Debian Codegraph @

doc/html/vbr.html

raw
1 <?xml version="1.0" encoding="UTF-8"?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
3 "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
4 <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
5 <head>
6 <meta http-equiv="Content-Type" content="application/xhtml+xml; charset=UTF-8" />
7 <meta name="generator" content="AsciiDoc 8.6.10" />
8 <title>TwoLAME: MPEG Audio Layer II VBR</title>
9 <link rel="stylesheet" href="./asciidoc.css" type="text/css" />
10
11
12 <link rel="stylesheet" href="./twolame.css" type="text/css" />
13 <script type="text/javascript" src="./asciidoc.js"></script>
14 <script type="text/javascript">
15 /*<![CDATA[*/
16 asciidoc.install();
17 /*]]>*/
18 </script>
19 </head>
20 <body class="article">
21 <div id="header">
22 <h1>TwoLAME: MPEG Audio Layer II VBR</h1>
23 <span id="revnumber">version 0.4.0</span>
24 </div>
25 <div id="content">
26 <div class="sect1">
27 <h2 id="_contents">Contents</h2>
28 <div class="sectionbody">
29 <div class="ulist"><ul>
30 <li>
31 <p>
32 Introduction
33 </p>
34 </li>
35 <li>
36 <p>
37 Usage
38 </p>
39 </li>
40 <li>
41 <p>
42 Bitrate Ranges for various Sampling frequencies
43 </p>
44 </li>
45 <li>
46 <p>
47 Why can&#8217;t the bitrate vary from 32kbps to 384kbps for every file?
48 </p>
49 </li>
50 <li>
51 <p>
52 Short Answer
53 </p>
54 </li>
55 <li>
56 <p>
57 Long Answer
58 </p>
59 </li>
60 <li>
61 <p>
62 Tech Stuff
63 </p>
64 </li>
65 </ul></div>
66 </div>
67 </div>
68 <div class="sect1">
69 <h2 id="_introduction">Introduction</h2>
70 <div class="sectionbody">
71 <div class="paragraph"><p>VBR mode works by selecting a different bitrate for each frame. Frames
72 which are harder to encode will be allocated more bits i.e. a higher bitrate.</p></div>
73 <div class="paragraph"><p>LayerII VBR is a complete hack - the ISO standard actually says that decoders are not
74 required to support it. As a hack, its implementation is a pain to try and understand.
75 If you&#8217;re mega-keen to get full range VBR working, either (a) send me money (b) grab the
76 ISO standard and a C compiler and email me.</p></div>
77 </div>
78 </div>
79 <div class="sect1">
80 <h2 id="_usage">Usage</h2>
81 <div class="sectionbody">
82 <div class="literalblock">
83 <div class="content">
84 <pre><code>twolame -v [level] inputfile outputfile.</code></pre>
85 </div></div>
86 <div class="paragraph"><p>A level of 5 works very well for me.</p></div>
87 <div class="paragraph"><p>The level value can is a measurement of quality - the higher
88 the level the higher the average bitrate of the resultant file.
89 See TECH STUFF for a better explanation of what the value does.</p></div>
90 <div class="paragraph"><p>The confusing part of my implementation of LayerII VBR is that it&#8217;s different from MP3 VBR.</p></div>
91 <div class="ulist"><ul>
92 <li>
93 <p>
94 The range of bitrates used is controlled by the input sampling frequency. (See below "Bitrate ranges")
95 </p>
96 </li>
97 <li>
98 <p>
99 The tendency to use higher bitrates is governed by the &lt;level&gt;.
100 </p>
101 </li>
102 </ul></div>
103 <div class="paragraph"><p>E.g. Say you have a 44.1kHz Stereo file. In VBR mode, the bitrate can range from 192 to 384 kbps.</p></div>
104 <div class="paragraph"><p>Using "-v -5" will force the encoder to favour the lower bitrate.</p></div>
105 <div class="paragraph"><p>Using "-v 5" will force the encoder to favour the upper bitrate.</p></div>
106 <div class="paragraph"><p>The value can actually be <strong>any</strong> int. -27, 233, 47. The larger the number, the greater
107 the bitrate bias.</p></div>
108 </div>
109 </div>
110 <div class="sect1">
111 <h2 id="_bitrate_ranges">Bitrate Ranges</h2>
112 <div class="sectionbody">
113 <div class="paragraph"><p>When making a VBR stream, the bitrate is only allowed to vary within
114 set limits</p></div>
115 <div class="literalblock">
116 <div class="content">
117 <pre><code>48kHz
118 Stereo: 112-384kbps Mono: 56-192kbps</code></pre>
119 </div></div>
120 <div class="literalblock">
121 <div class="content">
122 <pre><code>44.1kHz &amp; 32kHz
123 Stereo: 192-384kbps Mono: 96-192kbps</code></pre>
124 </div></div>
125 <div class="literalblock">
126 <div class="content">
127 <pre><code>24kHz, 22.05kHz &amp; 16kHz
128 Stereo/Mono: 8-160kbps</code></pre>
129 </div></div>
130 </div>
131 </div>
132 <div class="sect1">
133 <h2 id="_why_doesn_8217_t_the_vbr_mode_work_the_same_as_mp3vbr_the_short_answer">Why doesn&#8217;t the VBR mode work the same as MP3VBR? The Short Answer</h2>
134 <div class="sectionbody">
135 <div class="paragraph"><p><strong>Why can&#8217;t the bitrate vary from 32kbps to 384kbps for every file?</strong></p></div>
136 <div class="paragraph"><p>According to the standard (ISO/IEC 11172-3:1993) Section 2.4.2.3</p></div>
137 <div class="literalblock">
138 <div class="content">
139 <pre><code>"In order to provide the smallest possible delay and complexity, the
140 decoder is not required to support a continuously variable bitrate when
141 in layer I or II. Layer III supports variable bitrate by switching the
142 bitrate index."</code></pre>
143 </div></div>
144 <div class="literalblock">
145 <div class="content">
146 <pre><code>and</code></pre>
147 </div></div>
148 <div class="literalblock">
149 <div class="content">
150 <pre><code>"For Layer II, not all combinations of total bitrate and mode are allowed."</code></pre>
151 </div></div>
152 <div class="paragraph"><p>Hence, most LayerII coders would not have been written with VBR in mind, and
153 LayerII VBR is a hack. It works for limited cases. Getting it to work to
154 the same extent as MP3-style VBR will be a major hack.</p></div>
155 <div class="paragraph"><p>(If you <strong>really</strong> want better bitrate ranges, read "The Long Answer" and submit your mega-patch.)</p></div>
156 </div>
157 </div>
158 <div class="sect1">
159 <h2 id="_why_doesn_8217_t_the_vbr_mode_work_the_same_as_mp3vbr_the_long_answer">Why doesn&#8217;t the VBR mode work the same as MP3VBR? The Long Answer</h2>
160 <div class="sectionbody">
161 <div class="paragraph"><p><strong>Why can&#8217;t the bitrate vary from 32kbps to 384kbps for every file?</strong></p></div>
162 <div class="sect2">
163 <h3 id="_reason_1_the_standard_limits_the_range">Reason 1: The standard limits the range</h3>
164 <div class="paragraph"><p>As quoted above from the standard for 48/44.1/32kHz:</p></div>
165 <div class="literalblock">
166 <div class="content">
167 <pre><code>"For Layer II, not all combinations of total bitrate and mode are allowed. See
168 the following table."</code></pre>
169 </div></div>
170 <div class="literalblock">
171 <div class="content">
172 <pre><code>Bitrate Allowed Modes
173 (kbps)
174 32 mono only
175 48 mono only
176 56 mono only
177 64 all modes
178 80 mono only
179 96 all modes
180 112 all modes
181 128 all modes
182 160 all modes
183 192 all modes
184 224 stereo only
185 256 stereo only
186 320 stereo only
187 384 stereo only</code></pre>
188 </div></div>
189 <div class="paragraph"><p>So based upon this table alone, you <strong>could</strong> have VBR stereo encoding which varies
190 smoothly from 96 to 384kbps. Or you could have have VBR mono encoding which varies from
191 32 to 192kbps. But since the top and bottom bitrates don&#8217;t apply to all modes, it would
192 be impossible to have a stereo file encoded from 32 to 384 kbps.</p></div>
193 <div class="paragraph"><p>But this isn&#8217;t what is really limiting the allowable bitrate range - the bit allocation
194 tables are the major hurdle.</p></div>
195 </div>
196 <div class="sect2">
197 <h3 id="_reason_2_the_bit_allocation_tables_don_8217_t_allow_it">Reason 2: The bit allocation tables don&#8217;t allow it</h3>
198 <div class="paragraph"><p>From the standard, Section 2.4.3.3.1 "Bit allocation decoding"</p></div>
199 <div class="literalblock">
200 <div class="content">
201 <pre><code>"For different combinations of bitrate and sampling frequency, different bit
202 allocation tables exist.</code></pre>
203 </div></div>
204 <div class="paragraph"><p>These bit allocation tables are pre-determined tables (in Annex B of the standard) which
205 indicate</p></div>
206 <div class="ulist"><ul>
207 <li>
208 <p>
209 how many bits to read for the initial data (2,3 or 4)
210 </p>
211 </li>
212 <li>
213 <p>
214 these bits are then used as an index back into the table to
215 find the number of quantize levels for the samples in this subband
216 </p>
217 </li>
218 </ul></div>
219 <div class="paragraph"><p>But the table used (and hence the number of bits and the calculated index) are different
220 for different combinations of bitrate and sampling frequency.</p></div>
221 <div class="paragraph"><p>I will use TableB.2a as an example.</p></div>
222 <div class="paragraph"><p>Table B.2a Applies for the following combinations.</p></div>
223 <div class="literalblock">
224 <div class="content">
225 <pre><code>Sampling Freq Bitrates in (kbps/channel) [emphasis: this is a PER CHANNEL bitrate]
226 48 56, 64, 80, 96, 112, 128, 160, 192
227 44.1 56, 64, 80
228 32 56, 64, 80</code></pre>
229 </div></div>
230 <div class="paragraph"><p>If we have a STEREO 48kHz input file, and we use this table, then the bitrates
231 we could calculate from this would be 112, 128, 160, 192, 224, 256, 320 and 384 kbps.</p></div>
232 <div class="paragraph"><p>This table contains no information on how to encode stuff at bitrates less than 112kbps
233 (for a stereo file). You would have to load allocation table B.2c to encode stereo at
234 64kbps and 128kbps.</p></div>
235 <div class="paragraph"><p>Since it would be a MAJOR piece of hacking to get the different tables shifted in and out
236 during the encoding process, once an allocation table is loaded <strong>IT IS NOT CHANGED</strong>.</p></div>
237 <div class="paragraph"><p>Hence, the best table is picked at the start of the encoding process, and the encoder
238 is stuck with it for the rest of the encode.</p></div>
239 <div class="paragraph"><p>For twolame-02j, I have picked the table it loads for different
240 sampling frequencies in order to optimize the range of bitrates possible.</p></div>
241 <div class="literalblock">
242 <div class="content">
243 <pre><code>48 kHz - Table B.2a
244 Stereo Bitrate Range: 112 - 384
245 Mono Bitrate Range : 56 - 192</code></pre>
246 </div></div>
247 <div class="literalblock">
248 <div class="content">
249 <pre><code>44.1/32 kHz - Table B.2b
250 Stereo Bitrate Range: 192 - 384
251 Mono Bitrate Range: 96 - 192</code></pre>
252 </div></div>
253 <div class="literalblock">
254 <div class="content">
255 <pre><code>24/22.05/16 kHz - LSF Table (Standard ISO/IEC 13818.3:1995 Annex B, Table B.1)
256 There is only 1 table for the Lower Sampling Frequencies
257 All modes (mono and stereo) are allowable at all bitrates
258 So at the Lower Sampling Frequencies you *can* have a completely variable
259 bitrate over the entire range.</code></pre>
260 </div></div>
261 </div>
262 </div>
263 </div>
264 <div class="sect1">
265 <h2 id="_tech_stuff">Tech Stuff</h2>
266 <div class="sectionbody">
267 <div class="paragraph"><p>The VBR mode is mainly centered around the main_bit_allocation() and
268 a_bit_allocation() routines in encode.c.</p></div>
269 <div class="paragraph"><p>The limited range of VBR is due to my particular implementation which restricts
270 ranges to within one alloc table (see tables B.2a, B.2b, B.2c and B.2d in ISO 11172).
271 The VBR range for 32/44.1khz lies within B.2b, and the 48khz VBR lies within table B.2a.</p></div>
272 <div class="paragraph"><p>I&#8217;m not sure whether it is worth extending these ranges down to lower bitrates.
273 The work required to switch alloc tables <strong>during</strong> the encoding is major.</p></div>
274 <div class="paragraph"><p>In the case of silence, it might be worth doing a quick check for very low signals
275 and writing a pre-calculated <strong>blank</strong> 32kpbs frame. [probably also a lot of work].</p></div>
276 </div>
277 </div>
278 <div class="sect1">
279 <h2 id="_how_cbr_works">How CBR works</h2>
280 <div class="sectionbody">
281 <div class="ulist"><ul>
282 <li>
283 <p>
284 Use the psycho model to determine the MNRs for each subband
285 [MNR = the ratio of "masking" to "noise"]
286 (From an encoding perspective, a bigger MNR in a subband means that
287 it sounds better since the noise is more masked))
288 </p>
289 </li>
290 <li>
291 <p>
292 calculate the available data bits (adb) for this bitrate.
293 </p>
294 </li>
295 <li>
296 <p>
297 Based upon the MNR (Masking:Noise Ratio) values, allocate bits to each
298 subband
299 </p>
300 </li>
301 <li>
302 <p>
303 Keep increasing the bits to whichever subband currently has the min MNR
304 value until we have no bits left.
305 </p>
306 </li>
307 <li>
308 <p>
309 This mode does not guarentee that all the subbands are without noise
310 ie there may still be subbands with MNR less than 0.0 (noisy!)
311 </p>
312 </li>
313 </ul></div>
314 </div>
315 </div>
316 <div class="sect1">
317 <h2 id="_how_vbr_works">How VBR works</h2>
318 <div class="sectionbody">
319 <div class="ulist"><ul>
320 <li>
321 <p>
322 pretend we have lots of bits to spare, and work out the bits which would
323 raise the MNR in each subband to the level given by the argument on the
324 command line "-v [int]"
325 </p>
326 </li>
327 <li>
328 <p>
329 Pick the bitrate which has more bits than the required_bits we just calculated
330 </p>
331 </li>
332 <li>
333 <p>
334 calculate a_bit_allocation()
335 </p>
336 </li>
337 <li>
338 <p>
339 VBR "guarantees" that all subbands have MNR &gt; VBRLEVEL or that we have
340 reached the maximum bitrate.
341 </p>
342 </li>
343 </ul></div>
344 </div>
345 </div>
346 <div class="sect1">
347 <h2 id="_future">FUTURE</h2>
348 <div class="sectionbody">
349 <div class="ulist"><ul>
350 <li>
351 <p>
352 with this VBR mode, we know the bits aren&#8217;t going to run out, so we can
353 just assign them "greedily".
354 </p>
355 </li>
356 <li>
357 <p>
358 VBR_a_bit_allocation() is yet to be written :)
359 </p>
360 </li>
361 </ul></div>
362 </div>
363 </div>
364 </div>
365 <div id="footnotes"><hr /></div>
366 <div id="footer">
367 <div id="footer-text">
368 Version 0.4.0<br />
369 Last updated
370 2019-10-11 17:26:25 BST
371 </div>
372 </div>
373 </body>
374 </html>