2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnik
5 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
6 * of DV technical info.
9 * Copyright (c) 2002 Fabrice Bellard
11 * 50 Mbps (DVCPRO50) and 100 Mbps (DVCPRO HD) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
13 * Funded by BBC Research & Development
15 * This file is part of FFmpeg.
17 * FFmpeg is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU Lesser General Public
19 * License as published by the Free Software Foundation; either
20 * version 2.1 of the License, or (at your option) any later version.
22 * FFmpeg is distributed in the hope that it will be useful,
23 * but WITHOUT ANY WARRANTY; without even the implied warranty of
24 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
25 * Lesser General Public License for more details.
27 * You should have received a copy of the GNU Lesser General Public
28 * License along with FFmpeg; if not, write to the Free Software
29 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
34 #include "libavcodec/dv_profile.h"
35 #include "libavcodec/dvdata.h"
36 #include "libavutil/channel_layout.h"
37 #include "libavutil/intreadwrite.h"
38 #include "libavutil/mathematics.h"
39 #include "libavutil/timecode.h"
41 #include "libavutil/avassert.h"
43 struct DVDemuxContext {
44 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
45 AVFormatContext* fctx;
48 AVPacket audio_pkt[4];
49 uint8_t audio_buf[4][8192];
55 static inline uint16_t dv_audio_12to16(uint16_t sample)
57 uint16_t shift, result;
59 sample = (sample < 0x800) ? sample : sample | 0xf000;
60 shift = (sample & 0xf00) >> 8;
62 if (shift < 0x2 || shift > 0xd) {
64 } else if (shift < 0x8) {
66 result = (sample - (256 * shift)) << shift;
69 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
76 * This is the dumbest implementation of all -- it simply looks at
77 * a fixed offset and if pack isn't there -- fails. We might want
78 * to have a fallback mechanism for complete search of missing packs.
80 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
86 offs = (80*6 + 80*16*3 + 3);
88 case dv_audio_control:
89 offs = (80*6 + 80*16*4 + 3);
91 case dv_video_control:
92 offs = (80*5 + 48 + 5);
95 offs = (80*1 + 3 + 3);
101 return frame[offs] == t ? &frame[offs] : NULL;
104 static const int dv_audio_frequency[3] = {
109 * There's a couple of assumptions being made here:
110 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
111 * We can pass them upwards when libavcodec will be ready to deal with them.
112 * 2. We don't do software emphasis.
113 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
114 * are converted into 16bit linear ones.
116 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
117 const DVprofile *sys)
119 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
121 const uint8_t* as_pack;
124 as_pack = dv_extract_pack(frame, dv_audio_source);
125 if (!as_pack) /* No audio ? */
128 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
129 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
130 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
133 return -1; /* unsupported quantization */
135 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
136 return AVERROR_INVALIDDATA;
138 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
139 half_ch = sys->difseg_size / 2;
141 /* We work with 720p frames split in half, thus even frames have
142 * channels 0,1 and odd 2,3. */
143 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
145 if (ipcm + sys->n_difchan > (quant == 1 ? 2 : 4)) {
146 av_log(NULL, AV_LOG_ERROR, "too many dv pcm frames\n");
147 return AVERROR_INVALIDDATA;
150 /* for each DIF channel */
151 for (chan = 0; chan < sys->n_difchan; chan++) {
157 /* for each DIF segment */
158 for (i = 0; i < sys->difseg_size; i++) {
159 frame += 6 * 80; /* skip DIF segment header */
160 if (quant == 1 && i == half_ch) {
161 /* next stereo channel (12bit mode only) */
168 /* for each AV sequence */
169 for (j = 0; j < 9; j++) {
170 for (d = 8; d < 80; d += 2) {
171 if (quant == 0) { /* 16bit quantization */
172 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
176 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
177 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
178 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
180 } else { /* 12bit quantization */
181 lc = ((uint16_t)frame[d] << 4) |
182 ((uint16_t)frame[d+2] >> 4);
183 rc = ((uint16_t)frame[d+1] << 4) |
184 ((uint16_t)frame[d+2] & 0x0f);
185 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
186 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
188 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
192 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
193 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
194 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
195 (d - 8) / 3 * sys->audio_stride;
196 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
197 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
202 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
210 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
212 const uint8_t* as_pack;
213 int freq, stype, smpls, quant, i, ach;
215 as_pack = dv_extract_pack(frame, dv_audio_source);
216 if (!as_pack || !c->sys) { /* No audio ? */
221 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
222 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
223 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
224 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
226 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
227 av_log(c->fctx, AV_LOG_ERROR,
228 "Unrecognized audio sample rate index (%d)\n", freq);
233 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
238 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
239 ach = ((int[4]){ 1, 0, 2, 4})[stype];
240 if (ach == 1 && quant && freq == 2)
243 /* Dynamic handling of the audio streams in DV */
244 for (i = 0; i < ach; i++) {
246 c->ast[i] = avformat_new_stream(c->fctx, NULL);
249 avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
250 c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
251 c->ast[i]->codec->codec_id = AV_CODEC_ID_PCM_S16LE;
253 av_init_packet(&c->audio_pkt[i]);
254 c->audio_pkt[i].size = 0;
255 c->audio_pkt[i].data = c->audio_buf[i];
256 c->audio_pkt[i].stream_index = c->ast[i]->index;
257 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
259 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
260 c->ast[i]->codec->channels = 2;
261 c->ast[i]->codec->channel_layout = AV_CH_LAYOUT_STEREO;
262 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
263 c->ast[i]->start_time = 0;
267 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
270 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
272 const uint8_t* vsc_pack;
273 AVCodecContext* avctx;
278 avctx = c->vst->codec;
280 avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
281 c->sys->time_base.den);
282 avctx->time_base= c->sys->time_base;
284 /* finding out SAR is a little bit messy */
285 vsc_pack = dv_extract_pack(frame, dv_video_control);
286 apt = frame[4] & 0x07;
287 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
288 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
289 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
290 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
292 size = c->sys->frame_size;
297 static int dv_extract_timecode(DVDemuxContext* c, uint8_t* frame, char *tc)
299 const uint8_t *tc_pack;
301 // For PAL systems, drop frame bit is replaced by an arbitrary
302 // bit so its value should not be considered. Drop frame timecode
303 // is only relevant for NTSC systems.
304 int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
306 tc_pack = dv_extract_pack(frame, dv_timecode);
309 av_timecode_make_smpte_tc_string(tc, AV_RB32(tc_pack + 1), prevent_df);
314 * The following 3 functions constitute our interface to the world
317 DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
321 c = av_mallocz(sizeof(DVDemuxContext));
325 c->vst = avformat_new_stream(s, NULL);
332 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
333 c->vst->codec->codec_id = AV_CODEC_ID_DVVIDEO;
334 c->vst->codec->bit_rate = 25000000;
335 c->vst->start_time = 0;
340 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
345 for (i = 0; i < c->ach; i++) {
346 if (c->ast[i] && c->audio_pkt[i].size) {
347 *pkt = c->audio_pkt[i];
348 c->audio_pkt[i].size = 0;
357 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
358 uint8_t* buf, int buf_size, int64_t pos)
361 uint8_t *ppcm[4] = {0};
363 if (buf_size < DV_PROFILE_BYTES ||
364 !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
365 buf_size < c->sys->frame_size) {
366 return -1; /* Broken frame, or not enough data */
369 /* Queueing audio packet */
370 /* FIXME: in case of no audio/bad audio we have to do something */
371 size = dv_extract_audio_info(c, buf);
372 for (i = 0; i < c->ach; i++) {
373 c->audio_pkt[i].pos = pos;
374 c->audio_pkt[i].size = size;
375 c->audio_pkt[i].pts = c->abytes * 30000 * 8 / c->ast[i]->codec->bit_rate;
376 ppcm[i] = c->audio_buf[i];
379 dv_extract_audio(buf, ppcm, c->sys);
381 /* We work with 720p frames split in half, thus even frames have
382 * channels 0,1 and odd 2,3. */
383 if (c->sys->height == 720) {
385 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
387 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
394 /* Now it's time to return video packet */
395 size = dv_extract_video_info(c, buf);
400 pkt->flags |= AV_PKT_FLAG_KEY;
401 pkt->stream_index = c->vst->index;
402 pkt->pts = c->frames;
409 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
410 int64_t timestamp, int flags)
412 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
413 const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
415 int64_t size = avio_size(s->pb) - s->data_offset;
416 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
418 offset = sys->frame_size * timestamp;
420 if (size >= 0 && offset > max_offset) offset = max_offset;
421 else if (offset < 0) offset = 0;
423 return offset + s->data_offset;
426 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
428 c->frames= frame_offset;
431 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
432 (AVRational){8, c->ast[0]->codec->bit_rate});
434 av_log(c->fctx, AV_LOG_ERROR, "cannot adjust audio bytes\n");
436 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
437 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
440 /************************************************************
441 * Implementation of the easiest DV storage of all -- raw DV.
442 ************************************************************/
444 typedef struct RawDVContext {
445 DVDemuxContext* dv_demux;
446 uint8_t buf[DV_MAX_FRAME_SIZE];
449 static int dv_read_timecode(AVFormatContext *s) {
451 char timecode[AV_TIMECODE_STR_SIZE];
452 int64_t pos = avio_tell(s->pb);
454 // Read 3 DIF blocks: Header block and 2 Subcode blocks.
455 int partial_frame_size = 3 * 80;
456 uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
459 RawDVContext *c = s->priv_data;
460 ret = avio_read(s->pb, partial_frame, partial_frame_size);
464 if (ret < partial_frame_size) {
469 ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
471 av_dict_set(&s->metadata, "timecode", timecode, 0);
473 av_log(s, AV_LOG_ERROR, "Detected timecode is invalid\n");
476 av_free(partial_frame);
477 avio_seek(s->pb, pos, SEEK_SET);
481 static int dv_read_header(AVFormatContext *s)
483 unsigned state, marker_pos = 0;
484 RawDVContext *c = s->priv_data;
486 c->dv_demux = avpriv_dv_init_demux(s);
490 state = avio_rb32(s->pb);
491 while ((state & 0xffffff7f) != 0x1f07003f) {
492 if (url_feof(s->pb)) {
493 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
496 if (state == 0x003f0700 || state == 0xff3f0700)
497 marker_pos = avio_tell(s->pb);
498 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
499 avio_seek(s->pb, -163, SEEK_CUR);
500 state = avio_rb32(s->pb);
503 state = (state << 8) | avio_r8(s->pb);
505 AV_WB32(c->buf, state);
507 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) != DV_PROFILE_BYTES - 4 ||
508 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
511 c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
512 if (!c->dv_demux->sys) {
513 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
517 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
518 c->dv_demux->sys->time_base);
527 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
530 RawDVContext *c = s->priv_data;
532 size = avpriv_dv_get_packet(c->dv_demux, pkt);
535 int64_t pos = avio_tell(s->pb);
536 if (!c->dv_demux->sys)
538 size = c->dv_demux->sys->frame_size;
539 if (avio_read(s->pb, c->buf, size) <= 0)
542 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
548 static int dv_read_seek(AVFormatContext *s, int stream_index,
549 int64_t timestamp, int flags)
551 RawDVContext *r = s->priv_data;
552 DVDemuxContext *c = r->dv_demux;
553 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
555 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
558 ff_dv_offset_reset(c, offset / c->sys->frame_size);
562 static int dv_read_close(AVFormatContext *s)
564 RawDVContext *c = s->priv_data;
565 av_free(c->dv_demux);
569 static int dv_probe(AVProbeData *p)
571 unsigned state, marker_pos = 0;
574 int secondary_matches = 0;
579 state = AV_RB32(p->buf);
580 for (i = 4; i < p->buf_size; i++) {
581 if ((state & 0xffffff7f) == 0x1f07003f)
583 // any section header, also with seq/chan num != 0,
584 // should appear around every 12000 bytes, at least 10 per frame
585 if ((state & 0xff07ff7f) == 0x1f07003f)
587 if (state == 0x003f0700 || state == 0xff3f0700)
589 if (state == 0xff3f0701 && i - marker_pos == 80)
591 state = (state << 8) | p->buf[i];
594 if (matches && p->buf_size / matches < 1024*1024) {
595 if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
596 return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
597 return AVPROBE_SCORE_MAX/4;
602 #if CONFIG_DV_DEMUXER
603 AVInputFormat ff_dv_demuxer = {
605 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
606 .priv_data_size = sizeof(RawDVContext),
607 .read_probe = dv_probe,
608 .read_header = dv_read_header,
609 .read_packet = dv_read_packet,
610 .read_close = dv_read_close,
611 .read_seek = dv_read_seek,
612 .extensions = "dv,dif",