Atrac Pros/cons

[ROMBUSTERS]

Post what you like/dislike about and how it sounds, works, etc. DO NOT post about how much you hate having to convert your music, etc (however noting on the varying conversion times for the different CODECs can be done).

PROS:

-battery life: atrac uses less juice on average then MP3 or other compression systems and is thus able to allow for longer playback

-designed for music: atrac is basically the only compression system designed from the ground up for music playback

-constant improvements: the atrac encoders are improving not only with each new generation of hardware, but on the software side of things as well

CONS:

-propietary: sony has locked it down in their licensing system and it has not even made a scratch on freely available codecs

-DRM: music, locked in a box, lock in a vault pretty much sums of what you get when you use a propietary format with DRM

[bug80]

You pretty much summed it all up, I think.

For me the greatest pro is the soundquality. ATRAC 132 kbs is definitely more pleasant to listen to than MP3 128 kbs (and 160 kbs, with most audio).

Second comes the battery life. I think it's great my MD player continues playing for about 16 hours in LP2 mode (48 hours using a dry battery), while iPod owners have to recharge their unit every 7-8 hours

Something you didn't mention is, that ATRAC is exclusive to Sony has a great advantage. Because MP3 is an algorithm used by lots of software engineers, a lot of bad encoders exist on the market. For example encoders which don't use appropiate pre-processing before the actual encoding. This way, you never know if your downloaded mp3 (legal download, off course ) with a bitrate of 192 kbs sounds good, or sounds like crap.

[streaml1ne]

For me the greatest pro is the soundquality. ATRAC 132 kbs is definitely more pleasant to listen to than MP3 128 kbs (and 160 kbs, with most audio).

←

This is what baffles me about MP3 support on any Sony gear. Aside from the convenience of no conversion uploads what's the point. Bit for bit ATRAC beats MP3, to my ear at least, even using a good MP3 encoder like lame.

[ROMBUSTERS]

This is what baffles me about MP3 support on any Sony gear. Aside from the convenience of no conversion uploads what's the point. Bit for bit ATRAC beats MP3, to my ear at least, even using a good MP3 encoder like lame.

←

because people know wut MP3 is. if you went and asked Joe 40-year old about Atrac he'd probably mention the old 8-track

MP3 has the name and the standardization (although thats a joke as well)

[bug80]

if you went and asked Joe 40-year old about Atrac he'd probably mention the old 8-track

←

And he's probably right too.

[ROMBUSTERS]

i like the fact that Atrac can be altered and changed and yet still be backwards compatible. For example my old N505 netMD player can playback Atrac3 105kbps files. Not a big feat but a rate not mention whatsoever at the time of the units production.

Now if sony would just let us pick our own bit rates that'd be great

[bug80]

i like the fact that Atrac can be altered and changed and yet still be backwards compatible. For example my old N505 netMD player can playback Atrac3 105kbps files. Not a big feat but a rate not mention whatsoever at the time of the units production.

Now if sony would just let us pick our own bit rates that'd be great

←

Yes that would be great. I guess ATRAC 192 kbs will sound like MP3 256 kbs, or at least close to that.

[Damage]

You forget:

ATRAC wasn't originally designed with Battery Life in mind, but real time digital recording (along with DAT's PASC, which is now MP2 audio). It's only recently (Last two or three generation players) that Sony has looked into improving battery life. However, much of that, seemingly, comes from the cheaper and weaker AMP along with DSP improvements.

I'd say, overall, ATRAC's advantage came from real time digital recording first.

[ROMBUSTERS]

thats true

even to this day MD is basically your only real option for recording into a compressed format for the average consumer

[mmp64]

Good: It's gapless.

Bad: Can't normalize.

[1kyle]

because people know wut MP3 is. if you went and asked Joe 40-year old about Atrac he'd probably mention the old 8-track

Now Now then

I'm over 40 and can use computers / digital cameras / recording equipment / musical instruments etc probably a lot better than some younger members (and also have the money available to spend on decent equipment as well).

As far as 8 Track is concerned I actually remember these and didn't like them even then.

Ipods today IMO are "The 8 track" equivalent in todays terms --relatively inflexible, expensive, fairly easy to break and mark one out as either being a Nerd or a Geek.

A nice MH1 has a much better "Kewl" factor for any age group and ATRAC3 Plus --provided you don't get too hassled with Sony's almost paranoidal attempts at preventing digital copies beats the pants off any MP3.

Believe me Ipods are "Dinosaur Technology" --once some decent High end separates appear as well as HI-MD car radios the Ipod will be in terminal decline.

Perhaps you should read this thread

http://forums.minidisc.org/index.php?showtopic=7811

Cheers

-K

Edited January 22, 2005 by 1kyle

[nickjr]

<=========LOL, the first time I heard Atrac.......I Thought 8 track (36 yrs old)

[dex Otaku]

Lots of stuff on this topic.. but first, to dispel some myths:

1) atrac uses less power than mp3:

This is only an illusion. The amount of signal processing used to play back either signal will depend mostly on the bitrate, and will be pretty much equal between the two.

The place where you really save power is during playback of lower bitrates. A lower bitrate recording doesn't need to be accessed as fast or as often in order to be buffered for decoding - a simple example is that a single disc access at a fixed speed will read more of a 48kbps track than of a 256kbps track. This means you can read a whole bunch of 48kbps audio into the decoding buffer, shut off the disc, and then do everything from the buffer. With the disc not running, less power is consumed, so battery life is longer.

atrac/3/plus would seem to contradict this in one sense: there are more sub-bands and a higher resolution transform is used, so it would make sense that it takes more processing to decode it back to PCM than a lower-resolution format [like atrac or atrac3, or mp3] would take. More processing means more power consumption.

If the DSP is optimised specifically for atrac/3/plus it could counteract this, though. Comparing atrac/3/plus with mp3, though - I would expect that mp3 decodnig would be less processing-intensive to decode, meaning that even with format-specific decoder optimisations, atrac/3/plus might use the same amount of processing power as mp3 in the end.

2) ATRAC is more flexible because it can be altered while remaining backward-compatible.

This is also an illusion. As with literally -any- codec, the decoder is basically a standard set in stone at it s most basic level. The encoder, on the other hand, can be tuned over time to improve performance while maintaining a stream that the decoder doesn't know any different from one made by an older encoder.

Mp3 is probably the best example of this, actually. Because of the widespread adoption of the format, it has been taken and tuned to the point where it can far exceed its original specs. Lame and the like have done an excellent job of this, parallelling in many ways some of the many improvements made over the years to ATRAC SP as well as MDLP.

Enough for the myths, though.

Measurable differences between formats [such at all of the atrac formats' kown-distinct problems with pre-echo] do exist, and these are a concrete, objective thing. These can determine what formats are technically more accurate, though it is important to note that accuracy is not always what definitively says what is better to our ears.

As far as atrac/3/plus [all of the variations, and in fact, the same applies to all codecs] are concerned, "better sound" is mostly a matter of personal opinion, based on perception and personal preference. This is subjective, and as such, does not actually determine whether atrac/3/plus are better formats than anything else.

What it amounts to is that you can tune the encoder of any format as long as the stream it spits out is something that the decoder can recognise as valid. You can improve the encoder as much as you like, within the limits of what the decoder expects, and even older decoders will be able to play back the newer-version streams with improved fidelity.

One more thing - with the disclaimer that I am not an expert an audio encoding, so some of this might be incorrect even if it makes sense to me in terms of all the studying I've done on the subject.

The gapless nature of MD and HiMD are simply a matter of design intent. MD was made so that packets of audio matched those of CD, if I'm not mistaken - using a fixed length of samples encoded at a constant bitrate [CBR] for every block. Without doing this, both on-disc editing and gapless playback after editing would not have been possible.

It's important to note that gapless playback between tracks recorded in one go and gapless playback between tracks that have been edited are two different things.

Gapless playback in general requires that every decoded block end with a valid sample rather than padding data [as often happens with mp3 files].

Gapless playback on an editable format requires that every block have exactly the same length of samples. This may seem meaningless to many of you, but it's crucial to MD's editing capabilities. It's also a big part of the problem with atrac's pre-echo problems, since eliminating pre-echo requires encoding high transients with a different block length to make sure that the overlap between decoded blocks isn't so long that the echo is long enough to be obvious.

This is why a ripped CD will be gapless, while transcoded mp3s can not be.

Other than the original ATRAC codecs as implemented on MD, I don't know of any other audio codec than was designed specifically for gapless playback; the combination of variable block-lengths and variable bitrate encoding get in the way, for instance.

There is no actual reason why mp3 and other formats can't be gapless; all that's required is that everyone follow exactly the same standard, and for block lengths [transform window-lengths, really] always be constant.

The part about gapless playback that always gets me is that -all- players have to buffer the encoded data before decoding them for playback. All that's really required for gapless playback is that that playback buffer [not decoding buffer] be monitored for zero samples [padding], for those zeros to be tossed out, and for the playback buffering to seemlessly go from one track to the next. This is really simple to do [though it is processing intensive in a sense], and there's no reason why it shouldn't have been an option on every mp3 player in existence right from the beginning of their manufacture.

Maybe I'm oversimplifying, though. It's mostly a matter of engineering or design philosophy.

Most mp3 players work this way:

* start buffering the encoded stream

* decode the stream [play it back]

* when the end of the stream is reached, stop [i.e. turn off the decoder]

* start over again

whereas what I'm talking about requires this instead [which in theory works regardless of variable block or window lengths or VBR encoding]:

* ask the stream how long it is

* start buffering the stream

* decode/play back the stream

* when the stream reaches x blocks from its end, start monitoring the decoded data for padding [zeros]

* at the same time, start buffering the next stream [basically start from the beginning of this list again]

* remove zero samples from the end of the first stream and the beginning of the next stream

This may not make sense because it doesn't seem to have an end, but that's really the point - gapless playback never stops or turns off the decoder. It's constantly decoding, and runs in a sort of endless loop, rather than as a process that starts and ends discretely for every track.

This is complicated by many things, though. First and foremost is that metadata, specifically id3 tags in mp3s, are not always the same length, because there are no fixed standards for this. This means that the decoder has to be able to recognise what is actually audio data or not, which may seem pretty simple, but a lot of players simple don't bother to do this - hence the little glitches at the beginning and end of literally every track played back on most cheap players.

Add to this the fact that id3 v2 tags are placed at the beginning of every tracc, whereas v1 tags are at its end. Tags at the end can more easily be whatever length, because when the decoder reaches the beginning of the tag it simply ignores the rest. Tags placed at the beginning, however, have to be monitored until they end and the audio stream is reached. This is one of the reasons why there is actually an anti-id3 v2.x movement that has been around for a long time.

And. Yeah. There's my $0.02.

[ROMBUSTERS]

right the major misconception is that Atrac improves battery life in general, however because Atrac usually runs at lower bitrates then its MP3 equivelent (i.e. 132/105/ even 256 which is high but not the max by MP3 standards) it still does conserve battery power (especially when you consider decent sounding MP3s start at 192kbps)

[aeriyn]

right the major misconception is that Atrac improves battery life in general, however because Atrac usually runs at lower bitrates then its MP3 equivelent (i.e. 132/105/ even 256 which is high but not the max by MP3 standards) it still does conserve battery power (especially when you consider decent sounding MP3s start at 192kbps)

←

This is still subjective; I don't consider 132kbps ATRAC3 to be any better than LAME CBR 128kbps MP3. Both of them aren't good enough for me anymore (e.g. my equipment has gotten too picky as of late).

That's another thing it depends on. LP2 mode might be acceptible and hard to tell from Hi-SP with Sony EX71s unamped, but with Shure E5c amped, you better believe you can tell the difference. E5s aren't especially forgiving and amping them makes them even less so.

[1kyle]

This is still subjective; I don't consider 132kbps ATRAC3 to be any better than LAME CBR 128kbps MP3. Both of them aren't good enough for me anymore (e.g. my equipment has gotten too picky as of late).

That's another thing it depends on. LP2 mode might be acceptible and hard to tell from Hi-SP with Sony EX71s unamped, but with Shure E5c amped, you better believe you can tell the difference. E5s aren't especially forgiving and amping them makes them even less so.

←

PCM probably is the way to go but unfortunately most Car radios etc don't have facilities for playing this unless you plug in via a separate Line In --a hassle.

MD-LP2 mode is ususally easily good enouigh for the inside of sonewhere like a Car (and I have a decent BMW as well) althought HI-SP will be great when car radios can play it.

I never liked CD's in cars anyway --and am hoping that my current radio will last until a new unit comes out -- I'm surprised at the difficulty of finding card radios with Minidisc playing capability -- I have the old Kenwood KMD 673-R.

Another case of where Sony's marketing has been less than ideal --a copuple of minidiscs in your pocket is much better than fiddling around with CD's and stacking them in a changer in the boot (trunk for US guys) of your car.

As far as MP3 vs ATRAC goes -- LAME is absolutely HORRIBLE compared with HI-SP or even MD LP2 mode --believe me I've had this stuff on Oscilliscopes and the MP3 stuff is VERY ARTIFACTY. LP2 comes out quite cleanly surprisingly enough, HI-SP is around 95 -98% as good as PCM and unless you have top rate equipment (and I mean TOP RATE) and a good amped set of cans with the ears to match you probably won't be able to hear any difference except inside the walls of a first class professional recording studio.

For the "More Technically Challenged" -- an Oscilliscope essentially is a machine with a screen on it where you can display the sound wave and measure various characteristics. A sound wave is a complex combination of Sine Waves which shouldn't have any "Jaggies" or "artifacts"

Now even PCM can have some problems as PCM is pure digital -- and in order to get this into a form that our Ears can actually hear it has to be converted into analog Sine waves. There's complex mathematics called Fourier Ananlysis which shows how this is done -- but it's not a straight forward process. The hardware is called an A/D or Analog to Digital converter --it's simply doing "The Maths" in hardware. If the quality of the A/D converter is bad then you are still up the creek.

Now converting MP3's which are full of "noise and errors" the poor old A/D converter has a really hard time which is why you will never ever see MP3's being used in high end professional recording studios other than used for churning out relatively low level "consumer grade" quality material).

I'm not trying to flame anybody here --the MP3 format has done wonders for "portable music" and in general is sufficient for "The Masses" just as Macdonalds is too --but if you prefer French Gourmet you won't go to Macdonalds --same with sound --for me MP3 currently at any bit rate doesn't have it --but if YOU like it then good luck --in this game the only thing that matters ARE YOUR OWN EARS.

Cheers and happy listening.

-K

Edited January 23, 2005 by 1kyle

[bug80]

whereas what I'm talking about requires this instead [which in theory works regardless of variable block or window lengths or VBR encoding]:

* ask the stream how long it is

* start buffering the stream

* decode/play back the stream

* when the stream reaches x blocks from its end, start monitoring the decoded data for padding [zeros]

* at the same time, start buffering the next stream [basically start from the beginning of this list again]

* remove zero samples from the end of the first stream and the beginning of the next stream

←

Good story. To be honest, I really never thought of ATRAC being gapless (but in the end it all perfectly makes sense).Your proposed method of making MP3 players gapless is a good idea, however you will run into problems. As far as I know, all D/A converters depend on a fixed bufferlength in the timedomain. But, as soon as zeros are removed from the end from the current stream and the beginning of the next one, the length of the decoded data which is sent to the audiobuffer doesn't have this fixed length anymore, and guess what you get: a gap

Edited January 23, 2005 by bug80

[dex Otaku]

As far as MP3 vs ATRAC goes -- LAME is absolutely HORRIBLE compared with HI-SP or even MD LP2 mode --believe me I've had this stuff on Oscilliscopes and the MP3 stuff is VERY ARTIFACTY.  LP2 comes out quite cleanly surprisingly enough, HI-SP is around 95 -98% as good as PCM and unless you have top rate equipment (and I mean TOP RATE)  and a good amped set of cans  with the ears to match you probably won't be able to hear any difference except inside the walls of a first class professional recording studio.

Are you so sure about this? Lame encoded mp3s should in fact differ very little from HiSP.

I'm interested in what you're saying about oscilliscopes. Considering the fact that it should be impossible to see nearly -any- form of audible artifacting on an oscilliscope [that doesn't do series-sample and hold]. Also considering the fact that one would usually use various forms of spectrography to spot artifacting, not something that displays an instantaneous measurement of a waveform.

HiSP [in its first incarnation] has already been clearly demonstrated [through measurement] to not be as effective in encoding as ATRAC SP. It has also been shown to have the same ringing/pre-echo problems that the other forms of ATRAC do, to which regard are measurably outperformed by both the Lame and Fraunhofer mp3 encoders, due to mp3's slightly better abilities at handling high transients by switching between long and short blocks - something that [to my understanding, at least] ATRAC is bad at mostly due to the longer block lengths it uses.

And yes, I'm skipping a bunch of the background information that explains what long and short blocks are for, and how transform windows overlap which is part of what causes many artifacts, and on and on. I might also be using a bit of ambiguous terminology in talking about block lengths, but then, it's 5am.

For the "More Technically Challenged"  -- an Oscilliscope essentially is a machine with a screen on it where you can display the sound wave and measure various characteristics. A sound wave is a complex combination of Sine Waves which shouldn't have any "Jaggies" or "artifacts"

An oscilliscope is a simple device that displays a measurement of voltage over a set time period [amplitude vs. frequency].

Most forms of artifacting are harmonic, and as such, would bring little in the form of "jaggies" to any signal [ringing and pre-echo would in a certain sense, but the display of these on an oscilliscope would pass so quickly that there's no way you could hope to see any but the absolute worst examples].

Now even PCM can have some problems as PCM is pure digital -- and in order to get this into a form that our Ears can actually hear it has to be converted into analog Sine waves. There's complex mathematics called Fourier Ananlysis which shows how this is done -- but it's not a straight forward process. The hardware is called an A/D or Analog to Digital converter --it's simply doing "The Maths" in hardware. If the quality of the A/D converter is bad then you are still up the creek.

Fourier analysis is used to convert sampled amplitude data from the domain of time into that frequency.

This and related transforms [usually discrete or modified discrete cosine transforms] are employed in the kind of analysis used to convert PCM data into the various lossy compression formats.

Fourier analysis is not used at any point in PCM recording or playback, to my knowledge. The simplest A/D conversion is done using a resistor ladder network to convert variance in voltage to discrete steps which are measured at a given clock rate - the first giving bit depth, the second being the sampling rate. Others methods use variations of pulse width modulation [1-bit systems with extremely high sampling rates] which are then often converted to PCM for more common use. SACD's newer bitstream digital format is an example of one of these variations on PWM which forgoes the conversion to PCM.

PCM A/D converters also do not do any form of math; they are not signal processors. They simply measure an incoming signal and record the measured value once for every pulse of their clock.

Now converting MP3's which are full of "noise and errors" the poor old A/D converter has a really hard time which is why you will never ever see MP3's being used in high end professional recording studios other than used for churning out relatively low level "consumer grade" quality material).

Studios generally don't use any format that is lossy until the final step of the process - packaging the product for the consumer.

MP3s have nothing to do with A/D converters. A/D converters do not encode or decode lossy formats; they convert analogue signals into digital, most commonly using forms of PCM and PWM.

Furthermore, once compressed data is converted from the frequency domain back to that of time [i.e. to PCM], that signal is then converted to analogue by a D/A converter.

I'm seriously wondering where you got the idea that mp3s cause A/D converters trouble of some kind, since they not connected in any way whatsoever. Unless, of course, you've converted that mp3 to PCM, then to analogue using a D/A converter, then put the resulting analogue signal back into an A/D converter. At which point the A/D could care less whether the source was originally in mp3; it's simply converting analogue to digital to the best of its ability.

I'm not trying to flame anybody here --the MP3 format has done wonders for "portable music" and in general is sufficient for "The Masses" just as Macdonalds is too --but if you prefer French Gourmet you won't go to Macdonalds --same with sound --for me MP3 currently at any bit rate doesn't have it --but if YOU like it then good luck --in this game the only thing that matters ARE YOUR OWN EARS.

I'll agree with you here. Listening depends on perception; perception is subjective.

What matters most is whether -you- like what -you- are hearing.

I'm not trying to start a flamewar either; I'm simply pointing out that that your facts are more than merely suspect, but, for the most part, entirely wrong.

I make mistakes, too. And I occasionally profess things that are entirely wrong because I simply haven't learned enough yet. I won't hesitate to correct misinformation, though, because other people are learning from this, too. If they walk away with a head full of wrong ideas, that's a bad thing in my opinion.

I also welcome anyone to correct me if I'm wrong on anything. I would rather learn what's correct and be slightly humiliated for a few minutes than be wrong and not know about it.

[bug80]

Because perception is all what matters, I'm currently working on programming a blind test, in which the subject hears pairs of signals (each encoded with either MP3, ATRAC(+), AAC or OGG at different bitrates). After listening to one pair, the listener can point out which of the two he/she thinks sounds best.

Combining all results should lead to a "ranking".

[dex Otaku]

Because perception is all what matters, I'm currently working on programming a blind test, in which the subject hears pairs of signals (each encoded with either MP3, ATRAC(+), AAC or OGG at different bitrates). After listening to one pair, the listener can point out which of the two he/she thinks sounds best.

Combining all results should lead to a "ranking".

←

There is a freeware program used for ABX testing already. Poke around on hydrogenaudio, members there use it quite frequently for comparing codecs and such.

[bug80]

There is a freeware program used for ABX testing already.  Poke around on hydrogenaudio, members there use it quite frequently for comparing codecs and such.

←

That's great! I won't bother programming it myself, then.

What surprises me though, is that according to this test, ATRAC3 doesn't perform very well compared to other codecs

* EDIT * at a bitrate of 132 kbs, that is

Edited January 23, 2005 by bug80

[dex Otaku]

That's great! I won't bother programming it myself, then.

What surprises me though, is that according to this test, ATRAC3 doesn't perform very well compared to other codecs 

* EDIT * at a bitrate of 132 kbs, that is

←

ATRAC tends to fall towards the bottom of most independent ABX testing, from what I've seen.

Mind you - the tests I've seen used material encoded by SS, which has been shown to encode not nearly as well as the hardware codecs in MD or HiMD recorders. Especially with LP2.

I was totally put off LP2 because of SonicStage. A friend sent me copies of LP2 recordings she made with an MD recorder; the results were vastly superiour to any of the encoding I've heard SS do. Still, under most circumstance LP2 isn't enough for my tastes. I'd consider it well-suited for car listening, but I find it unbearable with 'phones. It also does quite well with mono recordings [thanks to joint stereo encoding].

[bug80]

ATRAC tends to fall towards the bottom of most independent ABX testing, from what I've seen. 

Mind you - the tests I've seen used material encoded by SS, which has been shown to encode not nearly as well as the hardware codecs in MD or HiMD recorders.  Especially with LP2.

I was totally put off LP2 because of SonicStage.  A friend sent me copies of LP2 recordings she made with an MD recorder; the results were vastly superiour to any of the encoding I've heard SS do.  Still, under most circumstance LP2 isn't enough for my tastes.  I'd consider it well-suited for car listening, but I find it unbearable with 'phones.  It also does quite well with mono recordings [thanks to joint stereo encoding].

←

They've used the best of the best encoders with the most optimal settings to encode their AAC, MP3, Vorbis, etc. files, while in the case of ATRAC, they only got one choice: Sonicstage (without optimalization settings). That explains a lot, yes.

[mrsoul]

PROS:

-designed for music: atrac is basically the only compression system designed from the ground up for music playback

←

Maybe I missed something but I thought Ogg Vorbis was designed specifically for music playback. Also, does FLAC count?

BTW, very good info in this thread. I just got a new MP3 car deck that I am enjoying (nice to have a few CDs on 1 CD but the gaps between tracks isn't cool but that is a buffer engineering oversight I think) but the only reason I have that new deck is the frontside AUX input for my Hi-MD

[Andy]

Fourier analysis is used to convert sampled amplitude data from the domain of time into that frequency.

This and related transforms [usually discrete or modified discrete cosine transforms] are employed in the kind of analysis used to convert PCM data into the various lossy compression formats.

This is correct. Recently, people have been playing around with wavelets for this sort of stuff (while fourier transforms transform into sine waves of differing phases and amplitudes, wavelets are used to transform into any set of functions, by choosing these functions well you can get much more compressable information). The PCM data is in the time domain (think just a plain waveform showing time vs amplitude), in this case, the energy of the signal is (usually) evenly distributed over the whole time, when you apply a fourier transform, you put the signal in the frequency domain, from here, usually most of the energy is in the lower bands, so it is easier to compress the signal. (Very simplified)

Fourier analysis is not used at any point in PCM recording or playback, to my knowledge. The simplest A/D conversion is done using a resistor ladder network to convert variance in voltage to discrete steps which are measured at a given clock rate - the first giving bit depth, the second being the sampling rate. Others methods use variations of pulse width modulation [1-bit systems with extremely high sampling rates] which are then often converted to PCM for more common use. SACD's newer bitstream digital format is an example of one of these variations on PWM which forgoes the conversion to PCM.

Also correct.

I believe SACD's system is the same principal as a 1-bit amp (both produce a waveform which just alternates between 2 fixed voltages, this waveform is then filtered to form the output). There are about 3 other methods I can think of for DA conversions, one is a resistor ladder (I think this is the most common method) which uses lots of resistors of the same resistance to form the final signal - convenient because all the resistors can be made with the same process and they won't differ much. I'm not sure if it is used anywhere, but some DA's just have 1 resistor per output voltage. I think there is also a method where you iterate through a few steps where each step you get closer to the desired voltage - I can't remember details, so it might be me making it up - I know there is a similar parallel for AD conversion, so I might be getting mixed up.

PCM A/D converters also do not do any form of math; they are not signal processors. They simply measure an incoming signal and record the measured value once for every pulse of their clock.

Also correct, except I think most AD converters use "progressive refinement", which requires them to have a input clock which is faster than the sample rate by a factor the the sample depth.

[jadeclaw]

Maybe I missed something but I thought Ogg Vorbis was designed specifically for music playback.  Also, does FLAC count? 

←

Almost all codecs had been designed for music, including OGG and MP3.

And FLAC of course too, however, FLAC is lossless like the well known ZIP.

There are codecs available which are specifically optimized for speech,

just check out Speex

[ROMBUSTERS]

Almost all codecs had been designed for music, including OGG and MP3.

And FLAC of course too, however, FLAC is lossless like the well known ZIP.

There are codecs available which are specifically optimized for speech,

just check out Speex

←

what i was refering to was that Atrac was basically the first codec that was in use for compressing music and that MP3 was made for general purposes and not just purely music.

[dex Otaku]

mp2 [PASC] and mp3 were in development as early as 1997, making their history pretty much concurrent if not slightly ahead of ATRAC's.

All "high-fidelity" codecs are considered general-purpose.

The MPEG 1 and 2 audio layers [l1, 22, and l3 which is mp3] were developed with things like DCC and DAB [digital radio] in mind, so I would surmise that they were made as much for music as ATRAC was.

ATRAC has been used other places, too, such as for Sony's SDDS film sound format [which was introduced as the same time as MD c.1992-3].

[ROMBUSTERS]

mp2 [PASC] and mp3 were in development as early as 1997, making their history pretty much concurrent if not slightly ahead of ATRAC's.

All "high-fidelity" codecs are considered general-purpose. 

The MPEG 1 and 2 audio layers [l1, 22, and l3 which is mp3] were developed with things like DCC and DAB [digital radio] in mind, so I would surmise that they were made as much for music as ATRAC was.

ATRAC has been used other places, too, such as for Sony's SDDS film sound format [which was introduced as the same time as MD c.1992-3].

←

you just contradicted yourself with your first dates there.

Sony MDs came out in what '91/92? and they have always employed Atrac obviously. Therefore how can "mp2 [PASC] and mp3 were in development as early as 1997, making their history pretty much concurrent if not slightly ahead of ATRAC's."

That would almost imply it was in use before Atrac, which was not true. Just to clarify.

Also taken from Wikipedia

"ATRAC (Adaptive TRansform Acoustic Coding) is an audio compression algorithm used to store information on Minidiscs and other Sony-branded audio players. First developed by Sony in 1991; the higher compression flavors of ATRAC3 and ATRAC3plus followed in 2000 and 2003, respectively."

not to start a flame or anything just wanted things to be clear, as im sure you know loads more MD info then I

Edited January 24, 2005 by ROMBUSTERS

[dex Otaku]

you just contradicted yourself with your first dates there.

Sony MDs came out in what '91/92? and they have always employed Atrac obviously. Therefore how can "mp2 [PASC] and mp3 were in development as early as 1997, making their history pretty much concurrent if not slightly ahead of ATRAC's."

That would almost imply it was in use before Atrac, which was not true. Just to clarify.

Hahahahahahahahahahaha. BRUTAL TYPO TIME!

That should have read *1987* not 1997.

MD reached the market in North America 1992-3. DCC was ostensibly released at the same time, using PASC which I had previously understood to be a codevelopment basis for Musicam and mpeg l2 audio. Oddly, though I have read this from multiple sources in the past, Wikipedia makes no mention of the PASC connection, so perhaps [assuming the Wikipedia article is more definitive, which isn't necessarily correct] my previous sources on this were dubious:

MPEG-1/2 Layer 2 encoding started in life as the Digital Audio Broadcast (DAB) project initiated by the Fraunhofer Society. This project was financed by the European Union as a part of the EUREKA research program where it was commonly known as EU-147.

EU-147 ran from 1987 to 1994. In 1991 there were two proposals available: Musicam (known as Layer II) and ASPEC (Adaptive Spectral Perceptual Entropy Coding) (with similarities to MP3). Musicam was chosen due to its simplicity and error resistance.

What I said was that the -development- of what became mp2 and mp3 predated ATRAC's, which your wikipedia quote establishes as having started in 1991. I was not sure of when -development- of ATRAC began.

Both algorithms were finalized in 1992 as part of MPEG-1, the first phase of work by MPEG, which resulted in the international standard ISO/International Electrotechnical Commission 11172-3, published in 1993.

For all intents and purposes, all three formats were established in the same year, if not in fact put into widespread use. In other words, they're all basically the same age.

not to start a flame or anything just wanted things to be clear, as im sure you know loads more MD info then I

Don't be so sure. I only started learning about MD about 9 months ago, though I had followed the initial release of both it and DCC with interest bordering on outrage [lossy compression being a good thing? Never!] back when I was in high school.

And heh, no flamewar. As I said in that previous post, corrections are welcome. Otherwise we never learn.

[ROMBUSTERS]

right but i think one thing your missing, if i remember what i read about MP3 correctly, is that your saying MP3 was released in '92 when i think ive read that was more or less MP2 (which started some internet trading but nothing major) until it was used (and abused) until it became MP3, later on in life of course. Oddly enough MP3 was only designed to reach close to MP2 at 196~256kbps at 128kbps. In reality (just like Atrac SP) MP1 is the best quality (although it runs at something like ~340kbps)

Edited January 24, 2005 by ROMBUSTERS

[dex Otaku]

right but i think one thing your missing, if i remember what i read about MP3 correctly, is that your saying MP3 was released in '92 when i think ive read that was more or less MP2 (which started some internet trading but nothing major) until it was used (and abused) until it became MP3, later on in life of course. Oddly enough MP3 was only designed to reach close to MP2 at 196~256kbps at 128kbps. In reality (just like Atrac SP) MP1 is the best quality (although it runs at something like ~340kbps)

←

Look up mp3 on wikipedia.

What became MP2 and MP3 are totally separate formats that were developed [more or less] concurrently.

MP2 did not become MP3.

Yes, MP3 was designed with the specific intent of achieving equivalent quality at 128kbps as MP2 did at 192kbps.

And it's not that MP3 was -released- in 1992, it's that the standard was ratified in 1992.

MPEG Layer-I was intended for use at 384kbps if I remember correctly [192 * 2]. I also believe [though this is at least partly subjective] that it's a misconception to assume that l1 audio is higher quality; the purpose of each codec was to achieve equivalent quality at lower bitrates which each successive layer; hence l1 at 384kbps should be of about the same quality as l2 at 192-256kbps, and l3 at 128-160kbps. In practice I've found these to be pretty arbitrary, really.

Using a given codec at higher than it's proposed "quality equivalent" bitrate should achieve higher quality, of course. It would thus make sense that l3 at 320kbps should have fewer artifacts than l1 at 384, though in practice it doesn't actually work this way as each codec has its own distinct flavours of artifacting.

In the early days of MP3, a lot of people [including myself] chose to use MP2 for "archival" encoding. At the time [around 1997] I had compared MP2 vs. MP3 at 256kbps and settled on the MP2 encoder. The probable reason for finding it better was that MP3 encoding hadn't evolved to the point that it has now. I still listen to some of those MP2 tracks, and they still sound just fine, incidentally.

Internet traffic in either did not become common until much later than the point when both MP2 and MP3 were ratified. In 1992 the fastest desktop computers still took over 10 hours to encode a single 4-minute stereo track into mp3, and hardware encoders were both expensive and uncommon. Realtime MP3 decoding was also not even possible in software with anything less than about a P100-based system. This has as much to do with the I/O capabilities of computers at that point as it does with their processing power.

MP3 and internet traffic in it is really a post-Pentium phenomenon. I could be wrong about this, but I've been on the net since about 1988 and I didn't know of anyone trading music online until early 1996 [when the P133 or thereabouts was state of the art]. Keep in mind that most computers at that point did not even have 1GB hard discs yet [they were around, yes, but not in the majority by any means], making the storage of music for compression highly impractical for the vast majority of users. 28.8 modems had only been around for a short time, as well. Broadband was uncommon in the extreme [i was peripherally involved in the testing of cable internet in New Brunswick at the end of 1996, as they were one of the testbeds for North America].

[Damage]

Re: MP2, Unless I'm mistaken, and I probably am, (it's either or) I think it was to be used in concert with Video CD, and as such, I believe it has one "fixed" bitrate to be used. Or, I tend to recall this from soething I read about a while back.

At any rate, I remember seeing and attempting to encode my own on a P100 (with 64Megs of EDO RAM and 4GB HD, that machine was one thousand three hundred thirty seven back in the day) and that process taking well over several hours. Using Win95's media player (and that capped at AM quality rendering out of my ignorance), I was rather impressed at AM rendered MP2/3 files, then find my surprise when I actually tried Winamp 0.9 or such.

But the whole MP3 thing came about into prominence around the same time as World Wide Web and such. The WWW was a great portale for finding random MP3s, then followed up by Napster. It's the confluence of these technologies that gave MP3 the staying power that it has even to this day.

[dex Otaku]

Heh. I still remember the transition between Gopher and the WWW.. using Lynx, around 1991 or something.

MP2 does not have a fixed bitrate, but in the case of VCD it was limited to a single rate in order to ensure that software could play on any player.

[bug80]

I've heard that the idea of compressing signals using quantization in the frequency domain is more than 100 years old (could be wrong, but I thought early 20th century). However, back than it took a hall full of people and about one day to calculate a single FFT

So the idea is old, but they didn't have the computerpower we have now.

* EDIT * This statement doesn't mean I am over 100 years old, mind you!

Edited January 24, 2005 by bug80