This is my first post around here, but I've been browsing this board for quite some time, trying to get familiar with whole ABX testing scene and reading very interesting comments of various codecs.
I'm sort of a Minidisc fan (yeah yeah I know
) and I've been using MD for some 7 years. I'm quite happy with the way ATRAC Type-R sounds, but since I have never seen a test involving ATRAC and others codecs, I decided to do mine.As I feel quite lazy, I'll just cut and paste my original post concerning this test on Minidisc.org's forums (the original post is over there):
QUOTE
As I suggested in an earlier thread, I wanted to provide a true blind-listening test involving ATRAC Type-R SP, ATRAC3+ @ 256 kbps, Vorbis, MP3 Lame and MPC.
Why? Because since Hi-MD came out with Sonicstage 2.0 and Hi-MD, some have been saying it was better than plain old SP, other saying SP was still superior.
I also included MP3, Vorbis and MPC because those three codecs are considered by most people to be far superior to ATRAC and ATRAC3/ATRAC3+.
The sample used is "Castanets.wav", a 6 sec long sample, probably the most common one as it's used in most of hydrogenaudio.org's ABX tests. On hydrogenaudio.org, there's already been testing done with this sample and ATRAC 4.5 and ATRAC Type-R. But as there's no way to digitally upload SP tracks to a PC, the transfer was always done via analog. Which, I believe, can only add distortion to the resulting sound and is not fair enough.
As I happen to own a Type-R MD deck (MDS-JB940) with a digital out, and a soundcard with digital in, I was able to record this track digitally on my PC. This makes this Type-R sample, I believe, the only one available on the Internet with the highest possible quality.
The samples below are to be used with ff123's famous "ABC/Hidden Reference Audio Comparison Tool", which is available at this page: http://ff123.net/abchr/abchr.html
Simply download the binary package, lauch the tool and open the "config.txt" file provided with my samples. You can then choose to do a ABC test (comparing and noting 5 different samples with respect to the original wav file) or do a ABX test (trying to identify the compressed sample against the original). Feel free to do both. I have not yet done any but will post the results afterwards so we can compare.
And for those of you who have never done such a testing, you do need to have a decent pair of headphones and be in a quiet environnment. Don't do the test with any pair of speakers, especially PC speakers, chances are you won't hear any difference.
Now here's how these samples were recorded and compressed.
The WAV file came directly from its source: http://www.pcabx.com/product/mds-jb920/
This is where it was first used I believe...
The ATRAC3+ @ 256 kbps file was encoded with Sonicstage 2.1 (I don't own a Hi-MD), then burned from within Sonicstage to an Audio CD from which a WAV file was created.
The MP3 Lame file was encoded with Lame 3.93.1, using the alt-extreme preset.
The Vorbis file was done using the Vorbis Megamix II encoder, based on libvorbis 1.1 RC1 (available here: http://www.rarewares.org/quantumknot/oggenc-megamix2.exe) and which is believed by many to be the encoder of choice for Vorbis. The encoding was used with quality 6 (-q6).
The MPC file was encoded with mppenc 1.14 (http://www.saunalahti.fi/~cse/mpc/encoders/mppenc-windows-1.14.zip), using the "insane" preset.
And finally, the ATRAC Type-R preset was done by burning the original wav sample to a CDRW, playing that CDRW on my CD deck, hooked up through digital coax to my MD deck, and then recorded the sample on a blank and brand new TDK RXG disk, and captured the sample on my PC through the optical out of the deck to the optical in of my soundcard. My soundcard itself is a Terratec Aureon fun, an entry-level card which doesn't have any 24 bit capabilities. Therefore, the output of the deck was set to 16 bits so the soundcard wouldn't have to do any dithering (which would have been poorly done given the value of the card). I used Nero's Wave Editor to record the sample from the Optical In.
All of the compressed samples were then converted back to WAV.
You can download the samples at this location (approx. 5 MB): http://perso.wanadoo.fr/rmstitanic/minidisc/samples.zip
Looking forward to reading your results
If anyone of you is willing to do a more complete test using other samples, feel free to send me the original WAV file and I'd be glad to create the Type-R sample using the same method as above.
Why? Because since Hi-MD came out with Sonicstage 2.0 and Hi-MD, some have been saying it was better than plain old SP, other saying SP was still superior.
I also included MP3, Vorbis and MPC because those three codecs are considered by most people to be far superior to ATRAC and ATRAC3/ATRAC3+.
The sample used is "Castanets.wav", a 6 sec long sample, probably the most common one as it's used in most of hydrogenaudio.org's ABX tests. On hydrogenaudio.org, there's already been testing done with this sample and ATRAC 4.5 and ATRAC Type-R. But as there's no way to digitally upload SP tracks to a PC, the transfer was always done via analog. Which, I believe, can only add distortion to the resulting sound and is not fair enough.
As I happen to own a Type-R MD deck (MDS-JB940) with a digital out, and a soundcard with digital in, I was able to record this track digitally on my PC. This makes this Type-R sample, I believe, the only one available on the Internet with the highest possible quality.
The samples below are to be used with ff123's famous "ABC/Hidden Reference Audio Comparison Tool", which is available at this page: http://ff123.net/abchr/abchr.html
Simply download the binary package, lauch the tool and open the "config.txt" file provided with my samples. You can then choose to do a ABC test (comparing and noting 5 different samples with respect to the original wav file) or do a ABX test (trying to identify the compressed sample against the original). Feel free to do both. I have not yet done any but will post the results afterwards so we can compare.
And for those of you who have never done such a testing, you do need to have a decent pair of headphones and be in a quiet environnment. Don't do the test with any pair of speakers, especially PC speakers, chances are you won't hear any difference.
Now here's how these samples were recorded and compressed.
The WAV file came directly from its source: http://www.pcabx.com/product/mds-jb920/
This is where it was first used I believe...
The ATRAC3+ @ 256 kbps file was encoded with Sonicstage 2.1 (I don't own a Hi-MD), then burned from within Sonicstage to an Audio CD from which a WAV file was created.
The MP3 Lame file was encoded with Lame 3.93.1, using the alt-extreme preset.
The Vorbis file was done using the Vorbis Megamix II encoder, based on libvorbis 1.1 RC1 (available here: http://www.rarewares.org/quantumknot/oggenc-megamix2.exe) and which is believed by many to be the encoder of choice for Vorbis. The encoding was used with quality 6 (-q6).
The MPC file was encoded with mppenc 1.14 (http://www.saunalahti.fi/~cse/mpc/encoders/mppenc-windows-1.14.zip), using the "insane" preset.
And finally, the ATRAC Type-R preset was done by burning the original wav sample to a CDRW, playing that CDRW on my CD deck, hooked up through digital coax to my MD deck, and then recorded the sample on a blank and brand new TDK RXG disk, and captured the sample on my PC through the optical out of the deck to the optical in of my soundcard. My soundcard itself is a Terratec Aureon fun, an entry-level card which doesn't have any 24 bit capabilities. Therefore, the output of the deck was set to 16 bits so the soundcard wouldn't have to do any dithering (which would have been poorly done given the value of the card). I used Nero's Wave Editor to record the sample from the Optical In.
All of the compressed samples were then converted back to WAV.
You can download the samples at this location (approx. 5 MB): http://perso.wanadoo.fr/rmstitanic/minidisc/samples.zip
Looking forward to reading your results
If anyone of you is willing to do a more complete test using other samples, feel free to send me the original WAV file and I'd be glad to create the Type-R sample using the same method as above.
Like I said, I look forward to your comments on this, and any criticism on how I recorded or encoded the samples are welcome, after all, this is my first try at this and it's probably far from perfect
(Did I really hear a difference between MPC --insane and the wav?!?! Maybe it was the tiramisu I just ate 
)