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
  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/avassert.h"
  38 #include "libavutil/fifo.h"
  39 #include "libavutil/mathematics.h"
  40 #include "libavutil/intreadwrite.h"
  41 #include "libavutil/opt.h"
  42 #include "libavutil/timecode.h"
  43
  44 #define MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32-bit audio
  45
  46 struct DVMuxContext {
  47     AVClass          *av_class;
  48     const AVDVProfile*  sys;           /* current DV profile, e.g.: 525/60, 625/50 */
  49     int               n_ast;         /* number of stereo audio streams (up to 2) */
  50     AVStream         *ast[4];        /* stereo audio streams */
  51     AVFifoBuffer     *audio_data[4]; /* FIFO for storing excessive amounts of PCM */
  52     int               frames;        /* current frame number */
  53     int64_t           start_time;    /* recording start time */
  54     int               has_audio;     /* frame under construction has audio */
  55     int               has_video;     /* frame under construction has video */
  56     uint8_t           frame_buf[DV_MAX_FRAME_SIZE]; /* frame under construction */
  57     AVTimecode        tc;            /* timecode context */
  58 };
  59
  60 static const int dv_aaux_packs_dist[12][9] = {
  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 },
  65     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  66     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  67     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  68     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  69     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  70     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  71     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
  72     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
  73 };
  74
  75 static int dv_audio_frame_size(const AVDVProfile* sys, int frame, int sample_rate)
  76 {
  77     if ((sys->time_base.den == 25 || sys->time_base.den == 50) && sys->time_base.num == 1) {
  78         if      (sample_rate == 32000) return 1280;
  79         else if (sample_rate == 44100) return 1764;
  80         else                           return 1920;
  81     }
  82
  83     av_assert0(sample_rate == 48000);
  84
  85     return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
  86                                             sizeof(sys->audio_samples_dist[0]))];
  87 }
  88
  89 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, int channel, int seq)
  90 {
  91     struct tm tc;
  92     time_t ct;
  93     uint32_t timecode;
  94     int audio_type = 0;
  95
  96     buf[0] = (uint8_t)pack_id;
  97     switch (pack_id) {
  98     case dv_timecode:
  99         timecode  = av_timecode_get_smpte_from_framenum(&c->tc, c->frames);
 100         timecode |= 1<<23 | 1<<15 | 1<<7 | 1<<6; // biphase and binary group flags
 101         AV_WB32(buf + 1, timecode);
 102         break;
 103     case dv_audio_source:  /* AAUX source pack */
 104         if (c->ast[channel]->codecpar->sample_rate == 44100) {
 105             audio_type = 1;
 106         } else if (c->ast[channel]->codecpar->sample_rate == 32000)
 107             audio_type = 2;
 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, c->ast[channel]->codecpar->sample_rate) -
 111                   c->sys->audio_min_samples[audio_type]);
 112                             /* # of samples      */
 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                  (seq >= c->sys->difseg_size/2); /* audio mode (1st or 2nd channel) */
 117         buf[3] = (1 << 7) | /* res               */
 118                  (1 << 6) | /* multi-language flag */
 119                  (c->sys->dsf << 5) | /*  system: 60fields/50fields */
 120                  (DV_PROFILE_IS_HD(c->sys) ? 0x3 : c->sys->video_stype ? 2 : 0); /* stype */
 121         buf[4] = (1 << 7) | /* emphasis: 1 -- off */
 122                  (0 << 6) | /* emphasis time constant: 0 -- reserved */
 123                  (audio_type << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
 124                   0;        /* quantization: 0 -- 16-bit linear, 1 -- 12-bit nonlinear */
 125
 126         break;
 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 */
 135                   7;
 136         buf[3] = (1 << 7) | /* direction: 1 -- forward */
 137                  (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0x20 : /* speed */
 138                                                        c->sys->ltc_divisor * 4);
 139         buf[4] = (1 << 7) | /* reserved -- always 1 */
 140                   0x7f;     /* genre category */
 141         break;
 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 */
 157         break;
 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 */
 174         break;
 175     default:
 176         buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
 177     }
 178     return 5;
 179 }
 180
 181 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
 182 {
 183     int i, j, d, of, size;
 184     size = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[channel]->codecpar->sample_rate);
 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], channel, i);
 190             for (d = 8; d < 80; d+=2) {
 191                 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
 192                 if (of*2 >= size)
 193                     continue;
 194
 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
 197             }
 198             frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 199         }
 200     }
 201 }
 202
 203 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
 204 {
 205     int j, k;
 206     uint8_t* buf;
 207     int seq = 0;
 208
 209     for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80, seq++) {
 210         /* DV subcode: 2nd and 3d DIFs */
 211         for (j = 80; j < 80 * 3; j += 80) {
 212             for (k = 6; k < 6 * 8; k += 8)
 213                 dv_write_pack(dv_timecode, c, &buf[j+k], 0, seq);
 214
 215             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 ? */
 216                 dv_write_pack(dv_video_recdate, c, &buf[j+14], 0, seq);
 217                 dv_write_pack(dv_video_rectime, c, &buf[j+22], 0, seq);
 218                 dv_write_pack(dv_video_recdate, c, &buf[j+38], 0, seq);
 219                 dv_write_pack(dv_video_rectime, c, &buf[j+46], 0, seq);
 220             }
 221         }
 222
 223         /* DV VAUX: 4th, 5th and 6th 3DIFs */
 224         for (j = 80*3 + 3; j < 80*6; j += 80) {
 225             dv_write_pack(dv_video_recdate, c, &buf[j+5* 2], 0, seq);
 226             dv_write_pack(dv_video_rectime, c, &buf[j+5* 3], 0, seq);
 227             dv_write_pack(dv_video_recdate, c, &buf[j+5*11], 0, seq);
 228             dv_write_pack(dv_video_rectime, c, &buf[j+5*12], 0, seq);
 229         }
 230     }
 231 }
 232
 233 /*
 234  * The following 3 functions constitute our interface to the world
 235  */
 236
 237 static int dv_assemble_frame(AVFormatContext *s,
 238                              DVMuxContext *c, AVStream* st,
 239                              uint8_t* data, int data_size, uint8_t** frame)
 240 {
 241     int i, reqasize;
 242
 243     *frame = &c->frame_buf[0];
 244
 245     switch (st->codecpar->codec_type) {
 246     case AVMEDIA_TYPE_VIDEO:
 247         /* FIXME: we have to have more sensible approach than this one */
 248         if (c->has_video)
 249             av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
 250         if (data_size != c->sys->frame_size) {
 251             av_log(s, AV_LOG_ERROR, "Unexpected frame size, %d != %d\n",
 252                    data_size, c->sys->frame_size);
 253             return AVERROR(ENOSYS);
 254         }
 255
 256         memcpy(*frame, data, c->sys->frame_size);
 257         c->has_video = 1;
 258         break;
 259     case AVMEDIA_TYPE_AUDIO:
 260         for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
 261
 262           /* FIXME: we have to have more sensible approach than this one */
 263         if (av_fifo_size(c->audio_data[i]) + data_size >= 100*MAX_AUDIO_FRAME_SIZE)
 264             av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
 265         av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
 266
 267         reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, st->codecpar->sample_rate);
 268
 269         /* Let us see if we've got enough audio for one DV frame. */
 270         c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
 271
 272         break;
 273     default:
 274         break;
 275     }
 276
 277     /* Let us see if we have enough data to construct one DV frame. */
 278     if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
 279         dv_inject_metadata(c, *frame);
 280         c->has_audio = 0;
 281         for (i=0; i < c->n_ast; i++) {
 282             dv_inject_audio(c, i, *frame);
 283             reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[i]->codecpar->sample_rate);
 284             av_fifo_drain(c->audio_data[i], reqasize);
 285             c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
 286         }
 287
 288         c->has_video = 0;
 289
 290         c->frames++;
 291
 292         return c->sys->frame_size;
 293     }
 294
 295     return 0;
 296 }
 297
 298 static DVMuxContext* dv_init_mux(AVFormatContext* s)
 299 {
 300     DVMuxContext *c = s->priv_data;
 301     AVStream *vst = NULL;
 302     int i;
 303
 304     /* we support at most 1 video and 2 audio streams */
 305     if (s->nb_streams > 5)
 306         return NULL;
 307
 308     /* We have to sort out where audio and where video stream is */
 309     for (i=0; i<s->nb_streams; i++) {
 310         switch (s->streams[i]->codecpar->codec_type) {
 311         case AVMEDIA_TYPE_VIDEO:
 312             if (vst) return NULL;
 313             vst = s->streams[i];
 314             break;
 315         case AVMEDIA_TYPE_AUDIO:
 316             if (c->n_ast > 1) return NULL;
 317             c->ast[c->n_ast++] = s->streams[i];
 318             break;
 319         default:
 320             goto bail_out;
 321         }
 322     }
 323
 324     /* Some checks -- DV format is very picky about its incoming streams */
 325     if (!vst || vst->codecpar->codec_id != AV_CODEC_ID_DVVIDEO)
 326         goto bail_out;
 327     for (i=0; i<c->n_ast; i++) {
 328         if (c->ast[i]) {
 329             if(c->ast[i]->codecpar->codec_id    != AV_CODEC_ID_PCM_S16LE ||
 330                c->ast[i]->codecpar->channels    != 2)
 331                 goto bail_out;
 332             if (c->ast[i]->codecpar->sample_rate != 48000 &&
 333                 c->ast[i]->codecpar->sample_rate != 44100 &&
 334                 c->ast[i]->codecpar->sample_rate != 32000    )
 335                 goto bail_out;
 336         }
 337     }
 338     c->sys = av_dv_codec_profile2(vst->codecpar->width, vst->codecpar->height,
 339                                   vst->codecpar->format, vst->time_base);
 340     if (!c->sys)
 341         goto bail_out;
 342
 343     if ((c->sys->time_base.den != 25 && c->sys->time_base.den != 50) || c->sys->time_base.num != 1) {
 344         if (c->ast[0] && c->ast[0]->codecpar->sample_rate != 48000)
 345             goto bail_out;
 346         if (c->ast[1] && c->ast[1]->codecpar->sample_rate != 48000)
 347             goto bail_out;
 348     }
 349
 350     if (((c->n_ast > 1) && (c->sys->n_difchan < 2)) ||
 351         ((c->n_ast > 2) && (c->sys->n_difchan < 4))) {
 352         /* only 2 stereo pairs allowed in 50Mbps mode */
 353         goto bail_out;
 354     }
 355
 356     /* Ok, everything seems to be in working order */
 357     c->frames     = 0;
 358     c->has_audio  = 0;
 359     c->has_video  = 0;
 360     ff_parse_creation_time_metadata(s, &c->start_time, 1);
 361
 362     for (i=0; i < c->n_ast; i++) {
 363         if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc_array(100, MAX_AUDIO_FRAME_SIZE))) {
 364             goto bail_out;
 365         }
 366     }
 367
 368     return c;
 369
 370 bail_out:
 371     return NULL;
 372 }
 373
 374 static int dv_write_header(AVFormatContext *s)
 375 {
 376     AVRational rate;
 377     DVMuxContext *dvc = s->priv_data;
 378     AVDictionaryEntry *tcr = av_dict_get(s->metadata, "timecode", NULL, 0);
 379
 380     if (!dv_init_mux(s)) {
 381         av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
 382                     "Make sure that you supply exactly two streams:\n"
 383                     "     video: 25fps or 29.97fps, audio: 2ch/48|44|32kHz/PCM\n"
 384                     "     (50Mbps allows an optional second audio stream)\n");
 385         return -1;
 386     }
 387     rate.num = dvc->sys->ltc_divisor;
 388     rate.den = 1;
 389     if (!tcr) { // no global timecode, look into the streams
 390         int i;
 391         for (i = 0; i < s->nb_streams; i++) {
 392             tcr = av_dict_get(s->streams[i]->metadata, "timecode", NULL, 0);
 393             if (tcr)
 394                 break;
 395         }
 396     }
 397     if (tcr && av_timecode_init_from_string(&dvc->tc, rate, tcr->value, s) >= 0)
 398         return 0;
 399     return av_timecode_init(&dvc->tc, rate, 0, 0, s);
 400 }
 401
 402 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
 403 {
 404     uint8_t* frame;
 405     int fsize;
 406
 407     fsize = dv_assemble_frame(s, s->priv_data, s->streams[pkt->stream_index],
 408                               pkt->data, pkt->size, &frame);
 409     if (fsize > 0) {
 410         avio_write(s->pb, frame, fsize);
 411     }
 412     return 0;
 413 }
 414
 415 /*
 416  * We might end up with some extra A/V data without matching counterpart.
 417  * E.g. video data without enough audio to write the complete frame.
 418  * Currently we simply drop the last frame. I don't know whether this
 419  * is the best strategy of all
 420  */
 421 static void dv_deinit(AVFormatContext *s)
 422 {
 423     DVMuxContext *c = s->priv_data;
 424
 425     for (int i = 0; i < c->n_ast; i++)
 426         av_fifo_freep(&c->audio_data[i]);
 427 }
 428
 429 AVOutputFormat ff_dv_muxer = {
 430     .name              = "dv",
 431     .long_name         = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
 432     .extensions        = "dv",
 433     .priv_data_size    = sizeof(DVMuxContext),
 434     .audio_codec       = AV_CODEC_ID_PCM_S16LE,
 435     .video_codec       = AV_CODEC_ID_DVVIDEO,
 436     .write_header      = dv_write_header,
 437     .write_packet      = dv_write_packet,
 438     .deinit            = dv_deinit,
 439 };