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/dv.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 AVDVProfile* 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;
75 static const uint8_t *dv_extract_pack(const uint8_t *frame, enum dv_pack_type t)
80 for (c = 0; c < 10; c++) {
83 if (c&1) offs = (80 * 6 + 80 * 16 * 0 + 3 + c*12000);
84 else offs = (80 * 6 + 80 * 16 * 3 + 3 + c*12000);
86 case dv_audio_control:
87 if (c&1) offs = (80 * 6 + 80 * 16 * 1 + 3 + c*12000);
88 else offs = (80 * 6 + 80 * 16 * 4 + 3 + c*12000);
90 case dv_video_control:
91 if (c&1) offs = (80 * 3 + 8 + c*12000);
92 else offs = (80 * 5 + 48 + 5 + c*12000);
95 offs = (80*1 + 3 + 3);
100 if (frame[offs] == t)
104 return frame[offs] == t ? &frame[offs] : NULL;
107 static const int dv_audio_frequency[3] = {
112 * There's a couple of assumptions being made here:
113 * 1. By default we silence erroneous (0x8000/16-bit 0x800/12-bit) audio samples.
114 * We can pass them upwards when libavcodec will be ready to deal with them.
115 * 2. We don't do software emphasis.
116 * 3. Audio is always returned as 16-bit linear samples: 12-bit nonlinear samples
117 * are converted into 16-bit linear ones.
119 static int dv_extract_audio(const uint8_t *frame, uint8_t **ppcm,
120 const AVDVProfile *sys)
122 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
124 const uint8_t *as_pack;
127 as_pack = dv_extract_pack(frame, dv_audio_source);
128 if (!as_pack) /* No audio ? */
131 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
132 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
133 quant = as_pack[4] & 0x07; /* 0 - 16-bit linear, 1 - 12-bit nonlinear */
136 return -1; /* unsupported quantization */
138 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
139 return AVERROR_INVALIDDATA;
141 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
142 half_ch = sys->difseg_size / 2;
144 /* We work with 720p frames split in half, thus even frames have
145 * channels 0,1 and odd 2,3. */
146 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
148 if (ipcm + sys->n_difchan > (quant == 1 ? 2 : 4)) {
149 av_log(NULL, AV_LOG_ERROR, "too many dv pcm frames\n");
150 return AVERROR_INVALIDDATA;
153 /* for each DIF channel */
154 for (chan = 0; chan < sys->n_difchan; chan++) {
160 /* for each DIF segment */
161 for (i = 0; i < sys->difseg_size; i++) {
162 frame += 6 * 80; /* skip DIF segment header */
163 if (quant == 1 && i == half_ch) {
164 /* next stereo channel (12-bit mode only) */
171 /* for each AV sequence */
172 for (j = 0; j < 9; j++) {
173 for (d = 8; d < 80; d += 2) {
174 if (quant == 0) { /* 16-bit quantization */
175 of = sys->audio_shuffle[i][j] +
176 (d - 8) / 2 * sys->audio_stride;
180 /* FIXME: maybe we have to admit that DV is a
182 pcm[of * 2] = frame[d + 1];
183 pcm[of * 2 + 1] = frame[d];
185 if (pcm[of * 2 + 1] == 0x80 && pcm[of * 2] == 0x00)
187 } else { /* 12-bit quantization */
188 lc = ((uint16_t)frame[d] << 4) |
189 ((uint16_t)frame[d + 2] >> 4);
190 rc = ((uint16_t)frame[d + 1] << 4) |
191 ((uint16_t)frame[d + 2] & 0x0f);
192 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
193 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
195 of = sys->audio_shuffle[i % half_ch][j] +
196 (d - 8) / 3 * sys->audio_stride;
200 /* FIXME: maybe we have to admit that DV is a
202 pcm[of * 2] = lc & 0xff;
203 pcm[of * 2 + 1] = lc >> 8;
204 of = sys->audio_shuffle[i % half_ch + half_ch][j] +
205 (d - 8) / 3 * sys->audio_stride;
206 /* FIXME: maybe we have to admit that DV is a
208 pcm[of * 2] = rc & 0xff;
209 pcm[of * 2 + 1] = rc >> 8;
214 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
222 static int dv_extract_audio_info(DVDemuxContext *c, const uint8_t *frame)
224 const uint8_t *as_pack;
225 int freq, stype, smpls, quant, i, ach;
227 as_pack = dv_extract_pack(frame, dv_audio_source);
228 if (!as_pack || !c->sys) { /* No audio ? */
233 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
234 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
235 stype = as_pack[3] & 0x1f; /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
236 quant = as_pack[4] & 0x07; /* 0 - 16-bit linear, 1 - 12-bit nonlinear */
238 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
239 av_log(c->fctx, AV_LOG_ERROR,
240 "Unrecognized audio sample rate index (%d)\n", freq);
245 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
250 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
251 ach = ((int[4]) { 1, 0, 2, 4 })[stype];
252 if (ach == 1 && quant && freq == 2)
255 /* Dynamic handling of the audio streams in DV */
256 for (i = 0; i < ach; i++) {
258 c->ast[i] = avformat_new_stream(c->fctx, NULL);
261 avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
262 c->ast[i]->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
263 c->ast[i]->codecpar->codec_id = AV_CODEC_ID_PCM_S16LE;
265 av_init_packet(&c->audio_pkt[i]);
266 c->audio_pkt[i].size = 0;
267 c->audio_pkt[i].data = c->audio_buf[i];
268 c->audio_pkt[i].stream_index = c->ast[i]->index;
269 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
271 c->ast[i]->codecpar->sample_rate = dv_audio_frequency[freq];
272 c->ast[i]->codecpar->channels = 2;
273 c->ast[i]->codecpar->channel_layout = AV_CH_LAYOUT_STEREO;
274 c->ast[i]->codecpar->bit_rate = 2 * dv_audio_frequency[freq] * 16;
275 c->ast[i]->start_time = 0;
279 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
282 static int dv_extract_video_info(DVDemuxContext *c, const uint8_t *frame)
284 const uint8_t *vsc_pack;
285 AVCodecParameters *par;
288 par = c->vst->codecpar;
290 avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
291 c->sys->time_base.den);
292 c->vst->avg_frame_rate = av_inv_q(c->vst->time_base);
294 /* finding out SAR is a little bit messy */
295 vsc_pack = dv_extract_pack(frame, dv_video_control);
296 apt = frame[4] & 0x07;
297 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
298 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
299 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
300 par->bit_rate = av_rescale_q(c->sys->frame_size,
301 (AVRational) { 8, 1 },
303 return c->sys->frame_size;
306 static int dv_extract_timecode(DVDemuxContext* c, const uint8_t* frame, char *tc)
308 const uint8_t *tc_pack;
310 // For PAL systems, drop frame bit is replaced by an arbitrary
311 // bit so its value should not be considered. Drop frame timecode
312 // is only relevant for NTSC systems.
313 int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
315 tc_pack = dv_extract_pack(frame, dv_timecode);
318 av_timecode_make_smpte_tc_string(tc, AV_RB32(tc_pack + 1), prevent_df);
322 /* The following 3 functions constitute our interface to the world */
324 DVDemuxContext *avpriv_dv_init_demux(AVFormatContext *s)
328 c = av_mallocz(sizeof(DVDemuxContext));
332 c->vst = avformat_new_stream(s, NULL);
339 c->vst->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
340 c->vst->codecpar->codec_id = AV_CODEC_ID_DVVIDEO;
341 c->vst->codecpar->bit_rate = 25000000;
342 c->vst->start_time = 0;
347 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
352 for (i = 0; i < c->ach; i++) {
353 if (c->ast[i] && c->audio_pkt[i].size) {
354 *pkt = c->audio_pkt[i];
355 c->audio_pkt[i].size = 0;
364 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
365 uint8_t *buf, int buf_size, int64_t pos)
368 uint8_t *ppcm[5] = { 0 };
370 if (buf_size < DV_PROFILE_BYTES ||
371 !(c->sys = av_dv_frame_profile(c->sys, buf, buf_size)) ||
372 buf_size < c->sys->frame_size) {
373 return -1; /* Broken frame, or not enough data */
376 /* Queueing audio packet */
377 /* FIXME: in case of no audio/bad audio we have to do something */
378 size = dv_extract_audio_info(c, buf);
379 for (i = 0; i < c->ach; i++) {
380 c->audio_pkt[i].pos = pos;
381 c->audio_pkt[i].size = size;
382 c->audio_pkt[i].pts = c->abytes * 30000 * 8 /
383 c->ast[i]->codecpar->bit_rate;
384 ppcm[i] = c->audio_buf[i];
387 dv_extract_audio(buf, ppcm, c->sys);
389 /* We work with 720p frames split in half, thus even frames have
390 * channels 0,1 and odd 2,3. */
391 if (c->sys->height == 720) {
393 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
395 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
402 /* Now it's time to return video packet */
403 size = dv_extract_video_info(c, buf);
408 pkt->flags |= AV_PKT_FLAG_KEY;
409 pkt->stream_index = c->vst->index;
410 pkt->pts = c->frames;
417 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
418 int64_t timestamp, int flags)
420 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
421 const int frame_size = c->sys->frame_size;
423 int64_t size = avio_size(s->pb) - s->internal->data_offset;
424 int64_t max_offset = ((size - 1) / frame_size) * frame_size;
426 offset = frame_size * timestamp;
428 if (size >= 0 && offset > max_offset)
433 return offset + s->internal->data_offset;
436 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
438 c->frames = frame_offset;
441 c->abytes = av_rescale_q(c->frames, c->sys->time_base,
442 (AVRational) { 8, c->ast[0]->codecpar->bit_rate });
444 av_log(c->fctx, AV_LOG_ERROR, "cannot adjust audio bytes\n");
446 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
447 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
450 /************************************************************
451 * Implementation of the easiest DV storage of all -- raw DV.
452 ************************************************************/
454 typedef struct RawDVContext {
455 DVDemuxContext *dv_demux;
456 uint8_t buf[DV_MAX_FRAME_SIZE];
459 static int dv_read_timecode(AVFormatContext *s) {
461 char timecode[AV_TIMECODE_STR_SIZE];
462 int64_t pos = avio_tell(s->pb);
464 // Read 3 DIF blocks: Header block and 2 Subcode blocks.
465 int partial_frame_size = 3 * 80;
466 uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
469 RawDVContext *c = s->priv_data;
471 return AVERROR(ENOMEM);
473 ret = avio_read(s->pb, partial_frame, partial_frame_size);
477 if (ret < partial_frame_size) {
482 ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
484 av_dict_set(&s->metadata, "timecode", timecode, 0);
486 av_log(s, AV_LOG_ERROR, "Detected timecode is invalid\n");
489 av_free(partial_frame);
490 avio_seek(s->pb, pos, SEEK_SET);
494 static int dv_read_header(AVFormatContext *s)
496 unsigned state, marker_pos = 0;
497 RawDVContext *c = s->priv_data;
499 c->dv_demux = avpriv_dv_init_demux(s);
503 state = avio_rb32(s->pb);
504 while ((state & 0xffffff7f) != 0x1f07003f) {
505 if (avio_feof(s->pb)) {
506 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
509 if (state == 0x003f0700 || state == 0xff3f0700)
510 marker_pos = avio_tell(s->pb);
511 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
512 avio_seek(s->pb, -163, SEEK_CUR);
513 state = avio_rb32(s->pb);
516 state = (state << 8) | avio_r8(s->pb);
518 AV_WB32(c->buf, state);
520 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) != DV_PROFILE_BYTES - 4 ||
521 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
524 c->dv_demux->sys = av_dv_frame_profile(c->dv_demux->sys,
527 if (!c->dv_demux->sys) {
528 av_log(s, AV_LOG_ERROR,
529 "Can't determine profile of DV input stream.\n");
533 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size,
534 (AVRational) { 8, 1 },
535 c->dv_demux->sys->time_base);
537 if (s->pb->seekable & AVIO_SEEKABLE_NORMAL)
543 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
546 RawDVContext *c = s->priv_data;
548 size = avpriv_dv_get_packet(c->dv_demux, pkt);
552 int64_t pos = avio_tell(s->pb);
553 if (!c->dv_demux->sys)
555 size = c->dv_demux->sys->frame_size;
556 ret = avio_read(s->pb, c->buf, size);
559 } else if (ret == 0) {
563 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
569 static int dv_read_seek(AVFormatContext *s, int stream_index,
570 int64_t timestamp, int flags)
572 RawDVContext *r = s->priv_data;
573 DVDemuxContext *c = r->dv_demux;
574 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
576 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
579 ff_dv_offset_reset(c, offset / c->sys->frame_size);
583 static int dv_read_close(AVFormatContext *s)
585 RawDVContext *c = s->priv_data;
586 av_freep(&c->dv_demux);
590 static int dv_probe(AVProbeData *p)
592 unsigned marker_pos = 0;
596 int secondary_matches = 0;
601 for (i = 0; i < p->buf_size-4; i++) {
602 unsigned state = AV_RB32(p->buf+i);
603 if ((state & 0x0007f840) == 0x00070000) {
604 // any section header, also with seq/chan num != 0,
605 // should appear around every 12000 bytes, at least 10 per frame
606 if ((state & 0xff07ff7f) == 0x1f07003f) {
608 if ((state & 0xffffff7f) == 0x1f07003f) {
614 if (state == 0x003f0700 || state == 0xff3f0700)
616 if (state == 0xff3f0701 && i - marker_pos == 80)
621 if (matches && p->buf_size / matches < 1024 * 1024) {
622 if (matches > 4 || firstmatch ||
623 (secondary_matches >= 10 &&
624 p->buf_size / secondary_matches < 24000))
625 // not max to avoid dv in mov to match
626 return AVPROBE_SCORE_MAX * 3 / 4;
627 return AVPROBE_SCORE_MAX / 4;
632 AVInputFormat ff_dv_demuxer = {
634 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
635 .priv_data_size = sizeof(RawDVContext),
636 .read_probe = dv_probe,
637 .read_header = dv_read_header,
638 .read_packet = dv_read_packet,
639 .read_close = dv_read_close,
640 .read_seek = dv_read_seek,
641 .extensions = "dv,dif",