🚀 go-pugleaf

md_fan@usa.net wrote in message <79bfu8$smh$1@nnrp1.dejanews.com>...
>In article <PSXt2.754$ob2.1685@news.flash.net>,
>  "Arny Kr�ger" <arnyk@flash.net> wrote:
>
>> ]>I would have asked "which is better" since you can get very good sound
from
>> >both formats.  ATRAC uses a 292k bitrate and most MP3s found on the Net
use a
>> >128k bitrate.  ATRAC is considerably cleaner, especially in the high end.
>> >When MP3s are encoded at 160k they start sounding pretty good and at 256k
>> >they get hard to distinguish from ATRAC.
>>
>> What experiments are these conclusions based on?
>
>Blind and semi-blind listening tests, which is the only way you can really
>test any perceptually coded format, no?


I've had some success at getting perceptual coders to misbehave using some of
the standard test (multi)tones I use to do tests at http://www.pcavtech.com .

For example, my tests of the Sony MDS-J902
http://www.pcavtech.com/play-rec/mds-j920-2/index.htm show audible frequency
response variations that the owner of the deck says correspond to his
experience with it. There is another as-yet not complete test of the MZ-R2
(http://www.pcavtech.com/play-rec/Sony_MZ-R2/index.htm) that also shows
potentially audible situations, and the owner of that machine says that what
he sees in the chart corresponds to his experience with it. OTOH, the
corresponding charts for the MDS JE 510
http://www.pcavtech.com/play-rec/Sony%20MDS-JE510/index.htm  (via digital I/O)
seem to be sufficiently precise that "catching" it should be a decent
listening test challenge. I think it implements a later version of ATRAC than
the others.

Not posted at PCAVTech are some casual technical evaluations of MP3 coders and
decoders, many of which show audible variations that should be in the audible
range, and some of which have been ABX'd and found to be audible.

Any of the suspect technical situations I've seen in MD devices should be
possible to represent with 100.00% technical accuracy on an ABX listening test
CD, such as the ones I'm using in the Digital Challenge series of CD's. I
imagine that a run-off between the best MP3 coders I've run into and the best
MD recorder/player would be of some interest...

Perceptual coders are sonically very complex things, compared to simple amps,
preamps, and CD players. My approach to testing them has included using tools
designed for loudspeaker testing, which has born some fruit. For example, Doug
Plumb's speaker evaluation software, ETF,
(http://www.etfacoustic.com/index.htm ) will show some kinds of misbehavior in
perceptual coders. But, due to their complexity, and the fact that their
problems are often in the range that is audible, albeit sometimes tending to
be subtle, listening tests are very relevant.

IMO, the magnum opus of perceptual coder tests is shown at
http://drogo.cselt.stet.it/mpeg/public/w2006.zip , which I believe that we
have discussed before.