step calculation for sampled sin sample

no-reply@allegro.cc (Ariesnl) — Thu, 08 Oct 2015 19:25:48 +0000

Calling for a math guru,
would be nice if someone could explain this

I need to find the stepsize to get a nice sin wave
and maybe other waves in the future. but I'm confused

Anyone ?

#SelectExpand
  1void CustomProgram::Tone(int a_nFrequency, int a_nDuration)
  2{
  3    unsigned int samples = (float(a_nDuration)/1000.0 * a_nFrequency)*100;
  4    unsigned long sample_size=al_get_channel_count(ALLEGRO_CHANNEL_CONF_1) *al_get_audio_depth_size(ALLEGRO_AUDIO_DEPTH_INT16);
  5    unsigned long bytes = samples * sample_size;
  6
  7    void *buff = NULL;
  8    buff = al_malloc(bytes);
  9
 10    al_destroy_sample(m_pSample);
 11    m_pSample = al_create_sample (buff,samples, a_nFrequency,ALLEGRO_AUDIO_DEPTH_INT16,ALLEGRO_CHANNEL_CONF_1,true);
 12
 13
 14    if (m_pSample!=NULL)
 15    {
 16        int16_t * ptr = (int16_t *) al_get_sample_data(m_pSample);
 17        double step = double(2 * PI * a_nDuration * a_nFrequency) / samples;    // <----- This one here !
 18        for (unsigned int i=0;i<samples;i++)
 19        {
 20             ptr[i]= sin(i*step) * 30000;
 21        }
 22        al_play_sample (m_pSample,1.0,0,1.0,ALLEGRO_PLAYMODE_ONCE,NULL);
 23    }
 24}

no-reply@allegro.cc (Arthur Kalliokoski) — Thu, 08 Oct 2015 20:02:50 +0000

You need to have a_nFrequency steps per second for a_nDuration seconds. The desired frequency of sine wave needs to be divided by a_nFrequency steps per second to get the increment to pass to the float that is input to the sin() function.

Let's say that a_nFrequency is 28,000 hertz, duration is 10 seconds, and you want a sin wave to produce concert A (440 hertz).

So you'd need 28,000 * 10 sample elements to put the sin values into, and the increment per sample element would be 440.0/28000.0 .

As Elias pointed out below, I forgot about the 2.0 * M_PI to get one complete cycle from sin().

no-reply@allegro.cc (Elias) — Thu, 08 Oct 2015 20:05:01 +0000

Well, just look at a sine wave:

y = sin(x)

If x runs from 0 to 2 * pi, you get one complete wave. Now multiply by 2 * pi:

y = sin(x * 2 * pi)

Now the full wave goes from 0 to 1. Now assume x is our time in seconds, and you want a 440 Hz tone:

y = sin(x * 2 * pi * 440);

Now each seconds there is 440 complete sine waves.

Now, instead of having the time in seconds, x is a sample position, 48000 samples make up one second:

y = sin(x * 2 * pi * 440 / 48000);

In your code, a_nFrequency is the 48000. So a_nDuration/samples is your tone's frequency. I'm a bit unsure about those variable names, I'd say you can simplify that function a lot and still have it do the same thing.

no-reply@allegro.cc (Ariesnl) — Fri, 09 Oct 2015 14:08:52 +0000

Thnx,
I think this is what it should look like.
I'll try to make some effects using harmonics , see what can be done with this.

Allegro 5 should definitely get some "record sample" or "audio input" functions

#SelectExpand
  1void CustomProgram::Tone(int a_nFrequency, int a_nVolume, int a_nDuration)
  2{
  int nSampleRate = 48000;
  if (a_nVolume > 255)
  {
      a_nVolume = 255;
  }
  else if (a_nVolume<0)
  {
      a_nVolume = 0;
  }
 12
  int nInternalVolume = (30000 * a_nVolume)/255;
  unsigned int samples = (nSampleRate * a_nDuration) / 1000.0;
  unsigned long sample_size=al_get_channel_count(ALLEGRO_CHANNEL_CONF_1) *al_get_audio_depth_size(ALLEGRO_AUDIO_DEPTH_INT16);
  unsigned long bytes = samples * sample_size;
 17
  void *buff = NULL;
  buff = al_malloc(bytes);
 20
  al_destroy_sample(m_pSample);
  m_pSample = al_create_sample (buff,samples, nSampleRate ,ALLEGRO_AUDIO_DEPTH_INT16,ALLEGRO_CHANNEL_CONF_1,true);
  double dFi = double(2 * PI * a_nFrequency) / nSampleRate;
 24
  if (m_pSample!=NULL)
  {
      int16_t * ptr = (int16_t *) al_get_sample_data(m_pSample);
      for (unsigned int i=0;i<samples;i++)
      {
           ptr[i]= sin(i * dFi) * nInternalVolume;
 31
      }
      al_play_sample (m_pSample,1.0,0,1.0,ALLEGRO_PLAYMODE_ONCE,NULL);
  }
 35}

no-reply@allegro.cc (Elias) — Fri, 09 Oct 2015 19:57:55 +0000

Looks good to me now, assuming a_nDuration is in ms.