libavformat/dv.c

   1 /*
   2  * General DV muxer/demuxer
   3  * Copyright (c) 2003 Roman Shaposhnick
   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  * This library is free software; you can redistribute it and/or
  12  * modify it under the terms of the GNU Lesser General Public
  13  * License as published by the Free Software Foundation; either
  14  * version 2 of the License, or (at your option) any later version.
  15  *
  16  * This library is distributed in the hope that it will be useful,
  17  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  18  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  19  * Lesser General Public License for more details.
  20  *
  21  * You should have received a copy of the GNU Lesser General Public
  22  * License along with this library; if not, write to the Free Software
  23  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  24  */
  25 #include <time.h>
  26 #include "avformat.h"
  27 #include "dvdata.h"
  28 #include "dv.h"
  29
  30 struct DVDemuxContext {
  31     AVFormatContext* fctx;
  32     AVStream*        vst;
  33     AVStream*        ast[2];
  34     AVPacket         audio_pkt[2];
  35     int              ach;
  36 };
  37
  38 struct DVMuxContext {
  39     const DVprofile*  sys;    /* Current DV profile. E.g.: 525/60, 625/50 */
  40     uint8_t     frame_buf[144000]; /* frame under contruction */
  41     FifoBuffer  audio_data;   /* Fifo for storing excessive amounts of PCM */
  42     int         frames;       /* Number of a current frame */
  43     time_t      start_time;   /* Start time of recording */
  44     uint8_t     aspect;       /* Aspect ID 0 - 4:3, 7 - 16:9 */
  45     int         has_audio;    /* frame under contruction has audio */
  46     int         has_video;    /* frame under contruction has video */
  47 };
  48
  49 enum dv_section_type {
  50      dv_sect_header  = 0x1f,
  51      dv_sect_subcode = 0x3f,
  52      dv_sect_vaux    = 0x56,
  53      dv_sect_audio   = 0x76,
  54      dv_sect_video   = 0x96,
  55 };
  56
  57 enum dv_pack_type {
  58      dv_header525     = 0x3f, /* see dv_write_pack for important details on */
  59      dv_header625     = 0xbf, /* these two packs */
  60      dv_timecode      = 0x13,
  61      dv_audio_source  = 0x50,
  62      dv_audio_control = 0x51,
  63      dv_audio_recdate = 0x52,
  64      dv_audio_rectime = 0x53,
  65      dv_video_source  = 0x60,
  66      dv_video_control = 0x61,
  67      dv_viedo_recdate = 0x62,
  68      dv_video_rectime = 0x63,
  69      dv_unknown_pack  = 0xff,
  70 };
  71
  72
  73
  74 /*
  75  * The reason why the following three big ugly looking tables are
  76  * here is my lack of DV spec IEC 61834. The tables were basically
  77  * constructed to make code that places packs in SSYB, VAUX and
  78  * AAUX blocks very simple and table-driven. They conform to the
  79  * SMPTE 314M and the output of my personal DV camcorder, neither
  80  * of which is sufficient for a reliable DV stream producing. Thus
  81  * while code is still in development I'll be gathering input from
  82  * people with different DV equipment and modifying the tables to
  83  * accommodate all the quirks. Later on, if possible, some of them
  84  * will be folded into smaller tables and/or switch-if logic. For
  85  * now, my only excuse is -- they don't eat up that much of a space.
  86  */
  87
  88 static const int dv_ssyb_packs_dist[12][6] = {
  89     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  90     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  91     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  92     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  93     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  94     { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
  95     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
  96     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
  97     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
  98     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
  99     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
 100     { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
 101 };
 102
 103 static const int dv_vaux_packs_dist[12][15] = {
 104     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 105       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 106     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 107       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 108     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 109       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 110     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 111       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 112     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 113       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 114     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 115       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 116     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 117       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 118     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 119       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 120     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 121       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 122     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 123       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 124     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 125       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 126     { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
 127       0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
 128 };
 129
 130 static const int dv_aaux_packs_dist[12][9] = {
 131     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 132     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 133     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 134     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 135     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 136     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 137     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 138     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 139     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 140     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 141     { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
 142     { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
 143 };
 144
 145 static inline uint16_t dv_audio_12to16(uint16_t sample)
 146 {
 147     uint16_t shift, result;
 148
 149     sample = (sample < 0x800) ? sample : sample | 0xf000;
 150     shift = (sample & 0xf00) >> 8;
 151
 152     if (shift < 0x2 || shift > 0xd) {
 153         result = sample;
 154     } else if (shift < 0x8) {
 155         shift--;
 156         result = (sample - (256 * shift)) << shift;
 157     } else {
 158         shift = 0xe - shift;
 159         result = ((sample + ((256 * shift) + 1)) << shift) - 1;
 160     }
 161
 162     return result;
 163 }
 164
 165 static int dv_audio_frame_size(const DVprofile* sys, int frame)
 166 {
 167     return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
 168                                             sizeof(sys->audio_samples_dist[0]))];
 169 }
 170
 171 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
 172 {
 173     struct tm tc;
 174     time_t ct;
 175     int ltc_frame;
 176
 177     buf[0] = (uint8_t)pack_id;
 178     switch (pack_id) {
 179     case dv_header525: /* I can't imagine why these two weren't defined as real */
 180     case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
 181           buf[1] = 0xf8 |               /* reserved -- always 1 */
 182                    (0 & 0x07);          /* APT: Track application ID */
 183           buf[2] = (0 << 7)    | /* TF1: audio data is 0 - valid; 1 - invalid */
 184                    (0x0f << 3) | /* reserved -- always 1 */
 185                    (0 & 0x07);   /* AP1: Audio application ID */
 186           buf[3] = (0 << 7)    | /* TF2: video data is 0 - valid; 1 - invalid */
 187                    (0x0f << 3) | /* reserved -- always 1 */
 188                    (0 & 0x07);   /* AP2: Video application ID */
 189           buf[4] = (0 << 7)    | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */
 190                    (0x0f << 3) | /* reserved -- always 1 */
 191                    (0 & 0x07);   /* AP3: Subcode application ID */
 192           break;
 193     case dv_timecode:
 194           ct = (time_t)(c->frames / ((float)c->sys->frame_rate /
 195                                      (float)c->sys->frame_rate_base));
 196           localtime_r(&ct, &tc);
 197           /*
 198            * LTC drop-frame frame counter drops two frames (0 and 1) every
 199            * minute, unless it is exactly divisible by 10
 200            */
 201           ltc_frame = (c->frames + 2*ct/60 - 2*ct/600) % c->sys->ltc_divisor;
 202           buf[1] = (0 << 7) | /* Color fame: 0 - unsync; 1 - sync mode */
 203                    (1 << 6) | /* Drop frame timecode: 0 - nondrop; 1 - drop */
 204                    ((ltc_frame / 10) << 4) | /* Tens of frames */
 205                    (ltc_frame % 10);         /* Units of frames */
 206           buf[2] = (1 << 7) | /* Biphase mark polarity correction: 0 - even; 1 - odd */
 207                    ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
 208                    (tc.tm_sec % 10);         /* Units of seconds */
 209           buf[3] = (1 << 7) | /* Binary group flag BGF0 */
 210                    ((tc.tm_min / 10) << 4) | /* Tens of minutes */
 211                    (tc.tm_min % 10);         /* Units of minutes */
 212           buf[4] = (1 << 7) | /* Binary group flag BGF2 */
 213                    (1 << 6) | /* Binary group flag BGF1 */
 214                    ((tc.tm_hour / 10) << 4) | /* Tens of hours */
 215                    (tc.tm_hour % 10);         /* Units of hours */
 216           break;
 217     case dv_audio_source:  /* AAUX source pack */
 218           buf[1] = (0 << 7) | /* locked mode       */
 219                    (1 << 6) | /* reserved -- always 1 */
 220                    (dv_audio_frame_size(c->sys, c->frames) -
 221                     c->sys->audio_min_samples[0]);
 222                               /* # of samples      */
 223           buf[2] = (0 << 7) | /* multi-stereo      */
 224                    (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
 225                    (0 << 4) | /* pair bit: 0 -- one pair of channels */
 226                     0;        /* audio mode        */
 227           buf[3] = (1 << 7) | /* res               */
 228                    (1 << 6) | /* multi-language flag */
 229                    (c->sys->dsf << 5) | /*  system: 60fields/50fields */
 230                     0;        /* definition: 0 -- SD (525/625) */
 231           buf[4] = (1 << 7) | /* emphasis: 1 -- off */
 232                    (0 << 6) | /* emphasis time constant: 0 -- reserved */
 233                    (0 << 3) | /* frequency: 0 -- 48Khz, 1 -- 44,1Khz, 2 -- 32Khz */
 234                     0;        /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
 235           break;
 236     case dv_audio_control:
 237           buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
 238                    (1 << 4) | /* input source: 1 -- digital input */
 239                    (3 << 2) | /* compression: 3 -- no information */
 240                     0;        /* misc. info/SMPTE emphasis off */
 241           buf[2] = (1 << 7) | /* recording start point: 1 -- no */
 242                    (1 << 6) | /* recording end point: 1 -- no */
 243                    (1 << 3) | /* recording mode: 1 -- original */
 244                     7;
 245           buf[3] = (1 << 7) | /* direction: 1 -- forward */
 246                     0x20;     /* speed */
 247           buf[4] = (1 << 7) | /* reserved -- always 1 */
 248                     0x7f;     /* genre category */
 249           break;
 250     case dv_audio_recdate:
 251     case dv_viedo_recdate:  /* VAUX recording date */
 252           ct = c->start_time + (time_t)(c->frames /
 253                ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
 254           localtime_r(&ct, &tc);
 255           buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
 256                          /* 0xff is very likely to be "unknown" */
 257           buf[2] = (3 << 6) | /* reserved -- always 1 */
 258                    ((tc.tm_mday / 10) << 4) | /* Tens of day */
 259                    (tc.tm_mday % 10);         /* Units of day */
 260           buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
 261                    (tc.tm_mon % 10);         /* Units of month */
 262           buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
 263                    (tc.tm_year % 10);                 /* Units of year */
 264           break;
 265     case dv_audio_rectime:  /* AAUX recording time */
 266     case dv_video_rectime:  /* VAUX recording time */
 267           ct = c->start_time + (time_t)(c->frames /
 268                ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
 269           localtime_r(&ct, &tc);
 270           buf[1] = (3 << 6) | /* reserved -- always 1 */
 271                    0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
 272           buf[2] = (1 << 7) | /* reserved -- always 1 */
 273                    ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
 274                    (tc.tm_sec % 10);         /* Units of seconds */
 275           buf[3] = (1 << 7) | /* reserved -- always 1 */
 276                    ((tc.tm_min / 10) << 4) | /* Tens of minutes */
 277                    (tc.tm_min % 10);         /* Units of minutes */
 278           buf[4] = (3 << 6) | /* reserved -- always 1 */
 279                    ((tc.tm_hour / 10) << 4) | /* Tens of hours */
 280                    (tc.tm_hour % 10);         /* Units of hours */
 281           break;
 282     case dv_video_source:
 283           buf[1] = 0xff; /* reserved -- always 1 */
 284           buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
 285                    (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
 286                    (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
 287                    0xf; /* reserved -- always 1 */
 288           buf[3] = (3 << 6) | /* reserved -- always 1 */
 289                    (c->sys->dsf << 5) | /*  system: 60fields/50fields */
 290                    0; /* signal type video compression */
 291           buf[4] = 0xff; /* VISC: 0xff -- no information */
 292           break;
 293     case dv_video_control:
 294           buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
 295                    0x3f; /* reserved -- always 1 */
 296           buf[2] = 0xc8 | /* reserved -- always b11001xxx */
 297                    c->aspect;
 298           buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
 299                    (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
 300                    (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
 301                    (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
 302                    0xc; /* reserved -- always b1100 */
 303           buf[4] = 0xff; /* reserved -- always 1 */
 304           break;
 305     default:
 306           buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
 307     }
 308     return 5;
 309 }
 310
 311 static inline int dv_write_dif_id(enum dv_section_type t, uint8_t seq_num,
 312                                   uint8_t dif_num, uint8_t* buf)
 313 {
 314     buf[0] = (uint8_t)t;    /* Section type */
 315     buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
 316              (0 << 3) |     /* FSC: for 50Mb/s 0 - first channel; 1 - second */
 317              7;             /* reserved -- always 1 */
 318     buf[2] = dif_num;       /* DIF block number Video: 0-134, Audio: 0-8 */
 319     return 3;
 320 }
 321
 322 static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
 323 {
 324     if (syb_num == 0 || syb_num == 6) {
 325         buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
 326                  (0<<4)  | /* AP3 (Subcode application ID) */
 327                  0x0f;     /* reserved -- always 1 */
 328     }
 329     else if (syb_num == 11) {
 330         buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
 331                  0x7f;     /* reserved -- always 1 */
 332     }
 333     else {
 334         buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
 335                  (0<<4)  | /* APT (Track application ID) */
 336                  0x0f;     /* reserved -- always 1 */
 337     }
 338     buf[1] = 0xf0 |            /* reserved -- always 1 */
 339              (syb_num & 0x0f); /* SSYB number 0 - 11   */
 340     buf[2] = 0xff;             /* reserved -- always 1 */
 341     return 3;
 342 }
 343
 344 static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
 345 {
 346     int i, j, k;
 347
 348     for (i = 0; i < c->sys->difseg_size; i++) {
 349        memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
 350
 351        /* DV header: 1DIF */
 352        buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
 353        buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
 354        buf += 72; /* unused bytes */
 355
 356        /* DV subcode: 2DIFs */
 357        for (j = 0; j < 2; j++) {
 358           buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
 359           for (k = 0; k < 6; k++) {
 360              buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
 361              buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
 362           }
 363           buf += 29; /* unused bytes */
 364        }
 365
 366        /* DV VAUX: 3DIFs */
 367        for (j = 0; j < 3; j++) {
 368           buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
 369           for (k = 0; k < 15 ; k++)
 370              buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
 371           buf += 2; /* unused bytes */
 372        }
 373
 374        /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
 375        for (j = 0; j < 135; j++) {
 376             if (j%15 == 0) {
 377                 buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
 378                 buf += dv_write_pack(dv_aaux_packs_dist[i][j/15], c, buf);
 379                 buf += 72; /* shuffled PCM audio */
 380             }
 381             buf += dv_write_dif_id(dv_sect_video, i, j, buf);
 382             buf += 77; /* 1 video macro block: 1 bytes control
 383                                                4 * 14 bytes Y 8x8 data
 384                                                    10 bytes Cr 8x8 data
 385                                                    10 bytes Cb 8x8 data */
 386        }
 387     }
 388 }
 389
 390 static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, uint8_t* frame_ptr)
 391 {
 392     int i, j, d, of;
 393     for (i = 0; i < c->sys->difseg_size; i++) {
 394        frame_ptr += 6 * 80; /* skip DIF segment header */
 395        for (j = 0; j < 9; j++) {
 396           for (d = 8; d < 80; d+=2) {
 397              of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
 398              frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
 399              frame_ptr[d+1] = pcm[of*2]; //        that DV is a big endian PCM
 400           }
 401           frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 402        }
 403     }
 404 }
 405
 406 static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
 407 {
 408     int i, j;
 409     int ptr = 0;
 410
 411     for (i = 0; i < c->sys->difseg_size; i++) {
 412        ptr += 6 * 80; /* skip DIF segment header */
 413        for (j = 0; j < 135; j++) {
 414             if (j%15 == 0)
 415                 ptr += 80; /* skip Audio DIF */
 416             ptr += 3;
 417             memcpy(frame_ptr + ptr, video_data + ptr, 77);
 418             ptr += 77;
 419        }
 420     }
 421 }
 422
 423 /*
 424  * This is the dumbest implementation of all -- it simply looks at
 425  * a fixed offset and if pack isn't there -- fails. We might want
 426  * to have a fallback mechanism for complete search of missing packs.
 427  */
 428 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
 429 {
 430     int offs;
 431
 432     switch (t) {
 433     case dv_audio_source:
 434           offs = (80*6 + 80*16*3 + 3);
 435           break;
 436     case dv_audio_control:
 437           offs = (80*6 + 80*16*4 + 3);
 438           break;
 439     case dv_video_control:
 440           offs = (80*5 + 48 + 5);
 441           break;
 442     default:
 443           return NULL;
 444     }
 445
 446     return (frame[offs] == t ? &frame[offs] : NULL);
 447 }
 448
 449 /*
 450  * There's a couple of assumptions being made here:
 451  * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
 452  *    We can pass them upwards when ffmpeg will be ready to deal with them.
 453  * 2. We don't do software emphasis.
 454  * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
 455  *    are converted into 16bit linear ones.
 456  */
 457 static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2)
 458 {
 459     int size, i, j, d, of, smpls, freq, quant, half_ch;
 460     uint16_t lc, rc;
 461     const DVprofile* sys;
 462     const uint8_t* as_pack;
 463
 464     as_pack = dv_extract_pack(frame, dv_audio_source);
 465     if (!as_pack)    /* No audio ? */
 466         return 0;
 467
 468     sys = dv_frame_profile(frame);
 469     smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
 470     freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
 471     quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
 472
 473     if (quant > 1)
 474         return -1; /* Unsupported quantization */
 475
 476     size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
 477     half_ch = sys->difseg_size/2;
 478
 479     /* for each DIF segment */
 480     for (i = 0; i < sys->difseg_size; i++) {
 481        frame += 6 * 80; /* skip DIF segment header */
 482        if (quant == 1 && i == half_ch) {
 483            if (!pcm2)
 484                break;
 485            else
 486                pcm = pcm2;
 487        }
 488
 489        for (j = 0; j < 9; j++) {
 490           for (d = 8; d < 80; d += 2) {
 491              if (quant == 0) {  /* 16bit quantization */
 492                  of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
 493                  if (of*2 >= size)
 494                      continue;
 495
 496                  pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
 497                  pcm[of*2+1] = frame[d]; //        that DV is a big endian PCM
 498                  if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
 499                      pcm[of*2+1] = 0;
 500               } else {           /* 12bit quantization */
 501                  lc = ((uint16_t)frame[d] << 4) |
 502                       ((uint16_t)frame[d+2] >> 4);
 503                  rc = ((uint16_t)frame[d+1] << 4) |
 504                       ((uint16_t)frame[d+2] & 0x0f);
 505                  lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
 506                  rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
 507
 508                  of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
 509                  if (of*2 >= size)
 510                      continue;
 511
 512                  pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
 513                  pcm[of*2+1] = lc >> 8; //        that DV is a big endian PCM
 514                  of = sys->audio_shuffle[i%half_ch+half_ch][j] +
 515                       (d - 8)/3 * sys->audio_stride;
 516                  pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
 517                  pcm[of*2+1] = rc >> 8; //        that DV is a big endian PCM
 518                  ++d;
 519               }
 520           }
 521
 522           frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 523         }
 524     }
 525
 526     return size;
 527 }
 528
 529 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
 530 {
 531     const uint8_t* as_pack;
 532     const DVprofile* sys;
 533     int freq, smpls, quant, i;
 534
 535     sys = dv_frame_profile(frame);
 536     as_pack = dv_extract_pack(frame, dv_audio_source);
 537     if (!as_pack || !sys) {    /* No audio ? */
 538         c->ach = 0;
 539         return -1;
 540     }
 541
 542     smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
 543     freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
 544     quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
 545     c->ach = (quant && freq == 2) ? 2 : 1;
 546
 547     /* The second stereo channel could appear in IEC 61834 stream only */
 548     if (c->ach == 2 && !c->ast[1]) {
 549         c->ast[1] = av_new_stream(c->fctx, 0);
 550         if (c->ast[1]) {
 551             c->ast[1]->codec.codec_type = CODEC_TYPE_AUDIO;
 552             c->ast[1]->codec.codec_id = CODEC_ID_PCM_S16LE;
 553         } else
 554             c->ach = 1;
 555     }
 556     for (i=0; i<c->ach; i++) {
 557        c->ast[i]->codec.sample_rate = dv_audio_frequency[freq];
 558        c->ast[i]->codec.channels = 2;
 559        c->ast[i]->codec.bit_rate = 2 * dv_audio_frequency[freq] * 16;
 560     }
 561
 562     return (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
 563 }
 564
 565 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
 566 {
 567     const DVprofile* sys;
 568     const uint8_t* vsc_pack;
 569     AVCodecContext* avctx;
 570     int apt, is16_9;
 571     int size = 0;
 572
 573     sys = dv_frame_profile(frame);
 574     if (sys) {
 575         avctx = &c->vst->codec;
 576
 577         avctx->frame_rate = sys->frame_rate;
 578         avctx->frame_rate_base = sys->frame_rate_base;
 579         avctx->width = sys->width;
 580         avctx->height = sys->height;
 581         avctx->pix_fmt = sys->pix_fmt;
 582
 583         vsc_pack = dv_extract_pack(frame, dv_video_control);
 584         apt = frame[4] & 0x07;
 585         is16_9 = (vsc_pack && (vsc_pack[2] & 0x07) == (apt?0x02:0x07));
 586         avctx->sample_aspect_ratio = sys->sar[is16_9];
 587
 588         size = sys->frame_size;
 589     }
 590     return size;
 591 }
 592
 593 /*
 594  * The following 6 functions constitute our interface to the world
 595  */
 596
 597 int dv_assemble_frame(DVMuxContext *c, AVStream* st,
 598                       const uint8_t* data, int data_size, uint8_t** frame)
 599 {
 600     uint8_t pcm[8192];
 601     int fsize, reqasize;
 602
 603     *frame = &c->frame_buf[0];
 604     if (c->has_audio && c->has_video) {  /* must be a stale frame */
 605         dv_format_frame(c, *frame);
 606         c->frames++;
 607         c->has_audio = c->has_video = 0;
 608     }
 609
 610     if (st->codec.codec_type == CODEC_TYPE_VIDEO) {
 611         /* FIXME: we have to have more sensible approach than this one */
 612         if (c->has_video)
 613             av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
 614
 615         dv_inject_video(c, data, *frame);
 616         c->has_video = 1;
 617         data_size = 0;
 618     }
 619
 620     reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
 621     fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
 622     if (st->codec.codec_type == CODEC_TYPE_AUDIO || (c->has_video && fsize >= reqasize)) {
 623         if (fsize + data_size >= reqasize && !c->has_audio) {
 624             if (fsize >= reqasize) {
 625                 fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
 626             } else {
 627                 fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);
 628                 memcpy(&pcm[fsize], &data[0], reqasize - fsize);
 629                 data += reqasize - fsize;
 630                 data_size -= reqasize - fsize;
 631             }
 632             dv_inject_audio(c, &pcm[0], *frame);
 633             c->has_audio = 1;
 634         }
 635
 636         /* FIXME: we have to have more sensible approach than this one */
 637         if (fifo_size(&c->audio_data, c->audio_data.rptr) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
 638             av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
 639         fifo_write(&c->audio_data, (uint8_t *)data, data_size, &c->audio_data.wptr);
 640     }
 641
 642     return (c->has_audio && c->has_video) ? c->sys->frame_size : 0;
 643 }
 644
 645 DVMuxContext* dv_init_mux(AVFormatContext* s)
 646 {
 647     DVMuxContext *c;
 648     AVStream *vst;
 649     AVStream *ast;
 650
 651     c = av_mallocz(sizeof(DVMuxContext));
 652     if (!c)
 653         return NULL;
 654
 655     if (s->nb_streams != 2)
 656         goto bail_out;
 657
 658     /* We have to sort out where audio and where video stream is */
 659     if (s->streams[0]->codec.codec_type == CODEC_TYPE_VIDEO &&
 660         s->streams[1]->codec.codec_type == CODEC_TYPE_AUDIO) {
 661        vst = s->streams[0];
 662        ast = s->streams[1];
 663     }
 664     else if (s->streams[1]->codec.codec_type == CODEC_TYPE_VIDEO &&
 665              s->streams[0]->codec.codec_type == CODEC_TYPE_AUDIO) {
 666            vst = s->streams[1];
 667            ast = s->streams[0];
 668     } else
 669         goto bail_out;
 670
 671     /* Some checks -- DV format is very picky about its incoming streams */
 672     if (vst->codec.codec_id != CODEC_ID_DVVIDEO ||
 673         ast->codec.codec_id != CODEC_ID_PCM_S16LE)
 674        goto bail_out;
 675     if (ast->codec.sample_rate != 48000 ||
 676         ast->codec.channels != 2)
 677        goto bail_out;
 678
 679     c->sys = dv_codec_profile(&vst->codec);
 680     if (!c->sys)
 681         goto bail_out;
 682
 683     /* Ok, everything seems to be in working order */
 684     c->frames = 0;
 685     c->has_audio = c->has_video = 0;
 686     c->start_time = time(NULL);
 687     c->aspect = 0; /* 4:3 is the default */
 688     if ((int)(av_q2d(vst->codec.sample_aspect_ratio) * vst->codec.width / vst->codec.height * 10) == 17) /* 16:9 */
 689         c->aspect = 0x07;
 690
 691     if (fifo_init(&c->audio_data, 100*AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0)
 692         goto bail_out;
 693
 694     dv_format_frame(c, &c->frame_buf[0]);
 695
 696     return c;
 697
 698 bail_out:
 699     av_free(c);
 700     return NULL;
 701 }
 702
 703 void dv_delete_mux(DVMuxContext *c)
 704 {
 705     fifo_free(&c->audio_data);
 706 }
 707
 708 DVDemuxContext* dv_init_demux(AVFormatContext *s)
 709 {
 710     DVDemuxContext *c;
 711
 712     c = av_mallocz(sizeof(DVDemuxContext));
 713     if (!c)
 714         return NULL;
 715
 716     c->vst = av_new_stream(s, 0);
 717     c->ast[0] = av_new_stream(s, 0);
 718     if (!c->vst || !c->ast[0])
 719         goto fail;
 720     c->fctx = s;
 721     c->ast[1] = NULL;
 722     c->ach = 0;
 723     c->audio_pkt[0].size = 0;
 724     c->audio_pkt[1].size = 0;
 725
 726     c->vst->codec.codec_type = CODEC_TYPE_VIDEO;
 727     c->vst->codec.codec_id = CODEC_ID_DVVIDEO;
 728     c->vst->codec.bit_rate = 25000000;
 729
 730     c->ast[0]->codec.codec_type = CODEC_TYPE_AUDIO;
 731     c->ast[0]->codec.codec_id = CODEC_ID_PCM_S16LE;
 732
 733     s->ctx_flags |= AVFMTCTX_NOHEADER;
 734
 735     return c;
 736
 737 fail:
 738     if (c->vst)
 739         av_free(c->vst);
 740     if (c->ast[0])
 741         av_free(c->ast[0]);
 742     av_free(c);
 743     return NULL;
 744 }
 745
 746 static void __destruct_pkt(struct AVPacket *pkt)
 747 {
 748     pkt->data = NULL; pkt->size = 0;
 749     return;
 750 }
 751
 752 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
 753 {
 754     int size = -1;
 755     int i;
 756
 757     for (i=0; i<c->ach; i++) {
 758        if (c->ast[i] && c->audio_pkt[i].size) {
 759            *pkt = c->audio_pkt[i];
 760            c->audio_pkt[i].size = 0;
 761            size = pkt->size;
 762            break;
 763        }
 764     }
 765
 766     return size;
 767 }
 768
 769 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
 770                       uint8_t* buf, int buf_size)
 771 {
 772     int size, i;
 773
 774     if (buf_size < 4 || buf_size < dv_frame_profile(buf)->frame_size)
 775         return -1;   /* Broken frame, or not enough data */
 776
 777     /* Queueing audio packet */
 778     /* FIXME: in case of no audio/bad audio we have to do something */
 779     size = dv_extract_audio_info(c, buf);
 780     c->audio_pkt[0].data = c->audio_pkt[1].data = NULL;
 781     for (i=0; i<c->ach; i++) {
 782        if (av_new_packet(&c->audio_pkt[i], size) < 0)
 783            return AVERROR_NOMEM;
 784        c->audio_pkt[i].stream_index = c->ast[i]->index;
 785     }
 786     dv_extract_audio(buf, c->audio_pkt[0].data, c->audio_pkt[1].data);
 787
 788     /* Now it's time to return video packet */
 789     size = dv_extract_video_info(c, buf);
 790     av_init_packet(pkt);
 791     pkt->destruct = __destruct_pkt;
 792     pkt->data     = buf;
 793     pkt->size     = size;
 794     pkt->flags   |= PKT_FLAG_KEY;
 795     pkt->stream_index = c->vst->id;
 796
 797     return size;
 798 }
 799
 800 /************************************************************
 801  * Implementation of the easiest DV storage of all -- raw DV.
 802  ************************************************************/
 803
 804 typedef struct RawDVContext {
 805     uint8_t         buf[144000];
 806     DVDemuxContext* dv_demux;
 807 } RawDVContext;
 808
 809 static int dv_read_header(AVFormatContext *s,
 810                           AVFormatParameters *ap)
 811 {
 812     RawDVContext *c = s->priv_data;
 813     c->dv_demux = dv_init_demux(s);
 814
 815     return c->dv_demux ? 0 : -1;
 816 }
 817
 818
 819 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
 820 {
 821     int size;
 822     RawDVContext *c = s->priv_data;
 823
 824     size = dv_get_packet(c->dv_demux, pkt);
 825
 826     if (size < 0) {
 827         if (get_buffer(&s->pb, c->buf, 4) <= 0)
 828             return -EIO;
 829
 830         size = dv_frame_profile(c->buf)->frame_size;
 831         if (get_buffer(&s->pb, c->buf + 4, size - 4) <= 0)
 832             return -EIO;
 833
 834         size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
 835     }
 836
 837     return size;
 838 }
 839
 840 static int dv_read_close(AVFormatContext *s)
 841 {
 842     RawDVContext *c = s->priv_data;
 843     av_free(c->dv_demux);
 844     return 0;
 845 }
 846
 847 static int dv_write_header(AVFormatContext *s)
 848 {
 849     s->priv_data = dv_init_mux(s);
 850     if (!s->priv_data) {
 851         av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
 852                     "Make sure that you supply exactly two streams:\n"
 853                     "     video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n");
 854         return -1;
 855     }
 856     return 0;
 857 }
 858
 859 static int dv_write_packet(struct AVFormatContext *s,
 860                            int stream_index,
 861                            const uint8_t *buf, int size, int64_t pts)
 862 {
 863     uint8_t* frame;
 864     int fsize;
 865
 866     fsize = dv_assemble_frame((DVMuxContext *)s->priv_data, s->streams[stream_index],
 867                               buf, size, &frame);
 868     if (fsize > 0) {
 869         put_buffer(&s->pb, frame, fsize);
 870         put_flush_packet(&s->pb);
 871     }
 872     return 0;
 873 }
 874
 875 /*
 876  * We might end up with some extra A/V data without matching counterpart.
 877  * E.g. video data without enough audio to write the complete frame.
 878  * Currently we simply drop the last frame. I don't know whether this
 879  * is the best strategy of all
 880  */
 881 static int dv_write_trailer(struct AVFormatContext *s)
 882 {
 883     dv_delete_mux((DVMuxContext *)s->priv_data);
 884     return 0;
 885 }
 886
 887 static AVInputFormat dv_iformat = {
 888     "dv",
 889     "DV video format",
 890     sizeof(RawDVContext),
 891     NULL,
 892     dv_read_header,
 893     dv_read_packet,
 894     dv_read_close,
 895     .extensions = "dv",
 896 };
 897
 898 static AVOutputFormat dv_oformat = {
 899     "dv",
 900     "DV video format",
 901     NULL,
 902     "dv",
 903     sizeof(DVMuxContext),
 904     CODEC_ID_PCM_S16LE,
 905     CODEC_ID_DVVIDEO,
 906     dv_write_header,
 907     dv_write_packet,
 908     dv_write_trailer,
 909 };
 910
 911 int dv_init(void)
 912 {
 913     av_register_input_format(&dv_iformat);
 914     av_register_output_format(&dv_oformat);
 915     return 0;
 916 }