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 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 "libavcodec/dvdata.h"
  34 #include "dv.h"
  35 #include "libavutil/fifo.h"
  36
  37 struct DVMuxContext {
  38     const DVprofile*  sys;           /* current DV profile, e.g.: 525/60, 625/50 */
  39     int               n_ast;         /* number of stereo audio streams (up to 2) */
  40     AVStream         *ast[2];        /* stereo audio streams */
  41     AVFifoBuffer     *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
  42     int               frames;        /* current frame number */
  43     time_t            start_time;    /* recording start time */
  44     int               has_audio;     /* frame under contruction has audio */
  45     int               has_video;     /* frame under contruction has video */
  46     uint8_t           frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
  47 };
  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
  64 static int dv_audio_frame_size(const DVprofile* sys, int frame)
  65 {
  66     return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
  67                                             sizeof(sys->audio_samples_dist[0]))];
  68 }
  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
  77     buf[0] = (uint8_t)pack_id;
  78     switch (pack_id) {
  79     case dv_timecode:
  80         ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
  81                                     c->sys->time_base.den, AV_ROUND_DOWN);
  82         brktimegm(ct, &tc);
  83         /*
  84          * LTC drop-frame frame counter drops two frames (0 and 1) every
  85          * minute, unless it is exactly divisible by 10
  86          */
  87         ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
  88         buf[1] = (0                 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
  89                  (1                 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
  90                  ((ltc_frame / 10)  << 4) | /* tens of frames */
  91                  (ltc_frame % 10);          /* units of frames */
  92         buf[2] = (1                 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
  93                  ((tc.tm_sec / 10)  << 4) | /* tens of seconds */
  94                  (tc.tm_sec % 10);          /* units of seconds */
  95         buf[3] = (1                 << 7) | /* binary group flag BGF0 */
  96                  ((tc.tm_min / 10)  << 4) | /* tens of minutes */
  97                  (tc.tm_min % 10);          /* units of minutes */
  98         buf[4] = (1                 << 7) | /* binary group flag BGF2 */
  99                  (1                 << 6) | /* binary group flag BGF1 */
 100                  ((tc.tm_hour / 10) << 4) | /* tens of hours */
 101                  (tc.tm_hour % 10);         /* units of hours */
 102         break;
 103     case dv_audio_source:  /* AAUX source pack */
 104         va_start(ap, buf);
 105         buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
 106                  (1 << 6) | /* reserved -- always 1 */
 107                  (dv_audio_frame_size(c->sys, c->frames) -
 108                   c->sys->audio_min_samples[0]);
 109                             /* # of samples      */
 110         buf[2] = (0 << 7) | /* multi-stereo      */
 111                  (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
 112                  (0 << 4) | /* pair bit: 0 -- one pair of channels */
 113                  !!va_arg(ap, int); /* audio mode        */
 114         buf[3] = (1 << 7) | /* res               */
 115                  (1 << 6) | /* multi-language flag */
 116                  (c->sys->dsf << 5) | /*  system: 60fields/50fields */
 117                  (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
 118         buf[4] = (1 << 7) | /* emphasis: 1 -- off */
 119                  (0 << 6) | /* emphasis time constant: 0 -- reserved */
 120                  (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
 121                   0;        /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
 122         va_end(ap);
 123         break;
 124     case dv_audio_control:
 125         buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
 126                  (1 << 4) | /* input source: 1 -- digital input */
 127                  (3 << 2) | /* compression: 3 -- no information */
 128                   0;        /* misc. info/SMPTE emphasis off */
 129         buf[2] = (1 << 7) | /* recording start point: 1 -- no */
 130                  (1 << 6) | /* recording end point: 1 -- no */
 131                  (1 << 3) | /* recording mode: 1 -- original */
 132                   7;
 133         buf[3] = (1 << 7) | /* direction: 1 -- forward */
 134                  (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
 135                                                        c->sys->ltc_divisor * 4);
 136         buf[4] = (1 << 7) | /* reserved -- always 1 */
 137                   0x7f;     /* genre category */
 138         break;
 139     case dv_audio_recdate:
 140     case dv_video_recdate:  /* VAUX recording date */
 141         ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
 142                                             c->sys->time_base.den, AV_ROUND_DOWN);
 143         brktimegm(ct, &tc);
 144         buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
 145                        /* 0xff is very likely to be "unknown" */
 146         buf[2] = (3 << 6) | /* reserved -- always 1 */
 147                  ((tc.tm_mday / 10) << 4) | /* Tens of day */
 148                  (tc.tm_mday % 10);         /* Units of day */
 149         buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
 150                  ((tc.tm_mon / 10) << 4) |    /* Tens of month */
 151                  (tc.tm_mon  % 10);           /* Units of month */
 152         buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
 153                  (tc.tm_year % 10);                 /* Units of year */
 154         break;
 155     case dv_audio_rectime:  /* AAUX recording time */
 156     case dv_video_rectime:  /* VAUX recording time */
 157         ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
 158                                                        c->sys->time_base.den, AV_ROUND_DOWN);
 159         brktimegm(ct, &tc);
 160         buf[1] = (3 << 6) | /* reserved -- always 1 */
 161                  0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
 162         buf[2] = (1 << 7) | /* reserved -- always 1 */
 163                  ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
 164                  (tc.tm_sec % 10);         /* Units of seconds */
 165         buf[3] = (1 << 7) | /* reserved -- always 1 */
 166                  ((tc.tm_min / 10) << 4) | /* Tens of minutes */
 167                  (tc.tm_min % 10);         /* Units of minutes */
 168         buf[4] = (3 << 6) | /* reserved -- always 1 */
 169                  ((tc.tm_hour / 10) << 4) | /* Tens of hours */
 170                  (tc.tm_hour % 10);         /* Units of hours */
 171         break;
 172     default:
 173         buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
 174     }
 175     return 5;
 176 }
 177
 178 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
 179 {
 180     int i, j, d, of, size;
 181     size = 4 * dv_audio_frame_size(c->sys, c->frames);
 182     frame_ptr += channel * c->sys->difseg_size * 150 * 80;
 183     for (i = 0; i < c->sys->difseg_size; i++) {
 184         frame_ptr += 6 * 80; /* skip DIF segment header */
 185         for (j = 0; j < 9; j++) {
 186             dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
 187             for (d = 8; d < 80; d+=2) {
 188                 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
 189                 if (of*2 >= size)
 190                     continue;
 191
 192                 frame_ptr[d]   = av_fifo_peek(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
 193                 frame_ptr[d+1] = av_fifo_peek(c->audio_data[channel], of*2);   //        that DV is a big-endian PCM
 194             }
 195             frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 196         }
 197     }
 198 }
 199
 200 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
 201 {
 202     int j, k;
 203     uint8_t* buf;
 204
 205     for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
 206         /* DV subcode: 2nd and 3d DIFs */
 207         for (j = 80; j < 80 * 3; j += 80) {
 208             for (k = 6; k < 6 * 8; k += 8)
 209                 dv_write_pack(dv_timecode, c, &buf[j+k]);
 210
 211             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 ? */
 212                 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
 213                 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
 214                 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
 215                 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
 216             }
 217         }
 218
 219         /* DV VAUX: 4th, 5th and 6th 3DIFs */
 220         for (j = 80*3 + 3; j < 80*6; j += 80) {
 221             dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
 222             dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
 223             dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
 224             dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
 225         }
 226     }
 227 }
 228
 229 /*
 230  * The following 3 functions constitute our interface to the world
 231  */
 232
 233 int dv_assemble_frame(DVMuxContext *c, AVStream* st,
 234                       uint8_t* data, int data_size, uint8_t** frame)
 235 {
 236     int i, reqasize;
 237
 238     *frame = &c->frame_buf[0];
 239     reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
 240
 241     switch (st->codec->codec_type) {
 242     case CODEC_TYPE_VIDEO:
 243         /* FIXME: we have to have more sensible approach than this one */
 244         if (c->has_video)
 245             av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
 246
 247         memcpy(*frame, data, c->sys->frame_size);
 248         c->has_video = 1;
 249         break;
 250     case CODEC_TYPE_AUDIO:
 251         for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
 252
 253           /* FIXME: we have to have more sensible approach than this one */
 254         if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
 255             av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
 256         av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
 257
 258         /* Let us see if we've got enough audio for one DV frame. */
 259         c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
 260
 261         break;
 262     default:
 263         break;
 264     }
 265
 266     /* Let us see if we have enough data to construct one DV frame. */
 267     if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
 268         dv_inject_metadata(c, *frame);
 269         c->has_audio = 0;
 270         for (i=0; i < c->n_ast; i++) {
 271             dv_inject_audio(c, i, *frame);
 272             av_fifo_drain(c->audio_data[i], reqasize);
 273             c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
 274         }
 275
 276         c->has_video = 0;
 277
 278         c->frames++;
 279
 280         return c->sys->frame_size;
 281     }
 282
 283     return 0;
 284 }
 285
 286 DVMuxContext* dv_init_mux(AVFormatContext* s)
 287 {
 288     DVMuxContext *c = s->priv_data;
 289     AVStream *vst = NULL;
 290     int i;
 291
 292     /* we support at most 1 video and 2 audio streams */
 293     if (s->nb_streams > 3)
 294         return NULL;
 295
 296     c->n_ast  = 0;
 297     c->ast[0] = c->ast[1] = NULL;
 298
 299     /* We have to sort out where audio and where video stream is */
 300     for (i=0; i<s->nb_streams; i++) {
 301         switch (s->streams[i]->codec->codec_type) {
 302         case CODEC_TYPE_VIDEO:
 303             if (vst) return NULL;
 304             vst = s->streams[i];
 305             break;
 306         case CODEC_TYPE_AUDIO:
 307             if (c->n_ast > 1) return NULL;
 308             c->ast[c->n_ast++] = s->streams[i];
 309             break;
 310         default:
 311             goto bail_out;
 312         }
 313     }
 314
 315     /* Some checks -- DV format is very picky about its incoming streams */
 316     if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
 317         goto bail_out;
 318     for (i=0; i<c->n_ast; i++) {
 319         if (c->ast[i] && (c->ast[i]->codec->codec_id    != CODEC_ID_PCM_S16LE ||
 320                           c->ast[i]->codec->sample_rate != 48000 ||
 321                           c->ast[i]->codec->channels    != 2))
 322             goto bail_out;
 323     }
 324     c->sys = dv_codec_profile(vst->codec);
 325     if (!c->sys)
 326         goto bail_out;
 327
 328     if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
 329         /* only 1 stereo pair is allowed in 25Mbps mode */
 330         goto bail_out;
 331     }
 332
 333     /* Ok, everything seems to be in working order */
 334     c->frames     = 0;
 335     c->has_audio  = 0;
 336     c->has_video  = 0;
 337     c->start_time = (time_t)s->timestamp;
 338
 339     for (i=0; i < c->n_ast; i++) {
 340         if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*AVCODEC_MAX_AUDIO_FRAME_SIZE))) {
 341             while (i > 0) {
 342                 i--;
 343                 av_fifo_free(c->audio_data[i]);
 344             }
 345             goto bail_out;
 346         }
 347     }
 348
 349     return c;
 350
 351 bail_out:
 352     return NULL;
 353 }
 354
 355 void dv_delete_mux(DVMuxContext *c)
 356 {
 357     int i;
 358     for (i=0; i < c->n_ast; i++)
 359         av_fifo_free(c->audio_data[i]);
 360 }
 361
 362 #if CONFIG_DV_MUXER
 363 static int dv_write_header(AVFormatContext *s)
 364 {
 365     if (!dv_init_mux(s)) {
 366         av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
 367                     "Make sure that you supply exactly two streams:\n"
 368                     "     video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
 369                     "     (50Mbps allows an optional second audio stream)\n");
 370         return -1;
 371     }
 372     return 0;
 373 }
 374
 375 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
 376 {
 377     uint8_t* frame;
 378     int fsize;
 379
 380     fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
 381                               pkt->data, pkt->size, &frame);
 382     if (fsize > 0) {
 383         put_buffer(s->pb, frame, fsize);
 384         put_flush_packet(s->pb);
 385     }
 386     return 0;
 387 }
 388
 389 /*
 390  * We might end up with some extra A/V data without matching counterpart.
 391  * E.g. video data without enough audio to write the complete frame.
 392  * Currently we simply drop the last frame. I don't know whether this
 393  * is the best strategy of all
 394  */
 395 static int dv_write_trailer(struct AVFormatContext *s)
 396 {
 397     dv_delete_mux(s->priv_data);
 398     return 0;
 399 }
 400
 401 AVOutputFormat dv_muxer = {
 402     "dv",
 403     NULL_IF_CONFIG_SMALL("DV video format"),
 404     NULL,
 405     "dv",
 406     sizeof(DVMuxContext),
 407     CODEC_ID_PCM_S16LE,
 408     CODEC_ID_DVVIDEO,
 409     dv_write_header,
 410     dv_write_packet,
 411     dv_write_trailer,
 412 };
 413 #endif /* CONFIG_DV_MUXER */