2 * ATRAC3+ compatible decoder
4 * Copyright (c) 2010-2013 Maxim Poliakovski
6 * This file is part of Libav.
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 * Sony ATRAC3+ compatible decoder.
27 * Container formats used to store its data:
28 * RIFF WAV (.at3) and Sony OpenMG (.oma, .aa3).
30 * Technical description of this codec can be found here:
31 * http://wiki.multimedia.cx/index.php?title=ATRAC3plus
33 * Kudos to Benjamin Larsson and Michael Karcher
34 * for their precious technical help!
40 #include "libavutil/channel_layout.h"
41 #include "libavutil/float_dsp.h"
46 #include "atrac3plus.h"
48 typedef struct ATRAC3PContext {
50 AVFloatDSPContext fdsp;
52 DECLARE_ALIGNED(32, float, samples)[2][ATRAC3P_FRAME_SAMPLES]; ///< quantized MDCT spectrum
53 DECLARE_ALIGNED(32, float, mdct_buf)[2][ATRAC3P_FRAME_SAMPLES]; ///< output of the IMDCT
54 DECLARE_ALIGNED(32, float, time_buf)[2][ATRAC3P_FRAME_SAMPLES]; ///< output of the gain compensation
55 DECLARE_ALIGNED(32, float, outp_buf)[2][ATRAC3P_FRAME_SAMPLES];
57 AtracGCContext gainc_ctx; ///< gain compensation context
59 FFTContext ipqf_dct_ctx; ///< IDCT context used by IPQF
61 Atrac3pChanUnitCtx *ch_units; ///< global channel units
63 int num_channel_blocks; ///< number of channel blocks
64 uint8_t channel_blocks[5]; ///< channel configuration descriptor
65 uint64_t my_channel_layout; ///< current channel layout
68 static av_cold int atrac3p_decode_close(AVCodecContext *avctx)
70 av_free(((ATRAC3PContext *)(avctx->priv_data))->ch_units);
75 static av_cold int set_channel_params(ATRAC3PContext *ctx,
76 AVCodecContext *avctx)
78 memset(ctx->channel_blocks, 0, sizeof(ctx->channel_blocks));
80 switch (avctx->channels) {
82 if (avctx->channel_layout != AV_CH_FRONT_LEFT)
83 avctx->channel_layout = AV_CH_LAYOUT_MONO;
85 ctx->num_channel_blocks = 1;
86 ctx->channel_blocks[0] = CH_UNIT_MONO;
89 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
90 ctx->num_channel_blocks = 1;
91 ctx->channel_blocks[0] = CH_UNIT_STEREO;
94 avctx->channel_layout = AV_CH_LAYOUT_SURROUND;
95 ctx->num_channel_blocks = 2;
96 ctx->channel_blocks[0] = CH_UNIT_STEREO;
97 ctx->channel_blocks[1] = CH_UNIT_MONO;
100 avctx->channel_layout = AV_CH_LAYOUT_4POINT0;
101 ctx->num_channel_blocks = 3;
102 ctx->channel_blocks[0] = CH_UNIT_STEREO;
103 ctx->channel_blocks[1] = CH_UNIT_MONO;
104 ctx->channel_blocks[2] = CH_UNIT_MONO;
107 avctx->channel_layout = AV_CH_LAYOUT_5POINT1_BACK;
108 ctx->num_channel_blocks = 4;
109 ctx->channel_blocks[0] = CH_UNIT_STEREO;
110 ctx->channel_blocks[1] = CH_UNIT_MONO;
111 ctx->channel_blocks[2] = CH_UNIT_STEREO;
112 ctx->channel_blocks[3] = CH_UNIT_MONO;
115 avctx->channel_layout = AV_CH_LAYOUT_6POINT1_BACK;
116 ctx->num_channel_blocks = 5;
117 ctx->channel_blocks[0] = CH_UNIT_STEREO;
118 ctx->channel_blocks[1] = CH_UNIT_MONO;
119 ctx->channel_blocks[2] = CH_UNIT_STEREO;
120 ctx->channel_blocks[3] = CH_UNIT_MONO;
121 ctx->channel_blocks[4] = CH_UNIT_MONO;
124 avctx->channel_layout = AV_CH_LAYOUT_7POINT1;
125 ctx->num_channel_blocks = 5;
126 ctx->channel_blocks[0] = CH_UNIT_STEREO;
127 ctx->channel_blocks[1] = CH_UNIT_MONO;
128 ctx->channel_blocks[2] = CH_UNIT_STEREO;
129 ctx->channel_blocks[3] = CH_UNIT_STEREO;
130 ctx->channel_blocks[4] = CH_UNIT_MONO;
133 av_log(avctx, AV_LOG_ERROR,
134 "Unsupported channel count: %d!\n", avctx->channels);
135 return AVERROR_INVALIDDATA;
141 static av_cold int atrac3p_decode_init(AVCodecContext *avctx)
143 ATRAC3PContext *ctx = avctx->priv_data;
146 if (!avctx->block_align) {
147 av_log(avctx, AV_LOG_ERROR, "block_align is not set\n");
148 return AVERROR(EINVAL);
151 avpriv_float_dsp_init(&ctx->fdsp, avctx->flags & AV_CODEC_FLAG_BITEXACT);
153 /* initialize IPQF */
154 ff_mdct_init(&ctx->ipqf_dct_ctx, 5, 1, 32.0 / 32768.0);
156 ff_atrac3p_init_imdct(avctx, &ctx->mdct_ctx);
158 ff_atrac_init_gain_compensation(&ctx->gainc_ctx, 6, 2);
160 ff_atrac3p_init_wave_synth();
162 if ((ret = set_channel_params(ctx, avctx)) < 0)
165 ctx->my_channel_layout = avctx->channel_layout;
167 ctx->ch_units = av_mallocz(sizeof(*ctx->ch_units) *
168 ctx->num_channel_blocks);
169 if (!ctx->ch_units) {
170 atrac3p_decode_close(avctx);
171 return AVERROR(ENOMEM);
174 for (i = 0; i < ctx->num_channel_blocks; i++) {
175 for (ch = 0; ch < 2; ch++) {
176 ctx->ch_units[i].channels[ch].ch_num = ch;
177 ctx->ch_units[i].channels[ch].wnd_shape = &ctx->ch_units[i].channels[ch].wnd_shape_hist[0][0];
178 ctx->ch_units[i].channels[ch].wnd_shape_prev = &ctx->ch_units[i].channels[ch].wnd_shape_hist[1][0];
179 ctx->ch_units[i].channels[ch].gain_data = &ctx->ch_units[i].channels[ch].gain_data_hist[0][0];
180 ctx->ch_units[i].channels[ch].gain_data_prev = &ctx->ch_units[i].channels[ch].gain_data_hist[1][0];
181 ctx->ch_units[i].channels[ch].tones_info = &ctx->ch_units[i].channels[ch].tones_info_hist[0][0];
182 ctx->ch_units[i].channels[ch].tones_info_prev = &ctx->ch_units[i].channels[ch].tones_info_hist[1][0];
185 ctx->ch_units[i].waves_info = &ctx->ch_units[i].wave_synth_hist[0];
186 ctx->ch_units[i].waves_info_prev = &ctx->ch_units[i].wave_synth_hist[1];
189 avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
194 static void decode_residual_spectrum(Atrac3pChanUnitCtx *ctx,
195 float out[2][ATRAC3P_FRAME_SAMPLES],
197 AVCodecContext *avctx)
199 int i, sb, ch, qu, nspeclines, RNG_index;
202 /* calculate RNG table index for each subband */
203 int sb_RNG_index[ATRAC3P_SUBBANDS] = { 0 };
205 if (ctx->mute_flag) {
206 for (ch = 0; ch < num_channels; ch++)
207 memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch]));
211 for (qu = 0, RNG_index = 0; qu < ctx->used_quant_units; qu++)
212 RNG_index += ctx->channels[0].qu_sf_idx[qu] +
213 ctx->channels[1].qu_sf_idx[qu];
215 for (sb = 0; sb < ctx->num_coded_subbands; sb++, RNG_index += 128)
216 sb_RNG_index[sb] = RNG_index & 0x3FC;
218 /* inverse quant and power compensation */
219 for (ch = 0; ch < num_channels; ch++) {
220 /* clear channel's residual spectrum */
221 memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch]));
223 for (qu = 0; qu < ctx->used_quant_units; qu++) {
224 src = &ctx->channels[ch].spectrum[ff_atrac3p_qu_to_spec_pos[qu]];
225 dst = &out[ch][ff_atrac3p_qu_to_spec_pos[qu]];
226 nspeclines = ff_atrac3p_qu_to_spec_pos[qu + 1] -
227 ff_atrac3p_qu_to_spec_pos[qu];
229 if (ctx->channels[ch].qu_wordlen[qu] > 0) {
230 q = ff_atrac3p_sf_tab[ctx->channels[ch].qu_sf_idx[qu]] *
231 ff_atrac3p_mant_tab[ctx->channels[ch].qu_wordlen[qu]];
232 for (i = 0; i < nspeclines; i++)
237 for (sb = 0; sb < ctx->num_coded_subbands; sb++)
238 ff_atrac3p_power_compensation(ctx, ch, &out[ch][0],
239 sb_RNG_index[sb], sb);
242 if (ctx->unit_type == CH_UNIT_STEREO) {
243 for (sb = 0; sb < ctx->num_coded_subbands; sb++) {
244 if (ctx->swap_channels[sb]) {
245 for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++)
246 FFSWAP(float, out[0][sb * ATRAC3P_SUBBAND_SAMPLES + i],
247 out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]);
250 /* flip coefficients' sign if requested */
251 if (ctx->negate_coeffs[sb])
252 for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++)
253 out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i] = -(out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]);
258 static void reconstruct_frame(ATRAC3PContext *ctx, Atrac3pChanUnitCtx *ch_unit,
259 int num_channels, AVCodecContext *avctx)
263 for (ch = 0; ch < num_channels; ch++) {
264 for (sb = 0; sb < ch_unit->num_subbands; sb++) {
265 /* inverse transform and windowing */
266 ff_atrac3p_imdct(&ctx->fdsp, &ctx->mdct_ctx,
267 &ctx->samples[ch][sb * ATRAC3P_SUBBAND_SAMPLES],
268 &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES],
269 (ch_unit->channels[ch].wnd_shape_prev[sb] << 1) +
270 ch_unit->channels[ch].wnd_shape[sb], sb);
272 /* gain compensation and overlapping */
273 ff_atrac_gain_compensation(&ctx->gainc_ctx,
274 &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES],
275 &ch_unit->prev_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES],
276 &ch_unit->channels[ch].gain_data_prev[sb],
277 &ch_unit->channels[ch].gain_data[sb],
278 ATRAC3P_SUBBAND_SAMPLES,
279 &ctx->time_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES]);
282 /* zero unused subbands in both output and overlapping buffers */
283 memset(&ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES],
285 (ATRAC3P_SUBBANDS - ch_unit->num_subbands) *
286 ATRAC3P_SUBBAND_SAMPLES *
287 sizeof(ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES]));
288 memset(&ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES],
290 (ATRAC3P_SUBBANDS - ch_unit->num_subbands) *
291 ATRAC3P_SUBBAND_SAMPLES *
292 sizeof(ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES]));
294 /* resynthesize and add tonal signal */
295 if (ch_unit->waves_info->tones_present ||
296 ch_unit->waves_info_prev->tones_present) {
297 for (sb = 0; sb < ch_unit->num_subbands; sb++)
298 if (ch_unit->channels[ch].tones_info[sb].num_wavs ||
299 ch_unit->channels[ch].tones_info_prev[sb].num_wavs) {
300 ff_atrac3p_generate_tones(ch_unit, &ctx->fdsp, ch, sb,
301 &ctx->time_buf[ch][sb * 128]);
305 /* subband synthesis and acoustic signal output */
306 ff_atrac3p_ipqf(&ctx->ipqf_dct_ctx, &ch_unit->ipqf_ctx[ch],
307 &ctx->time_buf[ch][0], &ctx->outp_buf[ch][0]);
310 /* swap window shape and gain control buffers. */
311 for (ch = 0; ch < num_channels; ch++) {
312 FFSWAP(uint8_t *, ch_unit->channels[ch].wnd_shape,
313 ch_unit->channels[ch].wnd_shape_prev);
314 FFSWAP(AtracGainInfo *, ch_unit->channels[ch].gain_data,
315 ch_unit->channels[ch].gain_data_prev);
316 FFSWAP(Atrac3pWavesData *, ch_unit->channels[ch].tones_info,
317 ch_unit->channels[ch].tones_info_prev);
320 FFSWAP(Atrac3pWaveSynthParams *, ch_unit->waves_info, ch_unit->waves_info_prev);
323 static int atrac3p_decode_frame(AVCodecContext *avctx, void *data,
324 int *got_frame_ptr, AVPacket *avpkt)
326 ATRAC3PContext *ctx = avctx->priv_data;
327 AVFrame *frame = data;
328 int i, ret, ch_unit_id, ch_block = 0, out_ch_index = 0, channels_to_process;
329 float **samples_p = (float **)frame->extended_data;
331 frame->nb_samples = ATRAC3P_FRAME_SAMPLES;
332 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
333 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
337 if ((ret = init_get_bits8(&ctx->gb, avpkt->data, avpkt->size)) < 0)
340 if (get_bits1(&ctx->gb)) {
341 av_log(avctx, AV_LOG_ERROR, "Invalid start bit!\n");
342 return AVERROR_INVALIDDATA;
345 while (get_bits_left(&ctx->gb) >= 2 &&
346 (ch_unit_id = get_bits(&ctx->gb, 2)) != CH_UNIT_TERMINATOR) {
347 if (ch_unit_id == CH_UNIT_EXTENSION) {
348 avpriv_report_missing_feature(avctx, "Channel unit extension");
349 return AVERROR_PATCHWELCOME;
351 if (ch_block >= ctx->num_channel_blocks ||
352 ctx->channel_blocks[ch_block] != ch_unit_id) {
353 av_log(avctx, AV_LOG_ERROR,
354 "Frame data doesn't match channel configuration!\n");
355 return AVERROR_INVALIDDATA;
358 ctx->ch_units[ch_block].unit_type = ch_unit_id;
359 channels_to_process = ch_unit_id + 1;
361 if ((ret = ff_atrac3p_decode_channel_unit(&ctx->gb,
362 &ctx->ch_units[ch_block],
367 decode_residual_spectrum(&ctx->ch_units[ch_block], ctx->samples,
368 channels_to_process, avctx);
369 reconstruct_frame(ctx, &ctx->ch_units[ch_block],
370 channels_to_process, avctx);
372 for (i = 0; i < channels_to_process; i++)
373 memcpy(samples_p[out_ch_index + i], ctx->outp_buf[i],
374 ATRAC3P_FRAME_SAMPLES * sizeof(**samples_p));
377 out_ch_index += channels_to_process;
382 return avctx->block_align;
385 AVCodec ff_atrac3p_decoder = {
386 .name = "atrac3plus",
387 .long_name = NULL_IF_CONFIG_SMALL("ATRAC3+ (Adaptive TRansform Acoustic Coding 3+)"),
388 .type = AVMEDIA_TYPE_AUDIO,
389 .id = AV_CODEC_ID_ATRAC3P,
390 .capabilities = CODEC_CAP_DR1,
391 .priv_data_size = sizeof(ATRAC3PContext),
392 .init = atrac3p_decode_init,
393 .init_static_data = ff_atrac3p_init_vlcs,
394 .close = atrac3p_decode_close,
395 .decode = atrac3p_decode_frame,