and standard deviation
n = length(x);
mean = sum(x)/n;
stdev = sqrt(sum((x-mean).^2)/n);
As shown in the example, a function can contain multiple output variables, which are
enclosed in square brackets. Input arguments are enclosed in parentheses. All vari-
ables inside the function file are local and are not shared with the calling workspace.
Only the M-file function can be converted to a C (or ) file. A script M-file
cannot be compiled into a C file and thus needs to be converted to a function M-file
first. Other M-files that cannot be compiled include an M-file that uses objects, an
M-file that uses input and eval to manipulate a workspace, and MATLAB built-in
functions. MATLAB supports C and compilers such as (1) a C compiler (LCC
C version 2.4), (2) Watcom versions 10.6 & 11.0, (3) Borland versions
5.0 onward, and (4) Microsoft Visual versions 5.0 and 6.0. The compiler can be
set up by issuing the following statement: mex -setup.
The compiler for building external interface files [MATLAB executable
(MEX)] is chosen from the list of installed C compilers. The user can perform the
following conversions:
To convert an M-file to a C-file and create a C MEX-file, use
mcc x <M-filename>
To convert an M-file to a C-file and create a Simulink S-file, use
mcc S <M-filename>
To convert an M-file to a C-file and create a stand-alone C program, use
mcc m <M-filename>
To convert an M-file to a C-file and create a stand-alone application, use
mcc p <M-filename>
A.1.5 Data Types
It is important to understand the data types used in MATLAB since they affect the
precision, the dynamic range, and quantization errors in the computation. By default,
all MATLAB computations are carried out in double-precision, floating-point format
C+ +
C+ +
C+ + C/C+ +
C+ +
C+ +
Crtl +Break Ctrl -C

TABLE A.3 Data Types Used in MATLAB (adapted from [1])
Data type Description
single Single precision, floating-point (32 bits)
double Double precision, floating-point (64 bits)
int8, uint8 Signed, unsigned integer (8 bits)
char Character array (or string)
cell Cell array has elements that contain other arrays
structure Structure array has field arrays and contains
other arrays
user class User-specified class
(64 bits). However, data can be stored in single precision (32 bits) or integer (8 bits, 16
bits, and 32 bits) format to reduce memory requirements. Table A.3 lists some funda-
mental data types.
Using the data type as a prefix for numbers or variables can specify the preci-
sion. For example,
a = [1 2 3] % a double precision array (default)
b = single(a) % convert to single precision array
c = uint8(a) % convert to unsigned 8-bit integer
Note that we can find the number of bytes used in representing the array by looking at
the workspace window. Besides representing numbers in the required format, we can
also display the numeric format without converting the data by specifying the display
format in .
MATLAB allows data to be grouped under a single variable known as the cell
array. For example, we can group three different variables as an array X and specify
the cell array using curly braces as follows:
X = {[1 2 3], 'hello', eye(3)} % define a cell array
X{1} % extract [1 2 3]
X{2} % extract 'hello'
An alternate form for specifying the preceding cell array by name is to use the
structure data type as follows:
X.num = [1 2 3]
X.char = 'hello'
X.matrix = eye(3)
56
File : Preference : Command Window : Text display : Numeric format

TABLE A.4 MATLAB Functions for Matrix
and Linear Algebra (adapted from [1])
Function Comment
norm Matrix or vector norm
rank Matrix rank
det Determinant of matrix
trace Sum of diagonal elements
null Null space
orth Orthogonalization
A.1.6 Useful Commands
Many useful MATLAB functions are commonly used for matrix and linear-algebra
operations, as summarized in Table A.4. MATLAB also provides several Fourier
analysis commands for signal processing and data analysis. These functions are listed
in Table A.5.
More signal-processing functions are discussed in the next section on MATLAB
toolboxes.
TABLE A.5 Fourier Analysis Function (adapted from [1])
Function Comment
fft FFT
fft2 Two-dimensional FFT
fftn N-dimensional FFT
ifft IFFT
ifft2 Two-dimensional IFFT
ifftn N-dimensional IFFT
abs Magnitude
angle Phase angle
unwrap Unwraps a phase angle in radian
fftshift Moves a zeroth lag to the center of the
spectrum
cplxpair Sorts numbers into complex-conjugate
pairs
nextpow2 Next higher power-of-two number

A.2 USING DIGITAL SIGNAL PROCESSING TOOLBOXES
AND INTERACTIVE TOOLS
As mentioned earlier, a MATLAB toolbox contains many application-specific M-files
to solve a particular problem. Commonly used DSP toolboxes include the Signal Pro-
cessing Toolbox [3], Filter Design Toolbox [4], Communication Toolbox [5], Image
Processing Toolbox [6], Wavelet Toolbox [7], etc. This section summarizes the signal
processing functions and tools in the Signal Processing Toolbox and Filter Design
Toolbox. The reader can refer to the user guides for more details.
A.2.1 Signal Processing Toolbox
The Signal Processing Toolbox contains many functions that perform signal-processing
algorithms, such as filter design and implementation, spectral analysis, windowing,
statistical signal processing, transforms, multirate signal processing, waveform gen-
eration, and other operations. In addition to these DSP functions, the toolbox also
contains two useful interactive tools: (1) the Signal Processing Tool, which provides
interactive tools for analyzing and filtering signals, and (2) the Filter Design and
Analysis Tool, which provides advanced filter-design tools for designing digital fil-
ters, quantizing filter coefficients, and analyzing quantization effects. These tools are
explained in the following sections.
Table A.6 lists some important signal-processing functions in the Signal Pro-
cessing Toolbox. This list is grouped under different categories. To learn about the
detailed usage of these functions, simply type
help function_name
to display the help menu for that particular function.
The signal-processing functions summarized in Table A.6, together with the
powerful graphical tools, provide a comprehensive tool for signal processing. The
Signal Processing Toolbox further provides an interactive tool that integrates the
functions with the GUI, a topic which is discussed in the next section.
A.2.2 Signal Processing Tool
SPTool provides several tools for use in analyzing signals, designing and analyzing
filters, filtering signals, and analyzing the spectrum of signals. The user can open this
tool by typing
sptool
in the MATLAB command window. The SPTool main window appears, as shown in
Fig. A.3.
Four windows can be accessed within SPTool:
1. The Signal Browser is used to view input signals. Signals from the workspace
or file can be loaded into SPTool by clicking on . An Import to SP-
Tool window appears, which allows the user to select data from the file or
File :Import
Section A.2 Using Digital Signal Processing Toolboxes and Interactive Tools A11
TABLE A.6 Signal-Processing Functions (adapted from [3])
Filter analysis Description
filternorm Two-norm or inf-norm of a digital filter
freqs Frequency response of an analog filter
freqspace Frequency spacing for a frequency response
freqz Frequency response of a digital filter
freqzplot Plots frequency-response data
grpdelay Group delay of a filter
impz Impulse response of a digital filter
unwrap Unwraps a phase angle
zplane Zero-pole plot
Filter
implementation Description
conv Convolution and polynomial multiplication
conv2 2-D convolution
deconv Deconvolution and polynomial division
fftfilt FFT-based FIR filtering using overlap-add
filter FIR or IIR filtering
filter2 2-D digital filtering
filtfilt Zero-phase digital filtering
filtic Initial condition of a transposed form-II IIR filter
latcfilt Lattice and lattice-ladder filtering
medfilt1 1-D median filtering
sgolayfilt SavitzkyGolay filtering
sosfilt Second-order (biquad) IIR filtering
upfirdn Upsample, FIR filtering, and downsample
FIR filter design Description
convmtx Convolution matrix
cremez Complex and nonlinear-phase equiripple FIR
fir1 Window-based FIR-filter design
fir2 Frequency sampling-based FIR-filter design
fircls Constrained least-square FIR-filter design (multiband),
fircls1 (lowpass and highpass)
firls Least-square linear-phase FIR-filter design
firrcos Raised-cosine FIR-filter design
intfilt Interpolation FIR-filter design
kaiserord FIR-filter design using a Kaiser window
remez Computes a ParksMcClellan optimal
remezord FIR-filter design and filter-order estimation
sgolay SavitzkyGolay filter design
IIR filter design Description
bilinear Bilinear transformation
butter Butterworth filter design
buttord Order and cutoff frequency for a Butterworth filter
(Continued)

IIR filter design Description
cheby1 Chebyshev Type I filter design
cheb1ord Order for a Chebyshev Type I filter
cheby2 Chebyshev Type II filter design
cheb2ord Order for a Chebyshev Type II filter
ellip Elliptic filter design
ellipord Minimum order for an elliptic filter
impinvar Impulse-invariance method
maxflat Generalized-digital Butterworth filter design
prony Pronys method for IIR-filter design
stmcb Linear model using a Steiglitz-McBride iteration
yulewalk Recursive digital-filter design (least-square)
Linear system
transform Description
latc2tf Converts lattice parameters to a transfer function
polystab Stabilizes a polynomial
polyscale Scales the roots of a polynomial
residuez z-transform partial-fraction expansion
sos2ss Converts a second-order section to a state-space form
sos2tf Converts a second-order section to a transfer function
sos2zp Converts a second-order section to a zero-pole-gain form
ss2sos Convert state-space parameters to a second-order form
ss2tf Converts a state-space to a transfer function
ss2zp Convert state-space parameters to a zero-pole-gain
tf2latc Converts a transfer function to a lattice-filter form
tf2sos Converts a transfer function to a second-order section
tf2ss Converts a transfer function to a state-space
tf2zp Converts a transfer function to a zero-pole-gain
zp2sos Converts a zero-pole-gain to a second-order form
zp2ss Converts a zero-pole-gain to a state-space form
zp2tf Converts a zero-pole-gain to a transfer function
Windows Description
bartlett Bartlett window
barthannwin Modified BartlettHanning window
blackman Blackman window
blackmanharris Minimum four-term BlackmanHarris window
bohmanwin Bohman window
boxcar Rectangular window
chebwin Chebyshev window
gausswin Gaussian window
hamming Hamming window
hann Hann (Hanning) window
Kaiser Kaiser window
nuttallwin Nuttall-defined four-term BlackmanHarris
TABLE A.6 (Continued)
(Continued)
Windows Description
triang Triangular window
tukeywin Tukey window
window Window-function gateway
Transforms Description
bitrevorder Permutes input into a bit-reversed order
czt Chirp z-transform
dct Discrete cosine transform
dftmtx DFT matrix
fft 1-D FFT
fft2 2-D FFT
fftshift Moves zero-th lag to the center of the spectrum
goertzel Second-order Goertzel algorithm
hilbert Computes an analytic signal using a Hilbert transform
idct Inverse-discrete cosine transform
ifft 1-D IFFT
ifft2 2-D IFFT
Cepstral analysis Description
cceps Complex cepstral analysis
icceps Inverse-complex cepstrum
rceps Real cepstrum, minimum phase reconstruction
Statistical and
spectrum analysis Description
cohere Estimates the magnitude-square coherence function
corrcoef Correlation coefficients
corrmtx Autocorrelation matrix
cov Covariance matrix
csd Cross-spectral density
pburg Power-spectrum density (PSD) estimate using Burgs method
pcov PSD estimate using a covariance method
peig PSD estimate using an eigenvector method
periodogram PSD estimate using a periodogram method
pmcov PSD estimate using a modified covariance method
pmtm PSD estimate using a Thomson multitaper method
pmusic PSD estimate using the MUSIC method
psdplot Plots PSD data
pwelch PSD estimate using Welchs method
pyulear PSD estimate using the YuleWalker autoregressive method
rooteig Frequency and power estimation using the eigenvector algorithm
rootmusic Frequency and power estimation using the MUSIC algorithm
tfe Transfer function estimate
xcorr Crosscorrelation function
xcorr2 2-D crosscorrelation
xcov Covariance function
(Continued)

Linear prediction Description
ac2rc Autocorrelation sequence to reflection coefficients
ac2poly Autocorrelation sequence to a prediction polynomial
is2rc Inverse sine parameters to reflection coefficients
lar2rc Log area ratios to reflection coefficients conversion
levinson LevinsonDurbin recursion
lpc Linear-predictive coefficients using autocorrelation
lsf2poly Line-spectral frequencies to prediction polynomial
poly2ac Prediction polynomial to an autocorrelation sequence
poly2lsf Prediction polynomial to line-spectral frequencies
poly2rc Prediction polynomial to reflection coefficients
rc2ac Reflection coefficients to an autocorrelation sequence
rc2is Reflection coefficients to inverse-sine parameters
rc2lar Reflection coefficients to log-area ratios
rc2poly Reflection coefficients to a prediction polynomial
rlevinson Reverse LevinsonDurbin recursion
schurrc Schur algorithm
Multirate signal
processing Description
decimate Resamples at a lower sampling rate (decimation)
downsample Downsamples an input signal
interp Resamples data at a higher sample rate (interpolation)
interp1 General 1-D interpolation
resample Changes the sampling rate by any rational factor
spline Cubic spline interpolation
upfirdn Upsample, FIR filtering, down sample
upsample Upsample input signal
Waveform
generation Description
chirp Swept-frequency cosine
diric Dirichlet or periodic-sinc function
gauspuls Gaussian-modulated sinusoidal pulse
gmonopuls Gaussian monopulse
pulstran Pulse train
rectpuls Sampled aperiodic rectangle
sawtooth Sawtooth (triangle) wave
sinc Since function
square Square wave
tripuls Sampled aperiodic triangle
vco Voltage-controlled oscillator
workspace. To view the signal, simply highlight it and click on View. The Signal
Browser window, shown in Fig. A.4., allows the user to zoom-in and zoom-out
from the signal, read the data values via markers, display the format, and even
play the selected signal using the computers speakers.
Figure A.3 SPTool window
Figure A.4 Signal Browser window
2. The Filter Designer is used for designing digital FIR and IIR filters. The user
simply clicks on the New icon for a new filter or the Edit icon for an existing fil-
ter under the Filter column in SPTool to open the Filter Designer window, as

Figure A.5 Filter Designer window
shown in Fig. A.5. The user can design lowpass, highpass, bandpass, and
bandstop filters using different filter-design algorithms. In addition, the user
can also design a filter using the Pole/Zero Editor to graphically place poles and
zeros in the z-plane. A useful feature is the ability to overlay the input spec-
trum onto the frequency response of the filter by clicking on the Frequency Mag-
nitude/Phase icon .
3. Once the filter has been designed, frequency specification and other filter
characteristics can be verified by using the Filter Viewer. Select the name of the
designed filter and click on the View icon under the Filter column in SPTool to
open the Filter Viewer window, as shown in Fig. A.6. The user can analyze the
filter in terms of its magnitude response, phase response, group delay, zero-
pole plot, impulse response, and step response.
When the filter characteristics have been confirmed, the user can then
select the input signal and the designed filter. Click on the Apply button to per-
form filtering and generate the output signal. The Apply Filter window appears,
as shown in Fig. A.7, and allows the user to specify the file name of the output
signal. The Algorithm list provides a choice of several filter structures.
4. The final GUI window is the Spectrum Viewer, as shown in Fig. A.8. The user
can view existing spectra by clicking on file names and then on the View button.
Select the signal and click on the Create button to view the Spectrum Viewer
window. The user can select one of the many spectral-estimation methods,
such as Burg, covariance, FFT, modified covariance, MUSIC, Welch, Yule-
Walker autoregressive (AR), etc., to implement the spectrum estimation. In
addition, the size of the FFT, window functions, and overlapping samples can
be selected to complete the power-spectrum density (PSD) estimation.
SPTool also provides a useful tool for exporting signals, filter parameters, and
spectra to the MATLAB workspace or files. These saved parameters are represent-
ed in MATLAB as a structure of signals, filters, and spectra. More information can
be found in the Signal Processing Toolbox Users Guide [3]. A step-by-step example
of using SPTool in designing an IIR filter is given in Chapter 2.
A.2.3 Filter Design Toolbox
The Filter Design Toolbox is a collection of tools that provides advanced techniques
for designing, simulating, and analyzing digital filters. It extends the capabilities of
Figure A.6 Filter Viewer window
Figure A.7 Apply Filter window

the Signal Processing Toolbox by adding filter structures and design methods for
complex real-time DSP applications. This toolbox also provides functions that sim-
plify the design of fixed-point filters and the analysis of quantization effects.
The toolbox provides the following key features for digital-filter designs:
1. Advanced FIR filter-design methods: The advanced equiripple FIR design auto-
matically determines the minimum filter order required. It also provides con-
strained-ripple, minimum-phase, extra-ripple, and maximal-ripple designs. In
addition, the least P-th norm FIR design allows the user to adjust the tradeoff be-
tween minimum-stopband energy and minimum order equiripple characteristics.
2. Advanced IIR filter-design methods: Allpass IIR filter design with arbitrary
group delay enables the equalization of nonlinear group delays of other IIR
filters to obtain an overall approximate linear-phase passband response.
Least-Pth-norm IIR design creates optimal IIR filters with arbitrary magni-
tude, and constrained least-P-th-norm IIR design constrains the maximum
radius of the filter poles to improve the robustness of the quantization.
3. Quantization: The toolbox provides quantization functions for signals, filters,
and FFTs. It also supports quantization of filter coefficients, including coeffi-
cients created using the Signal Processing Toolbox.
4. Analysis of quantized filters: The toolbox supports analysis of the frequency
response, zero-pole plot, impulse response, group delay, step response, and
phase response of quantized filters. In addition, it supports limit-cycle analysis
for fixed-point IIR filters.
It is important to emphasize that the Filter Design Toolbox supports design-
ing, simulating, and analyzing fixed-point filters for a wide precision range. It also
Figure A.8 Spectrum Viewer window
allows the user to compute quantized FFTs and IFFTs. These functions ease the task
of determining the effects of quantization on real-world designs. Quantization tools
allow the user to model the behavior of fixed-point filters and FFT algorithms pre-
cisely. The quantized algorithms in the toolbox exactly match the output of the algo-
rithms implemented on a fixed-point processor because the simulation is bit-true.
The Filter Design Toolbox includes a new GUI tool called FDATool. This tool
allows the user to design optimal FIR and IIR filters from scratch, import previous-
ly designed filters, quantize floating-point filters, and analyze quantization effects.
This tool is introduced in the next section.
A.2.4 Filter Design and Analysis Tool
This interactive tool provides several advanced techniques that support designing,
simulating, and analyzing fixed-point and floating-point filters for a wide range of
precision. This tool performs the following functions:
1. Designs filters
2. Converts filters between different structures
3. Quantizes filters
4. Quantizes data
5. Quantizes FFT and IFFT
6. Designs adaptive filters
In addition to the existing functions listed in Table A.6, FDATool contains the addi-
tional filter-design functions listed in Table A.7.
FDATool can be activated by typing
fdatool
in the MATLAB command window. The Filter Design & Analysis Tool window is
shown in Fig. A.9. This window includes tools similar to those shown in the Filter
TABLE A.7 Additional Filter Design Functions in FDATool
Filter design function Description
firlpnorm Designs minimax FIR filters using the
least-Pth algorithm
gremez Designs optimal FIR filters (Remez
exchange)
iirgrpdelay Designs IIR filters (specifies group delay
in the passband)
iirlpnorm Designs minimax IIR filters using the
least-Pth algorithm
iirlpnormc Designs minimax IIR filters using the
least-Pth algorithm, which restricts filter
poles and zeros within a fixed radius

Figure A.9 Filter Design & Analysis Tool window
Designer window in SPTool (Fig. A.5). The design steps and available features that
can be used to view filter characteristics are also similar. However, FDATool is a
more advanced filter-design tool that includes additional filter types such as differ-
entiator, Hilbert transform, multiband, arbitrary magnitude and group delay,
Nyquist, and raised-cosine. FDATool also has an additional option that allows the
default filter structure (direct-form II transposed) to be converted to different struc-
tures, such as direct-form I, direct-form II, direct-form I transposed, state-space, and
its lattice equivalents, as shown in Fig. A.10.
In addition, FDATool is a very powerful tool for investigating quantized filters.
Once a filter has been designed and verified, we can turn on the quantization mode
by clicking on the Set Quantization Parameters icon located at the bottom-left
side of window and then check the box Turn quantization on. The bottom panel of
the Filter Design & Analysis Tool window changes, as shown in Fig. A.11. We can
select (1) Mode, (2) Round mode, (3) Overflow mode, and (4) Format for represent-
ing coefficients, input, output, multiplicands, products, and sums of the filter. The
options and explanations for setting these quantizer properties are listed in Table A.8.
This new panel also provides a useful option for limiting filter coefficients to less
than 1 by scaling the input of the filter, which can be accomplished by clicking on the
Figure A.10 Convert Structure window
Optimization button (the meanings of these options are explained in Chapter 7). Fur-
thermore, the direct-form IIR filter can also be converted to a cascade of second-
order sections for a more stable implementation. This option can be activated by
clicking on to Second-Order sections. This option is only applicable
to IIR filters and is explained in more detail in Chapter 7.
In addition to these features, the new version of FDATool also provides an
option for frequency transformation, which can be accessed by clicking on the
Transform Filter icon . This option transforms the original designed filter to
another filter with different characteristics. It also contains a Filter Realization Wiz-
ard (Realize Model) that allows the user to create a fixed-point or floating-point
filter blockset, which can be used in the Simulink environment. Furthermore, filters
can also be imported from variables or discrete-time filter objects in the workspace
by clicking on the Import Filter icon .
Besides using FDATool in specifying the quantized filter, we can also con-
struct a quantized filter object using the function qfilt in MATLAB. For example,
Hq=qfilt('quantizer',{'float',[32 8],'round'})
Edit :Convert
Figure A.11 Setting quantization parameters in FDATool

TABLE A.8 Quantizer Properties
Quantizer property Description
Mode:
fixed Specifies fixed-point arithmetic (default)
ufixed Unsigned fixed-point calculations
float Specifies floating-point arithmetic
double Specifies double-precision, floating-point arithmetic
single Specifies single-precision, floating-point arithmetic
Round mode:
ceil Rounds a value to the nearest integer towards
convergent As in ceil (If tie, round down if the next-to-last bit is even;
up if odd)
fix Rounds a value to the nearest integer toward 0
floor Rounds a value to the nearest integer toward
round Rounds a value to the nearest integer (default): rounds a
negative number toward a positive number toward
and ties toward
Overflow mode:
saturate Sets the overflowed values to the maximum or minimum
values (default)
wrap Maps overflow values to the number range using modular
arithmetic
Format:
[wl,fl] (fixed) Default value is [16 15] for Q.15
Maximum wordlength wl = 53 bits, fl: fractional length
[wl,exp] (float) exp up to 11 bits; 64-bit >wl>exp, exp: exponent length
[32,8] (single) IEEE-754 single precision (exp = 8, fl = 23, sign = 1)
[64,11] (double) IEEE-754 double precision (exp = 11, fl = 52, sign = 1)
+
q
+
q
,
-
q
,
-
q
+
q
creates a quantized-filter object that is quantized to single-precision, floating-point
format and that uses round mode for rounding coefficients and arithmetic results.
The object is defined as a structure data type, and direct referencing can modify its
properties after the object is constructed. For example, to change the value of the
scaling factor at the input of the filter to 0.5, we can use the following command:
Hq.scalevalues = 0.5
FDATool also allows the user to construct the other quantized objects listed in
Table A.9. A detailed explanation can be found in the Filter Design Toolbox Users
Guide [4].
A useful report showing the minimum, maximum, number of overflows, num-
ber of underflows, and number of operations of the most recent application of F
(quantized filter object, quantized FFT object, or quantizer object) can be produced
by issuing the following command:
qreport(F)
Section A.3 Using Simulink A23

TABLE A.9 Quantizer, Quantized Filter, and Quantized FFT
Quantized function Description
qfft Constructor for quantized FFT objects
qfilt Constructor for quantized filter objects
quantizer Constructor for quantizer objects
unitquantizer Constructor for unit-quantizer objects
Finally, the latest FDATool also provides a set of adaptive-filter functions using dif-
ferent adaptive algorithms for updating the filter coefficients listed in Table 9.1.
Examples of using adaptive-filter functions are given in Chapter 9.
A.3 USING SIMULINK
Simulink provides a useful interactive interface for use in designing dynamic sys-
tems. This section introduces some basic functional blocks and their operations.
Detailed information can be found in the Simulink: Users Guide [8]. To start
Simulink from the MATLAB environment, type the following command:
simulink
or click on the Simulink icon at the top of the MATLAB window. A Simulink
Library Browser appears, which contains the main Simulink blocks and all of the
available Simulink blocksets, as shown in Fig. A.12. In this section, we examine the
main Simulink Block Library. The DSP Blockset [9] and the Fixed-Point Blockset
[10] are explained in Section A.4.
Right-click on Simulink, shown in Fig. A.12, and click on the menu Open the
Simulink Library to open the Simulink Library window, as shown in Fig. A.13. The
functional blocks (libraries) include the following:
1. Continuous contains a set of blocks that perform numeric derivative, integra-
tion, delay, etc.
2. Discrete contains a set of blocks that perform discrete-dynamic systems, zero-
order hold, delay, etc.
3. Look-up Tables contains a set of blocks that perform table lookups.
4. User-Defined Functions contains a set of MATLAB functions, S-functions, etc.
5. Math operations performs math functions such as scaling, dot product, prod-
uct, trigonometric operations, and logical operations, etc.
6. Discontinuities contains a set of blocks for quantizer, saturation, relay, switch,
nonlinear functions, etc.
7. Signals Routing manipulates the signal flow, storage and reading of data, etc.
8. Signal Attributes determines the data-type conversion line probing for data
width and sample time, specify attribute of signal line, etc.

Figure A.12 Simulink
Library Browser window
9. Sinks contains a set of destination blocks such as scope, files, workspace,
graphs, etc.
10. Sources contains all source elements such as signal generator, random num-
ber, noise, clock, etc.
11. Port & Subsystems contains useful blocks that allow external trigger, condi-
tional switching, iterative system, subsystem clock, I/O port, etc.
12. Model Verification checks the dynamic range of signals, input resolution, and
static upper and lower bounds, etc.
13. Model-Wide Utilities provides documentation of model and linearization of
running model.

Figure A.13 Simulink Library window
To begin the block-building process, simply click on the Create a new model
icon to open a new worksheet. In order to use these functional blocks, click on
the selected block icon and drag it into the opened worksheet. Most blocks (except
the sink and source blocks) have at least one input and one output node. We can con-
nect two blocks by clicking on the output node of the first block and dragging the
mouse to the input node of the second block. A quicker way of connecting two blocks
is to click on the first (source) block and hold down the Ctrl key while also clicking
on the second (destination) block. These two blocks are connected automatically.
In this section, we use a simple example to show the processes of designing
and building a simple dynamic digital system in Simulink from scratch. The com-
plete system is shown in Fig. A.14. The step-by-step procedure of building this com-
plete design involves the following steps:
Step 1: Click on in the Simulink Library Browser, or sim-
ply click on the Create a new model icon to open a new worksheet.
Step 2: Select a sinusoidal source by double-clicking on DSP Blockset, and
then click on DSP Sources. Select Sine Wave and drag it into the work-
sheet. Double-click on this block and set the following parameters:
Step 3: Select a white-noise source by clicking on Random Source (in
) and drag it into the worksheet. Dou-
ble-click on this block, and set the following parameters: Source
and Sample
Step 4: Attach a gain block to the output of each source. The gain block is
located in . Drag two gain blocks
into the worksheet, set the gain for the sinusoidal source to 1, and set
the gain for the noise source to 0.1. Link the output of the source to
the input of the gain block by clicking on the output node of the
Simulink :Math Operations :Gain
1/10,000.
time = Variance = 1, Mean = 0, Type = Gaussian,
DSP Blockset :DSP Sources
Amplitude = 1, Frequency = 1,000, and Sample time = 1/10,000.
File :New :Model

source block. Then, complete the connection by dragging the mouse
to the input node of the gain block, as shown in Fig. A.14.
Step 5: Combine the two signal sources using a summing block. Click on Sum
(in ) to select the summing block, and
drag it into the worksheet. Connect it to the outputs of the Gain
blocks using the method described in Step 4.
Step 6: Apply a digital bandpass filter to the output of the summing junction
to enhance the sinewave and attenuate the noise. The digital filter is
located at DSP Blockset. Click on Filtering, and double-click on Filter
Designs. Drag Digital Filter Design (FDATool block) to the worksheet
and double-click on the block to open a window similar to that shown
in Fig. A.9. We can specify the following parameters for designing the
bandpass filter: Filter
Equiripple,
and After entering all of the
parameters, click on the Design Filter button to start the filter design
and implementation.
Step 7: Examine the signals before and after the bandpass filter using a
Spectrum Scope. This scope can be found in
. Drag it to the worksheet. We can also view the output
from the summing junction (noisy sinewave) by using another
Spectrum Scope. Drag the second Spectrum Scope into the work-
sheet, click on the input node, and drag the mouse to the line between
the summing junction and the bandpass filter. Enable the buffer input
option of the Spectrum Scope, and use the default FFT length. This
configuration allows the user to compare the difference between the
original signal and the filtered signal.
Step 8: Set the Simulink parameters before running the simulation once
the Simulink model has been completed, as shown in Fig. A.14.
Click on Simulation and select Simulation parameters from the
menu to open the window shown in Fig. A.15. Set the parameters as
shown in Fig. A.15 for a 10-second simulation at a sampling fre-
quency of 10,000 Hz. Close the window and start the simulation by
clicking on the button . Two spectrum plots are displayed that
show that the processed signal has a 40 dB reduction in noise
power. To stop the simulation, simply click on the button .
We can save the simulation model shown in Fig. A.14 to a file bpf.mdl and use
the file later in the MATLAB environment by typing
bpf
In the second example, we open a dynamic system (from Simulink demo [8])
by typing
combfilter
DSP Sinks
DSP Blockset :
Astop2 = 40 dB. Apass = 1 dB, 40 dB,
Astop1 = Fstop2 = 1,100, Fpass2 = 1,050, Fpass1 = 950, 900,
Fstop1 = Fs = 10,000, Filter order = minimum order,
Design Method = FIR : Type = Bandpass,
Simulink :Math Operations

Figure A.14 A simple Simulink model for digital filtering
This Simulink model is shown in Fig. A.16. This example illustrates the differences
between digital filtering using single (upper path) and double (lower path) preci-
sions. We can click on the block and drag it around. Double-click on any block to
show either the internal blocks detail or open up the blocks parameters window.
For example, double-click on the Signal Waveform block to open the Block Parame-
ters menu, which allows the user to adjust the waveform type, amplitude, and fre-
quency of the signal. The user can change the default 1 Hz sinewave into a 2 Hz
squarewave with an amplitude of 50.
Figure A.15 Simulation
parameters to run the filter
simulation

Figure A.16 A Simulink demo showing single- and double-precision filtering
Several blocks can be grouped together to form a subsystem. The subsystem
can be created by selecting multiple blocks using Shift-click or by using the mouse
cursor to form a bounding box that includes these blocks. We then select Create sub-
system under the Edit menu to form a subsystem. The details of the subsystem can
be viewed by double-clicking on it. For example, in the preceding Simulink demo,
we can select the Signal Scaling block and the Multiply block to form a subsystem.
Double-click on this new subsystem to view the internal details.
The subsystem can be masked to allow the user to customize the dialog box
and icon for this subsystem. Click on the subsystem, and select Mask Subsystem
under the Edit menu. A Mask Editor window is opened, as shown in Fig. A.17. Click
on the Parameters pane, and then click on the Add icon to specify the attributes
of mask parameters including prompt, variables, and control type. In this example,
the parameters constant and c are specified in the Prompt and Variable fields,
respectively. Make sure that the Type entry is selected as edit and that both Evaluate
and Tunable are turned on. The block description is defined in the Documentation
pane, as shown in Fig. A.18. The Documentation menu specifies the dialog box once
the subsystem block is double-clicked.
Before starting Simulink, the simulation parameters must be specified by
selecting Simulation Parameters under the Simulation menu. The Simulation Para-
meters window appears, as shown in Fig. A.19. The start and stop time for the simu-
lation can be specified in the Start time (0 sec) and Stop time (4 sec) fields. Simulink
provides a number of solvers for the simulations. In DSP, the solver option is set to
Fixed-step (take the same step size during simulation) and discrete (no continuous
states). The Fixed step size field can be set to auto or to the sampling period of the

Figure A.18 Block documentation
Figure A.17 Mask editor
window
DSP system. Finally, the mode is set to SingleTasking (for a single-tasking system
with a single-sampling rate). Other available modes are MultiTasking (a multisam-
pling rate is used in the system) and Auto (automatically adjust between rates).
Click on OK and start the simulation by clicking on the play icon .
The Scope block displays the outputs from the reduced-precision and full-pre-
cision filters, and the error (middle plot) between these precisions is shown in Fig. A.20.
The user can also double-click on the Reduce Precision block to change the data
type to int8, int16, or int32, and observe the differences in comparison with the dou-
ble-precision comb filter. More information on the usage of various blocks and set-
tings can be found in the Simulink: Users Guide [8].

Figure A.19 Simulation Para-
meters window
Figure A.20 Scope display
Section A.4 Using Blocksets A31

Additional Simulink tools can be used for simulating a dynamic system. These
tools include:
1. Simulink Accelerator
2. Model Differencing tool
3. Profiler
4. Model Coverage tool
The Simulink Accelerator speeds up the execution of Simulink simulations.
This tool uses part of the Real-Time Workshop to create and compile C code to
replace the code that Simulink uses in normal mode. This mode can be selected by
clicking on the Simulink menu and selecting Accelerator. Alternatively, the user can
select Accelerator from the menu located in the middle of the toolbar .
The Model Differencing tool is available under the Tools menu of Simulink. It
finds and displays the differences between two Simulink models, and the differences
are reported in the Model Differences Report.
To activate the profiler, simply click on Profiler under Simulinks Tools menu.
The Simulink Profiler collects and profiles performance data while the model is
being simulated. When the simulation finishes, Simulink generates a report that
describes the time taken to execute each functional block. This feature allows the
user to identify the blocks or subsystems that require more time for execution and
thus are targets for optimization.
The Model Coverage tool validates the users models. It analyzes the execu-
tion of blocks that serve as decision points in the model. These blocks include
switch, triggered subsystem, enabled subsystem, absolute value, saturation, state-
flow charts, etc. The user can run the coverage tool by selecting Coverage Settings
from Simulinks Tools menu. When the Coverage Settings dialog box is displayed,
simply check Enable Coverage Reporting. Note that both model-coverage reporting
and acceleration mode cannot be enabled at the same time because Simulink dis-
ables coverage reporting if the accelerator option is enabled.
Several Simulink blocksets, such as DSP Blockset, Fixed-Point Blockset, Com-
munication Blockset, Code Division Multiple Access (CDMA) Reference Blockset,
and many others, are available to run in the Simulink environment. The following
section describes two important blocksets that are related to DSP applications.
A.4 USING BLOCKSETS
A blockset is a set of special blocks that are included in an application library for
handling the application tasks in Simulink. The most commonly used blocksets for
DSP applications are DSP and Fixed-Point blocksets [9, 10].
A.4.1 DSP Blockset
The DSP Blockset is a collection of signal-processing blocks for use with Simulink.
To access the DSP Blockset, type the following in the MATLAB command window:
dsplib

Figure A.21 Functions
available in the DSP
Blockset
Many important DSP algorithm blocks are included in the DSP Blockset Library.
As shown in Fig. A.21, this library includes the following blocks:
1. DSP Sources contains special sources for generating signals such as chirp, dis-
crete impulse, sinewave, clock, etc. It also contains blocks that capture signals
from workspaces, wave devices, and wave files.
2. DSP Sinks contains functional blocks for counter, matrix, time, vector, and
spectrum scope. It also contains a block for writing signals into workspaces,
wave devices, and wave files.
3. Filtering contains three main sections: (1) filter design (contains many FIR-
and IIR-filter designs and structures), (2) multirate filters (contains decima-
tion, interpolation, analysis and synthesis, and wavelet analysis and synthesis
filters), and (3) adaptive filters (contains LMS, Kalman, and RLS adaptive
filters).
4. Transforms contains a set of transforms that include FFT, IFFT, discrete cosine
transform (DCT), inverse discrete cosine transform (IDCT), and real and
complex cepstrum.
5. Signal Operations provides a set of commonly used DSP operations such as
convolution, upsample, downsample, integer and fractional delay, zero
padding, sample-and-hold, etc.
6. Estimation contains three sections: (1) linear prediction (contains autocorrela-
tion), (2) power-spectrum estimation (contains magnitude-square FFT, Burg,
covariance, Yule-Walker, and STFT), and (3) parametric estimation (Burg,
covariance, and Yule-Walker AR estimators).
7. Statistics contains a set of blocks for computing autocorrelation, correlation,
histogram, variance, median, mean, etc.

8. Math Functions contains three functional groups: (1) math operations (dB gain,
conversion, normalization, complex exponential, etc.), (2) matrix and linear
algebra (matrix factorization, matrix inversion, linear-system solvers, pseudo
inversion, matrix multiplication, matrix square, Toeplitz, etc.), and (3) polyno-
mial-function evaluation.
9. Quantizers contains several blocks for quantizer and uniform encoder and
decoder.
10. Signal Management contains four groups: (1) switches and counters, (2) buffer
for buffering and unbuffering signals, (3) matrix and signal indexing, and (4) sig-
nal attributes that convert the dimension of the signal, check signal and frame
status, etc.
11. Platform Specific I/O contains a set of blocks that take in sound signals from a
standard audio device and output to a standard audio device in real time. It
also contains a set of blocks that read from and write to a file with a standard
.wav format.
The DSP Blockset has the flexibility to handle single and multichannel signals,
as will as sample and frame-based signals. Therefore, the following four possible
types of input signal can be implemented in Simulink, as shown in Fig. A.22: (1) sam-
ple-based single-channel signals, (2) sample-based multichannel signals, (3) frame-
based single channel signals and, (4) frame-based multichannel signals.
Frame-based (block) processing can accelerate real-time systems, as
explained in Chapter 3. In real-time DSP systems, data acquisition can be carried
out at a higher rate when a block of samples is transferred to the processor and
when these block samples are processed at once. In contrast, sample processing
interrupts the processor at every sample. Therefore, frame-based processing mini-
mizes the overhead incurred in interrupting the processor. However, we have to
consider the latency introduced by frame-based processing in some applications, as
explained in Chapter 3.
In Simulink models, frame-based and sample-based signals are indicated by a
double line and single line respectively. A sample-based signal can be
converted to a frame-based signal by using the Buffer block, and an Unbuffer block
converts a frame-based signal back to a sample-based signal. The DSP Sources
Library provides a set of blocks for creating sample-based and frame-based signals.
Two different delays affect Simulink models: (1) computational delay and
(2) algorithmic delay. Computational delay depends on how fast the computer
hardware and software can execute a block of data. There are several methods to
reduce computational delay, such as using frame-based processing or using the
Real-Time Workshop to generate generic real-time code for specific hardware.
Algorithmic delay is intrinsic to the algorithm and is independent of the processor.
Algorithmic delay is commonly implemented in Simulink using an integer-delay
block. It can also occur in certain conditions known as tasking latency, which arises
from the synchronization requirements of Simulinks tasking mode. For example,
the multirate blocks operating in multitasking mode are subject to tasking latency.
More information on and examples of using the DSP Blockset can be found in
the DSP Blockset Users Guide [9].
1 :2, 1 Q2

1
1
1
1
2
2 2
3
3 3
4
4 4
(b) Sample-based, multichannel
(4 channels)
time
4
3
2
1
4
3
2
1
8
7
6
5
8
7
6
5
12
11
10
9 9
time
(d) Frame-based multichannel
(2 channels, 1 frame per channel,
4 samples per frame)
4
3
2
1
4
3
2
1
4
3
2
1
4
3
2
1
time
(c) Frame-based single-channel
(4 samples per frame)
1
2
3
4
time
(a) Sample-based, single-channel
Figure A.22 Different types of signal processing available in Simulink
A.4.2 Fixed-Point Blockset
The Fixed-Point Blockset [10] simulates effects commonly encountered in fixed-point
systems such as digital filtering. The Fixed-Point Blockset extends the functional block-
sets of the standard Simulink. This blockset allows the user to develop, simulate, ana-
lyze, and implement DSP systems using fixed-point arithmetic. A fixed-point
simulation is vital for resolving finite-wordlength problems associated with different
algorithms before entering the coding and implementation stages of the development
cycle. Such a simulation greatly enhances the accuracy of the code and speeds up sys-
tem-development time. Some important features of this blockset include the following:
1. Integer, fractional, and generalized fixed-point data types with a wordlength
from 1 to 128 bits

2. IEEE-754 format with single and double precisions, as well as non-IEEE
standards
3. Methods to control the overflow, scaling, and rounding of fixed-point data
4. An interface tool to profile the statistics of the simulation
5. Generation of C code (integer type only) for execution on a given embedded
processor
Figure A.23 shows the window that contains the blocks in the Fixed-Point
Blockset Library. This window can be opened by typing
fixpt
in the MATLAB command window. The window shows that the library contains the
following different blocks: (1) Math, (2) Data Type Conversion & Propagation, (3) Look-
Up tables, (4) Logic & Comparison, (5) Filters (for performing filtering), (6) Delays &
Holds, (7) Select (for selecting one from many inputs), (8) Nonlinear (for performing
different nonlinear threshold functions), (9) Calculus, (10) Bits, (11) Sources,
(12) FixPt GUI, (13) Demos, and (14) Edge Detect.
Figure A.23 Main window for the Fixed-Point Blockset Library

Configure Fixed-Point Blocks
The Fixed-Point Blockset provides an option for specifying the output data type via
the block dialog box. Possible options are listed in Table A.10. In the signed and
unsigned integer representation (sint, uint), the default binary point is to the right
of the LSB. In the unsigned fractional-data type (ufrac), the default binary point is to
the left of the MSB. In the signed fractional-data type (sfrac), the binary point is just
to the right of the MSB (sign bit). In order to provide flexibility in defining different
Q formats (or generalized fixed-point numbers) with a user-specified binary point,
the Fixed-Point Blockset provides generalized signed and unsigned fixed-point num-
bers (sfix, ufix), and the binary point is determined from the value in the output
scaling. For example, if the scaling factor is the binary point is located E bits left
of the LSB. The final data type is the floating-point data format, which can be repre-
sented by IEEE-754 single precision float('single'), IEEE-754 double precision
float('double'), and a non-IEEE format expressed as float(w,e) for a total of w
bits with e exponent bits, (w-e-1) fractional bits, and 1 sign bit.
The option of selecting output scaling is applicable to fixed-point data types.
There are two general scaling modes: binary point-only scaling and slope/bias scal-
ing. In binary point-only scaling mode, power-of-two scaling (e.g., where E is
unrestricted) is used. This scaling has the effect of moving the binary point to the
left by E bits. In slope/bias scaling mode, a slope of is realized, where a
bias of B bits is used to scale the integer number Q as Note that
specifies the binary point, and F is the fractional slope, which is normalized to
Slope/bias scaling is a superset and contains power-of-two scaling,
which can be considered under slope/bias scaling with and In gener-
al, slope/bias scaling mode allows more flexible scaling and maximizes usage of the
finite number of bits.
Fixed-point numbers can be rounded using the following rounding modes:
1. Zero, which rounds toward zero and is similar to the fix function in MATLAB.
2. Nearest, which rounds toward the nearest representable number and is similar
to the round function.
B = 0. F = 1
1 F 6 2.
2
E
SQ + B.
S = F2
E
2
-E
,
2
E
,
TABLE A.10 Definition of Output Data Types
Option Description
uint(n) n-bit unsigned integer
sint(n) n-bit signed integer
ufrac(n,g) n-bit unsigned fractional number
sfrac(n,g) n-bit signed fractional number
Note: the number of guard bits, g (optional), lies at the left
of the default binary point
ufix(n) n-bit unsigned fixed-point number
sfix(n) n-bit signed fixed-point number
float('single') IEEE-754 single-precision, floating-point number
float('double') IEEE-754 double-precision, floating-point number
float(w,e) Total bits w and exponent bits e

3. Ceiling, which rounds toward positive infinity and is similar to the ceil
function.
4. Floor, which rounds toward negative infinity and is similar to the floor
function.
Overflow handling in the Fixed-Point Blockset can be set by checking the
Saturate to max or min when overflow occurs check box to specify the use of satu-
rated mode. If unchecked, overflow wraps to the other end of the number range.
An example of an analog-to-digital model (shown in Fig. A.24) is used to illus-
trate the effects of using different wordlengths in representing digitized samples
with the Fixed-Point Blockset. To start the simulation, type
fxpadc
in the MATLAB command line. The user can double-click on any block and adjust
the default parameters. The Signal Generator block is configured to generate a
sinewave with double-precision amplitude on the interval The Zero-Order
Hold block simulates the sampling of the continuous sinewave, and the Dbl To FixPt1
block converts the double-precision, floating-point number to a fixed-point repre-
sentation. Once the desired parameters have been specified, the user can click on
the play button to start the simulation.
In the example, different fixed-point implementations using (1) Q.15
sfrac(16), (2) Q1.14 sfix(16) with scaling of and (3) 8-bit wordlength
sfix(8) with scaling of can be implemented individually inside the Dbl to FixPt1
block. The result for a Q.15 representation is shown in Fig. A.25(a), which is unable
to cover the entire range of the input signal due to the fact that the dynamic range of
a Q.15 number is By introducing an additional integer bit as in the
Q1.14 format, the dynamic range increases to but with reduced preci-
sion. The result is illustrated in Fig. A.25(b), which shows very little difference
between the double-precision and Q1.14 representation. A coarser quantization
[-2, +1.99],
[-1, +0.99].
2
-4
2
-14
,
[-2, +2].
Figure A.24 Simulink block diagram for a fixed-point implementation of ADC

(a)
(b)
(c)
Figure A.25 Different data represen-
tations: (a) Q.15 and double precision,
(b) Q1.14 and double precision, (c)
Q3.4 and double precision
Section A.5 MATLAB Link for Code Composer Studio A39

step is observed in Fig. A.25(c) for the 8-bit wordlength, where the binary point is
the fourth bit to the left of the rightmost bit.
C-Code Generation in Fixed-Point Blockset
After the system has been designed and analyzed using the Fixed-Point Blockset, C
code can be generated directly using the Real-Time Workshop [11]. The C code gen-
erated from the fixed-point block uses only integer types and can be used directly
on embedded fixed-point processors, which can be either a fixed-point or floating-
point architecture. The code can also be generated for testing on a rapid prototyping
system such as xPC [12] and a real-time window target [13]. It can also generate
code for non-real-time testing on a computer running on any supported operating
system. Please refer to reference [10] for detailed information on the step-by-step
generation of fixed-point C code.
A.5 MATLAB LINK FOR CODE COMPOSER STUDIO
MATLAB has introduced a new toolbox called MATLAB Link for CCS, which
establishes bidirectional links between MATLAB, CCS, and TMS320 processors
[14]. As shown in Fig. A.26, the MATLAB Link for CCS connects MATLAB to
Texas Instruments software and hardware. It allows the MATLAB user to create an
object that links to CCS and the real-time data exchange (RTDX) so that the user
can transfer data to and from the processor without halting it. In other words, the
user can open the CCS environment, download the program, communicate with the
DSP processor, access the processors registers and memories, and perform data log-
ging while the DSP processor is running. This capability allows the user to change a
parameter or variable in MATLAB and transfer the value into the running DSP
processor in order to tune and alter the algorithm in real time.
MATLAB
MATLAB Link for CCS
CCS
Function calls and
data manipulation
Debug RTDX
C2000/C5000/C6000 DSP processors Figure A.26 Block diagram of
MATLAB Link for CCS

The current version of MATLAB Link for CCS can link the TMS320C6701
Evaluation Module (EVM) TMS320C6711 DSK, C6000 and C5000 simulators, and
other DSP boards that are supported in the setup utility. In this section, we give a
brief overview of some important features of MATLAB Link for CCS and show
how to link it to the C5000 simulator.
First, we can check whether the link for CCS is installed in MATLAB by
typing
help ccslink
in the MATLAB command line. If the software is installed, it returns a list of com-
mands for analysis and debugging with CCS. Next, we can check the boards and
processors that are installed on the computer by typing
ccsboardinfo
which returns a list of boards and processors that are installed and recognized by
CCS.
In order to select the C5000 simulator, type
[boardnum,procnum]=boardprocsel
and select the C5400 (or C5500) simulator. Note that if only a single board (or sim-
ulator) is installed in the system, this step can be skipped. After successful selection
of the board or simulator, we can create the link between MATLAB and CCS by
typing the following MATLAB command:
cc = ccsdsp('boardnum',boardnum,'procnum',procnum);
We notice that CCS is placed in the background, and we can view the status by
typing
disp(cc)
In addition, we can set the visibility for the CCS window by typing
visible(cc,1)
The next step is to load a project file (e.g., ccstut_54xx.pjt) into CCS using the fol-
lowing MATLAB commands:
projfile = fullfile(matlabroot,'toolbox','ccslink','ccsdemos',
'ccstutorial', 'ccstut_54xx.pjt');
projpath = fileparts(projfile)
open(cc,projfile) % load a file into CCS
cd(cc,projpath) % change the working directory that CCS uses
However, we cannot build the CCS project in the MATLAB window. The user must
build the project in CCS by clicking on the Build-All icon in the CCS window. After
Suggested Readings A41

the project is built, we can load the ccstut_54xx.out file using the command
load (cc,'ccstut_54xx.out') % transfer program file to target
% processor
After loading, we can set the breakpoint, reset the program counter, and run the
program using the following commands:
halt(cc) % terminate the execution running on
% the target
restart(cc) % reset the PC to start of program
run(cc,'runtohalt',30) % run to breakpoint with timeout=30 sec
Data located in memory on the target processor can be transferred to the
MATLAB environment. For example, to read the data ddtav and idtav from CCS
to the MATLAB workspace, type the following commands:
ddtav = read(cc,address(cc,'ddat'),'single',4)
idtav = read(cc,address(cc,'idat'),'int16',4)
MATLAB supports several data types. In the preceding example, the data type
single is used instead of double, which is not supported by the C5000 simulator.
Besides reading from the target processor, the user can also write data to
memory on the target processor. For example, we can modify ddtav and idtav as
follows:
write(cc,address(cc,'ddat'),single([3.14 -10.3 exp(-2)
sin(pi/2)]));
write(cc,address(cc,'idat'),int16([1:4]));
In addition, DSP registers can be viewed and modified. A simple example that
changes the value of ACC A is shown as follows:
regal=cc.regread('AL','binary')
dec2hex(regal)
cc.regwrite('AL',hex2dec('683'),'binary')
dec2hex(cc.regread('AL','binary'))
More detailed information on MATLAB Link for CCS and how to use links
for RTDX, which requires actual hardware (EVM or DSK) to be hooked up to the
host computer, are documented in [14].
SUGGESTED READINGS
1 The MathWorks. Using MATLAB. Version 6, 2002.
2 The MathWorks. Using MATLAB Graphics. Version 6, 2002.
3 The MathWorks. Signal Processing Toolbox Users Guide. Version 6, 2002.

4 The MathWorks. Filter Design Toolbox Users Guide. Version 2, 2002.
5 The MathWorks. Communication Toolbox Users Guide. Version 2, 2002.
6 The MathWorks. Image Processing Toolbox Users Guide. Version 3, 2002.
7 The MathWorks. Wavelet Toolbox Users Guide. Version 2, 2002.
8 The MathWorks. Simulink: Users Guide. Version 5, 2002.
9 The MathWorks. DSP Blockset Users Guide: For Use with Simulink. Version 5, 2002.
10 The MathWorks. Fixed-Point Blockset Users Guide: For Use with Simulink. Version 4,
2002.
11 The MathWorks. Real-Time Workshop Users Guide: For Use with Simulink. Version 5,
2002.
12 The MathWorks. xPC Target Users Guide: For Use with Simulink. Version 2, 2002.
13 The MathWorks. Real-Time Window Target Users Guide: For Use with Simulink. Version
2.1, 2002.
14 The MathWorks. MATLAB Link for Code Composer Studio. Version 1, 2002

Matlab DSP

Uploaded by

Copyright:

Available Formats

Matlab DSP

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Matlab DSP

Uploaded by

Copyright:

Available Formats

A1

A2 Appendix A Introduction to MATLAB and Simulink

Section A.1 Using MATLAB A3

A4 Appendix A Introduction to MATLAB and Simulink

Section A.1 Using MATLAB A5

A6 Appendix A Introduction to MATLAB and Simulink

Section A.1 Using MATLAB A7

A8 Appendix A Introduction to MATLAB and Simulink

Section A.1 Using MATLAB A9

A10 Appendix A Introduction to MATLAB and Simulink

A12 Appendix A Introduction to MATLAB and Simulink

A14 Appendix A Introduction to MATLAB and Simulink

A16 Appendix A Introduction to MATLAB and Simulink

A18 Appendix A Introduction to MATLAB and Simulink

A20 Appendix A Introduction to MATLAB and Simulink

A22 Appendix A Introduction to MATLAB and Simulink

Section A.3 Using Simulink A23

A24 Appendix A Introduction to MATLAB and Simulink

Section A.3 Using Simulink A25

A26 Appendix A Introduction to MATLAB and Simulink

Section A.3 Using Simulink A27

A28 Appendix A Introduction to MATLAB and Simulink

Section A.3 Using Simulink A29

A30 Appendix A Introduction to MATLAB and Simulink

Section A.4 Using Blocksets A31

A32 Appendix A Introduction to MATLAB and Simulink

Section A.4 Using Blocksets A33

A34 Appendix A Introduction to MATLAB and Simulink

Section A.4 Using Blocksets A35

A36 Appendix A Introduction to MATLAB and Simulink

Section A.4 Using Blocksets A37

A38 Appendix A Introduction to MATLAB and Simulink

Section A.5 MATLAB Link for Code Composer Studio A39

A40 Appendix A Introduction to MATLAB and Simulink

Suggested Readings A41

A42 Appendix A Introduction to MATLAB and Simulink

You might also like