HydrogenAudio

Hydrogenaudio Forum => General Audio => Topic started by: Joe Bloggs on 2002-09-11 05:22:17

Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 05:22:17
You know what I'm talking about... the proponents of Blade back in the day, and those who say that MD is more 'musical' than mp3, etc...

It's almost like these people feel that whereas a well-tuned psychoacoustic encoder doesn't show any outright detectable artifacts, the bits and pieces they trimmed at the edge of human audibility add up to a loss of 'tonal purity' and 'musicality' whereas they would prefer a coder that sometimes let artifacts slip through but at the same time preserve the 'purity' of the fundamental tones in the music, etc.
Title: 'tonal Purity'? 'musicality'?
Post by: ff123 on 2002-09-11 06:45:17
Chris Johnson (webmaster of Airwindows) is still propagating this stuff about "tonal purity":

http://slashdot.org/comments.pl?sid=39509&cid=4221824 (http://slashdot.org/comments.pl?sid=39509&cid=4221824)

And people modded his post up as "insightful."  I think it is just nonsense.

BTW, when Chris talks about ABXing 256 kbit/s mp3 from the original, he's really referring to distinguishing castanets.wav encoded with FhG's slow codec, not about mp3 at 256 kbit/s in general.

Also BTW, when Chris talks about 94% confidence in a comparison of dithering vs. truncation at 16 bits, he neglected to mention that he actually continued past his initial score of 8/10 to get 9/16 overall.  After that, he tried another test with a different dither and got 10/16 on that one.

ff123
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 07:29:23
Gee, how do you know this shit... I mean when you fabricate ABX scores how come you'd let others know the real score too...
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 07:32:28
I just have trouble supposing that all these people say MD is superior to mp3 to their own ears and there's absolutely nothing it. Especially when somebody claims that he has heard the best of both mp3 and MD on high end systems.

Perhaps MD has a more 'pleasant' distortion?
Title: 'tonal Purity'? 'musicality'?
Post by: KikeG on 2002-09-11 08:26:09
Quote
Perhaps MD has a more 'pleasant' distortion?

Not to me, at least if if we compare with LAME -aps.

As I said at http://www.audio-illumination.org/forums/i...=3373&st=25&hl= (http://www.audio-illumination.org/forums/index.php?act=ST&f=1&t=3373&st=25&hl=),
MD distorts the sound in a quite more annoying way than MP3 --aps, according to a preliminar quick comparison with the later: pre-echo (castanets) , harsh and distorted highs (violins1, french) , losing of spatialness (timpani), timbral artifacts (violins2).

I will make a more thorough comparison with LAME -aps on these samples when I have time.
Title: 'tonal Purity'? 'musicality'?
Post by: JohnV on 2002-09-11 14:44:43
Yeah, forget proper block switching and short blocks, and suddenly you have higher "tonal purity".. what a crap. Pre-echo in no way makes the sound more "tonally pure"..

Well, it's no surprise that Chris Johnson thinks that you either choose between transient accuracy and tonal purity with mp3, after all, he's the author of airwindows test site... A test site that uses inverse mixing for measuring quality of psychoacoustic codecs.. 
Can something be more inaccurate than claims like these..?

Tord came up with Blade's tonal purity when he was incapable to implement proper block switching and use of short blocks. Unfortunately so many people bought it, not very surprising though that Chris Johnson bought it, after all, his tests have nothing to do with measuring of quality of psychoacoustic codecs..
Title: 'tonal Purity'? 'musicality'?
Post by: Cygnus X1 on 2002-09-11 16:01:46
I don't buy the whole "tonal purity" argument with any codec, be it MP3, MPC, ATRAC, whatever. If one cannot ABX a sample from the source, that means that the encoding will sound no worse than the wav. Thus, proponents of the tonal purity theory should try comparing the original CD with the compressed recording, not just by themselves (it's easy to "make" people think that one recording sounds better when in fact it is of lower quality). If they indeed cannot be told apart, that means that the encoding cannot be any less pure (tonally) than the original CD!

As for MD, I think it has its uses. I own both MD and MP3, and use MD while exercising/etc due to the good battery life and being able to swap media if I so choose. I usually recommend it to friends who want good battery life and cheap media, but beyond that, MP3 players are a much more flexible choice. However, I was able to ABX a whole slew of samples a couple of years ago, so I wouldn't think of using it in a serious listening environment. Like I've said before, ATRAC is an unimpressive algorithm IMO, considering the type of artifacts (crackling, pre-echo) it produces at 292kbps ... ....perhaps this goes to show what proper tuning of a codec can do, in that since ATRAC is proprietary, it cannot be "tweaked" like some have done with MP3 and AAC.
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 16:24:39
Can you ABX 'tonal purity' and 'musicality'?

Like I said in the subtitle...

I'm sure that at some level (obvious enough) you can, but do any of you subscribe to the idea that there are some things that you can't ABX but nevertheless feel? (I'm not sure where I lie on this matter  )
Title: 'tonal Purity'? 'musicality'?
Post by: JohnV on 2002-09-11 16:39:29
Well, what's bother me most in this "tonal purity" discussion, is that tonal purity would obviously refer to as little as possible audible quantization noise in non-transient sections of music. What does transient handling has got to do with this, like Tord and Chris Johnson imply?? How does proper transient handling possibly could automatically make an mp3-codec less tonally pure in non-transient sections of music?

What comes to Vorbis "tonal purity", before 1.0 its bitrate tripped often much too low with non-transient signals, like mouth-organ, making the sound very "tonally impure", meaning of course that the sound is harsh because of audible quantization noise due to too low bitrate. 
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 16:48:19
Yes, so is there any reason to believe that MD is more 'tonally pure' or 'musical' other than the fact that it has poorer transient handling? 

I suppose you could say that given the same average bitrate, if you used fewer bits to encode the attack of an instrument, you can use more bits to encode the steady phase.  Or if the developers neglected transient handling more, they could spend more time adding 'tonal purity'
Title: 'tonal Purity'? 'musicality'?
Post by: bryant on 2002-09-11 17:00:09
Quote
I'm sure that at some level (obvious enough) you can, but do any of you subscribe to the idea that there are some things that you can't ABX but nevertheless feel? (I'm not sure where I lie on this matter  )

I subscribe to the idea that there might be some things that you can't ABX but that actually affect the listening experience. Perhaps someday these will be accurately measured (be means other than ABX). Until then, however, it is impossible to prove either way whether such effects exist.

This was explored in great detail in the later part of the following thread, including a couple posts by my brother who is getting his PhD in cognitive psychology.

http://www.audio-illumination.org/forums/i...3abebcd0f71c0b7 (http://www.audio-illumination.org/forums/index.php?act=ST&f=1&t=2456&hl=duff)

It's kind of like believing in God; there's no proof either way but intelligent, thoughtful people still manage to have pretty strong opinions on both sides
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 17:01:51
Quote
Well, what's bother me most in this "tonal purity" discussion, is that tonal purity would obviously refer to as little as possible audible quantization noise in non-transient sections of music. What does transient handling has got to do with this, like Tord and Chris Johnson imply??

It's not only in non-transient sections. The interesting question is also what happens in the transient sections 

Of course, encoders like Bladeenc are simply horrible there, but the better encoders are not perfect there either. Notably MPC can(could?) be ABXed by me with what sounded like tonal problems in transients. Don't get me wrong - it still does better than other encoders there, but that's the only place it sometimes 'falls through' for me.
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 17:03:36
Quote
It's kind of like believing in God; there's no proof either way

Don't mistake faith for science.

If you want to discuss faith, there are 'audiophile' forums.
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 17:07:28
Hey! I'm an undergraduate in Cognitive Science! I would like to get in contact with your brother  ! Can you give me his ICQ or something? 

@Garf: so transients are the places where all encoders fail, even mpc? I would think the weakest link of mpc is 'tonal purity'  There have been reports of it bloating on *classical* to retain 'tonal purity'! 

So transients are the biggest trouble for all encoders, but transform coders fail because of pre-echo while mpc fails because it has not enough 'tonal purity'?
Title: 'tonal Purity'? 'musicality'?
Post by: bryant on 2002-09-11 17:21:32
Quote
Hey! I'm an undergraduate in Cognitive Science! I would like to get in contact with your brother  ! Can you give me his ICQ or something? 

That's cool! Yeah, no problem, I'm sure he'd like to hear from you. His email is gbryant "at" cats.ucsc.edu

He's on vacation in Spain for another week so don't expect to hear back right away.
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-11 17:30:57
Great, thanks B) ! Hey, what is his research field?
Title: 'tonal Purity'? 'musicality'?
Post by: NeoRenegade on 2002-09-11 17:36:21
People who subscribe to the idea of tonal purity remind me of the guy at http://www.timecube.com/ (http://www.timecube.com/)
Title: 'tonal Purity'? 'musicality'?
Post by: rjamorim on 2002-09-11 17:44:16
Quote
People who subscribe to the idea of tonal purity remind me of the guy at http://www.timecube.com/ (http://www.timecube.com/)

Huh... could somebody explain me what the heck is that?

I read one third of the page, and gave up.
Title: 'tonal Purity'? 'musicality'?
Post by: NeoRenegade on 2002-09-11 17:51:01
It's either one of the stupidest/funniest internet jokes of all time whcih the creator maintains even though it's a total waste of time, or the actual beliefs of a stark raving mad imbecile.
Title: 'tonal Purity'? 'musicality'?
Post by: ErikS on 2002-09-11 18:03:29
Quote
Well, what's bother me most in this "tonal purity" discussion, is that tonal purity would obviously refer to as little as possible audible quantization noise in non-transient sections of music. What does transient handling has got to do with this, like Tord and Chris Johnson imply?? How does proper transient handling possibly could automatically make an mp3-codec less tonally pure in non-transient sections of music?

I guess people just guess that tones must become 'impure' because you trade frecuency precision for time precision when swiching to short windows. So as Garf said the tonal impurity should be in the attack itself. I only wonder if it is really possible to detect warbling (or whatever it should sound like) in such a short period of time when there is a (broadband?) impulse at the same time?

Anyway I never heard such a thing, and until someone can point out a sample where it is easy to hear, I don't bother... Pre-echo an the other hand is really annoying sometimes.
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 18:05:41
Quote
I only wonder if it is really possible to detect warbling (or whatever it should sound like) in such a short period of time when there is a (broadband?) impulse at the same time?

Anyway I never heard such a thing, and until someone can point out a sample where it is easy to hear, I don't bother... Pre-echo an the other hand is really annoying sometimes.

I posted samples and ABX results about this in the past, you should be able to find them in the archive.
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 18:11:10
Quote
@Garf: so transients are the places where all encoders fail, even mpc? I would think the weakest link of mpc is 'tonal purity'  There have been reports of it bloating on *classical* to retain 'tonal purity'! 

So transients are the biggest trouble for all encoders, but transform coders fail because of pre-echo while mpc fails because it has not enough 'tonal purity'?

I think you misinterpreted what I said. MPC does excellent on transients. I have never heard any problems with it's 'tonal purity' (gah) either in normal music.

But in the few samples (like, 2 or 3) where I was able to ABX an MPC --standard (with a previous version) clip (out of the hundreds I tried), it was on a transient and it sounded like a tonal problem.

Comparing this to a transform codec is silly. MPC does much better there on every sample. The clips where MPC failed were clips where the transform coders had much much worse problems.
Title: 'tonal Purity'? 'musicality'?
Post by: ErikS on 2002-09-11 18:14:24
Quote
I posted samples and ABX results about this in the past, you should be able to find them in the archive.

whhooo... Any hints on what to search for? Both "ABX" and "MPC" would generate a lot of hits. (If you still remember anything of what you wrote...)
Title: 'tonal Purity'? 'musicality'?
Post by: ErikS on 2002-09-11 18:21:27
Quote
But in the few samples (like, 2 or 3) where I was able to ABX an MPC --standard (with a previous version) clip (out of the hundreds I tried), it was on a transient and it sounded like a tonal problem.

Is it the same problem Guruboolez heard with castanets? A small "piiu" after the castanet click. Or hihat sample like the one l3maniac mentioned a long time ago?

Edit: sorry 'bout the misspelled nick and the wrong quotation...
Title: 'tonal Purity'? 'musicality'?
Post by: guruboolez on 2002-09-11 18:24:46
Quote
Is it the same problem Gurublooz heard with castanets? A small "piiu" after the castanet click.

Not really after, but inside, or around/inside the attack.
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 18:27:09
Quote
Quote
I posted samples and ABX results about this in the past, you should be able to find them in the archive.

whhooo... Any hints on what to search for? Both "ABX" and "MPC" would generate a lot of hits. (If you still remember anything of what you wrote...)

Bleh. I couldn't find them in the archive myself
Title: 'tonal Purity'? 'musicality'?
Post by: JohnV on 2002-09-11 20:37:41
Quote
It's not only in non-transient sections. The interesting question is also what happens in the transient sections

Yes, but I was talking what people like Tord and Chris Johnson mean. They claim that you can either have good transient accuracy or good "tonal purity", but not both. But bad transient handling in no way makes a transient more "tonally pure", bad transient handling makes transients sound like s*it.
Tord and Chris Johnson obviously only refered to non-transient part of the signal, because there's no way anybody can say pre-echo makes something sound "tonally pure" in any sense..
Title: 'tonal Purity'? 'musicality'?
Post by: ErikS on 2002-09-11 20:43:12
Quote
Yes, but I was talking what people like Tord and Chris Johnson mean. They claim that you can either have good transient accuracy or good "tonal purity", but not both. But bad transient handling in no way makes a transient more "tonally pure", bad transient handling makes transients sound like s*it.
Tord and Chris Johnson obviously only refered to non-transient part of the signal, because there's no way anybody can say pre-echo makes something sound "tonally pure" in any sense..

Why not?
Title: 'tonal Purity'? 'musicality'?
Post by: Dibrom on 2002-09-11 21:35:59
Because for one, Pre-echo doesn't sound "pure", and it doesn't sound "tonal". 

Pre-echo usually sounds noisy, unstable (a bit warbly kind of), sometimes static-like, or it just gives the attacks a generally "gritty" and kind of dirty sound.

It doesn't in any way fit into any sort of definition I could imagine for "tonal purity".

The whole concept of tonal purity and this nonsense that Tord and Chris Johnson have been pushing is based simply on a misunderstanding of psychoacoustics and of lossy audio encoders which uses these principles.

As JohnV said, Tord came up with this theory to try to downplay Blade's weaknesses.  You can see this in the same manner in which he talks about joint stereo, etc.

And you can see this in Chris Johnsons tests because due to the methodology he uses, you can see that he completely misses the necessary colleration between psychoacoustic principles and how these affect the audibility of lost information (through selective degredation).  Instead, he tries to draw a flawed correlation between the visibility of "flaws" in low resolution spectrograms and the audibility of lost information.

See the problem here?  He's not testing aural quality, he's testing visual quality, and he somehow comes to the conclusion that whichever codec looks the best in a certain situation must feature the most "tonal purity". 

Yeah.. ok. 
Title: 'tonal Purity'? 'musicality'?
Post by: JohnV on 2002-09-11 21:42:31
Quote
Why not?

Why Not?? 
Because pre- or post-echo is audible distortion. There's nothing "tonally pure" in distorted transients, on the contrary. I can't think of that distorted transient with excess audible noise would be "tonally pure", that's the last thing I'd describe it.. 
Title: 'tonal Purity'? 'musicality'?
Post by: Garf on 2002-09-11 21:45:42
Quote
Because for one, Pre-echo doesn't sound "pure", and it doesn't sound "tonal".  

Pre-echo usually sounds noisy, unstable (a bit warbly kind of), sometimes static-like, or it just gives the attacks a generally "gritty" and kind of dirty sound.

It doesn't in any way fit into any sort of definition I could imagine for "tonal purity".

I more or less disagree. Even in a wide spectrum impulse sound (i.e. some kind cymbal) you brain will still be able to associate a certain tone or frequency with this sound. If this changes after encoding, I'd define it as a 'tonal impurity'  (a better word is much appreciated)

Edit: Uh, misunderstanding. OF COURSE preecho doesn't sound tonally pure. But even an impulse sound can have some tonality, which is what I was trying to explain.
Title: 'tonal Purity'? 'musicality'?
Post by: Dibrom on 2002-09-11 21:55:49
Quote
Edit: Uh, misunderstanding. OF COURSE preecho doesn't sound tonally pure. But even an impulse sound can have some tonality, which is what I was trying to explain.

  Yeah, I wasn't trying to say that impulses don't have tonality.  Rather, only that pre-echo is not "tonally pure".
Title: 'tonal Purity'? 'musicality'?
Post by: JohnV on 2002-09-11 21:56:59
Quote
Edit: Uh, misunderstanding. OF COURSE preecho doesn't sound tonally pure. But even an impulse sound can have some tonality, which is what I was trying to explain.

Yes, impulse sound can have tonality, but when it's affected by pre-echo, it's obviously not as tonal anymore for a listeners, since pre-echo-noise is now making it sound distorted and less tonal, maybe even totally different sounding. 
Title: 'tonal Purity'? 'musicality'?
Post by: Joe Bloggs on 2002-09-12 08:17:19
Quote
I think you misinterpreted what I said. MPC does excellent on transients. I have never heard any problems with it's 'tonal purity' (gah) either in normal music.

But in the few samples (like, 2 or 3) where I was able to ABX an MPC --standard (with a previous version) clip (out of the hundreds I tried), it was on a transient and it sounded like a tonal problem.

Comparing this to a transform codec is silly. MPC does much better there on every sample. The clips where MPC failed were clips where the transform coders had much much worse problems.

I know, mpc is much better than most any codecs out there qualitywise  But what I was trying to say is that it turns out that the places where mpc has troubles (although it usually overcomes them  ) is not as different from trouble spots for transform codecs after all
Title: 'tonal Purity'? 'musicality'?
Post by: 2Bdecided on 2002-09-12 09:09:31
My ears exhibit excellent transient response, and a high degree of tonal purity!

 

David.

http://privatewww.essex.ac.uk/~djmrob/thes...is/2.html#2_3_4 (http://privatewww.essex.ac.uk/~djmrob/thesis/2.html#2_3_4)
"...Unfortunately, this measurement is almost useless for audio quality assessment...."
Title: 'tonal Purity'? 'musicality'?
Post by: wkw on 2002-09-12 16:06:26
Yes attacks processing does have an impact on tone purities. There is a few official tests clips that specifically tailored for this. The clips Glockenspiel, gspi35_1 & gspi25_2 and also mc_sich2 is used to detect distortion of tone components when switched to short-block.. Of course, castanets is a bad test sample for tone purity during transients.

Short Blocks generally have bad tone separations compared to long block.
From fft analysis of audio signals, tone isolation decreases as the length of the fft decreases. 

For AAC, when switched to 8 short blocks, the pre-echo noise is no longer caused by all 8 short blocks but only 1 of the short blocks.. As a results the remaining short blocks could have very tonal components which could be wrongly "squashed".. Maintaining tonal purity during transients is a very delicate tasks.

To be more accurate, maintaining tone purity in region very near to the actual attack is a concern here..


wkw
SimplePortal 1.0.0 RC1 © 2008-2021