|
Chris@1
|
1 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
|
|
Chris@1
|
2 <html>
|
|
Chris@1
|
3 <head>
|
|
Chris@1
|
4
|
|
Chris@1
|
5 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-15"/>
|
|
Chris@1
|
6 <title>Ogg Vorbis Documentation</title>
|
|
Chris@1
|
7
|
|
Chris@1
|
8 <style type="text/css">
|
|
Chris@1
|
9 body {
|
|
Chris@1
|
10 margin: 0 18px 0 18px;
|
|
Chris@1
|
11 padding-bottom: 30px;
|
|
Chris@1
|
12 font-family: Verdana, Arial, Helvetica, sans-serif;
|
|
Chris@1
|
13 color: #333333;
|
|
Chris@1
|
14 font-size: .8em;
|
|
Chris@1
|
15 }
|
|
Chris@1
|
16
|
|
Chris@1
|
17 a {
|
|
Chris@1
|
18 color: #3366cc;
|
|
Chris@1
|
19 }
|
|
Chris@1
|
20
|
|
Chris@1
|
21 img {
|
|
Chris@1
|
22 border: 0;
|
|
Chris@1
|
23 }
|
|
Chris@1
|
24
|
|
Chris@1
|
25 #xiphlogo {
|
|
Chris@1
|
26 margin: 30px 0 16px 0;
|
|
Chris@1
|
27 }
|
|
Chris@1
|
28
|
|
Chris@1
|
29 #content p {
|
|
Chris@1
|
30 line-height: 1.4;
|
|
Chris@1
|
31 }
|
|
Chris@1
|
32
|
|
Chris@1
|
33 h1, h1 a, h2, h2 a, h3, h3 a, h4, h4 a {
|
|
Chris@1
|
34 font-weight: bold;
|
|
Chris@1
|
35 color: #ff9900;
|
|
Chris@1
|
36 margin: 1.3em 0 8px 0;
|
|
Chris@1
|
37 }
|
|
Chris@1
|
38
|
|
Chris@1
|
39 h1 {
|
|
Chris@1
|
40 font-size: 1.3em;
|
|
Chris@1
|
41 }
|
|
Chris@1
|
42
|
|
Chris@1
|
43 h2 {
|
|
Chris@1
|
44 font-size: 1.2em;
|
|
Chris@1
|
45 }
|
|
Chris@1
|
46
|
|
Chris@1
|
47 h3 {
|
|
Chris@1
|
48 font-size: 1.1em;
|
|
Chris@1
|
49 }
|
|
Chris@1
|
50
|
|
Chris@1
|
51 li {
|
|
Chris@1
|
52 line-height: 1.4;
|
|
Chris@1
|
53 }
|
|
Chris@1
|
54
|
|
Chris@1
|
55 #copyright {
|
|
Chris@1
|
56 margin-top: 30px;
|
|
Chris@1
|
57 line-height: 1.5em;
|
|
Chris@1
|
58 text-align: center;
|
|
Chris@1
|
59 font-size: .8em;
|
|
Chris@1
|
60 color: #888888;
|
|
Chris@1
|
61 clear: both;
|
|
Chris@1
|
62 }
|
|
Chris@1
|
63 </style>
|
|
Chris@1
|
64
|
|
Chris@1
|
65 </head>
|
|
Chris@1
|
66
|
|
Chris@1
|
67 <body>
|
|
Chris@1
|
68
|
|
Chris@1
|
69 <div id="xiphlogo">
|
|
Chris@1
|
70 <a href="http://www.xiph.org/"><img src="fish_xiph_org.png" alt="Fish Logo and Xiph.Org"/></a>
|
|
Chris@1
|
71 </div>
|
|
Chris@1
|
72
|
|
Chris@1
|
73 <h1>Ogg Vorbis stereo-specific channel coupling discussion</h1>
|
|
Chris@1
|
74
|
|
Chris@1
|
75 <h2>Abstract</h2>
|
|
Chris@1
|
76
|
|
Chris@1
|
77 <p>The Vorbis audio CODEC provides a channel coupling
|
|
Chris@1
|
78 mechanisms designed to reduce effective bitrate by both eliminating
|
|
Chris@1
|
79 interchannel redundancy and eliminating stereo image information
|
|
Chris@1
|
80 labeled inaudible or undesirable according to spatial psychoacoustic
|
|
Chris@1
|
81 models. This document describes both the mechanical coupling
|
|
Chris@1
|
82 mechanisms available within the Vorbis specification, as well as the
|
|
Chris@1
|
83 specific stereo coupling models used by the reference
|
|
Chris@1
|
84 <tt>libvorbis</tt> codec provided by xiph.org.</p>
|
|
Chris@1
|
85
|
|
Chris@1
|
86 <h2>Mechanisms</h2>
|
|
Chris@1
|
87
|
|
Chris@1
|
88 <p>In encoder release beta 4 and earlier, Vorbis supported multiple
|
|
Chris@1
|
89 channel encoding, but the channels were encoded entirely separately
|
|
Chris@1
|
90 with no cross-analysis or redundancy elimination between channels.
|
|
Chris@1
|
91 This multichannel strategy is very similar to the mp3's <em>dual
|
|
Chris@1
|
92 stereo</em> mode and Vorbis uses the same name for its analogous
|
|
Chris@1
|
93 uncoupled multichannel modes.</p>
|
|
Chris@1
|
94
|
|
Chris@1
|
95 <p>However, the Vorbis spec provides for, and Vorbis release 1.0 rc1 and
|
|
Chris@1
|
96 later implement a coupled channel strategy. Vorbis has two specific
|
|
Chris@1
|
97 mechanisms that may be used alone or in conjunction to implement
|
|
Chris@1
|
98 channel coupling. The first is <em>channel interleaving</em> via
|
|
Chris@1
|
99 residue backend type 2, and the second is <em>square polar
|
|
Chris@1
|
100 mapping</em>. These two general mechanisms are particularly well
|
|
Chris@1
|
101 suited to coupling due to the structure of Vorbis encoding, as we'll
|
|
Chris@1
|
102 explore below, and using both we can implement both totally
|
|
Chris@1
|
103 <em>lossless stereo image coupling</em> [bit-for-bit decode-identical
|
|
Chris@1
|
104 to uncoupled modes], as well as various lossy models that seek to
|
|
Chris@1
|
105 eliminate inaudible or unimportant aspects of the stereo image in
|
|
Chris@1
|
106 order to enhance bitrate. The exact coupling implementation is
|
|
Chris@1
|
107 generalized to allow the encoder a great deal of flexibility in
|
|
Chris@1
|
108 implementation of a stereo or surround model without requiring any
|
|
Chris@1
|
109 significant complexity increase over the combinatorially simpler
|
|
Chris@1
|
110 mid/side joint stereo of mp3 and other current audio codecs.</p>
|
|
Chris@1
|
111
|
|
Chris@1
|
112 <p>A particular Vorbis bitstream may apply channel coupling directly to
|
|
Chris@1
|
113 more than a pair of channels; polar mapping is hierarchical such that
|
|
Chris@1
|
114 polar coupling may be extrapolated to an arbitrary number of channels
|
|
Chris@1
|
115 and is not restricted to only stereo, quadraphonics, ambisonics or 5.1
|
|
Chris@1
|
116 surround. However, the scope of this document restricts itself to the
|
|
Chris@1
|
117 stereo coupling case.</p>
|
|
Chris@1
|
118
|
|
Chris@1
|
119 <a name="sqpm"></a>
|
|
Chris@1
|
120 <h3>Square Polar Mapping</h3>
|
|
Chris@1
|
121
|
|
Chris@1
|
122 <h4>maximal correlation</h4>
|
|
Chris@1
|
123
|
|
Chris@1
|
124 <p>Recall that the basic structure of a a Vorbis I stream first generates
|
|
Chris@1
|
125 from input audio a spectral 'floor' function that serves as an
|
|
Chris@1
|
126 MDCT-domain whitening filter. This floor is meant to represent the
|
|
Chris@1
|
127 rough envelope of the frequency spectrum, using whatever metric the
|
|
Chris@1
|
128 encoder cares to define. This floor is subtracted from the log
|
|
Chris@1
|
129 frequency spectrum, effectively normalizing the spectrum by frequency.
|
|
Chris@1
|
130 Each input channel is associated with a unique floor function.</p>
|
|
Chris@1
|
131
|
|
Chris@1
|
132 <p>The basic idea behind any stereo coupling is that the left and right
|
|
Chris@1
|
133 channels usually correlate. This correlation is even stronger if one
|
|
Chris@1
|
134 first accounts for energy differences in any given frequency band
|
|
Chris@1
|
135 across left and right; think for example of individual instruments
|
|
Chris@1
|
136 mixed into different portions of the stereo image, or a stereo
|
|
Chris@1
|
137 recording with a dominant feature not perfectly in the center. The
|
|
Chris@1
|
138 floor functions, each specific to a channel, provide the perfect means
|
|
Chris@1
|
139 of normalizing left and right energies across the spectrum to maximize
|
|
Chris@1
|
140 correlation before coupling. This feature of the Vorbis format is not
|
|
Chris@1
|
141 a convenient accident.</p>
|
|
Chris@1
|
142
|
|
Chris@1
|
143 <p>Because we strive to maximally correlate the left and right channels
|
|
Chris@1
|
144 and generally succeed in doing so, left and right residue is typically
|
|
Chris@1
|
145 nearly identical. We could use channel interleaving (discussed below)
|
|
Chris@1
|
146 alone to efficiently remove the redundancy between the left and right
|
|
Chris@1
|
147 channels as a side effect of entropy encoding, but a polar
|
|
Chris@1
|
148 representation gives benefits when left/right correlation is
|
|
Chris@1
|
149 strong.</p>
|
|
Chris@1
|
150
|
|
Chris@1
|
151 <h4>point and diffuse imaging</h4>
|
|
Chris@1
|
152
|
|
Chris@1
|
153 <p>The first advantage of a polar representation is that it effectively
|
|
Chris@1
|
154 separates the spatial audio information into a 'point image'
|
|
Chris@1
|
155 (magnitude) at a given frequency and located somewhere in the sound
|
|
Chris@1
|
156 field, and a 'diffuse image' (angle) that fills a large amount of
|
|
Chris@1
|
157 space simultaneously. Even if we preserve only the magnitude (point)
|
|
Chris@1
|
158 data, a detailed and carefully chosen floor function in each channel
|
|
Chris@1
|
159 provides us with a free, fine-grained, frequency relative intensity
|
|
Chris@1
|
160 stereo*. Angle information represents diffuse sound fields, such as
|
|
Chris@1
|
161 reverberation that fills the entire space simultaneously.</p>
|
|
Chris@1
|
162
|
|
Chris@1
|
163 <p>*<em>Because the Vorbis model supports a number of different possible
|
|
Chris@1
|
164 stereo models and these models may be mixed, we do not use the term
|
|
Chris@1
|
165 'intensity stereo' talking about Vorbis; instead we use the terms
|
|
Chris@1
|
166 'point stereo', 'phase stereo' and subcategories of each.</em></p>
|
|
Chris@1
|
167
|
|
Chris@1
|
168 <p>The majority of a stereo image is representable by polar magnitude
|
|
Chris@1
|
169 alone, as strong sounds tend to be produced at near-point sources;
|
|
Chris@1
|
170 even non-diffuse, fast, sharp echoes track very accurately using
|
|
Chris@1
|
171 magnitude representation almost alone (for those experimenting with
|
|
Chris@1
|
172 Vorbis tuning, this strategy works much better with the precise,
|
|
Chris@1
|
173 piecewise control of floor 1; the continuous approximation of floor 0
|
|
Chris@1
|
174 results in unstable imaging). Reverberation and diffuse sounds tend
|
|
Chris@1
|
175 to contain less energy and be psychoacoustically dominated by the
|
|
Chris@1
|
176 point sources embedded in them. Thus, we again tend to concentrate
|
|
Chris@1
|
177 more represented energy into a predictably smaller number of numbers.
|
|
Chris@1
|
178 Separating representation of point and diffuse imaging also allows us
|
|
Chris@1
|
179 to model and manipulate point and diffuse qualities separately.</p>
|
|
Chris@1
|
180
|
|
Chris@1
|
181 <h4>controlling bit leakage and symbol crosstalk</h4>
|
|
Chris@1
|
182
|
|
Chris@1
|
183 <p>Because polar
|
|
Chris@1
|
184 representation concentrates represented energy into fewer large
|
|
Chris@1
|
185 values, we reduce bit 'leakage' during cascading (multistage VQ
|
|
Chris@1
|
186 encoding) as a secondary benefit. A single large, monolithic VQ
|
|
Chris@1
|
187 codebook is more efficient than a cascaded book due to entropy
|
|
Chris@1
|
188 'crosstalk' among symbols between different stages of a multistage cascade.
|
|
Chris@1
|
189 Polar representation is a way of further concentrating entropy into
|
|
Chris@1
|
190 predictable locations so that codebook design can take steps to
|
|
Chris@1
|
191 improve multistage codebook efficiency. It also allows us to cascade
|
|
Chris@1
|
192 various elements of the stereo image independently.</p>
|
|
Chris@1
|
193
|
|
Chris@1
|
194 <h4>eliminating trigonometry and rounding</h4>
|
|
Chris@1
|
195
|
|
Chris@1
|
196 <p>Rounding and computational complexity are potential problems with a
|
|
Chris@1
|
197 polar representation. As our encoding process involves quantization,
|
|
Chris@1
|
198 mixing a polar representation and quantization makes it potentially
|
|
Chris@1
|
199 impossible, depending on implementation, to construct a coupled stereo
|
|
Chris@1
|
200 mechanism that results in bit-identical decompressed output compared
|
|
Chris@1
|
201 to an uncoupled encoding should the encoder desire it.</p>
|
|
Chris@1
|
202
|
|
Chris@1
|
203 <p>Vorbis uses a mapping that preserves the most useful qualities of
|
|
Chris@1
|
204 polar representation, relies only on addition/subtraction (during
|
|
Chris@1
|
205 decode; high quality encoding still requires some trig), and makes it
|
|
Chris@1
|
206 trivial before or after quantization to represent an angle/magnitude
|
|
Chris@1
|
207 through a one-to-one mapping from possible left/right value
|
|
Chris@1
|
208 permutations. We do this by basing our polar representation on the
|
|
Chris@1
|
209 unit square rather than the unit-circle.</p>
|
|
Chris@1
|
210
|
|
Chris@1
|
211 <p>Given a magnitude and angle, we recover left and right using the
|
|
Chris@1
|
212 following function (note that A/B may be left/right or right/left
|
|
Chris@1
|
213 depending on the coupling definition used by the encoder):</p>
|
|
Chris@1
|
214
|
|
Chris@1
|
215 <pre>
|
|
Chris@1
|
216 if(magnitude>0)
|
|
Chris@1
|
217 if(angle>0){
|
|
Chris@1
|
218 A=magnitude;
|
|
Chris@1
|
219 B=magnitude-angle;
|
|
Chris@1
|
220 }else{
|
|
Chris@1
|
221 B=magnitude;
|
|
Chris@1
|
222 A=magnitude+angle;
|
|
Chris@1
|
223 }
|
|
Chris@1
|
224 else
|
|
Chris@1
|
225 if(angle>0){
|
|
Chris@1
|
226 A=magnitude;
|
|
Chris@1
|
227 B=magnitude+angle;
|
|
Chris@1
|
228 }else{
|
|
Chris@1
|
229 B=magnitude;
|
|
Chris@1
|
230 A=magnitude-angle;
|
|
Chris@1
|
231 }
|
|
Chris@1
|
232 }
|
|
Chris@1
|
233 </pre>
|
|
Chris@1
|
234
|
|
Chris@1
|
235 <p>The function is antisymmetric for positive and negative magnitudes in
|
|
Chris@1
|
236 order to eliminate a redundant value when quantizing. For example, if
|
|
Chris@1
|
237 we're quantizing to integer values, we can visualize a magnitude of 5
|
|
Chris@1
|
238 and an angle of -2 as follows:</p>
|
|
Chris@1
|
239
|
|
Chris@1
|
240 <p><img src="squarepolar.png" alt="square polar"/></p>
|
|
Chris@1
|
241
|
|
Chris@1
|
242 <p>This representation loses or replicates no values; if the range of A
|
|
Chris@1
|
243 and B are integral -5 through 5, the number of possible Cartesian
|
|
Chris@1
|
244 permutations is 121. Represented in square polar notation, the
|
|
Chris@1
|
245 possible values are:</p>
|
|
Chris@1
|
246
|
|
Chris@1
|
247 <pre>
|
|
Chris@1
|
248 0, 0
|
|
Chris@1
|
249
|
|
Chris@1
|
250 -1,-2 -1,-1 -1, 0 -1, 1
|
|
Chris@1
|
251
|
|
Chris@1
|
252 1,-2 1,-1 1, 0 1, 1
|
|
Chris@1
|
253
|
|
Chris@1
|
254 -2,-4 -2,-3 -2,-2 -2,-1 -2, 0 -2, 1 -2, 2 -2, 3
|
|
Chris@1
|
255
|
|
Chris@1
|
256 2,-4 2,-3 ... following the pattern ...
|
|
Chris@1
|
257
|
|
Chris@1
|
258 ... 5, 1 5, 2 5, 3 5, 4 5, 5 5, 6 5, 7 5, 8 5, 9
|
|
Chris@1
|
259
|
|
Chris@1
|
260 </pre>
|
|
Chris@1
|
261
|
|
Chris@1
|
262 <p>...for a grand total of 121 possible values, the same number as in
|
|
Chris@1
|
263 Cartesian representation (note that, for example, <tt>5,-10</tt> is
|
|
Chris@1
|
264 the same as <tt>-5,10</tt>, so there's no reason to represent
|
|
Chris@1
|
265 both. 2,10 cannot happen, and there's no reason to account for it.)
|
|
Chris@1
|
266 It's also obvious that this mapping is exactly reversible.</p>
|
|
Chris@1
|
267
|
|
Chris@1
|
268 <h3>Channel interleaving</h3>
|
|
Chris@1
|
269
|
|
Chris@1
|
270 <p>We can remap and A/B vector using polar mapping into a magnitude/angle
|
|
Chris@1
|
271 vector, and it's clear that, in general, this concentrates energy in
|
|
Chris@1
|
272 the magnitude vector and reduces the amount of information to encode
|
|
Chris@1
|
273 in the angle vector. Encoding these vectors independently with
|
|
Chris@1
|
274 residue backend #0 or residue backend #1 will result in bitrate
|
|
Chris@1
|
275 savings. However, there are still implicit correlations between the
|
|
Chris@1
|
276 magnitude and angle vectors. The most obvious is that the amplitude
|
|
Chris@1
|
277 of the angle is bounded by its corresponding magnitude value.</p>
|
|
Chris@1
|
278
|
|
Chris@1
|
279 <p>Entropy coding the results, then, further benefits from the entropy
|
|
Chris@1
|
280 model being able to compress magnitude and angle simultaneously. For
|
|
Chris@1
|
281 this reason, Vorbis implements residue backend #2 which pre-interleaves
|
|
Chris@1
|
282 a number of input vectors (in the stereo case, two, A and B) into a
|
|
Chris@1
|
283 single output vector (with the elements in the order of
|
|
Chris@1
|
284 A_0, B_0, A_1, B_1, A_2 ... A_n-1, B_n-1) before entropy encoding. Thus
|
|
Chris@1
|
285 each vector to be coded by the vector quantization backend consists of
|
|
Chris@1
|
286 matching magnitude and angle values.</p>
|
|
Chris@1
|
287
|
|
Chris@1
|
288 <p>The astute reader, at this point, will notice that in the theoretical
|
|
Chris@1
|
289 case in which we can use monolithic codebooks of arbitrarily large
|
|
Chris@1
|
290 size, we can directly interleave and encode left and right without
|
|
Chris@1
|
291 polar mapping; in fact, the polar mapping does not appear to lend any
|
|
Chris@1
|
292 benefit whatsoever to the efficiency of the entropy coding. In fact,
|
|
Chris@1
|
293 it is perfectly possible and reasonable to build a Vorbis encoder that
|
|
Chris@1
|
294 dispenses with polar mapping entirely and merely interleaves the
|
|
Chris@1
|
295 channel. Libvorbis based encoders may configure such an encoding and
|
|
Chris@1
|
296 it will work as intended.</p>
|
|
Chris@1
|
297
|
|
Chris@1
|
298 <p>However, when we leave the ideal/theoretical domain, we notice that
|
|
Chris@1
|
299 polar mapping does give additional practical benefits, as discussed in
|
|
Chris@1
|
300 the above section on polar mapping and summarized again here:</p>
|
|
Chris@1
|
301
|
|
Chris@1
|
302 <ul>
|
|
Chris@1
|
303 <li>Polar mapping aids in controlling entropy 'leakage' between stages
|
|
Chris@1
|
304 of a cascaded codebook.</li>
|
|
Chris@1
|
305 <li>Polar mapping separates the stereo image
|
|
Chris@1
|
306 into point and diffuse components which may be analyzed and handled
|
|
Chris@1
|
307 differently.</li>
|
|
Chris@1
|
308 </ul>
|
|
Chris@1
|
309
|
|
Chris@1
|
310 <h2>Stereo Models</h2>
|
|
Chris@1
|
311
|
|
Chris@1
|
312 <h3>Dual Stereo</h3>
|
|
Chris@1
|
313
|
|
Chris@1
|
314 <p>Dual stereo refers to stereo encoding where the channels are entirely
|
|
Chris@1
|
315 separate; they are analyzed and encoded as entirely distinct entities.
|
|
Chris@1
|
316 This terminology is familiar from mp3.</p>
|
|
Chris@1
|
317
|
|
Chris@1
|
318 <h3>Lossless Stereo</h3>
|
|
Chris@1
|
319
|
|
Chris@1
|
320 <p>Using polar mapping and/or channel interleaving, it's possible to
|
|
Chris@1
|
321 couple Vorbis channels losslessly, that is, construct a stereo
|
|
Chris@1
|
322 coupling encoding that both saves space but also decodes
|
|
Chris@1
|
323 bit-identically to dual stereo. OggEnc 1.0 and later uses this
|
|
Chris@1
|
324 mode in all high-bitrate encoding.</p>
|
|
Chris@1
|
325
|
|
Chris@1
|
326 <p>Overall, this stereo mode is overkill; however, it offers a safe
|
|
Chris@1
|
327 alternative to users concerned about the slightest possible
|
|
Chris@1
|
328 degradation to the stereo image or archival quality audio.</p>
|
|
Chris@1
|
329
|
|
Chris@1
|
330 <h3>Phase Stereo</h3>
|
|
Chris@1
|
331
|
|
Chris@1
|
332 <p>Phase stereo is the least aggressive means of gracefully dropping
|
|
Chris@1
|
333 resolution from the stereo image; it affects only diffuse imaging.</p>
|
|
Chris@1
|
334
|
|
Chris@1
|
335 <p>It's often quoted that the human ear is deaf to signal phase above
|
|
Chris@1
|
336 about 4kHz; this is nearly true and a passable rule of thumb, but it
|
|
Chris@1
|
337 can be demonstrated that even an average user can tell the difference
|
|
Chris@1
|
338 between high frequency in-phase and out-of-phase noise. Obviously
|
|
Chris@1
|
339 then, the statement is not entirely true. However, it's also the case
|
|
Chris@1
|
340 that one must resort to nearly such an extreme demonstration before
|
|
Chris@1
|
341 finding the counterexample.</p>
|
|
Chris@1
|
342
|
|
Chris@1
|
343 <p>'Phase stereo' is simply a more aggressive quantization of the polar
|
|
Chris@1
|
344 angle vector; above 4kHz it's generally quite safe to quantize noise
|
|
Chris@1
|
345 and noisy elements to only a handful of allowed phases, or to thin the
|
|
Chris@1
|
346 phase with respect to the magnitude. The phases of high amplitude
|
|
Chris@1
|
347 pure tones may or may not be preserved more carefully (they are
|
|
Chris@1
|
348 relatively rare and L/R tend to be in phase, so there is generally
|
|
Chris@1
|
349 little reason not to spend a few more bits on them)</p>
|
|
Chris@1
|
350
|
|
Chris@1
|
351 <h4>example: eight phase stereo</h4>
|
|
Chris@1
|
352
|
|
Chris@1
|
353 <p>Vorbis may implement phase stereo coupling by preserving the entirety
|
|
Chris@1
|
354 of the magnitude vector (essential to fine amplitude and energy
|
|
Chris@1
|
355 resolution overall) and quantizing the angle vector to one of only
|
|
Chris@1
|
356 four possible values. Given that the magnitude vector may be positive
|
|
Chris@1
|
357 or negative, this results in left and right phase having eight
|
|
Chris@1
|
358 possible permutation, thus 'eight phase stereo':</p>
|
|
Chris@1
|
359
|
|
Chris@1
|
360 <p><img src="eightphase.png" alt="eight phase"/></p>
|
|
Chris@1
|
361
|
|
Chris@1
|
362 <p>Left and right may be in phase (positive or negative), the most common
|
|
Chris@1
|
363 case by far, or out of phase by 90 or 180 degrees.</p>
|
|
Chris@1
|
364
|
|
Chris@1
|
365 <h4>example: four phase stereo</h4>
|
|
Chris@1
|
366
|
|
Chris@1
|
367 <p>Similarly, four phase stereo takes the quantization one step further;
|
|
Chris@1
|
368 it allows only in-phase and 180 degree out-out-phase signals:</p>
|
|
Chris@1
|
369
|
|
Chris@1
|
370 <p><img src="fourphase.png" alt="four phase"/></p>
|
|
Chris@1
|
371
|
|
Chris@1
|
372 <h3>example: point stereo</h3>
|
|
Chris@1
|
373
|
|
Chris@1
|
374 <p>Point stereo eliminates the possibility of out-of-phase signal
|
|
Chris@1
|
375 entirely. Any diffuse quality to a sound source tends to collapse
|
|
Chris@1
|
376 inward to a point somewhere within the stereo image. A practical
|
|
Chris@1
|
377 example would be balanced reverberations within a large, live space;
|
|
Chris@1
|
378 normally the sound is diffuse and soft, giving a sonic impression of
|
|
Chris@1
|
379 volume. In point-stereo, the reverberations would still exist, but
|
|
Chris@1
|
380 sound fairly firmly centered within the image (assuming the
|
|
Chris@1
|
381 reverberation was centered overall; if the reverberation is stronger
|
|
Chris@1
|
382 to the left, then the point of localization in point stereo would be
|
|
Chris@1
|
383 to the left). This effect is most noticeable at low and mid
|
|
Chris@1
|
384 frequencies and using headphones (which grant perfect stereo
|
|
Chris@1
|
385 separation). Point stereo is is a graceful but generally easy to
|
|
Chris@1
|
386 detect degradation to the sound quality and is thus used in frequency
|
|
Chris@1
|
387 ranges where it is least noticeable.</p>
|
|
Chris@1
|
388
|
|
Chris@1
|
389 <h3>Mixed Stereo</h3>
|
|
Chris@1
|
390
|
|
Chris@1
|
391 <p>Mixed stereo is the simultaneous use of more than one of the above
|
|
Chris@1
|
392 stereo encoding models, generally using more aggressive modes in
|
|
Chris@1
|
393 higher frequencies, lower amplitudes or 'nearly' in-phase sound.</p>
|
|
Chris@1
|
394
|
|
Chris@1
|
395 <p>It is also the case that near-DC frequencies should be encoded using
|
|
Chris@1
|
396 lossless coupling to avoid frame blocking artifacts.</p>
|
|
Chris@1
|
397
|
|
Chris@1
|
398 <h3>Vorbis Stereo Modes</h3>
|
|
Chris@1
|
399
|
|
Chris@1
|
400 <p>Vorbis, as of 1.0, uses lossless stereo and a number of mixed modes
|
|
Chris@1
|
401 constructed out of lossless and point stereo. Phase stereo was used
|
|
Chris@1
|
402 in the rc2 encoder, but is not currently used for simplicity's sake. It
|
|
Chris@1
|
403 will likely be re-added to the stereo model in the future.</p>
|
|
Chris@1
|
404
|
|
Chris@1
|
405 <div id="copyright">
|
|
Chris@1
|
406 The Xiph Fish Logo is a
|
|
Chris@1
|
407 trademark (™) of Xiph.Org.<br/>
|
|
Chris@1
|
408
|
|
Chris@1
|
409 These pages © 1994 - 2005 Xiph.Org. All rights reserved.
|
|
Chris@1
|
410 </div>
|
|
Chris@1
|
411
|
|
Chris@1
|
412 </body>
|
|
Chris@1
|
413 </html>
|
|
Chris@1
|
414
|
|
Chris@1
|
415
|
|
Chris@1
|
416
|
|
Chris@1
|
417
|
|
Chris@1
|
418
|
|
Chris@1
|
419
|
