dvenc.c 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420
  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 FFmpeg.
  15. *
  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.
  20. *
  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.
  25. *
  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
  29. */
  30. #include <time.h>
  31. #include <stdarg.h>
  32. #include "avformat.h"
  33. #include "internal.h"
  34. #include "libavcodec/dvdata.h"
  35. #include "dv.h"
  36. #include "libavutil/fifo.h"
  37. #include "libavutil/mathematics.h"
  38. struct DVMuxContext {
  39. const DVprofile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
  40. int n_ast; /* number of stereo audio streams (up to 2) */
  41. AVStream *ast[2]; /* stereo audio streams */
  42. AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
  43. int frames; /* current frame number */
  44. int64_t start_time; /* recording start time */
  45. int has_audio; /* frame under contruction has audio */
  46. int has_video; /* frame under contruction has video */
  47. uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
  48. };
  49. static const int dv_aaux_packs_dist[12][9] = {
  50. { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  51. { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  52. { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  53. { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  54. { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  55. { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  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. };
  63. static int dv_audio_frame_size(const DVprofile* sys, int frame)
  64. {
  65. return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
  66. sizeof(sys->audio_samples_dist[0]))];
  67. }
  68. static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
  69. {
  70. struct tm tc;
  71. time_t ct;
  72. int ltc_frame;
  73. va_list ap;
  74. buf[0] = (uint8_t)pack_id;
  75. switch (pack_id) {
  76. case dv_timecode:
  77. ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
  78. c->sys->time_base.den, AV_ROUND_DOWN);
  79. brktimegm(ct, &tc);
  80. /*
  81. * LTC drop-frame frame counter drops two frames (0 and 1) every
  82. * minute, unless it is exactly divisible by 10
  83. */
  84. ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
  85. buf[1] = (0 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
  86. (1 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
  87. ((ltc_frame / 10) << 4) | /* tens of frames */
  88. (ltc_frame % 10); /* units of frames */
  89. buf[2] = (1 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
  90. ((tc.tm_sec / 10) << 4) | /* tens of seconds */
  91. (tc.tm_sec % 10); /* units of seconds */
  92. buf[3] = (1 << 7) | /* binary group flag BGF0 */
  93. ((tc.tm_min / 10) << 4) | /* tens of minutes */
  94. (tc.tm_min % 10); /* units of minutes */
  95. buf[4] = (1 << 7) | /* binary group flag BGF2 */
  96. (1 << 6) | /* binary group flag BGF1 */
  97. ((tc.tm_hour / 10) << 4) | /* tens of hours */
  98. (tc.tm_hour % 10); /* units of hours */
  99. break;
  100. case dv_audio_source: /* AAUX source pack */
  101. va_start(ap, buf);
  102. buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
  103. (1 << 6) | /* reserved -- always 1 */
  104. (dv_audio_frame_size(c->sys, c->frames) -
  105. c->sys->audio_min_samples[0]);
  106. /* # of samples */
  107. buf[2] = (0 << 7) | /* multi-stereo */
  108. (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
  109. (0 << 4) | /* pair bit: 0 -- one pair of channels */
  110. !!va_arg(ap, int); /* audio mode */
  111. buf[3] = (1 << 7) | /* res */
  112. (1 << 6) | /* multi-language flag */
  113. (c->sys->dsf << 5) | /* system: 60fields/50fields */
  114. (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
  115. buf[4] = (1 << 7) | /* emphasis: 1 -- off */
  116. (0 << 6) | /* emphasis time constant: 0 -- reserved */
  117. (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
  118. 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
  119. va_end(ap);
  120. break;
  121. case dv_audio_control:
  122. buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
  123. (1 << 4) | /* input source: 1 -- digital input */
  124. (3 << 2) | /* compression: 3 -- no information */
  125. 0; /* misc. info/SMPTE emphasis off */
  126. buf[2] = (1 << 7) | /* recording start point: 1 -- no */
  127. (1 << 6) | /* recording end point: 1 -- no */
  128. (1 << 3) | /* recording mode: 1 -- original */
  129. 7;
  130. buf[3] = (1 << 7) | /* direction: 1 -- forward */
  131. (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
  132. c->sys->ltc_divisor * 4);
  133. buf[4] = (1 << 7) | /* reserved -- always 1 */
  134. 0x7f; /* genre category */
  135. break;
  136. case dv_audio_recdate:
  137. case dv_video_recdate: /* VAUX recording date */
  138. ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
  139. c->sys->time_base.den, AV_ROUND_DOWN);
  140. brktimegm(ct, &tc);
  141. buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
  142. /* 0xff is very likely to be "unknown" */
  143. buf[2] = (3 << 6) | /* reserved -- always 1 */
  144. ((tc.tm_mday / 10) << 4) | /* Tens of day */
  145. (tc.tm_mday % 10); /* Units of day */
  146. buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
  147. ((tc.tm_mon / 10) << 4) | /* Tens of month */
  148. (tc.tm_mon % 10); /* Units of month */
  149. buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
  150. (tc.tm_year % 10); /* Units of year */
  151. break;
  152. case dv_audio_rectime: /* AAUX recording time */
  153. case dv_video_rectime: /* VAUX recording time */
  154. ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
  155. c->sys->time_base.den, AV_ROUND_DOWN);
  156. brktimegm(ct, &tc);
  157. buf[1] = (3 << 6) | /* reserved -- always 1 */
  158. 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
  159. buf[2] = (1 << 7) | /* reserved -- always 1 */
  160. ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
  161. (tc.tm_sec % 10); /* Units of seconds */
  162. buf[3] = (1 << 7) | /* reserved -- always 1 */
  163. ((tc.tm_min / 10) << 4) | /* Tens of minutes */
  164. (tc.tm_min % 10); /* Units of minutes */
  165. buf[4] = (3 << 6) | /* reserved -- always 1 */
  166. ((tc.tm_hour / 10) << 4) | /* Tens of hours */
  167. (tc.tm_hour % 10); /* Units of hours */
  168. break;
  169. default:
  170. buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
  171. }
  172. return 5;
  173. }
  174. static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
  175. {
  176. int i, j, d, of, size;
  177. size = 4 * dv_audio_frame_size(c->sys, c->frames);
  178. frame_ptr += channel * c->sys->difseg_size * 150 * 80;
  179. for (i = 0; i < c->sys->difseg_size; i++) {
  180. frame_ptr += 6 * 80; /* skip DIF segment header */
  181. for (j = 0; j < 9; j++) {
  182. dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
  183. for (d = 8; d < 80; d+=2) {
  184. of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
  185. if (of*2 >= size)
  186. continue;
  187. frame_ptr[d] = av_fifo_peek(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
  188. frame_ptr[d+1] = av_fifo_peek(c->audio_data[channel], of*2); // that DV is a big-endian PCM
  189. }
  190. frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
  191. }
  192. }
  193. }
  194. static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
  195. {
  196. int j, k;
  197. uint8_t* buf;
  198. for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
  199. /* DV subcode: 2nd and 3d DIFs */
  200. for (j = 80; j < 80 * 3; j += 80) {
  201. for (k = 6; k < 6 * 8; k += 8)
  202. dv_write_pack(dv_timecode, c, &buf[j+k]);
  203. 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 ? */
  204. dv_write_pack(dv_video_recdate, c, &buf[j+14]);
  205. dv_write_pack(dv_video_rectime, c, &buf[j+22]);
  206. dv_write_pack(dv_video_recdate, c, &buf[j+38]);
  207. dv_write_pack(dv_video_rectime, c, &buf[j+46]);
  208. }
  209. }
  210. /* DV VAUX: 4th, 5th and 6th 3DIFs */
  211. for (j = 80*3 + 3; j < 80*6; j += 80) {
  212. dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
  213. dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
  214. dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
  215. dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
  216. }
  217. }
  218. }
  219. /*
  220. * The following 3 functions constitute our interface to the world
  221. */
  222. static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
  223. uint8_t* data, int data_size, uint8_t** frame)
  224. {
  225. int i, reqasize;
  226. *frame = &c->frame_buf[0];
  227. reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
  228. switch (st->codec->codec_type) {
  229. case AVMEDIA_TYPE_VIDEO:
  230. /* FIXME: we have to have more sensible approach than this one */
  231. if (c->has_video)
  232. av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
  233. memcpy(*frame, data, c->sys->frame_size);
  234. c->has_video = 1;
  235. break;
  236. case AVMEDIA_TYPE_AUDIO:
  237. for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
  238. /* FIXME: we have to have more sensible approach than this one */
  239. if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
  240. av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
  241. av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
  242. /* Let us see if we've got enough audio for one DV frame. */
  243. c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
  244. break;
  245. default:
  246. break;
  247. }
  248. /* Let us see if we have enough data to construct one DV frame. */
  249. if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
  250. dv_inject_metadata(c, *frame);
  251. c->has_audio = 0;
  252. for (i=0; i < c->n_ast; i++) {
  253. dv_inject_audio(c, i, *frame);
  254. av_fifo_drain(c->audio_data[i], reqasize);
  255. c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
  256. }
  257. c->has_video = 0;
  258. c->frames++;
  259. return c->sys->frame_size;
  260. }
  261. return 0;
  262. }
  263. static DVMuxContext* dv_init_mux(AVFormatContext* s)
  264. {
  265. DVMuxContext *c = s->priv_data;
  266. AVStream *vst = NULL;
  267. AVDictionaryEntry *t;
  268. int i;
  269. /* we support at most 1 video and 2 audio streams */
  270. if (s->nb_streams > 3)
  271. return NULL;
  272. c->n_ast = 0;
  273. c->ast[0] = c->ast[1] = NULL;
  274. /* We have to sort out where audio and where video stream is */
  275. for (i=0; i<s->nb_streams; i++) {
  276. switch (s->streams[i]->codec->codec_type) {
  277. case AVMEDIA_TYPE_VIDEO:
  278. if (vst) return NULL;
  279. vst = s->streams[i];
  280. break;
  281. case AVMEDIA_TYPE_AUDIO:
  282. if (c->n_ast > 1) return NULL;
  283. c->ast[c->n_ast++] = s->streams[i];
  284. break;
  285. default:
  286. goto bail_out;
  287. }
  288. }
  289. /* Some checks -- DV format is very picky about its incoming streams */
  290. if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
  291. goto bail_out;
  292. for (i=0; i<c->n_ast; i++) {
  293. if (c->ast[i] && (c->ast[i]->codec->codec_id != CODEC_ID_PCM_S16LE ||
  294. c->ast[i]->codec->sample_rate != 48000 ||
  295. c->ast[i]->codec->channels != 2))
  296. goto bail_out;
  297. }
  298. c->sys = ff_dv_codec_profile(vst->codec);
  299. if (!c->sys)
  300. goto bail_out;
  301. if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
  302. /* only 1 stereo pair is allowed in 25Mbps mode */
  303. goto bail_out;
  304. }
  305. /* Ok, everything seems to be in working order */
  306. c->frames = 0;
  307. c->has_audio = 0;
  308. c->has_video = 0;
  309. #if FF_API_TIMESTAMP
  310. if (s->timestamp)
  311. c->start_time = s->timestamp;
  312. else
  313. #endif
  314. if (t = av_dict_get(s->metadata, "creation_time", NULL, 0))
  315. c->start_time = ff_iso8601_to_unix_time(t->value);
  316. for (i=0; i < c->n_ast; i++) {
  317. if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*AVCODEC_MAX_AUDIO_FRAME_SIZE))) {
  318. while (i > 0) {
  319. i--;
  320. av_fifo_free(c->audio_data[i]);
  321. }
  322. goto bail_out;
  323. }
  324. }
  325. return c;
  326. bail_out:
  327. return NULL;
  328. }
  329. static void dv_delete_mux(DVMuxContext *c)
  330. {
  331. int i;
  332. for (i=0; i < c->n_ast; i++)
  333. av_fifo_free(c->audio_data[i]);
  334. }
  335. static int dv_write_header(AVFormatContext *s)
  336. {
  337. if (!dv_init_mux(s)) {
  338. av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
  339. "Make sure that you supply exactly two streams:\n"
  340. " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
  341. " (50Mbps allows an optional second audio stream)\n");
  342. return -1;
  343. }
  344. return 0;
  345. }
  346. static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
  347. {
  348. uint8_t* frame;
  349. int fsize;
  350. fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
  351. pkt->data, pkt->size, &frame);
  352. if (fsize > 0) {
  353. avio_write(s->pb, frame, fsize);
  354. avio_flush(s->pb);
  355. }
  356. return 0;
  357. }
  358. /*
  359. * We might end up with some extra A/V data without matching counterpart.
  360. * E.g. video data without enough audio to write the complete frame.
  361. * Currently we simply drop the last frame. I don't know whether this
  362. * is the best strategy of all
  363. */
  364. static int dv_write_trailer(struct AVFormatContext *s)
  365. {
  366. dv_delete_mux(s->priv_data);
  367. return 0;
  368. }
  369. AVOutputFormat ff_dv_muxer = {
  370. .name = "dv",
  371. .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
  372. .extensions = "dv",
  373. .priv_data_size = sizeof(DVMuxContext),
  374. .audio_codec = CODEC_ID_PCM_S16LE,
  375. .video_codec = CODEC_ID_DVVIDEO,
  376. .write_header = dv_write_header,
  377. .write_packet = dv_write_packet,
  378. .write_trailer = dv_write_trailer,
  379. };