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 Libav.
16 * Libav 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 * Libav 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 Libav; 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"
37 #include "libavutil/fifo.h"
38 #include "libavutil/mathematics.h"
41 const DVprofile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
42 int n_ast; /* number of stereo audio streams (up to 2) */
43 AVStream *ast[2]; /* stereo audio streams */
44 AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
45 int frames; /* current frame number */
46 int64_t start_time; /* recording start time */
47 int has_audio; /* frame under contruction has audio */
48 int has_video; /* frame under contruction has video */
49 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
52 static const int dv_aaux_packs_dist[12][9] = {
53 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
54 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
55 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
56 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
57 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
58 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
59 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
60 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
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 },
67 static int dv_audio_frame_size(const DVprofile* sys, int frame)
69 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
70 sizeof(sys->audio_samples_dist[0]))];
73 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
80 buf[0] = (uint8_t)pack_id;
83 ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
84 c->sys->time_base.den, AV_ROUND_DOWN);
85 ff_brktimegm(ct, &tc);
87 * LTC drop-frame frame counter drops two frames (0 and 1) every
88 * minute, unless it is exactly divisible by 10
90 ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
91 buf[1] = (0 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
92 (1 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
93 ((ltc_frame / 10) << 4) | /* tens of frames */
94 (ltc_frame % 10); /* units of frames */
95 buf[2] = (1 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
96 ((tc.tm_sec / 10) << 4) | /* tens of seconds */
97 (tc.tm_sec % 10); /* units of seconds */
98 buf[3] = (1 << 7) | /* binary group flag BGF0 */
99 ((tc.tm_min / 10) << 4) | /* tens of minutes */
100 (tc.tm_min % 10); /* units of minutes */
101 buf[4] = (1 << 7) | /* binary group flag BGF2 */
102 (1 << 6) | /* binary group flag BGF1 */
103 ((tc.tm_hour / 10) << 4) | /* tens of hours */
104 (tc.tm_hour % 10); /* units of hours */
106 case dv_audio_source: /* AAUX source pack */
108 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
109 (1 << 6) | /* reserved -- always 1 */
110 (dv_audio_frame_size(c->sys, c->frames) -
111 c->sys->audio_min_samples[0]);
113 buf[2] = (0 << 7) | /* multi-stereo */
114 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
115 (0 << 4) | /* pair bit: 0 -- one pair of channels */
116 !!va_arg(ap, int); /* audio mode */
117 buf[3] = (1 << 7) | /* res */
118 (1 << 6) | /* multi-language flag */
119 (c->sys->dsf << 5) | /* system: 60fields/50fields */
120 (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
121 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
122 (0 << 6) | /* emphasis time constant: 0 -- reserved */
123 (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
124 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
127 case dv_audio_control:
128 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
129 (1 << 4) | /* input source: 1 -- digital input */
130 (3 << 2) | /* compression: 3 -- no information */
131 0; /* misc. info/SMPTE emphasis off */
132 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
133 (1 << 6) | /* recording end point: 1 -- no */
134 (1 << 3) | /* recording mode: 1 -- original */
136 buf[3] = (1 << 7) | /* direction: 1 -- forward */
137 (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
138 c->sys->ltc_divisor * 4);
139 buf[4] = (1 << 7) | /* reserved -- always 1 */
140 0x7f; /* genre category */
142 case dv_audio_recdate:
143 case dv_video_recdate: /* VAUX recording date */
144 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
145 c->sys->time_base.den, AV_ROUND_DOWN);
146 ff_brktimegm(ct, &tc);
147 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
148 /* 0xff is very likely to be "unknown" */
149 buf[2] = (3 << 6) | /* reserved -- always 1 */
150 ((tc.tm_mday / 10) << 4) | /* Tens of day */
151 (tc.tm_mday % 10); /* Units of day */
152 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
153 ((tc.tm_mon / 10) << 4) | /* Tens of month */
154 (tc.tm_mon % 10); /* Units of month */
155 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
156 (tc.tm_year % 10); /* Units of year */
158 case dv_audio_rectime: /* AAUX recording time */
159 case dv_video_rectime: /* VAUX recording time */
160 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
161 c->sys->time_base.den, AV_ROUND_DOWN);
162 ff_brktimegm(ct, &tc);
163 buf[1] = (3 << 6) | /* reserved -- always 1 */
164 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
165 buf[2] = (1 << 7) | /* reserved -- always 1 */
166 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
167 (tc.tm_sec % 10); /* Units of seconds */
168 buf[3] = (1 << 7) | /* reserved -- always 1 */
169 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
170 (tc.tm_min % 10); /* Units of minutes */
171 buf[4] = (3 << 6) | /* reserved -- always 1 */
172 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
173 (tc.tm_hour % 10); /* Units of hours */
176 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
181 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
183 int i, j, d, of, size;
184 size = 4 * dv_audio_frame_size(c->sys, c->frames);
185 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
186 for (i = 0; i < c->sys->difseg_size; i++) {
187 frame_ptr += 6 * 80; /* skip DIF segment header */
188 for (j = 0; j < 9; j++) {
189 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
190 for (d = 8; d < 80; d+=2) {
191 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
195 frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
196 frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
198 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
203 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
208 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
209 /* DV subcode: 2nd and 3d DIFs */
210 for (j = 80; j < 80 * 3; j += 80) {
211 for (k = 6; k < 6 * 8; k += 8)
212 dv_write_pack(dv_timecode, c, &buf[j+k]);
214 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 ? */
215 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
216 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
217 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
218 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
222 /* DV VAUX: 4th, 5th and 6th 3DIFs */
223 for (j = 80*3 + 3; j < 80*6; j += 80) {
224 dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
225 dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
226 dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
227 dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
233 * The following 3 functions constitute our interface to the world
236 static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
237 uint8_t* data, int data_size, uint8_t** frame)
241 *frame = &c->frame_buf[0];
242 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
244 switch (st->codec->codec_type) {
245 case AVMEDIA_TYPE_VIDEO:
246 /* FIXME: we have to have more sensible approach than this one */
248 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
250 memcpy(*frame, data, c->sys->frame_size);
253 case AVMEDIA_TYPE_AUDIO:
254 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
256 /* FIXME: we have to have more sensible approach than this one */
257 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
258 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
259 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
261 /* Let us see if we've got enough audio for one DV frame. */
262 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
269 /* Let us see if we have enough data to construct one DV frame. */
270 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
271 dv_inject_metadata(c, *frame);
273 for (i=0; i < c->n_ast; i++) {
274 dv_inject_audio(c, i, *frame);
275 av_fifo_drain(c->audio_data[i], reqasize);
276 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
283 return c->sys->frame_size;
289 static DVMuxContext* dv_init_mux(AVFormatContext* s)
291 DVMuxContext *c = s->priv_data;
292 AVStream *vst = NULL;
293 AVDictionaryEntry *t;
296 /* we support at most 1 video and 2 audio streams */
297 if (s->nb_streams > 3)
301 c->ast[0] = c->ast[1] = NULL;
303 /* We have to sort out where audio and where video stream is */
304 for (i=0; i<s->nb_streams; i++) {
305 switch (s->streams[i]->codec->codec_type) {
306 case AVMEDIA_TYPE_VIDEO:
307 if (vst) return NULL;
310 case AVMEDIA_TYPE_AUDIO:
311 if (c->n_ast > 1) return NULL;
312 c->ast[c->n_ast++] = s->streams[i];
319 /* Some checks -- DV format is very picky about its incoming streams */
320 if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
322 for (i=0; i<c->n_ast; i++) {
323 if (c->ast[i] && (c->ast[i]->codec->codec_id != CODEC_ID_PCM_S16LE ||
324 c->ast[i]->codec->sample_rate != 48000 ||
325 c->ast[i]->codec->channels != 2))
328 c->sys = avpriv_dv_codec_profile(vst->codec);
332 if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
333 /* only 1 stereo pair is allowed in 25Mbps mode */
337 /* Ok, everything seems to be in working order */
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 if (!dv_init_mux(s)) {
370 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
371 "Make sure that you supply exactly two streams:\n"
372 " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
373 " (50Mbps allows an optional second audio stream)\n");
379 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
384 fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
385 pkt->data, pkt->size, &frame);
387 avio_write(s->pb, frame, fsize);
394 * We might end up with some extra A/V data without matching counterpart.
395 * E.g. video data without enough audio to write the complete frame.
396 * Currently we simply drop the last frame. I don't know whether this
397 * is the best strategy of all
399 static int dv_write_trailer(struct AVFormatContext *s)
401 dv_delete_mux(s->priv_data);
405 AVOutputFormat ff_dv_muxer = {
407 .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
409 .priv_data_size = sizeof(DVMuxContext),
410 .audio_codec = CODEC_ID_PCM_S16LE,
411 .video_codec = CODEC_ID_DVVIDEO,
412 .write_header = dv_write_header,
413 .write_packet = dv_write_packet,
414 .write_trailer = dv_write_trailer,