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];
54 static inline uint16_t dv_audio_12to16(uint16_t sample)
56 uint16_t shift, result;
58 sample = (sample < 0x800) ? sample : sample | 0xf000;
59 shift = (sample & 0xf00) >> 8;
61 if (shift < 0x2 || shift > 0xd) {
63 } else if (shift < 0x8) {
65 result = (sample - (256 * shift)) << shift;
68 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
74 static const uint8_t *dv_extract_pack(const uint8_t *frame, enum dv_pack_type t)
79 for (c = 0; c < 10; c++) {
82 if (c&1) offs = (80 * 6 + 80 * 16 * 0 + 3 + c*12000);
83 else offs = (80 * 6 + 80 * 16 * 3 + 3 + c*12000);
85 case dv_audio_control:
86 if (c&1) offs = (80 * 6 + 80 * 16 * 1 + 3 + c*12000);
87 else offs = (80 * 6 + 80 * 16 * 4 + 3 + c*12000);
89 case dv_video_control:
90 if (c&1) offs = (80 * 3 + 8 + c*12000);
91 else offs = (80 * 5 + 48 + 5 + c*12000);
94 offs = (80*1 + 3 + 3);
103 return frame[offs] == t ? &frame[offs] : NULL;
106 static const int dv_audio_frequency[3] = {
111 * There's a couple of assumptions being made here:
112 * 1. By default we silence erroneous (0x8000/16-bit 0x800/12-bit) audio samples.
113 * We can pass them upwards when libavcodec will be ready to deal with them.
114 * 2. We don't do software emphasis.
115 * 3. Audio is always returned as 16-bit linear samples: 12-bit nonlinear samples
116 * are converted into 16-bit linear ones.
118 static int dv_extract_audio(const uint8_t *frame, uint8_t **ppcm,
119 const AVDVProfile *sys)
121 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
123 const uint8_t *as_pack;
126 as_pack = dv_extract_pack(frame, dv_audio_source);
127 if (!as_pack) /* No audio ? */
130 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
131 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
132 quant = as_pack[4] & 0x07; /* 0 - 16-bit linear, 1 - 12-bit nonlinear */
135 return -1; /* unsupported quantization */
137 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
138 return AVERROR_INVALIDDATA;
140 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
141 half_ch = sys->difseg_size / 2;
143 /* We work with 720p frames split in half, thus even frames have
144 * channels 0,1 and odd 2,3. */
145 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
147 if (ipcm + sys->n_difchan > (quant == 1 ? 2 : 4)) {
148 av_log(NULL, AV_LOG_ERROR, "too many dv pcm frames\n");
149 return AVERROR_INVALIDDATA;
152 /* for each DIF channel */
153 for (chan = 0; chan < sys->n_difchan; chan++) {
159 /* for each DIF segment */
160 for (i = 0; i < sys->difseg_size; i++) {
161 frame += 6 * 80; /* skip DIF segment header */
162 if (quant == 1 && i == half_ch) {
163 /* next stereo channel (12-bit mode only) */
170 /* for each AV sequence */
171 for (j = 0; j < 9; j++) {
172 for (d = 8; d < 80; d += 2) {
173 if (quant == 0) { /* 16-bit quantization */
174 of = sys->audio_shuffle[i][j] +
175 (d - 8) / 2 * sys->audio_stride;
179 /* FIXME: maybe we have to admit that DV is a
181 pcm[of * 2] = frame[d + 1];
182 pcm[of * 2 + 1] = frame[d];
184 if (pcm[of * 2 + 1] == 0x80 && pcm[of * 2] == 0x00)
186 } else { /* 12-bit quantization */
187 lc = ((uint16_t)frame[d] << 4) |
188 ((uint16_t)frame[d + 2] >> 4);
189 rc = ((uint16_t)frame[d + 1] << 4) |
190 ((uint16_t)frame[d + 2] & 0x0f);
191 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
192 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
194 of = sys->audio_shuffle[i % half_ch][j] +
195 (d - 8) / 3 * sys->audio_stride;
199 /* FIXME: maybe we have to admit that DV is a
201 pcm[of * 2] = lc & 0xff;
202 pcm[of * 2 + 1] = lc >> 8;
203 of = sys->audio_shuffle[i % half_ch + half_ch][j] +
204 (d - 8) / 3 * sys->audio_stride;
205 /* FIXME: maybe we have to admit that DV is a
207 pcm[of * 2] = rc & 0xff;
208 pcm[of * 2 + 1] = rc >> 8;
213 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
221 static int dv_extract_audio_info(DVDemuxContext *c, const uint8_t *frame)
223 const uint8_t *as_pack;
224 int freq, stype, smpls, quant, i, ach;
226 as_pack = dv_extract_pack(frame, dv_audio_source);
227 if (!as_pack || !c->sys) { /* No audio ? */
232 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
233 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
234 stype = as_pack[3] & 0x1f; /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
235 quant = as_pack[4] & 0x07; /* 0 - 16-bit linear, 1 - 12-bit nonlinear */
237 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
238 av_log(c->fctx, AV_LOG_ERROR,
239 "Unrecognized audio sample rate index (%d)\n", freq);
244 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
249 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
250 ach = ((int[4]) { 1, 0, 2, 4 })[stype];
251 if (ach == 1 && quant && freq == 2)
254 /* Dynamic handling of the audio streams in DV */
255 for (i = 0; i < ach; i++) {
257 c->ast[i] = avformat_new_stream(c->fctx, NULL);
260 avpriv_set_pts_info(c->ast[i], 64, c->sys->time_base.num, c->sys->time_base.den);
261 c->ast[i]->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
262 c->ast[i]->codecpar->codec_id = AV_CODEC_ID_PCM_S16LE;
264 av_init_packet(&c->audio_pkt[i]);
265 c->audio_pkt[i].size = 0;
266 c->audio_pkt[i].data = c->audio_buf[i];
267 c->audio_pkt[i].stream_index = c->ast[i]->index;
268 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
270 c->ast[i]->codecpar->sample_rate = dv_audio_frequency[freq];
271 c->ast[i]->codecpar->channels = 2;
272 c->ast[i]->codecpar->channel_layout = AV_CH_LAYOUT_STEREO;
273 c->ast[i]->codecpar->bit_rate = 2 * dv_audio_frequency[freq] * 16;
274 c->ast[i]->start_time = 0;
278 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
281 static int dv_extract_video_info(DVDemuxContext *c, const uint8_t *frame)
283 const uint8_t *vsc_pack;
284 AVCodecParameters *par;
287 par = c->vst->codecpar;
289 avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
290 c->sys->time_base.den);
291 c->vst->avg_frame_rate = av_inv_q(c->vst->time_base);
293 /* finding out SAR is a little bit messy */
294 vsc_pack = dv_extract_pack(frame, dv_video_control);
295 apt = frame[4] & 0x07;
296 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
297 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
298 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
299 par->bit_rate = av_rescale_q(c->sys->frame_size,
300 (AVRational) { 8, 1 },
302 return c->sys->frame_size;
305 static int dv_extract_timecode(DVDemuxContext* c, const uint8_t* frame, char *tc)
307 const uint8_t *tc_pack;
309 // For PAL systems, drop frame bit is replaced by an arbitrary
310 // bit so its value should not be considered. Drop frame timecode
311 // is only relevant for NTSC systems.
312 int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
314 tc_pack = dv_extract_pack(frame, dv_timecode);
317 av_timecode_make_smpte_tc_string2(tc, av_inv_q(c->sys->time_base), AV_RB32(tc_pack + 1), prevent_df, 1);
321 /* The following 3 functions constitute our interface to the world */
323 static int dv_init_demux(AVFormatContext *s, DVDemuxContext *c)
325 c->vst = avformat_new_stream(s, NULL);
327 return AVERROR(ENOMEM);
330 c->vst->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
331 c->vst->codecpar->codec_id = AV_CODEC_ID_DVVIDEO;
332 c->vst->codecpar->bit_rate = 25000000;
333 c->vst->start_time = 0;
338 DVDemuxContext *avpriv_dv_init_demux(AVFormatContext *s)
342 c = av_mallocz(sizeof(DVDemuxContext));
346 if (dv_init_demux(s, c)) {
354 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
359 for (i = 0; i < c->ach; i++) {
360 if (c->ast[i] && c->audio_pkt[i].size) {
361 *pkt = c->audio_pkt[i];
362 c->audio_pkt[i].size = 0;
371 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
372 uint8_t *buf, int buf_size, int64_t pos)
375 uint8_t *ppcm[5] = { 0 };
377 if (buf_size < DV_PROFILE_BYTES ||
378 !(c->sys = av_dv_frame_profile(c->sys, buf, buf_size)) ||
379 buf_size < c->sys->frame_size) {
380 return -1; /* Broken frame, or not enough data */
383 /* Queueing audio packet */
384 /* FIXME: in case of no audio/bad audio we have to do something */
385 size = dv_extract_audio_info(c, buf);
386 for (i = 0; i < c->ach; i++) {
387 c->audio_pkt[i].pos = pos;
388 c->audio_pkt[i].size = size;
389 c->audio_pkt[i].pts = (c->sys->height == 720) ? (c->frames & ~1) : c->frames;
390 ppcm[i] = c->audio_buf[i];
393 dv_extract_audio(buf, ppcm, c->sys);
395 /* We work with 720p frames split in half, thus even frames have
396 * channels 0,1 and odd 2,3. */
397 if (c->sys->height == 720) {
399 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
401 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
405 /* Now it's time to return video packet */
406 size = dv_extract_video_info(c, buf);
411 pkt->flags |= AV_PKT_FLAG_KEY;
412 pkt->stream_index = c->vst->index;
413 pkt->pts = c->frames;
420 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
421 int64_t timestamp, int flags)
423 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
424 const int frame_size = c->sys->frame_size;
426 int64_t size = avio_size(s->pb) - s->internal->data_offset;
427 int64_t max_offset = ((size - 1) / frame_size) * frame_size;
429 offset = frame_size * timestamp;
431 if (size >= 0 && offset > max_offset)
436 return offset + s->internal->data_offset;
439 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
441 c->frames = frame_offset;
442 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
443 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
446 /************************************************************
447 * Implementation of the easiest DV storage of all -- raw DV.
448 ************************************************************/
450 typedef struct RawDVContext {
451 DVDemuxContext dv_demux;
452 uint8_t buf[DV_MAX_FRAME_SIZE];
455 static int dv_read_timecode(AVFormatContext *s) {
457 char timecode[AV_TIMECODE_STR_SIZE];
458 int64_t pos = avio_tell(s->pb);
460 // Read 3 DIF blocks: Header block and 2 Subcode blocks.
461 #define PARTIAL_FRAME_SIZE (3 * 80)
462 uint8_t partial_frame[PARTIAL_FRAME_SIZE];
463 RawDVContext *c = s->priv_data;
465 ret = avio_read(s->pb, partial_frame, PARTIAL_FRAME_SIZE);
469 if (ret < PARTIAL_FRAME_SIZE) {
474 ret = dv_extract_timecode(&c->dv_demux, partial_frame, timecode);
476 av_dict_set(&s->metadata, "timecode", timecode, 0);
478 av_log(s, AV_LOG_ERROR, "Detected timecode is invalid\n");
481 avio_seek(s->pb, pos, SEEK_SET);
485 static int dv_read_header(AVFormatContext *s)
487 unsigned state, marker_pos = 0;
488 RawDVContext *c = s->priv_data;
491 if ((ret = dv_init_demux(s, &c->dv_demux)) < 0)
494 state = avio_rb32(s->pb);
495 while ((state & 0xffffff7f) != 0x1f07003f) {
496 if (avio_feof(s->pb)) {
497 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
498 return AVERROR_INVALIDDATA;
500 if (state == 0x003f0700 || state == 0xff3f0700)
501 marker_pos = avio_tell(s->pb);
502 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
503 avio_seek(s->pb, -163, SEEK_CUR);
504 state = avio_rb32(s->pb);
507 state = (state << 8) | avio_r8(s->pb);
509 AV_WB32(c->buf, state);
511 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) != DV_PROFILE_BYTES - 4 ||
512 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0) {
516 c->dv_demux.sys = av_dv_frame_profile(c->dv_demux.sys,
519 if (!c->dv_demux.sys) {
520 av_log(s, AV_LOG_ERROR,
521 "Can't determine profile of DV input stream.\n");
522 return AVERROR_INVALIDDATA;
525 s->bit_rate = av_rescale_q(c->dv_demux.sys->frame_size,
526 (AVRational) { 8, 1 },
527 c->dv_demux.sys->time_base);
529 if (s->pb->seekable & AVIO_SEEKABLE_NORMAL)
535 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
538 RawDVContext *c = s->priv_data;
540 size = avpriv_dv_get_packet(&c->dv_demux, pkt);
544 int64_t pos = avio_tell(s->pb);
545 if (!c->dv_demux.sys)
547 size = c->dv_demux.sys->frame_size;
548 ret = avio_read(s->pb, c->buf, size);
551 } else if (ret == 0) {
555 size = avpriv_dv_produce_packet(&c->dv_demux, pkt, c->buf, size, pos);
561 static int dv_read_seek(AVFormatContext *s, int stream_index,
562 int64_t timestamp, int flags)
564 RawDVContext *r = s->priv_data;
565 DVDemuxContext *c = &r->dv_demux;
566 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
568 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
571 ff_dv_offset_reset(c, offset / c->sys->frame_size);
575 static int dv_probe(const AVProbeData *p)
577 unsigned marker_pos = 0;
581 int secondary_matches = 0;
586 for (i = 0; i < p->buf_size-4; i++) {
587 unsigned state = AV_RB32(p->buf+i);
588 if ((state & 0x0007f840) == 0x00070000) {
589 // any section header, also with seq/chan num != 0,
590 // should appear around every 12000 bytes, at least 10 per frame
591 if ((state & 0xff07ff7f) == 0x1f07003f) {
593 if ((state & 0xffffff7f) == 0x1f07003f) {
599 if (state == 0x003f0700 || state == 0xff3f0700)
601 if (state == 0xff3f0701 && i - marker_pos == 80)
606 if (matches && p->buf_size / matches < 1024 * 1024) {
607 if (matches > 4 || firstmatch ||
608 (secondary_matches >= 10 &&
609 p->buf_size / secondary_matches < 24000))
610 // not max to avoid dv in mov to match
611 return AVPROBE_SCORE_MAX * 3 / 4;
612 return AVPROBE_SCORE_MAX / 4;
617 AVInputFormat ff_dv_demuxer = {
619 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
620 .priv_data_size = sizeof(RawDVContext),
621 .read_probe = dv_probe,
622 .read_header = dv_read_header,
623 .read_packet = dv_read_packet,
624 .read_seek = dv_read_seek,
625 .extensions = "dv,dif",