Colon operation in fixed-point

I am writing a function in 2020b that I then want to convert into fixed-point using the tool. I am defining some arrays by the colon operation, however when i use variables in the colon it does not like it. Yet using the exactly same numbers as numbers and not variables the tool says no and that it must be a fi or constant object. Why is this and how can i fix it?

[James Tursa]

Please post a small example code that reproduces the problem so we know exactly what you are attempting.

[Joshua Ford]

f0 = fi(-20000000);                             %Initial Frequency 
f1 = fi(20000000);                              %Target Frequency    
T =  fi(4e-6);                                  %Pulse Duration  
fs = 50000000;                                  %Sampling rate = 50MHz
Ts = fi(1/fs);                                  %Sampling time 
NoPulses = 512;                                 %Number of Pulses
Signal_Duration = 12e-6;                        %Time Duration of Signal

k = uint64((f1-f0)/5e-6);                       %Bandwidth = 40MHz
t = fi(0:Ts:T);                                 %Time Samples                       
theta1 = fi(f0.*t + (k/2)*t.^2, 1, 16, 7);
theta2 = fi(2*pi, 1, 16, 7);
theta = theta1*theta2;                          %Phase
Chirp = cos(theta) + 1i*sin(theta) ;            %Reference Chirp 

[Joshua Ford]

I have tried multiple variation of this and I am either getting an error stating that 'colon operation aren't supported for fiex-point operation' or that 'invalid v - value must be fi or constant'

Answer 1

[Kiran Kintali]

MATLAB HDL Coder workflow does support colon operator during fixed-point conversion and code generation. Please share a sample design file dut.m, testbench.m (caling dut) and a project file with fixed-point conversion settings.

[Joshua Ford]

f0 = fi(-20000000);                             %Initial Frequency 
f1 = fi(20000000);                              %Target Frequency    
T =  fi(4e-6);                                  %Pulse Duration  
fs = 50000000;                                  %Sampling rate = 50MHz
Ts = fi(1/fs);                                  %Sampling time 
NoPulses = 512;                                 %Number of Pulses
Signal_Duration = 12e-6;                        %Time Duration of Signal

k = uint64((f1-f0)/5e-6);                       %Bandwidth = 40MHz
t = fi(0:Ts:T);                                 %Time Samples                       
theta1 = fi(f0.*t + (k/2)*t.^2, 1, 16, 7);
theta2 = fi(2*pi, 1, 16, 7);
theta = theta1*theta2;                          %Phase
Chirp = cos(theta) + 1i*sin(theta) ;            %Reference Chirp 

[Joshua Ford]

I have tried multiple variation of this and I am either getting an error stating that 'colon operation aren't supported for fiex-point operation' or that 'invalid v - value must be fi or constant'

Answer 2

[Kiran Kintali]

Support for colon exists with fixed-point types according to documentation. https://www.mathworks.com/help/fixedpoint/ref/colon.html

Please reach out to https://www.mathworks.com/support.html for additional help specific to your usecase.

Answer 3

[Andy Bartlett]

fi colon integer values only

As the fi colon documention mentions, all colon operands must have integer values.

Strategy to generalize fi colon

The following function shows a strategy for getting colon behavior that will work with fi objects.

The key idea is create an integer valued vector [0, 1, 2, ... n-1].

Then multiply by a scalar and add a scalar to produce the desired colon-style vector.

function y = colonAlternative(uLo,uSpacing,uHi)
   %colonAlternative alternative to colon due to fi restrictions with colon
   
   % fi colon operator
   %   lo:spacing:hi
   % requires the all three arguments
   %   lo
   %   spacing
   %   hi
   % to have integer values
   %
   % To work around this create an integer valued vector
   % then multiply by the desired spacing
   % and add the desired starting point if it is not zero
   % Conceptually
   %    fiIntegerVec = (0:(nPts-1));
   %    finalVec = fiAnchor + fiIntegerVec.*fiSpacing
   %
   uDelta = uHi - uLo;
   nPtsMinusOne = round( uDelta/uSpacing );
   integerVec = 0:nPtsMinusOne;
   y = uLo + (uSpacing .* integerVec);    
end

Compatible Example for Fixed-Point Converter

The attached design file fiColonLookupV2.m and its associated testbench file test_fiColonLookupV2.m are compatible with Fixed-Point Converter. Use of the strategy shown makes sure that only integer valued fi objects are fed to the colon operator's operands.