3 * This code was developed as part of Google Summer of Code 2006.
4 * E-AC-3 support was added as part of Google Summer of Code 2007.
6 * Copyright (c) 2006 Kartikey Mahendra BHATT (bhattkm at gmail dot com)
7 * Copyright (c) 2007-2008 Bartlomiej Wolowiec <bartek.wolowiec@gmail.com>
8 * Copyright (c) 2007 Justin Ruggles <justin.ruggles@gmail.com>
10 * This file is part of FFmpeg.
12 * FFmpeg is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU Lesser General Public
14 * License as published by the Free Software Foundation; either
15 * version 2.1 of the License, or (at your option) any later version.
17 * FFmpeg is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20 * Lesser General Public License for more details.
22 * You should have received a copy of the GNU Lesser General Public
23 * License along with FFmpeg; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
32 #include "libavutil/crc.h"
33 #include "libavutil/opt.h"
35 #include "aac_ac3_parser.h"
36 #include "ac3_parser.h"
38 #include "ac3dec_data.h"
42 * table for ungrouping 3 values in 7 bits.
43 * used for exponents and bap=2 mantissas
45 static uint8_t ungroup_3_in_7_bits_tab[128][3];
48 /** tables for ungrouping mantissas */
49 static int b1_mantissas[32][3];
50 static int b2_mantissas[128][3];
51 static int b3_mantissas[8];
52 static int b4_mantissas[128][2];
53 static int b5_mantissas[16];
56 * Quantization table: levels for symmetric. bits for asymmetric.
57 * reference: Table 7.18 Mapping of bap to Quantizer
59 static const uint8_t quantization_tab[16] = {
61 5, 6, 7, 8, 9, 10, 11, 12, 14, 16
64 /** dynamic range table. converts codes to scale factors. */
65 static float dynamic_range_tab[256];
67 /** Adjustments in dB gain */
68 static const float gain_levels[9] = {
72 LEVEL_MINUS_1POINT5DB,
74 LEVEL_MINUS_4POINT5DB,
81 * Table for center mix levels
82 * reference: Section 5.4.2.4 cmixlev
84 static const uint8_t center_levels[4] = { 4, 5, 6, 5 };
87 * Table for surround mix levels
88 * reference: Section 5.4.2.5 surmixlev
90 static const uint8_t surround_levels[4] = { 4, 6, 7, 6 };
93 * Table for default stereo downmixing coefficients
94 * reference: Section 7.8.2 Downmixing Into Two Channels
96 static const uint8_t ac3_default_coeffs[8][5][2] = {
97 { { 2, 7 }, { 7, 2 }, },
99 { { 2, 7 }, { 7, 2 }, },
100 { { 2, 7 }, { 5, 5 }, { 7, 2 }, },
101 { { 2, 7 }, { 7, 2 }, { 6, 6 }, },
102 { { 2, 7 }, { 5, 5 }, { 7, 2 }, { 8, 8 }, },
103 { { 2, 7 }, { 7, 2 }, { 6, 7 }, { 7, 6 }, },
104 { { 2, 7 }, { 5, 5 }, { 7, 2 }, { 6, 7 }, { 7, 6 }, },
108 * Symmetrical Dequantization
109 * reference: Section 7.3.3 Expansion of Mantissas for Symmetrical Quantization
110 * Tables 7.19 to 7.23
113 symmetric_dequant(int code, int levels)
115 return ((code - (levels >> 1)) << 24) / levels;
119 * Initialize tables at runtime.
121 static av_cold void ac3_tables_init(void)
125 /* generate table for ungrouping 3 values in 7 bits
126 reference: Section 7.1.3 Exponent Decoding */
127 for(i=0; i<128; i++) {
128 ungroup_3_in_7_bits_tab[i][0] = i / 25;
129 ungroup_3_in_7_bits_tab[i][1] = (i % 25) / 5;
130 ungroup_3_in_7_bits_tab[i][2] = (i % 25) % 5;
133 /* generate grouped mantissa tables
134 reference: Section 7.3.5 Ungrouping of Mantissas */
135 for(i=0; i<32; i++) {
136 /* bap=1 mantissas */
137 b1_mantissas[i][0] = symmetric_dequant(ff_ac3_ungroup_3_in_5_bits_tab[i][0], 3);
138 b1_mantissas[i][1] = symmetric_dequant(ff_ac3_ungroup_3_in_5_bits_tab[i][1], 3);
139 b1_mantissas[i][2] = symmetric_dequant(ff_ac3_ungroup_3_in_5_bits_tab[i][2], 3);
141 for(i=0; i<128; i++) {
142 /* bap=2 mantissas */
143 b2_mantissas[i][0] = symmetric_dequant(ungroup_3_in_7_bits_tab[i][0], 5);
144 b2_mantissas[i][1] = symmetric_dequant(ungroup_3_in_7_bits_tab[i][1], 5);
145 b2_mantissas[i][2] = symmetric_dequant(ungroup_3_in_7_bits_tab[i][2], 5);
147 /* bap=4 mantissas */
148 b4_mantissas[i][0] = symmetric_dequant(i / 11, 11);
149 b4_mantissas[i][1] = symmetric_dequant(i % 11, 11);
151 /* generate ungrouped mantissa tables
152 reference: Tables 7.21 and 7.23 */
154 /* bap=3 mantissas */
155 b3_mantissas[i] = symmetric_dequant(i, 7);
157 for(i=0; i<15; i++) {
158 /* bap=5 mantissas */
159 b5_mantissas[i] = symmetric_dequant(i, 15);
162 /* generate dynamic range table
163 reference: Section 7.7.1 Dynamic Range Control */
164 for(i=0; i<256; i++) {
165 int v = (i >> 5) - ((i >> 7) << 3) - 5;
166 dynamic_range_tab[i] = powf(2.0f, v) * ((i & 0x1F) | 0x20);
172 * AVCodec initialization
174 static av_cold int ac3_decode_init(AVCodecContext *avctx)
176 AC3DecodeContext *s = avctx->priv_data;
180 if (avctx->drc_scale)
181 s->drc_scale = avctx->drc_scale;
184 ff_ac3_common_init();
186 ff_mdct_init(&s->imdct_256, 8, 1, 1.0);
187 ff_mdct_init(&s->imdct_512, 9, 1, 1.0);
188 ff_kbd_window_init(s->window, 5.0, 256);
189 dsputil_init(&s->dsp, avctx);
190 ff_ac3dsp_init(&s->ac3dsp, avctx->flags & CODEC_FLAG_BITEXACT);
191 ff_fmt_convert_init(&s->fmt_conv, avctx);
192 av_lfg_init(&s->dith_state, 0);
194 /* set scale value for float to int16 conversion */
195 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
197 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
199 s->mul_bias = 32767.0f;
200 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
203 /* allow downmixing to stereo or mono */
204 if (avctx->channels > 0 && avctx->request_channels > 0 &&
205 avctx->request_channels < avctx->channels &&
206 avctx->request_channels <= 2) {
207 avctx->channels = avctx->request_channels;
215 * Parse the 'sync info' and 'bit stream info' from the AC-3 bitstream.
216 * GetBitContext within AC3DecodeContext must point to
217 * the start of the synchronized AC-3 bitstream.
219 static int ac3_parse_header(AC3DecodeContext *s)
221 GetBitContext *gbc = &s->gbc;
224 /* read the rest of the bsi. read twice for dual mono mode. */
225 i = !(s->channel_mode);
227 skip_bits(gbc, 5); // skip dialog normalization
229 skip_bits(gbc, 8); //skip compression
231 skip_bits(gbc, 8); //skip language code
233 skip_bits(gbc, 7); //skip audio production information
236 skip_bits(gbc, 2); //skip copyright bit and original bitstream bit
238 /* skip the timecodes (or extra bitstream information for Alternate Syntax)
239 TODO: read & use the xbsi1 downmix levels */
241 skip_bits(gbc, 14); //skip timecode1 / xbsi1
243 skip_bits(gbc, 14); //skip timecode2 / xbsi2
245 /* skip additional bitstream info */
246 if (get_bits1(gbc)) {
247 i = get_bits(gbc, 6);
257 * Common function to parse AC-3 or E-AC-3 frame header
259 static int parse_frame_header(AC3DecodeContext *s)
264 err = avpriv_ac3_parse_header(&s->gbc, &hdr);
268 /* get decoding parameters from header info */
269 s->bit_alloc_params.sr_code = hdr.sr_code;
270 s->bitstream_mode = hdr.bitstream_mode;
271 s->channel_mode = hdr.channel_mode;
272 s->channel_layout = hdr.channel_layout;
273 s->lfe_on = hdr.lfe_on;
274 s->bit_alloc_params.sr_shift = hdr.sr_shift;
275 s->sample_rate = hdr.sample_rate;
276 s->bit_rate = hdr.bit_rate;
277 s->channels = hdr.channels;
278 s->fbw_channels = s->channels - s->lfe_on;
279 s->lfe_ch = s->fbw_channels + 1;
280 s->frame_size = hdr.frame_size;
281 s->center_mix_level = hdr.center_mix_level;
282 s->surround_mix_level = hdr.surround_mix_level;
283 s->num_blocks = hdr.num_blocks;
284 s->frame_type = hdr.frame_type;
285 s->substreamid = hdr.substreamid;
288 s->start_freq[s->lfe_ch] = 0;
289 s->end_freq[s->lfe_ch] = 7;
290 s->num_exp_groups[s->lfe_ch] = 2;
291 s->channel_in_cpl[s->lfe_ch] = 0;
294 if (hdr.bitstream_id <= 10) {
296 s->snr_offset_strategy = 2;
297 s->block_switch_syntax = 1;
298 s->dither_flag_syntax = 1;
299 s->bit_allocation_syntax = 1;
300 s->fast_gain_syntax = 0;
301 s->first_cpl_leak = 0;
304 memset(s->channel_uses_aht, 0, sizeof(s->channel_uses_aht));
305 return ac3_parse_header(s);
306 } else if (CONFIG_EAC3_DECODER) {
308 return ff_eac3_parse_header(s);
310 av_log(s->avctx, AV_LOG_ERROR, "E-AC-3 support not compiled in\n");
316 * Set stereo downmixing coefficients based on frame header info.
317 * reference: Section 7.8.2 Downmixing Into Two Channels
319 static void set_downmix_coeffs(AC3DecodeContext *s)
322 float cmix = gain_levels[center_levels[s->center_mix_level]];
323 float smix = gain_levels[surround_levels[s->surround_mix_level]];
326 for(i=0; i<s->fbw_channels; i++) {
327 s->downmix_coeffs[i][0] = gain_levels[ac3_default_coeffs[s->channel_mode][i][0]];
328 s->downmix_coeffs[i][1] = gain_levels[ac3_default_coeffs[s->channel_mode][i][1]];
330 if(s->channel_mode > 1 && s->channel_mode & 1) {
331 s->downmix_coeffs[1][0] = s->downmix_coeffs[1][1] = cmix;
333 if(s->channel_mode == AC3_CHMODE_2F1R || s->channel_mode == AC3_CHMODE_3F1R) {
334 int nf = s->channel_mode - 2;
335 s->downmix_coeffs[nf][0] = s->downmix_coeffs[nf][1] = smix * LEVEL_MINUS_3DB;
337 if(s->channel_mode == AC3_CHMODE_2F2R || s->channel_mode == AC3_CHMODE_3F2R) {
338 int nf = s->channel_mode - 4;
339 s->downmix_coeffs[nf][0] = s->downmix_coeffs[nf+1][1] = smix;
344 for(i=0; i<s->fbw_channels; i++) {
345 norm0 += s->downmix_coeffs[i][0];
346 norm1 += s->downmix_coeffs[i][1];
348 norm0 = 1.0f / norm0;
349 norm1 = 1.0f / norm1;
350 for(i=0; i<s->fbw_channels; i++) {
351 s->downmix_coeffs[i][0] *= norm0;
352 s->downmix_coeffs[i][1] *= norm1;
355 if(s->output_mode == AC3_CHMODE_MONO) {
356 for(i=0; i<s->fbw_channels; i++)
357 s->downmix_coeffs[i][0] = (s->downmix_coeffs[i][0] + s->downmix_coeffs[i][1]) * LEVEL_MINUS_3DB;
362 * Decode the grouped exponents according to exponent strategy.
363 * reference: Section 7.1.3 Exponent Decoding
365 static int decode_exponents(GetBitContext *gbc, int exp_strategy, int ngrps,
366 uint8_t absexp, int8_t *dexps)
368 int i, j, grp, group_size;
373 group_size = exp_strategy + (exp_strategy == EXP_D45);
374 for(grp=0,i=0; grp<ngrps; grp++) {
375 expacc = get_bits(gbc, 7);
376 dexp[i++] = ungroup_3_in_7_bits_tab[expacc][0];
377 dexp[i++] = ungroup_3_in_7_bits_tab[expacc][1];
378 dexp[i++] = ungroup_3_in_7_bits_tab[expacc][2];
381 /* convert to absolute exps and expand groups */
383 for(i=0,j=0; i<ngrps*3; i++) {
384 prevexp += dexp[i] - 2;
387 switch (group_size) {
388 case 4: dexps[j++] = prevexp;
389 dexps[j++] = prevexp;
390 case 2: dexps[j++] = prevexp;
391 case 1: dexps[j++] = prevexp;
398 * Generate transform coefficients for each coupled channel in the coupling
399 * range using the coupling coefficients and coupling coordinates.
400 * reference: Section 7.4.3 Coupling Coordinate Format
402 static void calc_transform_coeffs_cpl(AC3DecodeContext *s)
406 bin = s->start_freq[CPL_CH];
407 for (band = 0; band < s->num_cpl_bands; band++) {
408 int band_start = bin;
409 int band_end = bin + s->cpl_band_sizes[band];
410 for (ch = 1; ch <= s->fbw_channels; ch++) {
411 if (s->channel_in_cpl[ch]) {
412 int cpl_coord = s->cpl_coords[ch][band] << 5;
413 for (bin = band_start; bin < band_end; bin++) {
414 s->fixed_coeffs[ch][bin] = MULH(s->fixed_coeffs[CPL_CH][bin] << 4, cpl_coord);
416 if (ch == 2 && s->phase_flags[band]) {
417 for (bin = band_start; bin < band_end; bin++)
418 s->fixed_coeffs[2][bin] = -s->fixed_coeffs[2][bin];
427 * Grouped mantissas for 3-level 5-level and 11-level quantization
439 * Decode the transform coefficients for a particular channel
440 * reference: Section 7.3 Quantization and Decoding of Mantissas
442 static void ac3_decode_transform_coeffs_ch(AC3DecodeContext *s, int ch_index, mant_groups *m)
444 int start_freq = s->start_freq[ch_index];
445 int end_freq = s->end_freq[ch_index];
446 uint8_t *baps = s->bap[ch_index];
447 int8_t *exps = s->dexps[ch_index];
448 int *coeffs = s->fixed_coeffs[ch_index];
449 int dither = (ch_index == CPL_CH) || s->dither_flag[ch_index];
450 GetBitContext *gbc = &s->gbc;
453 for(freq = start_freq; freq < end_freq; freq++){
454 int bap = baps[freq];
459 mantissa = (av_lfg_get(&s->dith_state) & 0x7FFFFF) - 0x400000;
466 mantissa = m->b1_mant[m->b1];
469 int bits = get_bits(gbc, 5);
470 mantissa = b1_mantissas[bits][0];
471 m->b1_mant[1] = b1_mantissas[bits][1];
472 m->b1_mant[0] = b1_mantissas[bits][2];
479 mantissa = m->b2_mant[m->b2];
482 int bits = get_bits(gbc, 7);
483 mantissa = b2_mantissas[bits][0];
484 m->b2_mant[1] = b2_mantissas[bits][1];
485 m->b2_mant[0] = b2_mantissas[bits][2];
490 mantissa = b3_mantissas[get_bits(gbc, 3)];
495 mantissa = m->b4_mant;
498 int bits = get_bits(gbc, 7);
499 mantissa = b4_mantissas[bits][0];
500 m->b4_mant = b4_mantissas[bits][1];
505 mantissa = b5_mantissas[get_bits(gbc, 4)];
507 default: /* 6 to 15 */
508 /* Shift mantissa and sign-extend it. */
509 mantissa = get_sbits(gbc, quantization_tab[bap]);
510 mantissa <<= 24 - quantization_tab[bap];
513 coeffs[freq] = mantissa >> exps[freq];
518 * Remove random dithering from coupling range coefficients with zero-bit
519 * mantissas for coupled channels which do not use dithering.
520 * reference: Section 7.3.4 Dither for Zero Bit Mantissas (bap=0)
522 static void remove_dithering(AC3DecodeContext *s) {
525 for(ch=1; ch<=s->fbw_channels; ch++) {
526 if(!s->dither_flag[ch] && s->channel_in_cpl[ch]) {
527 for(i = s->start_freq[CPL_CH]; i<s->end_freq[CPL_CH]; i++) {
528 if(!s->bap[CPL_CH][i])
529 s->fixed_coeffs[ch][i] = 0;
535 static void decode_transform_coeffs_ch(AC3DecodeContext *s, int blk, int ch,
538 if (!s->channel_uses_aht[ch]) {
539 ac3_decode_transform_coeffs_ch(s, ch, m);
541 /* if AHT is used, mantissas for all blocks are encoded in the first
542 block of the frame. */
544 if (!blk && CONFIG_EAC3_DECODER)
545 ff_eac3_decode_transform_coeffs_aht_ch(s, ch);
546 for (bin = s->start_freq[ch]; bin < s->end_freq[ch]; bin++) {
547 s->fixed_coeffs[ch][bin] = s->pre_mantissa[ch][bin][blk] >> s->dexps[ch][bin];
553 * Decode the transform coefficients.
555 static void decode_transform_coeffs(AC3DecodeContext *s, int blk)
561 m.b1 = m.b2 = m.b4 = 0;
563 for (ch = 1; ch <= s->channels; ch++) {
564 /* transform coefficients for full-bandwidth channel */
565 decode_transform_coeffs_ch(s, blk, ch, &m);
566 /* tranform coefficients for coupling channel come right after the
567 coefficients for the first coupled channel*/
568 if (s->channel_in_cpl[ch]) {
570 decode_transform_coeffs_ch(s, blk, CPL_CH, &m);
571 calc_transform_coeffs_cpl(s);
574 end = s->end_freq[CPL_CH];
576 end = s->end_freq[ch];
579 s->fixed_coeffs[ch][end] = 0;
583 /* zero the dithered coefficients for appropriate channels */
588 * Stereo rematrixing.
589 * reference: Section 7.5.4 Rematrixing : Decoding Technique
591 static void do_rematrixing(AC3DecodeContext *s)
596 end = FFMIN(s->end_freq[1], s->end_freq[2]);
598 for(bnd=0; bnd<s->num_rematrixing_bands; bnd++) {
599 if(s->rematrixing_flags[bnd]) {
600 bndend = FFMIN(end, ff_ac3_rematrix_band_tab[bnd+1]);
601 for(i=ff_ac3_rematrix_band_tab[bnd]; i<bndend; i++) {
602 int tmp0 = s->fixed_coeffs[1][i];
603 s->fixed_coeffs[1][i] += s->fixed_coeffs[2][i];
604 s->fixed_coeffs[2][i] = tmp0 - s->fixed_coeffs[2][i];
611 * Inverse MDCT Transform.
612 * Convert frequency domain coefficients to time-domain audio samples.
613 * reference: Section 7.9.4 Transformation Equations
615 static inline void do_imdct(AC3DecodeContext *s, int channels)
619 for (ch=1; ch<=channels; ch++) {
620 if (s->block_switch[ch]) {
622 float *x = s->tmp_output+128;
624 x[i] = s->transform_coeffs[ch][2*i];
625 s->imdct_256.imdct_half(&s->imdct_256, s->tmp_output, x);
626 s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 128);
628 x[i] = s->transform_coeffs[ch][2*i+1];
629 s->imdct_256.imdct_half(&s->imdct_256, s->delay[ch-1], x);
631 s->imdct_512.imdct_half(&s->imdct_512, s->tmp_output, s->transform_coeffs[ch]);
632 s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 128);
633 memcpy(s->delay[ch-1], s->tmp_output+128, 128*sizeof(float));
639 * Downmix the output to mono or stereo.
641 void ff_ac3_downmix_c(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len)
646 for(i=0; i<len; i++) {
648 for(j=0; j<in_ch; j++) {
649 v0 += samples[j][i] * matrix[j][0];
650 v1 += samples[j][i] * matrix[j][1];
655 } else if(out_ch == 1) {
656 for(i=0; i<len; i++) {
658 for(j=0; j<in_ch; j++)
659 v0 += samples[j][i] * matrix[j][0];
666 * Upmix delay samples from stereo to original channel layout.
668 static void ac3_upmix_delay(AC3DecodeContext *s)
670 int channel_data_size = sizeof(s->delay[0]);
671 switch(s->channel_mode) {
672 case AC3_CHMODE_DUALMONO:
673 case AC3_CHMODE_STEREO:
674 /* upmix mono to stereo */
675 memcpy(s->delay[1], s->delay[0], channel_data_size);
677 case AC3_CHMODE_2F2R:
678 memset(s->delay[3], 0, channel_data_size);
679 case AC3_CHMODE_2F1R:
680 memset(s->delay[2], 0, channel_data_size);
682 case AC3_CHMODE_3F2R:
683 memset(s->delay[4], 0, channel_data_size);
684 case AC3_CHMODE_3F1R:
685 memset(s->delay[3], 0, channel_data_size);
687 memcpy(s->delay[2], s->delay[1], channel_data_size);
688 memset(s->delay[1], 0, channel_data_size);
694 * Decode band structure for coupling, spectral extension, or enhanced coupling.
695 * The band structure defines how many subbands are in each band. For each
696 * subband in the range, 1 means it is combined with the previous band, and 0
697 * means that it starts a new band.
699 * @param[in] gbc bit reader context
700 * @param[in] blk block number
701 * @param[in] eac3 flag to indicate E-AC-3
702 * @param[in] ecpl flag to indicate enhanced coupling
703 * @param[in] start_subband subband number for start of range
704 * @param[in] end_subband subband number for end of range
705 * @param[in] default_band_struct default band structure table
706 * @param[out] num_bands number of bands (optionally NULL)
707 * @param[out] band_sizes array containing the number of bins in each band (optionally NULL)
709 static void decode_band_structure(GetBitContext *gbc, int blk, int eac3,
710 int ecpl, int start_subband, int end_subband,
711 const uint8_t *default_band_struct,
712 int *num_bands, uint8_t *band_sizes)
714 int subbnd, bnd, n_subbands, n_bands=0;
716 uint8_t coded_band_struct[22];
717 const uint8_t *band_struct;
719 n_subbands = end_subband - start_subband;
721 /* decode band structure from bitstream or use default */
722 if (!eac3 || get_bits1(gbc)) {
723 for (subbnd = 0; subbnd < n_subbands - 1; subbnd++) {
724 coded_band_struct[subbnd] = get_bits1(gbc);
726 band_struct = coded_band_struct;
728 band_struct = &default_band_struct[start_subband+1];
730 /* no change in band structure */
734 /* calculate number of bands and band sizes based on band structure.
735 note that the first 4 subbands in enhanced coupling span only 6 bins
737 if (num_bands || band_sizes ) {
738 n_bands = n_subbands;
739 bnd_sz[0] = ecpl ? 6 : 12;
740 for (bnd = 0, subbnd = 1; subbnd < n_subbands; subbnd++) {
741 int subbnd_size = (ecpl && subbnd < 4) ? 6 : 12;
742 if (band_struct[subbnd-1]) {
744 bnd_sz[bnd] += subbnd_size;
746 bnd_sz[++bnd] = subbnd_size;
751 /* set optional output params */
753 *num_bands = n_bands;
755 memcpy(band_sizes, bnd_sz, n_bands);
759 * Decode a single audio block from the AC-3 bitstream.
761 static int decode_audio_block(AC3DecodeContext *s, int blk)
763 int fbw_channels = s->fbw_channels;
764 int channel_mode = s->channel_mode;
766 int different_transforms;
769 GetBitContext *gbc = &s->gbc;
770 uint8_t bit_alloc_stages[AC3_MAX_CHANNELS];
772 memset(bit_alloc_stages, 0, AC3_MAX_CHANNELS);
774 /* block switch flags */
775 different_transforms = 0;
776 if (s->block_switch_syntax) {
777 for (ch = 1; ch <= fbw_channels; ch++) {
778 s->block_switch[ch] = get_bits1(gbc);
779 if(ch > 1 && s->block_switch[ch] != s->block_switch[1])
780 different_transforms = 1;
784 /* dithering flags */
785 if (s->dither_flag_syntax) {
786 for (ch = 1; ch <= fbw_channels; ch++) {
787 s->dither_flag[ch] = get_bits1(gbc);
792 i = !(s->channel_mode);
795 s->dynamic_range[i] = ((dynamic_range_tab[get_bits(gbc, 8)]-1.0) *
797 } else if(blk == 0) {
798 s->dynamic_range[i] = 1.0f;
802 /* spectral extension strategy */
803 if (s->eac3 && (!blk || get_bits1(gbc))) {
804 s->spx_in_use = get_bits1(gbc);
806 int dst_start_freq, dst_end_freq, src_start_freq,
807 start_subband, end_subband;
809 /* determine which channels use spx */
810 if (s->channel_mode == AC3_CHMODE_MONO) {
811 s->channel_uses_spx[1] = 1;
813 for (ch = 1; ch <= fbw_channels; ch++)
814 s->channel_uses_spx[ch] = get_bits1(gbc);
817 /* get the frequency bins of the spx copy region and the spx start
819 dst_start_freq = get_bits(gbc, 2);
820 start_subband = get_bits(gbc, 3) + 2;
821 if (start_subband > 7)
822 start_subband += start_subband - 7;
823 end_subband = get_bits(gbc, 3) + 5;
825 end_subband += end_subband - 7;
826 dst_start_freq = dst_start_freq * 12 + 25;
827 src_start_freq = start_subband * 12 + 25;
828 dst_end_freq = end_subband * 12 + 25;
830 /* check validity of spx ranges */
831 if (start_subband >= end_subband) {
832 av_log(s->avctx, AV_LOG_ERROR, "invalid spectral extension "
833 "range (%d >= %d)\n", start_subband, end_subband);
836 if (dst_start_freq >= src_start_freq) {
837 av_log(s->avctx, AV_LOG_ERROR, "invalid spectral extension "
838 "copy start bin (%d >= %d)\n", dst_start_freq, src_start_freq);
842 s->spx_dst_start_freq = dst_start_freq;
843 s->spx_src_start_freq = src_start_freq;
844 s->spx_dst_end_freq = dst_end_freq;
846 decode_band_structure(gbc, blk, s->eac3, 0,
847 start_subband, end_subband,
848 ff_eac3_default_spx_band_struct,
852 for (ch = 1; ch <= fbw_channels; ch++) {
853 s->channel_uses_spx[ch] = 0;
854 s->first_spx_coords[ch] = 1;
859 /* spectral extension coordinates */
861 for (ch = 1; ch <= fbw_channels; ch++) {
862 if (s->channel_uses_spx[ch]) {
863 if (s->first_spx_coords[ch] || get_bits1(gbc)) {
865 int bin, master_spx_coord;
867 s->first_spx_coords[ch] = 0;
868 spx_blend = get_bits(gbc, 5) * (1.0f/32);
869 master_spx_coord = get_bits(gbc, 2) * 3;
871 bin = s->spx_src_start_freq;
872 for (bnd = 0; bnd < s->num_spx_bands; bnd++) {
874 int spx_coord_exp, spx_coord_mant;
875 float nratio, sblend, nblend, spx_coord;
877 /* calculate blending factors */
878 bandsize = s->spx_band_sizes[bnd];
879 nratio = ((float)((bin + (bandsize >> 1))) / s->spx_dst_end_freq) - spx_blend;
880 nratio = av_clipf(nratio, 0.0f, 1.0f);
881 nblend = sqrtf(3.0f * nratio); // noise is scaled by sqrt(3) to give unity variance
882 sblend = sqrtf(1.0f - nratio);
885 /* decode spx coordinates */
886 spx_coord_exp = get_bits(gbc, 4);
887 spx_coord_mant = get_bits(gbc, 2);
888 if (spx_coord_exp == 15) spx_coord_mant <<= 1;
889 else spx_coord_mant += 4;
890 spx_coord_mant <<= (25 - spx_coord_exp - master_spx_coord);
891 spx_coord = spx_coord_mant * (1.0f/(1<<23));
893 /* multiply noise and signal blending factors by spx coordinate */
894 s->spx_noise_blend [ch][bnd] = nblend * spx_coord;
895 s->spx_signal_blend[ch][bnd] = sblend * spx_coord;
899 s->first_spx_coords[ch] = 1;
904 /* coupling strategy */
905 if (s->eac3 ? s->cpl_strategy_exists[blk] : get_bits1(gbc)) {
906 memset(bit_alloc_stages, 3, AC3_MAX_CHANNELS);
908 s->cpl_in_use[blk] = get_bits1(gbc);
909 if (s->cpl_in_use[blk]) {
910 /* coupling in use */
911 int cpl_start_subband, cpl_end_subband;
913 if (channel_mode < AC3_CHMODE_STEREO) {
914 av_log(s->avctx, AV_LOG_ERROR, "coupling not allowed in mono or dual-mono\n");
918 /* check for enhanced coupling */
919 if (s->eac3 && get_bits1(gbc)) {
920 /* TODO: parse enhanced coupling strategy info */
921 av_log_missing_feature(s->avctx, "Enhanced coupling", 1);
925 /* determine which channels are coupled */
926 if (s->eac3 && s->channel_mode == AC3_CHMODE_STEREO) {
927 s->channel_in_cpl[1] = 1;
928 s->channel_in_cpl[2] = 1;
930 for (ch = 1; ch <= fbw_channels; ch++)
931 s->channel_in_cpl[ch] = get_bits1(gbc);
934 /* phase flags in use */
935 if (channel_mode == AC3_CHMODE_STEREO)
936 s->phase_flags_in_use = get_bits1(gbc);
938 /* coupling frequency range */
939 cpl_start_subband = get_bits(gbc, 4);
940 cpl_end_subband = s->spx_in_use ? (s->spx_src_start_freq - 37) / 12 :
941 get_bits(gbc, 4) + 3;
942 if (cpl_start_subband >= cpl_end_subband) {
943 av_log(s->avctx, AV_LOG_ERROR, "invalid coupling range (%d >= %d)\n",
944 cpl_start_subband, cpl_end_subband);
947 s->start_freq[CPL_CH] = cpl_start_subband * 12 + 37;
948 s->end_freq[CPL_CH] = cpl_end_subband * 12 + 37;
950 decode_band_structure(gbc, blk, s->eac3, 0, cpl_start_subband,
952 ff_eac3_default_cpl_band_struct,
953 &s->num_cpl_bands, s->cpl_band_sizes);
955 /* coupling not in use */
956 for (ch = 1; ch <= fbw_channels; ch++) {
957 s->channel_in_cpl[ch] = 0;
958 s->first_cpl_coords[ch] = 1;
960 s->first_cpl_leak = s->eac3;
961 s->phase_flags_in_use = 0;
963 } else if (!s->eac3) {
965 av_log(s->avctx, AV_LOG_ERROR, "new coupling strategy must be present in block 0\n");
968 s->cpl_in_use[blk] = s->cpl_in_use[blk-1];
971 cpl_in_use = s->cpl_in_use[blk];
973 /* coupling coordinates */
975 int cpl_coords_exist = 0;
977 for (ch = 1; ch <= fbw_channels; ch++) {
978 if (s->channel_in_cpl[ch]) {
979 if ((s->eac3 && s->first_cpl_coords[ch]) || get_bits1(gbc)) {
980 int master_cpl_coord, cpl_coord_exp, cpl_coord_mant;
981 s->first_cpl_coords[ch] = 0;
982 cpl_coords_exist = 1;
983 master_cpl_coord = 3 * get_bits(gbc, 2);
984 for (bnd = 0; bnd < s->num_cpl_bands; bnd++) {
985 cpl_coord_exp = get_bits(gbc, 4);
986 cpl_coord_mant = get_bits(gbc, 4);
987 if (cpl_coord_exp == 15)
988 s->cpl_coords[ch][bnd] = cpl_coord_mant << 22;
990 s->cpl_coords[ch][bnd] = (cpl_coord_mant + 16) << 21;
991 s->cpl_coords[ch][bnd] >>= (cpl_coord_exp + master_cpl_coord);
994 av_log(s->avctx, AV_LOG_ERROR, "new coupling coordinates must be present in block 0\n");
998 /* channel not in coupling */
999 s->first_cpl_coords[ch] = 1;
1003 if (channel_mode == AC3_CHMODE_STEREO && cpl_coords_exist) {
1004 for (bnd = 0; bnd < s->num_cpl_bands; bnd++) {
1005 s->phase_flags[bnd] = s->phase_flags_in_use? get_bits1(gbc) : 0;
1010 /* stereo rematrixing strategy and band structure */
1011 if (channel_mode == AC3_CHMODE_STEREO) {
1012 if ((s->eac3 && !blk) || get_bits1(gbc)) {
1013 s->num_rematrixing_bands = 4;
1014 if (cpl_in_use && s->start_freq[CPL_CH] <= 61) {
1015 s->num_rematrixing_bands -= 1 + (s->start_freq[CPL_CH] == 37);
1016 } else if (s->spx_in_use && s->spx_src_start_freq <= 61) {
1017 s->num_rematrixing_bands--;
1019 for(bnd=0; bnd<s->num_rematrixing_bands; bnd++)
1020 s->rematrixing_flags[bnd] = get_bits1(gbc);
1022 av_log(s->avctx, AV_LOG_WARNING, "Warning: new rematrixing strategy not present in block 0\n");
1023 s->num_rematrixing_bands = 0;
1027 /* exponent strategies for each channel */
1028 for (ch = !cpl_in_use; ch <= s->channels; ch++) {
1030 s->exp_strategy[blk][ch] = get_bits(gbc, 2 - (ch == s->lfe_ch));
1031 if(s->exp_strategy[blk][ch] != EXP_REUSE)
1032 bit_alloc_stages[ch] = 3;
1035 /* channel bandwidth */
1036 for (ch = 1; ch <= fbw_channels; ch++) {
1037 s->start_freq[ch] = 0;
1038 if (s->exp_strategy[blk][ch] != EXP_REUSE) {
1040 int prev = s->end_freq[ch];
1041 if (s->channel_in_cpl[ch])
1042 s->end_freq[ch] = s->start_freq[CPL_CH];
1043 else if (s->channel_uses_spx[ch])
1044 s->end_freq[ch] = s->spx_src_start_freq;
1046 int bandwidth_code = get_bits(gbc, 6);
1047 if (bandwidth_code > 60) {
1048 av_log(s->avctx, AV_LOG_ERROR, "bandwidth code = %d > 60\n", bandwidth_code);
1051 s->end_freq[ch] = bandwidth_code * 3 + 73;
1053 group_size = 3 << (s->exp_strategy[blk][ch] - 1);
1054 s->num_exp_groups[ch] = (s->end_freq[ch]+group_size-4) / group_size;
1055 if(blk > 0 && s->end_freq[ch] != prev)
1056 memset(bit_alloc_stages, 3, AC3_MAX_CHANNELS);
1059 if (cpl_in_use && s->exp_strategy[blk][CPL_CH] != EXP_REUSE) {
1060 s->num_exp_groups[CPL_CH] = (s->end_freq[CPL_CH] - s->start_freq[CPL_CH]) /
1061 (3 << (s->exp_strategy[blk][CPL_CH] - 1));
1064 /* decode exponents for each channel */
1065 for (ch = !cpl_in_use; ch <= s->channels; ch++) {
1066 if (s->exp_strategy[blk][ch] != EXP_REUSE) {
1067 s->dexps[ch][0] = get_bits(gbc, 4) << !ch;
1068 if (decode_exponents(gbc, s->exp_strategy[blk][ch],
1069 s->num_exp_groups[ch], s->dexps[ch][0],
1070 &s->dexps[ch][s->start_freq[ch]+!!ch])) {
1071 av_log(s->avctx, AV_LOG_ERROR, "exponent out-of-range\n");
1074 if(ch != CPL_CH && ch != s->lfe_ch)
1075 skip_bits(gbc, 2); /* skip gainrng */
1079 /* bit allocation information */
1080 if (s->bit_allocation_syntax) {
1081 if (get_bits1(gbc)) {
1082 s->bit_alloc_params.slow_decay = ff_ac3_slow_decay_tab[get_bits(gbc, 2)] >> s->bit_alloc_params.sr_shift;
1083 s->bit_alloc_params.fast_decay = ff_ac3_fast_decay_tab[get_bits(gbc, 2)] >> s->bit_alloc_params.sr_shift;
1084 s->bit_alloc_params.slow_gain = ff_ac3_slow_gain_tab[get_bits(gbc, 2)];
1085 s->bit_alloc_params.db_per_bit = ff_ac3_db_per_bit_tab[get_bits(gbc, 2)];
1086 s->bit_alloc_params.floor = ff_ac3_floor_tab[get_bits(gbc, 3)];
1087 for(ch=!cpl_in_use; ch<=s->channels; ch++)
1088 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 2);
1090 av_log(s->avctx, AV_LOG_ERROR, "new bit allocation info must be present in block 0\n");
1095 /* signal-to-noise ratio offsets and fast gains (signal-to-mask ratios) */
1096 if(!s->eac3 || !blk){
1097 if(s->snr_offset_strategy && get_bits1(gbc)) {
1100 csnr = (get_bits(gbc, 6) - 15) << 4;
1101 for (i = ch = !cpl_in_use; ch <= s->channels; ch++) {
1103 if (ch == i || s->snr_offset_strategy == 2)
1104 snr = (csnr + get_bits(gbc, 4)) << 2;
1105 /* run at least last bit allocation stage if snr offset changes */
1106 if(blk && s->snr_offset[ch] != snr) {
1107 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 1);
1109 s->snr_offset[ch] = snr;
1111 /* fast gain (normal AC-3 only) */
1113 int prev = s->fast_gain[ch];
1114 s->fast_gain[ch] = ff_ac3_fast_gain_tab[get_bits(gbc, 3)];
1115 /* run last 2 bit allocation stages if fast gain changes */
1116 if(blk && prev != s->fast_gain[ch])
1117 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 2);
1120 } else if (!s->eac3 && !blk) {
1121 av_log(s->avctx, AV_LOG_ERROR, "new snr offsets must be present in block 0\n");
1126 /* fast gain (E-AC-3 only) */
1127 if (s->fast_gain_syntax && get_bits1(gbc)) {
1128 for (ch = !cpl_in_use; ch <= s->channels; ch++) {
1129 int prev = s->fast_gain[ch];
1130 s->fast_gain[ch] = ff_ac3_fast_gain_tab[get_bits(gbc, 3)];
1131 /* run last 2 bit allocation stages if fast gain changes */
1132 if(blk && prev != s->fast_gain[ch])
1133 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 2);
1135 } else if (s->eac3 && !blk) {
1136 for (ch = !cpl_in_use; ch <= s->channels; ch++)
1137 s->fast_gain[ch] = ff_ac3_fast_gain_tab[4];
1140 /* E-AC-3 to AC-3 converter SNR offset */
1141 if (s->frame_type == EAC3_FRAME_TYPE_INDEPENDENT && get_bits1(gbc)) {
1142 skip_bits(gbc, 10); // skip converter snr offset
1145 /* coupling leak information */
1147 if (s->first_cpl_leak || get_bits1(gbc)) {
1148 int fl = get_bits(gbc, 3);
1149 int sl = get_bits(gbc, 3);
1150 /* run last 2 bit allocation stages for coupling channel if
1151 coupling leak changes */
1152 if(blk && (fl != s->bit_alloc_params.cpl_fast_leak ||
1153 sl != s->bit_alloc_params.cpl_slow_leak)) {
1154 bit_alloc_stages[CPL_CH] = FFMAX(bit_alloc_stages[CPL_CH], 2);
1156 s->bit_alloc_params.cpl_fast_leak = fl;
1157 s->bit_alloc_params.cpl_slow_leak = sl;
1158 } else if (!s->eac3 && !blk) {
1159 av_log(s->avctx, AV_LOG_ERROR, "new coupling leak info must be present in block 0\n");
1162 s->first_cpl_leak = 0;
1165 /* delta bit allocation information */
1166 if (s->dba_syntax && get_bits1(gbc)) {
1167 /* delta bit allocation exists (strategy) */
1168 for (ch = !cpl_in_use; ch <= fbw_channels; ch++) {
1169 s->dba_mode[ch] = get_bits(gbc, 2);
1170 if (s->dba_mode[ch] == DBA_RESERVED) {
1171 av_log(s->avctx, AV_LOG_ERROR, "delta bit allocation strategy reserved\n");
1174 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 2);
1176 /* channel delta offset, len and bit allocation */
1177 for (ch = !cpl_in_use; ch <= fbw_channels; ch++) {
1178 if (s->dba_mode[ch] == DBA_NEW) {
1179 s->dba_nsegs[ch] = get_bits(gbc, 3) + 1;
1180 for (seg = 0; seg < s->dba_nsegs[ch]; seg++) {
1181 s->dba_offsets[ch][seg] = get_bits(gbc, 5);
1182 s->dba_lengths[ch][seg] = get_bits(gbc, 4);
1183 s->dba_values[ch][seg] = get_bits(gbc, 3);
1185 /* run last 2 bit allocation stages if new dba values */
1186 bit_alloc_stages[ch] = FFMAX(bit_alloc_stages[ch], 2);
1189 } else if(blk == 0) {
1190 for(ch=0; ch<=s->channels; ch++) {
1191 s->dba_mode[ch] = DBA_NONE;
1195 /* Bit allocation */
1196 for(ch=!cpl_in_use; ch<=s->channels; ch++) {
1197 if(bit_alloc_stages[ch] > 2) {
1198 /* Exponent mapping into PSD and PSD integration */
1199 ff_ac3_bit_alloc_calc_psd(s->dexps[ch],
1200 s->start_freq[ch], s->end_freq[ch],
1201 s->psd[ch], s->band_psd[ch]);
1203 if(bit_alloc_stages[ch] > 1) {
1204 /* Compute excitation function, Compute masking curve, and
1205 Apply delta bit allocation */
1206 if (ff_ac3_bit_alloc_calc_mask(&s->bit_alloc_params, s->band_psd[ch],
1207 s->start_freq[ch], s->end_freq[ch],
1208 s->fast_gain[ch], (ch == s->lfe_ch),
1209 s->dba_mode[ch], s->dba_nsegs[ch],
1210 s->dba_offsets[ch], s->dba_lengths[ch],
1211 s->dba_values[ch], s->mask[ch])) {
1212 av_log(s->avctx, AV_LOG_ERROR, "error in bit allocation\n");
1216 if(bit_alloc_stages[ch] > 0) {
1217 /* Compute bit allocation */
1218 const uint8_t *bap_tab = s->channel_uses_aht[ch] ?
1219 ff_eac3_hebap_tab : ff_ac3_bap_tab;
1220 s->ac3dsp.bit_alloc_calc_bap(s->mask[ch], s->psd[ch],
1221 s->start_freq[ch], s->end_freq[ch],
1223 s->bit_alloc_params.floor,
1224 bap_tab, s->bap[ch]);
1228 /* unused dummy data */
1229 if (s->skip_syntax && get_bits1(gbc)) {
1230 int skipl = get_bits(gbc, 9);
1235 /* unpack the transform coefficients
1236 this also uncouples channels if coupling is in use. */
1237 decode_transform_coeffs(s, blk);
1239 /* TODO: generate enhanced coupling coordinates and uncouple */
1241 /* recover coefficients if rematrixing is in use */
1242 if(s->channel_mode == AC3_CHMODE_STEREO)
1245 /* apply scaling to coefficients (headroom, dynrng) */
1246 for(ch=1; ch<=s->channels; ch++) {
1247 float gain = s->mul_bias / 4194304.0f;
1248 if(s->channel_mode == AC3_CHMODE_DUALMONO) {
1249 gain *= s->dynamic_range[2-ch];
1251 gain *= s->dynamic_range[0];
1253 s->fmt_conv.int32_to_float_fmul_scalar(s->transform_coeffs[ch], s->fixed_coeffs[ch], gain, 256);
1256 /* apply spectral extension to high frequency bins */
1257 if (s->spx_in_use && CONFIG_EAC3_DECODER) {
1258 ff_eac3_apply_spectral_extension(s);
1261 /* downmix and MDCT. order depends on whether block switching is used for
1262 any channel in this block. this is because coefficients for the long
1263 and short transforms cannot be mixed. */
1264 downmix_output = s->channels != s->out_channels &&
1265 !((s->output_mode & AC3_OUTPUT_LFEON) &&
1266 s->fbw_channels == s->out_channels);
1267 if(different_transforms) {
1268 /* the delay samples have already been downmixed, so we upmix the delay
1269 samples in order to reconstruct all channels before downmixing. */
1275 do_imdct(s, s->channels);
1277 if(downmix_output) {
1278 s->dsp.ac3_downmix(s->output, s->downmix_coeffs, s->out_channels, s->fbw_channels, 256);
1281 if(downmix_output) {
1282 s->dsp.ac3_downmix(s->transform_coeffs+1, s->downmix_coeffs, s->out_channels, s->fbw_channels, 256);
1285 if(downmix_output && !s->downmixed) {
1287 s->dsp.ac3_downmix(s->delay, s->downmix_coeffs, s->out_channels, s->fbw_channels, 128);
1290 do_imdct(s, s->out_channels);
1297 * Decode a single AC-3 frame.
1299 static int ac3_decode_frame(AVCodecContext * avctx, void *data, int *data_size,
1302 const uint8_t *buf = avpkt->data;
1303 int buf_size = avpkt->size;
1304 AC3DecodeContext *s = avctx->priv_data;
1305 float *out_samples_flt = data;
1306 int16_t *out_samples_s16 = data;
1308 int data_size_orig, data_size_tmp;
1309 const uint8_t *channel_map;
1310 const float *output[AC3_MAX_CHANNELS];
1312 /* copy input buffer to decoder context to avoid reading past the end
1313 of the buffer, which can be caused by a damaged input stream. */
1314 if (buf_size >= 2 && AV_RB16(buf) == 0x770B) {
1315 // seems to be byte-swapped AC-3
1316 int cnt = FFMIN(buf_size, AC3_FRAME_BUFFER_SIZE) >> 1;
1317 s->dsp.bswap16_buf((uint16_t *)s->input_buffer, (const uint16_t *)buf, cnt);
1319 memcpy(s->input_buffer, buf, FFMIN(buf_size, AC3_FRAME_BUFFER_SIZE));
1320 buf = s->input_buffer;
1321 /* initialize the GetBitContext with the start of valid AC-3 Frame */
1322 init_get_bits(&s->gbc, buf, buf_size * 8);
1324 /* parse the syncinfo */
1325 data_size_orig = *data_size;
1327 err = parse_frame_header(s);
1331 case AAC_AC3_PARSE_ERROR_SYNC:
1332 av_log(avctx, AV_LOG_ERROR, "frame sync error\n");
1334 case AAC_AC3_PARSE_ERROR_BSID:
1335 av_log(avctx, AV_LOG_ERROR, "invalid bitstream id\n");
1337 case AAC_AC3_PARSE_ERROR_SAMPLE_RATE:
1338 av_log(avctx, AV_LOG_ERROR, "invalid sample rate\n");
1340 case AAC_AC3_PARSE_ERROR_FRAME_SIZE:
1341 av_log(avctx, AV_LOG_ERROR, "invalid frame size\n");
1343 case AAC_AC3_PARSE_ERROR_FRAME_TYPE:
1344 /* skip frame if CRC is ok. otherwise use error concealment. */
1345 /* TODO: add support for substreams and dependent frames */
1346 if(s->frame_type == EAC3_FRAME_TYPE_DEPENDENT || s->substreamid) {
1347 av_log(avctx, AV_LOG_ERROR, "unsupported frame type : skipping frame\n");
1348 return s->frame_size;
1350 av_log(avctx, AV_LOG_ERROR, "invalid frame type\n");
1354 av_log(avctx, AV_LOG_ERROR, "invalid header\n");
1358 /* check that reported frame size fits in input buffer */
1359 if (s->frame_size > buf_size) {
1360 av_log(avctx, AV_LOG_ERROR, "incomplete frame\n");
1361 err = AAC_AC3_PARSE_ERROR_FRAME_SIZE;
1362 } else if (avctx->err_recognition & AV_EF_CRCCHECK) {
1363 /* check for crc mismatch */
1364 if (av_crc(av_crc_get_table(AV_CRC_16_ANSI), 0, &buf[2], s->frame_size-2)) {
1365 av_log(avctx, AV_LOG_ERROR, "frame CRC mismatch\n");
1366 err = AAC_AC3_PARSE_ERROR_CRC;
1371 /* if frame is ok, set audio parameters */
1373 avctx->sample_rate = s->sample_rate;
1374 avctx->bit_rate = s->bit_rate;
1376 /* channel config */
1377 s->out_channels = s->channels;
1378 s->output_mode = s->channel_mode;
1380 s->output_mode |= AC3_OUTPUT_LFEON;
1381 if (avctx->request_channels > 0 && avctx->request_channels <= 2 &&
1382 avctx->request_channels < s->channels) {
1383 s->out_channels = avctx->request_channels;
1384 s->output_mode = avctx->request_channels == 1 ? AC3_CHMODE_MONO : AC3_CHMODE_STEREO;
1385 s->channel_layout = ff_ac3_channel_layout_tab[s->output_mode];
1387 avctx->channels = s->out_channels;
1388 avctx->channel_layout = s->channel_layout;
1390 s->loro_center_mix_level = gain_levels[ center_levels[s-> center_mix_level]];
1391 s->loro_surround_mix_level = gain_levels[surround_levels[s->surround_mix_level]];
1392 s->ltrt_center_mix_level = LEVEL_MINUS_3DB;
1393 s->ltrt_surround_mix_level = LEVEL_MINUS_3DB;
1394 /* set downmixing coefficients if needed */
1395 if(s->channels != s->out_channels && !((s->output_mode & AC3_OUTPUT_LFEON) &&
1396 s->fbw_channels == s->out_channels)) {
1397 set_downmix_coeffs(s);
1399 } else if (!s->out_channels) {
1400 s->out_channels = avctx->channels;
1401 if(s->out_channels < s->channels)
1402 s->output_mode = s->out_channels == 1 ? AC3_CHMODE_MONO : AC3_CHMODE_STEREO;
1404 /* set audio service type based on bitstream mode for AC-3 */
1405 avctx->audio_service_type = s->bitstream_mode;
1406 if (s->bitstream_mode == 0x7 && s->channels > 1)
1407 avctx->audio_service_type = AV_AUDIO_SERVICE_TYPE_KARAOKE;
1409 /* decode the audio blocks */
1410 channel_map = ff_ac3_dec_channel_map[s->output_mode & ~AC3_OUTPUT_LFEON][s->lfe_on];
1411 for (ch = 0; ch < s->out_channels; ch++)
1412 output[ch] = s->output[channel_map[ch]];
1413 data_size_tmp = s->num_blocks * 256 * avctx->channels;
1414 data_size_tmp *= avctx->sample_fmt == AV_SAMPLE_FMT_FLT ? sizeof(*out_samples_flt) : sizeof(*out_samples_s16);
1415 if (data_size_orig < data_size_tmp)
1417 *data_size = data_size_tmp;
1418 for (blk = 0; blk < s->num_blocks; blk++) {
1419 if (!err && decode_audio_block(s, blk)) {
1420 av_log(avctx, AV_LOG_ERROR, "error decoding the audio block\n");
1424 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1425 s->fmt_conv.float_interleave(out_samples_flt, output, 256,
1427 out_samples_flt += 256 * s->out_channels;
1429 s->fmt_conv.float_to_int16_interleave(out_samples_s16, output, 256,
1431 out_samples_s16 += 256 * s->out_channels;
1434 *data_size = s->num_blocks * 256 * avctx->channels *
1435 av_get_bytes_per_sample(avctx->sample_fmt);
1436 return FFMIN(buf_size, s->frame_size);
1440 * Uninitialize the AC-3 decoder.
1442 static av_cold int ac3_decode_end(AVCodecContext *avctx)
1444 AC3DecodeContext *s = avctx->priv_data;
1445 ff_mdct_end(&s->imdct_512);
1446 ff_mdct_end(&s->imdct_256);
1451 #define OFFSET(x) offsetof(AC3DecodeContext, x)
1452 #define PAR (AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM)
1453 static const AVOption options[] = {
1454 { "drc_scale", "percentage of dynamic range compression to apply", OFFSET(drc_scale), AV_OPT_TYPE_FLOAT, {1.0}, 0.0, 1.0, PAR },
1456 {"dmix_mode", "Preferred Stereo Downmix Mode", OFFSET(preferred_stereo_downmix), AV_OPT_TYPE_INT, {.dbl = -1 }, -1, 2, 0, "dmix_mode"},
1457 {"ltrt_cmixlev", "Lt/Rt Center Mix Level", OFFSET(ltrt_center_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, 0},
1458 {"ltrt_surmixlev", "Lt/Rt Surround Mix Level", OFFSET(ltrt_surround_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, 0},
1459 {"loro_cmixlev", "Lo/Ro Center Mix Level", OFFSET(loro_center_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, 0},
1460 {"loro_surmixlev", "Lo/Ro Surround Mix Level", OFFSET(loro_surround_mix_level), AV_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, 0},
1465 static const AVClass ac3_decoder_class = {
1466 .class_name = "AC3 decoder",
1467 .item_name = av_default_item_name,
1469 .version = LIBAVUTIL_VERSION_INT,
1472 AVCodec ff_ac3_decoder = {
1474 .type = AVMEDIA_TYPE_AUDIO,
1476 .priv_data_size = sizeof (AC3DecodeContext),
1477 .init = ac3_decode_init,
1478 .close = ac3_decode_end,
1479 .decode = ac3_decode_frame,
1480 .long_name = NULL_IF_CONFIG_SMALL("ATSC A/52A (AC-3)"),
1481 .sample_fmts = (const enum AVSampleFormat[]) {
1482 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1484 .priv_class = &ac3_decoder_class,
1487 #if CONFIG_EAC3_DECODER
1488 static const AVClass eac3_decoder_class = {
1489 .class_name = "E-AC3 decoder",
1490 .item_name = av_default_item_name,
1492 .version = LIBAVUTIL_VERSION_INT,
1494 AVCodec ff_eac3_decoder = {
1496 .type = AVMEDIA_TYPE_AUDIO,
1497 .id = CODEC_ID_EAC3,
1498 .priv_data_size = sizeof (AC3DecodeContext),
1499 .init = ac3_decode_init,
1500 .close = ac3_decode_end,
1501 .decode = ac3_decode_frame,
1502 .long_name = NULL_IF_CONFIG_SMALL("ATSC A/52B (AC-3, E-AC-3)"),
1503 .sample_fmts = (const enum AVSampleFormat[]) {
1504 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1506 .priv_class = &eac3_decoder_class,