Chris@1:
Chris@1: Terminology discussed in this document is based on common terminology Chris@1: associated with contemporary codecs such as MPEG I audio layer 3 Chris@1: (mp3). However, some differences in terminology are useful in the Chris@1: context of Vorbis as Vorbis functions somewhat differently than most Chris@1: current formats. For clarity, then, we describe a common terminology Chris@1: for discussion of Vorbis's and other formats' audio quality.
Chris@1: Chris@1:Objective fidelity is a measure, based on a computable, Chris@1: mechanical metric, of how carefully an output matches an input. For Chris@1: example, a stereo amplifier may claim to introduce less that .01% Chris@1: total harmonic distortion when amplifying an input signal; this claim Chris@1: is easy to verify given proper equipment, and any number of testers are Chris@1: likely to arrive at the same, exact results. One need not listen to Chris@1: the equipment to make this measurement.
Chris@1: Chris@1:However, given two amplifiers with identical, verifiable objective Chris@1: specifications, listeners may strongly prefer the sound quality of one Chris@1: over the other. This is actually the case in the decades old debate Chris@1: [some would say jihad] among audiophiles involving vacuum tube versus Chris@1: solid state amplifiers. There are people who can tell the difference, Chris@1: and strongly prefer one over the other despite seemingly identical, Chris@1: measurable quality. This preference is subjective and Chris@1: difficult to measure but nonetheless real.
Chris@1: Chris@1:Individual elements of subjective differences often can be qualified, Chris@1: but overall subjective quality generally is not measurable. Different Chris@1: observers are likely to disagree on the exact results of a subjective Chris@1: test as each observer's perspective differs. When measuring Chris@1: subjective qualities, the best one can hope for is average, empirical Chris@1: results that show statistical significance across a group.
Chris@1: Chris@1:Perceptual codecs are most concerned with subjective, not objective, Chris@1: quality. This is why evaluating a perceptual codec via distortion Chris@1: measures and sonograms alone is useless; these objective measures may Chris@1: provide insight into the quality or functioning of a codec, but cannot Chris@1: answer the much squishier subjective question, "Does it sound Chris@1: good?". The tube amplifier example is perhaps not the best as very few Chris@1: people can hear, or care to hear, the minute differences between tubes Chris@1: and transistors, whereas the subjective differences in perceptual Chris@1: codecs tend to be quite large even when objective differences are Chris@1: not.
Chris@1: Chris@1:Audio artifacts and loss of fidelity or more simply Chris@1: put, audio differences are not the same thing.
Chris@1: Chris@1:A loss of fidelity implies differences between the perceived input and Chris@1: output signal; it does not necessarily imply that the differences in Chris@1: output are displeasing or that the output sounds poor (although this Chris@1: is often the case). Tube amplifiers are not higher fidelity Chris@1: than modern solid state and digital systems. They simply produce a Chris@1: form of distortion and coloring that is either unnoticeable or actually Chris@1: pleasing to many ears.
Chris@1: Chris@1:As compared to an original signal using hard metrics, all perceptual Chris@1: codecs [ASPEC, ATRAC, MP3, WMA, AAC, TwinVQ, AC3 and Vorbis included] Chris@1: lose objective fidelity in order to reduce bitrate. This is fact. The Chris@1: idea is to lose fidelity in ways that cannot be perceived. However, Chris@1: most current streaming applications demand bitrates lower than what Chris@1: can be achieved by sacrificing only objective fidelity; this is also Chris@1: fact, despite whatever various company press releases might claim. Chris@1: Subjective fidelity eventually must suffer in one way or another.
Chris@1: Chris@1:The goal is to choose the best possible tradeoff such that the Chris@1: fidelity loss is graceful and not obviously noticeable. Most listeners Chris@1: of FM radio do not realize how much lower fidelity that medium is as Chris@1: compared to compact discs or DAT. However, when compared directly to Chris@1: source material, the difference is obvious. A cassette tape is lower Chris@1: fidelity still, and yet the degradation, relatively speaking, is Chris@1: graceful and generally easy not to notice. Compare this graceful loss Chris@1: of quality to an average 44.1kHz stereo mp3 encoded at 80 or 96kbps. Chris@1: The mp3 might actually be higher objective fidelity but subjectively Chris@1: sounds much worse.
Chris@1: Chris@1:Thus, when a CODEC must sacrifice subjective quality in order Chris@1: to satisfy a user's requirements, the result should be a Chris@1: difference that is generally either difficult to notice Chris@1: without comparison, or easy to ignore. An artifact, on the Chris@1: other hand, is an element introduced into the output that is Chris@1: immediately noticeable, obviously foreign, and undesired. The famous Chris@1: 'underwater' or 'twinkling' effect synonymous with low bitrate (or Chris@1: poorly encoded) mp3 is an example of an artifact. This Chris@1: working definition differs slightly from common usage, but the coined Chris@1: distinction between differences and artifacts is useful for our Chris@1: discussion.
Chris@1: Chris@1:The goal, when it is absolutely necessary to sacrifice subjective Chris@1: fidelity, is obviously to strive for differences and not artifacts. Chris@1: The vast majority of codecs today fail at this task miserably, Chris@1: predictably, and regularly in one way or another. Avoiding such Chris@1: failures when it is necessary to sacrifice subjective quality is a Chris@1: fundamental design objective of Vorbis and that objective is reflected Chris@1: in Vorbis's design and tuning.
Chris@1: Chris@1: