You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

418 lines
16KB

  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. #include "avformat.h"
  33. #include "internal.h"
  34. #include "libavcodec/dv_profile.h"
  35. #include "libavcodec/dv.h"
  36. #include "dv.h"
  37. #include "libavutil/fifo.h"
  38. #include "libavutil/mathematics.h"
  39. #define MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32bit audio
  40. struct DVMuxContext {
  41. const AVDVProfile* 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 construction has audio */
  48. int has_video; /* frame under construction has video */
  49. uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under construction */
  50. };
  51. static const int dv_aaux_packs_dist[12][9] = {
  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. { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  63. { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  64. };
  65. static int dv_audio_frame_size(const AVDVProfile* sys, int frame)
  66. {
  67. return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
  68. sizeof(sys->audio_samples_dist[0]))];
  69. }
  70. static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
  71. {
  72. struct tm tc;
  73. time_t ct;
  74. int ltc_frame;
  75. va_list ap;
  76. buf[0] = (uint8_t)pack_id;
  77. switch (pack_id) {
  78. case dv_timecode:
  79. ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
  80. c->sys->time_base.den, AV_ROUND_DOWN);
  81. ff_brktimegm(ct, &tc);
  82. /*
  83. * LTC drop-frame frame counter drops two frames (0 and 1) every
  84. * minute, unless it is exactly divisible by 10
  85. */
  86. ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
  87. buf[1] = (0 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
  88. (1 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
  89. ((ltc_frame / 10) << 4) | /* tens of frames */
  90. (ltc_frame % 10); /* units of frames */
  91. buf[2] = (1 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
  92. ((tc.tm_sec / 10) << 4) | /* tens of seconds */
  93. (tc.tm_sec % 10); /* units of seconds */
  94. buf[3] = (1 << 7) | /* binary group flag BGF0 */
  95. ((tc.tm_min / 10) << 4) | /* tens of minutes */
  96. (tc.tm_min % 10); /* units of minutes */
  97. buf[4] = (1 << 7) | /* binary group flag BGF2 */
  98. (1 << 6) | /* binary group flag BGF1 */
  99. ((tc.tm_hour / 10) << 4) | /* tens of hours */
  100. (tc.tm_hour % 10); /* units of hours */
  101. break;
  102. case dv_audio_source: /* AAUX source pack */
  103. va_start(ap, buf);
  104. buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
  105. (1 << 6) | /* reserved -- always 1 */
  106. (dv_audio_frame_size(c->sys, c->frames) -
  107. c->sys->audio_min_samples[0]);
  108. /* # of samples */
  109. buf[2] = (0 << 7) | /* multi-stereo */
  110. (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
  111. (0 << 4) | /* pair bit: 0 -- one pair of channels */
  112. !!va_arg(ap, int); /* audio mode */
  113. buf[3] = (1 << 7) | /* res */
  114. (1 << 6) | /* multi-language flag */
  115. (c->sys->dsf << 5) | /* system: 60fields/50fields */
  116. (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
  117. buf[4] = (1 << 7) | /* emphasis: 1 -- off */
  118. (0 << 6) | /* emphasis time constant: 0 -- reserved */
  119. (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
  120. 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
  121. va_end(ap);
  122. break;
  123. case dv_audio_control:
  124. buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
  125. (1 << 4) | /* input source: 1 -- digital input */
  126. (3 << 2) | /* compression: 3 -- no information */
  127. 0; /* misc. info/SMPTE emphasis off */
  128. buf[2] = (1 << 7) | /* recording start point: 1 -- no */
  129. (1 << 6) | /* recording end point: 1 -- no */
  130. (1 << 3) | /* recording mode: 1 -- original */
  131. 7;
  132. buf[3] = (1 << 7) | /* direction: 1 -- forward */
  133. (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0x20 : /* speed */
  134. c->sys->ltc_divisor * 4);
  135. buf[4] = (1 << 7) | /* reserved -- always 1 */
  136. 0x7f; /* genre category */
  137. break;
  138. case dv_audio_recdate:
  139. case dv_video_recdate: /* VAUX recording date */
  140. ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
  141. c->sys->time_base.den, AV_ROUND_DOWN);
  142. ff_brktimegm(ct, &tc);
  143. buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
  144. /* 0xff is very likely to be "unknown" */
  145. buf[2] = (3 << 6) | /* reserved -- always 1 */
  146. ((tc.tm_mday / 10) << 4) | /* Tens of day */
  147. (tc.tm_mday % 10); /* Units of day */
  148. buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
  149. ((tc.tm_mon / 10) << 4) | /* Tens of month */
  150. (tc.tm_mon % 10); /* Units of month */
  151. buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
  152. (tc.tm_year % 10); /* Units of year */
  153. break;
  154. case dv_audio_rectime: /* AAUX recording time */
  155. case dv_video_rectime: /* VAUX recording time */
  156. ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
  157. c->sys->time_base.den, AV_ROUND_DOWN);
  158. ff_brktimegm(ct, &tc);
  159. buf[1] = (3 << 6) | /* reserved -- always 1 */
  160. 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
  161. buf[2] = (1 << 7) | /* reserved -- always 1 */
  162. ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
  163. (tc.tm_sec % 10); /* Units of seconds */
  164. buf[3] = (1 << 7) | /* reserved -- always 1 */
  165. ((tc.tm_min / 10) << 4) | /* Tens of minutes */
  166. (tc.tm_min % 10); /* Units of minutes */
  167. buf[4] = (3 << 6) | /* reserved -- always 1 */
  168. ((tc.tm_hour / 10) << 4) | /* Tens of hours */
  169. (tc.tm_hour % 10); /* Units of hours */
  170. break;
  171. default:
  172. buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
  173. }
  174. return 5;
  175. }
  176. static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
  177. {
  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;
  187. if (of*2 >= size)
  188. continue;
  189. frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
  190. frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
  191. }
  192. frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
  193. }
  194. }
  195. }
  196. static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
  197. {
  198. int j, k;
  199. uint8_t* buf;
  200. for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
  201. /* DV subcode: 2nd and 3d DIFs */
  202. for (j = 80; j < 80 * 3; j += 80) {
  203. for (k = 6; k < 6 * 8; k += 8)
  204. dv_write_pack(dv_timecode, c, &buf[j+k]);
  205. 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 ? */
  206. dv_write_pack(dv_video_recdate, c, &buf[j+14]);
  207. dv_write_pack(dv_video_rectime, c, &buf[j+22]);
  208. dv_write_pack(dv_video_recdate, c, &buf[j+38]);
  209. dv_write_pack(dv_video_rectime, c, &buf[j+46]);
  210. }
  211. }
  212. /* DV VAUX: 4th, 5th and 6th 3DIFs */
  213. for (j = 80*3 + 3; j < 80*6; j += 80) {
  214. dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
  215. dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
  216. dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
  217. dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
  218. }
  219. }
  220. }
  221. /*
  222. * The following 3 functions constitute our interface to the world
  223. */
  224. static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
  225. uint8_t* data, int data_size, uint8_t** frame)
  226. {
  227. int i, reqasize;
  228. *frame = &c->frame_buf[0];
  229. reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
  230. switch (st->codec->codec_type) {
  231. case AVMEDIA_TYPE_VIDEO:
  232. /* FIXME: we have to have more sensible approach than this one */
  233. if (c->has_video)
  234. av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
  235. memcpy(*frame, data, c->sys->frame_size);
  236. c->has_video = 1;
  237. break;
  238. case AVMEDIA_TYPE_AUDIO:
  239. for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
  240. /* FIXME: we have to have more sensible approach than this one */
  241. if (av_fifo_size(c->audio_data[i]) + data_size >= 100*MAX_AUDIO_FRAME_SIZE)
  242. av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
  243. av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
  244. /* Let us see if we've got enough audio for one DV frame. */
  245. c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
  246. break;
  247. default:
  248. break;
  249. }
  250. /* Let us see if we have enough data to construct one DV frame. */
  251. if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
  252. dv_inject_metadata(c, *frame);
  253. c->has_audio = 0;
  254. for (i=0; i < c->n_ast; i++) {
  255. dv_inject_audio(c, i, *frame);
  256. av_fifo_drain(c->audio_data[i], reqasize);
  257. c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
  258. }
  259. c->has_video = 0;
  260. c->frames++;
  261. return c->sys->frame_size;
  262. }
  263. return 0;
  264. }
  265. static DVMuxContext* dv_init_mux(AVFormatContext* s)
  266. {
  267. DVMuxContext *c = s->priv_data;
  268. AVStream *vst = NULL;
  269. AVDictionaryEntry *t;
  270. int i;
  271. /* we support at most 1 video and 2 audio streams */
  272. if (s->nb_streams > 3)
  273. return NULL;
  274. c->n_ast = 0;
  275. c->ast[0] = c->ast[1] = NULL;
  276. /* We have to sort out where audio and where video stream is */
  277. for (i=0; i<s->nb_streams; i++) {
  278. switch (s->streams[i]->codec->codec_type) {
  279. case AVMEDIA_TYPE_VIDEO:
  280. if (vst) return NULL;
  281. vst = s->streams[i];
  282. break;
  283. case AVMEDIA_TYPE_AUDIO:
  284. if (c->n_ast > 1) return NULL;
  285. c->ast[c->n_ast++] = s->streams[i];
  286. break;
  287. default:
  288. goto bail_out;
  289. }
  290. }
  291. /* Some checks -- DV format is very picky about its incoming streams */
  292. if (!vst || vst->codec->codec_id != AV_CODEC_ID_DVVIDEO)
  293. goto bail_out;
  294. for (i=0; i<c->n_ast; i++) {
  295. if (c->ast[i] && (c->ast[i]->codec->codec_id != AV_CODEC_ID_PCM_S16LE ||
  296. c->ast[i]->codec->sample_rate != 48000 ||
  297. c->ast[i]->codec->channels != 2))
  298. goto bail_out;
  299. }
  300. c->sys = av_dv_codec_profile(vst->codec->width, vst->codec->height, vst->codec->pix_fmt);
  301. if (!c->sys)
  302. goto bail_out;
  303. if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
  304. /* only 1 stereo pair is allowed in 25Mbps mode */
  305. goto bail_out;
  306. }
  307. /* Ok, everything seems to be in working order */
  308. c->frames = 0;
  309. c->has_audio = 0;
  310. c->has_video = 0;
  311. if (t = av_dict_get(s->metadata, "creation_time", NULL, 0))
  312. c->start_time = ff_iso8601_to_unix_time(t->value);
  313. for (i=0; i < c->n_ast; i++) {
  314. if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*MAX_AUDIO_FRAME_SIZE))) {
  315. while (i > 0) {
  316. i--;
  317. av_fifo_free(c->audio_data[i]);
  318. }
  319. goto bail_out;
  320. }
  321. }
  322. return c;
  323. bail_out:
  324. return NULL;
  325. }
  326. static void dv_delete_mux(DVMuxContext *c)
  327. {
  328. int i;
  329. for (i=0; i < c->n_ast; i++)
  330. av_fifo_free(c->audio_data[i]);
  331. }
  332. static int dv_write_header(AVFormatContext *s)
  333. {
  334. if (!dv_init_mux(s)) {
  335. av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
  336. "Make sure that you supply exactly two streams:\n"
  337. " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
  338. " (50Mbps allows an optional second audio stream)\n");
  339. return -1;
  340. }
  341. return 0;
  342. }
  343. static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
  344. {
  345. uint8_t* frame;
  346. int fsize;
  347. fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
  348. pkt->data, pkt->size, &frame);
  349. if (fsize > 0) {
  350. avio_write(s->pb, frame, fsize);
  351. }
  352. return 0;
  353. }
  354. /*
  355. * We might end up with some extra A/V data without matching counterpart.
  356. * E.g. video data without enough audio to write the complete frame.
  357. * Currently we simply drop the last frame. I don't know whether this
  358. * is the best strategy of all
  359. */
  360. static int dv_write_trailer(struct AVFormatContext *s)
  361. {
  362. dv_delete_mux(s->priv_data);
  363. return 0;
  364. }
  365. AVOutputFormat ff_dv_muxer = {
  366. .name = "dv",
  367. .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
  368. .extensions = "dv",
  369. .priv_data_size = sizeof(DVMuxContext),
  370. .audio_codec = AV_CODEC_ID_PCM_S16LE,
  371. .video_codec = AV_CODEC_ID_DVVIDEO,
  372. .write_header = dv_write_header,
  373. .write_packet = dv_write_packet,
  374. .write_trailer = dv_write_trailer,
  375. };