* The simplest AC3 encoder
* Copyright (c) 2000 Fabrice Bellard.
*
- * This library is free software; you can redistribute it and/or
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
- * version 2 of the License, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
*
- * This library is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
*/
//#define DEBUG
//#define DEBUG_BITALLOC
+#include "libavutil/crc.h"
#include "avcodec.h"
-
+#include "bitstream.h"
#include "ac3.h"
typedef struct AC3EncodeContext {
int lfe_channel;
int bit_rate;
unsigned int sample_rate;
- unsigned int bsid;
+ unsigned int bitstream_id;
unsigned int frame_size_min; /* minimum frame size in case rounding is necessary */
unsigned int frame_size; /* current frame size in words */
- int halfratecod;
- unsigned int frmsizecod;
- unsigned int fscod; /* frequency */
- unsigned int acmod;
+ unsigned int bits_written;
+ unsigned int samples_written;
+ int sr_shift;
+ unsigned int frame_size_code;
+ unsigned int sr_code; /* frequency */
+ unsigned int channel_mode;
int lfe;
- unsigned int bsmod;
+ unsigned int bitstream_mode;
short last_samples[AC3_MAX_CHANNELS][256];
unsigned int chbwcod[AC3_MAX_CHANNELS];
int nb_coefs[AC3_MAX_CHANNELS];
-
+
/* bitrate allocation control */
- int sgaincod, sdecaycod, fdecaycod, dbkneecod, floorcod;
+ int slow_gain_code, slow_decay_code, fast_decay_code, db_per_bit_code, floor_code;
AC3BitAllocParameters bit_alloc;
- int csnroffst;
- int fgaincod[AC3_MAX_CHANNELS];
- int fsnroffst[AC3_MAX_CHANNELS];
+ int coarse_snr_offset;
+ int fast_gain_code[AC3_MAX_CHANNELS];
+ int fine_snr_offset[AC3_MAX_CHANNELS];
/* mantissa encoding */
int mant1_cnt, mant2_cnt, mant4_cnt;
} AC3EncodeContext;
-#include "ac3tab.h"
+static int16_t costab[64];
+static int16_t sintab[64];
+static int16_t xcos1[128];
+static int16_t xsin1[128];
#define MDCT_NBITS 9
#define N (1 << MDCT_NBITS)
/* new exponents are sent if their Norm 1 exceed this number */
#define EXP_DIFF_THRESHOLD 1000
-static void fft_init(int ln);
-static void ac3_crc_init(void);
-
static inline int16_t fix15(float a)
{
int v;
v = (int)(a * (float)(1 << 15));
if (v < -32767)
v = -32767;
- else if (v > 32767)
+ else if (v > 32767)
v = 32767;
return v;
}
-static inline int calc_lowcomp1(int a, int b0, int b1)
-{
- if ((b0 + 256) == b1) {
- a = 384 ;
- } else if (b0 > b1) {
- a = a - 64;
- if (a < 0) a=0;
- }
- return a;
-}
-
-static inline int calc_lowcomp(int a, int b0, int b1, int bin)
-{
- if (bin < 7) {
- if ((b0 + 256) == b1) {
- a = 384 ;
- } else if (b0 > b1) {
- a = a - 64;
- if (a < 0) a=0;
- }
- } else if (bin < 20) {
- if ((b0 + 256) == b1) {
- a = 320 ;
- } else if (b0 > b1) {
- a= a - 64;
- if (a < 0) a=0;
- }
- } else {
- a = a - 128;
- if (a < 0) a=0;
- }
- return a;
-}
-
-/* AC3 bit allocation. The algorithm is the one described in the AC3
- spec. */
-void ac3_parametric_bit_allocation(AC3BitAllocParameters *s, uint8_t *bap,
- int8_t *exp, int start, int end,
- int snroffset, int fgain, int is_lfe,
- int deltbae,int deltnseg,
- uint8_t *deltoffst, uint8_t *deltlen, uint8_t *deltba)
-{
- int bin,i,j,k,end1,v,v1,bndstrt,bndend,lowcomp,begin;
- int fastleak,slowleak,address,tmp;
- int16_t psd[256]; /* scaled exponents */
- int16_t bndpsd[50]; /* interpolated exponents */
- int16_t excite[50]; /* excitation */
- int16_t mask[50]; /* masking value */
-
- /* exponent mapping to PSD */
- for(bin=start;bin<end;bin++) {
- psd[bin]=(3072 - (exp[bin] << 7));
- }
-
- /* PSD integration */
- j=start;
- k=masktab[start];
- do {
- v=psd[j];
- j++;
- end1=bndtab[k+1];
- if (end1 > end) end1=end;
- for(i=j;i<end1;i++) {
- int c,adr;
- /* logadd */
- v1=psd[j];
- c=v-v1;
- if (c >= 0) {
- adr=c >> 1;
- if (adr > 255) adr=255;
- v=v + latab[adr];
- } else {
- adr=(-c) >> 1;
- if (adr > 255) adr=255;
- v=v1 + latab[adr];
- }
- j++;
- }
- bndpsd[k]=v;
- k++;
- } while (end > bndtab[k]);
-
- /* excitation function */
- bndstrt = masktab[start];
- bndend = masktab[end-1] + 1;
-
- if (bndstrt == 0) {
- lowcomp = 0;
- lowcomp = calc_lowcomp1(lowcomp, bndpsd[0], bndpsd[1]) ;
- excite[0] = bndpsd[0] - fgain - lowcomp ;
- lowcomp = calc_lowcomp1(lowcomp, bndpsd[1], bndpsd[2]) ;
- excite[1] = bndpsd[1] - fgain - lowcomp ;
- begin = 7 ;
- for (bin = 2; bin < 7; bin++) {
- if (!(is_lfe && bin == 6))
- lowcomp = calc_lowcomp1(lowcomp, bndpsd[bin], bndpsd[bin+1]) ;
- fastleak = bndpsd[bin] - fgain ;
- slowleak = bndpsd[bin] - s->sgain ;
- excite[bin] = fastleak - lowcomp ;
- if (!(is_lfe && bin == 6)) {
- if (bndpsd[bin] <= bndpsd[bin+1]) {
- begin = bin + 1 ;
- break ;
- }
- }
- }
-
- end1=bndend;
- if (end1 > 22) end1=22;
-
- for (bin = begin; bin < end1; bin++) {
- if (!(is_lfe && bin == 6))
- lowcomp = calc_lowcomp(lowcomp, bndpsd[bin], bndpsd[bin+1], bin) ;
-
- fastleak -= s->fdecay ;
- v = bndpsd[bin] - fgain;
- if (fastleak < v) fastleak = v;
-
- slowleak -= s->sdecay ;
- v = bndpsd[bin] - s->sgain;
- if (slowleak < v) slowleak = v;
-
- v=fastleak - lowcomp;
- if (slowleak > v) v=slowleak;
-
- excite[bin] = v;
- }
- begin = 22;
- } else {
- /* coupling channel */
- begin = bndstrt;
-
- fastleak = (s->cplfleak << 8) + 768;
- slowleak = (s->cplsleak << 8) + 768;
- }
-
- for (bin = begin; bin < bndend; bin++) {
- fastleak -= s->fdecay ;
- v = bndpsd[bin] - fgain;
- if (fastleak < v) fastleak = v;
- slowleak -= s->sdecay ;
- v = bndpsd[bin] - s->sgain;
- if (slowleak < v) slowleak = v;
-
- v=fastleak;
- if (slowleak > v) v = slowleak;
- excite[bin] = v;
- }
-
- /* compute masking curve */
-
- for (bin = bndstrt; bin < bndend; bin++) {
- v1 = excite[bin];
- tmp = s->dbknee - bndpsd[bin];
- if (tmp > 0) {
- v1 += tmp >> 2;
- }
- v=hth[bin >> s->halfratecod][s->fscod];
- if (v1 > v) v=v1;
- mask[bin] = v;
- }
-
- /* delta bit allocation */
-
- if (deltbae == 0 || deltbae == 1) {
- int band, seg, delta;
- band = 0 ;
- for (seg = 0; seg < deltnseg; seg++) {
- band += deltoffst[seg] ;
- if (deltba[seg] >= 4) {
- delta = (deltba[seg] - 3) << 7;
- } else {
- delta = (deltba[seg] - 4) << 7;
- }
- for (k = 0; k < deltlen[seg]; k++) {
- mask[band] += delta ;
- band++ ;
- }
- }
- }
-
- /* compute bit allocation */
-
- i = start ;
- j = masktab[start] ;
- do {
- v=mask[j];
- v -= snroffset ;
- v -= s->floor ;
- if (v < 0) v = 0;
- v &= 0x1fe0 ;
- v += s->floor ;
-
- end1=bndtab[j] + bndsz[j];
- if (end1 > end) end1=end;
-
- for (k = i; k < end1; k++) {
- address = (psd[i] - v) >> 5 ;
- if (address < 0) address=0;
- else if (address > 63) address=63;
- bap[i] = baptab[address];
- i++;
- }
- } while (end > bndtab[j++]) ;
-}
-
typedef struct IComplex {
short re,im;
} IComplex;
static void fft_init(int ln)
{
- int i, j, m, n;
+ int i, n;
float alpha;
n = 1 << ln;
costab[i] = fix15(cos(alpha));
sintab[i] = fix15(sin(alpha));
}
-
- for(i=0;i<n;i++) {
- m=0;
- for(j=0;j<ln;j++) {
- m |= ((i >> j) & 1) << (ln-j-1);
- }
- fft_rev[i]=m;
- }
}
/* butter fly op */
/* do a 2^n point complex fft on 2^ln points. */
static void fft(IComplex *z, int ln)
{
- int j, l, np, np2;
- int nblocks, nloops;
+ int j, l, np, np2;
+ int nblocks, nloops;
register IComplex *p,*q;
int tmp_re, tmp_im;
/* reverse */
for(j=0;j<np;j++) {
- int k;
- IComplex tmp;
- k = fft_rev[j];
- if (k < j) {
- tmp = z[k];
- z[k] = z[j];
- z[j] = tmp;
- }
+ int k = ff_reverse[j] >> (8 - ln);
+ if (k < j)
+ FFSWAP(IComplex, z[k], z[j]);
}
/* pass 0 */
p=&z[0];
j=(np >> 1);
do {
- BF(p[0].re, p[0].im, p[1].re, p[1].im,
+ BF(p[0].re, p[0].im, p[1].re, p[1].im,
p[0].re, p[0].im, p[1].re, p[1].im);
p+=2;
} while (--j != 0);
p=&z[0];
j=np >> 2;
do {
- BF(p[0].re, p[0].im, p[2].re, p[2].im,
+ BF(p[0].re, p[0].im, p[2].re, p[2].im,
p[0].re, p[0].im, p[2].re, p[2].im);
- BF(p[1].re, p[1].im, p[3].re, p[3].im,
+ BF(p[1].re, p[1].im, p[3].re, p[3].im,
p[1].re, p[1].im, p[3].im, -p[3].re);
p+=4;
} while (--j != 0);
BF(p->re, p->im, q->re, q->im,
p->re, p->im, q->re, q->im);
-
+
p++;
q++;
for(l = nblocks; l < np2; l += nblocks) {
static void mdct512(int32_t *out, int16_t *in)
{
int i, re, im, re1, im1;
- int16_t rot[N];
+ int16_t rot[N];
IComplex x[N/4];
/* shift to simplify computations */
rot[i] = -in[i + 3*N/4];
for(i=N/4;i<N;i++)
rot[i] = in[i - N/4];
-
+
/* pre rotation */
for(i=0;i<N/4;i++) {
re = ((int)rot[2*i] - (int)rot[N-1-2*i]) >> 1;
}
fft(x, MDCT_NBITS - 2);
-
+
/* post rotation */
for(i=0;i<N/4;i++) {
re = x[i].re;
{
int i, j;
int exp_diff;
-
+
/* estimate if the exponent variation & decide if they should be
reused in the next frame */
exp_strategy[0][ch] = EXP_NEW;
for(i=1;i<NB_BLOCKS;i++) {
exp_diff = calc_exp_diff(exp[i][ch], exp[i-1][ch], N/2);
-#ifdef DEBUG
+#ifdef DEBUG
av_log(NULL, AV_LOG_DEBUG, "exp_diff=%d\n", exp_diff);
#endif
if (exp_diff > EXP_DIFF_THRESHOLD)
exp_strategy[i][ch] = EXP_REUSE;
}
if (is_lfe)
- return;
+ return;
/* now select the encoding strategy type : if exponents are often
recoded, we use a coarse encoding */
exp_strategy[i][ch] = EXP_D15;
break;
}
- i = j;
+ i = j;
}
}
exp[i] = exp1[i];
}
}
-
+
/* update the exponents so that they are the ones the decoder will
decode. Return the number of bits used to code the exponents */
-static int encode_exp(uint8_t encoded_exp[N/2],
- uint8_t exp[N/2],
+static int encode_exp(uint8_t encoded_exp[N/2],
+ uint8_t exp[N/2],
int nb_exps,
int exp_strategy)
{
- int group_size, nb_groups, i, j, k, recurse, exp_min, delta;
+ int group_size, nb_groups, i, j, k, exp_min;
uint8_t exp1[N/2];
switch(exp_strategy) {
if (exp1[0] > 15)
exp1[0] = 15;
- /* Iterate until the delta constraints between each groups are
- satisfyed. I'm sure it is possible to find a better algorithm,
- but I am lazy */
- do {
- recurse = 0;
- for(i=1;i<=nb_groups;i++) {
- delta = exp1[i] - exp1[i-1];
- if (delta > 2) {
- /* if delta too big, we encode a smaller exponent */
- exp1[i] = exp1[i-1] + 2;
- } else if (delta < -2) {
- /* if delta is too small, we must decrease the previous
- exponent, which means we must recurse */
- recurse = 1;
- exp1[i-1] = exp1[i] + 2;
- }
- }
- } while (recurse);
-
+ /* Decrease the delta between each groups to within 2
+ * so that they can be differentially encoded */
+ for (i=1;i<=nb_groups;i++)
+ exp1[i] = FFMIN(exp1[i], exp1[i-1] + 2);
+ for (i=nb_groups-1;i>=0;i--)
+ exp1[i] = FFMIN(exp1[i], exp1[i+1] + 2);
+
/* now we have the exponent values the decoder will see */
encoded_exp[0] = exp1[0];
k = 1;
}
k += group_size;
}
-
+
#if defined(DEBUG)
av_log(NULL, AV_LOG_DEBUG, "exponents: strategy=%d\n", exp_strategy);
for(i=0;i<=nb_groups * group_size;i++) {
break;
case 1:
/* 3 mantissa in 5 bits */
- if (s->mant1_cnt == 0)
+ if (s->mant1_cnt == 0)
bits += 5;
if (++s->mant1_cnt == 3)
s->mant1_cnt = 0;
break;
case 2:
/* 3 mantissa in 7 bits */
- if (s->mant2_cnt == 0)
+ if (s->mant2_cnt == 0)
bits += 7;
if (++s->mant2_cnt == 3)
s->mant2_cnt = 0;
/* 2 mantissa in 7 bits */
if (s->mant4_cnt == 0)
bits += 7;
- if (++s->mant4_cnt == 2)
+ if (++s->mant4_cnt == 2)
s->mant4_cnt = 0;
break;
case 14:
}
+static void bit_alloc_masking(AC3EncodeContext *s,
+ uint8_t encoded_exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+ uint8_t exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS],
+ int16_t psd[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
+ int16_t mask[NB_BLOCKS][AC3_MAX_CHANNELS][50])
+{
+ int blk, ch;
+ int16_t band_psd[NB_BLOCKS][AC3_MAX_CHANNELS][50];
+
+ for(blk=0; blk<NB_BLOCKS; blk++) {
+ for(ch=0;ch<s->nb_all_channels;ch++) {
+ if(exp_strategy[blk][ch] == EXP_REUSE) {
+ memcpy(psd[blk][ch], psd[blk-1][ch], (N/2)*sizeof(int16_t));
+ memcpy(mask[blk][ch], mask[blk-1][ch], 50*sizeof(int16_t));
+ } else {
+ ff_ac3_bit_alloc_calc_psd(encoded_exp[blk][ch], 0,
+ s->nb_coefs[ch],
+ psd[blk][ch], band_psd[blk][ch]);
+ ff_ac3_bit_alloc_calc_mask(&s->bit_alloc, band_psd[blk][ch],
+ 0, s->nb_coefs[ch],
+ ff_ac3_fast_gain_tab[s->fast_gain_code[ch]],
+ ch == s->lfe_channel,
+ DBA_NONE, 0, NULL, NULL, NULL,
+ mask[blk][ch]);
+ }
+ }
+ }
+}
+
static int bit_alloc(AC3EncodeContext *s,
+ int16_t mask[NB_BLOCKS][AC3_MAX_CHANNELS][50],
+ int16_t psd[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
uint8_t bap[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
- uint8_t encoded_exp[NB_BLOCKS][AC3_MAX_CHANNELS][N/2],
- uint8_t exp_strategy[NB_BLOCKS][AC3_MAX_CHANNELS],
- int frame_bits, int csnroffst, int fsnroffst)
+ int frame_bits, int coarse_snr_offset, int fine_snr_offset)
{
int i, ch;
+ int snr_offset;
+
+ snr_offset = (((coarse_snr_offset - 15) << 4) + fine_snr_offset) << 2;
/* compute size */
for(i=0;i<NB_BLOCKS;i++) {
s->mant2_cnt = 0;
s->mant4_cnt = 0;
for(ch=0;ch<s->nb_all_channels;ch++) {
- ac3_parametric_bit_allocation(&s->bit_alloc,
- bap[i][ch], (int8_t *)encoded_exp[i][ch],
- 0, s->nb_coefs[ch],
- (((csnroffst-15) << 4) +
- fsnroffst) << 2,
- fgaintab[s->fgaincod[ch]],
- ch == s->lfe_channel,
- 2, 0, NULL, NULL, NULL);
- frame_bits += compute_mantissa_size(s, bap[i][ch],
+ ff_ac3_bit_alloc_calc_bap(mask[i][ch], psd[i][ch], 0,
+ s->nb_coefs[ch], snr_offset,
+ s->bit_alloc.floor, ff_ac3_bap_tab,
+ bap[i][ch]);
+ frame_bits += compute_mantissa_size(s, bap[i][ch],
s->nb_coefs[ch]);
}
}
#if 0
- printf("csnr=%d fsnr=%d frame_bits=%d diff=%d\n",
- csnroffst, fsnroffst, frame_bits,
+ printf("csnr=%d fsnr=%d frame_bits=%d diff=%d\n",
+ coarse_snr_offset, fine_snr_offset, frame_bits,
16 * s->frame_size - ((frame_bits + 7) & ~7));
#endif
return 16 * s->frame_size - frame_bits;
int frame_bits)
{
int i, ch;
- int csnroffst, fsnroffst;
+ int coarse_snr_offset, fine_snr_offset;
uint8_t bap1[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+ int16_t psd[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
+ int16_t mask[NB_BLOCKS][AC3_MAX_CHANNELS][50];
static int frame_bits_inc[8] = { 0, 0, 2, 2, 2, 4, 2, 4 };
/* init default parameters */
- s->sdecaycod = 2;
- s->fdecaycod = 1;
- s->sgaincod = 1;
- s->dbkneecod = 2;
- s->floorcod = 4;
- for(ch=0;ch<s->nb_all_channels;ch++)
- s->fgaincod[ch] = 4;
-
+ s->slow_decay_code = 2;
+ s->fast_decay_code = 1;
+ s->slow_gain_code = 1;
+ s->db_per_bit_code = 2;
+ s->floor_code = 4;
+ for(ch=0;ch<s->nb_all_channels;ch++)
+ s->fast_gain_code[ch] = 4;
+
/* compute real values */
- s->bit_alloc.fscod = s->fscod;
- s->bit_alloc.halfratecod = s->halfratecod;
- s->bit_alloc.sdecay = sdecaytab[s->sdecaycod] >> s->halfratecod;
- s->bit_alloc.fdecay = fdecaytab[s->fdecaycod] >> s->halfratecod;
- s->bit_alloc.sgain = sgaintab[s->sgaincod];
- s->bit_alloc.dbknee = dbkneetab[s->dbkneecod];
- s->bit_alloc.floor = floortab[s->floorcod];
-
+ s->bit_alloc.sr_code = s->sr_code;
+ s->bit_alloc.sr_shift = s->sr_shift;
+ s->bit_alloc.slow_decay = ff_ac3_slow_decay_tab[s->slow_decay_code] >> s->sr_shift;
+ s->bit_alloc.fast_decay = ff_ac3_fast_decay_tab[s->fast_decay_code] >> s->sr_shift;
+ s->bit_alloc.slow_gain = ff_ac3_slow_gain_tab[s->slow_gain_code];
+ s->bit_alloc.db_per_bit = ff_ac3_db_per_bit_tab[s->db_per_bit_code];
+ s->bit_alloc.floor = ff_ac3_floor_tab[s->floor_code];
+
/* header size */
frame_bits += 65;
- // if (s->acmod == 2)
+ // if (s->channel_mode == 2)
// frame_bits += 2;
- frame_bits += frame_bits_inc[s->acmod];
+ frame_bits += frame_bits_inc[s->channel_mode];
/* audio blocks */
for(i=0;i<NB_BLOCKS;i++) {
frame_bits += s->nb_channels * 2 + 2; /* blksw * c, dithflag * c, dynrnge, cplstre */
- if (s->acmod == 2)
+ if (s->channel_mode == AC3_CHMODE_STEREO) {
frame_bits++; /* rematstr */
+ if(i==0) frame_bits += 4;
+ }
frame_bits += 2 * s->nb_channels; /* chexpstr[2] * c */
- if (s->lfe)
- frame_bits++; /* lfeexpstr */
+ if (s->lfe)
+ frame_bits++; /* lfeexpstr */
for(ch=0;ch<s->nb_channels;ch++) {
if (exp_strategy[i][ch] != EXP_REUSE)
frame_bits += 6 + 2; /* chbwcod[6], gainrng[2] */
/* (fsnoffset[4] + fgaincod[4]) * c */
frame_bits += 2*4 + 3 + 6 + s->nb_all_channels * (4 + 3);
+ /* auxdatae, crcrsv */
+ frame_bits += 2;
+
/* CRC */
frame_bits += 16;
+ /* calculate psd and masking curve before doing bit allocation */
+ bit_alloc_masking(s, encoded_exp, exp_strategy, psd, mask);
+
/* now the big work begins : do the bit allocation. Modify the snr
offset until we can pack everything in the requested frame size */
- csnroffst = s->csnroffst;
- while (csnroffst >= 0 &&
- bit_alloc(s, bap, encoded_exp, exp_strategy, frame_bits, csnroffst, 0) < 0)
- csnroffst -= SNR_INC1;
- if (csnroffst < 0) {
- av_log(NULL, AV_LOG_ERROR, "Yack, Error !!!\n");
- return -1;
+ coarse_snr_offset = s->coarse_snr_offset;
+ while (coarse_snr_offset >= 0 &&
+ bit_alloc(s, mask, psd, bap, frame_bits, coarse_snr_offset, 0) < 0)
+ coarse_snr_offset -= SNR_INC1;
+ if (coarse_snr_offset < 0) {
+ av_log(NULL, AV_LOG_ERROR, "Bit allocation failed. Try increasing the bitrate.\n");
+ return -1;
}
- while ((csnroffst + SNR_INC1) <= 63 &&
- bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits,
- csnroffst + SNR_INC1, 0) >= 0) {
- csnroffst += SNR_INC1;
+ while ((coarse_snr_offset + SNR_INC1) <= 63 &&
+ bit_alloc(s, mask, psd, bap1, frame_bits,
+ coarse_snr_offset + SNR_INC1, 0) >= 0) {
+ coarse_snr_offset += SNR_INC1;
memcpy(bap, bap1, sizeof(bap1));
}
- while ((csnroffst + 1) <= 63 &&
- bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits, csnroffst + 1, 0) >= 0) {
- csnroffst++;
+ while ((coarse_snr_offset + 1) <= 63 &&
+ bit_alloc(s, mask, psd, bap1, frame_bits, coarse_snr_offset + 1, 0) >= 0) {
+ coarse_snr_offset++;
memcpy(bap, bap1, sizeof(bap1));
}
- fsnroffst = 0;
- while ((fsnroffst + SNR_INC1) <= 15 &&
- bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits,
- csnroffst, fsnroffst + SNR_INC1) >= 0) {
- fsnroffst += SNR_INC1;
+ fine_snr_offset = 0;
+ while ((fine_snr_offset + SNR_INC1) <= 15 &&
+ bit_alloc(s, mask, psd, bap1, frame_bits,
+ coarse_snr_offset, fine_snr_offset + SNR_INC1) >= 0) {
+ fine_snr_offset += SNR_INC1;
memcpy(bap, bap1, sizeof(bap1));
}
- while ((fsnroffst + 1) <= 15 &&
- bit_alloc(s, bap1, encoded_exp, exp_strategy, frame_bits,
- csnroffst, fsnroffst + 1) >= 0) {
- fsnroffst++;
+ while ((fine_snr_offset + 1) <= 15 &&
+ bit_alloc(s, mask, psd, bap1, frame_bits,
+ coarse_snr_offset, fine_snr_offset + 1) >= 0) {
+ fine_snr_offset++;
memcpy(bap, bap1, sizeof(bap1));
}
-
- s->csnroffst = csnroffst;
+
+ s->coarse_snr_offset = coarse_snr_offset;
for(ch=0;ch<s->nb_all_channels;ch++)
- s->fsnroffst[ch] = fsnroffst;
+ s->fine_snr_offset[ch] = fine_snr_offset;
#if defined(DEBUG_BITALLOC)
{
int j;
return 0;
}
-void ac3_common_init(void)
-{
- int i, j, k, l, v;
- /* compute bndtab and masktab from bandsz */
- k = 0;
- l = 0;
- for(i=0;i<50;i++) {
- bndtab[i] = l;
- v = bndsz[i];
- for(j=0;j<v;j++) masktab[k++]=i;
- l += v;
- }
- bndtab[50] = 0;
-}
-
-
-static int AC3_encode_init(AVCodecContext *avctx)
+static av_cold int AC3_encode_init(AVCodecContext *avctx)
{
int freq = avctx->sample_rate;
int bitrate = avctx->bit_rate;
AC3EncodeContext *s = avctx->priv_data;
int i, j, ch;
float alpha;
- static const uint8_t acmod_defs[6] = {
- 0x01, /* C */
- 0x02, /* L R */
- 0x03, /* L C R */
- 0x06, /* L R SL SR */
- 0x07, /* L C R SL SR */
- 0x07, /* L C R SL SR (+LFE) */
+ int bw_code;
+ static const uint8_t channel_mode_defs[6] = {
+ 0x01, /* C */
+ 0x02, /* L R */
+ 0x03, /* L C R */
+ 0x06, /* L R SL SR */
+ 0x07, /* L C R SL SR */
+ 0x07, /* L C R SL SR (+LFE) */
};
avctx->frame_size = AC3_FRAME_SIZE;
-
+
+ ac3_common_init();
+
/* number of channels */
if (channels < 1 || channels > 6)
- return -1;
- s->acmod = acmod_defs[channels - 1];
+ return -1;
+ s->channel_mode = channel_mode_defs[channels - 1];
s->lfe = (channels == 6) ? 1 : 0;
s->nb_all_channels = channels;
s->nb_channels = channels > 5 ? 5 : channels;
/* frequency */
for(i=0;i<3;i++) {
- for(j=0;j<3;j++)
- if ((ac3_freqs[j] >> i) == freq)
+ for(j=0;j<3;j++)
+ if ((ff_ac3_sample_rate_tab[j] >> i) == freq)
goto found;
}
return -1;
- found:
+ found:
s->sample_rate = freq;
- s->halfratecod = i;
- s->fscod = j;
- s->bsid = 8 + s->halfratecod;
- s->bsmod = 0; /* complete main audio service */
+ s->sr_shift = i;
+ s->sr_code = j;
+ s->bitstream_id = 8 + s->sr_shift;
+ s->bitstream_mode = 0; /* complete main audio service */
/* bitrate & frame size */
- bitrate /= 1000;
for(i=0;i<19;i++) {
- if ((ac3_bitratetab[i] >> s->halfratecod) == bitrate)
+ if ((ff_ac3_bitrate_tab[i] >> s->sr_shift)*1000 == bitrate)
break;
}
if (i == 19)
return -1;
s->bit_rate = bitrate;
- s->frmsizecod = i << 1;
- s->frame_size_min = (bitrate * 1000 * AC3_FRAME_SIZE) / (freq * 16);
- /* for now we do not handle fractional sizes */
+ s->frame_size_code = i << 1;
+ s->frame_size_min = ff_ac3_frame_size_tab[s->frame_size_code][s->sr_code];
+ s->bits_written = 0;
+ s->samples_written = 0;
s->frame_size = s->frame_size_min;
-
+
/* bit allocation init */
+ if(avctx->cutoff) {
+ /* calculate bandwidth based on user-specified cutoff frequency */
+ int cutoff = av_clip(avctx->cutoff, 1, s->sample_rate >> 1);
+ int fbw_coeffs = cutoff * 512 / s->sample_rate;
+ bw_code = av_clip((fbw_coeffs - 73) / 3, 0, 60);
+ } else {
+ /* use default bandwidth setting */
+ /* XXX: should compute the bandwidth according to the frame
+ size, so that we avoid annoying high frequency artifacts */
+ bw_code = 50;
+ }
for(ch=0;ch<s->nb_channels;ch++) {
/* bandwidth for each channel */
- /* XXX: should compute the bandwidth according to the frame
- size, so that we avoid anoying high freq artefacts */
- s->chbwcod[ch] = 50; /* sample bandwidth as mpeg audio layer 2 table 0 */
- s->nb_coefs[ch] = ((s->chbwcod[ch] + 12) * 3) + 37;
+ s->chbwcod[ch] = bw_code;
+ s->nb_coefs[ch] = bw_code * 3 + 73;
}
if (s->lfe) {
- s->nb_coefs[s->lfe_channel] = 7; /* fixed */
+ s->nb_coefs[s->lfe_channel] = 7; /* fixed */
}
/* initial snr offset */
- s->csnroffst = 40;
-
- ac3_common_init();
+ s->coarse_snr_offset = 40;
/* mdct init */
fft_init(MDCT_NBITS - 2);
xsin1[i] = fix15(-sin(alpha));
}
- ac3_crc_init();
-
avctx->coded_frame= avcodec_alloc_frame();
avctx->coded_frame->key_frame= 1;
put_bits(&s->pb, 16, 0x0b77); /* frame header */
put_bits(&s->pb, 16, 0); /* crc1: will be filled later */
- put_bits(&s->pb, 2, s->fscod);
- put_bits(&s->pb, 6, s->frmsizecod + (s->frame_size - s->frame_size_min));
- put_bits(&s->pb, 5, s->bsid);
- put_bits(&s->pb, 3, s->bsmod);
- put_bits(&s->pb, 3, s->acmod);
- if ((s->acmod & 0x01) && s->acmod != 0x01)
- put_bits(&s->pb, 2, 1); /* XXX -4.5 dB */
- if (s->acmod & 0x04)
- put_bits(&s->pb, 2, 1); /* XXX -6 dB */
- if (s->acmod == 0x02)
+ put_bits(&s->pb, 2, s->sr_code);
+ put_bits(&s->pb, 6, s->frame_size_code + (s->frame_size - s->frame_size_min));
+ put_bits(&s->pb, 5, s->bitstream_id);
+ put_bits(&s->pb, 3, s->bitstream_mode);
+ put_bits(&s->pb, 3, s->channel_mode);
+ if ((s->channel_mode & 0x01) && s->channel_mode != AC3_CHMODE_MONO)
+ put_bits(&s->pb, 2, 1); /* XXX -4.5 dB */
+ if (s->channel_mode & 0x04)
+ put_bits(&s->pb, 2, 1); /* XXX -6 dB */
+ if (s->channel_mode == AC3_CHMODE_STEREO)
put_bits(&s->pb, 2, 0); /* surround not indicated */
put_bits(&s->pb, 1, s->lfe); /* LFE */
put_bits(&s->pb, 5, 31); /* dialog norm: -31 db */
put_bits(&s->pb, 1, 1); /* original bitstream */
put_bits(&s->pb, 1, 0); /* no time code 1 */
put_bits(&s->pb, 1, 0); /* no time code 2 */
- put_bits(&s->pb, 1, 0); /* no addtional bit stream info */
+ put_bits(&s->pb, 1, 0); /* no additional bit stream info */
}
/* symetric quantization on 'levels' levels */
uint16_t *qmant1_ptr, *qmant2_ptr, *qmant4_ptr;
int delta0, delta1, delta2;
- for(ch=0;ch<s->nb_channels;ch++)
+ for(ch=0;ch<s->nb_channels;ch++)
put_bits(&s->pb, 1, 0); /* 512 point MDCT */
- for(ch=0;ch<s->nb_channels;ch++)
+ for(ch=0;ch<s->nb_channels;ch++)
put_bits(&s->pb, 1, 1); /* no dither */
put_bits(&s->pb, 1, 0); /* no dynamic range */
if (block_num == 0) {
put_bits(&s->pb, 1, 0); /* no new coupling strategy */
}
- if (s->acmod == 2)
+ if (s->channel_mode == AC3_CHMODE_STEREO)
{
- if(block_num==0)
- {
- /* first block must define rematrixing (rematstr) */
- put_bits(&s->pb, 1, 1);
-
- /* dummy rematrixing rematflg(1:4)=0 */
- for (rbnd=0;rbnd<4;rbnd++)
- put_bits(&s->pb, 1, 0);
- }
- else
- {
- /* no matrixing (but should be used in the future) */
- put_bits(&s->pb, 1, 0);
- }
+ if(block_num==0)
+ {
+ /* first block must define rematrixing (rematstr) */
+ put_bits(&s->pb, 1, 1);
+
+ /* dummy rematrixing rematflg(1:4)=0 */
+ for (rbnd=0;rbnd<4;rbnd++)
+ put_bits(&s->pb, 1, 0);
+ }
+ else
+ {
+ /* no matrixing (but should be used in the future) */
+ put_bits(&s->pb, 1, 0);
+ }
}
-#if defined(DEBUG)
+#if defined(DEBUG)
{
static int count = 0;
av_log(NULL, AV_LOG_DEBUG, "Block #%d (%d)\n", block_num, count++);
for(ch=0;ch<s->nb_channels;ch++) {
put_bits(&s->pb, 2, exp_strategy[ch]);
}
-
+
if (s->lfe) {
- put_bits(&s->pb, 1, exp_strategy[s->lfe_channel]);
+ put_bits(&s->pb, 1, exp_strategy[s->lfe_channel]);
}
for(ch=0;ch<s->nb_channels;ch++) {
if (exp_strategy[ch] != EXP_REUSE)
put_bits(&s->pb, 6, s->chbwcod[ch]);
}
-
+
/* exponents */
for (ch = 0; ch < s->nb_all_channels; ch++) {
switch(exp_strategy[ch]) {
group_size = 4;
break;
}
- nb_groups = (s->nb_coefs[ch] + (group_size * 3) - 4) / (3 * group_size);
+ nb_groups = (s->nb_coefs[ch] + (group_size * 3) - 4) / (3 * group_size);
p = encoded_exp[ch];
/* first exponent */
put_bits(&s->pb, 7, ((delta0 * 5 + delta1) * 5) + delta2);
}
- if (ch != s->lfe_channel)
- put_bits(&s->pb, 2, 0); /* no gain range info */
+ if (ch != s->lfe_channel)
+ put_bits(&s->pb, 2, 0); /* no gain range info */
}
/* bit allocation info */
baie = (block_num == 0);
put_bits(&s->pb, 1, baie);
if (baie) {
- put_bits(&s->pb, 2, s->sdecaycod);
- put_bits(&s->pb, 2, s->fdecaycod);
- put_bits(&s->pb, 2, s->sgaincod);
- put_bits(&s->pb, 2, s->dbkneecod);
- put_bits(&s->pb, 3, s->floorcod);
+ put_bits(&s->pb, 2, s->slow_decay_code);
+ put_bits(&s->pb, 2, s->fast_decay_code);
+ put_bits(&s->pb, 2, s->slow_gain_code);
+ put_bits(&s->pb, 2, s->db_per_bit_code);
+ put_bits(&s->pb, 3, s->floor_code);
}
/* snr offset */
put_bits(&s->pb, 1, baie); /* always present with bai */
if (baie) {
- put_bits(&s->pb, 6, s->csnroffst);
+ put_bits(&s->pb, 6, s->coarse_snr_offset);
for(ch=0;ch<s->nb_all_channels;ch++) {
- put_bits(&s->pb, 4, s->fsnroffst[ch]);
- put_bits(&s->pb, 3, s->fgaincod[ch]);
+ put_bits(&s->pb, 4, s->fine_snr_offset[ch]);
+ put_bits(&s->pb, 3, s->fast_gain_code[ch]);
}
}
-
+
put_bits(&s->pb, 1, 0); /* no delta bit allocation */
put_bits(&s->pb, 1, 0); /* no data to skip */
/* second pass : output the values */
for (ch = 0; ch < s->nb_all_channels; ch++) {
int b, q;
-
+
for(i=0;i<s->nb_coefs[ch];i++) {
q = qmant[ch][i];
b = bap[ch][i];
case 0:
break;
case 1:
- if (q != 128)
+ if (q != 128)
put_bits(&s->pb, 5, q);
break;
case 2:
- if (q != 128)
+ if (q != 128)
put_bits(&s->pb, 7, q);
break;
case 3:
}
}
-/* compute the ac3 crc */
-
#define CRC16_POLY ((1 << 0) | (1 << 2) | (1 << 15) | (1 << 16))
-static void ac3_crc_init(void)
-{
- unsigned int c, n, k;
-
- for(n=0;n<256;n++) {
- c = n << 8;
- for (k = 0; k < 8; k++) {
- if (c & (1 << 15))
- c = ((c << 1) & 0xffff) ^ (CRC16_POLY & 0xffff);
- else
- c = c << 1;
- }
- crc_table[n] = c;
- }
-}
-
-static unsigned int ac3_crc(uint8_t *data, int n, unsigned int crc)
-{
- int i;
- for(i=0;i<n;i++) {
- crc = (crc_table[data[i] ^ (crc >> 8)] ^ (crc << 8)) & 0xffff;
- }
- return crc;
-}
-
static unsigned int mul_poly(unsigned int a, unsigned int b, unsigned int poly)
{
unsigned int c;
assert(n >= 0);
if(n>0)
memset(pbBufPtr(&s->pb), 0, n);
-
+
/* Now we must compute both crcs : this is not so easy for crc1
because it is at the beginning of the data... */
frame_size_58 = (frame_size >> 1) + (frame_size >> 3);
- crc1 = ac3_crc(frame + 4, (2 * frame_size_58) - 4, 0);
+ crc1 = bswap_16(av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0,
+ frame + 4, 2 * frame_size_58 - 4));
/* XXX: could precompute crc_inv */
crc_inv = pow_poly((CRC16_POLY >> 1), (16 * frame_size_58) - 16, CRC16_POLY);
crc1 = mul_poly(crc_inv, crc1, CRC16_POLY);
- frame[2] = crc1 >> 8;
- frame[3] = crc1;
-
- crc2 = ac3_crc(frame + 2 * frame_size_58, (frame_size - frame_size_58) * 2 - 2, 0);
- frame[2*frame_size - 2] = crc2 >> 8;
- frame[2*frame_size - 1] = crc2;
+ AV_WB16(frame+2,crc1);
+
+ crc2 = bswap_16(av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0,
+ frame + 2 * frame_size_58,
+ (frame_size - frame_size_58) * 2 - 2));
+ AV_WB16(frame+2*frame_size-2,crc2);
// printf("n=%d frame_size=%d\n", n, frame_size);
return frame_size * 2;
unsigned char *frame, int buf_size, void *data)
{
AC3EncodeContext *s = avctx->priv_data;
- short *samples = data;
+ int16_t *samples = data;
int i, j, k, v, ch;
int16_t input_samples[N];
int32_t mdct_coef[NB_BLOCKS][AC3_MAX_CHANNELS][N/2];
for(j=0;j<N/2;j++) {
v = *sptr;
input_samples[j + N/2] = v;
- s->last_samples[ch][j] = v;
+ s->last_samples[ch][j] = v;
sptr += sinc;
}
/* apply the MDCT window */
for(j=0;j<N/2;j++) {
- input_samples[j] = MUL16(input_samples[j],
- ac3_window[j]) >> 15;
- input_samples[N-j-1] = MUL16(input_samples[N-j-1],
- ac3_window[j]) >> 15;
+ input_samples[j] = MUL16(input_samples[j],
+ ff_ac3_window[j]) >> 15;
+ input_samples[N-j-1] = MUL16(input_samples[N-j-1],
+ ff_ac3_window[j]) >> 15;
}
-
+
/* Normalize the samples to use the maximum available
precision */
v = 14 - log2_tab(input_samples, N);
if (v < 0)
v = 0;
- exp_samples[i][ch] = v - 8;
+ exp_samples[i][ch] = v - 9;
lshift_tab(input_samples, N, v);
/* do the MDCT */
mdct512(mdct_coef[i][ch], input_samples);
-
+
/* compute "exponents". We take into account the
normalization there */
for(j=0;j<N/2;j++) {
exp[i][ch][j] = e;
}
}
-
+
compute_exp_strategy(exp_strategy, exp, ch, ch == s->lfe_channel);
/* compute the exponents as the decoder will see them. The
j++;
}
frame_bits += encode_exp(encoded_exp[i][ch],
- exp[i][ch], s->nb_coefs[ch],
+ exp[i][ch], s->nb_coefs[ch],
exp_strategy[i][ch]);
/* copy encoded exponents for reuse case */
for(k=i+1;k<j;k++) {
- memcpy(encoded_exp[k][ch], encoded_exp[i][ch],
+ memcpy(encoded_exp[k][ch], encoded_exp[i][ch],
s->nb_coefs[ch] * sizeof(uint8_t));
}
i = j;
}
}
+ /* adjust for fractional frame sizes */
+ while(s->bits_written >= s->bit_rate && s->samples_written >= s->sample_rate) {
+ s->bits_written -= s->bit_rate;
+ s->samples_written -= s->sample_rate;
+ }
+ s->frame_size = s->frame_size_min + (s->bits_written * s->sample_rate < s->samples_written * s->bit_rate);
+ s->bits_written += s->frame_size * 16;
+ s->samples_written += AC3_FRAME_SIZE;
+
compute_bit_allocation(s, bap, encoded_exp, exp_strategy, frame_bits);
/* everything is known... let's output the frame */
output_frame_header(s, frame);
-
+
for(i=0;i<NB_BLOCKS;i++) {
- output_audio_block(s, exp_strategy[i], encoded_exp[i],
+ output_audio_block(s, exp_strategy[i], encoded_exp[i],
bap[i], mdct_coef[i], exp_samples[i], i);
}
return output_frame_end(s);
}
-static int AC3_encode_close(AVCodecContext *avctx)
+static av_cold int AC3_encode_close(AVCodecContext *avctx)
{
av_freep(&avctx->coded_frame);
return 0;
/*************************************************************************/
/* TEST */
+#undef random
#define FN (N/4)
void fft_test(void)
sum_re += in1[n].re * cos(a) - in1[n].im * sin(a);
sum_im += in1[n].re * sin(a) + in1[n].im * cos(a);
}
- printf("%3d: %6d,%6d %6.0f,%6.0f\n",
- k, in[k].re, in[k].im, sum_re / FN, sum_im / FN);
+ printf("%3d: %6d,%6d %6.0f,%6.0f\n",
+ k, in[k].re, in[k].im, sum_re / FN, sum_im / FN);
}
}
}
mdct512(output, input);
-
+
/* do it by hand */
for(k=0;k<N/2;k++) {
s = 0;
}
output1[k] = -2 * s / N;
}
-
+
err = 0;
emax = 0;
for(i=0;i<N/2;i++) {
unsigned char frame[AC3_MAX_CODED_FRAME_SIZE];
short samples[AC3_FRAME_SIZE];
int ret, i;
-
+
AC3_encode_init(&ctx, 44100, 64000, 1);
fft_test();
AC3_encode_frame,
AC3_encode_close,
NULL,
+ .long_name = NULL_IF_CONFIG_SMALL("ATSC A/52 / AC-3"),
};