0
$\begingroup$

this is a crosspost from stackoverflow as I figured there may be more who can help me here...

I've been writing a basic C program to generate a sine wave and write it out to STDOUT for piping into a filter which I am also going to write but I'm having issues.

I am able to generate what sounds like a good clean sine wave, but the frequency is off. When I listen to the output I am generating and compare to a true signal generators output, the frequency of mine is slightly too high.

The C code should generate raw 16 bit signed stereo wav @ 44100Hz for a frequency passed as an argument to the program, or 500Hz as default.

A matlab version of the code is here, (modded very slightly for 1 indexing in Matlab, WHY MATHWORKS WHY!?) minus the passing to STDOUT etc as I know that works fine

CHANNELS = 2; SAMPLING_RATE = 44100; NUM_SAMPLES = 512;

frequency = 1000;

%% BEGIN WHILE(1) LOOP HERE

output_buff = zeros(1,CHANNELS*NUM_SAMPLES);

for i = 1:2:CHANNELS*NUM_SAMPLES
    output_buff(i) = 30000 * sin(frequency * pi * (i-1)/NUM_SAMPLES);
    output_buff(i+1) = output_buff(i);
end

%% OUTPUT TO STDOUT AND REPEAT

I should add that this code runs (in the C version) inside a while true loop, generating a full output_buff of values, then pushing the buffer to STDOUT.

I have written some further testing code to see what's actually being generated as follows:

plot(1:CHANNELS*NUM_SAMPLES, output_buff)

output_buff = output_buff .* hanning(length(output_buff))';    
Y = fft(output_buff);    
Mag=abs(Y(1:length(output_buff)/2)).^2;    
[a,b]=max(Mag);    
% Result    
SAMPLING_RATE*b/length(output_buff)

When I run this script, I can see at the end the frequency of the generated signal is actually 1.0767e+03Hz... Close but no cigar...

I've tried tweaking some of the parameters but I have no idea what's wrong or how to make the generated frequency more accurate.

The C code is as follows:

#include <sys/types.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>

#include <unistd.h>

#define M_PI 3.1415926535897932384626433832795L
#define NUM_CHANNELS 2
#define NUM_SAMPLES 1024
#define SAMPLE_RATE 44100

double frequency = 500.0;

int main(int argc, char *argv[])
{   
    if (argc >= 2){
        frequency = atof(argv[1]);
    }

    int16_t samples[NUM_CHANNELS * NUM_SAMPLES];
    unsigned cbBuffer=sizeof(samples);

    while(1){
        for (int i = 0; i < cbBuffer; i+=NUM_CHANNELS){
            samples[i] = (int16_t) 30000 * sin(frequency * M_PI * (double)i /(double)(NUM_SAMPLES));
            samples[i+1] = samples[i];
        }

        int done=write(STDOUT_FILENO, samples, cbBuffer);
        if(done<0){
            fprintf(stderr, "%s : Write to stdout failed, error=%s.", argv[0], strerror(errno));
            exit(1);
        }else if(done!=cbBuffer){
            fprintf(stderr, "%s : Could not read requested number of bytes from stream.\n", argv[0]);
        }
    }

    return 0;
}

UPDATE:

I changed the C code as follows and I now get a nice smooth sine wave but the frequency is way too low

int main(int argc, char *argv[])
{   
    if (argc >= 2){
        frequency = atof(argv[1]);
    }

    int16_t samples[NUM_CHANNELS * NUM_SAMPLES];
    unsigned cbBuffer=sizeof(samples);
    int counter = 0;

    while(1){
        for (int i = 0; i < NUM_CHANNELS * NUM_SAMPLES; i+=NUM_CHANNELS){
            samples[i] = (int16_t) 3000 * sin(frequency * M_PI * (double)counter /(double)(SAMPLE_RATE));
            samples[i+1] = samples[i];
            counter++;
        }

        int done=write(STDOUT_FILENO, samples, cbBuffer);
        if(done<0){
            fprintf(stderr, "%s : Write to stdout failed, error=%s.", argv[0], strerror(errno));
            exit(1);
        }else if(done!=cbBuffer){
            fprintf(stderr, "%s : Could not read requested number of bytes from stream.\n", argv[0]);
        }
    }

    return 0;
}
$\endgroup$
3
  • $\begingroup$ you need to reference the sample rate during sine wave creation. Try x = 30000*sin(frequency*2*pi*(0:NUM_SAMPLES-1)'/SAMPLING_RATE) $\endgroup$
    – Hilmar
    Commented Feb 5, 2014 at 14:26
  • $\begingroup$ Thanks for the advice, however when I do this the end result is a sine wave with a strange sort of buzz noise over the top... Difficult thing to describe in text! $\endgroup$
    – rcbevans
    Commented Feb 5, 2014 at 14:31
  • $\begingroup$ A raw sample generated can be found at sharesend.com/k5nfklyn and played with sox --no-show-progress -t raw -e signed -b 16 -c 2 -r 44100 $\endgroup$
    – rcbevans
    Commented Feb 5, 2014 at 14:36

1 Answer 1

0
$\begingroup$

So I've managed to fix it. My problem was I wasn't properly taking into account the Sampling Rate of the produced raw wav file, and that by using the loop variable, there is a discontinuity for each iteration of the loop.

The final working code is as follows:

#include <sys/types.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>

#include <unistd.h>

#define M_PI 3.1415926535897932384626433832795L
#define NUM_CHANNELS 2
#define NUM_SAMPLES 512
#define SAMPLE_RATE 44100

double frequency = 500.0;

int main(int argc, char *argv[])
{   
    if (argc >= 2){
        frequency = atof(argv[1]);
    }

    int16_t samples[NUM_CHANNELS * NUM_SAMPLES];
    unsigned cbBuffer=sizeof(samples);
    int counter = 0;

    while(1){
        for (int i = 0; i < NUM_CHANNELS * NUM_SAMPLES; i+=NUM_CHANNELS){
            samples[i] = (int16_t) 3000 * sin(frequency * 2 * M_PI * (double)counter /(double)(SAMPLE_RATE));
            samples[i+1] = samples[i];
            counter++;
        }

        int done=write(STDOUT_FILENO, samples, cbBuffer);
        if(done<0){
            fprintf(stderr, "%s : Write to stdout failed, error=%s.", argv[0], strerror(errno));
            exit(1);
        }else if(done!=cbBuffer){
            fprintf(stderr, "%s : Could not read requested number of bytes from stream.\n", argv[0]);
        }
    }

    return 0;
}
$\endgroup$
1
  • $\begingroup$ You could also have used in your original version samples[2*i]=samples[2*i+1]=... to get the index right. If NUM_CHANNELS could ever be different from 2, you would need an inner loop to account for that variability. $\endgroup$ Commented Feb 5, 2014 at 19:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.