Skip to content

What is “Quality?”

Three words have been bandied about: Quality, Clarity, Fidelity.

Anyone remember what "Hi-Fi" actually means? <g>

"Quality" will always be the word to use. The FAQ should, IMO, seek to explain that it means three things at once.

 

Quality as "Fidelity" refers to how close to the original source material the MP3 is. That depends on:

a(1)  Treatment of Source
and/or 
a(2)  Accurate Ripping

b.  No "Normalization" 
That doesn't preclude the setting of a "record" level for analogue sources that avoids clipping of the loudest portions; but it does mean using that same setting for all the material from that source.

c.  Quality Encoder

and only lastly

d.  Bitrate

All producers of MP3s should aim to optimize (a)-(c). Choice of (d) then becomes the only "variable" that determines the fidelity of the final result to the source. Fidelity, then, is a mathematically precise matter. It is not one of "use your ears". That comes next.

 

Quality as "Clarity" refers to the sound of the final result, as "auditioned" (I presume this is the audio equivalent of "previewed") by the person making the MP3, as well as perceived by the intended audience.

A "Clear" MP3 is free from artifacts such as flanging and wooshing, has Dolby Pro Logic retained through Stereo encoding instead of Joint Stereo (if desired - I have tested low bitrate Stereo MP3s and a surprising amount of Pro Logic is retained even then), or sounds as stereophonic as it should if Joint Stereo, does not click or pop or jump, and generally sounds as good as it should *given its intended fidelity*. This is not a mathematical operation, it is a "use your ears" operation. The reason for this is that the output from an MP3 depends on the decoder in use, the sound-card/amplifier/speakers/ headphones/whatever, the personal preferences of the listener in terms of Treble/Bass/EQ, overall volume, "normalization" etc. A quiet track, ripped with a fidelity of nominal value "n", should sound as faithful to the original as that value "n" intends, when listened to under the same circumstances as the original.

 

Finally, Quality as "Quality" refers to the sum of Fidelity and Clarity, plus the little things that make for a "good" MP3. Good use of ID3 tags is part of it. Tracks that should run together when played back should do so with little or no discernible jump between them (within the limitations of the player). Mono material (and I still have yet to find a *fast* and reliable way of identifying such) should be encoded at mono. Whole CDs may have an SFV file to assist in ensuring that propagation has not broken any of it. NFO files (or /0 details) should be prepared with enough care to provide all the information they can/should.

Where does that leave us with 320kbps, 112kpbs, VBR, HQ?

Well, if prepared with care, the bit-rate reflects only the fidelity, not the overall quality. A 320kbps encode is of low quality if it is shipped out with no ID3 tags and exceeds the 75 MB per day guideline. A 112kbps encode is of high quality if it precisely reflects the intention to create a 112kbps encode, although it will be of intentionally low fidelity and some people may take issue with this. VBR is of low quality if it cannot be played correctly by all VBR-capable players.  Use of HQ will affect clarity, not fidelity (AIUI), while use of GoGo or Xing will affect fidelity.

 

So, what should the FAQ tell us?

1.  Fidelity can be controlled precisely by following the instructions given in the FAQ and its Tips sections or at external venues it recommends.

2.  The level of fidelity to aim for depends on your needs, on the nature of the source, and the target audience. The audiobooks group has an entirely different audience than the Beatles group, and their material should be prepared accordingly. (see below)

3.  Clarity is a matter of listening with a critical ear to each piece of encoded material, and nothing more. If it sounds as intended, it is good. If it sounds wrong (compared with the source), it is bad, and needs redoing.

4.  Quality is in the overall presentation, and details the level of fidelity (i.e. Was it a vinyl source? What's the bitrate?  What encoder?) as well as clarity (Are there source artifacts that make this a highly faithful reproduction of a poor source?) and other salient information.

 

Questions:

  • We should be seeking to maximize bandwidth. This means *appropriate* use of bitrate, encoding (S/DC/JS/M) and sampling frequency (44kHz, 32kHz etc). I would like to see a set of guidance notes produced that give recommended settings for different source types.For instance, only true Stereo preserves Dolby Pro Logic, but when is Joint Stereo appropriate? Is it, for example, when a Vocal Cut produces a decrease in overall amplitude of more than 15dB, or is it not that simple?

    Most people use 44kHz (except for spoken word). I have seen arguments that older material only needs 32kHz. Can this be justified and encapsulated in a set of guidelines?

    If the source is vinyl or tape, is there a maximum bitrate beyond which no further fidelity is to be gained, because of source limitations?

    Can a table of EBR (Effective Bit Rates) be produced as a set of guidelines (not rules) to follow?

  • There is, IMO, justification in suggesting that, *as a rule*, the encoded size of a post should reflect the 75MB daily cap as much as the intended fidelity. I have posted all my Stereo Blues CDs as 160 so that I can be sure of being within the cap every day, while achieving as high a fidelity as I can (given the essentially OOP nature of the material). For "normal" postings (i.e. not OOP, rarities, bootlegs etc.) should a maximum bitrate  be recommended to achieve the same? This especially applies to the  pre-release material, but also to anything that is genuinely available off the shelf.

 

Just my 0.02 Euros...

--
Prof. Eugene Q Thrung III