HOLY sh*t (by Bill Evans)

[Flecko]

I have (and always had) reasons to believe that MP3 might be better than uncompressed. And now, in the era of "HighRes is nothing", it kind of fits a few technical theories.

Quote

I listen to MP3's a lot because sometimes it's all I have, but don't really know what the signature of MP3 is.

I would say MP3 lose details, especially in the highs. They sound more dull than an original cd or flac. They do not sound bad, if they are made well. Just not as good as the original.

Sorry Peter... I only had the time to post the result of my comparison and give a quick opinion about the mp3 sound to GerardA. I didn't "planed" to argue against you. I just didn't read what you wrote about mp3 before.

[PeterSt]

Hello Adrian - Your observations are perfectly normal IMO. I just have some ideas. Maybe it doesn't work out at all ...

Peter

[crisnee]

Thanks Peter, for your in depth thoughts/explanation.

The thing about the high frequencies (I mean really high 20khz-100khz) was that they apparently made people feel better when they were present as opposed to not present. This was determined by some kind of a test (I think you were sort of referring to it) several years ago in Japan I think.

These high frequencies were the real thing (I'm supposing--I can't remember where I read the paper), as opposed to the distortion you're referring to in most (all?) present day hi-res recordings. So, in theory at least, hi-res (or call them recordings with extra-aural high frequencies present should be more enjoyable [if not discernable by ear] than nor-res recordings).

I'm not stating this as the case, but just as what seems logic would dictate. It's kind of an odd thing, as one doesn't hear it, but it does register as more pleasurable in the brain. Does this make sense to you? Or am I missing something? (I'm assuming you know about this test).

-Chris

[PeterSt]

Hi Chris,

Yes, those tests are known. The last one I read, I read with special care, thinking of what material they actually would have been listening to, but f what I recall from that test, is that some high frequency test tones were superimposed on normal music. Yes, people could perceive that, but that is different from better music.

It is my estimate that anyhow high frequencies will influence mind state; it may activate parts of the brain; make it more awake. At saying this I now recall it was exactly about that (scans of brain activity included).

The funny thing is, if we'd look back at those tests and try to determine the quality of the HiRes material the attenders would have been listening to, all is moot (by now).

Peter

PS: At least this article (Japanese test I think) looks like it (didn't read it, but looks interesting at a glance anyway) : http://jn.physiology.org/content/83/6/3548.full

Edit : Here is the PDF of that (reads better) : http://www.brainmusic.org/EducationalActivitiesFolder/Oohashi_HFCs2000.pdf

[PeterSt]

One more :

http://www.hificritic.com/downloads/Archive_A10.pdf

If you really want to know, this is (I think) the result of that Japanese study, now summarized in 347 pages 5-6 years after. Can't find the whole (e)Book, but I guess it can be bought somewhere.

This may (per today !) be more important than it seems, because it looks like a "tweak" for better or worse sound. Remember, those high frequencies just *are* in there, no matter what. This again needs more explanation, but what has been filtered beyond say 20Khz, will come back at you way more further in the frequency band. Think like that whatever is filtered out can't be thrown away. It has to be collected somewhere, and there it will nicely play for you again.

This is why it *is* important how your amplifiers deal with it, and next how your speakers do, once the amplifiers done there job. This now is why analogue filteres behind everything will be as important. Mind you, for better or for worse, and the "better with no filter" comes up when the brain would be stimulated by the high frequencies which *are* correlated to the music.

Peter

[crisnee]

Great, I was going to try to find the article(s), because it seems they might have some info that might apply to your thoughts, or be useful. So now I don't have to search for them, thanks.

By the way, Teresa doesn't listen to mp3's. I did some searching to see what her whole thing was about. I was really interested to read her rational for listening to hi-res and mp3 while excluding redbook. Unfortunately she excludes mp3's too. You must have misread something, or I don't know what. But too bad anyway.

-Chris

[PeterSt]

No no ... you are correct. She dismisses MP3 just the same, but found those listenable opposed to uncompressed Redbook (but not a preferred choice, just as I about never listen to MP3 (surprise)).

For others : she is  a Positive Feedback (review) writer, which btw to some only degraded that review site.

[crisnee]

Peter, this is the current conclusion from all the studies quoted from the Martin Collom article.

>>"Taken overall, and with some polite hesitation from the authors, the Nishiguchi results (These tests were done to further and even more carefully investigate/test what Oohashi tested--Chris) contradict Oohashi, (Oohashi is the Japanese test that initially confirmed sensitivity to 20khz--100khz and its pleasure enhancing quality --Chris) and in practical terms reiterate that 20kHz is entirely sufficient as a nominal limit for sound reproduction. This later result is attributed in part to the use of very steep band filtering in the experiment, to firmly separate the ‘audible’ energy band from the ‘inaudible’. Advance notice has now been given of further work by Oohashi et al9 which suggests that his previously reported phenomenon actually requires that the body and not just the ears be exposed to the ultrasonic sound field. If the body shielded from the ultrasonic component, then the perception of an extended bandwidth is no longer reported."<<

And here is Collom's conclusion in his own words

>>"Considering the evidence, it would seem that the case for a wider sound reproduction bandwidth is not proven.
 
3 EXTENDED BANDWIDTH, MASTERING AND REPLAY
Regardless of the audibility issue you may still argue advantages for the proposition. For origination a wider operational bandwidth may result in improved performance, or the equivalent of increased resolution, benefiting the established audible range. At present strong low pass filtering
is often employed to fit a 20kHz bandwidth to many digital audio chains especially MP3 portables, broadcast, and CD. The use of a recording medium of extended bandwidth makes possible more gentle filtering, with improved phase and impulse characteristics. For loudspeakers the effort to
extend bandwidth again may improve quality in the lower frequency range. Designing for better bandwidth should provide superior power response to higher frequencies, improving off-axis responses and consequently, sound quality. In addition diaphragm resonances that may have been closely proximate to the ‘audible’ band, may now be located well out of range. In Figure 2 the typical
low loss, multi-resonant output of a rigid ‘piston’ dome driver is compared with that for a wideband pure diamond type, clearly moving such problems well out of the way.
 
4 CONCLUSIONS
As an industry we need to maintain a healthy skepticism concerning marketing based performance claims of all kinds. Human perception is notoriously difficult to quantify especially when differences are small, and some results may be misleading. I believe that elements of the audio industry have exaggerated the necessity for the consumer to enjoy an extended reproducing bandwidth. Judging from this review of the subject, the case for an ultrasonic reproducing bandwidth is not proven, and 20kHz remains the practical limit for the chain as a whole.
 
However for origination and post-production, within reason, we should be encouraged to use the widest bandwidth possible to preserve the greatest information content. This will maximise archival quality."<<

Peter, I realize that you can read this from the links, but I thought others might be interested, and it's really not necessary that one dig through the details unless one has a deep interest in the subject.

-Chris

[crisnee]

So, what you got against mp3's anyway? I guess kids got no taste, but at least they got good ears.

-Chris

[PeterSt]

So, what you got against mp3's anyway?

Haha, that I should be against it, because we all know it is "compressed", which can not be for the better.

But once different reasons pop up (made up by myself), it becomes another matter. Or at least it can ...

Peter

[crisnee]

Alright I have the answer to it all--well if the Japanese had not left out one critical aspect to their test anyway.

If you're right Peter about the mess above 16khz and the mp3 solution, and the Japanese are right (sort of) about the pleasure of high frequencies. Ok, I know they refuted the pleasure, but only without the body. Last I heard, we do still come with bodies. So here's the solution:

A quasi mp3/16khz filter for the music, plus a random hi frequency generator played through a special separate super tweeter or two. The highs pleasure the body, the mp3 the ears, voila, musical heaven. Of course this would kill the hi-res business.

I wish they'd included a test with random high frequencies, because my guess is that it didn't matter what they were. After all the hi freq played by themselves registered pleasure in the brain, and the brain doesn't know that they belonged to the sound of the gamelan.

I really have to go to bed, I'm getting too carried away.

-Chris

[PeterSt]

Chris, thank you very much for this nice summ up of the conclusions. And no, today I did not read them again, but I did before (and had forgotten about them already).
But they *are* intersting to say the least.

Ok, I hope I am allowed to put a few things into "today's" (my own) perspective. Look :

This later result is attributed in part to the use of very steep band filtering in the experiment, to firmly separate the ‘audible’ energy band from the ‘inaudible’.

This is why I said in the earlier post that we may wonder what they were listening to. Btw, it is explicitly brought up by Colom himself in the last part of your post as well;

If this were to be about high transient music (I didn't read back on what exactly has been used), then these transients will have gone. See my graphs elsewhere. This by itself already implies two facts of importance :
1. The resulting now sine-like transients may be perceived as "more nice", although they would not represent the music.
2. When the impulse response (high transients) would have been sustained (which they were obviously not), everything can have happened, from WOW's (NOS1, sorry !) up to OUCH ! (sheer distortion because of not being able to cope further chain elements).

I won't say this makes all moot, but the test would be completely flawed in its base if only I have the right on my hand (which of course *I* presume).

Additionlly it would be obvious that any high transient stuff (beyond the audible band or not) can be perceived more easily. Within itself on one side, and because of being distortion when not done / worked out right on the other.
Also, I can't emphasize enough how much "different" the NOS1 sounds (which is today), and which for me obviously is to be compared to whatever they used back then. Additionally (and don't let this be a commercial !) I by now *know* what happens when normal filtering is let loose on music, which makes it so unlisteneble (again, by now and for me) that whatever will sneak in via the high frequency realm would make me dead sure such a test can *never* had worked. Never.

For origination a wider operational bandwidth may result in improved performance, or the equivalent of increased resolution, benefiting the established audible range.

This too is what I referred to earlier;
We know by now that *no* HighRes ofr good (digital mastering) existed at the time. So, the conclusion looks theoretically ok, but practically it could not have happened.
I must add to this that this will have been SACD, which is fairly different from DVD-A which we encounter these days (the HDTracks downloads). But *now* I don't know of any self respecting high end audio shop which found SACD to outbetter CD (after SACD being around for several years), and which is fairly different for opinion from all the shouting we read on the internet. I again refer to what I said about this all, everybody raving and such, until people started to produce AudaCity graphs about it ...

At present strong low pass filtering
is often employed to fit a 20kHz bandwidth to many digital audio chains especially MP3 portables

Good for me ! hehe

The use of a recording medium of extended bandwidth makes possible more gentle filtering, with improved phase and impulse characteristics.

And so I leave the filtering out at all.
No, this is not without anomalies within itself.

Judging from this review of the subject, the case for an ultrasonic reproducing bandwidth is not proven, and 20kHz remains the practical limit for the chain as a whole.

This surely needs to read back on the underlaying arguments, but it is what I am saying (but read back on those arguments !). All I add to it, is that 16KHz is far more practical, while at the same time 20KHz is some stupid nice even limit at the same time. Why not 20.3 or 19.55.
Do the test on your hearing abilities.

I'd go as far as trying a gentle normal filter which starts at 16KHz, but the problem is that any gentle filter will gently sneak into way lower frequencies along the way. But ok, since it is all easily measureable, it's easy to try. In the end that more gentle filter could start at 20KHz, so net it starts at 16KHz.

There is one additional matter which may only be about my own ignorance :
It is said that these common filters only ring in the area which is filtered out (for net result). I don't believe that, because I haven't seen anything that comes close to sustaining those transients, whatever I tried earlier on. This probably will be about the math that forms a transient, and where the steeper the transient, the more high frequencies it needs. This by itself implies that any higher frequencies filtered out will kill the steep transients, and there's no way out. So, now this is not about ringing anymore, but about the impulse still turning into a sine like thing.
If this is true, it still leaves the way MP3 filters as an alternative, to at least investigate.

Peter

[PeterSt]

I wish they'd included a test with random high frequencies, because my guess is that it didn't matter what they were.

Exactly. Or at least that is one piece of that puzzle I was talking about; People clearly (meaning : I sure believe that) perceive HighRes as better, which as clearly is about distortion beyond the audio band.

Indirect empirical proof.

Thanks and good night !
Peter

PS: And you may rethink about the (16/44.1) NOS/Filterless belief in general. Now *that* is distortion. But still. And it *is* music correlated. It may do something too. Something which may be fairly natural, net.

[PeterSt]

There is a related subject here too :

A few years back I started to talk about harmonics being able to be recreated in mid-air. Another strange thing, but back then I could hear it happening (after some improvements). Today I am not sure anymore whether this ever can be true (for exactly how I layed it out back then), but something else might be going on. And the twist for today is : in our brains.

If we take the "white papers" springing from Nyquist/Shannon Theorem (all what is band limited can be reconstructed to it's sheer original analogue infnite), possibly this can be worked out by our brains themselves. I won't (nor possibly am not able to) work this out here, but if you read those papers ...

... Then maybe it is possible that any too high frequency for our EARS to begin with, -and which must be about some actively operating filter !! (which is different from explicitly)- may work out the same way as we try to do that digitally.

This time though, it will be an analogue filter (which does not ring), and its steepness is determined by ... well, not even important (but may be phase related). It will be a natural thing. The point is, what is filtered here will be the harmonic result of the too rough stepping, that by itself being formed by high frequencies (because any square is formed by sines of way higher frequency). This does not mean the squares are filtered out (call that not-explicitly), but just the high frequencies are, with that the squares vanishing. Maybe completely, maybe to some degree (larger sines being the result on the latter).

When this really would be happening, there's no imaging behind the audio band to our ears (but our equipment will be as bugged by it) but more importantly, there will also no aliases *in* the audio band because of that (not being there in the first place).

What about that eh ?


PS: For those who don't know : I have always been intrigued by those liking NOS/Filterless, which was ever back the exact reason to buy an NOS DAC (which was 18/96 at the time), next being even more intrigued because I liked it myself, and always knew how most poor that technically measures (what about 30% THD easily !).
Officially it just can't be true that 16/44.1 NOS/Filterless sounds any good. Even with the tradeoff of ringing not being there, it seems hard to believe that we don't perceive that as sheer distortion.
By now I know that it sure makes a difference if we move the in-band aliases out of the way (things sound more natural compared to when they remain in), but not as much as measurement seems to depict.

Ok, back to work now.

[Flecko]

Exactly. Or at least that is one piece of that puzzle I was talking about; People clearly (meaning : I sure believe that) perceive HighRes as better, which as clearly is about distortion beyond the audio band.

I think one point is, that people might find hires better, because they think it should be better. Also, maybe they have better equipment for the hires files. I heard one guy saying he likes the DVD-A version of the black album the best, by far. The point is, that this specific version is more compressed then the others...So it cannot sound better and it doesn't.
If we look at the music material (records mastered to death) and what we can hear, we do not realy need the additional resolution and a higher bandwidth. But in a firts place good records and good equipment (A trivial statement, but nontheless true).