IPB

Welcome Guest ( Log In | Register )

2 Pages V  < 1 2  
Reply to this topicStart new topic
ATRAC3 superior or inferior to mp3 ?, Signal Analysis vs Subjective Perception
guruboolez
post Jul 13 2004, 16:02
Post #26





Group: Members (Donating)
Posts: 3474
Joined: 7-November 01
From: Strasbourg (France)
Member No.: 420



No, I don't have any MD player to compare software and hardware encoding. Anyway, there are serious chances that the encoder tested by the laboratory is different from consumers encoders. The potential of ATRAC3 is certainly superior to what we heard during Roberto's 128 test: SonicStage is optimized for high-speed transcoding, and is ~4 time faster than lame --preset abr/cbr on my computer, 2.5 time faster than mpc, and even faster than wma. I suppose that a slower encoder could be better (slightly or much better, I can't say).
Go to the top of the page
+Quote Post
Latexxx
post Jul 13 2004, 16:05
Post #27


A/V Moderator


Group: Members
Posts: 858
Joined: 12-May 03
From: Finland
Member No.: 6557



We should get some samples encoded using some hardware encoders which do it in real-time instead of 100x.
Go to the top of the page
+Quote Post
saratoga
post Jul 13 2004, 16:41
Post #28





Group: Members
Posts: 4715
Joined: 2-September 02
Member No.: 3264



QUOTE (Mystic_Image @ Jul 12 2004, 05:47 PM)
QUOTE (Gabriel @ Jul 12 2004, 03:54 PM)
QUOTE
It is quite right in the sense of a CODEC, but not in the sense I was talking about. I was addressing the fact that ATRAC3 had to be useable on a hardware platform 10 years old, i.e. MiniDisc

Wrong.
ATRAC3 is not usable on older minidisc players.
*



I apologize for my use of the word 'useable' in a different manner, as ATRAC3 is indeed not possible to LISTEN to on older hardware, but if you read my argument, what I am trying to say is that ATRAC3 still has to fit the MiniDisc standard mold.

So even on an old player, an ATRAC3 track *WILL PLAY*, though it will be mute, as opposed to simply being dysfunctional and causing errors or noise.

You do understand what my point is, right?
*



Thats nothing. MP3s have to play on the computer "mold". So while ATRAC may be 10 years old, this obviously makes MP3 over 50 years old laugh.gif

Heck they also play on electronics and devices with batteries. Maybe they're over a hundred years old!
Go to the top of the page
+Quote Post
Mystic_Image
post Jul 14 2004, 01:07
Post #29





Group: Members
Posts: 9
Joined: 9-July 04
Member No.: 15193



Clearly, someone hasn't extrapolated what I thought they would from what I argued... doh.

I obviously haven't communicated my ideas about engineering and design restrictions well enough.

You still seem to be saying something about MP3's being comparable to MiniDisc, claiming that I don't believe it should be done, when I've repeatedly said that such a thing is fine... under the right circumstances.

Quite simply, the design goals/requirements were quite different from most CODECs, and that means you have to take into account these differences in order to make a comparison that is more meaningful (i.e. where only one variable counts).

----

I should be able to get some samples of hardware encoded ATRAC and ATRAC3. I recorded a show I helped out with recently, and I used both modes on my N910 (with ATRAC Type-S). The only problem is trying to get it off the MiniDisc digitally as I don't have a standalone player with digital out.

One other issue we haven't covered yet is that there is apparently a noticeable difference between MD manufacturers... Sharp vs. Sony to be particular. I think they actually use different hardware encoders too...

This post has been edited by Mystic_Image: Jul 14 2004, 01:07
Go to the top of the page
+Quote Post
guruboolez
post Jul 14 2004, 01:11
Post #30





Group: Members (Donating)
Posts: 3474
Joined: 7-November 01
From: Strasbourg (France)
Member No.: 420



QUOTE (Mystic_Image @ Jul 14 2004, 01:07 AM)
The only problem is trying to get it off the MiniDisc digitally as I don't have a standalone player with digital out.
*

It's not a problem here. An analogic capture won't change the lowpass value (background noise will be easy to distinguish from the real musical informations). It's the only thing we need to check (in order to see if the frequency response (FR) showed in the laboratory test is comparable to the FR of the hardware encoder).

This post has been edited by guruboolez: Jul 14 2004, 01:12
Go to the top of the page
+Quote Post
Cygnus X1
post Jul 15 2004, 06:36
Post #31





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



A friend of mine who is quite dull in terms of cognitive functioning (to put it kindly) sold me a basically-unused MZ-N1 portable MD recorder for a mere pittance because he couldn't quite figure out how to use it laugh.gif Although I bought it with the goal of doing some live recording and documenting my diagnostic assessments for transcription (part of my PhD requirements), it dawned on me that with such a high-quality unit, I can record and upload samples of hardware-encoded ATRAC and ATRAC3 files. Since the MZ-N1 is considered by many in the MD community to be one of the "best sounding" units of all time, samples recorded on the unit would carry a bit more objective validity amongst those types (ok, maybe).

Please, I would appreciate fellow HA members to recommend a collection of maybe 20 "killer samples" that I will locate and encode to MD. The MD samples can be recorded using my standalone CD deck or my PowerBook's audio input (which is very quiet). I will then compress them with FLAC. This way, HA can have a permanent database of MD samples, both software and hardware encoded, that will help answer people's questions regarding the quality of ATRAC and ATRAC3 in an unbiased, objective fashion. In the future, I'd possibly even be open to conducting a listening test (probably with Roberto's help wink.gif ), so suggestions are welcome!

This post has been edited by Cygnus X1: Jul 15 2004, 06:39
Go to the top of the page
+Quote Post
Mystic_Image
post Jul 15 2004, 06:38
Post #32





Group: Members
Posts: 9
Joined: 9-July 04
Member No.: 15193



QUOTE (guruboolez @ Jul 13 2004, 07:11 PM)
QUOTE (Mystic_Image @ Jul 14 2004, 01:07 AM)
The only problem is trying to get it off the MiniDisc digitally as I don't have a standalone player with digital out.
*

It's not a problem here. An analogic capture won't change the lowpass value (background noise will be easy to distinguish from the real musical informations). It's the only thing we need to check (in order to see if the frequency response (FR) showed in the laboratory test is comparable to the FR of the hardware encoder).
*




Ok, I've done some preliminary tests using a Sony N910 portable, a Sony JE440 deck, and a Sharp IM-DR80 portable, all using digital-line in from a Denon DVD player (playing a standard CD) as a source and then played back and recorded into my computer at 16 bit, 48KHz on an M-Audio Revolution.

There's definitely something fishy. For all 3 devices, on standard ATRAC modes, the frequency response reaches 20 - 21KHz. However, when using ATRAC3, there's a cutoff around 18KHz.

Looking on Sony's sites, I can't find any MiniDisc device that has a frequency response rated above 20KHz, regardless of whether it is ES (high end) series or not... so something is most likely wrong with the report. The ATRAC3 132kbps frequency response in the report is identical (maybe even BETTER) than the wave file one, and seems to exceed 20KHz by far.
Go to the top of the page
+Quote Post
Cygnus X1
post Jul 15 2004, 06:44
Post #33





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



QUOTE
Ok, I've done some preliminary tests using a Sony N910 portable, a Sony JE440 deck, and a Sharp IM-DR80 portable, all using digital-line in from a Denon DVD player (playing a standard CD) as a source and then played back and recorded into my computer at 16 bit, 48KHz on an M-Audio Revolution.

There's definitely something fishy. For all 3 devices, on standard ATRAC modes, the frequency response reaches 20 - 21KHz. However, when using ATRAC3, there's a cutoff around 18KHz.

Looking on Sony's sites, I can't find any MiniDisc device that has a frequency response rated above 20KHz, regardless of whether it is ES (high end) series or not... so something is most likely wrong with the report. The ATRAC3 132kbps frequency response in the report is identical (maybe even BETTER) than the wave file one, and seems to exceed 20KHz by far.
*


I think I may have a theory on this. I remember reading a review on the MD Community forum in which somebody bubled about the quality of LP4, claiming that it had a "razor-flat" frequency response to 20Khz when using a test tone. However, we all know that when recording music, the upper response is probably around 14Khz at best (17.5Khz for LP2). This leads me to believe that ATRAC3 might use some sort of adaptive lowpass, which probably could be tweaked by Sony to look really nice for an easy sample. Does this make any sense?
Go to the top of the page
+Quote Post
ssamadhi97
post Jul 15 2004, 06:54
Post #34





Group: Developer (Donating)
Posts: 1203
Joined: 10-February 02
From: Endless Water
Member No.: 1305



QUOTE (Cygnus X1 @ Jul 15 2004, 07:36 AM)
Please, I would appreciate fellow HA members to recommend a collection of maybe 20 "killer samples" that I will locate and encode to MD.
*

Try Castanets for a good laugh. I've heard some MD encoded sample once, it was really pre-echo-ridden. Wonder if that's the best MD can do.. at least it was audibly worse than lame --aps


--------------------
A riddle is a short sword attached to the next 2000 years.
Go to the top of the page
+Quote Post
Cygnus X1
post Jul 15 2004, 07:09
Post #35





Group: Members (Donating)
Posts: 676
Joined: 5-June 02
From: New York
Member No.: 2224



QUOTE (ssamadhi97 @ Jul 15 2004, 12:54 AM)
QUOTE (Cygnus X1 @ Jul 15 2004, 07:36 AM)
Please, I would appreciate fellow HA members to recommend a collection of maybe 20 "killer samples" that I will locate and encode to MD.
*

Try Castanets for a good laugh. I've heard some MD encoded sample once, it was really pre-echo-ridden. Wonder if that's the best MD can do.. at least it was audibly worse than lame --aps
*



Yes, that was one of my first ideas...I had tried this a few years back with a ATRAC Type-R MD deck and the results were quite poor. Samples with trumpets or creaking doors or tolling bells also threw it off, IIRC.
Go to the top of the page
+Quote Post
Mystic_Image
post Jul 15 2004, 18:29
Post #36





Group: Members
Posts: 9
Joined: 9-July 04
Member No.: 15193



QUOTE (Cygnus X1 @ Jul 15 2004, 12:44 AM)
QUOTE
Ok, I've done some preliminary tests using a Sony N910 portable, a Sony JE440 deck, and a Sharp IM-DR80 portable, all using digital-line in from a Denon DVD player (playing a standard CD) as a source and then played back and recorded into my computer at 16 bit, 48KHz on an M-Audio Revolution.

There's definitely something fishy. For all 3 devices, on standard ATRAC modes, the frequency response reaches 20 - 21KHz. However, when using ATRAC3, there's a cutoff around 18KHz.

Looking on Sony's sites, I can't find any MiniDisc device that has a frequency response rated above 20KHz, regardless of whether it is ES (high end) series or not... so something is most likely wrong with the report. The ATRAC3 132kbps frequency response in the report is identical (maybe even BETTER) than the wave file one, and seems to exceed 20KHz by far.
*


I think I may have a theory on this. I remember reading a review on the MD Community forum in which somebody bubled about the quality of LP4, claiming that it had a "razor-flat" frequency response to 20Khz when using a test tone. However, we all know that when recording music, the upper response is probably around 14Khz at best (17.5Khz for LP2). This leads me to believe that ATRAC3 might use some sort of adaptive lowpass, which probably could be tweaked by Sony to look really nice for an easy sample. Does this make any sense?
*



It does makes sense. It shouldn't be too hard to detect extreme or limit cases where a low-pass should be changed. Still, I wouldn't expect it to go up to 22KHz when regular ATRAC won't do that either.

I'm generally interested in seeing comparisons between LAME, Fraunhofer and ATRAC3 on the same samples.
Go to the top of the page
+Quote Post
MugFunky
post Jul 15 2004, 19:07
Post #37





Group: Members
Posts: 493
Joined: 28-December 03
From: Melbourne, Aus
Member No.: 10767



QUOTE
Wow ! I would never have noticed this ! Are you sure that ATRAC3 lowpasses everything at 18 kHz ? Couldn't it be an apparent lowpass caused by the ATH, like the apparent 16 kHz lowpass of MP3 APS -Y ?


from what i read in the above links, an apparent lowpass would only be observed at 11025 Hz if it were the same kind of lowpass seen in mp3 (ie. occuring on a major band boundary). such lowpasses are seen in ATRAC with speech, but i haven't noticed them with ATRAC3 in most cases.

however, it's quite possible that pure tones would have a higher resolution in ATRAC/ATRAC3 as there is no transient variations to speak of, and a pure tone is an ideal situation for something like ATRAC. whether sony tweaked their encoder to deal separately with each piece of music could probably only be determined by getting hold of the original samples (preferably not decoded to PCM) and doing some serious comparisons with commercially available encoders.

hmmm... after reading that paper above, it's no wonder speech comes out all grotty in ATRAC - the bit allocation described there made me sick to my stomach.

This post has been edited by MugFunky: Jul 15 2004, 19:09
Go to the top of the page
+Quote Post
Latexxx
post Jul 15 2004, 21:26
Post #38


A/V Moderator


Group: Members
Posts: 858
Joined: 12-May 03
From: Finland
Member No.: 6557



QUOTE (MugFunky @ Jul 15 2004, 08:07 PM)
hmmm... after reading that paper above, it's no wonder speech comes out all grotty in ATRAC - the bit allocation described there made me sick to my stomach.
*

Hmm, the paper (http://www.minidisc.org/aes_atrac.html), which dates back to 1992, says: "ATRAC does not specify a bit allocation algorithm; any appropriate algorithm may be used." They might have figured out something during these 12 years (or not biggrin.gif ).
Go to the top of the page
+Quote Post
MugFunky
post Jul 16 2004, 16:13
Post #39





Group: Members
Posts: 493
Joined: 28-December 03
From: Melbourne, Aus
Member No.: 10767



hehe. from the sound of some ATRAC i've heard, and from reading the article, it seems bit-allocation is the culprit... i suppose a simple (bad) algo means cheaper manufacturing costs.
Go to the top of the page
+Quote Post

2 Pages V  < 1 2
Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 17th April 2014 - 00:18