OK, so REALLY, 44.1khz was never enough!

[Jon Hodgson]

Originally Posted by karyoky****

hi,

i am not sure i understand the reason why it would be less flat in a 96kHz bandwidth, or what significance that may have.

and i guess it matters which spectrum has more influence, but i don't understand if and / or when it is better for one or the other to dominate.

hopefully this is not completely unrelated to any possible topic.

Look at the spectrum of the pulse (third graph in the sequence I sent).

See how the line starts off almost horizontal then starts drop off, getting steeper as you go up in frequency till it hits zero (and then you get another bump and then another etc)?

So if you take just the very first part (20kHz worth), then it's as near as dammit flat, take a wider bit (96kHz worth) and you get a bit more of the drop.

[Jon Hodgson]

Originally Posted by karyoky****

hi,

yeah, i see that, i just don't see why it would be like that. of course i haven't really thought about it, either, so...............

is it because its not an actual theoretical impulse?

It is the shape it is in the frequency domain because it is the shape it is in the time domain, that's just how it is, you change the shape in one, you change the shape in the other.

A theoretical impulse would be flat to infinity, a rectangular pulse has a sinc shape in the frequency domain (which when plotted as a power spectra, which is the most useful view in this case, comes out like what you see).

For low frequencies the pulse's frequency spectrum is close enough to flat that it will serve as a reasonable substitute for the theoretical impulse.

is it because high frequencies represent "more energy" than low frequencies?

No

is this something i should worry or care about?

Not beyond understanding that for the frequency range we're interested in, that pulse contains pretty much equal amounts of every frequency from DC to the top.

[Fokus]

Originally Posted by Ethan Winer

As I understand it, the Kuncher paper says interchannel timing differences as small as 5 microseconds can be discerned in some cases by some people. He concluded that 44.1 KHz is not high enough to preserve that small a timing. JJ then showed with 'scope photos etc that 44.1 KHz is in fact high enough.

Sadly, that is not what Kunchur's alignment paper (as opposed to the 'filtering paper') said.

When the web community was alerted on this paper, oh it seems such a long time ago now, I was guilty too of immediately assuming that this was about interaural time differences. But then I read the thing, and no, it is about alignment differences between two vertically stacked sources, excited with a continuous 7Hz square.

But many made the incorrect assumptions of interaural delays, and that is IMO what caused so much of the misery at the Stereophile forum (apart of course of that number of, erm, uneducated and uneducatable specimens of homo sapiens). I'm afraid JJ made the same erroneous assumption...

stikestikestike

--

Now what I would like to see discussed, in serenity, is how both of Kunchur's experiments, a) 5-us misaligned squarewave sources and b) 5-us first-order filtered triangles piped into headphones, could lead to the
claim that because of people discerning down to 5us in both experiments the 44.1kHz sample rate is proven to be insufficient. Quite a leap if you ask me.

On an additional note, neither paper mentions the case of two pulses separated by 5us, but it turns up in the attacks, and then in Kunchur's FAQ where he claims again that 44.1kHz is not enough because it cannot keep the two pulses separated. Once again: please tell me about the aural relevance of these two pulses.


Werner

[Jon Hodgson]

Originally Posted by Fokus

Now what I would like to see discussed, in serenity, is how both of Kunchur's experiments, a) 5-us misaligned squarewave sources and b) 5-us first-order filtered triangles piped into headphones, could lead to the
claim that because of people discerning down to 5us in both experiments the 44.1kHz sample rate is proven to be insufficient. Quite a leap if you ask me.

On an additional note, neither paper mentions the case of two pulses separated by 5us, but it turns up in the attacks, and then in Kunchur's FAQ where he claims again that 44.1kHz is not enough because it cannot keep the two pulses separated. Once again: please tell me about the aural relevance of these two pulses.

Kunchur presents his papers as being about temporal resolution.

Yet the one attempt at an explanation of audibility of the tests that I've read in his paper (experiment 1), is not about temporal resolution, it is about non-linear amplitude responses and audibility of the artifacts thus generated (it looks a bit weak even then).

So the reasoning seems to be all over the place.

[jupiter8]

The really hilarious thing about this thread is that the articles the OP choose as "proof" for his theories has absolutely nothing to do with the subject.

Makes as much sense as Why 44.1 kHz isn't enough

[Jon Hodgson]

Originally Posted by jupiter8

The really hilarious thing about this thread is that the articles the OP choose as "proof" for his theories has absolutely nothing to do with the subject.

Makes as much sense as Why 44.1 kHz isn't enough

The cookie of the day is a turnover?

[Axon]

Originally Posted by Fokus

When the web community was alerted on this paper, oh it seems such a long time ago now, I was guilty too of immediately assuming that this was about interaural time differences.

O RLY? Which forums were alerted to this prior to Stereophile? (I'm curious)

On an additional note, neither paper mentions the case of two pulses separated by 5us, but it turns up in the attacks, and then in Kunchur's FAQ where he claims again that 44.1kHz is not enough because it cannot keep the two pulses separated. Once again: please tell me about the aural relevance of these two pulses.

The apparatus used in one of his tests - the speaker-misalignment one - can be modelled as an FIR filter with the 5us-spaced pulses. That said, yeah, everybody involved has been unusually quick in arguing about the pulses themselves, rather than the filtering involved. The two situations are analogous but they are not quite the same thing, in terms of the evaluation of the actual theses.

[Axon]

Originally Posted by Jon Hodgson

Kunchur presents his papers as being about temporal resolution.

Yet the one attempt at an explanation of audibility of the tests that I've read in his paper (experiment 1), is not about temporal resolution, it is about non-linear amplitude responses and audibility of the artifacts thus generated (it looks a bit weak even then).

So the reasoning seems to be all over the place.

You have no idea. In particular... check his trig. And then re-run it against all the actual data in the papers.

[Fokus]

Originally Posted by Axon

O RLY? Which forums were alerted to this prior to Stereophile?

I don't read Stereophile ;-) and I honestly don't remember at what place I got notified, but it seems an awful long time ago.

Check: S'phile started end of May. That IS an awful long time ago ;-)

Originally Posted by Axon

can be modelled as an FIR filter with the 5us-spaced pulses. That said, yeah, everybody involved has been unusually quick in arguing about the pulses themselves,

Sure, but K initially even omitted that, if I remember correctly from the alignment paper (it is such a tedious read that I won't like to delve back into it, thank *you* for that).

Werner

[Ethan Winer]

I have nothing of substance to contribute to this thread, but I wanted to say that I'm enjoying it immensely.

--Ethan

[Teddy Ray]

Originally Posted by Ethan Winer

I have nothing of substance to contribute to this thread, but I wanted to say that I'm enjoying it immensely.

--Ethan

Provocateur fuuck

[j_j]

Originally Posted by Jon Hodgson

Or to put it another way...

Unless you're planning a change of career to working in developing communications equipment, it's highly unlikely you need to worry about complex sampling, it exists, but so do moon rocks, and when's the last time you stubbed your toe on a moon rock?

Well, it's key to understanding how PQMF's, QMF's, and MDCT's work, and those filterbanks are kinda common in the audio coding realm.

Just for instance :p :p

[j_j]

Originally Posted by karyoky****

hi,

i am not sure i understand the reason why it would be less flat in a 96kHz bandwidth, or what significance that may have.

and i guess it matters which spectrum has more influence, but i don't understand if and / or when it is better for one or the other to dominate.

hopefully this is not completely unrelated to any possible topic.

No better, no worse, just "is".

[Jon Hodgson]

Originally Posted by j_j

Well, it's key to understanding how PQMF's, QMF's, and MDCT's work, and those filterbanks are kinda common in the audio coding realm.

Just for instance :p :p

Sure, and non baseband sampling can pop up too, so I'll confess to being a little extreme in my dismissal (especially since I've used the above in audio dsp myself).

But my point was that karyoky**** really doesn't need to concern himself with them for his needs, he's never going to develop a DSP algorithm, it's hard enough getting people to understand what they need to know without throwing in stuff they don't (more often than not at this level it leads to more misunderstandings and misconceptions).

[Bob Olhsson]

Frankly, I don't trust audio DSP developers to get their math right and I think healthy skepticism is the only thing that has improved the quality of digital audio.

[Jon Hodgson]

Originally Posted by Bob Olhsson

Frankly, I don't trust audio DSP developers to get their math right and I think healthy skepticism is the only thing that has improved the quality of digital audio.

What we're talking about here is not the maths of audio dsp developers, it's far more fundamental, wide ranging, and tried and tested than that.

Healthy skepticism is good, people should no more unquestionably accept the work or words of a dsp developer than they should that of an analog designer, or indeed that of a mastering engineer.

But what some people think is healthy skepticism is actually not understanding and getting it wrong.

[j_j]

Originally Posted by Bob Olhsson

Frankly, I don't trust audio DSP developers to get their math right and I think healthy skepticism is the only thing that has improved the quality of digital audio.

HUMPH! says the DSP guy.

But, you know, you do have a germ of truth in there, speed optimization types often throw out the baby with the bathwater.

[j_j]

Originally Posted by Jon Hodgson

But my point was that karyoky**** really doesn't need to concern himself with them for his needs, he's never going to develop a DSP algorithm, it's hard enough getting people to understand what they need to know without throwing in stuff they don't (more often than not at this level it leads to more misunderstandings and misconceptions).

I understand that, but there are a couple of people out there who will use the lack of mention to continue to spam their victims with "he or she is an idiot" if you don't at least introduce the concept.

There are some people out there who seem to have a vested interest in misteaching as much as possible, and they seem to be annoyingly persistant.

[j_j]

Originally Posted by karyoky****

hi,

i actually thought it was proper for jj to mention that there are some exceptions, and then to briefly explain how they didn't apply in the context.

Well, if you move on to 'how does MP3 work', presto, there you go, integer-band sampling with bandpass filters that isn't baseband sampling except in one band, followed by orthonormal filterbanks on top of all that.

To that we add the property of local aliasing cancellation ala QMF's. This can make your head hurt horribly the first time you see it.

[Jon Hodgson]

Originally Posted by j_j

I understand that, but there are a couple of people out there who will use the lack of mention to continue to spam their victims with "he or she is an idiot" if you don't at least introduce the concept.

There are some people out there who seem to have a vested interest in misteaching as much as possible, and they seem to be annoyingly persistant.

I can't say I disagree, but at the same time the chunks of my life that I've lost due to a carelessly placed "well you could theoretically also do x" or similar leading to someone latching onto it and requiring a long and often futile attempt to explain both the concept and why it's not relevant don't bear thinking about.

Damned if you do, damned if you don't.

[jayfrigo]

Mod note: Thanks to all for bringing this thread back from the brink. It has turned out to be informative, and actually touched on info not usually seen in these types of threads. JJ, I'm glad to see you here; welcome! Anyway, carry on.

[Jesse Graffam]

Originally Posted by uncajesse
I disagree. Hard analog clipping has the exact same problems with distortion. As far as aliasing...

Originally Posted by j_j

Analog clipping does not intrinsicly alias.
Digital clipping does.

BIG difference.

I was specifically not referring to aliasing, by talking about that in the paragraph after that you quoted, ala "As far as aliasing" dot dot dot. That could be confusing on second thought, sorry about that. I was talking about distortions that were not aliasing in that sentence.

Originally Posted by Jon Hodgson

You were, but uncajesse wasn't :-)

Correct. Analog being incapable of this, it's totally unfair to compare the two on this basis, and I was pointing out that more advanced forms of "basic" clipping exist in both digital and analog form. I wanted to point out that the comparison wasn't correct without making too big of a side-track into it.

Originally Posted by Fokus

Now what I would like to see discussed, in serenity, is how both of Kunchur's experiments, a) 5-us misaligned squarewave sources and b) 5-us first-order filtered triangles piped into headphones, could lead to the claim that because of people discerning down to 5us in both experiments the 44.1kHz sample rate is proven to be insufficient. Quite a leap if you ask me.

On an additional note, neither paper mentions the case of two pulses separated by 5us, but it turns up in the attacks, and then in Kunchur's FAQ where he claims again that 44.1kHz is not enough because it cannot keep the two pulses separated. Once again: please tell me about the aural relevance of these two pulses.

You make some good points there.

It's almost comical how none of the mayhem (mainly on Stereophile) would have happened if Kunchur would have bothered to hook up an ADC/DAC and tried auditioning several sampling rates through them with the same tests to see if and what sampling rate people stop hearing the difference.


And yeah, I'm with jay... nice to see how differently this has turned out than the incidious Stereophile version.