release/src/router/libvorbis/doc/framing.html

   1 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
   2 <html>
   3 <head>
   4
   5 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-15"/>
   6 <title>Ogg Vorbis Documentation</title>
   7
   8 <style type="text/css">
   9 body {
  10   margin: 0 18px 0 18px;
  11   padding-bottom: 30px;
  12   font-family: Verdana, Arial, Helvetica, sans-serif;
  13   color: #333333;
  14   font-size: .8em;
  15 }
  16
  17 a {
  18   color: #3366cc;
  19 }
  20
  21 img {
  22   border: 0;
  23 }
  24
  25 #xiphlogo {
  26   margin: 30px 0 16px 0;
  27 }
  28
  29 #content p {
  30   line-height: 1.4;
  31 }
  32
  33 h1, h1 a, h2, h2 a, h3, h3 a {
  34   font-weight: bold;
  35   color: #ff9900;
  36   margin: 1.3em 0 8px 0;
  37 }
  38
  39 h1 {
  40   font-size: 1.3em;
  41 }
  42
  43 h2 {
  44   font-size: 1.2em;
  45 }
  46
  47 h3 {
  48   font-size: 1.1em;
  49 }
  50
  51 li {
  52   line-height: 1.4;
  53 }
  54
  55 #copyright {
  56   margin-top: 30px;
  57   line-height: 1.5em;
  58   text-align: center;
  59   font-size: .8em;
  60   color: #888888;
  61   clear: both;
  62 }
  63 </style>
  64
  65 </head>
  66
  67 <body>
  68
  69 <div id="xiphlogo">
  70   <a href="http://www.xiph.org/"><img src="fish_xiph_org.png" alt="Fish Logo and Xiph.org"/></a>
  71 </div>
  72
  73 <h1>Ogg logical bitstream framing</h1>
  74
  75 <h2>Ogg bitstreams</h2>
  76
  77 <p>The Ogg transport bitstream is designed to provide framing, error
  78 protection and seeking structure for higher-level codec streams that
  79 consist of raw, unencapsulated data packets, such as the Vorbis audio
  80 codec or Theora video codec.</p>
  81
  82 <h2>Application example: Vorbis</h2>
  83
  84 <p>Vorbis encodes short-time blocks of PCM data into raw packets of
  85 bit-packed data. These raw packets may be used directly by transport
  86 mechanisms that provide their own framing and packet-separation
  87 mechanisms (such as UDP datagrams). For stream based storage (such as
  88 files) and transport (such as TCP streams or pipes), Vorbis uses the
  89 Ogg bitstream format to provide framing/sync, sync recapture
  90 after error, landmarks during seeking, and enough information to
  91 properly separate data back into packets at the original packet
  92 boundaries without relying on decoding to find packet boundaries.</p>
  93
  94 <h2>Design constraints for Ogg bitstreams</h2>
  95
  96 <ol>
  97 <li>True streaming; we must not need to seek to build a 100%
  98   complete bitstream.</li>
  99 <li>Use no more than approximately 1-2% of bitstream bandwidth for
 100   packet boundary marking, high-level framing, sync and seeking.</li>
 101 <li>Specification of absolute position within the original sample
 102   stream.</li>
 103 <li>Simple mechanism to ease limited editing, such as a simplified
 104   concatenation mechanism.</li>
 105 <li>Detection of corruption, recapture after error and direct, random
 106   access to data at arbitrary positions in the bitstream.</li>
 107 </ol>
 108
 109 <h2>Logical and Physical Bitstreams</h2>
 110
 111 <p>A <em>logical</em> Ogg bitstream is a contiguous stream of
 112 sequential pages belonging only to the logical bitstream. A
 113 <em>physical</em> Ogg bitstream is constructed from one or more
 114 than one logical Ogg bitstream (the simplest physical bitstream
 115 is simply a single logical bitstream). We describe below the exact
 116 formatting of an Ogg logical bitstream. Combining logical
 117 bitstreams into more complex physical bitstreams is described in the
 118 <a href="oggstream.html">Ogg bitstream overview</a>. The exact
 119 mapping of raw Vorbis packets into a valid Ogg Vorbis physical
 120 bitstream is described in the Vorbis I Specification.</p>
 121
 122 <h2>Bitstream structure</h2>
 123
 124 <p>An Ogg stream is structured by dividing incoming packets into
 125 segments of up to 255 bytes and then wrapping a group of contiguous
 126 packet segments into a variable length page preceded by a page
 127 header. Both the header size and page size are variable; the page
 128 header contains sizing information and checksum data to determine
 129 header/page size and data integrity.</p>
 130
 131 <p>The bitstream is captured (or recaptured) by looking for the beginning
 132 of a page, specifically the capture pattern. Once the capture pattern
 133 is found, the decoder verifies page sync and integrity by computing
 134 and comparing the checksum. At that point, the decoder can extract the
 135 packets themselves.</p>
 136
 137 <h3>Packet segmentation</h3>
 138
 139 <p>Packets are logically divided into multiple segments before encoding
 140 into a page. Note that the segmentation and fragmentation process is a
 141 logical one; it's used to compute page header values and the original
 142 page data need not be disturbed, even when a packet spans page
 143 boundaries.</p>
 144
 145 <p>The raw packet is logically divided into [n] 255 byte segments and a
 146 last fractional segment of &lt; 255 bytes. A packet size may well
 147 consist only of the trailing fractional segment, and a fractional
 148 segment may be zero length. These values, called "lacing values" are
 149 then saved and placed into the header segment table.</p>
 150
 151 <p>An example should make the basic concept clear:</p>
 152
 153 <pre>
 154 <tt>
 155 raw packet:
 156   ___________________________________________
 157  |______________packet data__________________| 753 bytes
 158
 159 lacing values for page header segment table: 255,255,243
 160 </tt>
 161 </pre>
 162
 163 <p>We simply add the lacing values for the total size; the last lacing
 164 value for a packet is always the value that is less than 255. Note
 165 that this encoding both avoids imposing a maximum packet size as well
 166 as imposing minimum overhead on small packets (as opposed to, eg,
 167 simply using two bytes at the head of every packet and having a max
 168 packet size of 32k. Small packets (&lt;255, the typical case) are
 169 penalized with twice the segmentation overhead). Using the lacing
 170 values as suggested, small packets see the minimum possible
 171 byte-aligned overheade (1 byte) and large packets, over 512 bytes or
 172 so, see a fairly constant ~.5% overhead on encoding space.</p>
 173
 174 <p>Note that a lacing value of 255 implies that a second lacing value
 175 follows in the packet, and a value of &lt; 255 marks the end of the
 176 packet after that many additional bytes. A packet of 255 bytes (or a
 177 multiple of 255 bytes) is terminated by a lacing value of 0:</p>
 178
 179 <pre><tt>
 180 raw packet:
 181   _______________________________
 182  |________packet data____________|          255 bytes
 183
 184 lacing values: 255, 0
 185 </tt></pre>
 186
 187 <p>Note also that a 'nil' (zero length) packet is not an error; it
 188 consists of nothing more than a lacing value of zero in the header.</p>
 189
 190 <h3>Packets spanning pages</h3>
 191
 192 <p>Packets are not restricted to beginning and ending within a page,
 193 although individual segments are, by definition, required to do so.
 194 Packets are not restricted to a maximum size, although excessively
 195 large packets in the data stream are discouraged; the Ogg
 196 bitstream specification strongly recommends nominal page size of
 197 approximately 4-8kB (large packets are foreseen as being useful for
 198 initialization data at the beginning of a logical bitstream).</p>
 199
 200 <p>After segmenting a packet, the encoder may decide not to place all the
 201 resulting segments into the current page; to do so, the encoder places
 202 the lacing values of the segments it wishes to belong to the current
 203 page into the current segment table, then finishes the page. The next
 204 page is begun with the first value in the segment table belonging to
 205 the next packet segment, thus continuing the packet (data in the
 206 packet body must also correspond properly to the lacing values in the
 207 spanned pages. The segment data in the first packet corresponding to
 208 the lacing values of the first page belong in that page; packet
 209 segments listed in the segment table of the following page must begin
 210 the page body of the subsequent page).</p>
 211
 212 <p>The last mechanic to spanning a page boundary is to set the header
 213 flag in the new page to indicate that the first lacing value in the
 214 segment table continues rather than begins a packet; a header flag of
 215 0x01 is set to indicate a continued packet. Although mandatory, it
 216 is not actually algorithmically necessary; one could inspect the
 217 preceding segment table to determine if the packet is new or
 218 continued. Adding the information to the packet_header flag allows a
 219 simpler design (with no overhead) that needs only inspect the current
 220 page header after frame capture. This also allows faster error
 221 recovery in the event that the packet originates in a corrupt
 222 preceding page, implying that the previous page's segment table
 223 cannot be trusted.</p>
 224
 225 <p>Note that a packet can span an arbitrary number of pages; the above
 226 spanning process is repeated for each spanned page boundary. Also a
 227 'zero termination' on a packet size that is an even multiple of 255
 228 must appear even if the lacing value appears in the next page as a
 229 zero-length continuation of the current packet. The header flag
 230 should be set to 0x01 to indicate that the packet spanned, even though
 231 the span is a nil case as far as data is concerned.</p>
 232
 233 <p>The encoding looks odd, but is properly optimized for speed and the
 234 expected case of the majority of packets being between 50 and 200
 235 bytes (note that it is designed such that packets of wildly different
 236 sizes can be handled within the model; placing packet size
 237 restrictions on the encoder would have only slightly simplified design
 238 in page generation and increased overall encoder complexity).</p>
 239
 240 <p>The main point behind tracking individual packets (and packet
 241 segments) is to allow more flexible encoding tricks that requiring
 242 explicit knowledge of packet size. An example is simple bandwidth
 243 limiting, implemented by simply truncating packets in the nominal case
 244 if the packet is arranged so that the least sensitive portion of the
 245 data comes last.</p>
 246
 247 <h3>Page header</h3>
 248
 249 <p>The headering mechanism is designed to avoid copying and re-assembly
 250 of the packet data (ie, making the packet segmentation process a
 251 logical one); the header can be generated directly from incoming
 252 packet data. The encoder buffers packet data until it finishes a
 253 complete page at which point it writes the header followed by the
 254 buffered packet segments.</p>
 255
 256 <h4>capture_pattern</h4>
 257
 258 <p>A header begins with a capture pattern that simplifies identifying
 259 pages; once the decoder has found the capture pattern it can do a more
 260 intensive job of verifying that it has in fact found a page boundary
 261 (as opposed to an inadvertent coincidence in the byte stream).</p>
 262
 263 <pre><tt>
 264  byte value
 265
 266   0  0x4f 'O'
 267   1  0x67 'g'
 268   2  0x67 'g'
 269   3  0x53 'S'
 270 </tt></pre>
 271
 272 <h4>stream_structure_version</h4>
 273
 274 <p>The capture pattern is followed by the stream structure revision:</p>
 275
 276 <pre><tt>
 277  byte value
 278
 279   4  0x00
 280 </tt></pre>
 281
 282 <h4>header_type_flag</h4>
 283
 284 <p>The header type flag identifies this page's context in the bitstream:</p>
 285
 286 <pre><tt>
 287  byte value
 288
 289   5  bitflags: 0x01: unset = fresh packet
 290                        set = continued packet
 291                0x02: unset = not first page of logical bitstream
 292                        set = first page of logical bitstream (bos)
 293                0x04: unset = not last page of logical bitstream
 294                        set = last page of logical bitstream (eos)
 295 </tt></pre>
 296
 297 <h4>absolute granule position</h4>
 298
 299 <p>(This is packed in the same way the rest of Ogg data is packed; LSb
 300 of LSB first. Note that the 'position' data specifies a 'sample'
 301 number (eg, in a CD quality sample is four octets, 16 bits for left
 302 and 16 bits for right; in video it would likely be the frame number.
 303 It is up to the specific codec in use to define the semantic meaning
 304 of the granule position value). The position specified is the total
 305 samples encoded after including all packets finished on this page
 306 (packets begun on this page but continuing on to the next page do not
 307 count). The rationale here is that the position specified in the
 308 frame header of the last page tells how long the data coded by the
 309 bitstream is. A truncated stream will still return the proper number
 310 of samples that can be decoded fully.</p>
 311
 312 <p>A special value of '-1' (in two's complement) indicates that no packets
 313 finish on this page.</p>
 314
 315 <pre><tt>
 316  byte value
 317
 318   6  0xXX LSB
 319   7  0xXX
 320   8  0xXX
 321   9  0xXX
 322  10  0xXX
 323  11  0xXX
 324  12  0xXX
 325  13  0xXX MSB
 326 </tt></pre>
 327
 328 <h4>stream serial number</h4>
 329
 330 <p>Ogg allows for separate logical bitstreams to be mixed at page
 331 granularity in a physical bitstream. The most common case would be
 332 sequential arrangement, but it is possible to interleave pages for
 333 two separate bitstreams to be decoded concurrently. The serial
 334 number is the means by which pages physical pages are associated with
 335 a particular logical stream. Each logical stream must have a unique
 336 serial number within a physical stream:</p>
 337
 338 <pre><tt>
 339  byte value
 340
 341  14  0xXX LSB
 342  15  0xXX
 343  16  0xXX
 344  17  0xXX MSB
 345 </tt></pre>
 346
 347 <h4>page sequence no</h4>
 348
 349 <p>Page counter; lets us know if a page is lost (useful where packets
 350 span page boundaries).</p>
 351
 352 <pre><tt>
 353  byte value
 354
 355  18  0xXX LSB
 356  19  0xXX
 357  20  0xXX
 358  21  0xXX MSB
 359 </tt></pre>
 360
 361 <h4>page checksum</h4>
 362
 363 <p>32 bit CRC value (direct algorithm, initial val and final XOR = 0,
 364 generator polynomial=0x04c11db7). The value is computed over the
 365 entire header (with the CRC field in the header set to zero) and then
 366 continued over the page. The CRC field is then filled with the
 367 computed value.</p>
 368
 369 <p>(A thorough discussion of CRC algorithms can be found in <a
 370 href="http://www.ross.net/crc/download/crc_v3.txt">"A
 371 Painless Guide to CRC Error Detection Algorithms"</a> by Ross
 372 Williams <a href="mailto:ross@ross.net">ross@ross.net</a>.)</p>
 373
 374 <pre><tt>
 375  byte value
 376
 377  22  0xXX LSB
 378  23  0xXX
 379  24  0xXX
 380  25  0xXX MSB
 381 </tt></pre>
 382
 383 <h4>page_segments</h4>
 384
 385 <p>The number of segment entries to appear in the segment table. The
 386 maximum number of 255 segments (255 bytes each) sets the maximum
 387 possible physical page size at 65307 bytes or just under 64kB (thus
 388 we know that a header corrupted so as destroy sizing/alignment
 389 information will not cause a runaway bitstream. We'll read in the
 390 page according to the corrupted size information that's guaranteed to
 391 be a reasonable size regardless, notice the checksum mismatch, drop
 392 sync and then look for recapture).</p>
 393
 394 <pre><tt>
 395  byte value
 396
 397  26 0x00-0xff (0-255)
 398 </tt></pre>
 399
 400 <h4>segment_table (containing packet lacing values)</h4>
 401
 402 <p>The lacing values for each packet segment physically appearing in
 403 this page are listed in contiguous order.</p>
 404
 405 <pre><tt>
 406  byte value
 407
 408  27 0x00-0xff (0-255)
 409  [...]
 410  n  0x00-0xff (0-255, n=page_segments+26)
 411 </tt></pre>
 412
 413 <p>Total page size is calculated directly from the known header size and
 414 lacing values in the segment table. Packet data segments follow
 415 immediately after the header.</p>
 416
 417 <p>Page headers typically impose a flat .25-.5% space overhead assuming
 418 nominal ~8k page sizes. The segmentation table needed for exact
 419 packet recovery in the streaming layer adds approximately .5-1%
 420 nominal assuming expected encoder behavior in the 44.1kHz, 128kbps
 421 stereo encodings.</p>
 422
 423 <div id="copyright">
 424   The Xiph Fish Logo is a
 425   trademark (&trade;) of Xiph.Org.<br/>
 426
 427   These pages &copy; 1994 - 2005 Xiph.Org. All rights reserved.
 428 </div>
 429
 430 </body>
 431 </html>