Music Atrac Can't Handle

[streaml1ne]

Listening tests can be annoying when you're hunting for the subtle artifacts left behind by whichever codec you're testing or whether one version of a codec is better than another. I think it would help everyone understand what artifacts people are hearing and zero in on them quicker if we posted examples. Maybe a small uncompressed or lossless compressed snippet of the original song (~10-20 seconds) around the problem area that we can compress with Sonic Stage and give a listen to. A description of what you hear as the artifact would probably go a long way helping others understand what certain artifacts sound like and how to detect them if they haven't already. If we can build a reference set of snippets where ATRAC blatantly fails we can judge better whether new releases of Sonic Stage show encoder improvements or not. Time codes, descriptions of your listening gear and environment, SS version or recorder hardware are probably all good things to include if you post a file.

(kurisu: If hosting small FLAC'd sound clips is a problem or this post should be somewhere else, work your mojo. I thought Tech Topics might be a good home, but figured I'd play it safe and put it here.)

[Christopher]

Feel free to use the attachment feature.

[dex Otaku]

SQAM is a reasonable place to start.

ATRAC in all its incarnations has known weaknesses which mainly revolve around the paramaters which make it editable [such as fixed encoding frame lengths and a fixed frame overlap length]. The most obvious result is the known problems with both ringing and pre-echo.

Basically -any- source material that contains fast/hard transients should be able to trip up the ATRAC/3/plus encoders, including te infamous castanets sample and basically any electronic music with loud, hard beats. High-frequency transients should be the most obvious to cause artifacts.

[streaml1ne]

http://ff123.net/training/training.html

This looks like a good start.

[BobS]

Plucked insturments such as a harpsichord are difficult for ATRAC also

[zerodB]

http://www.minidisc.org/atrac_breakdown.html

[dex Otaku]

Note also that due to there apparently being no lowpass filtration on the input to the ADC on many MD and HiMD units, a very easy way to trip up ATRAC/3/plus compression [or cause audible harmonics in PCM recordings] is to record sources that include high-intensity ultrasonics.

This is likely to blame for at least a couple of the examples on the page linked to above.

[Fast Eddie]

A description of what you hear as the artifact would probably go a long way helping others understand what certain artifacts sound like and how to detect them if they haven't already.

If they can't hear them, why help them to hear them? Unless they want to of course.

Ignorance is bliss isn't it?

Edited March 17, 2005 by Fast Eddie