Spectrum analyzation of WAV, 256kbps ATRAC3+, 132kbps ATRAC3

[Christopher]

Original information and pictures are here.

Conditions: Using TotalRecorder software and a MZ-NH1 with the NH1's direct USB playback capability, the fine folk at the Digital Audio Labratory [at av-watch] have done spectrum analyzation(s) comparing uncompressed WAV, 256kbps ATRAC3plus and 132kbps ATRAC3, with the ATRAC files being captured into wav files. The comparison tests, using WaveSpectra, [with an apparent focus on treble] are as follows: 1kHz, 20Hz, and some music.

1kHz tests:

-----------------

[uncompressed wav]

[256kbps ATRAC3plus aka "Hi-SP"]

[132kbps ATRAC3 aka "LP2"]

20kHz tests:

-----------------

[uncompressed wav]

[256kbps ATRAC3plus aka "Hi-SP"]

[132kbps ATRAC3 aka "LP2"]

Music tests:

-----------------

[uncompressed wav]

[256kbps ATRAC3plus aka "Hi-SP"]

[132kbps ATRAC3 aka "LP2"]

It's obvious that the differences between Hi-SP and PCM are nearly nil, amongst other things.

[Latexxx]

Unfortunately images don't tell us what it sounds like.

[Christopher]

You can simulate this test on your own with Sonicstage and the aforementioned softwares. :happy:

[Andy]

It's obvious that the differences between Hi-SP and PCM are nearly nil, amongst other things.

I don't think so.

1) For the tones, these are signals which you would expect to be compressed perfectly, since ATRAC does compession in the frequency domain, and a tone has a very simple breakdown in the frequency domain. (In fact, I'm quite dissapointed by the results of LP2 in this test)

2) The second test again is a signal which has a simple representation in the frequency domain.

3) Looking at the music samples, compare the blue channels of the PCM and HiSP. The blue in HiSP follows the red a lot more, while in the original, this doesn't happen so much (I'm mainly looking at the peaks in the mids). Now, look at the bass of the HiSP - notice how in the uncompressed version there are a few distinct peaks, where in the HiSP, these are a bit murkier (less distinction between them). The dropoff at the end is a little bit concerning, as I know that I can hear frequencies at 20KHz quite well, but at the same time, getting speakers to reproduce these would be quite tough.

Without listening, I'd guess that it sounds very good. But to call it identical to the uncompressed signal is wrong.

One of the good reasons for making this distinction is that if you want to do something whacky with the sound - eg, change the speed while keeping the same pitch, then suddenly all these things which ATRAC decided you wouldn't hear get pushed into the open. Processing lossily compressed data is bad.

For a fun example of this, if you have WinXP, backup a JPEG file and open it in the XP viewer (Fax/Picture viewer or something like that). Click on the rotation buttons lots of times (like 40+) and then compare the image you have with the original.

[jadeclaw]

If you click on " < Back number index > " on that AV-Watch page, you get a bunch of interesting reads and reviews. Despite the machine translation, that stuff is quite readable.

Second, if you want quickly test your Soundcard/recorder, hop over here:

http://audio.rightmark.org/

and download RMAA 5.3.

It sends a bunch of strange sounding test signals through the system and then reports the systems performance.

Tested mine, result: Analog part sucks like a black hole, digital part delivers 16 bit quality - as expected for a 16-bit card.

Record and analyse part can be separated, so that testing of recorders and encoders is possible.

Btw, the software is free.

And while rummaging through AV-Watch's reveiews, I stumbled upon this EMU 0404 review, which caught my eyes qualitywise and it's currently on sale for 99 bucks here.

[Christopher]

Without listening, I'd guess that it sounds very good. But to call it identical to the uncompressed signal is wrong.

Let me rephrase:

It's obvious that the differences between Hi-SP and PCM are nearly nil, amongst other things.

It's obvious that the differences between Hi-SP and PCM to the average consumer would be nearly nil, amongst other things.

Better? :happy:

[Andy]

Much better

Although I am a bit surprised by the apparent cross-talk there. Does Hi-SP now do joint stereo? SP didn't, so I would have guessed that Hi-SP would be just like it.

[anont]

It's obvious that the differences between Hi-SP and PCM to the average consumer would be nearly nil, amongst other things.

To be elitist, the difference between LP2 and PCM to the average consumer is nearly nil.

[joemango]

Is it me, or does there seem to be a huge difference between the input (red) and output (blue) of the uncompressed WAV? Is this due to the necessity of the analog input? Or was the test conducted with a digital source?

[Christopher]

All digital.

[joemango]

OK, so why the discrepancy, or am I just reading the chart wrong? Can it be that far off?

[hobgoblin]

so whats with the diffrence in input vs output db? did you turn output down or something? if not then wtf is going on?

oh and btw, that comment of andys about rotateing a jpeg sounds more like a bug in the software then something with the image compression as rotateing a image is just reading the same image data into a diffrent reference (vertical becomeing horisontal and left becomeing right and so on). altering speed while keeping pitch sounds more like stretching a image horisontaly while keeping the vertical stable...

[The Dude]

Red and blue are not input/output, they are the left and right channels of the stereo signal.

Regards, Dude

[Andy]

That rotating JPEG thing isn't a bug. It is what happens when you keep on recompressing something using lossy compression. Rotation was forcing it to do something to the image, which has an easy inverse, however, because lossy compression is lossy, the inverse of an operation won't necessarily be exactly the same as the original because of the compression and decomression. If it was done to a bitmap or a png (lossless compression), then it would work perfectly.

Say you want to record stuff to use it for sound effects, and one of the samples you take you want to slow it down lots and lower the pitch. If you do this, then what will happen is the spectogram graph will shift to the left, bringing all those high components into hearing range, suddenly you'll have a big hole in the sound, much earlier than if it was an uncompressed signal.

[hobgoblin]

then i follow your thinking on the image but you forgot to state that you want to write the rotated image to disk every time you do one 90 degree rotate. jepgs are only compressed when its written to disk, there is no reason to compress it while still in memory.

and lossy compression was never realy supposed to be a intermediary storage as i can see it. it have allways been a end user storage format for when you just want to bring out and listen or look at it but never manipulate it. i would realy love for all kinds of lossy formats to be removed but dont kid yourself into thinking that we can afford the increase in storage space that this will lead to. not by a long shot...

[Ppk3000]

then i follow your thinking on the image but you forgot to state that you want to write the rotated image to disk every time you do one 90 degree rotate. jepgs are only compressed when its written to disk, there is no reason to compress it while still in memory.

Andy referred to the (convinient for viewing, but crappy otherwise) "Windows Picture and Fax Viewer", the program (even warns that it) saves to disk everytime an image is rotated. I believe he intentionally used it as an example to demonstrate recompressing data multiple times.

Of course, just about any real imaging editing tool would never write to disk until the user saves the file, and the image is also "decompressed" while it's in the editing environment, so no unintentional compression artifacts would be added no matter how many times you rotated.

[hobgoblin]

ah, as i never have used that app and hope i never have to (pats his irfanview) i didnt know this. then the post makes a whole lot more sense and im sorry for the useless traffic...

[Andy]

and lossy compression was never realy supposed to be a intermediary storage as i can see it. it have allways been a end user storage format for when you just want to bring out and listen or look at it but never manipulate it. i would realy love for all kinds of lossy formats to be removed but dont kid yourself into thinking that we can afford the increase in storage space that this will lead to. not by a long shot...

Exactly. If you read my initial comment, it was giving reasons why you might want lossless compression over lossy compression even if you can't hear the difference. I have a tendency to be unclear with stuff that I write, which is probably the reason it has taken so many posts to work out what I'm saying

Offtopic Section:

- you need a (free) plugin to rotate jpegs losslessly in irfanview.

- Windows Picture and Fax Viewer is really crappy. It takes longer to load a picture using it than in paint (or at least it does on my computer). Irfanview makes me happy too

[MZ-NH1]

Just wondering, with my MZ-NH1, I am currently uploading all songs onto my MD as OpenMG Audio (ATRAC3plus) 64 kbps. Would there be much different in sound quality between this and the 256kbps? And which format is better ATRAC3 or ATRAC3plus? Obviously with the compression at 64, the files are smaller, and i am able to fit more songs onto 1 disc.

Happy to hear any comments

[jadeclaw]

Unless your ears are defective from working in an extremely noisy environment (e.g. Jet engine maintenance & testing ), you should hear a considerable difference between Hi-SP & Hi-LP. Hi-SP(256k) is far better and 99% CD-Quality. At the same bitrate Atrac3+ is better.

[MZ-NH1]

So would you say that OpenMG Audio (ATRAC3plus) 64 kbps is fine? Or would you suggest I upload at 256kbps?

[aeriyn]

So would you say that OpenMG Audio (ATRAC3plus) 64 kbps is fine? Or would you suggest I upload at 256kbps?

It all depends on what you're after. IMO, Hi-LP mangles most music into muddy, unrecognizeable blobs of sound, however it's perfectly acceptible for talk radio recordings, stand up comedy, audiobooks, etc.

For music, for anything with more complex instrumentation than your basic pop, I'd suggest Hi-SP. Since I do not own a Hi-MD unit myself (just basic old MD stuff) I use old SP mode, and I imagine that Hi-SP is pretty similar. Although several people have claimed that ATRAC 4.5R is superior to Hi-SP.

Since I have no Hi-MD equipment, testing that would be difficult for me.

[jadeclaw]

It's simple. Make some recordings in both modes and decide for yourself.

For me, 64k is not good enough.

[MZ-NH1]

Ok, thanks for that! I myself record in Hi-SP, because that records at 256kbps on the MD. When converting files onto the Hi-MD, it is also in Hi-SP at 64k. When I buy some 1GB discs, I think I'll upload everything to my MD as 256kbps... providing my tracks will fit on.

Thanks very much for all that info aeriyn and jadeclaw, very much appreciated

Edit: I am also annoyed and curious as to why Sony hasn’t allowed the 192kbps conversion rate, as it would be considered a big sound quality difference/gap between 64kbps and 256kbps. Anyone know the reason for this?

[MZ-NH1]

Listening to music on my Hi-MD at the moment. I have just changed all my music to 64kbps.... instead of the 48kbps I use to have. I must say that the quality isn't any better than 48, if not worse! I have also noticed that on the MD, it now says Hi-LP :| I am sure I had Hi-SP on before... how do I make sure it is Hi-SP and not LP?? Is it because I have gone to 64kbps, that it has changed to LP?

Thanks everyone for your help! Greatly appreciated

[MZ-NH1]

Listening to a Creed song now, and it is VERY static! :@ not too happy!

[aeriyn]

By default SonicStage transcodes everything to Hi-LP 64kbps ATRAC3+. In order for you to get it to transcode tracks to Hi-SP, you have to fiddle with the options a bit.

[MZ-NH1]

Oh ok, thanks for that. I'll have to have a look into that. Got exams now for 2 more weeks, so when they have all finished I’ll focus more on the MD and how it operates! Can't wait!

[minidisc3]

I've done some listening test on my NW-HD1.

The sound test below might be interesting to Hi-MD users as well.

I've tested ATRAC3plus 256 and ATRAC3 132 with the track Patience from George Michael's latest album Patience. This is a piano song with vocals. Piano music is very dynamic and the vocals are very pronounced and in the foreground. 'S' sounds in the vocals are somewhat overly pronounced on the original CD.

I used the original CD as a source and I used a 192 kb/s stereo encoded MP3 (Lame) as a source. All results below are compared to the original CD track. I listened with good quality headphones.

Results:

192 kb/s stereo encoded MP3 => ATRAC3plus 256:

Very crisp and full sound. Clear and very nice treble. Piano and voice lost some ground bass. Not as airy and 'open' sounding as the original, but still okay in this respect. Very attacking analytical sound. I liked it very much.

192 kb/s stereo encoded MP3 => ATRAC3 132:

Blurred sound and distorted treble, especially with the 'S' sounds in vocals. Loss in dynamics. The music looses its bite and is not as attacking as the CD original. Flat and not airy sounding. Somewhat sparkling background hissing. To me it sounds annoying. I found the same results when using this method with my MZ-N1.

CD => ATRAC3plus 256 (direct rip):

Very crisp and full sound. Clear and very nice treble. Ground bass and fullness of the original remains. Airy and 'open' sound; you can actually imagin yourself being in the same room as piano and singer! Very precise reproduction, I like it very much and I'm happy Hi-SP lives up to my expectations (until now). Very hard to tell the difference with the CD original!

CD => ATRAC3 132 (direct rip):

Now this was a real surprise! This sounds much better than the MP3 converted to ATRAC3 132. 'S' sounds in vocals are less pronounced but still overly pronounced. Treble was less blurred and more precise. Still loss of dynamics. Again it sounded a bit flat and not airy.

Overall conclusion:

I never really liked the Longplay mode of NetMD (ATRAC3 132) and this test only confirmed my earlier verdict on ATRAC3 132. The direct CD rip to ATRAC3 132 can be acceptable, that possibly depends on the type of music you're listening to. Dynamical music gives this codec a very hard time! This codec really tries to fool you at 132 kb/s, making you think everything is alright, but at the same time making you feel uncomfortable (is this what psychoaccoustics do to you?). I guess it will do when listening at a lower volume or in a noisy environment.

It's clear that when using ATRAC3 132 the source has to be a CD. The MP3 to ATRAC3 132 conversion makes listening almost unbearable. ATRAC3 132 artifacts seem to be far more annoying than MP3 192 kb/s artifacts.

ATRAC3plus 256 is a superb codec!

Ripping CD's to ATRAC3plus 256 absolutely gives the best results. Until now, I could not distinguish any difference to the CD original here.

ATRAC3plus 256 is definitely better than MP3 192 kb/s, it revealed all shortcomings of the MP3 sound.

Maybe now you understand why I didn't even bother testing the lower ATRAC3(plus) bitrates yet. I still might. If I do, I'll share my findings with you.

I don't even think it would be fair to negatively judge the lower bitrates, eventhough Sony claims 64 kb/s produces CD quality sound. If the low bitrates are not satisfying to you, just move up to a higher bitrate.

As far as the highest possible bitrates are concerned, they should live up to the price of 400 Euros for the HD1. This product is marketed as a high-end device and luckily it can produce brilliant sound depending on the choices YOU make.

I guess I'll have to re-rip my favourite CD's to ATRAC3plus 256...

Marco

[Christopher]

Excellent, Marco.

You should also participate in this thread: http://forums.minidisc.org/viewtopic.php?t=5950

[MZ-NH1]

Thanks HEAPS for that Marco, very much appreciated. At home, I normally use Windows Media Player to copy my songs from CD to PC, and then I use the import tool in Sonic Stage to import the music off the hard drive. I then load the songs onto my MD at ATRAC3plus 256. Are you saying the sound will be as good as CD quality if I do it this way?

Thanks for that research, was excellent!

Andrew

[aeriyn]

Thanks HEAPS for that Marco, very much appreciated.  At home, I normally use Windows Media Player to copy my songs from CD to PC, and then I use the import tool in Sonic Stage to import the music off the hard drive.  I then load the songs onto my MD at ATRAC3plus 256. Are you saying the sound will be as good as CD quality if I do it this way?

Thanks for that research, was excellent!

Andrew

Actually it will be much better if you use SonicStage or MD Simple Burner to rip directly from CD > ATRAC3+, since ripping into WMP is compressing it as MP3 or WMA.

[MZ-NH1]

Ok, thanks for that. If I did rip it to ATRAC3... that means I would end up playing all music through SonicStage, and only PC's with SonicStage would be able to listen to the music... I'll have to think about that one.

Thanks for that,

Andrew

[aeriyn]

Ok, thanks for that.  If I did rip it to ATRAC3... that means I would end up playing all music through SonicStage, and only PC's with SonicStage would be able to listen to the music...  I'll have to think about that one.

Thanks for that,

Andrew

What I tend to do with CDs that I own (not counting the ones that got stolen out of my car), is I will rip the CDs to MP3 320kbps CBR since I don't have the hard drive space for too many FLACs, and 320kbps is plenty small enough not to totally fill up my hard drive.

320kbps MP3 > ATRAC3+ 256kbps shouldn't be TOO bad, but CD > ATRAC3+ 256kbps would be a whole lot better. No transcoding involved.

[MZ-NH1]

Yeh I understand where your coming from. Thanks.

But I still say, all PC's would have to have SonicStage installed on them to play the files. And if I needed to burn a copy of a CD because it was stolen, then I would have to convert it to .wav...

[aeriyn]

I'm saying rip it twice. :laugh:

Time-consuming but it pays off. Use Simple Burner and SonicStage doesn't leave crap all over your HDD.

[MZ-NH1]

Kk, now worries. :rasp: Thanks

Since I got a pretty big hard drive (120GB), I think in future I'll copy the CD as ATRAC3 256 through SonicStage and through Windows Media Player as well.

Thanks very much for all that information both of you. My knowledge of my MD player has grow enourmously ever since I found out about this Forum site!

[minidisc3]

I should be (re)ripping all of my CD's straight to ATRAC3plus 256 with SonicStage, that gives the best results.

However, I have my whole collection already in MP3 format, worth about 25 GB. They're all in 192kb/s stereo. I can still remember how much time that took...

So instead, I decided to import the MP3's and convert them into Atrac3+ 256. Atrac3+ 256 is that good, it preserves the full quality of the MP3's.

After transferring them I delete the *.oma files from the PC with Windows Explorer. In this way the SS library will stay intact.

I'm so happy Sony took the chance to add this new codec with the introduction of Hi-MD (and the NW-HD1) of course.

I hope they will also offer ATRAC3+ 256 downloads from Connect when it becomes available in my country, but I doubt it...

Aeriyn, you said you're ripping your CD's to 320 kb/s MP3 and convert them into ATRAC3+ 256.

It's only my humble theory, but I think you'd better rip them to max. 256 kb/s MP3. I guess ATRAC3+ and MP3 work quite similar. When converting from 320 kb/s MP3 to 256 ATRAC3+ a lot of information needs to be thrown away, because of the lower bit rate of ATRAC3+ 256. This will not be the (so much) case when converting MP3 bitrates lower than 256 kb/s to ATRAC3+ 256.

Just do a quick test, I believe with 224 kb/s MP3's you'll hear no difference to the CD original. There should be no need to use 320 kb/s. It's more important you don't use the joint-stereo feature, just use the stereo setting (especially at high MP3 bitrates!).

[minidisc3]

There is one hughe advantage in ripping to MP3 before converting to ATRAC3(plus). You can normalize the MP3's to have higher volume output on your Network Walkman.

Mainly older CD's result in MP3's with low volume.

[aeriyn]

Aeriyn, you said you're ripping your CD's to 320 kb/s MP3 and convert them into ATRAC3+ 256.

It's only my humble theory, but I think you'd better rip them to max. 256 kb/s MP3. I guess ATRAC3+ and MP3 work quite similar. When converting from 320 kb/s MP3 to 256 ATRAC3+ a lot of information needs to be thrown away, because of the lower bit rate of ATRAC3+ 256. This will not be the (so much) case when converting MP3 bitrates lower than 256 kb/s to ATRAC3+ 256.

When transcoding, it's always better to start with the least amount of loss as possible. When you transcode from 256kbps MP3 to Hi-SP, it's going to throw away exactly as much information as it does when you rip directly from CD to Hi-SP.

However, with MP3, information has already been thrown away, so you get double the artifacts for the same low, low price. So if you absolutely have to transcode, it's better to do it with higher-bitrate files.

Also, Joint Stereo, if done in the m/s method, is lossless. Intensity stereo, however, is not.

There is one hughe advantage in ripping to MP3 before converting to ATRAC3(plus). You can normalize the MP3's to have higher volume output on your Network Walkman.

Mainly older CD's result in MP3's with low volume.

You want to fix the volume of all your MP3s to a certain rate, losslessly? Well, I have a nifty little program for you.

MP3Gain

Sets the gain (not peak level normalization, which is not lossless) of all your MP3s to an even level.

Oh, and you're welcome. :laugh: