encode.c

ISO mp3 sources (distribution 10) Layer 1/2/3, C Source, 512 k Sources of the Mpeg 1,2 layer 1,2

/**********************************************************************
 * ISO MPEG Audio Subgroup Software Simulation Group (1996)
 * ISO 13818-3 MPEG-2 Audio Encoder - Lower Sampling Frequency Extension
 *
 * $Id: encode.c,v 1.1 1996/02/14 04:04:23 rowlands Exp $
 *
 * $Log: encode.c,v $
 * Revision 1.1  1996/02/14 04:04:23  rowlands
 * Initial revision
 *
 * Received from Mike Coleman
 **********************************************************************/
/**********************************************************************
 *   date   programmers         comment                               *
 * 3/01/91  Douglas Wong,       start of version 1.1 records          *
 *          Davis Pan                                                 *
 * 3/06/91  Douglas Wong        rename: setup.h to endef.h            *
 *                                      efilter to enfilter           *
 *                                      ewindow to enwindow           *
 *                              integrated "quantizer", "scalefactor",*
 *                              and "transmission" files              *
 *                              update routine "window_subband"       *
 * 3/31/91  Bill Aspromonte     replaced read_filter by               *
 *                              create_an_filter                      *
 * 5/10/91  W. Joseph Carter    Ported to Macintosh and Unix.         *
 *                              Incorporated Jean-Georges Fritsch's   *
 *                              "bitstream.c" package.                *
 *                              Incorporated Bill Aspromonte's        *
 *                              filterbank coefficient matrix         *
 *                              calculation routines and added        *
 *                              roundoff to coincide with specs.      *
 *                              Modified to strictly adhere to        *
 *                              encoded bitstream specs, including    *
 *                              "Berlin changes".                     *
 *                              Modified PCM sound file handling to   *
 *                              process all incoming samples and fill *
 *                              out last encoded frame with zeros     *
 *                              (silence) if needed.                  *
 *                              Located and fixed numerous software   *
 *                              bugs and table data errors.           *
 * 19jun91  dpwe (Aware)        moved "alloc_*" reader to common.c    *
 *                              Globals sblimit, alloc replaced by new*
 *                              struct 'frame_params' passed as arg.  *
 *                              Added JOINT STEREO coding, layers I,II*
 *                              Affects: *_bit_allocation,            *
 *                              subband_quantization, encode_bit_alloc*
 *                              sample_encoding                       *
 * 6/10/91  Earle Jennings      modified II_subband_quantization to   *
 *                              resolve type cast problem for MS_DOS  *
 * 6/11/91  Earle Jennings      modified to avoid overflow on MS_DOS  *
 *                              in routine filter_subband             *
 * 7/10/91  Earle Jennings      port to MsDos from MacIntosh version  *
 * 8/ 8/91  Jens Spille         Change for MS-C6.00                   *
 *10/ 1/91  S.I. Sudharsanan,   Ported to IBM AIX platform.           *
 *          Don H. Lee,                                               *
 *          Peter W. Farrett                                          *
 *10/ 3/91  Don H. Lee          implemented CRC-16 error protection   *
 *                              newly introduced function encode_CRC  *
 *11/ 8/91  Kathy Wang          Documentation of code                 *
 *                              All variablenames are referred to     *
 *                              with surrounding pound (#) signs      *
 * 2/11/92  W. Joseph Carter    Ported new code to Macintosh.  Most   *
 *                              important fixes involved changing     *
 *                              16-bit ints to long or unsigned in    *
 *                              bit alloc routines for quant of 65535 *
 *                              and passing proper function args.     *
 *                              Removed "Other Joint Stereo" option   *
 *                              and made bitrate be total channel     *
 *                              bitrate, irrespective of the mode.    *
 *                              Fixed many small bugs & reorganized.  *
 * 6/16/92  Shaun Astarabadi    Changed I_scale_factor_calc() and     *
 *                              II_scale_factor_calc() to use scale   *
 *                              factor 0 thru 62 only and not to      *
 *                              encode index 63 into the bit stream.  *
 * 7/27/92  Mike Li             (re-)Port to MS-DOS                   *
 * 9/22/92  [e-mail]  Fixed _scale_factor_calc() defs       *
 * 3/31/93  Giogio Dimino       changed II_a_bit_allocation() from:   *
 *                              if( ad > ...) to if(ad >= ...)        *
 * 8/05/93  TEST                changed I_a_bit_allocation() from:    *
 *                              if( ad > ...) to if(ad >= ...)        *
 * 8/02/95  [e-mail]     Changed audio file reading code to    *
 *                              read samples big-endian               *
 *10/15/95  [e-mail]     Modified get_audio() for layer3-LSF   *
 **********************************************************************/
 
#include "common.h"
#include "encoder.h"

#ifdef MS_DOS
extern unsigned _stklen = 16384;
#endif


/*=======================================================================\
|                                                                       |
| This segment contains all the core routines of the encoder,           |
| except for the psychoacoustic models.                                 |
|                                                                       |
| The user can select either one of the two psychoacoustic              |
| models. Model I is a simple tonal and noise masking threshold         |
| generator, and Model II is a more sophisticated cochlear masking      |
| threshold generator. Model I is recommended for lower complexity      |
| applications whereas Model II gives better subjective quality at low  |
| bit rates.                                                            |
|                                                                       |
| Layers I and II of mono, stereo, and joint stereo modes are supported.|
| Routines associated with a given layer are prefixed by "I_" for layer |
| 1 and "II_" for layer 2.                                              |
\=======================================================================*/
 
/************************************************************************
*
* read_samples()
*
* PURPOSE:  reads the PCM samples from a file to the buffer
*
*  SEMANTICS:
* Reads #samples_read# number of shorts from #musicin# filepointer
* into #sample_buffer[]#.  Returns the number of samples read.
*
************************************************************************/

unsigned long read_samples(musicin, sample_buffer, num_samples, frame_size)
FILE *musicin;
short sample_buffer[2304];
unsigned long num_samples, frame_size;
{
    unsigned long samples_read;
    static unsigned long samples_to_read;
    static char init = TRUE;

    if (init) {
        samples_to_read = num_samples;
        init = FALSE;
    }
    if (samples_to_read >= frame_size)
        samples_read = frame_size;
    else
        samples_read = samples_to_read;
    if ((samples_read =
         fread(sample_buffer, sizeof(short), (int)samples_read, musicin)) == 0)
        printf("Hit end of audio data\n");
    /*
       Samples are big-endian. If this is a little-endian machine
       we must swap
     */
    if ( NativeByteOrder == order_unknown )
      {
	NativeByteOrder = DetermineByteOrder();
	if ( NativeByteOrder == order_unknown )
	  {
	    fprintf( stderr, "byte order not determined\n" );
	    exit( 1 );
	  }
      }
    if ( NativeByteOrder == order_littleEndian )
      SwapBytesInWords( sample_buffer, samples_read );

    samples_to_read -= samples_read;
    if (samples_read < frame_size && samples_read > 0) {
        printf("Insufficient PCM input for one frame - fillout with zeros\n");
        for (; samples_read < frame_size; sample_buffer[samples_read++] = 0);
        samples_to_read = 0;
    }
    return(samples_read);
}

/************************************************************************
*
* get_audio()
*
* PURPOSE:  reads a frame of audio data from a file to the buffer,
*   aligns the data for future processing, and separates the
*   left and right channels
*
*
************************************************************************/
 
unsigned long get_audio( musicin, buffer, num_samples, stereo, info )
FILE *musicin;
short FAR buffer[2][1152];
unsigned long num_samples;
int stereo;
layer *info;
{
    int j;
    short insamp[2304];
    unsigned long samples_read;
    int lay;
    lay = info->lay;

    if ( (lay == 3) && (info->version == 0) )
    {
	if ( stereo == 2 )
	{
	    samples_read = read_samples( musicin, insamp, num_samples,
					 (unsigned long) 1152 );
	    for ( j = 0; j < 576; j++ )
	    {
		buffer[0][j] = insamp[2 * j];
		buffer[1][j] = insamp[2 * j + 1];
	    }
	}
	else
	{
	    samples_read = read_samples( musicin, insamp, num_samples,
					 (unsigned long) 576 );
	    for ( j = 0; j < 576; j++ )
	    {
		buffer[0][j] = insamp[j];
		buffer[1][j] = 0;
	    }
	}
    }
    else
    {
	if (lay == 1){
	    if(stereo == 2){ /* layer 1, stereo */
		samples_read = read_samples(musicin, insamp, num_samples,
					    (unsigned long) 768);
		for(j=0;j<448;j++) {
		    if(j<64) {
			buffer[0][j] = buffer[0][j+384];
			buffer[1][j] = buffer[1][j+384];
		    }
		    else {
			buffer[0][j] = insamp[2*j-128];
			buffer[1][j] = insamp[2*j-127];
		    }
		}
	    }
	    else { /* layer 1, mono */
		samples_read = read_samples(musicin, insamp, num_samples,
					    (unsigned long) 384);
		for(j=0;j<448;j++){
		    if(j<64) {
			buffer[0][j] = buffer[0][j+384];
			buffer[1][j] = 0;
		    }
		    else {
			buffer[0][j] = insamp[j-64];
			buffer[1][j] = 0;
		    }
		}
	    }
	}
	else {
	    if(stereo == 2){ /* layer 2 (or 3), stereo */
		samples_read = read_samples(musicin, insamp, num_samples,
					    (unsigned long) 2304);
		for(j=0;j<1152;j++) {
		    buffer[0][j] = insamp[2*j];
		    buffer[1][j] = insamp[2*j+1];
		}
	    }
	    else { /* layer 2 (or 3), mono */
		samples_read = read_samples(musicin, insamp, num_samples,
					    (unsigned long) 1152);
		for(j=0;j<1152;j++){
		    buffer[0][j] = insamp[j];
		    buffer[1][j] = 0;
		}
	    }
	}
    }
    return(samples_read);
}
 
/************************************************************************
*
* read_ana_window()
*
* PURPOSE:  Reads encoder window file "enwindow" into array #ana_win#
*
************************************************************************/
 
void read_ana_window(ana_win)
double FAR ana_win[HAN_SIZE];
{
    int i,j[4];
    FILE *fp;
    double f[4];
    char t[150];
 
    if (!(fp = OpenTableFile("enwindow") ) ) {
       printf("Please check analysis window table 'enwindow'\n");
       exit(1);
    }
    for (i=0;i<512;i+=4) {
       fgets(t, 150, fp);
       sscanf(t,"C[%d] = %lf C[%d] = %lf C[%d] = %lf C[%d] = %lf\n",
              j, f,j+1,f+1,j+2,f+2,j+3,f+3);
       if (i==j[0]) {
          ana_win[i] = f[0];
          ana_win[i+1] = f[1];
          ana_win[i+2] = f[2];
          ana_win[i+3] = f[3];
       }
       else {
          printf("Check index in analysis window table\n");
          exit(1);
       }
       fgets(t,150,fp);
    }
    fclose(fp);
}

/************************************************************************
*
* window_subband()
*
* PURPOSE:  Overlapping window on PCM samples
*
* SEMANTICS:
* 32 16-bit pcm samples are scaled to fractional 2's complement and
* concatenated to the end of the window buffer #x#. The updated window
* buffer #x# is then windowed by the analysis window #c# to produce the
* windowed sample #z#
*
************************************************************************/
 
void window_subband(buffer, z, k)
short FAR **buffer;
double FAR z[HAN_SIZE];
int k;
{
    typedef double FAR XX[2][HAN_SIZE];
    static XX FAR *x;
    int i, j;
    static off[2] = {0,0};
    static char init = 0;
    static double FAR *c;
    if (!init) {
        c = (double FAR *) mem_alloc(sizeof(double) * HAN_SIZE, "window");
        read_ana_window(c);
        x = (XX FAR *) mem_alloc(sizeof(XX),"x");
        for (i=0;i<2;i++)
            for (j=0;j<HAN_SIZE;j++)
                (*x)[i][j] = 0;
        init = 1;
    }

    /* replace 32 oldest samples with 32 new samples */
    for (i=0;i<32;i++) (*x)[k][31-i+off[k]] = (double) *(*buffer)++/SCALE;
    /* shift samples into proper window positions */
    for (i=0;i<HAN_SIZE;i++) z[i] = (*x)[k][(i+off[k])&HAN_SIZE-1] * c[i];
    off[k] += 480;              /*offset is modulo (HAN_SIZE-1)*/
    off[k] &= HAN_SIZE-1;

}
 
/************************************************************************
*
* create_ana_filter()
*
* PURPOSE:  Calculates the analysis filter bank coefficients
*
* SEMANTICS:
* Calculates the analysis filterbank coefficients and rounds to the
* 9th decimal place accuracy of the filterbank tables in the ISO
* document.  The coefficients are stored in #filter#

************************************************************************/
 
void create_ana_filter(filter)