FIR filter not working as expected

Hi Vishal,

Sure I've attached the matlab script I've used to generate the coefficients here. I think it's just the same ezfir16.m from TI but from an old controlsuite.

Also the c code to use this is just using the TI library but below is how I use it.

Wait to hear from you!  Also let me know anything else you may need.

disp('ezFIR FILTER DESIGN SCRIPT');
if exist('ezOrder') == 0
   ezOrder=input('Input FIR Filter order(EVEN for BS and HP Filter) : ');
end
if exist('ezFres') == 0
	disp('Low Pass          : 1');
	disp('High Pass         : 2');
	disp('Band Pass         : 3');
	disp('Band Stop         : 4');
   disp('Fractional Delay  : 5');
   ezFres=input('Select Any one of the above Response              : ');
end
if exist('ezWinType') == 0
	if ezFres~=5
		disp('Hamming           : 1');
		disp('Hanning           : 2');
		disp('Bartlett          : 3');
		disp('Blackman          : 4');
		disp('Flat top          : 5');
      ezWinType=input('Select Any one of the above window                : ');
    end
end

if exist('ezFs') == 0
   ezFs=input('Enter the Sampling frequency                      : ');
end   

if ezFres~=5
   if exist('ezFc') == 0
      ezFc=input('Enter the corner frequency(Fc or [Fl Fu])         : ');
   end
else   
  	ezWinType = 1;
   ezFc = 0.25;
   if exist('ezDelay') == 0
      ezDelay=input('Enter the required delay (samples)                : ');
   end
end

if exist('ezFName') == 0
   ezFName=input('Enter the name of the file for coeff storage      : ','s');
end

% Design the Filter
if ezFres==1
    res='';
   elseif ezFres==2
      res='high';
   elseif ezFres==3
      res='';
   elseif ezFres==4
      res='stop';
   elseif ezFres==5
      res='fd';
end

if ezWinType==1
	   win=hamming(ezOrder+1);
   	winStr = 'Hamming';
   elseif ezWinType==2
      win=hanning(ezOrder+1);
   	winStr = 'Hanning';
   elseif ezWinType==3
      win=bartlett(ezOrder+1);
   	winStr = 'Bartlett';
   elseif ezWinType==4
      win=blackman(ezOrder+1);
      winStr = 'Blackman';
   elseif ezWinType==5
      win=flattop(ezOrder+1);
      winStr = 'Flattop';
   else
	  	winStr = '';
end

FcHz = ezFc;
nomFc=ezFc/(ezFs/2);                               % Normalise the frequency values
if ezFres~=5
   B=fir1(ezOrder,nomFc,res,win);
else
   B = lagrange(ezOrder, ezDelay);
end
Bi=B*32768;                                 % Coefficients in Q15 format
Bi=floor(Bi);
bsize=length(Bi);

for i=1:bsize                               % Saturate the coefficients for Q15 format
   if Bi(i)==32768
      Bi(i)=32767;
   end
end

if ezFres==1
    FirType='LP';
elseif ezFres==2
      FirType='HP';
elseif ezFres==3
     FirType='BP';
elseif ezFres==4
      FirType='BS';
elseif ezFres==5
      FirType='FD lagrange';
end
if length(FcHz) == 2
   FC = sprintf('%d-%d', FcHz(1), FcHz(2));
elseif length(FcHz) == 1
   FC = sprintf('%d', FcHz);
end

if ezFres == 5
   comment = sprintf('FIR16 %s order=%d Fs=%fHz Delay=%d samples', FirType, ezOrder, ezFs, ezDelay);    
else
   comment = sprintf('FIR16 %s order=%d %s Fs=%dHz Fc=%sHz', FirType, ezOrder, winStr, ezFs, FC); 
end
savefir16(ezFName,comment, Bi, ezOrder);
eval(['type ' ezFName]);
save B Bi;

% Plot the frequency response of the filter         
[H,f]=freqz(B,1,512,ezFs);
figure(1);
subplot(2,1,1);
plot(f,abs(H));
grid;
xlabel('Hertz');
ylabel('Magnitude Response');
subplot(2,1,2);
plot(f,unwrap(angle(H))*180/pi);
grid;
xlabel('Hertz');
ylabel('Phase (degrees)');
figure(2);
freqz(B,1,512,ezFs);

function h = lagrange(N, delay)
%LAGRANGE  h=lagrange(N,delay) returns order N FIR 
%          filter h which implements given delay 
%          (in samples).  For best results, 
%          delay should be near N/2 +/- 1.
n = 0:N;
h = ones(1,N+1);
for k = 0:N
    index = find(n ~= k);
    h(index) = h(index) *  (delay-k)./ (n(index)-k);
end

Thanks,

Ayaka

#include <fir.h>

// Copied from matlab script
/* FIR16 coefficients (FD_FILTER)
   FIR16 FD lagrange order=4 Fs=20000.000000Hz Delay=2.941176e+000 samples
   {155,-942,2913,31072,-432,} */
#define FIR_ORDER_FD 4
#define FD_COEFF {\
	2913,-28246958,2036334747}

FIR16 fir_FdRx = FIR16_DEFAULTS;

/* Define the Delay buffers and place in "firModFd" section */
#pragma DATA_SECTION(alFdRxBuf,"firModFd");
long alFdRxBuf[(FIR_ORDER_FD+2)/2];

/* Define Constant Co-efficient Array (used for Tx and Rx and place the .econst/.const section in
non-volatile memory */
#define FIR_ORDER_FD_SIZE ((FIR_ORDER_FD+2)/2)
const long FdCoeff_const[FIR_ORDER_FD_SIZE]= FD_COEFF;
long FdCoeff[FIR_ORDER_FD_SIZE];

// Initialisation ///////////////////
void init_filter(void)
{
	memcpy( FdCoeff, FdCoeff_const, FIR_ORDER_FD_SIZE * 2);
	fir_FdRx.dbuffer_ptr = alFdRxBuf;
	fir_FdRx.coeff_ptr=(long *)FdCoeff;
	fir_FdRx.order=FIR_ORDER_FD;
	fir_FdRx.init(&fir_FdRx);
}

// Usage ///////////////////////////
// This function gets called periodically and data is the ADC value collected at 20kHz
void modem_processRx(S16 data)
{
	// other codes here...

	fir_FdRx.input = data;
	fir_FdRx.calc(&fir_FdRx);

	// use fir_FdRx.output for the calculation followed...
}

Ok I think i found the problem. IF you place the section "coeffilt" in data memory i.e page = 1 and not page = 0 as it is in the example - it will work.  See the attached example

28069_FIXEDPT_FractionalDelayFIR_150209.zip

This matches with the MATLAB results:

% // TI File $Revision: /main/3 $
% // Checkin $Date: November 1, 2010   13:05:07 $
% // =====================================================================================
% //  This software is licensed for use with Texas Instruments C28x
% //  family DSCs.  This license was provided to you prior to installing
% //  the software.  You may review this license by consulting a copy of
% //  the agreement in the doc directory of this library.
% // -------------------------------------------------------------------------------------
% //          Copyright (C) 2010-2011 Texas Instruments, Incorporated.
% //                          All Rights Reserved.
% //======================================================================================
% //
% // 	FILE:    FixedPointFIR16.m
% //
% // 	TITLE:   DSP2833x Device Fixed Point FIR16 Test Program.   
% //
% //
% // 	ASSUMPTIONS:
% //
% //   	This matlab code is reference code for debugging FixedPoint DSP library
% //    16 bit FIR module and example project 2833x_FixedPoint_FIR16. 
% //
% //
% //   
% //    Watch Variables:
% //
% //      x_raw_sig         Input signal (float)
% //      fir_lpfxx_Q15		FIR filter coefficients (Q15)
% //      x_Q15             Input signal in Q15 format
% //      output            Filtered signal (Q15)
% //
% //######################################################################################
% // $TI Release: C28x Fixed-Point DSP Library v1.10 $
% // $Release Date: November 1, 2011 $
% //######################################################################################

clear all
close all

disp('==============================================================');
disp('==============================================================');
disp('this is Matlab version of C28x Fixed Point FIR16 code');
disp('The result is stored in array output Q15');
disp('The signal is stored in array x_raw_sig (float)');
disp('The signal in Q15 format is stored in array x_Q15');
disp('==============================================================');

RadStep = 0.1963495408494;

N=input('Please enter signal length N =');
halfSize=N-1;

Rad=[0:RadStep:(halfSize)*RadStep];

x_Q15 = 32768 * ((sin(Rad)+cos(Rad*2.3567))/2);


% x=Rx+j*Ix;
% x_raw_sig=x/2;

% x_Q15=2^15*x/2;

fir_lpf4_Q15=[155, -942, 2913, 31072, -432];

fir_lpf6_Q15=[-3368,347,17612,17612,347,-3368];

fir_lpf7_Q15=[-891,  -4096,   9083,  21178,   9083,  -4096,   -891];

fir_lpf32_Q15=[-31,    -61,     92,    122,   -186,   -248,    348,    449,   -607, -778,   1043,   1370,  -1909,  -2789,   4835,  14710,  14710,   4835, -2789,  -1909,   1370,   1043,   -778,   -607,    449,    348,   -248, -186,    122,     92,    -61,    -31];

%result=conv(fir_lpf6_Q15,x_Q15)*2^7/2^6/2^16;
%result=conv(fir_lpf7_Q15,x_Q15)*2^7/2^6/2^16;
% result=conv(fir_lpf32_Q15,x_Q15)*2^7/2^6/2^16;
result=conv(fir_lpf4_Q15,x_Q15)*2^7/2^6/2^16;

output=floor(result(1:N));

plot(output)