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/dvdata.h"
36 #include "libavcodec/timecode.h"
38 #include "libavutil/fifo.h"
39 #include "libavutil/mathematics.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/opt.h"
45 const DVprofile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
46 int n_ast; /* number of stereo audio streams (up to 2) */
47 AVStream *ast[2]; /* stereo audio streams */
48 AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
49 int frames; /* current frame number */
50 int64_t start_time; /* recording start time */
51 int has_audio; /* frame under contruction has audio */
52 int has_video; /* frame under contruction has video */
53 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
54 struct ff_timecode tc;
57 static const int dv_aaux_packs_dist[12][9] = {
58 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
59 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
60 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
61 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
62 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
63 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
64 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
65 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
66 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
67 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
68 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
69 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
72 static int dv_audio_frame_size(const DVprofile* sys, int frame)
74 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
75 sizeof(sys->audio_samples_dist[0]))];
78 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
86 buf[0] = (uint8_t)pack_id;
90 * LTC drop-frame frame counter drops two frames (0 and 1) every
91 * minute, unless it is exactly divisible by 10
93 ltc_frame = c->tc.start + c->frames;
95 ltc_frame = avpriv_framenum_to_drop_timecode(ltc_frame);
96 timecode = avpriv_framenum_to_smpte_timecode(ltc_frame, c->sys->ltc_divisor,
98 timecode |= 1<<23 | 1<<15 | 1<<7 | 1<<6; // biphase and binary group flags
99 AV_WB32(buf + 1, timecode);
101 case dv_audio_source: /* AAUX source pack */
103 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
104 (1 << 6) | /* reserved -- always 1 */
105 (dv_audio_frame_size(c->sys, c->frames) -
106 c->sys->audio_min_samples[0]);
108 buf[2] = (0 << 7) | /* multi-stereo */
109 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
110 (0 << 4) | /* pair bit: 0 -- one pair of channels */
111 !!va_arg(ap, int); /* audio mode */
112 buf[3] = (1 << 7) | /* res */
113 (1 << 6) | /* multi-language flag */
114 (c->sys->dsf << 5) | /* system: 60fields/50fields */
115 (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
116 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
117 (0 << 6) | /* emphasis time constant: 0 -- reserved */
118 (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
119 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
122 case dv_audio_control:
123 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
124 (1 << 4) | /* input source: 1 -- digital input */
125 (3 << 2) | /* compression: 3 -- no information */
126 0; /* misc. info/SMPTE emphasis off */
127 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
128 (1 << 6) | /* recording end point: 1 -- no */
129 (1 << 3) | /* recording mode: 1 -- original */
131 buf[3] = (1 << 7) | /* direction: 1 -- forward */
132 (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
133 c->sys->ltc_divisor * 4);
134 buf[4] = (1 << 7) | /* reserved -- always 1 */
135 0x7f; /* genre category */
137 case dv_audio_recdate:
138 case dv_video_recdate: /* VAUX recording date */
139 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
140 c->sys->time_base.den, AV_ROUND_DOWN);
142 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
143 /* 0xff is very likely to be "unknown" */
144 buf[2] = (3 << 6) | /* reserved -- always 1 */
145 ((tc.tm_mday / 10) << 4) | /* Tens of day */
146 (tc.tm_mday % 10); /* Units of day */
147 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
148 ((tc.tm_mon / 10) << 4) | /* Tens of month */
149 (tc.tm_mon % 10); /* Units of month */
150 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
151 (tc.tm_year % 10); /* Units of year */
153 case dv_audio_rectime: /* AAUX recording time */
154 case dv_video_rectime: /* VAUX recording time */
155 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
156 c->sys->time_base.den, AV_ROUND_DOWN);
158 buf[1] = (3 << 6) | /* reserved -- always 1 */
159 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
160 buf[2] = (1 << 7) | /* reserved -- always 1 */
161 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
162 (tc.tm_sec % 10); /* Units of seconds */
163 buf[3] = (1 << 7) | /* reserved -- always 1 */
164 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
165 (tc.tm_min % 10); /* Units of minutes */
166 buf[4] = (3 << 6) | /* reserved -- always 1 */
167 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
168 (tc.tm_hour % 10); /* Units of hours */
171 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
176 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
178 int i, j, d, of, size;
179 size = 4 * dv_audio_frame_size(c->sys, c->frames);
180 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
181 for (i = 0; i < c->sys->difseg_size; i++) {
182 frame_ptr += 6 * 80; /* skip DIF segment header */
183 for (j = 0; j < 9; j++) {
184 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
185 for (d = 8; d < 80; d+=2) {
186 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
190 frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
191 frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
193 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
198 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
203 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
204 /* DV subcode: 2nd and 3d DIFs */
205 for (j = 80; j < 80 * 3; j += 80) {
206 for (k = 6; k < 6 * 8; k += 8)
207 dv_write_pack(dv_timecode, c, &buf[j+k]);
209 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 ? */
210 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
211 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
212 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
213 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
217 /* DV VAUX: 4th, 5th and 6th 3DIFs */
218 for (j = 80*3 + 3; j < 80*6; j += 80) {
219 dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
220 dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
221 dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
222 dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
228 * The following 3 functions constitute our interface to the world
231 static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
232 uint8_t* data, int data_size, uint8_t** frame)
236 *frame = &c->frame_buf[0];
237 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
239 switch (st->codec->codec_type) {
240 case AVMEDIA_TYPE_VIDEO:
241 /* FIXME: we have to have more sensible approach than this one */
243 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
245 memcpy(*frame, data, c->sys->frame_size);
248 case AVMEDIA_TYPE_AUDIO:
249 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
251 /* FIXME: we have to have more sensible approach than this one */
252 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
253 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
254 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
256 /* Let us see if we've got enough audio for one DV frame. */
257 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
264 /* Let us see if we have enough data to construct one DV frame. */
265 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
266 dv_inject_metadata(c, *frame);
268 for (i=0; i < c->n_ast; i++) {
269 dv_inject_audio(c, i, *frame);
270 av_fifo_drain(c->audio_data[i], reqasize);
271 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
278 return c->sys->frame_size;
284 static DVMuxContext* dv_init_mux(AVFormatContext* s)
286 DVMuxContext *c = s->priv_data;
287 AVStream *vst = NULL;
288 AVDictionaryEntry *t;
291 /* we support at most 1 video and 2 audio streams */
292 if (s->nb_streams > 3)
296 c->ast[0] = c->ast[1] = NULL;
298 /* We have to sort out where audio and where video stream is */
299 for (i=0; i<s->nb_streams; i++) {
300 switch (s->streams[i]->codec->codec_type) {
301 case AVMEDIA_TYPE_VIDEO:
302 if (vst) return NULL;
305 case AVMEDIA_TYPE_AUDIO:
306 if (c->n_ast > 1) return NULL;
307 c->ast[c->n_ast++] = s->streams[i];
314 /* Some checks -- DV format is very picky about its incoming streams */
315 if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
317 for (i=0; i<c->n_ast; i++) {
318 if (c->ast[i] && (c->ast[i]->codec->codec_id != CODEC_ID_PCM_S16LE ||
319 c->ast[i]->codec->sample_rate != 48000 ||
320 c->ast[i]->codec->channels != 2))
323 c->sys = avpriv_dv_codec_profile(vst->codec);
327 if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
328 /* only 1 stereo pair is allowed in 25Mbps mode */
332 /* Ok, everything seems to be in working order */
338 c->start_time = s->timestamp;
341 if (t = av_dict_get(s->metadata, "creation_time", NULL, 0))
342 c->start_time = ff_iso8601_to_unix_time(t->value);
344 for (i=0; i < c->n_ast; i++) {
345 if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*AVCODEC_MAX_AUDIO_FRAME_SIZE))) {
348 av_fifo_free(c->audio_data[i]);
360 static void dv_delete_mux(DVMuxContext *c)
363 for (i=0; i < c->n_ast; i++)
364 av_fifo_free(c->audio_data[i]);
367 static int dv_write_header(AVFormatContext *s)
369 DVMuxContext *dvc = s->priv_data;
371 if (!dv_init_mux(s)) {
372 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
373 "Make sure that you supply exactly two streams:\n"
374 " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
375 " (50Mbps allows an optional second audio stream)\n");
379 dvc->tc.rate.num = dvc->sys->time_base.den;
380 dvc->tc.rate.den = dvc->sys->time_base.num;
381 if (avpriv_init_smpte_timecode(s, &dvc->tc) < 0)
387 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
392 fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
393 pkt->data, pkt->size, &frame);
395 avio_write(s->pb, frame, fsize);
402 * We might end up with some extra A/V data without matching counterpart.
403 * E.g. video data without enough audio to write the complete frame.
404 * Currently we simply drop the last frame. I don't know whether this
405 * is the best strategy of all
407 static int dv_write_trailer(struct AVFormatContext *s)
409 dv_delete_mux(s->priv_data);
413 static const AVClass class = {
415 .item_name = av_default_item_name,
416 .version = LIBAVUTIL_VERSION_INT,
417 .option = (const AVOption[]){
418 {TIMECODE_OPT(DVMuxContext, AV_OPT_FLAG_ENCODING_PARAM)},
423 AVOutputFormat ff_dv_muxer = {
425 .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
427 .priv_data_size = sizeof(DVMuxContext),
428 .audio_codec = CODEC_ID_PCM_S16LE,
429 .video_codec = CODEC_ID_DVVIDEO,
430 .write_header = dv_write_header,
431 .write_packet = dv_write_packet,
432 .write_trailer = dv_write_trailer,
433 .priv_class = &class,