2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnick
5 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
6 * of DV technical info.
9 * Copyright (c) 2002 Fabrice Bellard.
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.
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.
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
30 struct DVDemuxContext {
31 AVFormatContext* fctx;
34 AVPacket audio_pkt[2];
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 */
51 enum dv_section_type {
52 dv_sect_header = 0x1f,
53 dv_sect_subcode = 0x3f,
60 dv_header525 = 0x3f, /* see dv_write_pack for important details on */
61 dv_header625 = 0xbf, /* these two packs */
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,
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.
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 },
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 },
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 },
147 static inline uint16_t dv_audio_12to16(uint16_t sample)
149 uint16_t shift, result;
151 sample = (sample < 0x800) ? sample : sample | 0xf000;
152 shift = (sample & 0xf00) >> 8;
154 if (shift < 0x2 || shift > 0xd) {
156 } else if (shift < 0x8) {
158 result = (sample - (256 * shift)) << shift;
161 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
167 static int dv_audio_frame_size(const DVprofile* sys, int frame)
169 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
170 sizeof(sys->audio_samples_dist[0]))];
173 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
179 buf[0] = (uint8_t)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 */
196 ct = (time_t)(c->frames / ((float)c->sys->frame_rate /
197 (float)c->sys->frame_rate_base));
200 * LTC drop-frame frame counter drops two frames (0 and 1) every
201 * minute, unless it is exactly divisible by 10
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 */
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]);
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 */
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 */
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 */
247 buf[3] = (1 << 7) | /* direction: 1 -- forward */
249 buf[4] = (1 << 7) | /* reserved -- always 1 */
250 0x7f; /* genre category */
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));
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 + 1) / 10) << 4) | /* Tens of month */
264 ((tc.tm_mon + 1) % 10); /* Units of month */
265 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
266 (tc.tm_year % 10); /* Units of year */
268 case dv_audio_rectime: /* AAUX recording time */
269 case dv_video_rectime: /* VAUX recording time */
270 ct = c->start_time + (time_t)(c->frames /
271 ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
273 buf[1] = (3 << 6) | /* reserved -- always 1 */
274 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
275 buf[2] = (1 << 7) | /* reserved -- always 1 */
276 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
277 (tc.tm_sec % 10); /* Units of seconds */
278 buf[3] = (1 << 7) | /* reserved -- always 1 */
279 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
280 (tc.tm_min % 10); /* Units of minutes */
281 buf[4] = (3 << 6) | /* reserved -- always 1 */
282 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
283 (tc.tm_hour % 10); /* Units of hours */
285 case dv_video_source:
286 buf[1] = 0xff; /* reserved -- always 1 */
287 buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
288 (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
289 (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
290 0xf; /* reserved -- always 1 */
291 buf[3] = (3 << 6) | /* reserved -- always 1 */
292 (c->sys->dsf << 5) | /* system: 60fields/50fields */
293 0; /* signal type video compression */
294 buf[4] = 0xff; /* VISC: 0xff -- no information */
296 case dv_video_control:
297 buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
298 0x3f; /* reserved -- always 1 */
299 buf[2] = 0xc8 | /* reserved -- always b11001xxx */
301 buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
302 (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
303 (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
304 (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
305 0xc; /* reserved -- always b1100 */
306 buf[4] = 0xff; /* reserved -- always 1 */
309 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
314 static inline int dv_write_dif_id(enum dv_section_type t, uint8_t seq_num,
315 uint8_t dif_num, uint8_t* buf)
317 buf[0] = (uint8_t)t; /* Section type */
318 buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
319 (0 << 3) | /* FSC: for 50Mb/s 0 - first channel; 1 - second */
320 7; /* reserved -- always 1 */
321 buf[2] = dif_num; /* DIF block number Video: 0-134, Audio: 0-8 */
325 static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
327 if (syb_num == 0 || syb_num == 6) {
328 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
329 (0<<4) | /* AP3 (Subcode application ID) */
330 0x0f; /* reserved -- always 1 */
332 else if (syb_num == 11) {
333 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
334 0x7f; /* reserved -- always 1 */
337 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
338 (0<<4) | /* APT (Track application ID) */
339 0x0f; /* reserved -- always 1 */
341 buf[1] = 0xf0 | /* reserved -- always 1 */
342 (syb_num & 0x0f); /* SSYB number 0 - 11 */
343 buf[2] = 0xff; /* reserved -- always 1 */
347 static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
351 for (i = 0; i < c->sys->difseg_size; i++) {
352 memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
354 /* DV header: 1DIF */
355 buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
356 buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
357 buf += 72; /* unused bytes */
359 /* DV subcode: 2DIFs */
360 for (j = 0; j < 2; j++) {
361 buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
362 for (k = 0; k < 6; k++) {
363 buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
364 buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
366 buf += 29; /* unused bytes */
370 for (j = 0; j < 3; j++) {
371 buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
372 for (k = 0; k < 15 ; k++)
373 buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
374 buf += 2; /* unused bytes */
377 /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
378 for (j = 0; j < 135; j++) {
380 buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
381 buf += dv_write_pack(dv_aaux_packs_dist[i][j/15], c, buf);
382 buf += 72; /* shuffled PCM audio */
384 buf += dv_write_dif_id(dv_sect_video, i, j, buf);
385 buf += 77; /* 1 video macro block: 1 bytes control
386 4 * 14 bytes Y 8x8 data
388 10 bytes Cb 8x8 data */
393 static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, uint8_t* frame_ptr)
395 int i, j, d, of, size;
396 size = 4 * dv_audio_frame_size(c->sys, c->frames);
397 for (i = 0; i < c->sys->difseg_size; i++) {
398 frame_ptr += 6 * 80; /* skip DIF segment header */
399 for (j = 0; j < 9; j++) {
400 for (d = 8; d < 80; d+=2) {
401 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
405 frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
406 frame_ptr[d+1] = pcm[of*2]; // that DV is a big endian PCM
408 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
413 static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
418 for (i = 0; i < c->sys->difseg_size; i++) {
419 ptr += 6 * 80; /* skip DIF segment header */
420 for (j = 0; j < 135; j++) {
422 ptr += 80; /* skip Audio DIF */
424 memcpy(frame_ptr + ptr, video_data + ptr, 77);
431 * This is the dumbest implementation of all -- it simply looks at
432 * a fixed offset and if pack isn't there -- fails. We might want
433 * to have a fallback mechanism for complete search of missing packs.
435 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
440 case dv_audio_source:
441 offs = (80*6 + 80*16*3 + 3);
443 case dv_audio_control:
444 offs = (80*6 + 80*16*4 + 3);
446 case dv_video_control:
447 offs = (80*5 + 48 + 5);
453 return (frame[offs] == t ? &frame[offs] : NULL);
457 * There's a couple of assumptions being made here:
458 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
459 * We can pass them upwards when ffmpeg will be ready to deal with them.
460 * 2. We don't do software emphasis.
461 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
462 * are converted into 16bit linear ones.
464 static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2)
466 int size, i, j, d, of, smpls, freq, quant, half_ch;
468 const DVprofile* sys;
469 const uint8_t* as_pack;
471 as_pack = dv_extract_pack(frame, dv_audio_source);
472 if (!as_pack) /* No audio ? */
475 sys = dv_frame_profile(frame);
476 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
477 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
478 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
481 return -1; /* Unsupported quantization */
483 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
484 half_ch = sys->difseg_size/2;
486 /* for each DIF segment */
487 for (i = 0; i < sys->difseg_size; i++) {
488 frame += 6 * 80; /* skip DIF segment header */
489 if (quant == 1 && i == half_ch) {
496 for (j = 0; j < 9; j++) {
497 for (d = 8; d < 80; d += 2) {
498 if (quant == 0) { /* 16bit quantization */
499 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
503 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
504 pcm[of*2+1] = frame[d]; // that DV is a big endian PCM
505 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
507 } else { /* 12bit quantization */
508 lc = ((uint16_t)frame[d] << 4) |
509 ((uint16_t)frame[d+2] >> 4);
510 rc = ((uint16_t)frame[d+1] << 4) |
511 ((uint16_t)frame[d+2] & 0x0f);
512 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
513 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
515 of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
519 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
520 pcm[of*2+1] = lc >> 8; // that DV is a big endian PCM
521 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
522 (d - 8)/3 * sys->audio_stride;
523 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
524 pcm[of*2+1] = rc >> 8; // that DV is a big endian PCM
529 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
536 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
538 const uint8_t* as_pack;
539 const DVprofile* sys;
540 int freq, smpls, quant, i;
542 sys = dv_frame_profile(frame);
543 as_pack = dv_extract_pack(frame, dv_audio_source);
544 if (!as_pack || !sys) { /* No audio ? */
549 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
550 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
551 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
552 c->ach = (quant && freq == 2) ? 2 : 1;
554 /* The second stereo channel could appear in IEC 61834 stream only */
555 if (c->ach == 2 && !c->ast[1]) {
556 c->ast[1] = av_new_stream(c->fctx, 0);
558 av_set_pts_info(c->ast[1], 64, 1, 30000);
559 c->ast[1]->codec.codec_type = CODEC_TYPE_AUDIO;
560 c->ast[1]->codec.codec_id = CODEC_ID_PCM_S16LE;
564 for (i=0; i<c->ach; i++) {
565 c->ast[i]->codec.sample_rate = dv_audio_frequency[freq];
566 c->ast[i]->codec.channels = 2;
567 c->ast[i]->codec.bit_rate = 2 * dv_audio_frequency[freq] * 16;
570 return (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
573 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
575 const DVprofile* sys;
576 const uint8_t* vsc_pack;
577 AVCodecContext* avctx;
581 sys = dv_frame_profile(frame);
583 avctx = &c->vst->codec;
585 avctx->frame_rate = sys->frame_rate;
586 avctx->frame_rate_base = sys->frame_rate_base;
587 avctx->width = sys->width;
588 avctx->height = sys->height;
589 avctx->pix_fmt = sys->pix_fmt;
591 /* finding out SAR is a little bit messy */
592 vsc_pack = dv_extract_pack(frame, dv_video_control);
593 apt = frame[4] & 0x07;
594 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
595 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
596 avctx->sample_aspect_ratio = sys->sar[is16_9];
598 size = sys->frame_size;
604 * The following 6 functions constitute our interface to the world
607 int dv_assemble_frame(DVMuxContext *c, AVStream* st,
608 const uint8_t* data, int data_size, uint8_t** frame)
613 *frame = &c->frame_buf[0];
614 if (c->has_audio && c->has_video) { /* must be a stale frame */
615 dv_format_frame(c, *frame);
617 c->has_audio = c->has_video = 0;
620 if (st->codec.codec_type == CODEC_TYPE_VIDEO) {
621 /* FIXME: we have to have more sensible approach than this one */
623 av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
625 dv_inject_video(c, data, *frame);
630 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
631 fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
632 if (st->codec.codec_type == CODEC_TYPE_AUDIO || (c->has_video && fsize >= reqasize)) {
633 if (fsize + data_size >= reqasize && !c->has_audio) {
634 if (fsize >= reqasize) {
635 fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
637 fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);
638 memcpy(&pcm[fsize], &data[0], reqasize - fsize);
639 data += reqasize - fsize;
640 data_size -= reqasize - fsize;
642 dv_inject_audio(c, &pcm[0], *frame);
646 /* FIXME: we have to have more sensible approach than this one */
647 if (fifo_size(&c->audio_data, c->audio_data.rptr) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
648 av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
649 fifo_write(&c->audio_data, (uint8_t *)data, data_size, &c->audio_data.wptr);
652 return (c->has_audio && c->has_video) ? c->sys->frame_size : 0;
655 DVMuxContext* dv_init_mux(AVFormatContext* s)
661 c = av_mallocz(sizeof(DVMuxContext));
665 if (s->nb_streams != 2)
668 /* We have to sort out where audio and where video stream is */
669 if (s->streams[0]->codec.codec_type == CODEC_TYPE_VIDEO &&
670 s->streams[1]->codec.codec_type == CODEC_TYPE_AUDIO) {
674 else if (s->streams[1]->codec.codec_type == CODEC_TYPE_VIDEO &&
675 s->streams[0]->codec.codec_type == CODEC_TYPE_AUDIO) {
681 /* Some checks -- DV format is very picky about its incoming streams */
682 if (vst->codec.codec_id != CODEC_ID_DVVIDEO ||
683 ast->codec.codec_id != CODEC_ID_PCM_S16LE)
685 if (ast->codec.sample_rate != 48000 ||
686 ast->codec.channels != 2)
689 c->sys = dv_codec_profile(&vst->codec);
693 /* Ok, everything seems to be in working order */
695 c->has_audio = c->has_video = 0;
696 c->start_time = (time_t)s->timestamp;
697 c->aspect = 0; /* 4:3 is the default */
698 if ((int)(av_q2d(vst->codec.sample_aspect_ratio) * vst->codec.width / vst->codec.height * 10) == 17) /* 16:9 */
701 if (fifo_init(&c->audio_data, 100*AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0)
704 dv_format_frame(c, &c->frame_buf[0]);
713 void dv_delete_mux(DVMuxContext *c)
715 fifo_free(&c->audio_data);
718 DVDemuxContext* dv_init_demux(AVFormatContext *s)
722 c = av_mallocz(sizeof(DVDemuxContext));
726 c->vst = av_new_stream(s, 0);
727 c->ast[0] = av_new_stream(s, 0);
728 if (!c->vst || !c->ast[0])
730 av_set_pts_info(c->vst, 64, 1, 30000);
731 av_set_pts_info(c->ast[0], 64, 1, 30000);
738 c->audio_pkt[0].size = 0;
739 c->audio_pkt[1].size = 0;
741 c->vst->codec.codec_type = CODEC_TYPE_VIDEO;
742 c->vst->codec.codec_id = CODEC_ID_DVVIDEO;
743 c->vst->codec.bit_rate = 25000000;
745 c->ast[0]->codec.codec_type = CODEC_TYPE_AUDIO;
746 c->ast[0]->codec.codec_id = CODEC_ID_PCM_S16LE;
748 s->ctx_flags |= AVFMTCTX_NOHEADER;
761 static void __destruct_pkt(struct AVPacket *pkt)
763 pkt->data = NULL; pkt->size = 0;
767 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
772 for (i=0; i<c->ach; i++) {
773 if (c->ast[i] && c->audio_pkt[i].size) {
774 *pkt = c->audio_pkt[i];
775 c->audio_pkt[i].size = 0;
784 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
785 uint8_t* buf, int buf_size)
788 const DVprofile* sys = dv_frame_profile(buf);
790 if (buf_size < 4 || buf_size < sys->frame_size)
791 return -1; /* Broken frame, or not enough data */
793 /* Queueing audio packet */
794 /* FIXME: in case of no audio/bad audio we have to do something */
795 size = dv_extract_audio_info(c, buf);
796 c->audio_pkt[0].data = c->audio_pkt[1].data = NULL;
797 for (i=0; i<c->ach; i++) {
798 if (av_new_packet(&c->audio_pkt[i], size) < 0)
799 return AVERROR_NOMEM;
800 c->audio_pkt[i].stream_index = c->ast[i]->index;
801 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec.bit_rate;
802 c->audio_pkt[i].flags |= PKT_FLAG_KEY;
804 dv_extract_audio(buf, c->audio_pkt[0].data, c->audio_pkt[1].data);
807 /* Now it's time to return video packet */
808 size = dv_extract_video_info(c, buf);
810 pkt->destruct = __destruct_pkt;
813 pkt->flags |= PKT_FLAG_KEY;
814 pkt->stream_index = c->vst->id;
815 pkt->pts = c->frames * sys->frame_rate_base * (30000/sys->frame_rate);
822 int64_t dv_frame_offset(DVDemuxContext *c, int64_t timestamp)
824 const DVprofile* sys = dv_codec_profile(&c->vst->codec);
826 return sys->frame_size * ((timestamp * sys->frame_rate) /
827 (AV_TIME_BASE * sys->frame_rate_base));
830 /************************************************************
831 * Implementation of the easiest DV storage of all -- raw DV.
832 ************************************************************/
834 typedef struct RawDVContext {
836 DVDemuxContext* dv_demux;
839 static int dv_read_header(AVFormatContext *s,
840 AVFormatParameters *ap)
842 RawDVContext *c = s->priv_data;
843 c->dv_demux = dv_init_demux(s);
845 return c->dv_demux ? 0 : -1;
849 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
852 RawDVContext *c = s->priv_data;
854 size = dv_get_packet(c->dv_demux, pkt);
857 if (get_buffer(&s->pb, c->buf, 4) <= 0)
860 size = dv_frame_profile(c->buf)->frame_size;
861 if (get_buffer(&s->pb, c->buf + 4, size - 4) <= 0)
864 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
870 static int dv_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp)
872 RawDVContext *c = s->priv_data;
874 return url_fseek(&s->pb, dv_frame_offset(c->dv_demux, timestamp), SEEK_SET);
877 static int dv_read_close(AVFormatContext *s)
879 RawDVContext *c = s->priv_data;
880 av_free(c->dv_demux);
884 static int dv_write_header(AVFormatContext *s)
886 s->priv_data = dv_init_mux(s);
888 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
889 "Make sure that you supply exactly two streams:\n"
890 " video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n");
896 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
901 fsize = dv_assemble_frame((DVMuxContext *)s->priv_data, s->streams[pkt->stream_index],
902 pkt->data, pkt->size, &frame);
904 put_buffer(&s->pb, frame, fsize);
905 put_flush_packet(&s->pb);
911 * We might end up with some extra A/V data without matching counterpart.
912 * E.g. video data without enough audio to write the complete frame.
913 * Currently we simply drop the last frame. I don't know whether this
914 * is the best strategy of all
916 static int dv_write_trailer(struct AVFormatContext *s)
918 dv_delete_mux((DVMuxContext *)s->priv_data);
922 static AVInputFormat dv_iformat = {
925 sizeof(RawDVContext),
931 .extensions = "dv,dif",
934 static AVOutputFormat dv_oformat = {
939 sizeof(DVMuxContext),
949 av_register_input_format(&dv_iformat);
950 av_register_output_format(&dv_oformat);