lavc: export DV profile API used by muxer/demuxer as public
[libav.git] / libavformat / dvenc.c
1 /*
2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnik
4 *
5 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
6 * of DV technical info.
7 *
8 * Raw DV format
9 * Copyright (c) 2002 Fabrice Bellard
10 *
11 * 50 Mbps (DVCPRO50) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
13 *
14 * This file is part of Libav.
15 *
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.
20 *
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.
25 *
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
29 */
30 #include <time.h>
31 #include <stdarg.h>
32
33 #include "avformat.h"
34 #include "internal.h"
35 #include "libavcodec/dv_profile.h"
36 #include "libavcodec/dv.h"
37 #include "dv.h"
38 #include "libavutil/fifo.h"
39 #include "libavutil/mathematics.h"
40
41 #define MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32bit audio
42
43 struct DVMuxContext {
44 const AVDVProfile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
45 int n_ast; /* number of stereo audio streams (up to 2) */
46 AVStream *ast[2]; /* stereo audio streams */
47 AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
48 int frames; /* current frame number */
49 int64_t start_time; /* recording start time */
50 int has_audio; /* frame under construction has audio */
51 int has_video; /* frame under construction has video */
52 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under construction */
53 };
54
55 static const int dv_aaux_packs_dist[12][9] = {
56 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
57 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
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 };
69
70 static int dv_audio_frame_size(const AVDVProfile* sys, int frame)
71 {
72 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
73 sizeof(sys->audio_samples_dist[0]))];
74 }
75
76 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
77 {
78 struct tm tc;
79 time_t ct;
80 int ltc_frame;
81 va_list ap;
82
83 buf[0] = (uint8_t)pack_id;
84 switch (pack_id) {
85 case dv_timecode:
86 ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
87 c->sys->time_base.den, AV_ROUND_DOWN);
88 ff_brktimegm(ct, &tc);
89 /*
90 * LTC drop-frame frame counter drops two frames (0 and 1) every
91 * minute, unless it is exactly divisible by 10
92 */
93 ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
94 buf[1] = (0 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
95 (1 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
96 ((ltc_frame / 10) << 4) | /* tens of frames */
97 (ltc_frame % 10); /* units of frames */
98 buf[2] = (1 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
99 ((tc.tm_sec / 10) << 4) | /* tens of seconds */
100 (tc.tm_sec % 10); /* units of seconds */
101 buf[3] = (1 << 7) | /* binary group flag BGF0 */
102 ((tc.tm_min / 10) << 4) | /* tens of minutes */
103 (tc.tm_min % 10); /* units of minutes */
104 buf[4] = (1 << 7) | /* binary group flag BGF2 */
105 (1 << 6) | /* binary group flag BGF1 */
106 ((tc.tm_hour / 10) << 4) | /* tens of hours */
107 (tc.tm_hour % 10); /* units of hours */
108 break;
109 case dv_audio_source: /* AAUX source pack */
110 va_start(ap, buf);
111 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
112 (1 << 6) | /* reserved -- always 1 */
113 (dv_audio_frame_size(c->sys, c->frames) -
114 c->sys->audio_min_samples[0]);
115 /* # of samples */
116 buf[2] = (0 << 7) | /* multi-stereo */
117 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
118 (0 << 4) | /* pair bit: 0 -- one pair of channels */
119 !!va_arg(ap, int); /* audio mode */
120 buf[3] = (1 << 7) | /* res */
121 (1 << 6) | /* multi-language flag */
122 (c->sys->dsf << 5) | /* system: 60fields/50fields */
123 (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
124 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
125 (0 << 6) | /* emphasis time constant: 0 -- reserved */
126 (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
127 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
128 va_end(ap);
129 break;
130 case dv_audio_control:
131 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
132 (1 << 4) | /* input source: 1 -- digital input */
133 (3 << 2) | /* compression: 3 -- no information */
134 0; /* misc. info/SMPTE emphasis off */
135 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
136 (1 << 6) | /* recording end point: 1 -- no */
137 (1 << 3) | /* recording mode: 1 -- original */
138 7;
139 buf[3] = (1 << 7) | /* direction: 1 -- forward */
140 (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0x20 : /* speed */
141 c->sys->ltc_divisor * 4);
142 buf[4] = (1 << 7) | /* reserved -- always 1 */
143 0x7f; /* genre category */
144 break;
145 case dv_audio_recdate:
146 case dv_video_recdate: /* VAUX recording date */
147 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
148 c->sys->time_base.den, AV_ROUND_DOWN);
149 ff_brktimegm(ct, &tc);
150 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
151 /* 0xff is very likely to be "unknown" */
152 buf[2] = (3 << 6) | /* reserved -- always 1 */
153 ((tc.tm_mday / 10) << 4) | /* Tens of day */
154 (tc.tm_mday % 10); /* Units of day */
155 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
156 ((tc.tm_mon / 10) << 4) | /* Tens of month */
157 (tc.tm_mon % 10); /* Units of month */
158 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
159 (tc.tm_year % 10); /* Units of year */
160 break;
161 case dv_audio_rectime: /* AAUX recording time */
162 case dv_video_rectime: /* VAUX recording time */
163 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
164 c->sys->time_base.den, AV_ROUND_DOWN);
165 ff_brktimegm(ct, &tc);
166 buf[1] = (3 << 6) | /* reserved -- always 1 */
167 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
168 buf[2] = (1 << 7) | /* reserved -- always 1 */
169 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
170 (tc.tm_sec % 10); /* Units of seconds */
171 buf[3] = (1 << 7) | /* reserved -- always 1 */
172 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
173 (tc.tm_min % 10); /* Units of minutes */
174 buf[4] = (3 << 6) | /* reserved -- always 1 */
175 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
176 (tc.tm_hour % 10); /* Units of hours */
177 break;
178 default:
179 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
180 }
181 return 5;
182 }
183
184 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
185 {
186 int i, j, d, of, size;
187 size = 4 * dv_audio_frame_size(c->sys, c->frames);
188 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
189 for (i = 0; i < c->sys->difseg_size; i++) {
190 frame_ptr += 6 * 80; /* skip DIF segment header */
191 for (j = 0; j < 9; j++) {
192 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
193 for (d = 8; d < 80; d+=2) {
194 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
195 if (of*2 >= size)
196 continue;
197
198 frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
199 frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
200 }
201 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
202 }
203 }
204 }
205
206 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
207 {
208 int j, k;
209 uint8_t* buf;
210
211 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
212 /* DV subcode: 2nd and 3d DIFs */
213 for (j = 80; j < 80 * 3; j += 80) {
214 for (k = 6; k < 6 * 8; k += 8)
215 dv_write_pack(dv_timecode, c, &buf[j+k]);
216
217 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 ? */
218 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
219 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
220 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
221 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
222 }
223 }
224
225 /* DV VAUX: 4th, 5th and 6th 3DIFs */
226 for (j = 80*3 + 3; j < 80*6; j += 80) {
227 dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
228 dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
229 dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
230 dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
231 }
232 }
233 }
234
235 /*
236 * The following 3 functions constitute our interface to the world
237 */
238
239 static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
240 uint8_t* data, int data_size, uint8_t** frame)
241 {
242 int i, reqasize;
243
244 *frame = &c->frame_buf[0];
245 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
246
247 switch (st->codec->codec_type) {
248 case AVMEDIA_TYPE_VIDEO:
249 /* FIXME: we have to have more sensible approach than this one */
250 if (c->has_video)
251 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
252
253 memcpy(*frame, data, c->sys->frame_size);
254 c->has_video = 1;
255 break;
256 case AVMEDIA_TYPE_AUDIO:
257 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
258
259 /* FIXME: we have to have more sensible approach than this one */
260 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*MAX_AUDIO_FRAME_SIZE)
261 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
262 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
263
264 /* Let us see if we've got enough audio for one DV frame. */
265 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
266
267 break;
268 default:
269 break;
270 }
271
272 /* Let us see if we have enough data to construct one DV frame. */
273 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
274 dv_inject_metadata(c, *frame);
275 c->has_audio = 0;
276 for (i=0; i < c->n_ast; i++) {
277 dv_inject_audio(c, i, *frame);
278 av_fifo_drain(c->audio_data[i], reqasize);
279 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
280 }
281
282 c->has_video = 0;
283
284 c->frames++;
285
286 return c->sys->frame_size;
287 }
288
289 return 0;
290 }
291
292 static DVMuxContext* dv_init_mux(AVFormatContext* s)
293 {
294 DVMuxContext *c = s->priv_data;
295 AVStream *vst = NULL;
296 AVDictionaryEntry *t;
297 int i;
298
299 /* we support at most 1 video and 2 audio streams */
300 if (s->nb_streams > 3)
301 return NULL;
302
303 c->n_ast = 0;
304 c->ast[0] = c->ast[1] = NULL;
305
306 /* We have to sort out where audio and where video stream is */
307 for (i=0; i<s->nb_streams; i++) {
308 switch (s->streams[i]->codec->codec_type) {
309 case AVMEDIA_TYPE_VIDEO:
310 if (vst) return NULL;
311 vst = s->streams[i];
312 break;
313 case AVMEDIA_TYPE_AUDIO:
314 if (c->n_ast > 1) return NULL;
315 c->ast[c->n_ast++] = s->streams[i];
316 break;
317 default:
318 goto bail_out;
319 }
320 }
321
322 /* Some checks -- DV format is very picky about its incoming streams */
323 if (!vst || vst->codec->codec_id != AV_CODEC_ID_DVVIDEO)
324 goto bail_out;
325 for (i=0; i<c->n_ast; i++) {
326 if (c->ast[i] && (c->ast[i]->codec->codec_id != AV_CODEC_ID_PCM_S16LE ||
327 c->ast[i]->codec->sample_rate != 48000 ||
328 c->ast[i]->codec->channels != 2))
329 goto bail_out;
330 }
331 c->sys = av_dv_codec_profile(vst->codec->width, vst->codec->height, vst->codec->pix_fmt);
332 if (!c->sys)
333 goto bail_out;
334
335 if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
336 /* only 1 stereo pair is allowed in 25Mbps mode */
337 goto bail_out;
338 }
339
340 /* Ok, everything seems to be in working order */
341 c->frames = 0;
342 c->has_audio = 0;
343 c->has_video = 0;
344 if (t = av_dict_get(s->metadata, "creation_time", NULL, 0))
345 c->start_time = ff_iso8601_to_unix_time(t->value);
346
347 for (i=0; i < c->n_ast; i++) {
348 if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*MAX_AUDIO_FRAME_SIZE))) {
349 while (i > 0) {
350 i--;
351 av_fifo_free(c->audio_data[i]);
352 }
353 goto bail_out;
354 }
355 }
356
357 return c;
358
359 bail_out:
360 return NULL;
361 }
362
363 static void dv_delete_mux(DVMuxContext *c)
364 {
365 int i;
366 for (i=0; i < c->n_ast; i++)
367 av_fifo_free(c->audio_data[i]);
368 }
369
370 static int dv_write_header(AVFormatContext *s)
371 {
372 if (!dv_init_mux(s)) {
373 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
374 "Make sure that you supply exactly two streams:\n"
375 " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
376 " (50Mbps allows an optional second audio stream)\n");
377 return -1;
378 }
379 return 0;
380 }
381
382 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
383 {
384 uint8_t* frame;
385 int fsize;
386
387 fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
388 pkt->data, pkt->size, &frame);
389 if (fsize > 0) {
390 avio_write(s->pb, frame, fsize);
391 }
392 return 0;
393 }
394
395 /*
396 * We might end up with some extra A/V data without matching counterpart.
397 * E.g. video data without enough audio to write the complete frame.
398 * Currently we simply drop the last frame. I don't know whether this
399 * is the best strategy of all
400 */
401 static int dv_write_trailer(struct AVFormatContext *s)
402 {
403 dv_delete_mux(s->priv_data);
404 return 0;
405 }
406
407 AVOutputFormat ff_dv_muxer = {
408 .name = "dv",
409 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
410 .extensions = "dv",
411 .priv_data_size = sizeof(DVMuxContext),
412 .audio_codec = AV_CODEC_ID_PCM_S16LE,
413 .video_codec = AV_CODEC_ID_DVVIDEO,
414 .write_header = dv_write_header,
415 .write_packet = dv_write_packet,
416 .write_trailer = dv_write_trailer,
417 };