In my last test, I compared a castanets sample encoded with ATRAC Type-R (the latest incarnation of the ATRAC SP codec for MiniDisc) and ATRAC 4.5. ATRAC 4.5 was awful, which should come as no surprise, as the version is from circa 1996 and is not used anymore. ATRAC Type-R, while much improved, exhibited smearing in the guitar section (according to test done by Guruboolez) and pre-echo on certain attacks (heard by your's truly).

Guruboolez uploaded the same castanet sample encoded with the newest ATRAC algorithm, ATRAC3plus @256kbps, which is to replace ATRAC Type-R (SP mode) in the new 1GB Hi-MD machines. The sample can be found here. To my knowledge, this should represent the first ABX results for ATRAC3plus, and there will hopefully be more!

Sample A: Castanets - Sony Atrac3+ 256kbps.wav
Sample B: Castanets.wav

1 of 1, p = 0.5
2 of 2, p = 0.25
3 of 3, p = 0.125
4 of 4, p = 0.062
5 of 5, p = 0.031
6 of 6, p = 0.015
7 of 7, p = 0.0070
8 of 8, p = 0.0030
9 of 9, p = 0.0010
10 of 10, p = <.001
11 of 11, p = <.001
12 of 12, p = <.001
13 of 13, p = <.001
14 of 14, p = <.001
15 of 15, p = <.001
16 of 16, p = <.001
17 of 17, p = <.001
18 of 18, p = <.001
19 of 19, p = <.001
20 of 20, p = <.001

Sample A: Castanets - Sony Atrac3+ 256kbps.wav
Sample B: ATRAC-R.wav

1 of 1, p = 0.5
2 of 2, p = 0.25
3 of 3, p = 0.125
4 of 4, p = 0.062
5 of 5, p = 0.031
6 of 6, p = 0.015
7 of 7, p = 0.0070
8 of 8, p = 0.0030
9 of 9, p = 0.0010
10 of 10, p = <.001
11 of 11, p = <.001
12 of 12, p = <.001
13 of 13, p = <.001
14 of 14, p = <.001
15 of 15, p = <.001
16 of 16, p = <.001
17 of 17, p = <.001
18 of 18, p = <.001
19 of 19, p = <.001
20 of 20, p = <.001

RESULTS: ATRAC3plus at 256kbps is not transparent with castanets, p<.001. ATRAC3plus is distinguishable from (and significantly worse than) ATRAC Type-R with castanets, p<.001.


Comments: ATRAC3plus is definitely distinguishable from both the uncompressed file and the ATRAC Type-R compressed file. ATRAC3plus has a nasty, side-to-side, metallic pre-echo effect (almost like sparkling) between 2-4 seconds (where the castanets are being played). Attacks are definitely not as sharp as the original, either. I did not listen for the guitar smearing that Guruboolez mentioned earlier, so that might be something of interest for another test. In any case, not very encouraging for those lusting after Hi-MD machines.

I would appreciate other samples from people who have access to SonicStage 2.0 and the ATRAC3plus codec. If you submit anything, please upload a .zip archive containing FLAC copies of the original file and the decoded ATRAC3plus sample. Thanks

I can confirm your results.

foo_abx v1.2 report
foobar2000 v0.8.2
2004/08/01 13:08:34

File A: file://F:\projektit\atrac\Castanets - Sony Atrac3+ 256kbps.ape
File B: file://F:\projektit\atrac\castanets.wav

13:08:36 : Test started.
13:09:11 : 01/01 50.0%
13:09:35 : 02/02 25.0%
13:09:49 : 03/03 12.5%
13:10:41 : 04/04 6.3%
13:11:00 : 05/05 3.1%
13:11:12 : 06/06 1.6%
13:11:25 : 07/07 0.8%
13:11:46 : 07/08 3.5%
13:11:59 : 08/09 2.0%
13:12:08 : 09/10 1.1%
13:12:26 : 09/11 3.3%
13:12:38 : 10/12 1.9%
13:12:52 : 11/13 1.1%
13:13:07 : 12/14 0.6%
13:13:09 : Test finished.

----------
Total: 12/14 (0.6%)

It doesn't sound bad when it isn't compared to original, but it is clearly distinguishable.

And that is perhaps the primary draw of MD in regards to sound quality...blissful ignorance. I still like my Sharp MD-MS702 unit! Thank goodness I haven't done any ABX tests, I might have discarded it long ago.

I don't know that what I heard could be decribed as pleasant or blissful. Between the 2-4 second mark, the artifact I heard could best be described as billions of tiny little CO2 bubbles rising to the top of a cola in a metal glass. Or perhaps my ears are just a bit off

It'd say that it sounds like the castanet clangs would be at double speed.

I did a new test today because I was requested to. I used the original atrac3plus sample which can be found from the first message and I did a new one using new SonicStage 3.0. Here are my results.


It appears that Sony has done something. I've attached the new sample.

I'm truly sorry for being so late, but I completely forgot your intervention. I've just downloaded the encoded you've provided (thanks ), quickly tested it on unoptimal conditions (notebook and AC 97 soundchip), and... I'm disapointed. The encoding suffers from the same kind of distortions introduced by SonicStage 2 (IIRC) and atrac3plus, i.e. strong metalic noise on castanets + smearing audible on first guitar note. My ABX score:





I've also compared this new encoding to an atrac3plus one made with SonicStage 2, and if I've noticed some progress (same kind of irritating noise, but even stronger), the difference doesn't really impress me.

Fortunately, Sony have decided at last to support MP3 directly on their MD units...

Anyway, thanks for providing samples and making listening tests possible