00001 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
00002 <html>
00003 <head>
00004
00005 <meta http-equiv="Content-Type" content="text/html; charset=iso-8859-15"/>
00006 <title>Ogg Vorbis Documentation</title>
00007
00008 <style type="text/css">
00009 body {
00010 margin: 0 18px 0 18px;
00011 padding-bottom: 30px;
00012 font-family: Verdana, Arial, Helvetica, sans-serif;
00013 color: #333333;
00014 font-size: .8em;
00015 }
00016
00017 a {
00018 color: #3366cc;
00019 }
00020
00021 img {
00022 border: 0;
00023 }
00024
00025 #xiphlogo {
00026 margin: 30px 0 16px 0;
00027 }
00028
00029 #content p {
00030 line-height: 1.4;
00031 }
00032
00033 h1, h1 a, h2, h2 a, h3, h3 a {
00034 font-weight: bold;
00035 color: #ff9900;
00036 margin: 1.3em 0 8px 0;
00037 }
00038
00039 h1 {
00040 font-size: 1.3em;
00041 }
00042
00043 h2 {
00044 font-size: 1.2em;
00045 }
00046
00047 h3 {
00048 font-size: 1.1em;
00049 }
00050
00051 li {
00052 line-height: 1.4;
00053 }
00054
00055 #copyright {
00056 margin-top: 30px;
00057 line-height: 1.5em;
00058 text-align: center;
00059 font-size: .8em;
00060 color: #888888;
00061 clear: both;
00062 }
00063 </style>
00064
00065 </head>
00066
00067 <body>
00068
00069 <div id="xiphlogo">
00070 <a href="http://www.xiph.org/"><img src="fish_xiph_org.png" alt="Fish Logo and Xiph.org"/></a>
00071 </div>
00072
00073 <h1>Ogg Vorbis: Fidelity measurement and terminology discussion</h1>
00074
00075 <p>Terminology discussed in this document is based on common terminology
00076 associated with contemporary codecs such as MPEG I audio layer 3
00077 (mp3). However, some differences in terminology are useful in the
00078 context of Vorbis as Vorbis functions somewhat differently than most
00079 current formats. For clarity, then, we describe a common terminology
00080 for discussion of Vorbis's and other formats' audio quality.</p>
00081
00082 <h2>Subjective and Objective</h2>
00083
00084 <p><em>Objective</em> fidelity is a measure, based on a computable,
00085 mechanical metric, of how carefully an output matches an input. For
00086 example, a stereo amplifier may claim to introduce less that .01%
00087 total harmonic distortion when amplifying an input signal; this claim
00088 is easy to verify given proper equipment, and any number of testers are
00089 likely to arrive at the same, exact results. One need not listen to
00090 the equipment to make this measurement.</p>
00091
00092 <p>However, given two amplifiers with identical, verifiable objective
00093 specifications, listeners may strongly prefer the sound quality of one
00094 over the other. This is actually the case in the decades old debate
00095 [some would say jihad] among audiophiles involving vacuum tube versus
00096 solid state amplifiers. There are people who can tell the difference,
00097 and strongly prefer one over the other despite seemingly identical,
00098 measurable quality. This preference is <em>subjective</em> and
00099 difficult to measure but nonetheless real.</p>
00100
00101 <p>Individual elements of subjective differences often can be qualified,
00102 but overall subjective quality generally is not measurable. Different
00103 observers are likely to disagree on the exact results of a subjective
00104 test as each observer's perspective differs. When measuring
00105 subjective qualities, the best one can hope for is average, empirical
00106 results that show statistical significance across a group.</p>
00107
00108 <p>Perceptual codecs are most concerned with subjective, not objective,
00109 quality. This is why evaluating a perceptual codec via distortion
00110 measures and sonograms alone is useless; these objective measures may
00111 provide insight into the quality or functioning of a codec, but cannot
00112 answer the much squishier subjective question, "Does it sound
00113 good?". The tube amplifier example is perhaps not the best as very few
00114 people can hear, or care to hear, the minute differences between tubes
00115 and transistors, whereas the subjective differences in perceptual
00116 codecs tend to be quite large even when objective differences are
00117 not.</p>
00118
00119 <h2>Fidelity, Artifacts and Differences</h2>
00120
00121 <p>Audio <em>artifacts</em> and loss of fidelity or more simply
00122 put, audio <em>differences</em> are not the same thing.</p>
00123
00124 <p>A loss of fidelity implies differences between the perceived input and
00125 output signal; it does not necessarily imply that the differences in
00126 output are displeasing or that the output sounds poor (although this
00127 is often the case). Tube amplifiers are <em>not</em> higher fidelity
00128 than modern solid state and digital systems. They simply produce a
00129 form of distortion and coloring that is either unnoticeable or actually
00130 pleasing to many ears.</p>
00131
00132 <p>As compared to an original signal using hard metrics, all perceptual
00133 codecs [ASPEC, ATRAC, MP3, WMA, AAC, TwinVQ, AC3 and Vorbis included]
00134 lose objective fidelity in order to reduce bitrate. This is fact. The
00135 idea is to lose fidelity in ways that cannot be perceived. However,
00136 most current streaming applications demand bitrates lower than what
00137 can be achieved by sacrificing only objective fidelity; this is also
00138 fact, despite whatever various company press releases might claim.
00139 Subjective fidelity eventually must suffer in one way or another.</p>
00140
00141 <p>The goal is to choose the best possible tradeoff such that the
00142 fidelity loss is graceful and not obviously noticeable. Most listeners
00143 of FM radio do not realize how much lower fidelity that medium is as
00144 compared to compact discs or DAT. However, when compared directly to
00145 source material, the difference is obvious. A cassette tape is lower
00146 fidelity still, and yet the degradation, relatively speaking, is
00147 graceful and generally easy not to notice. Compare this graceful loss
00148 of quality to an average 44.1kHz stereo mp3 encoded at 80 or 96kbps.
00149 The mp3 might actually be higher objective fidelity but subjectively
00150 sounds much worse.</p>
00151
00152 <p>Thus, when a CODEC <em>must</em> sacrifice subjective quality in order
00153 to satisfy a user's requirements, the result should be a
00154 <em>difference</em> that is generally either difficult to notice
00155 without comparison, or easy to ignore. An <em>artifact</em>, on the
00156 other hand, is an element introduced into the output that is
00157 immediately noticeable, obviously foreign, and undesired. The famous
00158 'underwater' or 'twinkling' effect synonymous with low bitrate (or
00159 poorly encoded) mp3 is an example of an <em>artifact</em>. This
00160 working definition differs slightly from common usage, but the coined
00161 distinction between differences and artifacts is useful for our
00162 discussion.</p>
00163
00164 <p>The goal, when it is absolutely necessary to sacrifice subjective
00165 fidelity, is obviously to strive for differences and not artifacts.
00166 The vast majority of codecs today fail at this task miserably,
00167 predictably, and regularly in one way or another. Avoiding such
00168 failures when it is necessary to sacrifice subjective quality is a
00169 fundamental design objective of Vorbis and that objective is reflected
00170 in Vorbis's design and tuning.</p>
00171
00172 <div id="copyright">
00173 The Xiph Fish Logo is a
00174 trademark (™) of Xiph.Org.<br/>
00175
00176 These pages © 1994 - 2005 Xiph.Org. All rights reserved.
00177 </div>
00178
00179 </body>
00180 </html>
