1 /*
2 * Bluetooth low-complexity, subband codec (SBC)
3 *
4 * Copyright (C) 2017 Aurelien Jacobs <aurel@gnuage.org>
5 * Copyright (C) 2012-2013 Intel Corporation
6 * Copyright (C) 2008-2010 Nokia Corporation
7 * Copyright (C) 2004-2010 Marcel Holtmann <marcel@holtmann.org>
8 * Copyright (C) 2004-2005 Henryk Ploetz <henryk@ploetzli.ch>
9 * Copyright (C) 2005-2008 Brad Midgley <bmidgley@xmission.com>
10 *
11 * This file is part of FFmpeg.
12 *
13 * FFmpeg is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU Lesser General Public
15 * License as published by the Free Software Foundation; either
16 * version 2.1 of the License, or (at your option) any later version.
17 *
18 * FFmpeg is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * Lesser General Public License for more details.
22 *
23 * You should have received a copy of the GNU Lesser General Public
24 * License along with FFmpeg; if not, write to the Free Software
25 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 */
27
28 /**
29 * @file
30 * SBC encoder implementation
31 */
32
33 #include "libavutil/channel_layout.h"
34 #include "libavutil/opt.h"
35 #include "avcodec.h"
36 #include "codec_internal.h"
37 #include "encode.h"
38 #include "profiles.h"
39 #include "put_bits.h"
40 #include "sbc.h"
41 #include "sbcdsp.h"
42
43 typedef struct SBCEncContext {
44 AVClass *class;
45 int64_t max_delay;
46 int msbc;
47 DECLARE_ALIGNED(SBC_ALIGN, struct sbc_frame, frame);
48 DECLARE_ALIGNED(SBC_ALIGN, SBCDSPContext, dsp);
49 } SBCEncContext;
50
sbc_analyze_audio(SBCDSPContext * s,struct sbc_frame * frame)51 static int sbc_analyze_audio(SBCDSPContext *s, struct sbc_frame *frame)
52 {
53 int ch, blk;
54 int16_t *x;
55
56 switch (frame->subbands) {
57 case 4:
58 for (ch = 0; ch < frame->channels; ch++) {
59 x = &s->X[ch][s->position - 4 *
60 s->increment + frame->blocks * 4];
61 for (blk = 0; blk < frame->blocks;
62 blk += s->increment) {
63 s->sbc_analyze_4s(
64 s, x,
65 frame->sb_sample_f[blk][ch],
66 frame->sb_sample_f[blk + 1][ch] -
67 frame->sb_sample_f[blk][ch]);
68 x -= 4 * s->increment;
69 }
70 }
71 return frame->blocks * 4;
72
73 case 8:
74 for (ch = 0; ch < frame->channels; ch++) {
75 x = &s->X[ch][s->position - 8 *
76 s->increment + frame->blocks * 8];
77 for (blk = 0; blk < frame->blocks;
78 blk += s->increment) {
79 s->sbc_analyze_8s(
80 s, x,
81 frame->sb_sample_f[blk][ch],
82 frame->sb_sample_f[blk + 1][ch] -
83 frame->sb_sample_f[blk][ch]);
84 x -= 8 * s->increment;
85 }
86 }
87 return frame->blocks * 8;
88
89 default:
90 return AVERROR(EIO);
91 }
92 }
93
94 /*
95 * Packs the SBC frame from frame into the memory in avpkt.
96 * Returns the length of the packed frame.
97 */
sbc_pack_frame(AVPacket * avpkt,struct sbc_frame * frame,int joint,int msbc)98 static size_t sbc_pack_frame(AVPacket *avpkt, struct sbc_frame *frame,
99 int joint, int msbc)
100 {
101 PutBitContext pb;
102
103 /* Will copy the header parts for CRC-8 calculation here */
104 uint8_t crc_header[11] = { 0 };
105 int crc_pos;
106
107 uint32_t audio_sample;
108
109 int ch, sb, blk; /* channel, subband, block and bit counters */
110 int bits[2][8]; /* bits distribution */
111 uint32_t levels[2][8]; /* levels are derived from that */
112 uint32_t sb_sample_delta[2][8];
113
114 if (msbc) {
115 avpkt->data[0] = MSBC_SYNCWORD;
116 avpkt->data[1] = 0;
117 avpkt->data[2] = 0;
118 } else {
119 avpkt->data[0] = SBC_SYNCWORD;
120
121 avpkt->data[1] = (frame->frequency & 0x03) << 6;
122 avpkt->data[1] |= (((frame->blocks >> 2) - 1) & 0x03) << 4;
123 avpkt->data[1] |= (frame->mode & 0x03) << 2;
124 avpkt->data[1] |= (frame->allocation & 0x01) << 1;
125 avpkt->data[1] |= ((frame->subbands == 8) & 0x01) << 0;
126
127 avpkt->data[2] = frame->bitpool;
128
129 if (frame->bitpool > frame->subbands << (4 + (frame->mode == STEREO
130 || frame->mode == JOINT_STEREO)))
131 return -5;
132 }
133
134 /* Can't fill in crc yet */
135 crc_header[0] = avpkt->data[1];
136 crc_header[1] = avpkt->data[2];
137 crc_pos = 16;
138
139 init_put_bits(&pb, avpkt->data + 4, avpkt->size);
140
141 if (frame->mode == JOINT_STEREO) {
142 put_bits(&pb, frame->subbands, joint);
143 crc_header[crc_pos >> 3] = joint;
144 crc_pos += frame->subbands;
145 }
146
147 for (ch = 0; ch < frame->channels; ch++) {
148 for (sb = 0; sb < frame->subbands; sb++) {
149 put_bits(&pb, 4, frame->scale_factor[ch][sb] & 0x0F);
150 crc_header[crc_pos >> 3] <<= 4;
151 crc_header[crc_pos >> 3] |= frame->scale_factor[ch][sb] & 0x0F;
152 crc_pos += 4;
153 }
154 }
155
156 /* align the last crc byte */
157 if (crc_pos % 8)
158 crc_header[crc_pos >> 3] <<= 8 - (crc_pos % 8);
159
160 avpkt->data[3] = ff_sbc_crc8(frame->crc_ctx, crc_header, crc_pos);
161
162 ff_sbc_calculate_bits(frame, bits);
163
164 for (ch = 0; ch < frame->channels; ch++) {
165 for (sb = 0; sb < frame->subbands; sb++) {
166 levels[ch][sb] = ((1 << bits[ch][sb]) - 1) <<
167 (32 - (frame->scale_factor[ch][sb] +
168 SCALE_OUT_BITS + 2));
169 sb_sample_delta[ch][sb] = (uint32_t) 1 <<
170 (frame->scale_factor[ch][sb] +
171 SCALE_OUT_BITS + 1);
172 }
173 }
174
175 for (blk = 0; blk < frame->blocks; blk++) {
176 for (ch = 0; ch < frame->channels; ch++) {
177 for (sb = 0; sb < frame->subbands; sb++) {
178
179 if (bits[ch][sb] == 0)
180 continue;
181
182 audio_sample = ((uint64_t) levels[ch][sb] *
183 (sb_sample_delta[ch][sb] +
184 frame->sb_sample_f[blk][ch][sb])) >> 32;
185
186 put_bits(&pb, bits[ch][sb], audio_sample);
187 }
188 }
189 }
190
191 flush_put_bits(&pb);
192
193 return put_bytes_output(&pb);
194 }
195
sbc_encode_init(AVCodecContext * avctx)196 static int sbc_encode_init(AVCodecContext *avctx)
197 {
198 SBCEncContext *sbc = avctx->priv_data;
199 struct sbc_frame *frame = &sbc->frame;
200
201 if (avctx->profile == FF_PROFILE_SBC_MSBC)
202 sbc->msbc = 1;
203
204 if (sbc->msbc) {
205 if (avctx->ch_layout.nb_channels != 1) {
206 av_log(avctx, AV_LOG_ERROR, "mSBC require mono channel.\n");
207 return AVERROR(EINVAL);
208 }
209
210 if (avctx->sample_rate != 16000) {
211 av_log(avctx, AV_LOG_ERROR, "mSBC require 16 kHz samplerate.\n");
212 return AVERROR(EINVAL);
213 }
214
215 frame->mode = SBC_MODE_MONO;
216 frame->subbands = 8;
217 frame->blocks = MSBC_BLOCKS;
218 frame->allocation = SBC_AM_LOUDNESS;
219 frame->bitpool = 26;
220
221 avctx->frame_size = 8 * MSBC_BLOCKS;
222 } else {
223 int d;
224
225 if (avctx->global_quality > 255*FF_QP2LAMBDA) {
226 av_log(avctx, AV_LOG_ERROR, "bitpool > 255 is not allowed.\n");
227 return AVERROR(EINVAL);
228 }
229
230 if (avctx->ch_layout.nb_channels == 1) {
231 frame->mode = SBC_MODE_MONO;
232 if (sbc->max_delay <= 3000 || avctx->bit_rate > 270000)
233 frame->subbands = 4;
234 else
235 frame->subbands = 8;
236 } else {
237 if (avctx->bit_rate < 180000 || avctx->bit_rate > 420000)
238 frame->mode = SBC_MODE_JOINT_STEREO;
239 else
240 frame->mode = SBC_MODE_STEREO;
241 if (sbc->max_delay <= 4000 || avctx->bit_rate > 420000)
242 frame->subbands = 4;
243 else
244 frame->subbands = 8;
245 }
246 /* sbc algorithmic delay is ((blocks + 10) * subbands - 2) / sample_rate */
247 frame->blocks = av_clip(((sbc->max_delay * avctx->sample_rate + 2)
248 / (1000000 * frame->subbands)) - 10, 4, 16) & ~3;
249
250 frame->allocation = SBC_AM_LOUDNESS;
251
252 d = frame->blocks * ((frame->mode == SBC_MODE_DUAL_CHANNEL) + 1);
253 frame->bitpool = (((avctx->bit_rate * frame->subbands * frame->blocks) / avctx->sample_rate)
254 - 4 * frame->subbands * avctx->ch_layout.nb_channels
255 - (frame->mode == SBC_MODE_JOINT_STEREO)*frame->subbands - 32 + d/2) / d;
256 if (avctx->global_quality > 0)
257 frame->bitpool = avctx->global_quality / FF_QP2LAMBDA;
258
259 avctx->frame_size = 4*((frame->subbands >> 3) + 1) * 4*(frame->blocks >> 2);
260 }
261
262 for (int i = 0; avctx->codec->supported_samplerates[i]; i++)
263 if (avctx->sample_rate == avctx->codec->supported_samplerates[i])
264 frame->frequency = i;
265
266 frame->channels = avctx->ch_layout.nb_channels;
267 frame->codesize = frame->subbands * frame->blocks * avctx->ch_layout.nb_channels * 2;
268 frame->crc_ctx = av_crc_get_table(AV_CRC_8_EBU);
269
270 memset(&sbc->dsp.X, 0, sizeof(sbc->dsp.X));
271 sbc->dsp.position = (SBC_X_BUFFER_SIZE - frame->subbands * 9) & ~7;
272 sbc->dsp.increment = sbc->msbc ? 1 : 4;
273 ff_sbcdsp_init(&sbc->dsp);
274
275 return 0;
276 }
277
sbc_encode_frame(AVCodecContext * avctx,AVPacket * avpkt,const AVFrame * av_frame,int * got_packet_ptr)278 static int sbc_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
279 const AVFrame *av_frame, int *got_packet_ptr)
280 {
281 SBCEncContext *sbc = avctx->priv_data;
282 struct sbc_frame *frame = &sbc->frame;
283 uint8_t joint = frame->mode == SBC_MODE_JOINT_STEREO;
284 uint8_t dual = frame->mode == SBC_MODE_DUAL_CHANNEL;
285 int ret, j = 0;
286
287 int frame_length = 4 + (4 * frame->subbands * frame->channels) / 8
288 + ((frame->blocks * frame->bitpool * (1 + dual)
289 + joint * frame->subbands) + 7) / 8;
290
291 /* input must be large enough to encode a complete frame */
292 if (av_frame->nb_samples * frame->channels * 2 < frame->codesize)
293 return 0;
294
295 if ((ret = ff_get_encode_buffer(avctx, avpkt, frame_length, 0)) < 0)
296 return ret;
297
298 /* Select the needed input data processing function and call it */
299 if (frame->subbands == 8)
300 sbc->dsp.position = sbc->dsp.sbc_enc_process_input_8s(
301 sbc->dsp.position, av_frame->data[0], sbc->dsp.X,
302 frame->subbands * frame->blocks, frame->channels);
303 else
304 sbc->dsp.position = sbc->dsp.sbc_enc_process_input_4s(
305 sbc->dsp.position, av_frame->data[0], sbc->dsp.X,
306 frame->subbands * frame->blocks, frame->channels);
307
308 sbc_analyze_audio(&sbc->dsp, &sbc->frame);
309
310 if (frame->mode == JOINT_STEREO)
311 j = sbc->dsp.sbc_calc_scalefactors_j(frame->sb_sample_f,
312 frame->scale_factor,
313 frame->blocks,
314 frame->subbands);
315 else
316 sbc->dsp.sbc_calc_scalefactors(frame->sb_sample_f,
317 frame->scale_factor,
318 frame->blocks,
319 frame->channels,
320 frame->subbands);
321 emms_c();
322 sbc_pack_frame(avpkt, frame, j, sbc->msbc);
323
324 *got_packet_ptr = 1;
325 return 0;
326 }
327
328 #define OFFSET(x) offsetof(SBCEncContext, x)
329 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
330 static const AVOption options[] = {
331 { "sbc_delay", "set maximum algorithmic latency",
332 OFFSET(max_delay), AV_OPT_TYPE_DURATION, {.i64 = 13000}, 1000,13000, AE },
333 { "msbc", "use mSBC mode (wideband speech mono SBC)",
334 OFFSET(msbc), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AE },
335 FF_AVCTX_PROFILE_OPTION("msbc", NULL, AUDIO, FF_PROFILE_SBC_MSBC)
336 { NULL },
337 };
338
339 static const AVClass sbc_class = {
340 .class_name = "sbc encoder",
341 .item_name = av_default_item_name,
342 .option = options,
343 .version = LIBAVUTIL_VERSION_INT,
344 };
345
346 const FFCodec ff_sbc_encoder = {
347 .p.name = "sbc",
348 .p.long_name = NULL_IF_CONFIG_SMALL("SBC (low-complexity subband codec)"),
349 .p.type = AVMEDIA_TYPE_AUDIO,
350 .p.id = AV_CODEC_ID_SBC,
351 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_SMALL_LAST_FRAME,
352 .priv_data_size = sizeof(SBCEncContext),
353 .init = sbc_encode_init,
354 FF_CODEC_ENCODE_CB(sbc_encode_frame),
355 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE,
356 #if FF_API_OLD_CHANNEL_LAYOUT
357 .p.channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
358 AV_CH_LAYOUT_STEREO, 0},
359 #endif
360 .p.ch_layouts = (const AVChannelLayout[]) { AV_CHANNEL_LAYOUT_MONO,
361 AV_CHANNEL_LAYOUT_STEREO,
362 { 0 } },
363 .p.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16,
364 AV_SAMPLE_FMT_NONE },
365 .p.supported_samplerates = (const int[]) { 16000, 32000, 44100, 48000, 0 },
366 .p.priv_class = &sbc_class,
367 .p.profiles = NULL_IF_CONFIG_SMALL(ff_sbc_profiles),
368 };
369