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"
38 #include "libavutil/avassert.h"
40 struct DVDemuxContext {
41 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
42 AVFormatContext* fctx;
45 AVPacket audio_pkt[4];
46 uint8_t audio_buf[4][8192];
52 static inline uint16_t dv_audio_12to16(uint16_t sample)
54 uint16_t shift, result;
56 sample = (sample < 0x800) ? sample : sample | 0xf000;
57 shift = (sample & 0xf00) >> 8;
59 if (shift < 0x2 || shift > 0xd) {
61 } else if (shift < 0x8) {
63 result = (sample - (256 * shift)) << shift;
66 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
73 * This is the dumbest implementation of all -- it simply looks at
74 * a fixed offset and if pack isn't there -- fails. We might want
75 * to have a fallback mechanism for complete search of missing packs.
77 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
83 offs = (80*6 + 80*16*3 + 3);
85 case dv_audio_control:
86 offs = (80*6 + 80*16*4 + 3);
88 case dv_video_control:
89 offs = (80*5 + 48 + 5);
92 offs = (80*1 + 3 + 3);
98 return frame[offs] == t ? &frame[offs] : NULL;
102 * There's a couple of assumptions being made here:
103 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
104 * We can pass them upwards when libavcodec will be ready to deal with them.
105 * 2. We don't do software emphasis.
106 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
107 * are converted into 16bit linear ones.
109 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
110 const DVprofile *sys)
112 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
114 const uint8_t* as_pack;
117 as_pack = dv_extract_pack(frame, dv_audio_source);
118 if (!as_pack) /* No audio ? */
121 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
122 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
123 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
126 return -1; /* unsupported quantization */
128 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
129 half_ch = sys->difseg_size / 2;
131 /* We work with 720p frames split in half, thus even frames have
132 * channels 0,1 and odd 2,3. */
133 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
135 /* for each DIF channel */
136 for (chan = 0; chan < sys->n_difchan; chan++) {
141 /* for each DIF segment */
142 for (i = 0; i < sys->difseg_size; i++) {
143 frame += 6 * 80; /* skip DIF segment header */
144 if (quant == 1 && i == half_ch) {
145 /* next stereo channel (12bit mode only) */
152 /* for each AV sequence */
153 for (j = 0; j < 9; j++) {
154 for (d = 8; d < 80; d += 2) {
155 if (quant == 0) { /* 16bit quantization */
156 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
160 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
161 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
162 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
164 } else { /* 12bit quantization */
165 lc = ((uint16_t)frame[d] << 4) |
166 ((uint16_t)frame[d+2] >> 4);
167 rc = ((uint16_t)frame[d+1] << 4) |
168 ((uint16_t)frame[d+2] & 0x0f);
169 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
170 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
172 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
176 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
177 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
178 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
179 (d - 8) / 3 * sys->audio_stride;
180 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
181 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
186 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
190 /* next stereo channel (50Mbps and 100Mbps only) */
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 bcd2int(uint8_t bcd)
283 if (low > 9 || high > 9)
285 return low + 10*high;
288 static int dv_extract_timecode(DVDemuxContext* c, uint8_t* frame, char tc[32])
290 int hh, mm, ss, ff, drop_frame;
291 const uint8_t *tc_pack;
293 tc_pack = dv_extract_pack(frame, dv_timecode);
297 ff = bcd2int(tc_pack[1] & 0x3f);
298 ss = bcd2int(tc_pack[2] & 0x7f);
299 mm = bcd2int(tc_pack[3] & 0x7f);
300 hh = bcd2int(tc_pack[4] & 0x3f);
301 drop_frame = tc_pack[1] >> 6 & 0x1;
303 if (ff < 0 || ss < 0 || mm < 0 || hh < 0)
306 // For PAL systems, drop frame bit is replaced by an arbitrary
307 // bit so its value should not be considered. Drop frame timecode
308 // is only relevant for NTSC systems.
309 if(c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50) {
313 snprintf(tc, 32, "%02d:%02d:%02d%c%02d",
314 hh, mm, ss, drop_frame ? ';' : ':', ff);
319 * The following 3 functions constitute our interface to the world
322 DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
326 c = av_mallocz(sizeof(DVDemuxContext));
330 c->vst = avformat_new_stream(s, NULL);
338 memset(c->ast, 0, sizeof(c->ast));
343 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
344 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
345 c->vst->codec->bit_rate = 25000000;
346 c->vst->start_time = 0;
351 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
356 for (i = 0; i < c->ach; i++) {
357 if (c->ast[i] && c->audio_pkt[i].size) {
358 *pkt = c->audio_pkt[i];
359 c->audio_pkt[i].size = 0;
368 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
369 uint8_t* buf, int buf_size, int64_t pos)
372 uint8_t *ppcm[4] = {0};
374 if (buf_size < DV_PROFILE_BYTES ||
375 !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
376 buf_size < c->sys->frame_size) {
377 return -1; /* Broken frame, or not enough data */
380 /* Queueing audio packet */
381 /* FIXME: in case of no audio/bad audio we have to do something */
382 size = dv_extract_audio_info(c, buf);
383 for (i = 0; i < c->ach; i++) {
384 c->audio_pkt[i].pos = pos;
385 c->audio_pkt[i].size = size;
386 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
387 ppcm[i] = c->audio_buf[i];
390 dv_extract_audio(buf, ppcm, c->sys);
392 /* We work with 720p frames split in half, thus even frames have
393 * channels 0,1 and odd 2,3. */
394 if (c->sys->height == 720) {
396 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
398 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->id;
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 DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
426 int64_t size = avio_size(s->pb) - s->data_offset;
427 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
429 offset = sys->frame_size * timestamp;
431 if (size >= 0 && offset > max_offset) offset = max_offset;
432 else if (offset < 0) offset = 0;
434 return offset + s->data_offset;
437 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
439 c->frames= frame_offset;
441 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
442 (AVRational){8, c->ast[0]->codec->bit_rate});
443 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
444 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
447 /************************************************************
448 * Implementation of the easiest DV storage of all -- raw DV.
449 ************************************************************/
451 typedef struct RawDVContext {
452 DVDemuxContext* dv_demux;
453 uint8_t buf[DV_MAX_FRAME_SIZE];
456 static int dv_read_timecode(AVFormatContext *s) {
459 int64_t pos = avio_tell(s->pb);
461 // Read 3 DIF blocks: Header block and 2 Subcode blocks.
462 int partial_frame_size = 3 * 80;
463 uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
466 RawDVContext *c = s->priv_data;
467 ret = avio_read(s->pb, partial_frame, partial_frame_size);
471 if (ret < partial_frame_size) {
476 ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
478 av_dict_set(&s->metadata, "timecode", timecode, 0);
480 av_log(s, AV_LOG_ERROR, "Detected timecode is invalid");
483 av_free(partial_frame);
484 avio_seek(s->pb, pos, SEEK_SET);
488 static int dv_read_header(AVFormatContext *s)
490 unsigned state, marker_pos = 0;
491 RawDVContext *c = s->priv_data;
493 c->dv_demux = avpriv_dv_init_demux(s);
497 state = avio_rb32(s->pb);
498 while ((state & 0xffffff7f) != 0x1f07003f) {
499 if (url_feof(s->pb)) {
500 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
503 if (state == 0x003f0700 || state == 0xff3f0700)
504 marker_pos = avio_tell(s->pb);
505 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
506 avio_seek(s->pb, -163, SEEK_CUR);
507 state = avio_rb32(s->pb);
510 state = (state << 8) | avio_r8(s->pb);
512 AV_WB32(c->buf, state);
514 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
515 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
518 c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
519 if (!c->dv_demux->sys) {
520 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
524 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
525 c->dv_demux->sys->time_base);
534 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
537 RawDVContext *c = s->priv_data;
539 size = avpriv_dv_get_packet(c->dv_demux, pkt);
542 int64_t pos = avio_tell(s->pb);
543 if (!c->dv_demux->sys)
545 size = c->dv_demux->sys->frame_size;
546 if (avio_read(s->pb, c->buf, size) <= 0)
549 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
555 static int dv_read_seek(AVFormatContext *s, int stream_index,
556 int64_t timestamp, int flags)
558 RawDVContext *r = s->priv_data;
559 DVDemuxContext *c = r->dv_demux;
560 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
562 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
565 dv_offset_reset(c, offset / c->sys->frame_size);
569 static int dv_read_close(AVFormatContext *s)
571 RawDVContext *c = s->priv_data;
572 av_free(c->dv_demux);
576 static int dv_probe(AVProbeData *p)
578 unsigned state, marker_pos = 0;
581 int secondary_matches = 0;
586 state = AV_RB32(p->buf);
587 for (i = 4; i < p->buf_size; i++) {
588 if ((state & 0xffffff7f) == 0x1f07003f)
590 // any section header, also with seq/chan num != 0,
591 // should appear around every 12000 bytes, at least 10 per frame
592 if ((state & 0xff07ff7f) == 0x1f07003f)
594 if (state == 0x003f0700 || state == 0xff3f0700)
596 if (state == 0xff3f0701 && i - marker_pos == 80)
598 state = (state << 8) | p->buf[i];
601 if (matches && p->buf_size / matches < 1024*1024) {
602 if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
603 return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
604 return AVPROBE_SCORE_MAX/4;
609 #if CONFIG_DV_DEMUXER
610 AVInputFormat ff_dv_demuxer = {
612 .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
613 .priv_data_size = sizeof(RawDVContext),
614 .read_probe = dv_probe,
615 .read_header = dv_read_header,
616 .read_packet = dv_read_packet,
617 .read_close = dv_read_close,
618 .read_seek = dv_read_seek,
619 .extensions = "dv,dif",