Include toolame-dab as library

1 files changed, 1224 insertions, 0 deletions

diff --git a/libtoolame-dab/encode.c b/libtoolame-dab/encode.c
new file mode 100644
index 0000000..fa9bd4a
--- /dev/null
+++ b/libtoolame-dab/encode.c
@@ -0,0 +1,1224 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <math.h>
+#include "common.h"
+#include "encoder.h"
+#include "bitstream.h"
+#include "availbits.h"
+#include "encode.h"
+
+int vbrstats[15] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+
+/*  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.  */
+
+/************************************************************************
+* encode_info()
+*
+* PURPOSE:  Puts the syncword and header information on the output
+* bitstream.
+*
+************************************************************************/
+
+void encode_info (frame_info * frame, Bit_stream_struc * bs)
+{
+  frame_header *header = frame->header;
+
+  putbits (bs, 0xfff, 12);	/* syncword 12 bits */
+  put1bit (bs, header->version);	/* ID        1 bit  */
+  putbits (bs, 4 - header->lay, 2);	/* layer     2 bits */
+  put1bit (bs, !header->error_protection);	/* bit set => no err prot */
+  putbits (bs, header->bitrate_index, 4);
+  putbits (bs, header->sampling_frequency, 2);
+  put1bit (bs, header->padding);
+  put1bit (bs, header->extension);	/* private_bit */
+  putbits (bs, header->mode, 2);
+  putbits (bs, header->mode_ext, 2);
+  put1bit (bs, header->copyright);
+  put1bit (bs, header->original);
+  putbits (bs, header->emphasis, 2);
+}
+
+/************************************************************************
+*
+* combine_LR   (Layer II)
+*
+* PURPOSE:Combines left and right channels into a mono channel
+*
+* SEMANTICS:  The average of left and right subband samples is put into
+* #joint_sample#
+*
+*
+************************************************************************/
+
+void combine_LR (double sb_sample[2][3][SCALE_BLOCK][SBLIMIT],
+		 double joint_sample[3][SCALE_BLOCK][SBLIMIT], int sblimit)
+{				/* make a filtered mono for joint stereo */
+  int sb, smp, sufr;
+
+  for (sb = 0; sb < sblimit; ++sb)
+    for (smp = 0; smp < SCALE_BLOCK; ++smp)
+      for (sufr = 0; sufr < 3; ++sufr)
+	joint_sample[sufr][smp][sb] =
+	  .5 * (sb_sample[0][sufr][smp][sb] + sb_sample[1][sufr][smp][sb]);
+}
+
+/************************************************************************
+*
+* scale_factor_calc    (Layer II)
+*
+* PURPOSE:For each subband, calculate the scale factor for each set
+* of the 12 subband samples
+*
+* SEMANTICS:  Pick the scalefactor #multiple[]# just larger than the
+* absolute value of the peak subband sample of 12 samples,
+* and store the corresponding scalefactor index in #scalar#.
+*
+* Layer II has three sets of 12-subband samples for a given
+* subband.
+*
+************************************************************************/
+
+#define PDS1
+#ifdef PDS1
+void scale_factor_calc (double sb_sample[][3][SCALE_BLOCK][SBLIMIT],
+			unsigned int scalar[][3][SBLIMIT], int nch,
+			int sblimit)
+{
+  /* Optimized to use binary search instead of linear scan through the
+     scalefactor table; guarantees to find scalefactor in only 5
+     jumps/comparisons and not in {0 (lin. best) to 63 (lin. worst)}.
+     Scalefactors for subbands > sblimit are no longer computed.
+     Uses a single sblimit-loop.
+     Patrick De Smet Oct 1999.
+   */
+  int k, t;
+  /* Using '--' loops to avoid possible "cmp value + bne/beq" compiler  */
+  /* inefficiencies. Below loops should compile to "bne/beq" only code  */
+  for (k = nch; k--;)
+    for (t = 3; t--;) {
+      int i;
+      for (i = sblimit; i--;) {
+	int j;
+	unsigned int l;
+	register double temp;
+	unsigned int scale_fac;
+	/* Determination of max. over each set of 12 subband samples:  */
+	/* PDS TODO: maybe this could/should ??!! be integrated into   */
+	/* the subband filtering routines?                             */
+	register double cur_max = fabs (sb_sample[k][t][SCALE_BLOCK - 1][i]);
+	for (j = SCALE_BLOCK - 1; j--;) {
+	  if ((temp = fabs (sb_sample[k][t][j][i])) > cur_max)
+	    cur_max = temp;
+	}
+	/* PDS: binary search in the scalefactor table: */
+	/* This is the real speed up: */
+	for (l = 16, scale_fac = 32; l; l >>= 1) {
+	  if (cur_max <= multiple[scale_fac])
+	    scale_fac += l;
+	  else
+	    scale_fac -= l;
+	}
+	if (cur_max > multiple[scale_fac])
+	  scale_fac--;
+	scalar[k][t][i] = scale_fac;
+      }
+    }
+}
+#else
+void scale_factor_calc (sb_sample, scalar, nch, sblimit)
+     double sb_sample[][3][SCALE_BLOCK][SBLIMIT];
+     unsigned int scalar[][3][SBLIMIT];
+     int nch, sblimit;
+{
+  int i, j, k, t;
+  double s[SBLIMIT];
+
+  for (k = 0; k < nch; k++)
+    for (t = 0; t < 3; t++) {
+      for (i = 0; i < sblimit; i++)
+	for (j = 1, s[i] = fabs (sb_sample[k][t][0][i]); j < SCALE_BLOCK; j++)
+	  if (fabs (sb_sample[k][t][j][i]) > s[i])
+	    s[i] = fabs (sb_sample[k][t][j][i]);
+
+      for (i = 0; i < sblimit; i++)
+	for (j = SCALE_RANGE - 2, scalar[k][t][i] = 0; j >= 0; j--)	/* $A 6/16/92 */
+	  if (s[i] <= multiple[j]) {
+	    scalar[k][t][i] = j;
+	    break;
+	  }
+      for (i = sblimit; i < SBLIMIT; i++)
+	scalar[k][t][i] = SCALE_RANGE - 1;
+    }
+}
+
+#endif
+/************************************************************************
+*
+* pick_scale  (Layer II)
+*
+* PURPOSE:For each subband, puts the smallest scalefactor of the 3
+* associated with a frame into #max_sc#.  This is used
+* used by Psychoacoustic Model I.
+* (I would recommend changin max_sc to min_sc)
+*
+************************************************************************/
+
+void pick_scale (unsigned int scalar[2][3][SBLIMIT], frame_info * frame,
+		 double max_sc[2][SBLIMIT])
+{
+  int i, j, k, max;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+
+  for (k = 0; k < nch; k++)
+    for (i = 0; i < sblimit; max_sc[k][i] = multiple[max], i++)
+      for (j = 1, max = scalar[k][0][i]; j < 3; j++)
+	if (max > scalar[k][j][i])
+	  max = scalar[k][j][i];
+  for (i = sblimit; i < SBLIMIT; i++)
+    max_sc[0][i] = max_sc[1][i] = 1E-20;
+}
+
+/************************************************************************
+*
+* transmission_pattern (Layer II only)
+*
+* PURPOSE:For a given subband, determines whether to send 1, 2, or
+* all 3 of the scalefactors, and fills in the scalefactor
+* select information accordingly
+*
+* SEMANTICS:  The subbands and channels are classified based on how much
+* the scalefactors changes over its three values (corresponding
+* to the 3 sets of 12 samples per subband).  The classification
+* will send 1 or 2 scalefactors instead of three if the scalefactors
+* do not change much.  The scalefactor select information,
+* #scfsi#, is filled in accordingly.
+*
+************************************************************************/
+
+void transmission_pattern (unsigned int scalar[2][3][SBLIMIT],
+			   unsigned int scfsi[2][SBLIMIT],
+			   frame_info * frame)
+{
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int dscf[2];
+  int class[2], i, j, k;
+  static int pattern[5][5] = { {0x123, 0x122, 0x122, 0x133, 0x123},
+  {0x113, 0x111, 0x111, 0x444, 0x113},
+  {0x111, 0x111, 0x111, 0x333, 0x113},
+  {0x222, 0x222, 0x222, 0x333, 0x123},
+  {0x123, 0x122, 0x122, 0x133, 0x123}
+  };
+
+  for (k = 0; k < nch; k++)
+    for (i = 0; i < sblimit; i++) {
+      dscf[0] = (scalar[k][0][i] - scalar[k][1][i]);
+      dscf[1] = (scalar[k][1][i] - scalar[k][2][i]);
+      for (j = 0; j < 2; j++) {
+	if (dscf[j] <= -3)
+	  class[j] = 0;
+	else if (dscf[j] > -3 && dscf[j] < 0)
+	  class[j] = 1;
+	else if (dscf[j] == 0)
+	  class[j] = 2;
+	else if (dscf[j] > 0 && dscf[j] < 3)
+	  class[j] = 3;
+	else
+	  class[j] = 4;
+      }
+      switch (pattern[class[0]][class[1]]) {
+      case 0x123:
+	scfsi[k][i] = 0;
+	break;
+      case 0x122:
+	scfsi[k][i] = 3;
+	scalar[k][2][i] = scalar[k][1][i];
+	break;
+      case 0x133:
+	scfsi[k][i] = 3;
+	scalar[k][1][i] = scalar[k][2][i];
+	break;
+      case 0x113:
+	scfsi[k][i] = 1;
+	scalar[k][1][i] = scalar[k][0][i];
+	break;
+      case 0x111:
+	scfsi[k][i] = 2;
+	scalar[k][1][i] = scalar[k][2][i] = scalar[k][0][i];
+	break;
+      case 0x222:
+	scfsi[k][i] = 2;
+	scalar[k][0][i] = scalar[k][2][i] = scalar[k][1][i];
+	break;
+      case 0x333:
+	scfsi[k][i] = 2;
+	scalar[k][0][i] = scalar[k][1][i] = scalar[k][2][i];
+	break;
+      case 0x444:
+	scfsi[k][i] = 2;
+	if (scalar[k][0][i] > scalar[k][2][i])
+	  scalar[k][0][i] = scalar[k][2][i];
+	scalar[k][1][i] = scalar[k][2][i] = scalar[k][0][i];
+      }
+    }
+}
+
+/************************************************************************
+*
+* encode_scale (Layer II)
+*
+* PURPOSE:The encoded scalar factor information is arranged and
+* queued into the output fifo to be transmitted.
+*
+* For Layer II, the three scale factors associated with
+* a given subband and channel are transmitted in accordance
+* with the scfsi, which is transmitted first.
+*
+************************************************************************/
+
+void
+encode_scale (unsigned int bit_alloc[2][SBLIMIT],
+	      unsigned int scfsi[2][SBLIMIT],
+	      unsigned int scalar[2][3][SBLIMIT], frame_info * frame,
+	      Bit_stream_struc * bs)
+{
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int i, j, k;
+
+  for (i = 0; i < sblimit; i++)
+    for (k = 0; k < nch; k++)
+      if (bit_alloc[k][i])
+	putbits (bs, scfsi[k][i], 2);
+
+  for (i = 0; i < sblimit; i++)
+    for (k = 0; k < nch; k++)
+      if (bit_alloc[k][i])	/* above jsbound, bit_alloc[0][i] == ba[1][i] */
+	switch (scfsi[k][i]) {
+	case 0:
+	  for (j = 0; j < 3; j++)
+	    putbits (bs, scalar[k][j][i], 6);
+	  break;
+	case 1:
+	case 3:
+	  putbits (bs, scalar[k][0][i], 6);
+	  putbits (bs, scalar[k][2][i], 6);
+	  break;
+	case 2:
+	  putbits (bs, scalar[k][0][i], 6);
+	}
+}
+
+/*=======================================================================\
+|                                                                        |
+|      The following routines are done after the masking threshold       |
+| has been calculated by the fft analysis routines in the Psychoacoustic |
+| model. Using the MNR calculated, the actual number of bits allocated   |
+| to each subband is found iteratively.                                  |
+|                                                                        |
+\=======================================================================*/
+
+/************************************************************************
+*
+* bits_for_nonoise (Layer II)
+*
+* PURPOSE:Returns the number of bits required to produce a
+* mask-to-noise ratio better or equal to the noise/no_noise threshold.
+*
+* SEMANTICS:
+* bbal = # bits needed for encoding bit allocation
+* bsel = # bits needed for encoding scalefactor select information
+* banc = # bits needed for ancillary data (header info included)
+*
+* For each subband and channel, will add bits until one of the
+* following occurs:
+* - Hit maximum number of bits we can allocate for that subband
+* - MNR is better than or equal to the minimum masking level
+*   (NOISY_MIN_MNR)
+* Then the bits required for scalefactors, scfsi, bit allocation,
+* and the subband samples are tallied (#req_bits#) and returned.
+*
+* (NOISY_MIN_MNR) is the smallest MNR a subband can have before it is
+* counted as 'noisy' by the logic which chooses the number of JS
+* subbands.
+*
+* Joint stereo is supported.
+*
+************************************************************************/
+
+static double snr[18] = { 0.00, 7.00, 11.00, 16.00, 20.84,
+  25.28, 31.59, 37.75, 43.84,
+  49.89, 55.93, 61.96, 67.98, 74.01,
+  80.03, 86.05, 92.01, 98.01
+};
+
+int bits_for_nonoise (double perm_smr[2][SBLIMIT],
+		      unsigned int scfsi[2][SBLIMIT], frame_info * frame)
+{
+  int sb, ch, ba;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+  int req_bits = 0, bbal = 0, berr = 0, banc = 32;
+  int maxAlloc, sel_bits, sc_bits, smp_bits;
+  static int sfsPerScfsi[] = { 3, 2, 1, 2 };	/* lookup # sfs per scfsi */
+
+  /* added 92-08-11 shn */
+  if (frame->header->error_protection)
+    berr = 16;
+  else
+    berr = 0;
+
+  for (sb = 0; sb < jsbound; ++sb)
+    bbal += nch * (*alloc)[sb][0].bits;
+  for (sb = jsbound; sb < sblimit; ++sb)
+    bbal += (*alloc)[sb][0].bits;
+  req_bits = banc + bbal + berr;
+
+  for (sb = 0; sb < sblimit; ++sb)
+    for (ch = 0; ch < ((sb < jsbound) ? nch : 1); ++ch) {
+      maxAlloc = (1 << (*alloc)[sb][0].bits) - 1;
+      sel_bits = sc_bits = smp_bits = 0;
+      for (ba = 0; ba < maxAlloc - 1; ++ba)
+	if ((-perm_smr[ch][sb] +
+	     snr[(*alloc)[sb][ba].quant + ((ba > 0) ? 1 : 0)]) >=
+	    NOISY_MIN_MNR)
+	  break;		/* we found enough bits */
+      if (nch == 2 && sb >= jsbound)	/* check other JS channel */
+	for (; ba < maxAlloc - 1; ++ba)
+	  if ((-perm_smr[1 - ch][sb] +
+	       snr[(*alloc)[sb][ba].quant + ((ba > 0) ? 1 : 0)]) >=
+	      NOISY_MIN_MNR)
+	    break;
+      if (ba > 0) {
+	smp_bits =
+	  SCALE_BLOCK * ((*alloc)[sb][ba].group * (*alloc)[sb][ba].bits);
+	/* scale factor bits required for subband */
+	sel_bits = 2;
+	sc_bits = 6 * sfsPerScfsi[scfsi[ch][sb]];
+	if (nch == 2 && sb >= jsbound) {
+	  /* each new js sb has L+R scfsis */
+	  sel_bits += 2;
+	  sc_bits += 6 * sfsPerScfsi[scfsi[1 - ch][sb]];
+	}
+	req_bits += smp_bits + sel_bits + sc_bits;
+      }
+    }
+  return req_bits;
+}
+
+int VBR_bits_for_nonoise (double perm_smr[2][SBLIMIT],
+			  unsigned int scfsi[2][SBLIMIT],
+			  frame_info * frame, int vbrlevel)
+{
+  int sb, ch, ba;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+  int req_bits = 0, bbal = 0, berr = 0, banc = 32;
+  int maxAlloc, sel_bits, sc_bits, smp_bits;
+  static int sfsPerScfsi[] = { 3, 2, 1, 2 };	/* lookup # sfs per scfsi */
+
+  /* added 92-08-11 shn */
+  if (frame->header->error_protection)
+    berr = 16;
+  else
+    berr = 0;
+
+  for (sb = 0; sb < jsbound; ++sb)
+    bbal += nch * (*alloc)[sb][0].bits;
+  for (sb = jsbound; sb < sblimit; ++sb)
+    bbal += (*alloc)[sb][0].bits;
+  req_bits = banc + bbal + berr;
+
+  for (sb = 0; sb < sblimit; ++sb)
+    for (ch = 0; ch < ((sb < jsbound) ? nch : 1); ++ch) {
+      maxAlloc = (1 << (*alloc)[sb][0].bits) - 1;
+      sel_bits = sc_bits = smp_bits = 0;
+      for (ba = 0; ba < maxAlloc - 1; ++ba)
+	/* The change between this function and the normal one is that the MIN_MNR is increased by the vbrlevel */
+	if ((-perm_smr[ch][sb] +
+	     snr[(*alloc)[sb][ba].quant + ((ba > 0) ? 1 : 0)]) >=
+	    NOISY_MIN_MNR + vbrlevel)
+	  break;		/* we found enough bits */
+      if (nch == 2 && sb >= jsbound)	/* check other JS channel */
+	for (; ba < maxAlloc - 1; ++ba)
+	  if ((-perm_smr[1 - ch][sb] +
+	       snr[(*alloc)[sb][ba].quant + ((ba > 0) ? 1 : 0)]) >=
+	      NOISY_MIN_MNR + vbrlevel)
+	    break;
+      if (ba > 0) {
+	smp_bits =
+	  SCALE_BLOCK * ((*alloc)[sb][ba].group * (*alloc)[sb][ba].bits);
+	/* scale factor bits required for subband */
+	sel_bits = 2;
+	sc_bits = 6 * sfsPerScfsi[scfsi[ch][sb]];
+	if (nch == 2 && sb >= jsbound) {
+	  /* each new js sb has L+R scfsis */
+	  sel_bits += 2;
+	  sc_bits += 6 * sfsPerScfsi[scfsi[1 - ch][sb]];
+	}
+	req_bits += smp_bits + sel_bits + sc_bits;
+      }
+    }
+  return req_bits;
+}
+
+/************************************************************************
+*
+* main_bit_allocation  (Layer II)
+*
+* PURPOSE:For joint stereo mode, determines which of the 4 joint
+* stereo modes is needed.  Then calls *_a_bit_allocation(), which
+* allocates bits for each of the subbands until there are no more bits
+* left, or the MNR is at the noise/no_noise threshold.
+*
+* SEMANTICS:
+*
+* For joint stereo mode, joint stereo is changed to stereo if
+* there are enough bits to encode stereo at or better than the
+* no-noise threshold (NOISY_MIN_MNR).  Otherwise, the system
+* iteratively allocates less bits by using joint stereo until one
+* of the following occurs:
+* - there are no more noisy subbands (MNR >= NOISY_MIN_MNR)
+* - mode_ext has been reduced to 0, which means that all but the
+*   lowest 4 subbands have been converted from stereo to joint
+*   stereo, and no more subbands may be converted
+*
+*     This function calls *_bits_for_nonoise() and *_a_bit_allocation().
+*
+************************************************************************/
+void main_bit_allocation (double perm_smr[2][SBLIMIT],
+			  unsigned int scfsi[2][SBLIMIT],
+			  unsigned int bit_alloc[2][SBLIMIT], int *adb,
+			  frame_info * frame, options * glopts)
+{
+  int noisy_sbs;
+  int mode, mode_ext, lay;
+  int rq_db;			/* av_db = *adb; Not Used MFC Nov 99 */
+
+  /* these are the tables which specify the limits within which the VBR can vary 
+     You can't vary outside these ranges, otherwise a new alloc table would have to 
+     be loaded in the middle of encoding. This VBR hack is dodgy - the standard
+     says that LayerII decoders don't have to support a variable bitrate, but Layer3
+     decoders must do so. Hence, it is unlikely that a compliant layer2 decoder would be 
+     written to dynmically change allocation tables. *BUT* a layer3 encoder might handle it
+     by default meaning we could switch tables mid-encode and enjoy a wider range of bitrates
+     for the VBR encoding. 
+     None of this needs to be done for LSF, since there is only *one* possible alloc table in LSF 
+     MFC Feb 2003 */
+  int vbrlimits[2][3][2] = {
+    /* MONO */
+    { /* 44 */ {6, 10},
+     /* 48 */ {3, 10},
+     /* 32 */ {6, 10}},
+    /* STEREO */
+    { /* 44 */ {10, 14},
+     /* 48 */ {7, 14},
+     /* 32 */ {10, 14}}
+  };
+
+  static int init = 0;
+  static int lower = 10, upper = 10;
+  static int bitrateindextobits[15] =
+    { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
+  int guessindex = 0;
+
+  if (init == 0) {
+    int nch = 1;
+    int sfreq;
+    frame_header *header = frame->header;
+    init++;
+    if (header->version == 0) {
+      /* LSF: so can use any bitrate index from 1->15 */
+      lower = 1;
+      upper = 14;
+    } else {
+      if (frame->actual_mode == MPG_MD_MONO)
+	nch = 0;
+      sfreq = header->sampling_frequency;
+      lower = vbrlimits[nch][sfreq][0];
+      upper = vbrlimits[nch][sfreq][1];
+    }
+    if (glopts->verbosity > 2)
+      fprintf (stdout, "VBR bitrate index limits [%i -> %i]\n", lower, upper);
+
+    {				
+      /* set up a conversion table for bitrateindex->bits for this version/sampl freq 
+	 This will be used to find the best bitrate to cope with the number of bits that
+	 are needed (as determined by VBR_bits_for_nonoise) */
+      int brindex;
+      frame_header *header = frame->header;
+      for (brindex = lower; brindex <= upper; brindex++) {
+	bitrateindextobits[brindex] =
+	  (int) (1152.0 / s_freq[header->version][header->sampling_frequency]) *
+	  ((double) bitrate[header->version][brindex]);
+      }
+    }
+
+  }
+
+  if ((mode = frame->actual_mode) == MPG_MD_JOINT_STEREO) {
+    frame->header->mode = MPG_MD_STEREO;
+    frame->header->mode_ext = 0;
+    frame->jsbound = frame->sblimit;
+    if ((rq_db = bits_for_nonoise (perm_smr, scfsi, frame)) > *adb) {
+      frame->header->mode = MPG_MD_JOINT_STEREO;
+      mode_ext = 4;		/* 3 is least severe reduction */
+      lay = frame->header->lay;
+      do {
+	--mode_ext;
+	frame->jsbound = js_bound (mode_ext);
+	rq_db = bits_for_nonoise (perm_smr, scfsi, frame);
+      }
+      while ((rq_db > *adb) && (mode_ext > 0));
+      frame->header->mode_ext = mode_ext;
+    }				/* well we either eliminated noisy sbs or mode_ext == 0 */
+  }
+
+  /* decide on which bit allocation method to use */
+  if (glopts->vbr == FALSE) {
+    /* Just do the old bit allocation method */
+    noisy_sbs = a_bit_allocation (perm_smr, scfsi, bit_alloc, adb, frame);
+  } else {			
+    /* do the VBR bit allocation method */
+    frame->header->bitrate_index = lower;
+    *adb = available_bits (frame->header, glopts);
+    {
+      int brindex;
+      int found = FALSE;
+
+      /* Work out how many bits are needed for there to be no noise (ie all MNR > 0.0 + VBRLEVEL) */
+      int req =
+	VBR_bits_for_nonoise (perm_smr, scfsi, frame, glopts->vbrlevel);
+
+      /* Look up this value in the bitrateindextobits table to find what bitrate we should use for 
+         this frame */
+      for (brindex = lower; brindex <= upper; brindex++) {
+	if (bitrateindextobits[brindex] > req) {
+	  /* this method always *overestimates* the bits that are needed
+	     i.e. it will usually  guess right but
+	     when it's wrong it'll guess a higher bitrate than actually required.
+	     e.g. on "messages from earth" track 6, the guess was 
+	     wrong on 75/36341 frames. each time it guessed higher. 
+	     MFC Feb 2003 */
+	  guessindex = brindex;
+	  found = TRUE;
+	  break;
+	}
+      }
+      /* Just for sanity */
+      if (found == FALSE)
+	guessindex = upper;
+    }
+
+    frame->header->bitrate_index = guessindex;
+    *adb = available_bits (frame->header, glopts);
+
+    /* update the statistics */
+    vbrstats[frame->header->bitrate_index]++;
+
+    if (glopts->verbosity > 2) {
+      /* print out the VBR stats every 1000th frame */
+      static int count = 0;
+      int i;
+      if ((count++ % 1000) == 0) {
+	for (i = 1; i < 15; i++)
+	  fprintf (stdout, "%4i ", vbrstats[i]);
+	fprintf (stdout, "\n");
+      }
+
+      /* Print out *every* frames bitrateindex, bits required, and bits available at this bitrate */
+      if (glopts->verbosity > 5)
+	fprintf (stdout,
+		 "> bitrate index %2i has %i bits available to encode the %i bits\n",
+		 frame->header->bitrate_index, *adb,
+		 VBR_bits_for_nonoise (perm_smr, scfsi, frame,
+				       glopts->vbrlevel));
+
+    }
+
+    noisy_sbs =
+      VBR_bit_allocation (perm_smr, scfsi, bit_alloc, adb, frame, glopts);
+  }
+}
+
+void VBR_maxmnr (double mnr[2][SBLIMIT], char used[2][SBLIMIT], int sblimit,
+		 int nch, int *min_sb, int *min_ch, options * glopts)
+{
+  int i, k;
+  double small;
+
+  small = 999999.0;
+  *min_sb = -1;
+  *min_ch = -1;
+  for (k = 0; k < nch; ++k)
+    for (i = 0; i < sblimit; i++)
+      if (used[k][i] != 2 && small > mnr[k][i]) {
+	small = mnr[k][i];
+	*min_sb = i;
+	*min_ch = k;
+      }
+}
+/********************
+MFC Feb 2003
+VBR_bit_allocation is different to the normal a_bit_allocation in that
+it is known beforehand that there are definitely enough bits to do what we 
+have to - i.e. a bitrate was specificially chosen in main_bit_allocation so
+that we have enough bits to encode what we have to.
+This function should take that into account and just greedily assign
+the bits, rather than fussing over the minimum MNR subband - we know
+each subband gets its required bits, why quibble?
+This function doesn't chew much CPU, so I haven't made any attempt
+to do this yet.
+*********************/
+int
+VBR_bit_allocation (double perm_smr[2][SBLIMIT],
+		    unsigned int scfsi[2][SBLIMIT],
+		    unsigned int bit_alloc[2][SBLIMIT], int *adb,
+		    frame_info * frame, options * glopts)
+{
+  int i, min_ch, min_sb, oth_ch, k, increment, scale, seli, ba;
+  int bspl, bscf, bsel, ad, bbal = 0;
+  double mnr[2][SBLIMIT];
+  char used[2][SBLIMIT];
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+  static char init = 0;
+  static int banc = 32, berr = 0;
+  static int sfsPerScfsi[] = { 3, 2, 1, 2 };	/* lookup # sfs per scfsi */
+
+  if (!init) {
+    init = 1;
+    if (frame->header->error_protection)
+      berr = 16;		/* added 92-08-11 shn */
+  }
+
+  for (i = 0; i < jsbound; ++i)
+    bbal += nch * (*alloc)[i][0].bits;
+  for (i = jsbound; i < sblimit; ++i)
+    bbal += (*alloc)[i][0].bits;
+  *adb -= bbal + berr + banc;
+  ad = *adb;
+
+  for (i = 0; i < sblimit; i++)
+    for (k = 0; k < nch; k++) {
+      mnr[k][i] = snr[0] - perm_smr[k][i];
+      bit_alloc[k][i] = 0;
+      used[k][i] = 0;
+    }
+  bspl = bscf = bsel = 0;
+
+  do {
+    /* locate the subband with minimum SMR */
+    VBR_maxmnr (mnr, used, sblimit, nch, &min_sb, &min_ch, glopts);
+
+    if (min_sb > -1) {		/* there was something to find */
+      /* find increase in bit allocation in subband [min] */
+      increment =
+	SCALE_BLOCK * ((*alloc)[min_sb][bit_alloc[min_ch][min_sb] + 1].group *
+		       (*alloc)[min_sb][bit_alloc[min_ch][min_sb] + 1].bits);
+      if (used[min_ch][min_sb])
+	increment -=
+	  SCALE_BLOCK * ((*alloc)[min_sb][bit_alloc[min_ch][min_sb]].group *
+			 (*alloc)[min_sb][bit_alloc[min_ch][min_sb]].bits);
+
+      /* scale factor bits required for subband [min] */
+      oth_ch = 1 - min_ch;	/* above js bound, need both chans */
+      if (used[min_ch][min_sb])
+	scale = seli = 0;
+      else {			/* this channel had no bits or scfs before */
+	seli = 2;
+	scale = 6 * sfsPerScfsi[scfsi[min_ch][min_sb]];
+	if (nch == 2 && min_sb >= jsbound) {
+	  /* each new js sb has L+R scfsis */
+	  seli += 2;
+	  scale += 6 * sfsPerScfsi[scfsi[oth_ch][min_sb]];
+	}
+      }
+
+      /* check to see enough bits were available for */
+      /* increasing resolution in the minimum band */
+      if (ad >= bspl + bscf + bsel + seli + scale + increment) {
+	ba = ++bit_alloc[min_ch][min_sb];	/* next up alloc */
+	bspl += increment;	/* bits for subband sample */
+	bscf += scale;		/* bits for scale factor */
+	bsel += seli;		/* bits for scfsi code */
+	used[min_ch][min_sb] = 1;	/* subband has bits */
+	mnr[min_ch][min_sb] =
+	  -perm_smr[min_ch][min_sb] + snr[(*alloc)[min_sb][ba].quant + 1];
+	/* Check if subband has been fully allocated max bits */
+	if (ba >= (1 << (*alloc)[min_sb][0].bits) - 1)
+	  used[min_ch][min_sb] = 2;	/* don't let this sb get any more bits */
+      } else
+	used[min_ch][min_sb] = 2;	/* can't increase this alloc */
+
+      if (min_sb >= jsbound && nch == 2) {
+	/* above jsbound, alloc applies L+R */
+	ba = bit_alloc[oth_ch][min_sb] = bit_alloc[min_ch][min_sb];
+	used[oth_ch][min_sb] = used[min_ch][min_sb];
+	mnr[oth_ch][min_sb] =
+	  -perm_smr[oth_ch][min_sb] + snr[(*alloc)[min_sb][ba].quant + 1];
+      }
+
+    }
+  }
+  while (min_sb > -1);		/* until could find no channel */
+
+  /* Calculate the number of bits left */
+  ad -= bspl + bscf + bsel;
+  *adb = ad;
+  for (k = 0; k < nch; k++)
+    for (i = sblimit; i < SBLIMIT; i++)
+      bit_alloc[k][i] = 0;
+
+  return 0;
+}
+
+/************************************************************************
+*
+* a_bit_allocation (Layer II)
+*
+* PURPOSE:Adds bits to the subbands with the lowest mask-to-noise
+* ratios, until the maximum number of bits for the subband has
+* been allocated.
+*
+* SEMANTICS:
+* 1. Find the subband and channel with the smallest MNR (#min_sb#,
+*    and #min_ch#)
+* 2. Calculate the increase in bits needed if we increase the bit
+*    allocation to the next higher level
+* 3. If there are enough bits available for increasing the resolution
+*    in #min_sb#, #min_ch#, and the subband has not yet reached its
+*    maximum allocation, update the bit allocation, MNR, and bits
+    available accordingly
+* 4. Repeat until there are no more bits left, or no more available
+*    subbands. (A subband is still available until the maximum
+*    number of bits for the subband has been allocated, or there
+*    aren't enough bits to go to the next higher resolution in the
+    subband.)
+*
+************************************************************************/
+
+void maxmnr (double mnr[2][SBLIMIT], char used[2][SBLIMIT], int sblimit,
+	     int nch, int *min_sb, int *min_ch)
+{
+  int i, k;
+  double small;
+
+  small = 999999.0;
+  *min_sb = -1;
+  *min_ch = -1;
+  for (k = 0; k < nch; ++k)
+    for (i = 0; i < sblimit; i++)
+      if (used[k][i] != 2 && small > mnr[k][i]) {
+	small = mnr[k][i];
+	*min_sb = i;
+	*min_ch = k;
+      }
+}
+
+int a_bit_allocation (double perm_smr[2][SBLIMIT],
+		      unsigned int scfsi[2][SBLIMIT],
+		      unsigned int bit_alloc[2][SBLIMIT], int *adb,
+		      frame_info * frame)
+{
+  int i, min_ch, min_sb, oth_ch, k, increment, scale, seli, ba;
+  int bspl, bscf, bsel, ad, bbal = 0;
+  double mnr[2][SBLIMIT];
+  char used[2][SBLIMIT];
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+  static char init = 0;
+  static int banc = 32, berr = 0;
+  static int sfsPerScfsi[] = { 3, 2, 1, 2 };	/* lookup # sfs per scfsi */
+
+#define CHECKITERx
+#ifdef CHECKITER
+  int count=0;
+#endif
+
+  if (!init) {
+    init = 1;
+    if (frame->header->error_protection)
+      berr = 16;		/* added 92-08-11 shn */
+  }
+
+  for (i = 0; i < jsbound; ++i)
+    bbal += nch * (*alloc)[i][0].bits;
+  for (i = jsbound; i < sblimit; ++i)
+    bbal += (*alloc)[i][0].bits;
+  *adb -= bbal + berr + banc;
+  ad = *adb;
+
+  for (i = 0; i < sblimit; i++)
+    for (k = 0; k < nch; k++) {
+      mnr[k][i] = snr[0] - perm_smr[k][i];
+      bit_alloc[k][i] = 0;
+      used[k][i] = 0;
+    }
+  bspl = bscf = bsel = 0;
+
+  do {
+#ifdef CHECKITER
+    count++;
+#endif
+    /* locate the subband with minimum SMR */
+    maxmnr (mnr, used, sblimit, nch, &min_sb, &min_ch);
+
+    if (min_sb > -1) {		/* there was something to find */
+      /* find increase in bit allocation in subband [min] */
+      increment =
+	SCALE_BLOCK * ((*alloc)[min_sb][bit_alloc[min_ch][min_sb] + 1].group *
+		       (*alloc)[min_sb][bit_alloc[min_ch][min_sb] + 1].bits);
+      if (used[min_ch][min_sb])
+	increment -=
+	  SCALE_BLOCK * ((*alloc)[min_sb][bit_alloc[min_ch][min_sb]].group *
+			 (*alloc)[min_sb][bit_alloc[min_ch][min_sb]].bits);
+
+      /* scale factor bits required for subband [min] */
+      oth_ch = 1 - min_ch;	/* above js bound, need both chans */
+      if (used[min_ch][min_sb])
+	scale = seli = 0;
+      else {			/* this channel had no bits or scfs before */
+	seli = 2;
+	scale = 6 * sfsPerScfsi[scfsi[min_ch][min_sb]];
+	if (nch == 2 && min_sb >= jsbound) {
+	  /* each new js sb has L+R scfsis */
+	  seli += 2;
+	  scale += 6 * sfsPerScfsi[scfsi[oth_ch][min_sb]];
+	}
+      }
+
+      /* check to see enough bits were available for */
+      /* increasing resolution in the minimum band */
+      if (ad >= bspl + bscf + bsel + seli + scale + increment) {
+	ba = ++bit_alloc[min_ch][min_sb];	/* next up alloc */
+	bspl += increment;	/* bits for subband sample */
+	bscf += scale;		/* bits for scale factor */
+	bsel += seli;		/* bits for scfsi code */
+	used[min_ch][min_sb] = 1;	/* subband has bits */
+	mnr[min_ch][min_sb] =
+	  -perm_smr[min_ch][min_sb] + snr[(*alloc)[min_sb][ba].quant + 1];
+	/* Check if subband has been fully allocated max bits */
+	if (ba >= (1 << (*alloc)[min_sb][0].bits) - 1)
+	  used[min_ch][min_sb] = 2;	/* don't let this sb get any more bits */
+      } else
+	used[min_ch][min_sb] = 2;	/* can't increase this alloc */
+
+      if (min_sb >= jsbound && nch == 2) {
+	/* above jsbound, alloc applies L+R */
+	ba = bit_alloc[oth_ch][min_sb] = bit_alloc[min_ch][min_sb];
+	used[oth_ch][min_sb] = used[min_ch][min_sb];
+	mnr[oth_ch][min_sb] =
+	  -perm_smr[oth_ch][min_sb] + snr[(*alloc)[min_sb][ba].quant + 1];
+      }
+
+    }
+  }
+  while (min_sb > -1);		/* until could find no channel */
+
+  /* Calculate the number of bits left */
+  ad -= bspl + bscf + bsel;
+  *adb = ad;
+  for (k = 0; k < nch; k++)
+    for (i = sblimit; i < SBLIMIT; i++)
+      bit_alloc[k][i] = 0;
+
+#ifdef USELESSCODE
+  /* this function is declared to return an INT, which is meant to be a count
+     of the subbands which are still noisy. But, the return value is ignored,
+     so why bother? Is the count of noisy_sbs useful as any sort of 
+     quality measure? Leave this in, until I'm sure that noisy_sbs couldn't
+     be used for something
+     MFC Feb 2003 */
+
+  noisy_sbs = 0;		/* calc worst noise in case */
+  for (k = 0; k < nch; ++k) {
+    for (i = 0; i < sblimit; i++) {
+      if (mnr[k][i] < NOISY_MIN_MNR)
+	++noisy_sbs;		/* noise is not masked */
+    }
+  }
+  return noisy_sbs;
+#endif
+#ifdef CHECKITER
+  fprintf(stdout,"a bit alloc %i\n", count);
+#endif
+  return 0;
+}
+
+/************************************************************************
+*
+* subband_quantization (Layer II)
+*
+* PURPOSE:Quantizes subband samples to appropriate number of bits
+*
+* SEMANTICS:  Subband samples are divided by their scalefactors, which
+ makes the quantization more efficient. The scaled samples are
+* quantized by the function a*x+b, where a and b are functions of
+* the number of quantization levels. The result is then truncated
+* to the appropriate number of bits and the MSB is inverted.
+*
+* Note that for fractional 2's complement, inverting the MSB for a
+ negative number x is equivalent to adding 1 to it.
+*
+************************************************************************/
+#define PDS3
+#ifdef PDS3
+static double a[17] = {
+  0.750000000, 0.625000000, 0.875000000, 0.562500000, 0.937500000,
+  0.968750000, 0.984375000, 0.992187500, 0.996093750, 0.998046875,
+  0.999023438, 0.999511719, 0.999755859, 0.999877930, 0.999938965,
+  0.999969482, 0.999984741
+};
+
+static double b[17] = {
+  -0.250000000, -0.375000000, -0.125000000, -0.437500000, -0.062500000,
+  -0.031250000, -0.015625000, -0.007812500, -0.003906250, -0.001953125,
+  -0.000976563, -0.000488281, -0.000244141, -0.000122070, -0.000061035,
+  -0.000030518, -0.000015259
+};
+
+static unsigned int pds_quant_bits[17] = {
+  /* for a number of quantization steps; */
+  /*    3,     5,    7,    9,    15,
+     31,    63,  127,  255,   511,
+     1023,  2047, 4095, 8191, 16383,
+     32767, 65535 
+   */
+  /* below we need : */
+  2, 4, 4, 8, 8,
+  16, 32, 64, 128, 256,
+  512, 1024, 2048, 4096, 8192,
+  16384, 32768
+};
+/* to retain succesfull quant */
+/* This is only a quick and dirty tric to speed up ISO code */
+/* In below quant routine : also rewrote loops to decrement */
+/* Added/changed by Patrick De Smet, Nov. 1999 */
+
+/* PDS TODO: maybe it is faster not to store pds_quant_bits */
+/* but rather store (char) n, and use (1L shift left n) ;   */
+/* is a shift faster than loading unsigned int from array ? */
+
+void
+subband_quantization (unsigned int scalar[2][3][SBLIMIT],
+		      double sb_samples[2][3][SCALE_BLOCK][SBLIMIT],
+		      unsigned int j_scale[3][SBLIMIT],
+		      double j_samps[3][SCALE_BLOCK][SBLIMIT],
+		      unsigned int bit_alloc[2][SBLIMIT],
+		      unsigned int sbband[2][3][SCALE_BLOCK][SBLIMIT],
+		      frame_info * frame)
+{
+  int i, j, k, s, qnt, sig;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  double d;
+  al_table *alloc = frame->alloc;
+
+  for (s = 3; s--;)
+    for (j = SCALE_BLOCK; j--;)
+      for (i = sblimit; i--;)
+	for (k = ((i < jsbound) ? nch : 1); k--;)
+	  if (bit_alloc[k][i]) {
+	    /* scale and quantize FLOATing point sample */
+	    if (nch == 2 && i >= jsbound)	/* use j-stereo samples */
+	      d = j_samps[s][j][i] / multiple[j_scale[s][i]];
+	    else
+	      d = sb_samples[k][s][j][i] / multiple[scalar[k][s][i]];
+	    if (fabs(d) > 1.0)
+	      fprintf (stderr, "Not scaled properly %d %d %d %d\n", k, s, j,
+		       i);
+	    qnt = (*alloc)[i][bit_alloc[k][i]].quant;
+	    d = d * a[qnt] + b[qnt];
+	    /* extract MSB N-1 bits from the FLOATing point sample */
+	    if (d >= 0)
+	      sig = 1;
+	    else {
+	      sig = 0;
+	      d += 1.0;
+	    }
+	    sbband[k][s][j][i] =
+	      (unsigned int) (d * (double) (pds_quant_bits[qnt]));
+	    /* tag the inverted sign bit to sbband at position N */
+	    /* The bit inversion is a must for grouping with 3,5,9 steps
+	       so it is done for all subbands */
+	    if (sig)
+	      sbband[k][s][j][i] |= (pds_quant_bits[qnt]);
+	  }
+  for (s = 3; s--;)
+    for (j = sblimit; j < SBLIMIT; j++)
+      for (i = SCALE_BLOCK; i--;)
+	for (k = nch; k--;)
+	  sbband[k][s][i][j] = 0;
+}
+#else
+
+static double a[17] = {
+  0.750000000, 0.625000000, 0.875000000, 0.562500000, 0.937500000,
+  0.968750000, 0.984375000, 0.992187500, 0.996093750, 0.998046875,
+  0.999023438, 0.999511719, 0.999755859, 0.999877930, 0.999938965,
+  0.999969482, 0.999984741
+};
+
+static double b[17] = {
+  -0.250000000, -0.375000000, -0.125000000, -0.437500000, -0.062500000,
+  -0.031250000, -0.015625000, -0.007812500, -0.003906250, -0.001953125,
+  -0.000976563, -0.000488281, -0.000244141, -0.000122070, -0.000061035,
+  -0.000030518, -0.000015259
+};
+
+void
+subband_quantization (unsigned int scalar[2][3][SBLIMIT],
+		      double sb_samples[2][3][SCALE_BLOCK][SBLIMIT],
+		      unsigned int j_scale[3][SBLIMIT],
+		      double j_samps[3][SCALE_BLOCK][SBLIMIT],
+		      unsigned int bit_alloc[2][SBLIMIT],
+		      unsigned int sbband[2][3][SCALE_BLOCK][SBLIMIT],
+		      frame_info * frame)
+{
+  int i, j, k, s, n, qnt, sig;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  unsigned int stps;
+  double d;
+  al_table *alloc = frame->alloc;
+
+  for (s = 0; s < 3; s++)
+    for (j = 0; j < SCALE_BLOCK; j++)
+      for (i = 0; i < sblimit; i++)
+	for (k = 0; k < ((i < jsbound) ? nch : 1); k++)
+	  if (bit_alloc[k][i]) {
+	    /* scale and quantize FLOATing point sample */
+	    if (nch == 2 && i >= jsbound)	/* use j-stereo samples */
+	      d = j_samps[s][j][i] / multiple[j_scale[s][i]];
+	    else
+	      d = sb_samples[k][s][j][i] / multiple[scalar[k][s][i]];
+	    if (mod (d) > 1.0)
+	      fprintf (stderr, "Not scaled properly %d %d %d %d\n", k, s, j,
+		       i);
+	    qnt = (*alloc)[i][bit_alloc[k][i]].quant;
+	    d = d * a[qnt] + b[qnt];
+	    /* extract MSB N-1 bits from the FLOATing point sample */
+	    if (d >= 0)
+	      sig = 1;
+	    else {
+	      sig = 0;
+	      d += 1.0;
+	    }
+	    n = 0;
+	    stps = (*alloc)[i][bit_alloc[k][i]].steps;
+	    while ((1L << n) < stps)
+	      n++;
+	    n--;
+	    sbband[k][s][j][i] = (unsigned int) (d * (double) (1L << n));
+	    /* tag the inverted sign bit to sbband at position N */
+	    /* The bit inversion is a must for grouping with 3,5,9 steps
+	       so it is done for all subbands */
+	    if (sig)
+	      sbband[k][s][j][i] |= 1 << n;
+	  }
+
+  for (k = 0; k < nch; k++)
+    for (s = 0; s < 3; s++)
+      for (i = 0; i < SCALE_BLOCK; i++)
+	for (j = sblimit; j < SBLIMIT; j++)
+	  sbband[k][s][i][j] = 0;
+}
+#endif
+
+/*************************************************************************
+* encode_bit_alloc (Layer II)
+*
+* PURPOSE:Writes bit allocation information onto bitstream
+*
+* Layer II uses 4,3,2, or 0 bits depending on the
+* quantization table used.
+*
+************************************************************************/
+
+void encode_bit_alloc (unsigned int bit_alloc[2][SBLIMIT],
+		       frame_info * frame, Bit_stream_struc * bs)
+{
+  int i, k;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+
+  for (i = 0; i < sblimit; i++)
+    for (k = 0; k < ((i < jsbound) ? nch : 1); k++)
+      putbits (bs, bit_alloc[k][i], (*alloc)[i][0].bits);
+}
+
+/************************************************************************
+*
+* sample_encoding  (Layer II)
+*
+* PURPOSE:Put one frame of subband samples on to the bitstream
+*
+* SEMANTICS:  The number of bits allocated per sample is read from
+* the bit allocation information #bit_alloc#.  Layer 2
+* supports writing grouped samples for quantization steps
+* that are not a power of 2.
+*
+************************************************************************/
+
+void sample_encoding (unsigned int sbband[2][3][SCALE_BLOCK][SBLIMIT],
+		      unsigned int bit_alloc[2][SBLIMIT],
+		      frame_info * frame, Bit_stream_struc * bs)
+{
+  unsigned int temp;
+  unsigned int i, j, k, s, x, y;
+  int nch = frame->nch;
+  int sblimit = frame->sblimit;
+  int jsbound = frame->jsbound;
+  al_table *alloc = frame->alloc;
+
+  for (s = 0; s < 3; s++)
+    for (j = 0; j < SCALE_BLOCK; j += 3)
+      for (i = 0; i < sblimit; i++)
+	for (k = 0; k < ((i < jsbound) ? nch : 1); k++)
+	  if (bit_alloc[k][i]) {
+	    if ((*alloc)[i][bit_alloc[k][i]].group == 3) {
+	      for (x = 0; x < 3; x++)
+		putbits (bs, sbband[k][s][j + x][i],
+			 (*alloc)[i][bit_alloc[k][i]].bits);
+	    } else {
+	      y = (*alloc)[i][bit_alloc[k][i]].steps;
+	      temp =
+		sbband[k][s][j][i] + sbband[k][s][j + 1][i] * y +
+		sbband[k][s][j + 2][i] * y * y;
+	      putbits (bs, temp, (*alloc)[i][bit_alloc[k][i]].bits);
+	    }
+	  }
+}
+
+/************************************************************************
+*
+* encode_CRC
+*
+************************************************************************/
+
+void encode_CRC (unsigned int crc, Bit_stream_struc * bs)
+{
+  putbits (bs, crc, 16);
+}