Theory About Atrac Format

[Mr Sim]

A friend of mine told me that the atrac format of music is simple a .wave file where any sound that cannot be heard by the human earis simply cut out.

and that is how they save on so much file size and still retain such a high quality

My question is:

1. Is this all true?

2. Is this theory kind of close?

or

3. Do I have to find this guy and put him back in a cage at the Zoo because he's to wacky to be aloud to live in our current society?

[Fast Eddie]

Well that's the rough jist of any compressed sound format, it trys to remove the bulk of what is not heard.

[Guest tony wong]

Well that's the rough jist of any compressed sound format, it trys to remove the bulk of what is not heard.

←

that's not true

most of the compression will be done on patterns rather than the parts of the sound u won't be able to hear

only for mp3 it is true

compression itself doesn't necessary mean lossy

there is both loss-less and lossy compression technic in this world

[Fast Eddie]

That's a bit of a crazy sig you've got there

Forget Hi-MD, the equipment except the highest model, is too bulky

Equally strange.

Edited February 11, 2005 by Fast Eddie

[dex Otaku]

All lossy audio compression formats use basically the same set of techniques and theories to do their job.

The techniques incorporate established research done on how human hearing works, such as the absolute threshold of hearing and multiple forms of masking [one sound obscuring another].

There are many variations on how to use these theories, and many ways to implement them, such as using different mathematical techniques [mostly variations in discrete transforms/fourier analysis].

In all cases, lossy audio compression works by analysing the incoming signal and basically throwing away parts that are not likely to be missed - the parts people aren't likely to hear anyway.

Distortion [artifacts] occur when signals get too complex for the resolution of the encoder, such as when multiple instruments with complex harmonic overtones are playing at the same time.

WAV audio is uncompressed PCM. PCM data is not even of the same domain as most lossily-compressed formats; PCM is amplitude vs. time, whereas the majority of lossy formats store data in some variant of amplitude vs. frequency.

[Guest tony wong]

Well that's the rough jist of any compressed sound format, it trys to remove the bulk of what is not heard.

←

I would make a 2nd comment, that,

ur words only applied to wave format

anything missing from wave format(PCM) should be hearable for human

that is, it is not not heard

I do know why ATRAC exist is mainly for its algrothm, and not "cutting out the part that won't be noticeable"

that's why people will choose to use MD instead of mp3

purely because mp3 compression will cut out many noticeable parts from the music

for what u've said, may fits more to mp3

but not md(ATRAC)

[dex Otaku]

anything missing from wave format(PCM) should be hearable for human

At least 60 years of psychoacoustic studies prove you wrong on this.

I do know why ATRAC exist is mainly for its algrothm, and not "cutting out the part that won't be noticeable"

that's why people will choose to use MD instead of mp3

purely because mp3 compression will cut out many noticeable parts from the music

mp3 and all variations of ATRAC use the same principles to do the same job. Reread my previous reply [above].

Perhaps it's too technical, I don't know, but you're completely incorrect on this.

[Guest tony wong]

At least 60 years of psychoacoustic studies prove you wrong on this.

mp3 and all variations of ATRAC use the same principles to do the same job.  Reread my previous reply [above]. 

Perhaps it's too technical, I don't know, but you're completely incorrect on this.

←

unnoticeable is totally different from "not heard"

[dex Otaku]

unnoticeable is totally different from "not heard"

←

This is a specious semantic argument, but in any case, the point was that ATRAC and mp3 operate on the same basis - data reduction based on perceptual coding - which throws out data that is not likely to be perceived.

[Guest tony wong]

This is a specious semantic argument, but in any case, the point was that ATRAC and mp3 operate on the same basis - data reduction based on perceptual coding - which throws out data that is not likely to be perceived.

←

totally wrong

ATRAC is basically focusing more on advanced technic on pure compression

where sound quality is always the first thing to concern

where in contrast, mp3 is only a side product of MPEG

its sim is to provide sound for the MPEG video

sound quality have never been a concern

as, they always like to do what u have mentioned, "cut out parts of the sound that is not heard"

and, yes,

these are not "not heard"

just not that noticeable

[dex Otaku]

totally wrong

(sighs heavily)

Look, Mr. Wong. I'm honestly not looking to start a flamewar with you. What I am looking to do is make sure that the information presented on this forum is as accurate as possible, within the limits of my knowledge. That said - if you're so interested in proving your points, why not try doing some research first? As only a couple of suggestions: Wikipedia has excellent entries on the various MPEG formats and layers, as well as ATRAC; minidisc.org's links and research sections have pointers to plenty of information about the various flavours of ATRAC.

Please feel free to do some reading to back up your claims, rather than flippantly misinforming your fellow forum members.

ATRAC is basically focusing more on advanced technic on pure compression where sound quality is always the first thing to concern

Indeed it is. As it is with almost all of the other codecs and compression formats out there. Whether the developers succeed in their goal is another matter.

See further down for my thoughts on "sound quality".

where in contrast, mp3 is only a side product of MPEG

Categorically incorrect. MP3 came about as a result of combining previous research on perceptual data reduction that was being developed initially as an offering for digital broadcast radio.

Their goal was to achieve similar quality at 128kbps as layer II audio [musicam] did at 192kbps. Whether they managed to do this is a matter of personal opinion, but that was their goal. Considering the fact that what became layer II and layer III were developed in the late 80s and early 90s, I think they did an admirable job.

its sim is to provide sound for the MPEG video

Also categorically incorrect. As I said, what became MP3 was originally developed with the intent of being used for digital radio. MPEG adopted and ratified FhG's codec as layer III for MPEG-1 because it fit the group's purposes.

The format did not originally have anything specifically to do with either video or MPEG. It's inclusion in the MPEG spec is incidental to its original reasons for being.

ATRAC itself was introduced at almost the same time for both MD and as used for the optical film soundtrack format, SDDS.

Does that mean ATRAC was made for film soundtracks? No.

Further - to the best of my knowledge, MPEG does not work by commisioning encoding formats.

MPEG is a group that ratifies standards. They do not make anything, per se. They simply bring together a large group of media engineers to assess formats for inclusion in the standards they propose to the rest of the world. It's up to the rest of the world [or, at least, major media distributors and equipment makers] to make up their own minds whether they want to listen to MPEG or not.

sound quality have never been a concern

If sound quality had never been a concern, they never would have started developing it in the first place. As goes with every other codec. Again, whether they achieve their goal is another matter, but their intent is this:

Create an an algorithm to fit a specific sound quality expectation and bandwidth requirement. This need not [and often is not] be attached to any particular physical medium [such as minidisc, CD, DVD, &c.] or other information type [such as film or video].

I already stated what their goal was, so I won't repeat it.

The other half of this is the use of "sound quality". Particular compression formats are usually made to compromise between sound quality and bandwidth; part of the development cycle is to figure out what kind of quality is their actual goal.

The original ATRAC at 292kbps was not significantly better than either layer II or layer III audio [at their target bitrates for interleaving with video], and the goal for all three was as I stated above: to achieve the desired quality within a particular bandwidth requirement.

I would assert [though this is purely opinion] that ATRAC, MP2, and MP3 all originally had very close to the same goals in terms of *desired* fidelity.

In the case of both MP3 and ATRAC, their encoding algorithms have been tuned and retuned over the years to improve quality quite dramatically. I would agree that ATRAC at 292kbps exceeds MP3 at 128kbps, but that's to be expected considering the limitations imposed on both [by their decoders] and the large difference in bandwidth. Many would argue that today's MP3 encoders at 256kbps or greater are better than ATRAC, however. This is a matter of opinion, though.

as, they always like to do what u have mentioned, "cut out parts of the sound that is not heard"

and, yes,

these are not "not heard"

just not that noticeable

Back to your semantic argument, I see.

Technically, as I already digressed in the previous reply, to 'hear' is for sound to be transduced by your ear and received by your brain. Whether your brain is able to perceive every part of the sound is another matter, hence my using 'perceived' back there.

Most people do not make that distinction about the word 'hear' though. That is why I used the word in the first place.

There is a reason they are both called perceptual coders, but I've already stated what that was, so I'll let it rest now.

[Fast Eddie]

Mr Wong, it's obvious your're wrong here. Accept defeat beofe you show yourself up any further!

[dex Otaku]

An interesting side-note after further reading: it seems that the people who did the work for FhG on what became the MPEG codecs were a part of the MPEG group. So - while I'd still say that MPEG don't actually make anything themselves, it's important to remember that their members do.

[Guest tony wong]

a simple question can end up the discussion :

do u think mp3@320 can beat ATRAC(version 1.0)?

[Guest tony wong]

An interesting side-note after further reading: it seems that the people who did the work for FhG on what became the MPEG codecs were a part of the MPEG group.  So - while I'd still say that MPEG don't actually make anything themselves, it's important to remember that their members do.

←

that's the reason for the bad quality of such xPEG product

[dex Otaku]

a simple question can end up the discussion :

do u think mp3@320 can beat ATRAC(version 1.0)?

←

This is a matter of personal opinion, and as such my answer to it isn't definitive in any way. Neither would yours be.

Still: properly-encoded MP3 does have certain advantages over ATRAC in basically all of its variations, including SP. In a measurable sense, MP3 has fewer problems with ringing and pre-echo, which all ATRAC versions have difficulty with due to limitations imposed on it [related in part to ATRAC's requirements of editing and gapless playback - features I think offset the issue in a way that almost cancels it out].

ATRAC SP [v3.5 and newer, I believe] has the advantage of higher apparent bit-depth on decoding, so it would probably have a higher measurable dynamic range from the right source material. [Note that as transform-related formats, neither has a bit-depth in any conventional, PCM-related sense].

I consider MP3 at 320kbps transparent with most material in the same way that ATRAC is. If forced to choose between the ATRAC SP and MP3@256 or 320kbps, I would choose MP3 100% of the time simply because of its wide compatibility. ATRAC is a closed system that is not supported outside of a limited range of dedicated hardware, making it next to useless in a true portability sense.

that's the reason for the bad quality of such xPEG product

Evidently you really have something again the MPEG. Really though, this is no different than Head-Fi dogma #7, or saying that "ATRAC sucks because it was made by Sony."

It may be your opinion, but it doesn't amount to anything more than that [just as my and anyone else's opinion doesn't, either].

Also, again: please consider using the EDIT feature to add to your posts. Also, please consider shortening your signature, or at least removing the photo from it. Every one-liner you post is huge in screenlength.

[bug80]

dex Otaku is right.

By the way, recent listening tests point out that latest versions of the OGG, MP3, AAC, MPC and AAC algorithms all beat ATRAC in terms of sound quality at 128 kb/s.

However, they haven't tested the built-in encoders of MD units. Too bad, because I heard those encoders perform better than the encoder that comes with SonicStage. Why don't they use that algorithm in SonicStage anyway? Is it because of speed? I don't mind to wait a little longer for encoding when the resulting ATRAC files will sound way better!

[ROMBUSTERS]

Back to the original question in the very first post (in case its answer got missed during this giant mess):

an uncompressed PCM/WAV file is what you would find on a CD

to compress this file every lossy (MP3/Atrac/OGG/etc) compression system strips out portions of the audio at high frequencies that re not considered audible to the human ear (i.e. you shouldnt miss them). Also the encoders stress louder noises more than quieter ones because again you shouldn't miss them. In this way they can loose (lossy) a lot of audio info and shrink the file size down.

The next step is to employ mathematical formulas to compress the data that is left over even further. This is done in a similar manner to compression system that are used for data files on a PC (although never the exact same). Taking the audio data to binary you have sequences of 1s and 0s. One of the ways (i believe reading on a sony website) that Atrac compresses the audio is that for every 2 adjacent 1s they subsitute for just a single 1. For example:

Original sequence: 10011110011110

Compressed sequence: 1001100110

Obviously it is more complicated then that but you get the general idea.

This along with many other means is what actually compresses the original wave file to the compressed format.

The various codecs like MP3 and Atrac and Atrac3+ even, take different steps and ways of compressing the information which is why an MP3 at 256kbps will end up being a different file size and sounding (slightly) different then Atrac3+ at 256kbps.

However, they haven't tested the built-in encoders of MD units. Too bad, because I heard those encoders perform better than the encoder that comes with SonicStage. Why don't they use that algorithm in SonicStage anyway? Is it because of speed? I don't mind to wait a little longer for encoding when the resulting ATRAC files will sound way better!

←

the old theory used to be of speed but no one is really sure anymore

[aeriyn]

It's pretty much a general consensus that the encoders used by SonicStage are not nearly as efficient as the ATRAC ICs in MD recorders. I don't think it has to do so much with speed as with Sony's general incompetence when it comes to making computer software.

Note: Sound Forge was not made by Sony; it was made by the Sonic Foundry people. Sony took over the program and did some things that most long-time Sound Forge users didn't especially like. Sound Forge is exempt from the ideal that Sony isn't very good at making software products because Sony didn't make it.

[atrain]

on the other hand instead of making crappy software they bought out a dedicated creator in steinberg.

have to say i'm quite impressed with sony spinning off it's software writing for md/nw to the connect company. lets hope that a perspective from outside the corprate bunker provides more intuitive software, eh?

[Guest tony wong]

I would like to point out one thing :

wav itself is already selected range of so called "hearable sound"

the arguments I replied to Fast Eddie was only to this

and I only wished anyone can email to Sony, ask how many percent of ATRAC is done on lossless compression

how many percent is done on "cutting out of unhearable sound"

that would answer this thread's primary question over ATRAC compression

[aeriyn]

Actually, the thing about PCM is that it's only recording perceived audible frequency response... i.e. 20-20,000Hz, which is techically not the entire audible frequency range, hence the existence of high-rez formats like SACD and DVD-A.

But... there are many, many things that PCM picks up that are either 1) inaudible to most humans or 2) won't be noticed or missed when removed or 3) both. Psychoacoustic models like the ones used for ATRAC and other lossless encoders try to determine which sounds we can hear, which ones we can't and which ones we won't miss when they're gone. So PCM is not always what we can hear, or notice if it's missing or not.

In that aspect ATRAC is just like MP3 or AAC or vorbis or MPC or MP2 or AC3 or any other of the countless lossy encoding methods out there. ATRAC has advantages and disadvantages like any other audio compression codec. Primary advantages include nonlinear editing and gapless playback, but it suffers from pre-echo issues as dex pointed out earlier. It's a slightly better low (128kbps) bitrate encoder than MP3 and a far better very low bitrate (~64kbps) encoder than MP3. Largely, for general purpose use however, MP3 is better simply because nearly anything can use it.

ATRAC isn't a lossless codec and no part of it is lossless because it uses these psychoacoustic models to remove pieces of the signal (hence the term lossy).

[Guest tony wong]

But... there are many, many things that PCM picks up that are either 1) inaudible to most humans or 2) won't be noticed or missed when removed or 3) both. Psychoacoustic models like the ones used for ATRAC and other lossless encoders try to determine which sounds we can hear, which ones we can't and which ones we won't miss when they're gone. So PCM is not always what we can hear, or notice if it's missing or not.

yes, that's correct

so I use the word "noticeable" in the first place

and by the way....

see u have mentioned about dvd audio and sacd.....

they really extend the range?

[aeriyn]

see u have mentioned about dvd audio and sacd.....

they really extend the range?

←

Well, yes, the formats support higher resolutions and bit depths, in the sense of DVD-A anyhow. Super Audio CD is actually a 1-bit system, Direct Stream Digital that has a sampling rate up in the MHz range, while DVD-A is 24-bit, 192kHz resolution. So yes, the maximum effective dynamic range and frequency responses of these different formats can be much greater than that of old-school Redbook Audio (16 bit 44.1kHz resolution).

[streaml1ne]

yes, that's correct

so I use the word "noticeable" in the first place

and by the way....

see u have mentioned about dvd audio and sacd.....

they really extend the range?

←

http://www.daisy-laser.com/products/SACD/sacd.htm

http://www.digitalaudioguide.com/faq/dvd-audio/faq_intro.htm

Very much so...

[Guest NRen2k5]

a simple question can end up the discussion :

do u think mp3@320 can beat ATRAC(version 1.0)?

←

With ease. It would not even be necessary to perform a blind test.

ATRAC 1.0 was terrible. Even 132kbps "LP2" ATRAC3 will perform better than ATRAC 1.0 on most material.

[Guest NRen2k5]

Well, yes, the formats support higher resolutions and bit depths, in the sense of DVD-A anyhow. Super Audio CD is actually a 1-bit system, Direct Stream Digital that has a sampling rate up in the MHz range, while DVD-A is 24-bit, 192kHz resolution. So yes, the maximum effective dynamic range and frequency responses of these different formats can be much greater than that of old-school Redbook Audio (16 bit 44.1kHz resolution).

←

Yup.

If I recall correctly, though, CD-audio has a resolution of 44.1kHz and a range of 22.05kHz... DVD-A has several possible resolutions and a range of 24kHz... SACD's resolution is theoretically 1MHz and I can only imagine what its range is.

[bug80]

Yup.

If I recall correctly, though, CD-audio has a resolution of 44.1kHz and a range of 22.05kHz... DVD-A has several possible resolutions and a range of 24kHz... SACD's resolution is theoretically 1MHz and I can only imagine what its range is.

←

I hope I'm correct:

CD-Audio: Frequency range 0 - 22.05 kHz, dynamics 96 dB (16 bits per sample)

DVD-A: Freq range 0 - 96 kHz, dynamics 144 dB (24 bit)

SACD: Freq range 0 - 100 kHz, dynamics 120 dB (equivalent to 20 bit)

Remarks:

* I believe the standard frequency range for CD-audio is 5 - 22050 kHz

* The sampling frequency is twice the nyquist frequency for CD and DVD-A, and 2.8 MHz for SACD

* Because SACD uses 2.8 MHz / 1 bit sampling, I use the term equivalent to 20 bit.

* The values for DVD-A are the maximum possible values

more info on DVD-A

whitepaper on SACD

Edited February 22, 2005 by bug80

[Guest tony wong]

personally, I do say there is no equivalent for DSD to PCM

whereas, DSD emphasize very much on recording in a DSD studio

that is, it is a bit system all along the way from recording studio to ur home audio

DSD itself is "no modulation"

the primary recording, mothertape, is already the SACD