DVCPRO50 support.
[libav.git] / libavformat / dv.c
1 /*
2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnik
4 *
5 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
6 * of DV technical info.
7 *
8 * Raw DV format
9 * Copyright (c) 2002 Fabrice Bellard.
10 *
11 * 50 Mbps (DVCPRO50) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
13 *
14 * This library is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU Lesser General Public
16 * License as published by the Free Software Foundation; either
17 * version 2 of the License, or (at your option) any later version.
18 *
19 * This library is distributed in the hope that it will be useful,
20 * but WITHOUT ANY WARRANTY; without even the implied warranty of
21 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 * Lesser General Public License for more details.
23 *
24 * You should have received a copy of the GNU Lesser General Public
25 * License along with this library; if not, write to the Free Software
26 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 */
28 #include <time.h>
29 #include "avformat.h"
30 #include "dvdata.h"
31 #include "dv.h"
32
33 struct DVDemuxContext {
34 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
35 AVFormatContext* fctx;
36 AVStream* vst;
37 AVStream* ast[2];
38 AVPacket audio_pkt[2];
39 uint8_t audio_buf[2][8192];
40 int ach;
41 int frames;
42 uint64_t abytes;
43 };
44
45 struct DVMuxContext {
46 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
47 int n_ast; /* Number of stereo audio streams (up to 2) */
48 AVStream *ast[2]; /* Stereo audio streams */
49 FifoBuffer audio_data[2]; /* Fifo for storing excessive amounts of PCM */
50 int frames; /* Number of a current frame */
51 time_t start_time; /* Start time of recording */
52 uint8_t aspect; /* Aspect ID 0 - 4:3, 7 - 16:9 */
53 int has_audio; /* frame under contruction has audio */
54 int has_video; /* frame under contruction has video */
55 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
56 };
57
58 enum dv_section_type {
59 dv_sect_header = 0x1f,
60 dv_sect_subcode = 0x3f,
61 dv_sect_vaux = 0x56,
62 dv_sect_audio = 0x76,
63 dv_sect_video = 0x96,
64 };
65
66 enum dv_pack_type {
67 dv_header525 = 0x3f, /* see dv_write_pack for important details on */
68 dv_header625 = 0xbf, /* these two packs */
69 dv_timecode = 0x13,
70 dv_audio_source = 0x50,
71 dv_audio_control = 0x51,
72 dv_audio_recdate = 0x52,
73 dv_audio_rectime = 0x53,
74 dv_video_source = 0x60,
75 dv_video_control = 0x61,
76 dv_video_recdate = 0x62,
77 dv_video_rectime = 0x63,
78 dv_unknown_pack = 0xff,
79 };
80
81
82
83 /*
84 * The reason why the following three big ugly looking tables are
85 * here is my lack of DV spec IEC 61834. The tables were basically
86 * constructed to make code that places packs in SSYB, VAUX and
87 * AAUX blocks very simple and table-driven. They conform to the
88 * SMPTE 314M and the output of my personal DV camcorder, neither
89 * of which is sufficient for a reliable DV stream producing. Thus
90 * while code is still in development I'll be gathering input from
91 * people with different DV equipment and modifying the tables to
92 * accommodate all the quirks. Later on, if possible, some of them
93 * will be folded into smaller tables and/or switch-if logic. For
94 * now, my only excuse is -- they don't eat up that much of a space.
95 */
96
97 static const int dv_ssyb_packs_dist[12][6] = {
98 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
99 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
100 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
101 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
102 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
103 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
104 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
105 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
106 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
107 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
108 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
109 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
110 };
111
112 static const int dv_vaux_packs_dist[12][15] = {
113 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
114 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
115 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
116 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
117 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
118 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
119 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
120 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
121 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
122 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
123 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
124 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
125 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
126 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
127 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
128 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
129 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
130 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
131 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
132 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
133 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
134 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
135 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
136 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
137 };
138
139 static const int dv_aaux_packs_dist[12][9] = {
140 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
141 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
142 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
143 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
144 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
145 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
146 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
147 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
148 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
149 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
150 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
151 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
152 };
153
154 static inline uint16_t dv_audio_12to16(uint16_t sample)
155 {
156 uint16_t shift, result;
157
158 sample = (sample < 0x800) ? sample : sample | 0xf000;
159 shift = (sample & 0xf00) >> 8;
160
161 if (shift < 0x2 || shift > 0xd) {
162 result = sample;
163 } else if (shift < 0x8) {
164 shift--;
165 result = (sample - (256 * shift)) << shift;
166 } else {
167 shift = 0xe - shift;
168 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
169 }
170
171 return result;
172 }
173
174 static int dv_audio_frame_size(const DVprofile* sys, int frame)
175 {
176 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
177 sizeof(sys->audio_samples_dist[0]))];
178 }
179
180 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
181 {
182 struct tm tc;
183 time_t ct;
184 int ltc_frame;
185
186 /* Its hard to tell what SMPTE requires w.r.t. APT, but Quicktime needs it.
187 * We set it based on pix_fmt value but it really should be per DV profile */
188 int apt = (c->sys->pix_fmt == PIX_FMT_YUV422P ? 1 : 0);
189
190 buf[0] = (uint8_t)pack_id;
191 switch (pack_id) {
192 case dv_header525: /* I can't imagine why these two weren't defined as real */
193 case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
194 buf[1] = 0xf8 | /* reserved -- always 1 */
195 (apt & 0x07); /* APT: Track application ID */
196 buf[2] = (0 << 7) | /* TF1: audio data is 0 - valid; 1 - invalid */
197 (0x0f << 3) | /* reserved -- always 1 */
198 (apt & 0x07); /* AP1: Audio application ID */
199 buf[3] = (0 << 7) | /* TF2: video data is 0 - valid; 1 - invalid */
200 (0x0f << 3) | /* reserved -- always 1 */
201 (apt & 0x07); /* AP2: Video application ID */
202 buf[4] = (0 << 7) | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */
203 (0x0f << 3) | /* reserved -- always 1 */
204 (apt & 0x07); /* AP3: Subcode application ID */
205 break;
206 case dv_timecode:
207 ct = (time_t)(c->frames / ((float)c->sys->frame_rate /
208 (float)c->sys->frame_rate_base));
209 brktimegm(ct, &tc);
210 /*
211 * LTC drop-frame frame counter drops two frames (0 and 1) every
212 * minute, unless it is exactly divisible by 10
213 */
214 ltc_frame = (c->frames + 2*ct/60 - 2*ct/600) % c->sys->ltc_divisor;
215 buf[1] = (0 << 7) | /* Color fame: 0 - unsync; 1 - sync mode */
216 (1 << 6) | /* Drop frame timecode: 0 - nondrop; 1 - drop */
217 ((ltc_frame / 10) << 4) | /* Tens of frames */
218 (ltc_frame % 10); /* Units of frames */
219 buf[2] = (1 << 7) | /* Biphase mark polarity correction: 0 - even; 1 - odd */
220 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
221 (tc.tm_sec % 10); /* Units of seconds */
222 buf[3] = (1 << 7) | /* Binary group flag BGF0 */
223 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
224 (tc.tm_min % 10); /* Units of minutes */
225 buf[4] = (1 << 7) | /* Binary group flag BGF2 */
226 (1 << 6) | /* Binary group flag BGF1 */
227 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
228 (tc.tm_hour % 10); /* Units of hours */
229 break;
230 case dv_audio_source: /* AAUX source pack */
231 buf[1] = (0 << 7) | /* locked mode */
232 (1 << 6) | /* reserved -- always 1 */
233 (dv_audio_frame_size(c->sys, c->frames) -
234 c->sys->audio_min_samples[0]);
235 /* # of samples */
236 buf[2] = (0 << 7) | /* multi-stereo */
237 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
238 (0 << 4) | /* pair bit: 0 -- one pair of channels */
239 0; /* audio mode */
240 buf[3] = (1 << 7) | /* res */
241 (1 << 6) | /* multi-language flag */
242 (c->sys->dsf << 5) | /* system: 60fields/50fields */
243 (apt << 1);/* definition: 0 -- 25Mbps, 2 -- 50Mbps */
244 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
245 (0 << 6) | /* emphasis time constant: 0 -- reserved */
246 (0 << 3) | /* frequency: 0 -- 48Khz, 1 -- 44,1Khz, 2 -- 32Khz */
247 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
248 break;
249 case dv_audio_control:
250 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
251 (1 << 4) | /* input source: 1 -- digital input */
252 (3 << 2) | /* compression: 3 -- no information */
253 0; /* misc. info/SMPTE emphasis off */
254 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
255 (1 << 6) | /* recording end point: 1 -- no */
256 (1 << 3) | /* recording mode: 1 -- original */
257 7;
258 buf[3] = (1 << 7) | /* direction: 1 -- forward */
259 0x20; /* speed */
260 buf[4] = (1 << 7) | /* reserved -- always 1 */
261 0x7f; /* genre category */
262 break;
263 case dv_audio_recdate:
264 case dv_video_recdate: /* VAUX recording date */
265 ct = c->start_time + (time_t)(c->frames /
266 ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
267 brktimegm(ct, &tc);
268 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
269 /* 0xff is very likely to be "unknown" */
270 buf[2] = (3 << 6) | /* reserved -- always 1 */
271 ((tc.tm_mday / 10) << 4) | /* Tens of day */
272 (tc.tm_mday % 10); /* Units of day */
273 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
274 ((tc.tm_mon / 10) << 4) | /* Tens of month */
275 (tc.tm_mon % 10); /* Units of month */
276 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
277 (tc.tm_year % 10); /* Units of year */
278 break;
279 case dv_audio_rectime: /* AAUX recording time */
280 case dv_video_rectime: /* VAUX recording time */
281 ct = c->start_time + (time_t)(c->frames /
282 ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
283 brktimegm(ct, &tc);
284 buf[1] = (3 << 6) | /* reserved -- always 1 */
285 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
286 buf[2] = (1 << 7) | /* reserved -- always 1 */
287 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
288 (tc.tm_sec % 10); /* Units of seconds */
289 buf[3] = (1 << 7) | /* reserved -- always 1 */
290 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
291 (tc.tm_min % 10); /* Units of minutes */
292 buf[4] = (3 << 6) | /* reserved -- always 1 */
293 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
294 (tc.tm_hour % 10); /* Units of hours */
295 break;
296 case dv_video_source:
297 buf[1] = 0xff; /* reserved -- always 1 */
298 buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
299 (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
300 (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
301 0xf; /* reserved -- always 1 */
302 buf[3] = (3 << 6) | /* reserved -- always 1 */
303 (c->sys->dsf << 5) | /* system: 60fields/50fields */
304 (apt << 2); /* signal type video compression */
305 buf[4] = 0xff; /* VISC: 0xff -- no information */
306 break;
307 case dv_video_control:
308 buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
309 0x3f; /* reserved -- always 1 */
310 buf[2] = 0xc8 | /* reserved -- always b11001xxx */
311 c->aspect;
312 buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
313 (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
314 (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
315 (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
316 0xc; /* reserved -- always b1100 */
317 buf[4] = 0xff; /* reserved -- always 1 */
318 break;
319 default:
320 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
321 }
322 return 5;
323 }
324
325 static inline int dv_write_dif_id(enum dv_section_type t, uint8_t chan_num, uint8_t seq_num,
326 uint8_t dif_num, uint8_t* buf)
327 {
328 buf[0] = (uint8_t)t; /* Section type */
329 buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
330 (chan_num << 3) | /* FSC: for 50Mb/s 0 - first channel; 1 - second */
331 7; /* reserved -- always 1 */
332 buf[2] = dif_num; /* DIF block number Video: 0-134, Audio: 0-8 */
333 return 3;
334 }
335
336 static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
337 {
338 if (syb_num == 0 || syb_num == 6) {
339 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
340 (0<<4) | /* AP3 (Subcode application ID) */
341 0x0f; /* reserved -- always 1 */
342 }
343 else if (syb_num == 11) {
344 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
345 0x7f; /* reserved -- always 1 */
346 }
347 else {
348 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
349 (0<<4) | /* APT (Track application ID) */
350 0x0f; /* reserved -- always 1 */
351 }
352 buf[1] = 0xf0 | /* reserved -- always 1 */
353 (syb_num & 0x0f); /* SSYB number 0 - 11 */
354 buf[2] = 0xff; /* reserved -- always 1 */
355 return 3;
356 }
357
358 static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
359 {
360 int chan, i, j, k;
361
362 for (chan = 0; chan < c->sys->n_difchan; chan++) {
363 for (i = 0; i < c->sys->difseg_size; i++) {
364 memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
365
366 /* DV header: 1DIF */
367 buf += dv_write_dif_id(dv_sect_header, chan, i, 0, buf);
368 buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
369 buf += 72; /* unused bytes */
370
371 /* DV subcode: 2DIFs */
372 for (j = 0; j < 2; j++) {
373 buf += dv_write_dif_id(dv_sect_subcode, chan, i, j, buf);
374 for (k = 0; k < 6; k++) {
375 buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
376 buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
377 }
378 buf += 29; /* unused bytes */
379 }
380
381 /* DV VAUX: 3DIFs */
382 for (j = 0; j < 3; j++) {
383 buf += dv_write_dif_id(dv_sect_vaux, chan, i, j, buf);
384 for (k = 0; k < 15 ; k++)
385 buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
386 buf += 2; /* unused bytes */
387 }
388
389 /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
390 for (j = 0; j < 135; j++) {
391 if (j%15 == 0) {
392 memset(buf, 0xff, 80);
393 buf += dv_write_dif_id(dv_sect_audio, chan, i, j/15, buf);
394 buf += 77; /* audio control & shuffled PCM audio */
395 }
396 buf += dv_write_dif_id(dv_sect_video, chan, i, j, buf);
397 buf += 77; /* 1 video macro block: 1 bytes control
398 4 * 14 bytes Y 8x8 data
399 10 bytes Cr 8x8 data
400 10 bytes Cb 8x8 data */
401 }
402 }
403 }
404 }
405
406 static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, int channel, uint8_t* frame_ptr)
407 {
408 int i, j, d, of, size;
409 size = 4 * dv_audio_frame_size(c->sys, c->frames);
410 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
411 for (i = 0; i < c->sys->difseg_size; i++) {
412 frame_ptr += 6 * 80; /* skip DIF segment header */
413 for (j = 0; j < 9; j++) {
414 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3]);
415 for (d = 8; d < 80; d+=2) {
416 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
417 if (of*2 >= size)
418 continue;
419
420 frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
421 frame_ptr[d+1] = pcm[of*2]; // that DV is a big endian PCM
422 }
423 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
424 }
425 }
426 }
427
428 static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
429 {
430 int chan, i, j;
431 int ptr = 0;
432
433 for (chan = 0; chan < c->sys->n_difchan; chan++) {
434 for (i = 0; i < c->sys->difseg_size; i++) {
435 ptr += 6 * 80; /* skip DIF segment header */
436 for (j = 0; j < 135; j++) {
437 if (j%15 == 0)
438 ptr += 80; /* skip Audio DIF */
439 ptr += 3;
440 memcpy(frame_ptr + ptr, video_data + ptr, 77);
441 ptr += 77;
442 }
443 }
444 }
445 }
446
447 /*
448 * This is the dumbest implementation of all -- it simply looks at
449 * a fixed offset and if pack isn't there -- fails. We might want
450 * to have a fallback mechanism for complete search of missing packs.
451 */
452 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
453 {
454 int offs;
455
456 switch (t) {
457 case dv_audio_source:
458 offs = (80*6 + 80*16*3 + 3);
459 break;
460 case dv_audio_control:
461 offs = (80*6 + 80*16*4 + 3);
462 break;
463 case dv_video_control:
464 offs = (80*5 + 48 + 5);
465 break;
466 default:
467 return NULL;
468 }
469
470 return (frame[offs] == t ? &frame[offs] : NULL);
471 }
472
473 /*
474 * There's a couple of assumptions being made here:
475 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
476 * We can pass them upwards when ffmpeg will be ready to deal with them.
477 * 2. We don't do software emphasis.
478 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
479 * are converted into 16bit linear ones.
480 */
481 static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2,
482 const DVprofile *sys)
483 {
484 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
485 uint16_t lc, rc;
486 const uint8_t* as_pack;
487
488 as_pack = dv_extract_pack(frame, dv_audio_source);
489 if (!as_pack) /* No audio ? */
490 return 0;
491
492 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
493 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
494 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
495
496 if (quant > 1)
497 return -1; /* Unsupported quantization */
498
499 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
500 half_ch = sys->difseg_size/2;
501
502 /* for each DIF channel */
503 for (chan = 0; chan < sys->n_difchan; chan++) {
504 /* for each DIF segment */
505 for (i = 0; i < sys->difseg_size; i++) {
506 frame += 6 * 80; /* skip DIF segment header */
507 if (quant == 1 && i == half_ch) {
508 /* next stereo channel (12bit mode only) */
509 if (!pcm2)
510 break;
511 else
512 pcm = pcm2;
513 }
514
515 /* for each AV sequence */
516 for (j = 0; j < 9; j++) {
517 for (d = 8; d < 80; d += 2) {
518 if (quant == 0) { /* 16bit quantization */
519 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
520 if (of*2 >= size)
521 continue;
522
523 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
524 pcm[of*2+1] = frame[d]; // that DV is a big endian PCM
525 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
526 pcm[of*2+1] = 0;
527 } else { /* 12bit quantization */
528 lc = ((uint16_t)frame[d] << 4) |
529 ((uint16_t)frame[d+2] >> 4);
530 rc = ((uint16_t)frame[d+1] << 4) |
531 ((uint16_t)frame[d+2] & 0x0f);
532 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
533 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
534
535 of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
536 if (of*2 >= size)
537 continue;
538
539 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
540 pcm[of*2+1] = lc >> 8; // that DV is a big endian PCM
541 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
542 (d - 8)/3 * sys->audio_stride;
543 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
544 pcm[of*2+1] = rc >> 8; // that DV is a big endian PCM
545 ++d;
546 }
547 }
548
549 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
550 }
551 }
552
553 /* next stereo channel (50Mbps only) */
554 if(!pcm2)
555 break;
556 pcm = pcm2;
557 }
558
559 return size;
560 }
561
562 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
563 {
564 const uint8_t* as_pack;
565 int freq, stype, smpls, quant, i, ach;
566
567 as_pack = dv_extract_pack(frame, dv_audio_source);
568 if (!as_pack || !c->sys) { /* No audio ? */
569 c->ach = 0;
570 return 0;
571 }
572
573 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
574 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
575 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH */
576 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
577
578 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
579 ach = (stype == 2 || (quant && (freq == 2))) ? 2 : 1;
580
581 /* Dynamic handling of the audio streams in DV */
582 for (i=0; i<ach; i++) {
583 if (!c->ast[i]) {
584 c->ast[i] = av_new_stream(c->fctx, 0);
585 if (!c->ast[i])
586 break;
587 av_set_pts_info(c->ast[i], 64, 1, 30000);
588 c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
589 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
590
591 av_init_packet(&c->audio_pkt[i]);
592 c->audio_pkt[i].size = 0;
593 c->audio_pkt[i].data = c->audio_buf[i];
594 c->audio_pkt[i].stream_index = c->ast[i]->index;
595 c->audio_pkt[i].flags |= PKT_FLAG_KEY;
596 }
597 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
598 c->ast[i]->codec->channels = 2;
599 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
600 c->ast[i]->start_time = 0;
601 }
602 c->ach = i;
603
604 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
605 }
606
607 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
608 {
609 const uint8_t* vsc_pack;
610 AVCodecContext* avctx;
611 int apt, is16_9;
612 int size = 0;
613
614 if (c->sys) {
615 avctx = c->vst->codec;
616
617 av_set_pts_info(c->vst, 64, c->sys->frame_rate_base, c->sys->frame_rate);
618 avctx->time_base= (AVRational){c->sys->frame_rate_base, c->sys->frame_rate};
619 if(!avctx->width){
620 avctx->width = c->sys->width;
621 avctx->height = c->sys->height;
622 }
623 avctx->pix_fmt = c->sys->pix_fmt;
624
625 /* finding out SAR is a little bit messy */
626 vsc_pack = dv_extract_pack(frame, dv_video_control);
627 apt = frame[4] & 0x07;
628 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
629 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
630 avctx->sample_aspect_ratio = c->sys->sar[is16_9];
631 avctx->bit_rate = av_rescale(c->sys->frame_size * 8,
632 c->sys->frame_rate,
633 c->sys->frame_rate_base);
634 size = c->sys->frame_size;
635 }
636 return size;
637 }
638
639 /*
640 * The following 6 functions constitute our interface to the world
641 */
642
643 int dv_assemble_frame(DVMuxContext *c, AVStream* st,
644 const uint8_t* data, int data_size, uint8_t** frame)
645 {
646 uint8_t pcm[8192];
647 int i;
648
649 *frame = &c->frame_buf[0];
650 if (c->has_audio && c->has_video &&
651 (c->has_audio == -1 || c->has_audio == c->n_ast)) {
652 /* must be a stale frame */
653 dv_format_frame(c, *frame);
654 c->frames++;
655 if (c->has_audio > 0)
656 c->has_audio = 0;
657 c->has_video = 0;
658 }
659
660 if (st->codec->codec_type == CODEC_TYPE_VIDEO) {
661 /* FIXME: we have to have more sensible approach than this one */
662 if (c->has_video)
663 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
664
665 dv_inject_video(c, data, *frame);
666 c->has_video = 1;
667 data_size = 0;
668 if (c->has_audio < 0)
669 goto out;
670 }
671
672 for (i = 0; i < c->n_ast; i++) {
673 int reqasize, fsize;
674 if (st != c->ast[i])
675 continue;
676 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
677 fsize = fifo_size(&c->audio_data[i], c->audio_data[i].rptr);
678 if (st->codec->codec_type == CODEC_TYPE_AUDIO ||
679 (c->has_video && fsize >= reqasize)) {
680 if (fsize + data_size >= reqasize && (c->has_audio < c->n_ast)) {
681 if (fsize >= reqasize) {
682 fifo_read(&c->audio_data[i], &pcm[0], reqasize, &c->audio_data[i].rptr);
683 } else {
684 fifo_read(&c->audio_data[i], &pcm[0], fsize, &c->audio_data[i].rptr);
685 memcpy(&pcm[fsize], &data[0], reqasize - fsize);
686 data += reqasize - fsize;
687 data_size -= reqasize - fsize;
688 }
689 dv_inject_audio(c, &pcm[0], i, *frame);
690 c->has_audio += 1;
691 }
692 /* FIXME: we have to have more sensible approach than this one */
693 if (fifo_size(&c->audio_data[i], c->audio_data[i].rptr) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
694 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
695 fifo_write(&c->audio_data[i], (uint8_t *)data, data_size, &c->audio_data[i].wptr);
696 }
697 }
698 out:
699 return ((c->has_audio == -1 || c->has_audio == c->n_ast) && c->has_video) ? c->sys->frame_size : 0;
700 }
701
702 DVMuxContext* dv_init_mux(AVFormatContext* s)
703 {
704 DVMuxContext *c;
705 AVStream *vst = NULL;
706 int i;
707
708 /* we support at most 1 video and 2 audio streams */
709 if (s->nb_streams > 3)
710 return NULL;
711
712 c = av_mallocz(sizeof(DVMuxContext));
713 if (!c)
714 return NULL;
715
716 c->n_ast = 0;
717 c->ast[0] = c->ast[1] = NULL;
718
719 /* We have to sort out where audio and where video stream is */
720 for (i=0; i<s->nb_streams; i++) {
721 switch (s->streams[i]->codec->codec_type) {
722 case CODEC_TYPE_VIDEO:
723 vst = s->streams[i];
724 break;
725 case CODEC_TYPE_AUDIO:
726 c->ast[c->n_ast++] = s->streams[i];
727 break;
728 default:
729 goto bail_out;
730 }
731 }
732
733 /* Some checks -- DV format is very picky about its incoming streams */
734 if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
735 goto bail_out;
736 for (i=0; i<c->n_ast; i++) {
737 if (c->ast[i] && (c->ast[i]->codec->codec_id != CODEC_ID_PCM_S16LE ||
738 c->ast[i]->codec->sample_rate != 48000 ||
739 c->ast[i]->codec->channels != 2))
740 goto bail_out;
741 }
742 c->sys = dv_codec_profile(vst->codec);
743 if (!c->sys)
744 goto bail_out;
745
746 if((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
747 /* only 1 stereo pair is allowed in 25Mbps mode */
748 goto bail_out;
749 }
750
751 /* Ok, everything seems to be in working order */
752 c->frames = 0;
753 c->has_audio = c->n_ast ? 0 : -1;
754 c->has_video = 0;
755 c->start_time = (time_t)s->timestamp;
756 c->aspect = 0; /* 4:3 is the default */
757 if ((int)(av_q2d(vst->codec->sample_aspect_ratio) * vst->codec->width / vst->codec->height * 10) == 17) /* 16:9 */
758 c->aspect = 0x07;
759
760 for (i=0; i<c->n_ast; i++) {
761 if (c->ast[i] && fifo_init(&c->audio_data[i], 100*AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0) {
762 while (i>0) {
763 i--;
764 fifo_free(&c->audio_data[i]);
765 }
766 goto bail_out;
767 }
768 }
769
770 dv_format_frame(c, &c->frame_buf[0]);
771
772 return c;
773
774 bail_out:
775 av_free(c);
776 return NULL;
777 }
778
779 void dv_delete_mux(DVMuxContext *c)
780 {
781 int i;
782 for (i=0; i < c->n_ast; i++)
783 fifo_free(&c->audio_data[i]);
784 }
785
786 DVDemuxContext* dv_init_demux(AVFormatContext *s)
787 {
788 DVDemuxContext *c;
789
790 c = av_mallocz(sizeof(DVDemuxContext));
791 if (!c)
792 return NULL;
793
794 c->vst = av_new_stream(s, 0);
795 if (!c->vst) {
796 av_free(c);
797 return NULL;
798 }
799
800 c->sys = NULL;
801 c->fctx = s;
802 c->ast[0] = c->ast[1] = NULL;
803 c->ach = 0;
804 c->frames = 0;
805 c->abytes = 0;
806
807 c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
808 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
809 c->vst->codec->bit_rate = 25000000;
810 c->vst->start_time = 0;
811
812 return c;
813 }
814
815 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
816 {
817 int size = -1;
818 int i;
819
820 for (i=0; i<c->ach; i++) {
821 if (c->ast[i] && c->audio_pkt[i].size) {
822 *pkt = c->audio_pkt[i];
823 c->audio_pkt[i].size = 0;
824 size = pkt->size;
825 break;
826 }
827 }
828
829 return size;
830 }
831
832 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
833 uint8_t* buf, int buf_size)
834 {
835 int size, i;
836
837 if (buf_size < 4 || buf_size < c->sys->frame_size)
838 return -1; /* Broken frame, or not enough data */
839
840 /* Queueing audio packet */
841 /* FIXME: in case of no audio/bad audio we have to do something */
842 size = dv_extract_audio_info(c, buf);
843 for (i=0; i<c->ach; i++) {
844 c->audio_pkt[i].size = size;
845 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
846 }
847 dv_extract_audio(buf, c->audio_buf[0], c->audio_buf[1], c->sys);
848 c->abytes += size;
849
850 /* Now it's time to return video packet */
851 size = dv_extract_video_info(c, buf);
852 av_init_packet(pkt);
853 pkt->data = buf;
854 pkt->size = size;
855 pkt->flags |= PKT_FLAG_KEY;
856 pkt->stream_index = c->vst->id;
857 pkt->pts = c->frames;
858
859 c->frames++;
860
861 return size;
862 }
863
864 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
865 int64_t timestamp, int flags)
866 {
867 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
868 const DVprofile* sys = dv_codec_profile(c->vst->codec);
869 int64_t offset;
870 int64_t size = url_fsize(&s->pb);
871 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
872
873 offset = sys->frame_size * timestamp;
874
875 if (offset > max_offset) offset = max_offset;
876 else if (offset < 0) offset = 0;
877
878 return offset;
879 }
880
881 void dv_flush_audio_packets(DVDemuxContext *c)
882 {
883 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
884 }
885
886 /************************************************************
887 * Implementation of the easiest DV storage of all -- raw DV.
888 ************************************************************/
889
890 typedef struct RawDVContext {
891 DVDemuxContext* dv_demux;
892 uint8_t buf[DV_MAX_FRAME_SIZE];
893 } RawDVContext;
894
895 static int dv_read_header(AVFormatContext *s,
896 AVFormatParameters *ap)
897 {
898 RawDVContext *c = s->priv_data;
899
900 c->dv_demux = dv_init_demux(s);
901 if (!c->dv_demux)
902 return -1;
903
904 if (get_buffer(&s->pb, c->buf, DV_PROFILE_BYTES) <= 0 ||
905 url_fseek(&s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
906 return AVERROR_IO;
907
908 c->dv_demux->sys = dv_frame_profile(c->buf);
909 s->bit_rate = av_rescale(c->dv_demux->sys->frame_size * 8,
910 c->dv_demux->sys->frame_rate,
911 c->dv_demux->sys->frame_rate_base);
912
913 return 0;
914 }
915
916
917 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
918 {
919 int size;
920 RawDVContext *c = s->priv_data;
921
922 size = dv_get_packet(c->dv_demux, pkt);
923
924 if (size < 0) {
925 size = c->dv_demux->sys->frame_size;
926 if (get_buffer(&s->pb, c->buf, size) <= 0)
927 return AVERROR_IO;
928
929 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
930 }
931
932 return size;
933 }
934
935 static int dv_read_seek(AVFormatContext *s, int stream_index,
936 int64_t timestamp, int flags)
937 {
938 RawDVContext *r = s->priv_data;
939 DVDemuxContext *c = r->dv_demux;
940 int64_t offset= dv_frame_offset(s, c, timestamp, flags);
941
942 c->frames= offset / c->sys->frame_size;
943 if (c->ach)
944 c->abytes= av_rescale(c->frames,
945 c->ast[0]->codec->bit_rate * (int64_t)c->sys->frame_rate_base,
946 8*c->sys->frame_rate);
947
948 dv_flush_audio_packets(c);
949 return url_fseek(&s->pb, offset, SEEK_SET);
950 }
951
952 static int dv_read_close(AVFormatContext *s)
953 {
954 RawDVContext *c = s->priv_data;
955 av_free(c->dv_demux);
956 return 0;
957 }
958
959 #ifdef CONFIG_MUXERS
960 static int dv_write_header(AVFormatContext *s)
961 {
962 s->priv_data = dv_init_mux(s);
963 if (!s->priv_data) {
964 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
965 "Make sure that you supply exactly two streams:\n"
966 " video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n"
967 " (50Mbps allows an optional second audio stream)\n");
968 return -1;
969 }
970 return 0;
971 }
972
973 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
974 {
975 uint8_t* frame;
976 int fsize;
977
978 fsize = dv_assemble_frame((DVMuxContext *)s->priv_data, s->streams[pkt->stream_index],
979 pkt->data, pkt->size, &frame);
980 if (fsize > 0) {
981 put_buffer(&s->pb, frame, fsize);
982 put_flush_packet(&s->pb);
983 }
984 return 0;
985 }
986
987 /*
988 * We might end up with some extra A/V data without matching counterpart.
989 * E.g. video data without enough audio to write the complete frame.
990 * Currently we simply drop the last frame. I don't know whether this
991 * is the best strategy of all
992 */
993 static int dv_write_trailer(struct AVFormatContext *s)
994 {
995 dv_delete_mux((DVMuxContext *)s->priv_data);
996 return 0;
997 }
998 #endif /* CONFIG_MUXERS */
999
1000 static AVInputFormat dv_iformat = {
1001 "dv",
1002 "DV video format",
1003 sizeof(RawDVContext),
1004 NULL,
1005 dv_read_header,
1006 dv_read_packet,
1007 dv_read_close,
1008 dv_read_seek,
1009 .extensions = "dv,dif",
1010 };
1011
1012 #ifdef CONFIG_MUXERS
1013 static AVOutputFormat dv_oformat = {
1014 "dv",
1015 "DV video format",
1016 NULL,
1017 "dv",
1018 sizeof(DVMuxContext),
1019 CODEC_ID_PCM_S16LE,
1020 CODEC_ID_DVVIDEO,
1021 dv_write_header,
1022 dv_write_packet,
1023 dv_write_trailer,
1024 };
1025 #endif
1026
1027 int ff_dv_init(void)
1028 {
1029 av_register_input_format(&dv_iformat);
1030 #ifdef CONFIG_MUXERS
1031 av_register_output_format(&dv_oformat);
1032 #endif
1033 return 0;
1034 }