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/dvdata.h"
35 #include "libavutil/intreadwrite.h"
36 #include "libavutil/mathematics.h"
37 #include "libavutil/timecode.h"
39 #include "libavutil/avassert.h"
41 struct DVDemuxContext {
42 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
43 AVFormatContext* fctx;
46 AVPacket audio_pkt[4];
47 uint8_t audio_buf[4][8192];
53 static inline uint16_t dv_audio_12to16(uint16_t sample)
55 uint16_t shift, result;
57 sample = (sample < 0x800) ? sample : sample | 0xf000;
58 shift = (sample & 0xf00) >> 8;
60 if (shift < 0x2 || shift > 0xd) {
62 } else if (shift < 0x8) {
64 result = (sample - (256 * shift)) << shift;
67 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
74 * This is the dumbest implementation of all -- it simply looks at
75 * a fixed offset and if pack isn't there -- fails. We might want
76 * to have a fallback mechanism for complete search of missing packs.
78 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
84 offs = (80*6 + 80*16*3 + 3);
86 case dv_audio_control:
87 offs = (80*6 + 80*16*4 + 3);
89 case dv_video_control:
90 offs = (80*5 + 48 + 5);
93 offs = (80*1 + 3 + 3);
99 return frame[offs] == t ? &frame[offs] : NULL;
103 * There's a couple of assumptions being made here:
104 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
105 * We can pass them upwards when libavcodec will be ready to deal with them.
106 * 2. We don't do software emphasis.
107 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
108 * are converted into 16bit linear ones.
110 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
111 const DVprofile *sys)
113 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
115 const uint8_t* as_pack;
118 as_pack = dv_extract_pack(frame, dv_audio_source);
119 if (!as_pack) /* No audio ? */
122 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
123 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
124 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
127 return -1; /* unsupported quantization */
129 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
130 half_ch = sys->difseg_size / 2;
132 /* We work with 720p frames split in half, thus even frames have
133 * channels 0,1 and odd 2,3. */
134 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
136 /* for each DIF channel */
137 for (chan = 0; chan < sys->n_difchan; chan++) {
143 /* for each DIF segment */
144 for (i = 0; i < sys->difseg_size; i++) {
145 frame += 6 * 80; /* skip DIF segment header */
146 if (quant == 1 && i == half_ch) {
147 /* next stereo channel (12bit mode only) */
154 /* for each AV sequence */
155 for (j = 0; j < 9; j++) {
156 for (d = 8; d < 80; d += 2) {
157 if (quant == 0) { /* 16bit quantization */
158 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
162 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
163 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
164 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
166 } else { /* 12bit quantization */
167 lc = ((uint16_t)frame[d] << 4) |
168 ((uint16_t)frame[d+2] >> 4);
169 rc = ((uint16_t)frame[d+1] << 4) |
170 ((uint16_t)frame[d+2] & 0x0f);
171 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
172 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
174 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
178 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
179 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
180 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
181 (d - 8) / 3 * sys->audio_stride;
182 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
183 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
188 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
196 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
198 const uint8_t* as_pack;
199 int freq, stype, smpls, quant, i, ach;
201 as_pack = dv_extract_pack(frame, dv_audio_source);
202 if (!as_pack || !c->sys) { /* No audio ? */
207 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
208 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
209 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
210 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
213 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
218 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
219 ach = ((int[4]){ 1, 0, 2, 4})[stype];
220 if (ach == 1 && quant && freq == 2)
223 /* Dynamic handling of the audio streams in DV */
224 for (i = 0; i < ach; i++) {
226 c->ast[i] = avformat_new_stream(c->fctx, NULL);
229 avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
230 c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
231 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
233 av_init_packet(&c->audio_pkt[i]);
234 c->audio_pkt[i].size = 0;
235 c->audio_pkt[i].data = c->audio_buf[i];
236 c->audio_pkt[i].stream_index = c->ast[i]->index;
237 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
239 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
240 c->ast[i]->codec->channels = 2;
241 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
242 c->ast[i]->start_time = 0;
246 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
249 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
251 const uint8_t* vsc_pack;
252 AVCodecContext* avctx;
257 avctx = c->vst->codec;
259 avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
260 c->sys->time_base.den);
261 avctx->time_base= c->sys->time_base;
263 avcodec_set_dimensions(avctx, c->sys->width, c->sys->height);
264 avctx->pix_fmt = c->sys->pix_fmt;
266 /* finding out SAR is a little bit messy */
267 vsc_pack = dv_extract_pack(frame, dv_video_control);
268 apt = frame[4] & 0x07;
269 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
270 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
271 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
272 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
274 size = c->sys->frame_size;
279 static int dv_extract_timecode(DVDemuxContext* c, uint8_t* frame, char *tc)
281 const uint8_t *tc_pack;
283 // For PAL systems, drop frame bit is replaced by an arbitrary
284 // bit so its value should not be considered. Drop frame timecode
285 // is only relevant for NTSC systems.
286 int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
288 tc_pack = dv_extract_pack(frame, dv_timecode);
291 av_timecode_make_smpte_tc_string(tc, AV_RB32(tc_pack + 1), prevent_df);
296 * The following 3 functions constitute our interface to the world
299 DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
303 c = av_mallocz(sizeof(DVDemuxContext));
307 c->vst = avformat_new_stream(s, NULL);
315 memset(c->ast, 0, sizeof(c->ast));
320 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
321 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
322 c->vst->codec->bit_rate = 25000000;
323 c->vst->start_time = 0;
328 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
333 for (i = 0; i < c->ach; i++) {
334 if (c->ast[i] && c->audio_pkt[i].size) {
335 *pkt = c->audio_pkt[i];
336 c->audio_pkt[i].size = 0;
345 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
346 uint8_t* buf, int buf_size, int64_t pos)
349 uint8_t *ppcm[4] = {0};
351 if (buf_size < DV_PROFILE_BYTES ||
352 !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
353 buf_size < c->sys->frame_size) {
354 return -1; /* Broken frame, or not enough data */
357 /* Queueing audio packet */
358 /* FIXME: in case of no audio/bad audio we have to do something */
359 size = dv_extract_audio_info(c, buf);
360 for (i = 0; i < c->ach; i++) {
361 c->audio_pkt[i].pos = pos;
362 c->audio_pkt[i].size = size;
363 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
364 ppcm[i] = c->audio_buf[i];
367 dv_extract_audio(buf, ppcm, c->sys);
369 /* We work with 720p frames split in half, thus even frames have
370 * channels 0,1 and odd 2,3. */
371 if (c->sys->height == 720) {
373 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
375 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
382 /* Now it's time to return video packet */
383 size = dv_extract_video_info(c, buf);
388 pkt->flags |= AV_PKT_FLAG_KEY;
389 pkt->stream_index = c->vst->id;
390 pkt->pts = c->frames;
397 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
398 int64_t timestamp, int flags)
400 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
401 const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
403 int64_t size = avio_size(s->pb) - s->data_offset;
404 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
406 offset = sys->frame_size * timestamp;
408 if (size >= 0 && offset > max_offset) offset = max_offset;
409 else if (offset < 0) offset = 0;
411 return offset + s->data_offset;
414 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
416 c->frames= frame_offset;
418 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
419 (AVRational){8, c->ast[0]->codec->bit_rate});
420 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
421 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
424 /************************************************************
425 * Implementation of the easiest DV storage of all -- raw DV.
426 ************************************************************/
428 typedef struct RawDVContext {
429 DVDemuxContext* dv_demux;
430 uint8_t buf[DV_MAX_FRAME_SIZE];
433 static int dv_read_timecode(AVFormatContext *s) {
435 char timecode[AV_TIMECODE_STR_SIZE];
436 int64_t pos = avio_tell(s->pb);
438 // Read 3 DIF blocks: Header block and 2 Subcode blocks.
439 int partial_frame_size = 3 * 80;
440 uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
443 RawDVContext *c = s->priv_data;
444 ret = avio_read(s->pb, partial_frame, partial_frame_size);
448 if (ret < partial_frame_size) {
453 ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
455 av_dict_set(&s->metadata, "timecode", timecode, 0);
457 av_log(s, AV_LOG_ERROR, "Detected timecode is invalid");
460 av_free(partial_frame);
461 avio_seek(s->pb, pos, SEEK_SET);
465 static int dv_read_header(AVFormatContext *s)
467 unsigned state, marker_pos = 0;
468 RawDVContext *c = s->priv_data;
470 c->dv_demux = avpriv_dv_init_demux(s);
474 state = avio_rb32(s->pb);
475 while ((state & 0xffffff7f) != 0x1f07003f) {
476 if (url_feof(s->pb)) {
477 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
480 if (state == 0x003f0700 || state == 0xff3f0700)
481 marker_pos = avio_tell(s->pb);
482 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
483 avio_seek(s->pb, -163, SEEK_CUR);
484 state = avio_rb32(s->pb);
487 state = (state << 8) | avio_r8(s->pb);
489 AV_WB32(c->buf, state);
491 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
492 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
495 c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
496 if (!c->dv_demux->sys) {
497 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
501 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
502 c->dv_demux->sys->time_base);
511 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
514 RawDVContext *c = s->priv_data;
516 size = avpriv_dv_get_packet(c->dv_demux, pkt);
519 int64_t pos = avio_tell(s->pb);
520 if (!c->dv_demux->sys)
522 size = c->dv_demux->sys->frame_size;
523 if (avio_read(s->pb, c->buf, size) <= 0)
526 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
532 static int dv_read_seek(AVFormatContext *s, int stream_index,
533 int64_t timestamp, int flags)
535 RawDVContext *r = s->priv_data;
536 DVDemuxContext *c = r->dv_demux;
537 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
539 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
542 dv_offset_reset(c, offset / c->sys->frame_size);
546 static int dv_read_close(AVFormatContext *s)
548 RawDVContext *c = s->priv_data;
549 av_free(c->dv_demux);
553 static int dv_probe(AVProbeData *p)
555 unsigned state, marker_pos = 0;
558 int secondary_matches = 0;
563 state = AV_RB32(p->buf);
564 for (i = 4; i < p->buf_size; i++) {
565 if ((state & 0xffffff7f) == 0x1f07003f)
567 // any section header, also with seq/chan num != 0,
568 // should appear around every 12000 bytes, at least 10 per frame
569 if ((state & 0xff07ff7f) == 0x1f07003f)
571 if (state == 0x003f0700 || state == 0xff3f0700)
573 if (state == 0xff3f0701 && i - marker_pos == 80)
575 state = (state << 8) | p->buf[i];
578 if (matches && p->buf_size / matches < 1024*1024) {
579 if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
580 return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
581 return AVPROBE_SCORE_MAX/4;
586 #if CONFIG_DV_DEMUXER
587 AVInputFormat ff_dv_demuxer = {
589 .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
590 .priv_data_size = sizeof(RawDVContext),
591 .read_probe = dv_probe,
592 .read_header = dv_read_header,
593 .read_packet = dv_read_packet,
594 .read_close = dv_read_close,
595 .read_seek = dv_read_seek,
596 .extensions = "dv,dif",