Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5 (Read 18069 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Since pre-echo seems to be one of the easier artifacts for me to hear (and I'm no golden ear, let me tell you), I decided to try yet another castanets test, this time using my MZ-N1 portable MiniDisc recorder. I recently bought it for the price of a movie ticket, and it is a neat toy, though I  have trouble taking ATRAC seriously for anything but voice and field recordings.

I compared the original sample to a) a sample encoded with ATRAC 4.5, both found here, and b) the original encoded via S/PDIF into my MZ-N1, which has the newer ATRAC Type-R codec. Then I captured the recording off the MD with Audacity, normalized it to the same level as the other two files, and compared.

1R = C:\Documents and Settings\Angelo\Desktop\castanets_mds-jb920_digi24f.wav

---------------------------------------
General Comments:

---------------------------------------
ABX Results:
Original vs C:\Documents and Settings\Angelo\Desktop\castanets_mds-jb920_digi24f.wav
    16 out of 16, pval < 0.001

This was very, very easy: lots of smearing, flanging, and in general, a "duller" sound.

ATRAC Type-R vs. Original

New trial was sucessful: see edited post below

Now for Type-R vs. 4.5:


1L = C:\Documents and Settings\Angelo\Desktop\ATRAC-R.wav

---------------------------------------
General Comments:

---------------------------------------
ABX Results:
Original vs C:\Documents and Settings\Angelo\Desktop\ATRAC-R.wav
    16 out of 16, pval < 0.001

Again, very easy. The Type-R encoding sounds very good to my ears in comparison to the 4.5 encoding; sounds are much sharper and the flanging is gone.

Result: ATRAC-R shows some improvement over 4.5 in terms of pre-echo speficic to castanets. Since this is only 1 sample, it cannot be generalized any further.

I'm very surprised that Type-R seems to have improved over 4.5, at least on this one sample. Now, please don't read this as an endorsement for MD  Don't get me wrong; I love my minidisc as a fun toy, but I am certain once I get to Fatboy and Creaking and Waiting, things will not be so nice for Type-R.

To be continued......

 

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #1
Interesting report. Thanks 
But is it possible for you to upload some samples?

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #2
Last night, I became quite fatigued and had trouble with my last ABX test, ATRAC-R vs. the original. Also, I hadn't considered the fact that jet-engine-like fan noise from my powerbook trying to run VPC and a billion other things at once might interfere with my ability to hear things  Thus, I have re-conducted that particular ABX test, and the results are much different from those of last night.

Sample A: ATRAC-R.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
4 of 5, p = 0.187
5 of 6, p = 0.109
6 of 7, p = 0.062
7 of 8, p = 0.035
8 of 9, p = 0.019
9 of 10, p = 0.01
10 of 11, p = 0.0050
11 of 12, p = 0.0030
11 of 13, p = 0.011
12 of 14, p = 0.0060
13 of 15, p = 0.0030
14 of 16, p = 0.0020

(I used the "quick ABX" function of Java ABC/HR this time, as to eliminate fan noise from trying to run Windows XP in emulation.)

This was a bit harder than my AAC vs. LAME test, in that I twice guessed incorrectly. Thus, I achieved significance at the p<.01 level, but not p<.001 as with the other two tests. The sound difference is somewhat subtle here: ATRAC-R sounded duller and "fatter" on some of the clicks, which is what I listened for when I could concentrate hard enough. On ATRAC 4.5, I listened for smearing and flanging from left to right, which were very difficult to percieve with Type-R.

Result: ATRAC Type-R, while much better than the 4.5 encoding, still is not transparent on the castanets sample, suggesting that pre-echo is still an issue. I look forward to trying out some other samples with ATRAC Type-R, since there seems to be precious little empirical data on how it actually performs. Most of the (limited) listening tests done have been with 4.5, and many audiophiles still claim that ATRAC Type-R can sound better than CDDA, a claim which I intend to disprove 

(Edit: grammar)

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #3
Quote
Interesting report. Thanks 
But is it possible for you to upload some samples?
[{POST_SNAPBACK}][/a]


Your wish is my command:[a href="http://www.hydrogenaudio.org/forums/index.php?showtopic=24487]samples.[/url] With your excellent hearing, Guruboolez, I'm interested in seeing your comments.

(Edit: added samples link)

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #4
Quote
(...) and many audiophiles still claim that ATRAC Type-R can sound better than CDDA, (...)
[a href="index.php?act=findpost&pid=230086"][{POST_SNAPBACK}][/a]

Never heard an audiophile guy claiming that. By "audiophile", do you have in mind the childish and basic supporters of MD formats used to post their claims and dreams in dedicated board?

P.S. I'm downloading your sample, and test it myself

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #5
Quote
Quote
(...) and many audiophiles still claim that ATRAC Type-R can sound better than CDDA, (...)
[a href="index.php?act=findpost&pid=230086"][{POST_SNAPBACK}][/a]

Never heard an audiophile guy claiming that. By "audiophile", do you have in mind the childish and basic supporters of MD formats used to post their claims and dreams in dedicated board?

P.S. I'm downloading your sample, and test it myself
[a href="index.php?act=findpost&pid=230112"][{POST_SNAPBACK}][/a]


Yes, I mean them, as well as the snake-oil loving, green-marker toting, $1000 cable peddling guys at the local stereo shop. Somebody once claimed that a CD -> MD transfer via S/PDIF, played out of their 24-bit Hybrid Pulse DAC on their MD deck sounded "so much better" than the original CD. It's statements like these that require good listening tests and solid data to debunk. I may not have great ears, but I hope to do some more ABX'ing in the next weeks before school starts again. I haven't done any ABXing in a long time, but forgot how much I enjoy generating results that I think my fellow HA regulars might find useful.

Any suggestions for killer samples? 

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #6
They may like the coloration of ATRAC artifacts?
It's like bass-head peoples

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #7
Quote
They may like the coloration of ATRAC artifacts?
It's like bass-head peoples
[a href="index.php?act=findpost&pid=230124"][{POST_SNAPBACK}][/a]


Oh boy, what do I like? An artifact? Yes! But what music is? Is not it an artifact entirely?
Could I just claim that I correct for loudness because I do not wish to damage my ears with high sound levels? Or could I just claim, that some bass boost restores the loss occured during recording/transmission process?
The only thing I will not claim is that ATRAC is a better format than the original, but I can like whatever fits my taste

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #8
I've compared the reference file to the encoding you've uploaded (thanks). With Replaygain (0.35 dB difference between two files) enabled.
Pre-echo is very low with this encoding (4 first attacks). But before that, the guitar annoys me: smearing was immediately audible. Guitar has something sticky, like "chewing-gum". Lame 320 kbps have a similar problem, but it's really subtle in comparison. Here, ATRAC can't reproduce with precision the original guitar sound.
There's also some harshness in the second part of the file (bunch of castanets attack). It's an artifact I can't hear in lame for exemple.

Code: [Select]
foo_abx v1.2 report
foobar2000 v0.8.3
2004/07/30 13:00:04

File A: file://C:\ABX\castanets.wav
File B: file://C:\AUDIO\3_ CODECS\_atrac\ATRAC-R.flac

13:00:05 : Test started.
13:00:13 : 01/01  50.0%
13:00:24 : 02/02  25.0%
13:00:29 : 03/03  12.5%
13:00:34 : 04/04  6.3%
13:00:39 : 05/05  3.1%
13:00:43 : 06/06  1.6%
13:00:48 : 07/07  0.8%
13:00:52 : 08/08  0.4%
13:00:57 : 09/09  0.2%
13:01:01 : 10/10  0.1%
13:01:07 : 11/11  0.0%
13:01:11 : 12/12  0.0%
13:01:15 : 13/13  0.0%
13:01:20 : 14/14  0.0%
13:01:33 : 15/15  0.0%
13:01:37 : 16/16  0.0%
13:01:41 : 17/17  0.0%
13:01:44 : 18/18  0.0%
13:01:53 : 19/19  0.0%
13:01:56 : 20/20  0.0%
13:01:57 : Test finished.

----------
Total: 20/20 (0.0%)


To be short (comments based on this sample only):
- on sharp and distinct attacks, ATRAC doesn't suffer from pre-echo
- on sharp but dense attacks, ATRAC could introduce some artifacts
- on less precise attacks (guitar with background noise), smearing is annoying.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #9
Quote
To be short (comments based on this sample only):
- on sharp and distinct attacks, ATRAC doesn't suffer from pre-echo
- on sharp but dense attacks, ATRAC could introduce some artifacts
- on less precise attacks (guitar with background noise), smearing is annoying.
[a href="index.php?act=findpost&pid=230290"][{POST_SNAPBACK}][/a]


Thanks for your observations; they seem to corroborate my own, although you certainly have better hearing than I do  I did notice the smearing you mention, though for the latest test I focused on the material between 2 and 3.5 seconds (the "main" section of castanet playing). Thus, I was paying more attention to the precise attacks, as I seem most sensitive to pre-echo in those cases (maybe because it's easy to hear?).

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #10
Anybody know where I may find the original castanets sample?

I would like to use it for testing but have not been able to locate any orignal files that will cause pre-echo.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #11
I can still hear (abx 8/8, easy) pre-echo on the ATRAC-R sample. I just compared it with LAME 3.90.2 --aps, and found pre-echo on the castanets part more audible in the ATRAC-R sample. ABC/HR results:

1L = castanets_mp3_aps.wav
2R = castanets_mds-jb920_digi24f.wav
3L = castanets-ATRAC-R.wav

---------------------------------------
1L File: castanets_mp3_aps.wav
1L Rating: 4.3
1L Comment:
---------------------------------------
2R File: castanets_mds-jb920_digi24f.wav
2R Rating: 3.0
2R Comment:
---------------------------------------
3L File: castanets-ATRAC-R.wav
3L Rating: 4.0
3L Comment:
---------------------------------------

The castanets_mds-jb920_digi24f.wav is old ATRAC 4.5.

I used Sony MDR-7506 headphones.

However, I compared --aps and ATRAC-R yesterday with cheap headphones, and with those, --aps pre-echo was more audible, I guess because the cheap headphones had poorer frequency response (more "thin" sounding).  Note that with those headphones, the ATRAC 4.5 pre-echo didn't sound much worse than ATRAC-R one or --aps ones.

Edit: on my monitor speakers, ATRAC-R pre-echo seems to be even more noticeable than LAME --aps. No ABC/HR, though.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #12
Atrac3plus (Hi-MD) @ 256 kbps would be insteresting too.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #13
Now, I've tried ABXing the first guitar chord at the sample. I could ABX in case of LAME 3.90.2 --aps: 19/25  p=0.7%. But in case of ATRAC-R, I can't latch on any difference. With ATRAC 4.5 sample,I ABXed it 12/14  p=0.6%.

So, I guess quality of ATRAC-R is indeed comparable with that of LAME --aps.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #14
Quote
So, I guess quality of ATRAC-R is indeed comparable with that of LAME --aps.
[a href="index.php?act=findpost&pid=230531"][{POST_SNAPBACK}][/a]


I'm not sure that this would be the case for other samples like Fatboy, creaking, 41_sec, applause, sichia, etc. I plan on testing these soon, but I need to know what went wrong with the analog capture of the samples so we don't have the out-of-phase, different channel volume problem again.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #15
Quote
Anybody know where I may find the original castanets sample?

I would like to use it for testing but have not been able to locate any orignal files that will cause pre-echo.
[{POST_SNAPBACK}][/a]


[a href="http://www.pcabx.com/product/mds-jb920/]PC/ABX[/url] has samples of the original, uncompressed castanets clip and one encoded with ATRAC 4.5, which is quite nasty. I'm glad that I didn't buy a Sony MXD3 4x CD -> MD dubbing deck when I saw a good deal on it; it uses 4.5 rather than Type-R.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #16
I believe it was Sony who made the claim that ATRAC was comparable to CD when ATRAC 3.5 came out.

But ATRAC Type-R is about 5 years old by now.  Maybe we should ABX it against Xing?   

MD was so cool once upon a time (I still have 4 MD players).  Now all I need is my Rio Karma.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #17
Quote
Anybody know where I may find the original castanets sample?
[a href="index.php?act=findpost&pid=230519"][{POST_SNAPBACK}][/a]


I just posted the corresponding original at the ATRAC samples thread.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #18
Quote
I believe it was Sony who made the claim that ATRAC was comparable to CD when ATRAC 3.5 came out.

But ATRAC Type-R is about 5 years old by now.  Maybe we should ABX it against Xing?   

MD was so cool once upon a time (I still have 4 MD players).  Now all I need is my Rio Karma.
[a href="index.php?act=findpost&pid=230659"][{POST_SNAPBACK}][/a]


Actually, it's at least 6 years old...the first Type-R decks started coming out in mid-1998.

As far as Sony's (and minidisc.org's) claims of CD-like sound quality, I don't think that they are all that off. For casual listening with an untrained ear, MD probably is completely indistinguishable from CD. Half the music I've seen on friends' computers is in 64kbps WMA, after all 

Despite all its shortcomings, I still do like MD for certain uses, for the simple fact that it's like having a little digital tape recorder that can record voice, music, etc. on little cheap discs. Maybe I haven't left the 80's yet, but there's still a part of me that is scared to record lectures and important things on flash chips...I like that feeling of having a physical, removable medium. For listening to music, who cares...if my MP3 player's HDD crashes, I have all the music on DVD's.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #19
Well, the last thing I want is to set a flame war, but need to say that ATRAC Type-R recordings sound quite different when you listen to them from the deck (in my case Sony MDS-JB930) and when the ATRAC recording is converted to wavs. It may be due to the DACs used, but from my experience I fooled a lot of people doing blind tests (NAD C320 amplifier and DALI speakers were used). Nobody could get through 6/10 when compared CD with MD... the listeners were musicians and people who earn their bread with music, i.e. technicians, recording engineers and similar.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #20
Quote
Well, the last thing I want is to set a flame war, but need to say that ATRAC Type-R recordings sound quite different when you listen to them from the deck (in my case Sony MDS-JB930)

Care to make a recording through the digital output?

Say, the samples used in the 128 kbps test?

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #21
Moneo: Sorry, but we, who work or used to work professionally with the music tend to listen to music and its impact as a whole, not just "critical samples"... But it doesn't change anything. Mini Disk recordings, especially those from better decks, really sound awfully good when replayed from the equipment, where they were recorded. And that's the point and therefore I think a lot of people, including musicians, love their Mini Disks  And that's why there's a lot of users, who use it even home for listening to music, not only in portable devices.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #22
I used to be able to "fool" many people with my mid-end Type-R deck and good speakers, but that's easily explainable: a) most people, even musicians, are not "trained" to hear artifacts in compressed music, and sometime don't even notice obvious ones (like those of WMA). b) most of the material we play for other people is easy to encode, not killer samples. c) artifacts are best heard over headphones, not speakers. I could never point out pre-echo or anything else wrong with MD until I put on a pair of phones!

Also, I'm not sure that I follow the whole argument that recordings sound better when played by the deck as opposed to being captured by another device. They should sound the same! An encoding coming out of the deck goes through the same DAC, whether you listen to it in your stereo or capture it to wav. Artifacts are artifacts...pre-echo is caused by encoder limitations, not decoder ones.

Pre-echo II: Castanets, ATRAC-R vs. ATRAC 4.5

Reply #23
Hi there, I know that this thread is obsolete but I want to give it a try...

Would there anybody (especially Cygnus X1) be interested in an ABX for Castanets with ATRAC Type-R vs. SHARP ATRAC 8? (respectively 24-Bit ATRAC used on the legendary 1Bit digital amp portables.) I own a IM-DR410 so I could load the Castanets-Sample in SHARP ATRAC 8 up, if anyone is interested. Wish to know which ist the best ATRAC-Encoder for standard-MD recordings (SP.)