FFmpeg: libavformat/dvenc.c Source File

00001 /*
00002  * General DV muxer/demuxer
00003  * Copyright (c) 2003 Roman Shaposhnik
00004  *
00005  * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
00006  * of DV technical info.
00007  *
00008  * Raw DV format
00009  * Copyright (c) 2002 Fabrice Bellard
00010  *
00011  * 50 Mbps (DVCPRO50) support
00012  * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
00013  *
00014  * This file is part of FFmpeg.
00015  *
00016  * FFmpeg is free software; you can redistribute it and/or
00017  * modify it under the terms of the GNU Lesser General Public
00018  * License as published by the Free Software Foundation; either
00019  * version 2.1 of the License, or (at your option) any later version.
00020  *
00021  * FFmpeg is distributed in the hope that it will be useful,
00022  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00023  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00024  * Lesser General Public License for more details.
00025  *
00026  * You should have received a copy of the GNU Lesser General Public
00027  * License along with FFmpeg; if not, write to the Free Software
00028  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00029  */
00030 #include <time.h>
00031 #include <stdarg.h>
00032 
00033 #include "avformat.h"
00034 #include "internal.h"
00035 #include "libavcodec/dvdata.h"
00036 #include "dv.h"
00037 #include "libavutil/fifo.h"
00038 
00039 struct DVMuxContext {
00040     const DVprofile*  sys;           /* current DV profile, e.g.: 525/60, 625/50 */
00041     int               n_ast;         /* number of stereo audio streams (up to 2) */
00042     AVStream         *ast[2];        /* stereo audio streams */
00043     AVFifoBuffer     *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
00044     int               frames;        /* current frame number */
00045     time_t            start_time;    /* recording start time */
00046     int               has_audio;     /* frame under contruction has audio */
00047     int               has_video;     /* frame under contruction has video */
00048     uint8_t           frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
00049 };
00050 
00051 static const int dv_aaux_packs_dist[12][9] = {
00052     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00053     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00054     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00055     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00056     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00057     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00058     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00059     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00060     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00061     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00062     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
00063     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
00064 };
00065 
00066 static int dv_audio_frame_size(const DVprofile* sys, int frame)
00067 {
00068     return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
00069                                             sizeof(sys->audio_samples_dist[0]))];
00070 }
00071 
00072 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
00073 {
00074     struct tm tc;
00075     time_t ct;
00076     int ltc_frame;
00077     va_list ap;
00078 
00079     buf[0] = (uint8_t)pack_id;
00080     switch (pack_id) {
00081     case dv_timecode:
00082         ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
00083                                     c->sys->time_base.den, AV_ROUND_DOWN);
00084         brktimegm(ct, &tc);
00085         /*
00086          * LTC drop-frame frame counter drops two frames (0 and 1) every
00087          * minute, unless it is exactly divisible by 10
00088          */
00089         ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
00090         buf[1] = (0                 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
00091                  (1                 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
00092                  ((ltc_frame / 10)  << 4) | /* tens of frames */
00093                  (ltc_frame % 10);          /* units of frames */
00094         buf[2] = (1                 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
00095                  ((tc.tm_sec / 10)  << 4) | /* tens of seconds */
00096                  (tc.tm_sec % 10);          /* units of seconds */
00097         buf[3] = (1                 << 7) | /* binary group flag BGF0 */
00098                  ((tc.tm_min / 10)  << 4) | /* tens of minutes */
00099                  (tc.tm_min % 10);          /* units of minutes */
00100         buf[4] = (1                 << 7) | /* binary group flag BGF2 */
00101                  (1                 << 6) | /* binary group flag BGF1 */
00102                  ((tc.tm_hour / 10) << 4) | /* tens of hours */
00103                  (tc.tm_hour % 10);         /* units of hours */
00104         break;
00105     case dv_audio_source:  /* AAUX source pack */
00106         va_start(ap, buf);
00107         buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
00108                  (1 << 6) | /* reserved -- always 1 */
00109                  (dv_audio_frame_size(c->sys, c->frames) -
00110                   c->sys->audio_min_samples[0]);
00111                             /* # of samples      */
00112         buf[2] = (0 << 7) | /* multi-stereo      */
00113                  (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
00114                  (0 << 4) | /* pair bit: 0 -- one pair of channels */
00115                  !!va_arg(ap, int); /* audio mode        */
00116         buf[3] = (1 << 7) | /* res               */
00117                  (1 << 6) | /* multi-language flag */
00118                  (c->sys->dsf << 5) | /*  system: 60fields/50fields */
00119                  (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
00120         buf[4] = (1 << 7) | /* emphasis: 1 -- off */
00121                  (0 << 6) | /* emphasis time constant: 0 -- reserved */
00122                  (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
00123                   0;        /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
00124         va_end(ap);
00125         break;
00126     case dv_audio_control:
00127         buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
00128                  (1 << 4) | /* input source: 1 -- digital input */
00129                  (3 << 2) | /* compression: 3 -- no information */
00130                   0;        /* misc. info/SMPTE emphasis off */
00131         buf[2] = (1 << 7) | /* recording start point: 1 -- no */
00132                  (1 << 6) | /* recording end point: 1 -- no */
00133                  (1 << 3) | /* recording mode: 1 -- original */
00134                   7;
00135         buf[3] = (1 << 7) | /* direction: 1 -- forward */
00136                  (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
00137                                                        c->sys->ltc_divisor * 4);
00138         buf[4] = (1 << 7) | /* reserved -- always 1 */
00139                   0x7f;     /* genre category */
00140         break;
00141     case dv_audio_recdate:
00142     case dv_video_recdate:  /* VAUX recording date */
00143         ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
00144                                             c->sys->time_base.den, AV_ROUND_DOWN);
00145         brktimegm(ct, &tc);
00146         buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
00147                        /* 0xff is very likely to be "unknown" */
00148         buf[2] = (3 << 6) | /* reserved -- always 1 */
00149                  ((tc.tm_mday / 10) << 4) | /* Tens of day */
00150                  (tc.tm_mday % 10);         /* Units of day */
00151         buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
00152                  ((tc.tm_mon / 10) << 4) |    /* Tens of month */
00153                  (tc.tm_mon  % 10);           /* Units of month */
00154         buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
00155                  (tc.tm_year % 10);                 /* Units of year */
00156         break;
00157     case dv_audio_rectime:  /* AAUX recording time */
00158     case dv_video_rectime:  /* VAUX recording time */
00159         ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
00160                                                        c->sys->time_base.den, AV_ROUND_DOWN);
00161         brktimegm(ct, &tc);
00162         buf[1] = (3 << 6) | /* reserved -- always 1 */
00163                  0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
00164         buf[2] = (1 << 7) | /* reserved -- always 1 */
00165                  ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
00166                  (tc.tm_sec % 10);         /* Units of seconds */
00167         buf[3] = (1 << 7) | /* reserved -- always 1 */
00168                  ((tc.tm_min / 10) << 4) | /* Tens of minutes */
00169                  (tc.tm_min % 10);         /* Units of minutes */
00170         buf[4] = (3 << 6) | /* reserved -- always 1 */
00171                  ((tc.tm_hour / 10) << 4) | /* Tens of hours */
00172                  (tc.tm_hour % 10);         /* Units of hours */
00173         break;
00174     default:
00175         buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
00176     }
00177     return 5;
00178 }
00179 
00180 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
00181 {
00182     int i, j, d, of, size;
00183     size = 4 * dv_audio_frame_size(c->sys, c->frames);
00184     frame_ptr += channel * c->sys->difseg_size * 150 * 80;
00185     for (i = 0; i < c->sys->difseg_size; i++) {
00186         frame_ptr += 6 * 80; /* skip DIF segment header */
00187         for (j = 0; j < 9; j++) {
00188             dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
00189             for (d = 8; d < 80; d+=2) {
00190                 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
00191                 if (of*2 >= size)
00192                     continue;
00193 
00194                 frame_ptr[d]   = av_fifo_peek(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
00195                 frame_ptr[d+1] = av_fifo_peek(c->audio_data[channel], of*2);   //        that DV is a big-endian PCM
00196             }
00197             frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
00198         }
00199     }
00200 }
00201 
00202 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
00203 {
00204     int j, k;
00205     uint8_t* buf;
00206 
00207     for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
00208         /* DV subcode: 2nd and 3d DIFs */
00209         for (j = 80; j < 80 * 3; j += 80) {
00210             for (k = 6; k < 6 * 8; k += 8)
00211                 dv_write_pack(dv_timecode, c, &buf[j+k]);
00212 
00213             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 ? */
00214                 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
00215                 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
00216                 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
00217                 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
00218             }
00219         }
00220 
00221         /* DV VAUX: 4th, 5th and 6th 3DIFs */
00222         for (j = 80*3 + 3; j < 80*6; j += 80) {
00223             dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
00224             dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
00225             dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
00226             dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
00227         }
00228     }
00229 }
00230 
00231 /*
00232  * The following 3 functions constitute our interface to the world
00233  */
00234 
00235 static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
00236                              uint8_t* data, int data_size, uint8_t** frame)
00237 {
00238     int i, reqasize;
00239 
00240     *frame = &c->frame_buf[0];
00241     reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
00242 
00243     switch (st->codec->codec_type) {
00244     case AVMEDIA_TYPE_VIDEO:
00245         /* FIXME: we have to have more sensible approach than this one */
00246         if (c->has_video)
00247             av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
00248 
00249         memcpy(*frame, data, c->sys->frame_size);
00250         c->has_video = 1;
00251         break;
00252     case AVMEDIA_TYPE_AUDIO:
00253         for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
00254 
00255           /* FIXME: we have to have more sensible approach than this one */
00256         if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
00257             av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
00258         av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
00259 
00260         /* Let us see if we've got enough audio for one DV frame. */
00261         c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
00262 
00263         break;
00264     default:
00265         break;
00266     }
00267 
00268     /* Let us see if we have enough data to construct one DV frame. */
00269     if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
00270         dv_inject_metadata(c, *frame);
00271         c->has_audio = 0;
00272         for (i=0; i < c->n_ast; i++) {
00273             dv_inject_audio(c, i, *frame);
00274             av_fifo_drain(c->audio_data[i], reqasize);
00275             c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
00276         }
00277 
00278         c->has_video = 0;
00279 
00280         c->frames++;
00281 
00282         return c->sys->frame_size;
00283     }
00284 
00285     return 0;
00286 }
00287 
00288 static DVMuxContext* dv_init_mux(AVFormatContext* s)
00289 {
00290     DVMuxContext *c = s->priv_data;
00291     AVStream *vst = NULL;
00292     int i;
00293 
00294     /* we support at most 1 video and 2 audio streams */
00295     if (s->nb_streams > 3)
00296         return NULL;
00297 
00298     c->n_ast  = 0;
00299     c->ast[0] = c->ast[1] = NULL;
00300 
00301     /* We have to sort out where audio and where video stream is */
00302     for (i=0; i<s->nb_streams; i++) {
00303         switch (s->streams[i]->codec->codec_type) {
00304         case AVMEDIA_TYPE_VIDEO:
00305             if (vst) return NULL;
00306             vst = s->streams[i];
00307             break;
00308         case AVMEDIA_TYPE_AUDIO:
00309             if (c->n_ast > 1) return NULL;
00310             c->ast[c->n_ast++] = s->streams[i];
00311             break;
00312         default:
00313             goto bail_out;
00314         }
00315     }
00316 
00317     /* Some checks -- DV format is very picky about its incoming streams */
00318     if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
00319         goto bail_out;
00320     for (i=0; i<c->n_ast; i++) {
00321         if (c->ast[i] && (c->ast[i]->codec->codec_id    != CODEC_ID_PCM_S16LE ||
00322                           c->ast[i]->codec->sample_rate != 48000 ||
00323                           c->ast[i]->codec->channels    != 2))
00324             goto bail_out;
00325     }
00326     c->sys = ff_dv_codec_profile(vst->codec);
00327     if (!c->sys)
00328         goto bail_out;
00329 
00330     if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
00331         /* only 1 stereo pair is allowed in 25Mbps mode */
00332         goto bail_out;
00333     }
00334 
00335     /* Ok, everything seems to be in working order */
00336     c->frames     = 0;
00337     c->has_audio  = 0;
00338     c->has_video  = 0;
00339     c->start_time = (time_t)s->timestamp;
00340 
00341     for (i=0; i < c->n_ast; i++) {
00342         if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*AVCODEC_MAX_AUDIO_FRAME_SIZE))) {
00343             while (i > 0) {
00344                 i--;
00345                 av_fifo_free(c->audio_data[i]);
00346             }
00347             goto bail_out;
00348         }
00349     }
00350 
00351     return c;
00352 
00353 bail_out:
00354     return NULL;
00355 }
00356 
00357 static void dv_delete_mux(DVMuxContext *c)
00358 {
00359     int i;
00360     for (i=0; i < c->n_ast; i++)
00361         av_fifo_free(c->audio_data[i]);
00362 }
00363 
00364 static int dv_write_header(AVFormatContext *s)
00365 {
00366     if (!dv_init_mux(s)) {
00367         av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
00368                     "Make sure that you supply exactly two streams:\n"
00369                     "     video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
00370                     "     (50Mbps allows an optional second audio stream)\n");
00371         return -1;
00372     }
00373     return 0;
00374 }
00375 
00376 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
00377 {
00378     uint8_t* frame;
00379     int fsize;
00380 
00381     fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
00382                               pkt->data, pkt->size, &frame);
00383     if (fsize > 0) {
00384         avio_write(s->pb, frame, fsize);
00385         avio_flush(s->pb);
00386     }
00387     return 0;
00388 }
00389 
00390 /*
00391  * We might end up with some extra A/V data without matching counterpart.
00392  * E.g. video data without enough audio to write the complete frame.
00393  * Currently we simply drop the last frame. I don't know whether this
00394  * is the best strategy of all
00395  */
00396 static int dv_write_trailer(struct AVFormatContext *s)
00397 {
00398     dv_delete_mux(s->priv_data);
00399     return 0;
00400 }
00401 
00402 AVOutputFormat ff_dv_muxer = {
00403     "dv",
00404     NULL_IF_CONFIG_SMALL("DV video format"),
00405     NULL,
00406     "dv",
00407     sizeof(DVMuxContext),
00408     CODEC_ID_PCM_S16LE,
00409     CODEC_ID_DVVIDEO,
00410     dv_write_header,
00411     dv_write_packet,
00412     dv_write_trailer,
00413 };