Statistics
| Branch: | Revision:

ffmpeg / libavformat / dv.c @ 3ea1ce21

History | View | Annotate | Download (32 KB)

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
    int              frames;
37
    uint64_t         abytes;
38
};
39

    
40
struct DVMuxContext {
41
    const DVprofile*  sys;    /* Current DV profile. E.g.: 525/60, 625/50 */
42
    uint8_t     frame_buf[144000]; /* frame under contruction */
43
    FifoBuffer  audio_data;   /* Fifo for storing excessive amounts of PCM */
44
    int         frames;       /* Number of a current frame */
45
    time_t      start_time;   /* Start time of recording */
46
    uint8_t     aspect;       /* Aspect ID 0 - 4:3, 7 - 16:9 */
47
    int         has_audio;    /* frame under contruction has audio */
48
    int         has_video;    /* frame under contruction has video */
49
};
50

    
51
enum dv_section_type {
52
     dv_sect_header  = 0x1f,
53
     dv_sect_subcode = 0x3f,
54
     dv_sect_vaux    = 0x56,
55
     dv_sect_audio   = 0x76,
56
     dv_sect_video   = 0x96,
57
};
58

    
59
enum dv_pack_type {
60
     dv_header525     = 0x3f, /* see dv_write_pack for important details on */ 
61
     dv_header625     = 0xbf, /* these two packs */
62
     dv_timecode      = 0x13,
63
     dv_audio_source  = 0x50,
64
     dv_audio_control = 0x51,
65
     dv_audio_recdate = 0x52,
66
     dv_audio_rectime = 0x53,
67
     dv_video_source  = 0x60,
68
     dv_video_control = 0x61,
69
     dv_viedo_recdate = 0x62,
70
     dv_video_rectime = 0x63,
71
     dv_unknown_pack  = 0xff,
72
};
73

    
74

    
75

    
76
/*
77
 * The reason why the following three big ugly looking tables are
78
 * here is my lack of DV spec IEC 61834. The tables were basically 
79
 * constructed to make code that places packs in SSYB, VAUX and 
80
 * AAUX blocks very simple and table-driven. They conform to the
81
 * SMPTE 314M and the output of my personal DV camcorder, neither
82
 * of which is sufficient for a reliable DV stream producing. Thus
83
 * while code is still in development I'll be gathering input from
84
 * people with different DV equipment and modifying the tables to
85
 * accommodate all the quirks. Later on, if possible, some of them
86
 * will be folded into smaller tables and/or switch-if logic. For 
87
 * now, my only excuse is -- they don't eat up that much of a space.
88
 */
89

    
90
static const int dv_ssyb_packs_dist[12][6] = {
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, 0x13, 0x13, 0x13, 0x13, 0x13 },
96
    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
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
    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
102
    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
103
};
104

    
105
static const int dv_vaux_packs_dist[12][15] = {
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
    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
129
      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
130
};
131

    
132
static const int dv_aaux_packs_dist[12][9] = {
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
    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
144
    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
145
};
146

    
147
static inline uint16_t dv_audio_12to16(uint16_t sample)
148
{
149
    uint16_t shift, result;
150
    
151
    sample = (sample < 0x800) ? sample : sample | 0xf000;
152
    shift = (sample & 0xf00) >> 8;
153

    
154
    if (shift < 0x2 || shift > 0xd) {
155
        result = sample;
156
    } else if (shift < 0x8) {
157
        shift--;
158
        result = (sample - (256 * shift)) << shift;
159
    } else {
160
        shift = 0xe - shift;
161
        result = ((sample + ((256 * shift) + 1)) << shift) - 1;
162
    }
163

    
164
    return result;
165
}
166

    
167
static int dv_audio_frame_size(const DVprofile* sys, int frame)
168
{
169
    return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
170
                                            sizeof(sys->audio_samples_dist[0]))];
171
}
172

    
173
static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
174
{
175
    struct tm tc;
176
    time_t ct;
177
    int ltc_frame;
178

    
179
    buf[0] = (uint8_t)pack_id;
180
    switch (pack_id) {
181
    case dv_header525: /* I can't imagine why these two weren't defined as real */
182
    case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
183
          buf[1] = 0xf8 |               /* reserved -- always 1 */
184
                   (0 & 0x07);          /* APT: Track application ID */
185
          buf[2] = (0 << 7)    | /* TF1: audio data is 0 - valid; 1 - invalid */
186
                   (0x0f << 3) | /* reserved -- always 1 */
187
                   (0 & 0x07);   /* AP1: Audio application ID */
188
          buf[3] = (0 << 7)    | /* TF2: video data is 0 - valid; 1 - invalid */  
189
                   (0x0f << 3) | /* reserved -- always 1 */
190
                   (0 & 0x07);   /* AP2: Video application ID */
191
          buf[4] = (0 << 7)    | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */ 
192
                   (0x0f << 3) | /* reserved -- always 1 */
193
                   (0 & 0x07);   /* AP3: Subcode application ID */
194
          break;
195
    case dv_timecode:
196
          ct = (time_t)(c->frames / ((float)c->sys->frame_rate / 
197
                                     (float)c->sys->frame_rate_base));
198
          localtime_r(&ct, &tc);
199
          /* 
200
           * LTC drop-frame frame counter drops two frames (0 and 1) every 
201
           * minute, unless it is exactly divisible by 10
202
           */
203
          ltc_frame = (c->frames + 2*ct/60 - 2*ct/600) % c->sys->ltc_divisor;
204
          buf[1] = (0 << 7) | /* Color fame: 0 - unsync; 1 - sync mode */
205
                   (1 << 6) | /* Drop frame timecode: 0 - nondrop; 1 - drop */
206
                   ((ltc_frame / 10) << 4) | /* Tens of frames */
207
                   (ltc_frame % 10);         /* Units of frames */
208
          buf[2] = (1 << 7) | /* Biphase mark polarity correction: 0 - even; 1 - odd */
209
                   ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
210
                   (tc.tm_sec % 10);         /* Units of seconds */
211
          buf[3] = (1 << 7) | /* Binary group flag BGF0 */
212
                   ((tc.tm_min / 10) << 4) | /* Tens of minutes */
213
                   (tc.tm_min % 10);         /* Units of minutes */
214
          buf[4] = (1 << 7) | /* Binary group flag BGF2 */
215
                   (1 << 6) | /* Binary group flag BGF1 */
216
                   ((tc.tm_hour / 10) << 4) | /* Tens of hours */
217
                   (tc.tm_hour % 10);         /* Units of hours */
218
          break;
219
    case dv_audio_source:  /* AAUX source pack */
220
          buf[1] = (0 << 7) | /* locked mode       */
221
                   (1 << 6) | /* reserved -- always 1 */
222
                   (dv_audio_frame_size(c->sys, c->frames) -
223
                    c->sys->audio_min_samples[0]);
224
                              /* # of samples      */
225
          buf[2] = (0 << 7) | /* multi-stereo      */
226
                   (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
227
                   (0 << 4) | /* pair bit: 0 -- one pair of channels */
228
                    0;        /* audio mode        */
229
          buf[3] = (1 << 7) | /* res               */
230
                   (1 << 6) | /* multi-language flag */
231
                   (c->sys->dsf << 5) | /*  system: 60fields/50fields */
232
                    0;        /* definition: 0 -- SD (525/625) */
233
          buf[4] = (1 << 7) | /* emphasis: 1 -- off */
234
                   (0 << 6) | /* emphasis time constant: 0 -- reserved */
235
                   (0 << 3) | /* frequency: 0 -- 48Khz, 1 -- 44,1Khz, 2 -- 32Khz */
236
                    0;        /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */                            
237
          break;
238
    case dv_audio_control:
239
          buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
240
                   (1 << 4) | /* input source: 1 -- digital input */
241
                   (3 << 2) | /* compression: 3 -- no information */
242
                    0;        /* misc. info/SMPTE emphasis off */
243
          buf[2] = (1 << 7) | /* recording start point: 1 -- no */
244
                   (1 << 6) | /* recording end point: 1 -- no */
245
                   (1 << 3) | /* recording mode: 1 -- original */
246
                    7;         
247
          buf[3] = (1 << 7) | /* direction: 1 -- forward */
248
                    0x20;     /* speed */
249
          buf[4] = (1 << 7) | /* reserved -- always 1 */
250
                    0x7f;     /* genre category */
251
          break;
252
    case dv_audio_recdate:
253
    case dv_viedo_recdate:  /* VAUX recording date */
254
          ct = c->start_time + (time_t)(c->frames / 
255
               ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
256
          localtime_r(&ct, &tc);
257
          buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
258
                         /* 0xff is very likely to be "unknown" */
259
          buf[2] = (3 << 6) | /* reserved -- always 1 */
260
                   ((tc.tm_mday / 10) << 4) | /* Tens of day */
261
                   (tc.tm_mday % 10);         /* Units of day */
262
          buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
263
                   (tc.tm_mon % 10);         /* Units of month */
264
          buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
265
                   (tc.tm_year % 10);                 /* Units of year */
266
          break;
267
    case dv_audio_rectime:  /* AAUX recording time */
268
    case dv_video_rectime:  /* VAUX recording time */
269
          ct = c->start_time + (time_t)(c->frames / 
270
               ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
271
          localtime_r(&ct, &tc);
272
          buf[1] = (3 << 6) | /* reserved -- always 1 */
273
                   0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
274
          buf[2] = (1 << 7) | /* reserved -- always 1 */ 
275
                   ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
276
                   (tc.tm_sec % 10);         /* Units of seconds */
277
          buf[3] = (1 << 7) | /* reserved -- always 1 */
278
                   ((tc.tm_min / 10) << 4) | /* Tens of minutes */
279
                   (tc.tm_min % 10);         /* Units of minutes */
280
          buf[4] = (3 << 6) | /* reserved -- always 1 */ 
281
                   ((tc.tm_hour / 10) << 4) | /* Tens of hours */
282
                   (tc.tm_hour % 10);         /* Units of hours */
283
          break;
284
    case dv_video_source:
285
          buf[1] = 0xff; /* reserved -- always 1 */
286
          buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
287
                   (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
288
                   (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
289
                   0xf; /* reserved -- always 1 */
290
          buf[3] = (3 << 6) | /* reserved -- always 1 */
291
                   (c->sys->dsf << 5) | /*  system: 60fields/50fields */
292
                   0; /* signal type video compression */
293
          buf[4] = 0xff; /* VISC: 0xff -- no information */
294
          break;
295
    case dv_video_control:
296
          buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
297
                   0x3f; /* reserved -- always 1 */
298
          buf[2] = 0xc8 | /* reserved -- always b11001xxx */
299
                   c->aspect;
300
          buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
301
                   (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
302
                   (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
303
                   (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
304
                   0xc; /* reserved -- always b1100 */
305
          buf[4] = 0xff; /* reserved -- always 1 */
306
          break;
307
    default:
308
          buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
309
    }
310
    return 5;
311
}
312

    
313
static inline int dv_write_dif_id(enum dv_section_type t, uint8_t seq_num, 
314
                                  uint8_t dif_num, uint8_t* buf)
315
{
316
    buf[0] = (uint8_t)t;    /* Section type */
317
    buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
318
             (0 << 3) |     /* FSC: for 50Mb/s 0 - first channel; 1 - second */
319
             7;             /* reserved -- always 1 */
320
    buf[2] = dif_num;       /* DIF block number Video: 0-134, Audio: 0-8 */
321
    return 3;
322
}
323

    
324
static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
325
{
326
    if (syb_num == 0 || syb_num == 6) {
327
        buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
328
                 (0<<4)  | /* AP3 (Subcode application ID) */
329
                 0x0f;     /* reserved -- always 1 */
330
    } 
331
    else if (syb_num == 11) {
332
        buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
333
                 0x7f;     /* reserved -- always 1 */
334
    }
335
    else {
336
        buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
337
                 (0<<4)  | /* APT (Track application ID) */
338
                 0x0f;     /* reserved -- always 1 */
339
    }
340
    buf[1] = 0xf0 |            /* reserved -- always 1 */
341
             (syb_num & 0x0f); /* SSYB number 0 - 11   */
342
    buf[2] = 0xff;             /* reserved -- always 1 */
343
    return 3;
344
}
345

    
346
static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
347
{
348
    int i, j, k;
349
    
350
    for (i = 0; i < c->sys->difseg_size; i++) {
351
       memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
352
       
353
       /* DV header: 1DIF */
354
       buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
355
       buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
356
       buf += 72; /* unused bytes */
357
                                   
358
       /* DV subcode: 2DIFs */
359
       for (j = 0; j < 2; j++) {
360
          buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
361
          for (k = 0; k < 6; k++) {
362
             buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
363
             buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
364
          }
365
          buf += 29; /* unused bytes */
366
       }
367
       
368
       /* DV VAUX: 3DIFs */
369
       for (j = 0; j < 3; j++) {
370
          buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
371
          for (k = 0; k < 15 ; k++)
372
             buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
373
          buf += 2; /* unused bytes */
374
       } 
375
       
376
       /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
377
       for (j = 0; j < 135; j++) {
378
            if (j%15 == 0) {
379
                buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
380
                buf += dv_write_pack(dv_aaux_packs_dist[i][j/15], c, buf);
381
                buf += 72; /* shuffled PCM audio */
382
            }
383
            buf += dv_write_dif_id(dv_sect_video, i, j, buf);
384
            buf += 77; /* 1 video macro block: 1 bytes control
385
                                               4 * 14 bytes Y 8x8 data
386
                                                   10 bytes Cr 8x8 data
387
                                                   10 bytes Cb 8x8 data */
388
       }
389
    }
390
}
391

    
392
static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, uint8_t* frame_ptr)
393
{
394
    int i, j, d, of, size;
395
    size = 4 * dv_audio_frame_size(c->sys, c->frames);
396
    for (i = 0; i < c->sys->difseg_size; i++) {
397
       frame_ptr += 6 * 80; /* skip DIF segment header */
398
       for (j = 0; j < 9; j++) {
399
          for (d = 8; d < 80; d+=2) {
400
             of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
401
             if (of*2 >= size)
402
                 continue;
403
             
404
             frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
405
             frame_ptr[d+1] = pcm[of*2]; //        that DV is a big endian PCM       
406
          }
407
          frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
408
       }
409
    }
410
}
411

    
412
static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
413
{
414
    int i, j;
415
    int ptr = 0;
416

    
417
    for (i = 0; i < c->sys->difseg_size; i++) {
418
       ptr += 6 * 80; /* skip DIF segment header */
419
       for (j = 0; j < 135; j++) {
420
            if (j%15 == 0)
421
                ptr += 80; /* skip Audio DIF */
422
            ptr += 3;
423
            memcpy(frame_ptr + ptr, video_data + ptr, 77);
424
            ptr += 77;
425
       }
426
    }
427
}
428

    
429
/* 
430
 * This is the dumbest implementation of all -- it simply looks at
431
 * a fixed offset and if pack isn't there -- fails. We might want
432
 * to have a fallback mechanism for complete search of missing packs.
433
 */
434
static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
435
{
436
    int offs;
437
    
438
    switch (t) {
439
    case dv_audio_source:
440
          offs = (80*6 + 80*16*3 + 3);
441
          break;
442
    case dv_audio_control:
443
          offs = (80*6 + 80*16*4 + 3);
444
          break;
445
    case dv_video_control:
446
          offs = (80*5 + 48 + 5);
447
          break;
448
    default:
449
          return NULL;
450
    }   
451

    
452
    return (frame[offs] == t ? &frame[offs] : NULL);
453
}
454

    
455
/* 
456
 * There's a couple of assumptions being made here:
457
 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
458
 *    We can pass them upwards when ffmpeg will be ready to deal with them.
459
 * 2. We don't do software emphasis.
460
 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
461
 *    are converted into 16bit linear ones.
462
 */
463
static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2)
464
{
465
    int size, i, j, d, of, smpls, freq, quant, half_ch;
466
    uint16_t lc, rc;
467
    const DVprofile* sys;
468
    const uint8_t* as_pack;
469
    
470
    as_pack = dv_extract_pack(frame, dv_audio_source);
471
    if (!as_pack)    /* No audio ? */
472
        return 0;
473
   
474
    sys = dv_frame_profile(frame);
475
    smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
476
    freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
477
    quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
478
    
479
    if (quant > 1)
480
        return -1; /* Unsupported quantization */
481

    
482
    size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
483
    half_ch = sys->difseg_size/2;
484

    
485
    /* for each DIF segment */
486
    for (i = 0; i < sys->difseg_size; i++) {
487
       frame += 6 * 80; /* skip DIF segment header */
488
       if (quant == 1 && i == half_ch) {
489
           if (!pcm2)
490
               break;
491
           else
492
               pcm = pcm2;
493
       }
494

    
495
       for (j = 0; j < 9; j++) {
496
          for (d = 8; d < 80; d += 2) {
497
             if (quant == 0) {  /* 16bit quantization */
498
                 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
499
                 if (of*2 >= size)
500
                     continue;
501
                 
502
                 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
503
                 pcm[of*2+1] = frame[d]; //        that DV is a big endian PCM
504
                 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
505
                     pcm[of*2+1] = 0;
506
              } else {           /* 12bit quantization */
507
                 lc = ((uint16_t)frame[d] << 4) | 
508
                      ((uint16_t)frame[d+2] >> 4);
509
                 rc = ((uint16_t)frame[d+1] << 4) |
510
                      ((uint16_t)frame[d+2] & 0x0f);
511
                 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
512
                 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
513

    
514
                 of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
515
                 if (of*2 >= size)
516
                     continue;
517

    
518
                 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
519
                 pcm[of*2+1] = lc >> 8; //        that DV is a big endian PCM
520
                 of = sys->audio_shuffle[i%half_ch+half_ch][j] + 
521
                      (d - 8)/3 * sys->audio_stride;
522
                 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
523
                 pcm[of*2+1] = rc >> 8; //        that DV is a big endian PCM
524
                 ++d;
525
              }
526
          }
527
                
528
          frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
529
        }
530
    }
531

    
532
    return size;
533
}
534

    
535
static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
536
{
537
    const uint8_t* as_pack;
538
    const DVprofile* sys;
539
    int freq, smpls, quant, i;
540

    
541
    sys = dv_frame_profile(frame);
542
    as_pack = dv_extract_pack(frame, dv_audio_source);
543
    if (!as_pack || !sys) {    /* No audio ? */
544
        c->ach = 0;
545
        return -1;
546
    }
547
    
548
    smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
549
    freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
550
    quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
551
    c->ach = (quant && freq == 2) ? 2 : 1;
552

    
553
    /* The second stereo channel could appear in IEC 61834 stream only */
554
    if (c->ach == 2 && !c->ast[1]) {
555
        c->ast[1] = av_new_stream(c->fctx, 0);
556
        if (c->ast[1]) {
557
            av_set_pts_info(c->ast[1], 64, 1, 30000);
558
            c->ast[1]->codec.codec_type = CODEC_TYPE_AUDIO;
559
            c->ast[1]->codec.codec_id = CODEC_ID_PCM_S16LE;
560
        } else
561
            c->ach = 1;
562
    }
563
    for (i=0; i<c->ach; i++) {
564
       c->ast[i]->codec.sample_rate = dv_audio_frequency[freq];
565
       c->ast[i]->codec.channels = 2;
566
       c->ast[i]->codec.bit_rate = 2 * dv_audio_frequency[freq] * 16;
567
    }
568
    
569
    return (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
570
}
571

    
572
static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
573
{
574
    const DVprofile* sys; 
575
    const uint8_t* vsc_pack;
576
    AVCodecContext* avctx;
577
    int apt, is16_9;
578
    int size = 0;
579
    
580
    sys = dv_frame_profile(frame);
581
    if (sys) {
582
        avctx = &c->vst->codec;
583
        
584
        avctx->frame_rate = sys->frame_rate;
585
        avctx->frame_rate_base = sys->frame_rate_base;
586
        avctx->width = sys->width;
587
        avctx->height = sys->height;
588
        avctx->pix_fmt = sys->pix_fmt;
589
        
590
        /* finding out SAR is a little bit messy */
591
        vsc_pack = dv_extract_pack(frame, dv_video_control);
592
        apt = frame[4] & 0x07;
593
        is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
594
                               (!apt && (vsc_pack[2] & 0x07) == 0x07)));
595
        avctx->sample_aspect_ratio = sys->sar[is16_9];
596
        
597
        size = sys->frame_size;
598
    }
599
    return size;
600
}
601

    
602
/* 
603
 * The following 6 functions constitute our interface to the world
604
 */
605

    
606
int dv_assemble_frame(DVMuxContext *c, AVStream* st,
607
                      const uint8_t* data, int data_size, uint8_t** frame)
608
{
609
    uint8_t pcm[8192];
610
    int fsize, reqasize;
611
   
612
    *frame = &c->frame_buf[0];
613
    if (c->has_audio && c->has_video) {  /* must be a stale frame */
614
        dv_format_frame(c, *frame);
615
        c->frames++;
616
        c->has_audio = c->has_video = 0;
617
    }
618
    
619
    if (st->codec.codec_type == CODEC_TYPE_VIDEO) {
620
        /* FIXME: we have to have more sensible approach than this one */
621
        if (c->has_video)
622
            av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
623
            
624
        dv_inject_video(c, data, *frame);
625
        c->has_video = 1;
626
        data_size = 0;
627
    } 
628
    
629
    reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
630
    fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
631
    if (st->codec.codec_type == CODEC_TYPE_AUDIO || (c->has_video && fsize >= reqasize)) { 
632
        if (fsize + data_size >= reqasize && !c->has_audio) {
633
            if (fsize >= reqasize) {
634
                fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
635
            } else {
636
                fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);
637
                memcpy(&pcm[fsize], &data[0], reqasize - fsize);
638
                data += reqasize - fsize;
639
                data_size -= reqasize - fsize;
640
            }
641
            dv_inject_audio(c, &pcm[0], *frame);
642
            c->has_audio = 1;
643
        }
644
    
645
        /* FIXME: we have to have more sensible approach than this one */
646
        if (fifo_size(&c->audio_data, c->audio_data.rptr) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
647
            av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
648
        fifo_write(&c->audio_data, (uint8_t *)data, data_size, &c->audio_data.wptr);
649
    }
650

    
651
    return (c->has_audio && c->has_video) ? c->sys->frame_size : 0;
652
}
653

    
654
DVMuxContext* dv_init_mux(AVFormatContext* s)
655
{
656
    DVMuxContext *c;
657
    AVStream *vst;
658
    AVStream *ast;
659

    
660
    c = av_mallocz(sizeof(DVMuxContext));
661
    if (!c)
662
        return NULL;
663

    
664
    if (s->nb_streams != 2)
665
        goto bail_out;
666

    
667
    /* We have to sort out where audio and where video stream is */
668
    if (s->streams[0]->codec.codec_type == CODEC_TYPE_VIDEO &&
669
        s->streams[1]->codec.codec_type == CODEC_TYPE_AUDIO) {
670
       vst = s->streams[0];
671
       ast = s->streams[1];
672
    }
673
    else if (s->streams[1]->codec.codec_type == CODEC_TYPE_VIDEO &&
674
             s->streams[0]->codec.codec_type == CODEC_TYPE_AUDIO) {
675
           vst = s->streams[1];
676
           ast = s->streams[0];
677
    } else
678
        goto bail_out;
679
  
680
    /* Some checks -- DV format is very picky about its incoming streams */
681
    if (vst->codec.codec_id != CODEC_ID_DVVIDEO ||
682
        ast->codec.codec_id != CODEC_ID_PCM_S16LE)
683
       goto bail_out;
684
    if (ast->codec.sample_rate != 48000 ||
685
        ast->codec.channels != 2)
686
       goto bail_out;
687

    
688
    c->sys = dv_codec_profile(&vst->codec);
689
    if (!c->sys)
690
        goto bail_out;
691
    
692
    /* Ok, everything seems to be in working order */
693
    c->frames = 0;
694
    c->has_audio = c->has_video = 0;
695
    c->start_time = (time_t)s->timestamp;
696
    c->aspect = 0; /* 4:3 is the default */
697
    if ((int)(av_q2d(vst->codec.sample_aspect_ratio) * vst->codec.width / vst->codec.height * 10) == 17) /* 16:9 */ 
698
        c->aspect = 0x07;
699

    
700
    if (fifo_init(&c->audio_data, 100*AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0)
701
        goto bail_out;
702

    
703
    dv_format_frame(c, &c->frame_buf[0]);
704

    
705
    return c;
706
    
707
bail_out:
708
    av_free(c);
709
    return NULL;
710
}
711

    
712
void dv_delete_mux(DVMuxContext *c)
713
{    
714
    fifo_free(&c->audio_data);
715
}
716

    
717
DVDemuxContext* dv_init_demux(AVFormatContext *s)
718
{
719
    DVDemuxContext *c;
720

    
721
    c = av_mallocz(sizeof(DVDemuxContext));
722
    if (!c)
723
        return NULL;
724

    
725
    c->vst = av_new_stream(s, 0);
726
    c->ast[0] = av_new_stream(s, 0);
727
    if (!c->vst || !c->ast[0])
728
        goto fail;
729
    av_set_pts_info(c->vst, 64, 1, 30000);
730
    av_set_pts_info(c->ast[0], 64, 1, 30000);
731

    
732
    c->fctx = s;
733
    c->ast[1] = NULL;
734
    c->ach = 0;
735
    c->frames = 0;
736
    c->abytes = 0;
737
    c->audio_pkt[0].size = 0;
738
    c->audio_pkt[1].size = 0;
739
    
740
    c->vst->codec.codec_type = CODEC_TYPE_VIDEO;
741
    c->vst->codec.codec_id = CODEC_ID_DVVIDEO;
742
    c->vst->codec.bit_rate = 25000000;
743
    
744
    c->ast[0]->codec.codec_type = CODEC_TYPE_AUDIO;
745
    c->ast[0]->codec.codec_id = CODEC_ID_PCM_S16LE;
746
   
747
    s->ctx_flags |= AVFMTCTX_NOHEADER; 
748
    
749
    return c;
750
    
751
fail:
752
    if (c->vst)
753
        av_free(c->vst);
754
    if (c->ast[0])
755
        av_free(c->ast[0]);
756
    av_free(c);
757
    return NULL;
758
}
759

    
760
static void __destruct_pkt(struct AVPacket *pkt)
761
{
762
    pkt->data = NULL; pkt->size = 0;
763
    return;
764
}
765

    
766
int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
767
{
768
    int size = -1;
769
    int i;
770

    
771
    for (i=0; i<c->ach; i++) {
772
       if (c->ast[i] && c->audio_pkt[i].size) {
773
           *pkt = c->audio_pkt[i];
774
           c->audio_pkt[i].size = 0;
775
           size = pkt->size;
776
           break;
777
       }
778
    }
779

    
780
    return size;
781
}
782

    
783
int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt, 
784
                      uint8_t* buf, int buf_size)
785
{
786
    int size, i;
787
    const DVprofile* sys = dv_frame_profile(buf);
788
   
789
    if (buf_size < 4 || buf_size < sys->frame_size)
790
        return -1;   /* Broken frame, or not enough data */
791

    
792
    /* Queueing audio packet */
793
    /* FIXME: in case of no audio/bad audio we have to do something */
794
    size = dv_extract_audio_info(c, buf);
795
    c->audio_pkt[0].data = c->audio_pkt[1].data = NULL;
796
    for (i=0; i<c->ach; i++) {
797
       if (av_new_packet(&c->audio_pkt[i], size) < 0)
798
           return AVERROR_NOMEM;
799
       c->audio_pkt[i].stream_index = c->ast[i]->index;
800
       c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec.bit_rate;
801
       c->audio_pkt[i].flags |= PKT_FLAG_KEY;
802
    }
803
    dv_extract_audio(buf, c->audio_pkt[0].data, c->audio_pkt[1].data);
804
    c->abytes += size;
805
    
806
    /* Now it's time to return video packet */
807
    size = dv_extract_video_info(c, buf);
808
    av_init_packet(pkt);
809
    pkt->destruct = __destruct_pkt;
810
    pkt->data     = buf;
811
    pkt->size     = size; 
812
    pkt->flags   |= PKT_FLAG_KEY;
813
    pkt->stream_index = c->vst->id;
814
    pkt->pts      = c->frames * sys->frame_rate_base * (30000/sys->frame_rate);
815
    
816
    c->frames++;
817

    
818
    return size;
819
}
820
                           
821
int64_t dv_frame_offset(DVDemuxContext *c, int64_t timestamp)
822
{
823
    const DVprofile* sys = dv_codec_profile(&c->vst->codec);
824

    
825
    return sys->frame_size * ((timestamp * sys->frame_rate) / 
826
                              (AV_TIME_BASE * sys->frame_rate_base));
827
}
828

    
829
/************************************************************
830
 * Implementation of the easiest DV storage of all -- raw DV.
831
 ************************************************************/
832
 
833
typedef struct RawDVContext {
834
    uint8_t         buf[144000];
835
    DVDemuxContext* dv_demux;
836
} RawDVContext;
837

    
838
static int dv_read_header(AVFormatContext *s,
839
                          AVFormatParameters *ap)
840
{
841
    RawDVContext *c = s->priv_data;
842
    c->dv_demux = dv_init_demux(s);
843
   
844
    return c->dv_demux ? 0 : -1;
845
}
846

    
847

    
848
static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
849
{
850
    int size;
851
    RawDVContext *c = s->priv_data; 
852
  
853
    size = dv_get_packet(c->dv_demux, pkt);
854
    
855
    if (size < 0) {
856
        if (get_buffer(&s->pb, c->buf, 4) <= 0) 
857
            return AVERROR_IO;
858
    
859
        size = dv_frame_profile(c->buf)->frame_size;
860
        if (get_buffer(&s->pb, c->buf + 4, size - 4) <= 0)
861
            return AVERROR_IO;
862

    
863
        size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
864
    } 
865
    
866
    return size;
867
}
868

    
869
static int dv_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp)
870
{
871
    RawDVContext *c = s->priv_data;
872

    
873
    return url_fseek(&s->pb, dv_frame_offset(c->dv_demux, timestamp), SEEK_SET);
874
}
875

    
876
static int dv_read_close(AVFormatContext *s)
877
{
878
    RawDVContext *c = s->priv_data;
879
    av_free(c->dv_demux);
880
    return 0;
881
}
882

    
883
static int dv_write_header(AVFormatContext *s)
884
{
885
    s->priv_data = dv_init_mux(s);
886
    if (!s->priv_data) {
887
        av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
888
                    "Make sure that you supply exactly two streams:\n"
889
                    "     video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n");
890
        return -1;
891
    }
892
    return 0;
893
}
894

    
895
static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
896
{
897
    uint8_t* frame;
898
    int fsize;
899
   
900
    fsize = dv_assemble_frame((DVMuxContext *)s->priv_data, s->streams[pkt->stream_index],
901
                              pkt->data, pkt->size, &frame);
902
    if (fsize > 0) {
903
        put_buffer(&s->pb, frame, fsize); 
904
        put_flush_packet(&s->pb);
905
    } 
906
    return 0;
907
}
908

    
909
/* 
910
 * We might end up with some extra A/V data without matching counterpart.
911
 * E.g. video data without enough audio to write the complete frame.
912
 * Currently we simply drop the last frame. I don't know whether this 
913
 * is the best strategy of all
914
 */
915
static int dv_write_trailer(struct AVFormatContext *s)
916
{
917
    dv_delete_mux((DVMuxContext *)s->priv_data);
918
    return 0;
919
}
920

    
921
static AVInputFormat dv_iformat = {
922
    "dv",
923
    "DV video format",
924
    sizeof(RawDVContext),
925
    NULL,
926
    dv_read_header,
927
    dv_read_packet,
928
    dv_read_close,
929
    dv_read_seek,
930
    .extensions = "dv,dif",
931
};
932

    
933
static AVOutputFormat dv_oformat = {
934
    "dv",
935
    "DV video format",
936
    NULL,
937
    "dv",
938
    sizeof(DVMuxContext),
939
    CODEC_ID_PCM_S16LE,
940
    CODEC_ID_DVVIDEO,
941
    dv_write_header,
942
    dv_write_packet,
943
    dv_write_trailer,
944
};
945

    
946
int ff_dv_init(void)
947
{
948
    av_register_input_format(&dv_iformat);
949
    av_register_output_format(&dv_oformat);
950
    return 0;
951
}