Browse code

* Phase 1 for DV encoding support. Muxing/demuxing of DV streams is now possible. For example you can do:

ffmpeg -i i.dv -i audio_track.mp3 -map 0.0 -map 1.0 \
-vcodec copy -acodec pcm_s16le out.dv

* Preparations for getting rid of DVAUDIO codec, DV stream really
contains PCM audio, so there's no codec needed if we have a
decent demuxer.

* Providing entry points for dv1394 write support.

Originally committed as revision 2174 to svn://svn.ffmpeg.org/ffmpeg/trunk

Roman Shaposhnik authored on 2003/08/29 04:53:47
Showing 4 changed files
... ...
@@ -13,7 +13,7 @@ PPOBJS=
13 13
 
14 14
 # mux and demuxes
15 15
 OBJS+=mpeg.o mpegts.o mpegtsenc.o ffm.o crc.o img.o raw.o rm.o \
16
-      avienc.o avidec.o wav.o swf.o au.o gif.o mov.o mpjpeg.o dv.o \
16
+      avienc.o avidec.o wav.o swf.o au.o gif.o mov.o mpjpeg.o dvcore.o dv.o \
17 17
       yuv4mpeg.o 4xm.o flvenc.o flvdec.o movenc.o
18 18
 
19 19
 ifeq ($(CONFIG_RISKY),yes)
... ...
@@ -17,13 +17,11 @@
17 17
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
18 18
  */
19 19
 #include "avformat.h"
20
-
21
-#define NTSC_FRAME_SIZE 120000
22
-#define PAL_FRAME_SIZE  144000
20
+#include "dvcore.h"
23 21
 
24 22
 typedef struct DVDemuxContext {
25 23
     int       is_audio;
26
-    uint8_t   buf[PAL_FRAME_SIZE];
24
+    uint8_t   buf[144000];    
27 25
     int       size;
28 26
 } DVDemuxContext;
29 27
 
... ...
@@ -39,7 +37,7 @@ static int dv_read_header(AVFormatContext *s,
39 39
         return AVERROR_NOMEM;
40 40
     vst->codec.codec_type = CODEC_TYPE_VIDEO;
41 41
     vst->codec.codec_id = CODEC_ID_DVVIDEO;
42
-
42
+    vst->codec.bit_rate = 25000000;
43 43
 
44 44
     ast = av_new_stream(s, 1);
45 45
     if (!ast)
... ...
@@ -60,18 +58,14 @@ static void __destruct_pkt(struct AVPacket *pkt)
60 60
 
61 61
 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
62 62
 {
63
-    int ret, dsf;
63
+    int ret;
64 64
     DVDemuxContext *c = s->priv_data;
65 65
     
66 66
     if (!c->is_audio) {
67 67
         ret = get_buffer(&s->pb, c->buf, 4);
68 68
         if (ret <= 0) 
69 69
             return -EIO;
70
-        dsf = c->buf[3] & 0x80;
71
-        if (!dsf)
72
-            c->size = NTSC_FRAME_SIZE;
73
-        else
74
-            c->size = PAL_FRAME_SIZE;
70
+	c->size = dv_frame_profile(&c->buf[0])->frame_size;
75 71
 	
76 72
 	ret = get_buffer(&s->pb, c->buf + 4, c->size - 4);
77 73
 	if (ret <= 0)
... ...
@@ -94,20 +88,16 @@ static int dv_read_close(AVFormatContext *s)
94 94
     return 0;
95 95
 }
96 96
 
97
-static AVInputFormat dv_iformat = {
98
-    "dv",
99
-    "DV video format",
100
-    sizeof(DVDemuxContext),
101
-    NULL,
102
-    dv_read_header,
103
-    dv_read_packet,
104
-    dv_read_close,
105
-    .extensions = "dv",
106
-};
107
-
108
-
109 97
 int dv_write_header(struct AVFormatContext *s)
110 98
 {
99
+    DVMuxContext *c = s->priv_data;
100
+    
101
+    if (s->nb_streams != 2 || dv_core_init(c, s->streams) != 0) {
102
+        fprintf(stderr, "Can't initialize DV format!\n"
103
+		    "Make sure that you supply exactly two streams:\n"
104
+		    "     video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n");
105
+	return -1;
106
+    }
111 107
     return 0;
112 108
 }
113 109
 
... ...
@@ -115,24 +105,52 @@ int dv_write_packet(struct AVFormatContext *s,
115 115
                      int stream_index,
116 116
                      unsigned char *buf, int size, int force_pts)
117 117
 {
118
-    put_buffer(&s->pb, buf, size);
119
-    put_flush_packet(&s->pb);
118
+    DVMuxContext *c = s->priv_data;
119
+   
120
+    if (stream_index == c->vst)
121
+        dv_assemble_frame(c, buf, NULL, 0);
122
+    else
123
+        dv_assemble_frame(c, NULL, buf, size);
124
+   
125
+    if (c->has_audio && c->has_video) {
126
+        put_buffer(&s->pb, &c->frame_buf[0], c->sys->frame_size);
127
+        put_flush_packet(&s->pb);
128
+    }
129
+    
120 130
     return 0;
121 131
 }
122 132
 
133
+/* 
134
+ * We might end up with some extra A/V data without matching counterpart.
135
+ * E.g. video data without enough audio to write the complete frame.
136
+ * Currently we simply drop the last frame. I don't know whether this 
137
+ * is the best strategy of all
138
+ */
123 139
 int dv_write_trailer(struct AVFormatContext *s)
124 140
 {
141
+    dv_core_delete((DVMuxContext *)s->priv_data);
125 142
     return 0;
126 143
 }
127 144
 
145
+static AVInputFormat dv_iformat = {
146
+    "dv",
147
+    "DV video format",
148
+    sizeof(DVDemuxContext),
149
+    NULL,
150
+    dv_read_header,
151
+    dv_read_packet,
152
+    dv_read_close,
153
+    .extensions = "dv",
154
+};
155
+
128 156
 AVOutputFormat dv_oformat = {
129 157
     "dv",
130 158
     "DV video format",
131 159
     NULL,
132 160
     "dv",
133
-    0,
161
+    sizeof(DVMuxContext),
162
+    CODEC_ID_PCM_S16LE,
134 163
     CODEC_ID_DVVIDEO,
135
-    CODEC_ID_DVAUDIO,
136 164
     dv_write_header,
137 165
     dv_write_packet,
138 166
     dv_write_trailer,
139 167
new file mode 100644
... ...
@@ -0,0 +1,668 @@
0
+/* 
1
+ * DV format muxer/demuxer
2
+ * Copyright (c) 2003 Roman Shaposhnik
3
+ *
4
+ * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
5
+ * of DV technical info, and SMPTE 314M specification.
6
+ *
7
+ * This library is free software; you can redistribute it and/or
8
+ * modify it under the terms of the GNU Lesser General Public
9
+ * License as published by the Free Software Foundation; either
10
+ * version 2 of the License, or (at your option) any later version.
11
+ *
12
+ * This library is distributed in the hope that it will be useful,
13
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
+ * Lesser General Public License for more details.
16
+ *
17
+ * You should have received a copy of the GNU Lesser General Public
18
+ * License along with this library; if not, write to the Free Software
19
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
20
+ */
21
+#include "avformat.h"
22
+#include "dvcore.h"
23
+
24
+/*
25
+ * The reason why the following three big ugly looking tables are
26
+ * here is my lack of DV spec IEC 61834. The tables were basically 
27
+ * constructed to make code that places packs in SSYB, VAUX and 
28
+ * AAUX blocks very simple and table-driven. They conform to the
29
+ * SMPTE 314M and the output of my personal DV camcorder, neither
30
+ * of which is sufficient for a reliable DV stream producing. Thus
31
+ * while code is still in development I'll be gathering input from
32
+ * people with different DV equipment and modifying the tables to
33
+ * accommodate all the quirks. Later on, if possible, some of them
34
+ * will be folded into smaller tables and/or switch-if logic. For 
35
+ * now, my only excuse is -- they don't eat up that much of a space.
36
+ */
37
+
38
+static const int dv_ssyb_packs_dist[12][6] = {
39
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
40
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
41
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
42
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
43
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
44
+    { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
45
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
46
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
47
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
48
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
49
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
50
+    { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
51
+};
52
+
53
+static const int dv_vaux_packs_dist[12][15] = {
54
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
55
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
56
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
57
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
58
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
59
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
60
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
61
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
62
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
63
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
64
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
65
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
66
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
67
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
68
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
69
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
70
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
71
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
72
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
73
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
74
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
75
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
76
+    { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff, 
77
+      0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
78
+};
79
+
80
+static const int dv_aaux_packs_dist[12][9] = {
81
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
82
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
83
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
84
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
85
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
86
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
87
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
88
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
89
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
90
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
91
+    { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
92
+    { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
93
+};
94
+
95
+enum dv_section_type {
96
+     dv_sect_header  = 0x1f,
97
+     dv_sect_subcode = 0x3f,
98
+     dv_sect_vaux    = 0x56,
99
+     dv_sect_audio   = 0x76,
100
+     dv_sect_video   = 0x96,
101
+};
102
+
103
+enum dv_pack_type {
104
+     dv_header525     = 0x3f, /* see dv_write_pack for important details on */ 
105
+     dv_header625     = 0xbf, /* these two packs */
106
+     dv_timecode      = 0x13,
107
+     dv_audio_source  = 0x50,
108
+     dv_audio_control = 0x51,
109
+     dv_audio_recdate = 0x52,
110
+     dv_audio_rectime = 0x53,
111
+     dv_video_source  = 0x60,
112
+     dv_video_control = 0x61,
113
+     dv_viedo_recdate = 0x62,
114
+     dv_video_rectime = 0x63,
115
+     dv_unknown_pack  = 0xff,
116
+};
117
+
118
+static const uint16_t dv_audio_shuffle525[10][9] = {
119
+  {  0, 30, 60, 20, 50, 80, 10, 40, 70 }, /* 1st channel */
120
+  {  6, 36, 66, 26, 56, 86, 16, 46, 76 },
121
+  { 12, 42, 72,  2, 32, 62, 22, 52, 82 },
122
+  { 18, 48, 78,  8, 38, 68, 28, 58, 88 },
123
+  { 24, 54, 84, 14, 44, 74,  4, 34, 64 },
124
+  
125
+  {  1, 31, 61, 21, 51, 81, 11, 41, 71 }, /* 2nd channel */
126
+  {  7, 37, 67, 27, 57, 87, 17, 47, 77 },
127
+  { 13, 43, 73,  3, 33, 63, 23, 53, 83 },
128
+  { 19, 49, 79,  9, 39, 69, 29, 59, 89 },
129
+  { 25, 55, 85, 15, 45, 75,  5, 35, 65 },
130
+};
131
+
132
+static const uint16_t dv_audio_shuffle625[12][9] = {
133
+  {   0,  36,  72,  26,  62,  98,  16,  52,  88}, /* 1st channel */
134
+  {   6,  42,  78,  32,  68, 104,  22,  58,  94},
135
+  {  12,  48,  84,   2,  38,  74,  28,  64, 100},
136
+  {  18,  54,  90,   8,  44,  80,  34,  70, 106},
137
+  {  24,  60,  96,  14,  50,  86,   4,  40,  76},  
138
+  {  30,  66, 102,  20,  56,  92,  10,  46,  82},
139
+	
140
+  {   1,  37,  73,  27,  63,  99,  17,  53,  89}, /* 2nd channel */
141
+  {   7,  43,  79,  33,  69, 105,  23,  59,  95},
142
+  {  13,  49,  85,   3,  39,  75,  29,  65, 101},
143
+  {  19,  55,  91,   9,  45,  81,  35,  71, 107},
144
+  {  25,  61,  97,  15,  51,  87,   5,  41,  77},  
145
+  {  31,  67, 103,  21,  57,  93,  11,  47,  83},
146
+};
147
+
148
+static const int dv_audio_frequency[3] = {
149
+    48000, 44100, 32000,
150
+};
151
+    
152
+const DVprofile dv_profiles[2] = {
153
+    { .dsf = 0,
154
+      .frame_size = 120000,        /* 525/60 system (NTSC) */
155
+      .difseg_size = 10,
156
+      .frame_rate = 30000,
157
+      .ltc_divisor = 30,
158
+      .frame_rate_base = 1001,
159
+      .height = 480,
160
+/*    .video_place = dv_place_411, */
161
+      .audio_stride = 90,
162
+      .audio_min_samples = { 1580, 1452, 1053 }, /* for 48, 44.1 and 32Khz */
163
+      .audio_samples_dist = { 1602, 1601, 1602, 1601, 1602 },
164
+      .audio_shuffle = dv_audio_shuffle525,
165
+    }, 
166
+    { .dsf = 1,
167
+      .frame_size = 144000,        /* 625/50 system (PAL) */
168
+      .difseg_size = 12,
169
+      .frame_rate = 25,
170
+      .frame_rate_base = 1,
171
+      .ltc_divisor = 25,
172
+      .height = 576,
173
+/*    .video_place = dv_place_420, */
174
+      .audio_stride = 108,
175
+      .audio_min_samples = { 1896, 1742, 1264 }, /* for 48, 44.1 and 32Khz */
176
+      .audio_samples_dist = { 1920, 1920, 1920, 1920, 1920 },
177
+      .audio_shuffle = dv_audio_shuffle525,
178
+     }
179
+};
180
+
181
+static inline uint16_t dv_audio_12to16(uint16_t sample)
182
+{
183
+    uint16_t shift, result;
184
+    
185
+    sample = (sample < 0x800) ? sample : sample | 0xf000;
186
+    shift = (sample & 0xf00) >> 8;
187
+
188
+    if (shift < 0x2 || shift > 0xd) {
189
+	result = sample;
190
+    } else if (shift < 0x8) {
191
+        shift--;
192
+	result = (sample - (256 * shift)) << shift;
193
+    } else {
194
+	shift = 0xe - shift;
195
+	result = ((sample + ((256 * shift) + 1)) << shift) - 1;
196
+    }
197
+
198
+    return result;
199
+}
200
+
201
+static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
202
+{
203
+    struct tm tc;
204
+    time_t ct;
205
+    int ltc_frame;
206
+
207
+    buf[0] = (uint8_t)pack_id;
208
+    switch (pack_id) {
209
+    case dv_header525: /* I can't imagine why these two weren't defined as real */
210
+    case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
211
+          buf[1] = 0xf8 |               /* reserved -- always 1 */
212
+	           (0 & 0x07);          /* APT: Track application ID */
213
+          buf[2] = (0 << 7)    | /* TF1: audio data is 0 - valid; 1 - invalid */
214
+	           (0x0f << 3) | /* reserved -- always 1 */
215
+		   (0 & 0x07);   /* AP1: Audio application ID */
216
+          buf[3] = (0 << 7)    | /* TF2: video data is 0 - valid; 1 - invalid */  
217
+	           (0x0f << 3) | /* reserved -- always 1 */
218
+		   (0 & 0x07);   /* AP2: Video application ID */
219
+          buf[4] = (0 << 7)    | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */ 
220
+	           (0x0f << 3) | /* reserved -- always 1 */
221
+		   (0 & 0x07);   /* AP3: Subcode application ID */
222
+	  break;
223
+    case dv_timecode:
224
+          ct = (time_t)(c->frames / ((float)c->sys->frame_rate / 
225
+                                     (float)c->sys->frame_rate_base));
226
+          localtime_r(&ct, &tc);
227
+          /* 
228
+           * LTC drop-frame frame counter drops two frames (0 and 1) every 
229
+           * minute, unless it is exactly divisible by 10
230
+           */
231
+          ltc_frame = (c->frames + 2*ct/60 - 2*ct/600) % c->sys->ltc_divisor;
232
+	  buf[1] = (0 << 7) | /* Color fame: 0 - unsync; 1 - sync mode */
233
+		   (1 << 6) | /* Drop frame timecode: 0 - nondrop; 1 - drop */
234
+		   ((ltc_frame / 10) << 4) | /* Tens of frames */
235
+		   (ltc_frame % 10);         /* Units of frames */
236
+	  buf[2] = (1 << 7) | /* Biphase mark polarity correction: 0 - even; 1 - odd */
237
+		   ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
238
+		   (tc.tm_sec % 10);         /* Units of seconds */
239
+	  buf[3] = (1 << 7) | /* Binary group flag BGF0 */
240
+		   ((tc.tm_min / 10) << 4) | /* Tens of minutes */
241
+		   (tc.tm_min % 10);         /* Units of minutes */
242
+	  buf[4] = (1 << 7) | /* Binary group flag BGF2 */
243
+		   (1 << 6) | /* Binary group flag BGF1 */
244
+	           ((tc.tm_hour / 10) << 4) | /* Tens of hours */
245
+		   (tc.tm_hour % 10);         /* Units of hours */
246
+          break;
247
+    case dv_audio_source:  /* AAUX source pack */
248
+          buf[1] = (0 << 7) | /* locked mode       */
249
+                   (1 << 6) | /* reserved -- always 1 */
250
+	           (dv_audio_frame_size(c->sys, c->frames) -
251
+		    c->sys->audio_min_samples[0]);
252
+	                      /* # of samples      */
253
+          buf[2] = (0 << 7) | /* multi-stereo      */
254
+                   (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
255
+                   (0 << 4) | /* pair bit: 0 -- one pair of channels */
256
+	            0;        /* audio mode        */
257
+          buf[3] = (1 << 7) | /* res               */
258
+                   (1 << 6) | /* multi-language flag */
259
+	           (c->sys->dsf << 5) | /*  system: 60fields/50fields */
260
+	            0;        /* definition: 0 -- SD (525/625) */
261
+          buf[4] = (1 << 7) | /* emphasis: 1 -- off */
262
+                   (0 << 6) | /* emphasis time constant: 0 -- reserved */
263
+	           (0 << 3) | /* frequency: 0 -- 48Khz, 1 -- 44,1Khz, 2 -- 32Khz */
264
+                    0;        /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */			    
265
+          break;
266
+    case dv_audio_control:
267
+          buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
268
+                   (1 << 4) | /* input source: 1 -- digital input */
269
+	           (3 << 2) | /* compression: 3 -- no information */
270
+	            0;        /* misc. info/SMPTE emphasis off */
271
+          buf[2] = (1 << 7) | /* recording start point: 1 -- no */
272
+                   (1 << 6) | /* recording end point: 1 -- no */
273
+	           (1 << 3) | /* recording mode: 1 -- original */
274
+	            7;         
275
+          buf[3] = (1 << 7) | /* direction: 1 -- forward */
276
+                    0x20;     /* speed */
277
+          buf[4] = (1 << 7) | /* reserved -- always 1 */
278
+                    0x7f;     /* genre category */
279
+	  break;
280
+    case dv_audio_recdate:
281
+    case dv_viedo_recdate:  /* VAUX recording date */
282
+          ct = c->start_time + (time_t)(c->frames / 
283
+	       ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
284
+          localtime_r(&ct, &tc);
285
+	  buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
286
+	                 /* 0xff is very likely to be "unknown" */
287
+	  buf[2] = (3 << 6) | /* reserved -- always 1 */
288
+		   ((tc.tm_mday / 10) << 4) | /* Tens of day */
289
+		   (tc.tm_mday % 10);         /* Units of day */
290
+	  buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
291
+		   (tc.tm_mon % 10);         /* Units of month */
292
+	  buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
293
+		   (tc.tm_year % 10);                 /* Units of year */
294
+          break;
295
+    case dv_audio_rectime:  /* AAUX recording time */
296
+    case dv_video_rectime:  /* VAUX recording time */
297
+          ct = c->start_time + (time_t)(c->frames / 
298
+	       ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
299
+          localtime_r(&ct, &tc);
300
+	  buf[1] = (3 << 6) | /* reserved -- always 1 */
301
+		   0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
302
+	  buf[2] = (1 << 7) | /* reserved -- always 1 */ 
303
+		   ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
304
+		   (tc.tm_sec % 10);         /* Units of seconds */
305
+	  buf[3] = (1 << 7) | /* reserved -- always 1 */
306
+		   ((tc.tm_min / 10) << 4) | /* Tens of minutes */
307
+		   (tc.tm_min % 10);         /* Units of minutes */
308
+	  buf[4] = (3 << 6) | /* reserved -- always 1 */ 
309
+	           ((tc.tm_hour / 10) << 4) | /* Tens of hours */
310
+		   (tc.tm_hour % 10);         /* Units of hours */
311
+	  break;
312
+    case dv_video_source:
313
+	  buf[1] = 0xff; /* reserved -- always 1 */
314
+	  buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
315
+		   (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
316
+		   (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
317
+		   0xf; /* reserved -- always 1 */
318
+	  buf[3] = (3 << 6) | /* reserved -- always 1 */
319
+		   (c->sys->dsf << 5) | /*  system: 60fields/50fields */
320
+		   0; /* signal type video compression */
321
+	  buf[4] = 0xff; /* VISC: 0xff -- no information */
322
+          break;
323
+    case dv_video_control:
324
+	  buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
325
+		   0x3f; /* reserved -- always 1 */
326
+	  buf[2] = 0xc8 | /* reserved -- always b11001xxx */
327
+		   c->aspect;
328
+	  buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
329
+		   (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
330
+		   (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
331
+		   (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
332
+		   0xc; /* reserved -- always b1100 */
333
+	  buf[4] = 0xff; /* reserved -- always 1 */
334
+          break;
335
+    default:
336
+          buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
337
+    }
338
+    return 5;
339
+}
340
+
341
+static inline int dv_write_dif_id(enum dv_section_type t, uint8_t seq_num, 
342
+                                  uint8_t dif_num, uint8_t* buf)
343
+{
344
+    buf[0] = (uint8_t)t;    /* Section type */
345
+    buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
346
+	     (0 << 3) |     /* FSC: for 50Mb/s 0 - first channel; 1 - second */
347
+	     7;             /* reserved -- always 1 */
348
+    buf[2] = dif_num;       /* DIF block number Video: 0-134, Audio: 0-8 */
349
+    return 3;
350
+}
351
+
352
+static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
353
+{
354
+    if (syb_num == 0 || syb_num == 6) {
355
+	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
356
+		 (0<<4)  | /* AP3 (Subcode application ID) */
357
+		 0x0f;     /* reserved -- always 1 */
358
+    } 
359
+    else if (syb_num == 11) {
360
+	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
361
+                 0x7f;     /* reserved -- always 1 */
362
+    }
363
+    else {
364
+	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
365
+                 (0<<4)  | /* APT (Track application ID) */
366
+		 0x0f;     /* reserved -- always 1 */
367
+    }
368
+    buf[1] = 0xf0 |            /* reserved -- always 1 */
369
+             (syb_num & 0x0f); /* SSYB number 0 - 11   */
370
+    buf[2] = 0xff;             /* reserved -- always 1 */
371
+    return 3;
372
+}
373
+
374
+void dv_format_frame(DVMuxContext *c, uint8_t* buf)
375
+{
376
+    int i, j, k;
377
+    
378
+    for (i = 0; i < c->sys->difseg_size; i++) {
379
+       memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
380
+       
381
+       /* DV header: 1DIF */
382
+       buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
383
+       buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
384
+       buf += 72; /* unused bytes */
385
+				   
386
+       /* DV subcode: 2DIFs */
387
+       for (j = 0; j < 2; j++) {
388
+          buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
389
+	  for (k = 0; k < 6; k++) {
390
+	     buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
391
+	     buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
392
+	  }
393
+	  buf += 29; /* unused bytes */
394
+       }
395
+       
396
+       /* DV VAUX: 3DIFs */
397
+       for (j = 0; j < 3; j++) {
398
+	  buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
399
+	  for (k = 0; k < 15 ; k++)
400
+	     buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
401
+	  buf += 2; /* unused bytes */
402
+       } 
403
+       
404
+       /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
405
+       for (j = 0; j < 135; j++) {
406
+            if (j%15 == 0) {
407
+	        buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
408
+	        buf += dv_write_pack(dv_aaux_packs_dist[i][j/15], c, buf);
409
+		buf += 72; /* shuffled PCM audio */
410
+	    }
411
+	    buf += dv_write_dif_id(dv_sect_video, i, j, buf);
412
+	    buf += 77; /* 1 video macro block: 1 bytes control
413
+			                       4 * 14 bytes Y 8x8 data
414
+			                           10 bytes Cr 8x8 data
415
+						   10 bytes Cb 8x8 data */
416
+       }
417
+    }
418
+}
419
+
420
+void dv_inject_audio(DVMuxContext *c, uint8_t* pcm, uint8_t* frame_ptr)
421
+{
422
+    int i, j, d, of;
423
+    for (i = 0; i < c->sys->difseg_size; i++) {
424
+       frame_ptr += 6 * 80; /* skip DIF segment header */
425
+       for (j = 0; j < 9; j++) {
426
+          for (d = 8; d < 80; d+=2) {
427
+	     of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
428
+	     frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
429
+	     frame_ptr[d+1] = pcm[of*2]; //        that DV is a big endian PCM       
430
+          }
431
+          frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
432
+       }
433
+    }
434
+}
435
+
436
+void dv_inject_video(DVMuxContext *c, uint8_t* video_data, uint8_t* frame_ptr)
437
+{
438
+    int i, j;
439
+    int ptr = 0;
440
+
441
+    for (i = 0; i < c->sys->difseg_size; i++) {
442
+       ptr += 6 * 80; /* skip DIF segment header */
443
+       for (j = 0; j < 135; j++) {
444
+            if (j%15 == 0)
445
+	        ptr += 80; /* skip Audio DIF */
446
+	    ptr += 3;
447
+	    memcpy(frame_ptr + ptr, video_data + ptr, 77);
448
+	    ptr += 77;
449
+       }
450
+    }
451
+}
452
+
453
+int dv_audio_frame_size(const DVprofile* sys, int frame)
454
+{
455
+    return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
456
+		                            sizeof(sys->audio_samples_dist[0]))];
457
+}
458
+
459
+const DVprofile* dv_frame_profile(uint8_t* frame)
460
+{
461
+    return &dv_profiles[!!(frame[3] & 0x80)]; /* Header, DSF flag */
462
+}
463
+
464
+/* 
465
+ * This is the dumbest implementation of all -- it simply looks at
466
+ * a fixed offset and if pack isn't there -- fails. We might want
467
+ * to have a fallback mechanism for complete search of missing packs.
468
+ */
469
+const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
470
+{
471
+    int offs;
472
+    
473
+    switch (t) {
474
+    case dv_audio_source:
475
+          offs = (80*6 + 80*16*3 + 3);
476
+	  break;
477
+    case dv_audio_control:
478
+          offs = (80*6 + 80*16*4 + 3);
479
+	  break;
480
+    case dv_video_control:
481
+          offs = (80*5 + 48 + 5);
482
+          break;
483
+    default:
484
+          return NULL;
485
+    }   
486
+
487
+    return (frame[offs] == t ? &frame[offs] : NULL);
488
+}
489
+
490
+/* 
491
+ * There's a couple of assumptions being made here:
492
+ * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
493
+ *    We can pass them upwards when ffmpeg will be ready to deal with them.
494
+ * 2. We don't do software emphasis.
495
+ * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
496
+ *    are converted into 16bit linear ones.
497
+ */
498
+int dv_extract_audio(uint8_t* frame, uint8_t* pcm, AVCodecContext* avctx)
499
+{
500
+    int size, i, j, d, of, smpls, freq, quant;
501
+    uint16_t lc, rc;
502
+    const DVprofile* sys;
503
+    const uint8_t* as_pack;
504
+    
505
+    as_pack = dv_extract_pack(frame, dv_audio_source);
506
+    if (!as_pack)    /* No audio ? */
507
+	return 0;
508
+   
509
+    sys = dv_frame_profile(frame);
510
+    smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
511
+    freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
512
+    quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
513
+    
514
+    if (quant > 1)
515
+	return -1; /* Unsupported quantization */
516
+
517
+    avctx->sample_rate = dv_audio_frequency[freq];
518
+    avctx->channels = 2;
519
+    avctx->bit_rate = avctx->channels * avctx->sample_rate * 16;
520
+    // What about:
521
+    // avctx->frame_size =
522
+   
523
+    size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
524
+
525
+    /* for each DIF segment */
526
+    for (i = 0; i < sys->difseg_size; i++) {
527
+       frame += 6 * 80; /* skip DIF segment header */
528
+       for (j = 0; j < 9; j++) {
529
+          for (d = 8; d < 80; d += 2) {
530
+	     if (quant == 0) {  /* 16bit quantization */
531
+		 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
532
+		 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
533
+		 pcm[of*2+1] = frame[d]; //        that DV is a big endian PCM
534
+		 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
535
+		     pcm[of*2+1] = 0;
536
+	      } else {           /* 12bit quantization */
537
+		 if (i >= sys->difseg_size/2)
538
+	             goto out;  /* We're not doing 4ch at this time */
539
+		       
540
+		 lc = ((uint16_t)frame[d] << 4) | 
541
+		      ((uint16_t)frame[d+2] >> 4);
542
+		 rc = ((uint16_t)frame[d+1] << 4) |
543
+	              ((uint16_t)frame[d+2] & 0x0f);
544
+		 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
545
+		 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
546
+
547
+		 of = sys->audio_shuffle[i][j] + (d - 8)/3 * sys->audio_stride;
548
+		 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
549
+		 pcm[of*2+1] = lc >> 8; //        that DV is a big endian PCM
550
+		 of = sys->audio_shuffle[i+sys->difseg_size/2][j] + 
551
+		      (d - 8)/3 * sys->audio_stride;
552
+		 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
553
+		 pcm[of*2+1] = rc >> 8; //        that DV is a big endian PCM
554
+		 ++d;
555
+	      }
556
+	  }
557
+		
558
+	  frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
559
+        }
560
+    }
561
+
562
+out:
563
+    return size;
564
+}
565
+
566
+/* FIXME: The following three functions could be underengineered ;-) */
567
+void dv_assemble_frame(DVMuxContext *c, uint8_t* video, uint8_t* audio, int asize)
568
+{
569
+    uint8_t pcm[8192];
570
+    uint8_t* frame = &c->frame_buf[0];
571
+    int fsize, reqasize;
572
+   
573
+    if (c->has_audio && c->has_video) {  /* must be a stale frame */
574
+        dv_format_frame(c, frame);
575
+	c->frames++;
576
+	c->has_audio = c->has_video = 0;
577
+    }
578
+    
579
+    if (video) {
580
+        /* FIXME: we have to have more sensible approach than this one */
581
+	if (c->has_video)
582
+	    fprintf(stderr, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
583
+	    
584
+        dv_inject_video(c, video, frame);
585
+	c->has_video = 1;
586
+    } 
587
+    if (audio) {
588
+        reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
589
+        fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
590
+	if (fsize + asize >= reqasize) {
591
+            if (fsize >= reqasize) {
592
+	        fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
593
+            } else {
594
+	        fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);
595
+                memcpy(&pcm[fsize], &audio[0], reqasize - fsize);
596
+		audio += reqasize - fsize;
597
+		asize -= reqasize - fsize;
598
+	    }
599
+	    dv_inject_audio(c, &pcm[0], frame);
600
+	    c->has_audio = 1;
601
+	}
602
+    
603
+        /* FIXME: we have to have more sensible approach than this one */
604
+        if (fifo_size(&c->audio_data, c->audio_data.rptr) + asize >= AVCODEC_MAX_AUDIO_FRAME_SIZE)
605
+	    fprintf(stderr, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
606
+	fifo_write(&c->audio_data, audio, asize, &c->audio_data.wptr);
607
+    }
608
+}
609
+
610
+int dv_core_init(DVMuxContext *c, AVStream *streams[])
611
+{
612
+    /* We have to sort out where audio and where video stream is */
613
+    if (streams[0]->codec.codec_type == CODEC_TYPE_VIDEO &&
614
+        streams[1]->codec.codec_type == CODEC_TYPE_AUDIO) {
615
+       c->vst = 0;
616
+       c->ast = 1;
617
+    }
618
+    else if (streams[1]->codec.codec_type == CODEC_TYPE_VIDEO &&
619
+             streams[0]->codec.codec_type == CODEC_TYPE_AUDIO) {
620
+	    c->vst = 1;
621
+ 	    c->ast = 0;
622
+    } else
623
+        goto bail_out;
624
+  
625
+    /* Some checks -- DV format is very picky about its incoming streams */
626
+    if (streams[c->vst]->codec.codec_id != CODEC_ID_DVVIDEO ||
627
+	streams[c->ast]->codec.codec_id != CODEC_ID_PCM_S16LE)
628
+       goto bail_out;
629
+    if (streams[c->ast]->codec.sample_rate != 48000 ||
630
+	streams[c->ast]->codec.channels != 2)
631
+       goto bail_out;
632
+	
633
+    if (streams[c->vst]->codec.frame_rate == 25 &&
634
+        streams[c->vst]->codec.frame_rate_base == 1) {
635
+        /* May be we have to pick sys for every frame */
636
+	c->sys = &dv_profiles[1];
637
+    }
638
+    else if (streams[c->vst]->codec.frame_rate == 30000 &&
639
+             streams[c->vst]->codec.frame_rate_base == 1001) {
640
+	/* May be we have to pick sys for every frame */
641
+	c->sys = &dv_profiles[0];
642
+    } else
643
+        goto bail_out;
644
+    
645
+    /* Ok, everything seems to be in working order */
646
+    c->frames = 0;
647
+    c->has_audio = c->has_video = 0;
648
+    c->start_time = time(NULL);
649
+    c->aspect = 0; /* 4:3 is the default */
650
+    if (streams[c->vst]->codec.aspect_ratio == 16.0 / 9.0)
651
+        c->aspect = 0x07;
652
+
653
+    if (fifo_init(&c->audio_data, AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0)
654
+        goto bail_out;
655
+
656
+    dv_format_frame(c, &c->frame_buf[0]);
657
+
658
+    return 0;
659
+    
660
+bail_out:
661
+    return -1;
662
+}
663
+
664
+void dv_core_delete(DVMuxContext *c)
665
+{    
666
+    fifo_free(&c->audio_data);
667
+}
0 668
new file mode 100644
... ...
@@ -0,0 +1,73 @@
0
+/* 
1
+ * DV format muxer/demuxer
2
+ * Copyright (c) 2003 Roman Shaposhnik
3
+ *
4
+ * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
5
+ * of DV technical info, and SMPTE 314M specification.
6
+ *
7
+ * This library is free software; you can redistribute it and/or
8
+ * modify it under the terms of the GNU Lesser General Public
9
+ * License as published by the Free Software Foundation; either
10
+ * version 2 of the License, or (at your option) any later version.
11
+ *
12
+ * This library is distributed in the hope that it will be useful,
13
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
14
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15
+ * Lesser General Public License for more details.
16
+ *
17
+ * You should have received a copy of the GNU Lesser General Public
18
+ * License along with this library; if not, write to the Free Software
19
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
20
+ */
21
+#include <time.h>
22
+#include "avformat.h"
23
+
24
+/* 
25
+ * DVprofile is used to express the differences between various 
26
+ * DV flavors. For now it's primarily used for differentiating
27
+ * 525/60 and 625/50, but the plans are to use it for various
28
+ * DV specs as well (e.g. SMPTE314M vs. IEC 61834).
29
+ */
30
+typedef struct DVprofile {
31
+    int            dsf;                  /* value of the dsf in the DV header */
32
+    int            frame_size;           /* total size of one frame in bytes */
33
+    int            difseg_size;          /* number of DIF segments */
34
+    int            frame_rate;      
35
+    int            frame_rate_base;
36
+    int            ltc_divisor;          /* FPS from the LTS standpoint */
37
+    int            height;               /* picture height in pixels */
38
+    uint16_t       *video_place;         /* positions of all DV macro blocks */
39
+    
40
+    int            audio_stride;         /* size of audio_shuffle table */
41
+    int            audio_min_samples[3]; /* min ammount of audio samples */
42
+                                         /* for 48Khz, 44.1Khz and 32Khz */
43
+    int            audio_samples_dist[5];/* how many samples are supposed to be */
44
+                                         /* in each frame in a 5 frames window */
45
+    const uint16_t (*audio_shuffle)[9];  /* PCM shuffling table */
46
+} DVprofile;
47
+
48
+typedef struct DVMuxContext {
49
+    const DVprofile*  sys;    /* Current DV profile. E.g.: 525/60, 625/50 */
50
+    uint8_t     frame_buf[144000]; /* frame under contruction */
51
+    FifoBuffer  audio_data;   /* Fifo for storing excessive amounts of PCM */
52
+    int         frames;       /* Number of a current frame */
53
+    time_t      start_time;   /* Start time of recording */
54
+    uint8_t     aspect;       /* Aspect ID 0 - 4:3, 7 - 16:9 */
55
+    int         ast, vst;     /* Audio and Video stream indecies */
56
+    int         has_audio;    /* frame under contruction has audio */
57
+    int         has_video;    /* frame under contruction has video */
58
+} DVMuxContext;
59
+
60
+void dv_format_frame(DVMuxContext *, uint8_t*);
61
+void dv_inject_audio(DVMuxContext *, uint8_t*, uint8_t*);
62
+void dv_inject_video(DVMuxContext *, uint8_t*, uint8_t*);
63
+
64
+int  dv_extract_audio(uint8_t*, uint8_t*, AVCodecContext*);
65
+
66
+int  dv_audio_frame_size(const DVprofile*, int);
67
+
68
+void dv_assemble_frame(DVMuxContext *, uint8_t*, uint8_t*, int);
69
+int  dv_core_init(DVMuxContext *, AVStream*[]);
70
+void dv_core_delete(DVMuxContext *);
71
+
72
+const DVprofile* dv_frame_profile(uint8_t*);