Band Limited PWM Generator

Type : PWM generator
References : Posted by paul_sernine75 AT hotmail DOT fr
Notes :
This is a commented and deobfuscated version of my 1st April fish. It is based on a tutorial code by Thierry Rochebois. I just translated and added comments.


Paul Sernine.
Code :
// SelfPMpwm.cpp

// Antialised PWM oscillator

// Based on a tutorial code by Thierry Rochebois (98).
// Itself inspired by US patent 4249447 by Norio Tomisawa (81).
// Comments added/translated by P.Sernine (06).

// This program generates a 44100Hz-raw-PCM-mono-wavefile.
// It is based on Tomisawa's self-phase-modulated sinewave generators.
// Rochebois uses a common phase accumulator to feed two half-Tomisawa-
// oscillators. Each half-Tomisawa-oscillator generates a bandlimited
// sawtooth (band limitation depending on the feedback coeff B).
// These half oscillators are phase offseted according to the desired
// pulse width. They are finally combined to obtain the PW signal.
// Note: the anti-"hunting" filter is a critical feature of a good
// implementation of Tomisawa's method.
#include <math.h>
#include <stdio.h>
const float pi=3.14159265359f;
int main()
  float freq,dphi; //!< frequency (Hz) and phase increment(rad/sample)
  float dphif=0;   //!< filtered (anti click) phase increment
  float phi=-pi;   //!< phase
  float Y0=0,Y1=0; //!< feedback memories
  float PW=pi;     //!< pulse width ]0,2pi[
  float B=2.3f;    //!< feedback coef
  FILE *f=fopen("SelfPMpwm.pcm","wb");
  // séquence ('a'=mi=E)
  // you can edit this if you prefer another melody.
  static char seq[]="aiakahiafahadfaiakahiahafahadf"; //!< sequence
  int note=sizeof(seq)-2;  //!< note number in the sequence
  int octave=0;     //!< octave number
  float env,envf=0; //!< envelopped and filtered envelopped
  for(int ns=0;ns<8*(sizeof(seq)-1)*44100/6;ns++)
//waveform control --------------------------------------------------
    //freq=27.5f*powf(2.0f,8*ns/(8*30*44100.0f/6)); //sweep
    //freq*=(1.0f+0.01f*sinf(ns*0.0015f));        //vibrato
    dphi=freq*(pi/22050.0f);   // phase increment
    //notes and enveloppe trigger
      if(note>=(sizeof(seq)-1))// sequence loop
      env=1; //env set
      //PW=pi*(0.4+0.5f*(rand()%1000)/1000.0f); //random PW
    env*=0.9998f;              // exp enveloppe
    envf+=0.1f*(env-envf);     // de-clicked enveloppe
    B=1.0f;                    // feedback coefficient
    //try this for a nice bass sound:
    //B*=envf*envf;              // feedback controlled by enveloppe
    B*=2.3f*(1-0.0001f*freq);  // feedback limitation

//waveform generation -----------------------------------------------
    //Common phase
    phi+=dphif;                 // phase increment
      phi-=2*pi;               // phase wrapping

    // "phase"    half Tomisawa generator 0
    // B*Y0 -> self phase modulation
    float out0=cosf(phi+B*Y0); // half-output 0
    Y0=0.5f*(out0+Y0);         // anti "hunting" filter

    // "phase+PW" half Tomisawa generator 1
    // B*Y1 -> self phase modulation
    // PW   -> phase offset
    float out1=cosf(phi+B*Y1+PW); // half-output 1
    Y1=0.5f*(out1+Y1);            // anti "hunting" filter

    // combination, enveloppe and output
    short s=short(15000.0f*(out0-out1)*envf);
    fwrite(&s,2,1,f);          // file output
  return 0;

from : ---
comment : Did anyone try this? How is the antialiasing compared to applying phaserror between two oscs in zerocross, one aliasing the other not (but pitcherror). Best Regards, Arif Ove Karlsen.

from : kasper756[AT]unique-papers[DOT]com
comment : Try to search something related to this good post with new search engine <a href="">Bing</a> by MS itself.

from : Jessica33[AT]unique-papers[DOT]com
comment : It is beyond question that the experienced <a href="">essay order</a> service could compose the premium quality <a href="">already written essays</a> for people which do not know how to write. This can be the most simple way for them, I do guess!

from : LilyuQ[AT]unique-papers[DOT]com
comment : I want thank you so much for the good enough topic close to this good post! Moreover, you know, that the <a href="">freelance writers </a> will take it for the essays.

from : AmeliaCR[AT]unique-papers[DOT]com
comment : You are very master and your outcome just about this good post supposes to be superior. Would you continue your investigation? I have to order the graduate thesis or <a href="">thesis writing</a> from you.

from : Yu22Odry[AT]unique-papers[DOT]com
comment : Really good way to receive an academic success was to purchase the useful <a href="">literature essay writing</a> or stuff about this topic, or just get the information about <a href="">how it works</a> selecting the essay writing service.

from : penrose[AT]alumni[DOT]princeton[DOT]edu
comment : The implementation certainly produces aliased waveforms -- they are glaring on a spectrogram at -60dB and faint at -30dB. But it is a remarkably efficient algorithm. The aliasing can be mitigated somewhat by using a smaller feedback coefficient.