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) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
14 * This file is part of FFmpeg.
16 * FFmpeg is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU Lesser General Public
18 * License as published by the Free Software Foundation; either
19 * version 2.1 of the License, or (at your option) any later version.
21 * FFmpeg is distributed in the hope that it will be useful,
22 * but WITHOUT ANY WARRANTY; without even the implied warranty of
23 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
24 * Lesser General Public License for more details.
26 * You should have received a copy of the GNU Lesser General Public
27 * License along with FFmpeg; if not, write to the Free Software
28 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
35 #include "libavcodec/dv_profile.h"
36 #include "libavcodec/dv.h"
38 #include "libavutil/fifo.h"
39 #include "libavutil/mathematics.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/opt.h"
42 #include "libavutil/timecode.h"
44 #define MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32-bit audio
48 const AVDVProfile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
49 int n_ast; /* number of stereo audio streams (up to 2) */
50 AVStream *ast[2]; /* stereo audio streams */
51 AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
52 int frames; /* current frame number */
53 int64_t start_time; /* recording start time */
54 int has_audio; /* frame under construction has audio */
55 int has_video; /* frame under construction has video */
56 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under construction */
57 AVTimecode tc; /* timecode context */
60 static const int dv_aaux_packs_dist[12][9] = {
61 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
62 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
63 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
64 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
65 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
66 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
67 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
68 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
69 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
70 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
71 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
72 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
75 static int dv_audio_frame_size(const AVDVProfile* sys, int frame, int sample_rate)
77 if ((sys->time_base.den == 25 || sys->time_base.den == 50) && sys->time_base.num == 1) {
78 if (sample_rate == 32000) return 1280;
79 else if (sample_rate == 44100) return 1764;
83 av_assert0(sample_rate == 48000);
85 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
86 sizeof(sys->audio_samples_dist[0]))];
89 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
98 buf[0] = (uint8_t)pack_id;
101 timecode = av_timecode_get_smpte_from_framenum(&c->tc, c->frames);
102 timecode |= 1<<23 | 1<<15 | 1<<7 | 1<<6; // biphase and binary group flags
103 AV_WB32(buf + 1, timecode);
105 case dv_audio_source: /* AAUX source pack */
107 channel = va_arg(ap, int);
108 if (c->ast[channel]->codecpar->sample_rate == 44100) {
110 } else if (c->ast[channel]->codecpar->sample_rate == 32000)
112 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
113 (1 << 6) | /* reserved -- always 1 */
114 (dv_audio_frame_size(c->sys, c->frames, c->ast[channel]->codecpar->sample_rate) -
115 c->sys->audio_min_samples[audio_type]);
117 buf[2] = (0 << 7) | /* multi-stereo */
118 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
119 (0 << 4) | /* pair bit: 0 -- one pair of channels */
120 !!va_arg(ap, int); /* audio mode */
121 buf[3] = (1 << 7) | /* res */
122 (1 << 6) | /* multi-language flag */
123 (c->sys->dsf << 5) | /* system: 60fields/50fields */
124 (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
125 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
126 (0 << 6) | /* emphasis time constant: 0 -- reserved */
127 (audio_type << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
128 0; /* quantization: 0 -- 16-bit linear, 1 -- 12-bit nonlinear */
132 case dv_audio_control:
133 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
134 (1 << 4) | /* input source: 1 -- digital input */
135 (3 << 2) | /* compression: 3 -- no information */
136 0; /* misc. info/SMPTE emphasis off */
137 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
138 (1 << 6) | /* recording end point: 1 -- no */
139 (1 << 3) | /* recording mode: 1 -- original */
141 buf[3] = (1 << 7) | /* direction: 1 -- forward */
142 (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0x20 : /* speed */
143 c->sys->ltc_divisor * 4);
144 buf[4] = (1 << 7) | /* reserved -- always 1 */
145 0x7f; /* genre category */
147 case dv_audio_recdate:
148 case dv_video_recdate: /* VAUX recording date */
149 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
150 c->sys->time_base.den, AV_ROUND_DOWN);
151 ff_brktimegm(ct, &tc);
152 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
153 /* 0xff is very likely to be "unknown" */
154 buf[2] = (3 << 6) | /* reserved -- always 1 */
155 ((tc.tm_mday / 10) << 4) | /* Tens of day */
156 (tc.tm_mday % 10); /* Units of day */
157 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
158 ((tc.tm_mon / 10) << 4) | /* Tens of month */
159 (tc.tm_mon % 10); /* Units of month */
160 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
161 (tc.tm_year % 10); /* Units of year */
163 case dv_audio_rectime: /* AAUX recording time */
164 case dv_video_rectime: /* VAUX recording time */
165 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
166 c->sys->time_base.den, AV_ROUND_DOWN);
167 ff_brktimegm(ct, &tc);
168 buf[1] = (3 << 6) | /* reserved -- always 1 */
169 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
170 buf[2] = (1 << 7) | /* reserved -- always 1 */
171 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
172 (tc.tm_sec % 10); /* Units of seconds */
173 buf[3] = (1 << 7) | /* reserved -- always 1 */
174 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
175 (tc.tm_min % 10); /* Units of minutes */
176 buf[4] = (3 << 6) | /* reserved -- always 1 */
177 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
178 (tc.tm_hour % 10); /* Units of hours */
181 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
186 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
188 int i, j, d, of, size;
189 size = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[channel]->codecpar->sample_rate);
190 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
191 for (i = 0; i < c->sys->difseg_size; i++) {
192 frame_ptr += 6 * 80; /* skip DIF segment header */
193 for (j = 0; j < 9; j++) {
194 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], channel, i >= c->sys->difseg_size/2);
195 for (d = 8; d < 80; d+=2) {
196 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
200 frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
201 frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
203 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
208 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
213 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
214 /* DV subcode: 2nd and 3d DIFs */
215 for (j = 80; j < 80 * 3; j += 80) {
216 for (k = 6; k < 6 * 8; k += 8)
217 dv_write_pack(dv_timecode, c, &buf[j+k]);
219 if (((long)(buf-frame)/(c->sys->frame_size/(c->sys->difseg_size*c->sys->n_difchan))%c->sys->difseg_size) > 5) { /* FIXME: is this really needed ? */
220 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
221 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
222 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
223 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
227 /* DV VAUX: 4th, 5th and 6th 3DIFs */
228 for (j = 80*3 + 3; j < 80*6; j += 80) {
229 dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
230 dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
231 dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
232 dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
238 * The following 3 functions constitute our interface to the world
241 static int dv_assemble_frame(AVFormatContext *s,
242 DVMuxContext *c, AVStream* st,
243 uint8_t* data, int data_size, uint8_t** frame)
247 *frame = &c->frame_buf[0];
249 switch (st->codecpar->codec_type) {
250 case AVMEDIA_TYPE_VIDEO:
251 /* FIXME: we have to have more sensible approach than this one */
253 av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
254 if (data_size != c->sys->frame_size) {
255 av_log(s, AV_LOG_ERROR, "Unexpected frame size, %d != %d\n",
256 data_size, c->sys->frame_size);
257 return AVERROR(ENOSYS);
260 memcpy(*frame, data, c->sys->frame_size);
263 case AVMEDIA_TYPE_AUDIO:
264 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
266 /* FIXME: we have to have more sensible approach than this one */
267 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*MAX_AUDIO_FRAME_SIZE)
268 av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
269 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
271 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, st->codecpar->sample_rate);
273 /* Let us see if we've got enough audio for one DV frame. */
274 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
281 /* Let us see if we have enough data to construct one DV frame. */
282 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
283 dv_inject_metadata(c, *frame);
285 for (i=0; i < c->n_ast; i++) {
286 dv_inject_audio(c, i, *frame);
287 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[i]->codecpar->sample_rate);
288 av_fifo_drain(c->audio_data[i], reqasize);
289 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
296 return c->sys->frame_size;
302 static DVMuxContext* dv_init_mux(AVFormatContext* s)
304 DVMuxContext *c = s->priv_data;
305 AVStream *vst = NULL;
308 /* we support at most 1 video and 2 audio streams */
309 if (s->nb_streams > 3)
313 c->ast[0] = c->ast[1] = NULL;
315 /* We have to sort out where audio and where video stream is */
316 for (i=0; i<s->nb_streams; i++) {
317 switch (s->streams[i]->codecpar->codec_type) {
318 case AVMEDIA_TYPE_VIDEO:
319 if (vst) return NULL;
322 case AVMEDIA_TYPE_AUDIO:
323 if (c->n_ast > 1) return NULL;
324 c->ast[c->n_ast++] = s->streams[i];
331 /* Some checks -- DV format is very picky about its incoming streams */
332 if (!vst || vst->codecpar->codec_id != AV_CODEC_ID_DVVIDEO)
334 for (i=0; i<c->n_ast; i++) {
336 if(c->ast[i]->codecpar->codec_id != AV_CODEC_ID_PCM_S16LE ||
337 c->ast[i]->codecpar->channels != 2)
339 if (c->ast[i]->codecpar->sample_rate != 48000 &&
340 c->ast[i]->codecpar->sample_rate != 44100 &&
341 c->ast[i]->codecpar->sample_rate != 32000 )
345 c->sys = av_dv_codec_profile2(vst->codecpar->width, vst->codecpar->height,
346 vst->codecpar->format, vst->time_base);
350 if ((c->sys->time_base.den != 25 && c->sys->time_base.den != 50) || c->sys->time_base.num != 1) {
351 if (c->ast[0] && c->ast[0]->codecpar->sample_rate != 48000)
353 if (c->ast[1] && c->ast[1]->codecpar->sample_rate != 48000)
357 if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
358 /* only 1 stereo pair is allowed in 25Mbps mode */
362 /* Ok, everything seems to be in working order */
366 ff_parse_creation_time_metadata(s, &c->start_time, 1);
368 for (i=0; i < c->n_ast; i++) {
369 if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc_array(100, MAX_AUDIO_FRAME_SIZE))) {
372 av_fifo_freep(&c->audio_data[i]);
384 static void dv_delete_mux(DVMuxContext *c)
387 for (i=0; i < c->n_ast; i++)
388 av_fifo_freep(&c->audio_data[i]);
391 static int dv_write_header(AVFormatContext *s)
394 DVMuxContext *dvc = s->priv_data;
395 AVDictionaryEntry *tcr = av_dict_get(s->metadata, "timecode", NULL, 0);
397 if (!dv_init_mux(s)) {
398 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
399 "Make sure that you supply exactly two streams:\n"
400 " video: 25fps or 29.97fps, audio: 2ch/48|44|32kHz/PCM\n"
401 " (50Mbps allows an optional second audio stream)\n");
404 rate.num = dvc->sys->ltc_divisor;
406 if (!tcr) { // no global timecode, look into the streams
408 for (i = 0; i < s->nb_streams; i++) {
409 tcr = av_dict_get(s->streams[i]->metadata, "timecode", NULL, 0);
414 if (tcr && av_timecode_init_from_string(&dvc->tc, rate, tcr->value, s) >= 0)
416 return av_timecode_init(&dvc->tc, rate, 0, 0, s);
419 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
424 fsize = dv_assemble_frame(s, s->priv_data, s->streams[pkt->stream_index],
425 pkt->data, pkt->size, &frame);
427 avio_write(s->pb, frame, fsize);
433 * We might end up with some extra A/V data without matching counterpart.
434 * E.g. video data without enough audio to write the complete frame.
435 * Currently we simply drop the last frame. I don't know whether this
436 * is the best strategy of all
438 static int dv_write_trailer(struct AVFormatContext *s)
440 dv_delete_mux(s->priv_data);
444 AVOutputFormat ff_dv_muxer = {
446 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
448 .priv_data_size = sizeof(DVMuxContext),
449 .audio_codec = AV_CODEC_ID_PCM_S16LE,
450 .video_codec = AV_CODEC_ID_DVVIDEO,
451 .write_header = dv_write_header,
452 .write_packet = dv_write_packet,
453 .write_trailer = dv_write_trailer,