2 * DCA compatible decoder
3 * Copyright (C) 2004 Gildas Bazin
4 * Copyright (C) 2004 Benjamin Zores
5 * Copyright (C) 2006 Benjamin Larsson
6 * Copyright (C) 2007 Konstantin Shishkov
8 * This file is part of FFmpeg.
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
35 #include "bitstream.h"
40 /** DCA syncwords, also used for bitstream type detection */
42 #define DCA_MARKER_RAW_BE 0x7FFE8001
43 #define DCA_MARKER_RAW_LE 0xFE7F0180
44 #define DCA_MARKER_14B_BE 0x1FFFE800
45 #define DCA_MARKER_14B_LE 0xFF1F00E8
50 #define DCA_PRIM_CHANNELS_MAX (5)
51 #define DCA_SUBBANDS (32)
52 #define DCA_ABITS_MAX (32) /* Should be 28 */
53 #define DCA_SUBSUBFAMES_MAX (4)
54 #define DCA_LFE_MAX (3)
70 #define DCA_DOLBY 101 /* FIXME */
72 #define DCA_CHANNEL_BITS 6
73 #define DCA_CHANNEL_MASK 0x3F
77 #define HEADER_SIZE 14
78 #define CONVERT_BIAS 384
80 #define DCA_MAX_FRAME_SIZE 16383
84 int offset; ///< code values offset
85 int maxbits[8]; ///< max bits in VLC
86 int wrap; ///< wrap for get_vlc2()
87 VLC vlc[8]; ///< actual codes
90 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
91 static BitAlloc dca_tmode; ///< transition mode VLCs
92 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
93 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
95 /** Pre-calculated cosine modulation coefs for the QMF */
96 static float cos_mod[544];
98 static int av_always_inline get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx)
100 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
104 AVCodecContext *avctx;
106 int frame_type; ///< type of the current frame
107 int samples_deficit; ///< deficit sample count
108 int crc_present; ///< crc is present in the bitstream
109 int sample_blocks; ///< number of PCM sample blocks
110 int frame_size; ///< primary frame byte size
111 int amode; ///< audio channels arrangement
112 int sample_rate; ///< audio sampling rate
113 int bit_rate; ///< transmission bit rate
115 int downmix; ///< embedded downmix enabled
116 int dynrange; ///< embedded dynamic range flag
117 int timestamp; ///< embedded time stamp flag
118 int aux_data; ///< auxiliary data flag
119 int hdcd; ///< source material is mastered in HDCD
120 int ext_descr; ///< extension audio descriptor flag
121 int ext_coding; ///< extended coding flag
122 int aspf; ///< audio sync word insertion flag
123 int lfe; ///< low frequency effects flag
124 int predictor_history; ///< predictor history flag
125 int header_crc; ///< header crc check bytes
126 int multirate_inter; ///< multirate interpolator switch
127 int version; ///< encoder software revision
128 int copy_history; ///< copy history
129 int source_pcm_res; ///< source pcm resolution
130 int front_sum; ///< front sum/difference flag
131 int surround_sum; ///< surround sum/difference flag
132 int dialog_norm; ///< dialog normalisation parameter
134 /* Primary audio coding header */
135 int subframes; ///< number of subframes
136 int prim_channels; ///< number of primary audio channels
137 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
138 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
139 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
140 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
141 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
142 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
143 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
144 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
146 /* Primary audio coding side information */
147 int subsubframes; ///< number of subsubframes
148 int partial_samples; ///< partial subsubframe samples count
149 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
150 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
151 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
152 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
153 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
154 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
155 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
156 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
157 int dynrange_coef; ///< dynamic range coefficient
159 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
161 float lfe_data[2 * DCA_SUBSUBFAMES_MAX * DCA_LFE_MAX *
162 2 /*history */ ]; ///< Low frequency effect data
163 int lfe_scale_factor;
165 /* Subband samples history (for ADPCM) */
166 float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
167 float subband_fir_hist[DCA_PRIM_CHANNELS_MAX][512];
168 float subband_fir_noidea[DCA_PRIM_CHANNELS_MAX][64];
170 int output; ///< type of output
171 int bias; ///< output bias
173 DECLARE_ALIGNED_16(float, samples[1536]); /* 6 * 256 = 1536, might only need 5 */
174 DECLARE_ALIGNED_16(int16_t, tsamples[1536]);
176 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE];
177 int dca_buffer_size; ///< how much data is in the dca_buffer
180 /* Current position in DCA frame */
181 int current_subframe;
182 int current_subsubframe;
184 int debug_flag; ///< used for suppressing repeated error messages output
188 static void dca_init_vlcs()
190 static int vlcs_inited = 0;
196 dca_bitalloc_index.offset = 1;
197 dca_bitalloc_index.wrap = 1;
198 for (i = 0; i < 5; i++)
199 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
200 bitalloc_12_bits[i], 1, 1,
201 bitalloc_12_codes[i], 2, 2, 1);
202 dca_scalefactor.offset = -64;
203 dca_scalefactor.wrap = 2;
204 for (i = 0; i < 5; i++)
205 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
206 scales_bits[i], 1, 1,
207 scales_codes[i], 2, 2, 1);
208 dca_tmode.offset = 0;
210 for (i = 0; i < 4; i++)
211 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
213 tmode_codes[i], 2, 2, 1);
215 for(i = 0; i < 10; i++)
216 for(j = 0; j < 7; j++){
217 if(!bitalloc_codes[i][j]) break;
218 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
219 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
220 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
222 bitalloc_bits[i][j], 1, 1,
223 bitalloc_codes[i][j], 2, 2, 1);
228 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
231 *dst++ = get_bits(gb, bits);
234 static int dca_parse_frame_header(DCAContext * s)
237 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
238 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
239 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
241 s->bias = CONVERT_BIAS;
243 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
246 get_bits(&s->gb, 32);
249 s->frame_type = get_bits(&s->gb, 1);
250 s->samples_deficit = get_bits(&s->gb, 5) + 1;
251 s->crc_present = get_bits(&s->gb, 1);
252 s->sample_blocks = get_bits(&s->gb, 7) + 1;
253 s->frame_size = get_bits(&s->gb, 14) + 1;
254 if (s->frame_size < 95)
256 s->amode = get_bits(&s->gb, 6);
257 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
260 s->bit_rate = dca_bit_rates[get_bits(&s->gb, 5)];
264 s->downmix = get_bits(&s->gb, 1);
265 s->dynrange = get_bits(&s->gb, 1);
266 s->timestamp = get_bits(&s->gb, 1);
267 s->aux_data = get_bits(&s->gb, 1);
268 s->hdcd = get_bits(&s->gb, 1);
269 s->ext_descr = get_bits(&s->gb, 3);
270 s->ext_coding = get_bits(&s->gb, 1);
271 s->aspf = get_bits(&s->gb, 1);
272 s->lfe = get_bits(&s->gb, 2);
273 s->predictor_history = get_bits(&s->gb, 1);
275 /* TODO: check CRC */
277 s->header_crc = get_bits(&s->gb, 16);
279 s->multirate_inter = get_bits(&s->gb, 1);
280 s->version = get_bits(&s->gb, 4);
281 s->copy_history = get_bits(&s->gb, 2);
282 s->source_pcm_res = get_bits(&s->gb, 3);
283 s->front_sum = get_bits(&s->gb, 1);
284 s->surround_sum = get_bits(&s->gb, 1);
285 s->dialog_norm = get_bits(&s->gb, 4);
287 /* FIXME: channels mixing levels */
288 s->output = DCA_STEREO;
291 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
292 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
293 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
294 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
295 s->sample_blocks, s->sample_blocks * 32);
296 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
297 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
298 s->amode, dca_channels[s->amode]);
299 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i (%i Hz)\n",
300 s->sample_rate, dca_sample_rates[s->sample_rate]);
301 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i (%i bits/s)\n",
302 s->bit_rate, dca_bit_rates[s->bit_rate]);
303 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
304 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
305 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
306 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
307 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
308 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
309 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
310 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
311 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
312 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
313 s->predictor_history);
314 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
315 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
317 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
318 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
319 av_log(s->avctx, AV_LOG_DEBUG,
320 "source pcm resolution: %i (%i bits/sample)\n",
321 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
322 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
323 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
324 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
325 av_log(s->avctx, AV_LOG_DEBUG, "\n");
328 /* Primary audio coding header */
329 s->subframes = get_bits(&s->gb, 4) + 1;
330 s->prim_channels = get_bits(&s->gb, 3) + 1;
333 for (i = 0; i < s->prim_channels; i++) {
334 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
335 if (s->subband_activity[i] > DCA_SUBBANDS)
336 s->subband_activity[i] = DCA_SUBBANDS;
338 for (i = 0; i < s->prim_channels; i++) {
339 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
340 if (s->vq_start_subband[i] > DCA_SUBBANDS)
341 s->vq_start_subband[i] = DCA_SUBBANDS;
343 get_array(&s->gb, s->joint_intensity, s->prim_channels, 3);
344 get_array(&s->gb, s->transient_huffman, s->prim_channels, 2);
345 get_array(&s->gb, s->scalefactor_huffman, s->prim_channels, 3);
346 get_array(&s->gb, s->bitalloc_huffman, s->prim_channels, 3);
348 /* Get codebooks quantization indexes */
349 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
350 for (j = 1; j < 11; j++)
351 for (i = 0; i < s->prim_channels; i++)
352 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
354 /* Get scale factor adjustment */
355 for (j = 0; j < 11; j++)
356 for (i = 0; i < s->prim_channels; i++)
357 s->scalefactor_adj[i][j] = 1;
359 for (j = 1; j < 11; j++)
360 for (i = 0; i < s->prim_channels; i++)
361 if (s->quant_index_huffman[i][j] < thr[j])
362 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
364 if (s->crc_present) {
365 /* Audio header CRC check */
366 get_bits(&s->gb, 16);
369 s->current_subframe = 0;
370 s->current_subsubframe = 0;
373 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
374 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
375 for(i = 0; i < s->prim_channels; i++){
376 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
377 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
378 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
379 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
380 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
381 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
382 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
383 for (j = 0; j < 11; j++)
384 av_log(s->avctx, AV_LOG_DEBUG, " %i",
385 s->quant_index_huffman[i][j]);
386 av_log(s->avctx, AV_LOG_DEBUG, "\n");
387 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
388 for (j = 0; j < 11; j++)
389 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
390 av_log(s->avctx, AV_LOG_DEBUG, "\n");
398 static inline int get_scale(GetBitContext *gb, int level, int index, int value)
401 /* huffman encoded */
402 value += get_bitalloc(gb, &dca_scalefactor, index);
404 value = get_bits(gb, level + 1);
408 static int dca_subframe_header(DCAContext * s)
410 /* Primary audio coding side information */
413 s->subsubframes = get_bits(&s->gb, 2) + 1;
414 s->partial_samples = get_bits(&s->gb, 3);
415 for (j = 0; j < s->prim_channels; j++) {
416 for (k = 0; k < s->subband_activity[j]; k++)
417 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
420 /* Get prediction codebook */
421 for (j = 0; j < s->prim_channels; j++) {
422 for (k = 0; k < s->subband_activity[j]; k++) {
423 if (s->prediction_mode[j][k] > 0) {
424 /* (Prediction coefficient VQ address) */
425 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
430 /* Bit allocation index */
431 for (j = 0; j < s->prim_channels; j++) {
432 for (k = 0; k < s->vq_start_subband[j]; k++) {
433 if (s->bitalloc_huffman[j] == 6)
434 s->bitalloc[j][k] = get_bits(&s->gb, 5);
435 else if (s->bitalloc_huffman[j] == 5)
436 s->bitalloc[j][k] = get_bits(&s->gb, 4);
439 get_bitalloc(&s->gb, &dca_bitalloc_index, j);
442 if (s->bitalloc[j][k] > 26) {
443 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
444 // j, k, s->bitalloc[j][k]);
450 /* Transition mode */
451 for (j = 0; j < s->prim_channels; j++) {
452 for (k = 0; k < s->subband_activity[j]; k++) {
453 s->transition_mode[j][k] = 0;
454 if (s->subsubframes > 1 &&
455 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
456 s->transition_mode[j][k] =
457 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
462 for (j = 0; j < s->prim_channels; j++) {
463 uint32_t *scale_table;
466 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
468 if (s->scalefactor_huffman[j] == 6)
469 scale_table = (uint32_t *) scale_factor_quant7;
471 scale_table = (uint32_t *) scale_factor_quant6;
473 /* When huffman coded, only the difference is encoded */
476 for (k = 0; k < s->subband_activity[j]; k++) {
477 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
478 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], j, scale_sum);
479 s->scale_factor[j][k][0] = scale_table[scale_sum];
482 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
483 /* Get second scale factor */
484 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], j, scale_sum);
485 s->scale_factor[j][k][1] = scale_table[scale_sum];
490 /* Joint subband scale factor codebook select */
491 for (j = 0; j < s->prim_channels; j++) {
492 /* Transmitted only if joint subband coding enabled */
493 if (s->joint_intensity[j] > 0)
494 s->joint_huff[j] = get_bits(&s->gb, 3);
497 /* Scale factors for joint subband coding */
498 for (j = 0; j < s->prim_channels; j++) {
501 /* Transmitted only if joint subband coding enabled */
502 if (s->joint_intensity[j] > 0) {
504 source_channel = s->joint_intensity[j] - 1;
506 /* When huffman coded, only the difference is encoded
507 * (is this valid as well for joint scales ???) */
509 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
510 scale = get_scale(&s->gb, s->joint_huff[j], j, 0);
511 scale += 64; /* bias */
512 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
515 if (!s->debug_flag & 0x02) {
516 av_log(s->avctx, AV_LOG_DEBUG,
517 "Joint stereo coding not supported\n");
518 s->debug_flag |= 0x02;
523 /* Stereo downmix coefficients */
524 if (s->prim_channels > 2 && s->downmix) {
525 for (j = 0; j < s->prim_channels; j++) {
526 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
527 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
531 /* Dynamic range coefficient */
533 s->dynrange_coef = get_bits(&s->gb, 8);
535 /* Side information CRC check word */
536 if (s->crc_present) {
537 get_bits(&s->gb, 16);
541 * Primary audio data arrays
544 /* VQ encoded high frequency subbands */
545 for (j = 0; j < s->prim_channels; j++)
546 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
547 /* 1 vector -> 32 samples */
548 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
550 /* Low frequency effect data */
553 int lfe_samples = 2 * s->lfe * s->subsubframes;
556 for (j = lfe_samples; j < lfe_samples * 2; j++) {
557 /* Signed 8 bits int */
558 s->lfe_data[j] = get_sbits(&s->gb, 8);
561 /* Scale factor index */
562 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
564 /* Quantization step size * scale factor */
565 lfe_scale = 0.035 * s->lfe_scale_factor;
567 for (j = lfe_samples; j < lfe_samples * 2; j++)
568 s->lfe_data[j] *= lfe_scale;
572 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes);
573 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
575 for (j = 0; j < s->prim_channels; j++) {
576 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
577 for (k = 0; k < s->subband_activity[j]; k++)
578 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
579 av_log(s->avctx, AV_LOG_DEBUG, "\n");
581 for (j = 0; j < s->prim_channels; j++) {
582 for (k = 0; k < s->subband_activity[j]; k++)
583 av_log(s->avctx, AV_LOG_DEBUG,
584 "prediction coefs: %f, %f, %f, %f\n",
585 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
586 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
587 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
588 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
590 for (j = 0; j < s->prim_channels; j++) {
591 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
592 for (k = 0; k < s->vq_start_subband[j]; k++)
593 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
594 av_log(s->avctx, AV_LOG_DEBUG, "\n");
596 for (j = 0; j < s->prim_channels; j++) {
597 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
598 for (k = 0; k < s->subband_activity[j]; k++)
599 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
600 av_log(s->avctx, AV_LOG_DEBUG, "\n");
602 for (j = 0; j < s->prim_channels; j++) {
603 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
604 for (k = 0; k < s->subband_activity[j]; k++) {
605 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
606 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
607 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
608 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
610 av_log(s->avctx, AV_LOG_DEBUG, "\n");
612 for (j = 0; j < s->prim_channels; j++) {
613 if (s->joint_intensity[j] > 0) {
614 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
615 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
616 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
617 av_log(s->avctx, AV_LOG_DEBUG, "\n");
620 if (s->prim_channels > 2 && s->downmix) {
621 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
622 for (j = 0; j < s->prim_channels; j++) {
623 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
624 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
626 av_log(s->avctx, AV_LOG_DEBUG, "\n");
628 for (j = 0; j < s->prim_channels; j++)
629 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
630 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
632 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
633 for (j = lfe_samples; j < lfe_samples * 2; j++)
634 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
635 av_log(s->avctx, AV_LOG_DEBUG, "\n");
642 static void qmf_32_subbands(DCAContext * s, int chans,
643 float samples_in[32][8], float *samples_out,
644 float scale, float bias)
648 float praXin[33], *raXin = &praXin[1];
650 float *subband_fir_hist = s->subband_fir_hist[chans];
651 float *subband_fir_hist2 = s->subband_fir_noidea[chans];
653 int chindex = 0, subindex;
658 if (!s->multirate_inter) /* Non-perfect reconstruction */
659 prCoeff = (float *) fir_32bands_nonperfect;
660 else /* Perfect reconstruction */
661 prCoeff = (float *) fir_32bands_perfect;
663 /* Reconstructed channel sample index */
664 for (subindex = 0; subindex < 8; subindex++) {
665 float t1, t2, sum[16], diff[16];
667 /* Load in one sample from each subband and clear inactive subbands */
668 for (i = 0; i < s->subband_activity[chans]; i++)
669 raXin[i] = samples_in[i][subindex];
673 /* Multiply by cosine modulation coefficients and
674 * create temporary arrays SUM and DIFF */
675 for (j = 0, k = 0; k < 16; k++) {
678 for (i = 0; i < 16; i++, j++){
679 t1 += (raXin[2 * i] + raXin[2 * i + 1]) * cos_mod[j];
680 t2 += (raXin[2 * i] + raXin[2 * i - 1]) * cos_mod[j + 256];
688 for (k = 0; k < 16; k++)
689 subband_fir_hist[k] = cos_mod[j++] * sum[k];
690 for (k = 0; k < 16; k++)
691 subband_fir_hist[32-k-1] = cos_mod[j++] * diff[k];
693 /* Multiply by filter coefficients */
694 for (k = 31, i = 0; i < 32; i++, k--)
695 for (j = 0; j < 512; j += 64){
696 subband_fir_hist2[i] += prCoeff[i+j] * ( subband_fir_hist[i+j] - subband_fir_hist[j+k]);
697 subband_fir_hist2[i+32] += prCoeff[i+j+32]*(-subband_fir_hist[i+j] - subband_fir_hist[j+k]);
700 /* Create 32 PCM output samples */
701 for (i = 0; i < 32; i++)
702 samples_out[chindex++] = subband_fir_hist2[i] * scale + bias;
704 /* Update working arrays */
705 memmove(&subband_fir_hist[32], &subband_fir_hist[0], (512 - 32) * sizeof(float));
706 memmove(&subband_fir_hist2[0], &subband_fir_hist2[32], 32 * sizeof(float));
707 memset(&subband_fir_hist2[32], 0, 32 * sizeof(float));
711 static void lfe_interpolation_fir(int decimation_select,
712 int num_deci_sample, float *samples_in,
713 float *samples_out, float scale,
716 /* samples_in: An array holding decimated samples.
717 * Samples in current subframe starts from samples_in[0],
718 * while samples_in[-1], samples_in[-2], ..., stores samples
719 * from last subframe as history.
721 * samples_out: An array holding interpolated samples
724 int decifactor, k, j;
725 const float *prCoeff;
727 int interp_index = 0; /* Index to the interpolated samples */
730 /* Select decimation filter */
731 if (decimation_select == 1) {
733 prCoeff = lfe_fir_128;
736 prCoeff = lfe_fir_64;
739 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
740 /* One decimated sample generates decifactor interpolated ones */
741 for (k = 0; k < decifactor; k++) {
743 //FIXME the coeffs are symetric, fix that
744 for (j = 0; j < 512 / decifactor; j++)
745 rTmp += samples_in[deciindex - j] * prCoeff[k + j * decifactor];
746 samples_out[interp_index++] = rTmp / scale + bias;
751 /* downmixing routines */
752 #define MIX_REAR1(samples, si1) \
753 samples[i] += samples[si1]; \
754 samples[i+256] += samples[si1];
756 #define MIX_REAR2(samples, si1, si2) \
757 samples[i] += samples[si1]; \
758 samples[i+256] += samples[si2];
760 #define MIX_FRONT3(samples) \
762 samples[i] += samples[i+256]; \
763 samples[i+256] = samples[i+512] + t;
765 #define DOWNMIX_TO_STEREO(op1, op2) \
766 for(i = 0; i < 256; i++){ \
771 static void dca_downmix(float *samples, int srcfmt)
779 case DCA_STEREO_TOTAL:
780 case DCA_STEREO_SUMDIFF:
782 av_log(NULL, 0, "Not implemented!\n");
787 DOWNMIX_TO_STEREO(MIX_FRONT3(samples),);
790 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + 512),);
793 DOWNMIX_TO_STEREO(MIX_FRONT3(samples),
794 MIX_REAR1(samples, i + 768));
797 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + 512, i + 768),);
800 DOWNMIX_TO_STEREO(MIX_FRONT3(samples),
801 MIX_REAR2(samples, i + 768, i + 1024));
807 /* Very compact version of the block code decoder that does not use table
808 * look-up but is slightly slower */
809 static int decode_blockcode(int code, int levels, int *values)
812 int offset = (levels - 1) >> 1;
814 for (i = 0; i < 4; i++) {
815 values[i] = (code % levels) - offset;
822 av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n");
827 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
828 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
830 static int dca_subsubframe(DCAContext * s)
833 int subsubframe = s->current_subsubframe;
835 float *quant_step_table;
838 float subband_samples[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
844 /* Select quantization step size table */
845 if (s->bit_rate == 0x1f)
846 quant_step_table = (float *) lossless_quant_d;
848 quant_step_table = (float *) lossy_quant_d;
850 for (k = 0; k < s->prim_channels; k++) {
851 for (l = 0; l < s->vq_start_subband[k]; l++) {
854 /* Select the mid-tread linear quantizer */
855 int abits = s->bitalloc[k][l];
857 float quant_step_size = quant_step_table[abits];
861 * Determine quantization index code book and its type
864 /* Select quantization index code book */
865 int sel = s->quant_index_huffman[k][abits];
868 * Extract bits from the bit stream
871 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
872 }else if(abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
875 int block_code1, block_code2, size, levels;
878 size = abits_sizes[abits-1];
879 levels = abits_levels[abits-1];
881 block_code1 = get_bits(&s->gb, size);
882 /* FIXME Should test return value */
883 decode_blockcode(block_code1, levels, block);
884 block_code2 = get_bits(&s->gb, size);
885 decode_blockcode(block_code2, levels, &block[4]);
886 for (m = 0; m < 8; m++)
887 subband_samples[k][l][m] = block[m];
890 for (m = 0; m < 8; m++)
891 subband_samples[k][l][m] = get_sbits(&s->gb, abits - 3);
895 for (m = 0; m < 8; m++)
896 subband_samples[k][l][m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
899 /* Deal with transients */
900 if (s->transition_mode[k][l] &&
901 subsubframe >= s->transition_mode[k][l])
902 rscale = quant_step_size * s->scale_factor[k][l][1];
904 rscale = quant_step_size * s->scale_factor[k][l][0];
906 rscale *= s->scalefactor_adj[k][sel];
908 for (m = 0; m < 8; m++)
909 subband_samples[k][l][m] *= rscale;
912 * Inverse ADPCM if in prediction mode
914 if (s->prediction_mode[k][l]) {
916 for (m = 0; m < 8; m++) {
917 for (n = 1; n <= 4; n++)
919 subband_samples[k][l][m] +=
920 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
921 subband_samples[k][l][m - n] / 8192);
922 else if (s->predictor_history)
923 subband_samples[k][l][m] +=
924 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
925 s->subband_samples_hist[k][l][m - n +
932 * Decode VQ encoded high frequencies
934 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
935 /* 1 vector -> 32 samples but we only need the 8 samples
936 * for this subsubframe. */
939 if (!s->debug_flag & 0x01) {
940 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
941 s->debug_flag |= 0x01;
944 for (m = 0; m < 8; m++) {
945 subband_samples[k][l][m] =
946 high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
948 * (float) s->scale_factor[k][l][0] / 16.0;
953 /* Check for DSYNC after subsubframe */
954 if (s->aspf || subsubframe == s->subsubframes - 1) {
955 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
957 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
960 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
964 /* Backup predictor history for adpcm */
965 for (k = 0; k < s->prim_channels; k++)
966 for (l = 0; l < s->vq_start_subband[k]; l++)
967 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
968 4 * sizeof(subband_samples[0][0][0]));
970 /* 32 subbands QMF */
971 for (k = 0; k < s->prim_channels; k++) {
972 /* static float pcm_to_double[8] =
973 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
974 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * k],
975 2.0 / 3 /*pcm_to_double[s->source_pcm_res] */ ,
981 if (s->prim_channels > dca_channels[s->output & DCA_CHANNEL_MASK]) {
982 dca_downmix(s->samples, s->amode);
985 /* Generate LFE samples for this subsubframe FIXME!!! */
986 if (s->output & DCA_LFE) {
987 int lfe_samples = 2 * s->lfe * s->subsubframes;
988 int i_channels = dca_channels[s->output & DCA_CHANNEL_MASK];
990 lfe_interpolation_fir(s->lfe, 2 * s->lfe,
991 s->lfe_data + lfe_samples +
992 2 * s->lfe * subsubframe,
993 &s->samples[256 * i_channels],
995 /* Outputs 20bits pcm samples */
1002 static int dca_subframe_footer(DCAContext * s)
1004 int aux_data_count = 0, i;
1008 * Unpack optional information
1012 get_bits(&s->gb, 32);
1015 aux_data_count = get_bits(&s->gb, 6);
1017 for (i = 0; i < aux_data_count; i++)
1018 get_bits(&s->gb, 8);
1020 if (s->crc_present && (s->downmix || s->dynrange))
1021 get_bits(&s->gb, 16);
1023 lfe_samples = 2 * s->lfe * s->subsubframes;
1024 for (i = 0; i < lfe_samples; i++) {
1025 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1032 * Decode a dca frame block
1034 * @param s pointer to the DCAContext
1037 static int dca_decode_block(DCAContext * s)
1041 if (s->current_subframe >= s->subframes) {
1042 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1043 s->current_subframe, s->subframes);
1047 if (!s->current_subsubframe) {
1049 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1051 /* Read subframe header */
1052 if (dca_subframe_header(s))
1056 /* Read subsubframe */
1058 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1060 if (dca_subsubframe(s))
1064 s->current_subsubframe++;
1065 if (s->current_subsubframe >= s->subsubframes) {
1066 s->current_subsubframe = 0;
1067 s->current_subframe++;
1069 if (s->current_subframe >= s->subframes) {
1071 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1073 /* Read subframe footer */
1074 if (dca_subframe_footer(s))
1082 * Convert bitstream to one representation based on sync marker
1084 static int dca_convert_bitstream(uint8_t * src, int src_size, uint8_t * dst,
1089 uint16_t *ssrc = (uint16_t *) src, *sdst = (uint16_t *) dst;
1094 case DCA_MARKER_RAW_BE:
1095 memcpy(dst, src, FFMIN(src_size, max_size));
1096 return FFMIN(src_size, max_size);
1097 case DCA_MARKER_RAW_LE:
1098 for (i = 0; i < (FFMIN(src_size, max_size) + 1) >> 1; i++)
1099 *sdst++ = bswap_16(*ssrc++);
1100 return FFMIN(src_size, max_size);
1101 case DCA_MARKER_14B_BE:
1102 case DCA_MARKER_14B_LE:
1103 init_put_bits(&pb, dst, max_size);
1104 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1105 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1106 put_bits(&pb, 14, tmp);
1108 flush_put_bits(&pb);
1109 return (put_bits_count(&pb) + 7) >> 3;
1116 * Main frame decoding function
1117 * FIXME add arguments
1119 static int dca_decode_frame(AVCodecContext * avctx,
1120 void *data, int *data_size,
1121 uint8_t * buf, int buf_size)
1125 int16_t *samples = data;
1126 DCAContext *s = avctx->priv_data;
1130 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer, DCA_MAX_FRAME_SIZE);
1131 if (s->dca_buffer_size == -1) {
1132 av_log(avctx, AV_LOG_ERROR, "Not a DCA frame\n");
1136 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1137 if (dca_parse_frame_header(s) < 0) {
1138 //seems like the frame is corrupt, try with the next one
1141 //set AVCodec values with parsed data
1142 avctx->sample_rate = s->sample_rate;
1143 avctx->channels = 2; //FIXME
1144 avctx->bit_rate = s->bit_rate;
1146 channels = dca_channels[s->output];
1147 if(*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels)
1150 for (i = 0; i < (s->sample_blocks / 8); i++) {
1151 dca_decode_block(s);
1152 s->dsp.float_to_int16(s->tsamples, s->samples, 256 * channels);
1153 /* interleave samples */
1154 for (j = 0; j < 256; j++) {
1155 for (k = 0; k < channels; k++)
1156 samples[k] = s->tsamples[j + k * 256];
1157 samples += channels;
1159 *data_size += 256 * sizeof(int16_t) * channels;
1168 * Build the cosine modulation tables for the QMF
1170 * @param s pointer to the DCAContext
1173 static void pre_calc_cosmod(DCAContext * s)
1176 static int cosmod_inited = 0;
1178 if(cosmod_inited) return;
1179 for (j = 0, k = 0; k < 16; k++)
1180 for (i = 0; i < 16; i++)
1181 cos_mod[j++] = cos((2 * i + 1) * (2 * k + 1) * M_PI / 64);
1183 for (k = 0; k < 16; k++)
1184 for (i = 0; i < 16; i++)
1185 cos_mod[j++] = cos((i) * (2 * k + 1) * M_PI / 32);
1187 for (k = 0; k < 16; k++)
1188 cos_mod[j++] = 0.25 / (2 * cos((2 * k + 1) * M_PI / 128));
1190 for (k = 0; k < 16; k++)
1191 cos_mod[j++] = -0.25 / (2.0 * sin((2 * k + 1) * M_PI / 128));
1198 * DCA initialization
1200 * @param avctx pointer to the AVCodecContext
1203 static int dca_decode_init(AVCodecContext * avctx)
1205 DCAContext *s = avctx->priv_data;
1211 dsputil_init(&s->dsp, avctx);
1216 AVCodec dca_decoder = {
1218 .type = CODEC_TYPE_AUDIO,
1220 .priv_data_size = sizeof(DCAContext),
1221 .init = dca_decode_init,
1222 .decode = dca_decode_frame,
1225 #ifdef CONFIG_DCA_PARSER
1227 typedef struct DCAParseContext {
1229 uint32_t lastmarker;
1232 #define IS_MARKER(state, i, buf, buf_size) \
1233 ((state == DCA_MARKER_14B_LE && (i < buf_size-2) && (buf[i+1] & 0xF0) == 0xF0 && buf[i+2] == 0x07) \
1234 || (state == DCA_MARKER_14B_BE && (i < buf_size-2) && buf[i+1] == 0x07 && (buf[i+2] & 0xF0) == 0xF0) \
1235 || state == DCA_MARKER_RAW_LE || state == DCA_MARKER_RAW_BE)
1238 * finds the end of the current frame in the bitstream.
1239 * @return the position of the first byte of the next frame, or -1
1241 static int dca_find_frame_end(DCAParseContext * pc1, const uint8_t * buf,
1246 ParseContext *pc = &pc1->pc;
1248 start_found = pc->frame_start_found;
1253 for (i = 0; i < buf_size; i++) {
1254 state = (state << 8) | buf[i];
1255 if (IS_MARKER(state, i, buf, buf_size)) {
1256 if (pc1->lastmarker && state == pc1->lastmarker) {
1259 } else if (!pc1->lastmarker) {
1261 pc1->lastmarker = state;
1268 for (; i < buf_size; i++) {
1269 state = (state << 8) | buf[i];
1270 if (state == pc1->lastmarker && IS_MARKER(state, i, buf, buf_size)) {
1271 pc->frame_start_found = 0;
1277 pc->frame_start_found = start_found;
1279 return END_NOT_FOUND;
1282 static int dca_parse_init(AVCodecParserContext * s)
1284 DCAParseContext *pc1 = s->priv_data;
1286 pc1->lastmarker = 0;
1290 static int dca_parse(AVCodecParserContext * s,
1291 AVCodecContext * avctx,
1292 uint8_t ** poutbuf, int *poutbuf_size,
1293 const uint8_t * buf, int buf_size)
1295 DCAParseContext *pc1 = s->priv_data;
1296 ParseContext *pc = &pc1->pc;
1299 if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
1302 next = dca_find_frame_end(pc1, buf, buf_size);
1304 if (ff_combine_frame(pc, next, (uint8_t **) & buf, &buf_size) < 0) {
1310 *poutbuf = (uint8_t *) buf;
1311 *poutbuf_size = buf_size;
1315 AVCodecParser dca_parser = {
1317 sizeof(DCAParseContext),
1322 #endif /* CONFIG_DCA_PARSER */