| Columns Retired Columns & Blogs |
| Loudspeakers Amplification Digital Sources | |
Analog Sources Accessories Featured | |
Music Columns Retired Columns | |
Show Reports | |
Features Latest News Community | |
Resources Subscriptions |
© 2024 StereophileAVTech Media Americas Inc., USA
All rights reserved
You claimed that one can see 5us from a 16/44 machine, now you are changing your story?
Secondly, your little math/graph example tells nothing about whether the graph is actually true. We have caught others he manipulating room graphs, then changing setups from two speakers to one speaker, then claiming that the mic, which was close to two 1k reference mics, was suddenly 19db off etc. So we don't know if your graph is accurate or manipulated. But we do know something from the picture below, courtesy of Pyramix.
As far as your propagation comments, the recording mics are usually near the instruments, not 20 feet way, although this can occur. Enough said.
I do have a pix to show the audience/public, courtesy of Pyramix. Here is comparison picture with a 3us input pulse applied to various machines. Next to the analog input pulse is a 16/48khz cd analog output.
Notice the 16/48khz is severly lacking in reproducing a 3us pulse. Since a filter is not 100% effective the narrower tip, a little output of some higher frequencies. The wider body portion (most of the pulse output, most of which is below 20khz) are lower frequency artifacts not present in the original 3us analog input pulse. 5us will give a little more output but not much. Don't expect 40us to be that accurate either. Notice the higher samplings to the right are better, but only DSD was able to replicate the original input pulse. So the machines produced the artifacts plus, artificial, ringing.
Anyone in the audience think 16/44 is enough for music and is the highest fidelity?
Not surprising why you refuse to sign your own name to your own posts.
Take care.
Just the facts, just the facts. By the way, remember in the "dishonesty of sighted listening" string in which I replied to arny that 4 had either not posted before, or posted for months or years before. Well I forgot Krabapple. He made his first post as well on that string. So 5 suddenly entered the string after the first post baiting the viewers in to argue.
Just thought the audience would want to know.
Which you don't meantion that you were a lead researcher at Bell labs.
http://home.comcast.net/~retired_old_jj
That, sir, is simple misinterpretation, at best. What's more, your non-quote of something I didn't say is hopelessly vague.
I have said, repeatedly, that you can get 5 microsecond TIME RESOLUTION in interchannel time delay. I've also proven it. Axon has proven it to you.
This is not the same as whatever it is that you want to mislead people into thinking I said. What you claim I said is not a quote, and is hopelessly vague and ill-constructed.
When you can finally get around to addressing the actual facts of the matter, let us know, please.
Hi Axon,
Please check the picture in my above post.
Take care Axon.
Yep, just like I've repeatedly told you.
Err, not really. I would suggest, again, that you refer to Morrison's book on Fourier Analysis. While you are not completely wrong (any filter must have some leakage), that is not the primary reason for the shape of the pulse.
Um, the lower frequencies are indeed present in the 5 microsecond (or 3, what-have-you) pulse. Again, I refer you to Morrison, where you can discover very early on the spectrum of a rectangular pulse. When the high frequencies are removed by the filter, the lower frequencies are more visible. You can see this yourself by taking any linear algebra package (matlab, octave, scipy for instance) and building up the pulse from DC upwards. This will also show you the real reason for the shape of the pulse, because you can build it then without any issue of filter leakage.
Your pretty graphs are proving what I've already told you.
What they don't do is address what ELSE I've told you about the ability to resolve two pulses (5 microseconds is not necessary, but you can use it if you want) with a 5 microsecond offset, in two channels.
If you have two of your filters, and a standard function generator that will allow you to delay a second pulse by the necessary 5 microseconds, you can prove this to yourself.
The pulses you delay by 5 microseconds do not have to be 5 microseconds long, however, you could just as well make them 5 seconds long. You'll still see the shift in the signal at the onset.
And that is what the quotes at the beginning of this thread deny, and incorrectly so.
You again confuse interchannel delay with pulse length.
Pete, there is a moderate to large legal difference between checking his home page for his name and someone claiming J_J is JJ Johnston,,, and J_J actually posting his legal name to his post.
Just thought I would clear that up in case you missed a previous post.
Take care.
Just for you know.
Thanks Michigan. I sent an email off to Dr. Kunkur a few days ago and he actually replied, which was quite gracious since I am sure he has a busy schedule and exams. He informed me that he is preparing a post for me to put up. This should be accomplished sometime next week.
I am bowing out since Dr. Kunkur can easily defend himself in his post.
Take care.
Why? Do you want to sue me for teaching you how sampling really works?
It will be enlightening to see what you've told him others have said, vs. what others have actually said, based on their reliance of your (and others) third party quotes of his work here.
Well if one has noticed, J_J goes from outright accusing Dr. Kunkur of not knowing basic sampling to attempts to discredit Dr. by using two channels and using pulses shifted by 10us. After that fails, J_J attempts to use straight 10us pulses. When that does not work, J_J attempts to use special pulses and antialiasing and other tricks to stretch the pulse.
Anything to discredit Dr. Kunkur's findings that higher bandwidth is necessary.
Let's see. First JJ made this comment.
Of course the first attempt is to discredit Dr. Kunkur by using faulty math.
Next, JJ vainly attempts to discredit by cochlea comments and then using two channels shifted 10us apart.
Finally J_J tries the old manipulated pulse trick.
What happened to the music at the beginning of the thread??
I present a comparison between 3us analog and different formats I state
and J_J states
From my earliest posts, notice I have been the one who has continually demonstrated that 5us will not be faithfully reproduced from a 16/44 machine,
yet now J_J suddenly claims he told me that repeatedly.
Well what did J_J state earlier?
Notice the flipflop, just the opposite he says now.
Hard to be honest J_J.
Well like I said, no wonder J_J continually refuses to post his legal name.
Gonna have to agree with jj 100% on this again. In particular, I think it's pretty obvious that he never stated what you are interpreting him to have said about the whole 5us thing (in terms of pulse widths.
The wider body portion represented in the 16/48 pulse is in the original analog pulse. If you run that ideal analog pulse through an analog lowpass filter that has (hypothetically) has the same frequency response and phase response as the antialias filter on the 16/48 sample, you'll wind up with exactly the same signal.
I must relish the irony of people misinterpreting of a waveform plot to justify a subjective position of sound quality that is disproven by signal processing theory. I say "people" because you're not the first person to misinterpret that Pyramix plot. It's just a thoroughly misleading plot to begin with, actually.
Sure!
The importance of the observed ringing and frequency response etc of 16/44 is largely a psychoacoustic question. And I am not aware of any controlled tests demonstrating such things are audible. But I am aware of a preponderance of evidence (both theoretical and experimental) suggesting that it isn't audible. My own personal listening experience tells me it's not audible.
"Time resolution" is a weird nebulous conecpt that people like to bandy about a lot in this field, but is surprisingly difficult to measure in a way that has meaning. You, like Dr Kunchur, and like femlid in the SH.tv thread I linked to, have this idea in your head of how to measure it without really understanding that such a measurement is meaningless in any audible sense. In femlid's case, he frankly just did not comprehend the signal processing matter involved (but a few extremely patient people finally figured out how to explain it to him - in a followup thread at HA, of all places!) Based on what seem to be your misconstrusions of basic signal processing topics, I'm beginning to think I should put you in the same boat.
Only one notion of measuring "time resolution" has any real meaning in an audible sense, I believe: Given two superimposed analog signals, and a system they pass through (composed of an ADC->DAC here), and a consistent way to compare the quality of the system output compared to the input, what is the minimum delay that can be applied to one signal which yields an acceptable quality in the system output? That is, how well can the system "resolve" small differences in timing between two audio events?
The exact delay number depends on the nature of the signal and the method of comparing quality. But as far as digital audio is concerned, this delay will generally decrease as sampling period decreases, and will decrease as quantization level decreases. (Which jj mentioned in a different thread quite some time ago.) So it is true that "time resolution" is definitively higher for higher res formats as compared to 16/44. But these delays are, invariably, much much smaller than 5us.
This is not a put down, but I doubt if you understand the rigors Dr. Kurkun has gone through to produce those papers. I will trust him any time over the likes of what I have seen here. For one, he has no conflicts of interest like some here.
Nice job of quoting me out of context, there, "sasaudio".
An attempt, with entirely good math, which is entirely justified. And there's more where that came from. This quote from "Probing the temporal resolution and bandwidth of human hearing" demonstrates a profound and (IMHO) embarrassing ignorance of bandlimited interpolation, and in fact, openly asserts that all converters are NOS filterless (emphasis mine):
Except that you have failed to produce a conflict of interest for anyone here, and you haven't replied to the question somebody else asked about experimental funding. (An issue that I have no information one way or the other on.)
You mention work. I'm trying to convey some very, very basic information from my 30+ years of hard work, and you seem quite unwilling to credit that in any fashion. It seems quite clear that you will only support work that supports your prior belief.
Yes, it is hard, long work. But quotes like the one Axon just produced seem, from the quoted material again, to very seriously call the interpretation of the results into question. This is what I've said over and over again, that has you so very upset.
As to your claims about what I've said about 44.1 systems, it was rather less than clever of you to take a quote and try to spin it as something it wasn't. The context of this discussion from the very beginning has been interchannel delay. Your thoroughly specious removal of the quote from context in an attempt to professionally vilify me is quite unethical, and shows, I think, your enormous lack of equity, giving you the most credit possible.
Further, it appears to me that you are more concerned with being insulting and offensive than actually learning about sampling. I will wish Axon good luck here, and await what the good Doctor has to say.
If, say, it turns out he's tilting at a straw man of your creation, I will cheerfully point that out. I doubt he'll be any more pleased than I would be to find out that somebody was leading me on. So be careful out there.
Once again, 5 us resolution does not mean 5 us pulse width. Your claim is otherwise, so stop trying to pass off your misunderstanding as mine.
I saw that one too, and was just shaking my head. The signals he used for his tests were square waves, low-pass filtered by a single-pole analog filter. Such a waveform is not strictly bandlimited in the sense that the sampling theorem requires. Producing non-bandlimited signals is really out of the scope of what traditional DSP is meant to do (or at least traditional DSP as I know it). The anti-imaging filters will take out the stuff that makes the signal non-periodic, but then the signal will be bandlimited. It appears that he did not want a bandlimited signal for his tests, so DSP is not really the right tool for that job. A screwdriver is a great tool for what it's designed to do, but it's not so good for hammering nails. If you insist on hammering nails with it, it's not the screwdriver's fault that it doesn't work so well.
Aargh. It sounds like a big train wreck in the making actually.
Um, the author raised it and used it in his justification for his conclusion, so sorry, yes, it is part of the subject of the paper.
Of course. How much energy is in the 3 us pulse. Imagine that. What's more, how much of the pulse energy is outside the passband, too? Imagine that as well.
Please, PLEASE buy yourself a copy of Morrison's "Fourier Analysis". You will shortly find out, reading it, where you have gone wrong.
....
Maybe we should try a different tack.
Would this make any more sense to you if I pointed out that both pulse signals are, in fact, infinitely long in time? Because you're not seeming to understand that the 3us analog pulse has the 48khz pulse inside of it. (Or more formally, it's the sum of the 48khz pulse and the >48khz stopband content.)
To say that the 48khz pulse is intrinsically wider and shorter is to be distracted by superficialities. I could make that 3us pulse a lot wider with an allpass filter, but it wouldn't really mean anything.
True. But with me, if you kiss me at midnight, I turn into one.
I was indeed goofing around with Jan, because though I forget what he wrote now, it made me laugh at the time. But then I have to maintain my sense of humour with these so-called "objectivists", because they can get pretty cruel and ugly. jj insults my name on a regular basis, and twice on Sundays. When he isn't referring to me as "Frog Boy", Ethan enjoys calling me a "Toadstool". I think he thinks it's a type of frog. It's not. It's a type of mushroom, which some frogs are known to sit on. But not this frog, so he's just perpetuating stereotypes. I prefer my Aalto. Then there's Krueger, who will not hesitate to call me a "slimy little amphibian". Maybe I'm just too sensitive, but there seems to be a dig at personal hygiene in there somewhere. Oh, you should have seen how "funny" he thought it was, describing the pleasure he took in slaughtering my cousins and relatives in South Florida with the front bumper of his car. You know what wasn't so funny? The fact that I had to go into therapy after reading that. I must have cried for like a solid week. So yeah, sometimes dealing with objectivists from Hydrogen Audio is like living out one of those jangly twangly country tunes, that go "I have to laugh to keep from crying...."**
About the "scary Santa" photo... I didn't really see myself as knocking jj's looks because one, I was reflecting more about his demeanour, and two, I do not assume the person in the photo is the person on this forum who calls himself "j_j". Just as j_j" does not assume I am who I say I am, I do not assume he is who he says he is.
**(And please boys, NO blender jokes!)
Axon, keep in mind I am not involved with this information on a daily basis and struggle to keep up when things get deep into digital theory and operation. (Which I see as two somewhat different things depending on which theory you start with.) But, from my perspective, the analog input does not look like the 48k signal. Please, if you can, explain how the 16/48 pulse is "in" the analog input signal. I would say only the DSD has come close to actually replicating the analog source.
Isn't the "nature of the signal" what we should be discussing? When the signal is a signal - a pulse - and when the signal is music seems to be bandied about with convenience by the objectivists to suit their needs for a desired outcome.
Isn't "higher time resolution" sufficient reason to advocate higher resolution formats?
I asked this question days ago and both jj and arnie ignored the question. I assumed they ignored it for obvious reasons - they didn't have a good answer.
Possibly you can answer this for me. Why would we not want higher resolution digital audio? I see no particular reason to stay with a system (44/48kHz) that has been proven to be inferior to what is presently available. The argument over the last 17 pages seems to be what is the absolute bare minimum for audio resolution. Shouldn't we want more than the bare minimum?
This is pretty artless in its execution and simply disingenuous from someone who has dodged every question put to him in this thread. jj, when you start answering questions put to you, you can expect others to do likewise. You constantly demand answers from others while ignoring all requests sent your way. That's a nice weaseling technique on the Sunday morning talk shows but it doesn't help anyone on an audio forum.
Don't accuse someone of something you are guilty of doing yourself. If you are only here to win, then I have the same words for you as I had for arnie.
I admit to being at least a day behind on trying to read linked articles but from what I have read so far Axon introduced another forum's pages that still seem to have some doubts about whether what you've presented is completely factual or not just what you - and others - want to see as a result. This doesn't appear to be an issue that has a clear resolution, only opinions on both sides. And from what I can tell so far, what theory indicates and math alone proves and what exists in the real world can be quite dissimilar.
Again, jj, I remind you that you are quite good at doing just that - repeatedly.
I would bet most readers would agree you are equally if not more insulting and have little interest in sharing information. My preference is for sas who has remained calm and not made wild accusations or simply declared the other person "pathetic". Steve has introduced linked articles to support his words while you have said go read a book and then added "not that you will inderstand it".
Let's consider the instance when you were asked to provide that simple "yes" or "no" answer to whether you ran tests for higher resolution. It took three days of insults on your part before you finally came up with the out that this was priviledged information.
Three days to decide whether something was priveledged information?
Really?!
jj, don't tell someone else they are more interested in insults because other than arnie there is no one on this thread more willing to resort to insults and feigned indignation than you.
The answer lies in Fourier analysis. I directed you (and others) to several good books, since you are unwilling to accept the facts from me, but rather than actually doing some work, you just continue to lie, insult, and defame.
You have no cause to complain that I'm "insulting". None whatsoever. I've just pointed out the unethical nature you exhibit (along with Frog and sasaudio), and you find the facts of your own behavior insulting.
If you do not wish to be shown to be a liar and a demigog, then I would suggest that you cease being one.
In addition to your seemingly intentional, completely unsubstantiated defamation, you are now shown to be lazy and unwilling to undertake the learning necessary to actually understand what's going on. Rather than taking on the homework necessary to digest the situation, you complain you got no answer, when you most surely were pointed to several places where you could learn the answer, which is hardly something to be conveyed in one day or page. What did you do, you ignored the answer and claim you didn't get it. Pathetic.
Now, I'll try to help you with the "appearance" part of this pulse.
It consists of an infinite number of frequencies. When it goes through a filter, some of those frequencies are removed to below noise level, and some are barely or not at all affected. This splits the signal into two parts, that which went through the filter, and that which did not.
When you start to add back (using Fourier analysis) the parts that did not, your pulse will start to shape up, bit by bit, until it recovers the original shape.
It really is that simple, you know. Fourier analysis applies for any kind of signal that is finite in at least one of length or energy, and all audio signals are both.
Screw off, jj.
"Pathetic" is your middle name.
lol, yes it will be enlightening!
The internet is like a comedy show, really!
I see jj chimed in, but since I already wrote this by my lonesome...
Thanks for the compliment earlier, Jan, although if my blood pressure wasn't naturally so high recently, I might take more of an interest to artificially raising it ;-)
Heh, yeah, I guess I could characterize my statement as "obtuse".
Let's step back and go to theoretical land for a while. Signal A is an "analog" pulse has infinite rise and fall time (a Dirac delta - "analog" in quotes because it's not physically realizable). Signal B is signal A filtered at 24khz as would be done in a 48khz ADC (doesn't really matter if B travels through an ADC-DAC or not, given they would be ideal too). The filters are infinitely sharp.
Look at the two signals in the frequency domain rather than the time domain. This is where you get into Fourier analysis (and more specifically Fourier transforms). The Fourier transform of signal A, as a function of frequency, is actually simply 1 - equal amplitude at all frequencies (up to infinity). If you've ever seen a spectrum analyzer jump up at all frequencies when a mic hits something or there's a massive pop on an LP, it's the same basic idea. The Fourier transform of B is 1 up to the filter frequency (24khz) and 0 after that.
Now, everything you can do with signals in the time domain, you can (generally) do in the frequency domain - summing two Fourier transforms gives you the same result as if you summed the time-domain equivalents, then computed the Fourier transform. So it's pretty obvious here that if you take the transform of signal B, and add a little something else - that is, a signal with amplitude 1 at frequencies from 24khz to infinity - you are left with the Fourier transform of signal A. That's what I mean when I say that the 48khz pulse really is in the original pulse.
If your brain is exploding, try to keep in mind the nature of the original pulse: it is represented by an infinite number of sine waves. The fact that it is only nonzero at one point doesn't change the fact that it is composed of functions that stretch from -infinity to infinity. When you see the wide lobe of the 48khz pulse, you're really seeing a signal which is cancelled out in signal A by the higher-frequency sine waves. I believe one of Dan Lavry's pdfs of sampling theory has a really cool diagram of how adding more and more sine waves makes a pulse thinner and thinner. Same idea.
(All of this can be restated in terms of periodic functions, and discrete functions, rather than non-time-bounded functions, but the math gets a lot more complicated and a lot harder to explain, so the head-explosion risk probably won't go down.)
In terms of delays... delaying a signal in the time domain causes a different (but entirely predictable) shift in the phase of the Fourier transform. You can shift the original pulse in signal A by however many fractions of a femtosecond you want, and as long as the quantization in your ADC/DAC is sufficiently high res, signal B will track the shift perfectly. Wide-looking pulses can still shift, predictably, in very small amounts.
To completely cop out on your question, we can afford to get away with that trick, because Nyquist's Theorem guarentees that (and I'm restating this extremely loosely) whatever we prove for pulses will work just as well for "real" music.
But to respond more directly. How do you quantify any of this for "real" music? It's complicated enough talking about it as it is for pulses. For all the talk on this thread, I don't think I've seen anybody describe (let alone hypothesize any clear numerical way of evaluating "time resolution" while modulating bit depth or sampling rate.
I think it's dangerous to think of "time resolution" as is being fought over here as truly analogous to jitter. With jitter you really can, hypothetically, look at the output of the zero-order hold converter (even when it's hiding inside a delta sigma DAC and an antialiasing filter) and see how jitter affects the timing of the transitions. There's nothing like that happening with bitdepth or sample rate changes.
But actually, when I said "nature of the signal", I kinda thought I packed insurance for this objection anyway. That is... if you do define some sort of meaningful metric for evaluating time resolution (even with real music!) (good luck!), the specific numbers that are going to pop out of the "time resolution estimator" are going to depend on what you feed into the system. Just as a trivial example: If you're talking (say) signal peaks, measuring with a pulse of peak 1 will yield a better result than measuring with a pulse of peak 0.1. But if you try to do an apples-to-apples comparison with what Dr. Kuncher was testing, all of this is going to be kind of moot, because I'm pretty certain you're going to get results way below 5us anyway.
Not if the cost involved far outweighs the benefit. If 5us is the best that can be detected, what benefit is there to move from a format which can encode 5us, with lots of performance margin to spare, to some format like DVD-A which probably has a time resolution in perhaps the middle picoseconds?
I sorta agree that this sort of objection can be overblown. If everybody has the space and bandwidth and converters for 24/192 then fire away. But we don't. When I'm asked to pay a significant premium for high-res content as presently exists, I think the vanishingly small benefit from improved time resolution in high res is obscenely cost-ineffective, and to significantly justify high res on those grounds is actually kinda insulting.
Designing a system aiming for (say) 10x the limits of human perception is justifiable. Designing it for 10,000x is not.
Ultimately, what this boils down to is that I sincerely believe that high res audio has a very great risk of harming the industry rather than helping it. A lot of money has been spend on rather expensive high res kit by studios, producers, labels, consumers.... it's obviously gone down in price a lot lately , but I just shudder when I think of how many studios bought into DSD, DxD, 24/192, Pro Tools HD.... and I fear a huge amount of that investment will ultimately be wasted, and the industry is simply too weak for that to not be a big deal. So many people have latched on to high res as this messianical thing that will save high end audio, but frankly, I think many people are full of it. And the sooner they disavow themselves of that notion, the sooner they will take their money and spend it on things that actually matter, and matter a very great deal, to high quality sound. Like better masters. Or better speakers and listening areas. Or maybe fewer high end places would go under. I kind of dream that (and yes I know I am being openly offensive and self-centered here) if audiophiles in general had a better grasp of what really mattered, maybe, just maybe, the industry wouldn't be in as so much trouble, and audiophiles wouldn't be picked on as much, and people would actually listen to what they/we have to say instead of just buying Bose, etc.
Just for grins I fired up my 1 mHz 32 bit digital audio system and used it to generate a pulse that was 5 clocks, or 5 microseconds long and with a positive amplitude of 0.8 of full scale I then downsampled it to 44/32.
The results were a damped positive pulse about 2.5 clocks long with an amplitude of about 15% of FS.
I conclude that based on a real world experiment, JJ is right and SAS is wrong.
Just for grins I fired up my 1 mHz 32 bit digital audio system and used it to generate a stereo pulse that was 5 clocks, or 5 microseconds long and with a positive amplitude of 0.8 of full scale I time shifted the right channel so that it lagged the left channel by 1 sample, or 1 uSec.
I then downsampled the 1 MHz signal to 44/32. If 44/32 were incapable of resolving signals that varied by 1 usec, then I would obtain two identical signals. If 44/32 were capable of resolving signals that varied by 1 usec, then I would obtain two different signals, and one would lag the other.
Furthermore, if 44 KHz sampling were incapable of resolving a 1 uSec difference, subtracting the two signals would produce a zero result. If 44 KHz sampling were capable of resolving a 1 uSec difference, subtracting the two signals would produce a significant result.
Visual inspection showed that the L & R signals were different. I also subtracted them and obtained a significant difference.
I conclude that based on a real world experiment, JJ is right and Dr Kunchur is wrong.
You and I have a very different definition of "chimed".
OK, you added a word to my vocabulary; "femtosecond". I have absolutely no use for it but I now know what it means.
But I still don't get this ...
To me that says you are inducing magic. Or you want me to believe you can.
You are adding something that isn't there. Articles have indicated frequency response to well over 100kHz and I fully understand the analog signal containing an infinite number of sinewaves. And yet you just reach into thin air to get those frequencies into your equation. From where?
Am I misunderstanding Nyquist theory when I believe that 48k sampling can do no better 24kHz response in raw numbers? So where are you getting these add ons of information above 24kHz? How do you know they are what should be present in the analog signal? Isn't that what makes the difference between the 48k signal and the DSD signal?
But just like watts, bits are cheap. IMO this is where the "scientists" again mix their realities - sort of like using music when it suits their cause and pulses the next. It is whatever suits your needs at the moment you are debating or measuring.
If making audio better is not the goal, why not just stop now? Yeah, I get the marketing of something new that costs money and is outdated in two years time but the industry has done that to itself - better audio seldom has much to do with the marketing of audio. Why not design for 10k times what is required if that ultimately brings down the cost of 10k times - as with DSD? Better is better, right? This doesn't seem to be any different than anything else in audio over the last sixty years. Early adapters risk owning expensive paerweights but they are needed to bring down the cost for the second and third and fourth stream buyers. The trick is to guess correctly.
Not to turn this into a digital vs. analog debate but one of the things people find most interesting about analog playback is the amount of information that exists in that 60 year old groove waiting to be extracted. IMO most of those systems were not designed by people who thought, "Well, this is good enough." If the idea is we only need to design to "X" level and then let cost constrain what might have been better, then there wouldn't seem to be the same future for digital as analog now enjoys. Sooner or later an entire group of "good enoughs" will be useless. Bits are cheap and getting cheaper.
A Science Channel program on the history of the space flights indicated there was as much processing power on the Apollo spacecraft as we now have in an average cell phone. What was on the ground is the equivalent of an average PC today and it took up an entire room and then some in the late 1960's.
You and I disagree on this. It isn't the average audiophile who is shutting the doors on the high end shops. I've never known anyone interested in high quality audio who doesn't have a story to tell about opening a friend's or neighbor's ears to what music played through a good system sounds like.
And, yes, I know I'm being offensive when I say it, but possibly if the anti-audiophile community was not so dead set on arguing for what is good enough and paid more attention to the audiophiles, if that happened, then maybe more people would be educated as to why Bose is a poor investment. As is Bose is "good enough" for them.
There's far more to it than that but people don't know squat about audiophiles until they are introduced to the high end. What they do know about is the guy in the lab coat who promises them "Perfect Sound Forever" and then spends the next twenty five years defending what is good enough while in their labs they go about creating something that is even less good. Hanging anything on the audiophile when there is the guy in the lab coat telling the consumer not to pay attention to the audiophile is, IMO, one thing that is damaging the industry. The average audiophile only wants better music reproduction and not a fight over how to get there.
Yes, exactly, that's what this audiophile wants, to see you stop fighting about how to get to better audio, by paying attention to the remaining (many) obvious problems in sound reproduction, and ceasing to argue about mistaken beliefs about Fourier analysis.
So why ARE you fighting, now?
Which one of us has patents, papers, etc, on trying to create a much stronger verisimilitude between recorded music and the original venue (when such exists), and to create the sensation of a real, live, good-sounding venue when no such thing exists, and then overlaying THAT over your own listening room acoustics via well known psychoacoustics (if John is reading, yes, I finally figured out why PSR worked better than expected). Oh, that would be me.
So, yeah, I'm working on that.
Et tu, Vigne?
Et tu, Frog?
I've left out SAS because it would appear he is doing a businesslike job of developing euphonic amplifiers, which is a perfectly reasonable thing to be doing. He just needs to avoid arguing Fourier Analysis and/or sampling theory. On the other hand, Morrison's book would probably not be that hard a read, if he's designing tube amps already... Seriously.
There are several differences between analog and digital in that regard. Analog is hard to build, impliment, and care for, because it lacks the recovery characteristics of digital recording. In analog, you MUST try hard. It's the only thing that stands a chance of working. It's also part of the charm of analog, because it is part art and part science, in making the flaws that physics requires it to have be acceptable.
In digital processing, this used to be the case, and is still a concern in the ADC and DAC stages, which must handle analog signals extremely well. However, techniques like sigma-delta (or delta-sigma, I've personally watched Jim Candy and Joe Condon going on about that) allow the conversion aspects (Yes, you certainly must get the analog part right!!!) to proceed in stable fashions that don't drift or create problems. Of course, then other problems (clock issues, for instance) are created. So there is less, considerably less very sensitive hardware in the digital reproduction chain. But there is some. Asking me about grounding and bypassing in some consumer CD players is only to be done in a bar with Guinness. Please.
Having said all that, I'm on the record as saying that 44.1 kHz is right on the hairy edge. It goes back to the whole issue of the antialias and antiimaging filters. There is a simple result of Fourier analysis that says that for a filter (note, this is a filter, different thing, different rules) that frequency resolution (in Hz) times time resolution (in seconds) is greater than or equal to either 1/2 (for known vs. unknown situations) or 1 (unknown vs. unknown) (yes, that's a bit obtuse, and there is a whole book chapter betwixt the beginning and end of that sentence). The point is that if one is going to have a filter gain of 1 at 20kHz, and zero (approximately, of course) at 22050 Hz, then the length of the main (i.e. biggest values) lobe of the filter that does this must be 2/2050 seconds wide, AT MINIMUM. That is long enough that there is potential interaction between the shortest (high frequency) cochlear filters and the antialiasing filter (in a nonlinear sense, not a linear sense). At 48kHz, this is probably ameliorated. At 64kHz, with appropriate (i.e. not brick wall) filters, this potential problem (nobody has proven it is a problem yet, and it is fiendishly hard to test) is just gone.
So, I actually agree in some sense, 44.1kHz is pushing it a bit, however this has nothing whatsoever to do with the actual system bandwidth, and everything to do with nonlinear interactions between the cochlear filters in the human ear and the antialiasing filter. Were this potential (and remember, for now it remains potential only) problem addressed, the bandwidth would not grow, only the filter transition band would grow.
Even though an effect has never been demonstrated at the filter lengths in a typical delta-sigma DAC, my own engineering sensibilities would prefer a solution where this wasn't going to be an issue.
This effect, in terms of pre-echo, HAS been demonstrated in some poorly designed hardware resamplers (that are no longer on the market), so the effect does begin to matter at some filter length and quality of design. The limit to this problem lies somewhere between a gradual rolloff filter rolling off from 20kHz to 32kHz, and a really bad 44.1 to 48 upsampling filter. No, not much help, I'm afraid, that's a lot of space to cover.
But as you are probably going to point out, no, they didn't ask me about the sampling rate. At the time I was still a junior guy at Bell Labs, and not even on the Philips Research radar screen.
At higher sampling rates, there are also other considerable problems, in terms of word lengths required for calculation, etc. That's an entire thread on resolution issues in DSP in and of itself. They are fixable by "more hardware" that gets into the custom arithmetic unit range, which is also known as $$$$.
He's talking about the reproduced analog signal and its transform, which is no longer sampled, and therefore no longer bandwidth limited. (although there WILL be little or no energy above fs/2, mathematically the little or none can be calculated. In the sampled domain, the concept of those frequencies literally does not exist independently of the frequencies between -fs/2 and fs/2.
Linear system, remember? Difference between the analog INPUT and analog OUTPUT is what he's talking about.
He's specified the input signal as a Dirac delta. So he knows the input signal exactly, well, to the limits of physics, which is not a problem here.
The DSD signal has frequency response up to 50kHz or so, (you do know that there is in fact an analog filter on the output at about 50kHz to keep the delta-sigma noise that is extremely high level and extremely ultrasonic out of your electronics, where it would not be a good thing to find, yes?) so you will see a "sharper" edge on the impulse.
This is simply a different difference between the input and output, either in spectrum or time domain. It's a linear system. if x = a - b, then Transform(x)=Transform(a)-Transform(b). n.b. multiplication works differently, so don't go there. Well not unless you want to, in which case please head to my digital filter design deck on the tutorial site I've pointed to recently.
Question for you: Do you have any of octave (from sourceforge), Matlab, or scipy?
Nobody is talking about stopping the serach for better sound. For many of us the whole discussion is about which way to go to obtain the largest improvement.
Many of us learned that extending bandwidth beyond 22 KHz is highly unproductive, about 30 years ago. About 8 years I learned quite clearly that the point of incredibly rapidly diminishing returns is more like 15-16 KHz.
The good news is that there are many other more productive paths that are at this point only lightly travelled upon.
There is a big difference between data, information, and music. Not all data is information, and not all information is music. Given that we had cartridges with response out to 30+KHz in the late 1960s, and preamps with dynamic range no worse than we have today, it is safe to say that everything that could be wrested from the grooves of vinyl was known and being wrested, some 40 years ago.
There's a reason why the CD was invented - it was well known and generally agreed that vinyl was a dead end.
Agreed . If you believe what they wrote and published in those days, they thought that "Vinyl has been fully exploited" and "Vinyl is definately not good enough - let's do digital".
And they were right - whatever sample rate and bit depth we might need to do good audio, its available to just about anybody with a computer and maybe $100-200 for a good computer audio interface.
LOL, that reminds me of those ads for colon cleansers that use scientific sounding claims. Hey, you can't have it both ways. If someone wants to sell stuff based on beliefs that's okay. But when they try to bring in science to justify their claims, they better be prepared for real scientists to object.
--Ethan
Arny, if you can post a screen-cap you'll have a check-mate.
--Ethan
It would be nice, in my opinion...to see credentials from folks that challenge published physicists. I am too ignorant to know about the details of this stuff(way over my head) but Dr. Kuncher is quite obviously an expert..
I had no idea that this thread would be so long..Glad I saw those papers!
Let's take a look at J_Js comment:
This brings up some interesting points. But first, the public needs to know that one is allowed to edit their posts back three days and does not have to post the edit time. So don't be surprised if we are accused of misquoting or using a quote out of context.
1) Ncdrawal post a simple link to Dr. Kunkur's papers. Of course the question is why? Of course to "examine" it. I sure have never heard of Dr. Kunkur nor his papers. I bet most have not. So what is the purpose of examining the papers?
2) Well, the first thing we see is the attacks on Dr. Kunkur's credibility, and thus his papers, in the first two pages of this thread.
arny:
JJ:
3) When does one see a real professional scientist not invite the other side, Dr. Kunkur, to the table to discuss the issues? Yet arny and J_J did not. Not one attempt.
Instead, especially J_J, arny resorted to backstabbing Dr. Kunkur and his credentials in an attempt to discredit him (and of course his papers). It was not until later J_J asks me to contact Dr. Kunkur. So why did not J_J not contact Dr. Kunkur himself, and much earlier?
How about the requirement for J_J to reveal his legal name and his email address when corresponding to Dr. Kunkur? We have seen on every string that J_J has refused to post his legal name and distance himself, legally, from his own posts. So if he won't post his legal name to us, he certainly will not to Dr. Kunkur.
So Dr. Kunkur was never contacted by arny nor j_j so he could reply to the charges arny and J_J made against him (and his papers). Instead they go behind his back.
When was the last time a real professional resorted to such tatics. I was the one who voluntarily contacted Dr. Kunkur some days ago.
4) While Dr. Kunkur was discussing what the ear could detect and relate it to bandwidth, j_j manipulated the subject from music and the need for high bandwidth to first, 5us tests. When that failed, J_J and lately axon attempted to use special "stretched" 5us tests, and other manipulations etc; anything to discredit Dr. Kunkur. This, of course, all without Dr. Kunkur's knowledge and of course, ability to respond.
5) And not only are they attacking Dr. Kunkur (and his papers) without his knowledge, but J_J, Xenophanes, Axon, Krabapple etc will not sign their legal names to their own posts. We have no idea who Xenophanes, axon, Krabapple, and ncdrawal are. But the no names are quick to attack another, behind his back, without notice.
Is this the way Hydrogen audio forum and the Audio Engineering society function?
Take care gents.
(ps. I edited to renumber since I had two points labeled "2".)
Purely anecdotal so of no consequence.
I have already invited Dr. Kuncher to participate. Hopefully he will not be turned away by the fools in the peanut gallery.
I'm flattered to be included with jj, Axon, and Krabapple!
I should point out that it is misleading to suggest that j-j somehow attacked Dr. Kunchur's research, papers, and credibility in general. That is far too general. What j-j said was,
"Of course a 44.1kHz/16 bit system can resolve time to well under 5 microseconds. So something has gone wrong and as quoted, without surrounding context, the quote looks fundamentally ignorant of the basics of sampling theory."
Did Dr. Kunchur do any tests to justify his remarks about the time resolution of CD technology?
Agreed. But I think the reason we're not seeing these credentials, is because those who are busy here making these ludicrous stabs in the dark at challenging the highly credentialed Dr. Kuncher's published studies; which includes Ethan Winer, Arnold Krueger, "Axon", "Krabapple", "Xenophanes", "j_j"**, could not even come close to matching his credentials as a real scientist. (**But not you, SAS was mistaken about that)
Forget publishing their findings, forget the peer-reviewed process... heck, forget even reading their opponent's studies! These self-professed audio forum "objectivists", speaking from the peanut gallery, actually declare themselves to have superior knowledge of the science of audio and hearing, over Dr. Kunchur, and others who did similar studies with similar findings. And within the scope of what, two days?, they announce that they have rendered his years-long study into the effects of ultrasonics on human hearing null and void. Simply by what their brilliant accomplishments hacking away with their PC software came up with last night. Oh, and they might have prepared a chart on Flickr to show you! This is what we call in the scientific community, a "joke". But wouldn't it be great if science worked the way these guys imagine it does? Heck, we'd ALL be scientists! "Hey everybody, look what I came up with on my coffee break! This proves Einstein was wrong! If you doubt me, it's because you're not smart enough to understand it! "
It is silly to take these guys seriously and try to compare them with real scientists like Kunchur, because for one, none of them are scientists in the first place. They just like to play one on audio forums. Children like to play with Matchbox cars. It would be just as silly to assume that because they do, they are the equivalent of professional racers. None of these guys are even audio engineers, like SAS is. The only thing these guys seem to be professional at is the art of skepticism. And the one thing they all have in common, is an anti-high end agenda whereby they live on audio forums to dismiss anything that conflicts with their extremely narrow world view of the reproduction of sound and music.
Dr. Kunchur's studies could hardly be associated with the high end community, but still, his conclusion runs counter to their agenda, so their job is to simply "appear" to "debunk" it. Since none of them have even tried to contact the Dr. to present their version, and the Dr. is not here to defend his own paper, they get free reign to do that. There is one particularly condescending loud-mouth in this group who tries to intimidate his opponents on this forum with the idea that he is more credentialled and knows more about audio than anyone on the planet, and by some degree of bloated arrogance and narcissism I have not seen in my lifetime, even has the temerity to demand in his posts that everyone acknowledge this! The first problem with that is, this character refuses to post under his own name and as far as I know, has never posted under his real name. He refuses to state his professional affiliation in his posts. He refuses to even state his own real name when asked! He talks about himself in the third person, and his actual sanity is highly suspect, according to the testimonies of many. So naturally, he could be anyone, and I can't say who he really is.
I know who Steve is though, because he does sign all his posts with his real name and professional affiliation. So anyone can contact him by phone and find out who he is. From his posts, he is a calm and rational person, who's credentials are well known, but who never tries to intimidate and bully other members with appeals to authority. And as Steve inferred, what kind of self-respecting scientist won't even state his own name when asked on a discussion forum? So I think that all pretty much tells you how this group of armchair scientists we have seen in this thread, compares to real practicing scientists who are sane, rational, doctored professors like Dr. M.N. Kunchur. Who happens to have a lengthy portfolio of published papers and lectures on the science of human hearing, and who you can be sure as a busy professional attending lectures, symposiums, doing research and teaching at a university, does not spend his life posting to audio forums like this one, under an assumed name! Especially not to blast the hard work of others, based only on misunderstood out-of-context quotes taken from their studies.
So, measurements of hardware doing a task are "anecdotal", you say?
Are you SURE you want to go there?
It would be good to know what he actually intended to say, as opposed to what he has been quoted as saying here.
It would also be good of you to note that I have repeatedly said 'as quoted here'. There's a reason for that, it's possible that somebody misquoted him, or played the same kind of games with quotes that sasaudio tried to pull about "resolution" vs. actually passing a 5 microsecond pulse.
It is time for you to stop lying outright about people's credentials.
It's telling that you have nothing but defamatory, obviously false ad-hominem attacks to offer. There are two potential obvious reasons for this, the first being that you know for a fact I'm 100% right on the technical issues, and the second that you haven't a clue, and you're just here to derail discussions.
Since you've already admitted that you intend to harrass me, and intentional derail discussions, I think we're pretty much entitled to conclude you are nothing more than an irrational troll who wishes to squelch actual information and discussion.
The quote cited specifically addressed interaural time resolution.
I specifically discussed the ability of a 44.1/16 system to provide the time resolution attributed to your author.
A lie on your part. The sheer dishonesty of your repeated attempts to change the history of this discussion is really quite evident, and you should give it up, and admit that you are the one who introduced irrelevant nattering about "5 microsecond pulses" into a discussion, directly related to the quote from your author, about 5 microsecond TIME RESOLUTION. It is you who introduced this derail after you were shown factually wrong about the time resolution issue, something that it is, by the way, time you acknowlege.
Any fault lies with the one who cited specific paragraphs as evidence. That's the only responsible party here. Someone quoted a paragraph, that, as quoted, appears to have a basic misunderdstanding of time resolution in a PCM system. That person presented this, not your author. The fault, if that author said something different, lies entirely at the quoter's own feet, and failure to respond to this simply constitutes abandonment of the author and the (miss?) use of citations.
If Dr. Kunkur has a gripe with anyone (which is to say if it turns out someone took his commments out of context), it will be that someone.
Obviously, you have no understanding of the mathematics, you've proven that over and over again, so at this point, you would indeed be well advised to go buy the book I've pointed at you, and read it cover to cover.
Some self-education on your part, along with a massively improved attitude and the removal of the sequoia balanced on your shoulder, would go a long way toward helping you recover some of the horrible hit you've done to your credibility.
Sure where is your proof? You demand proof from us, but fail to provide proof when we reequest it. Have you actually worked and tested in ultrasonics?
I see you lack as much integrity with us as you have with Dr. Kunkur. And you expect people to trust you.
----------
Here is the post you tried to sneak around J_J. Can you answer it this time or are you going to let the public see under your facade?
Let's take a look at J_Js comment:
This brings up some interesting points. But first, the public needs to know that one is allowed to edit their posts back three days and does not have to post the edit time. So don't be surprised if we are accused of misquoting or using a quote out of context.
1) Ncdrawal post a simple link to Dr. Kunkur's papers. Of course the question is why? Of course to "examine" it. I sure have never heard of Dr. Kunkur nor his papers. I bet most have not. So what is the purpose of examining the papers?
2) Well, the first thing we see is the attacks on Dr. Kunkur's credibility, and thus his papers, in the first two pages of this thread.
arny:
JJ:
3) When does one see a real professional scientist not invite the other side, Dr. Kunkur, to the table to discuss the issues? Yet arny and J_J did not. Not one attempt.
Instead, especially J_J, arny resorted to backstabbing Dr. Kunkur and his credentials in an attempt to discredit him (and of course his papers). It was not until later J_J asks me to contact Dr. Kunkur. So why did not J_J not contact Dr. Kunkur himself, and much earlier?
How about the requirement for J_J to reveal his legal name and his email address when corresponding to Dr. Kunkur? We have seen on every string that J_J has refused to post his legal name and distance himself, legally, from his own posts. So if he won't post his legal name to us, he certainly will not to Dr. Kunkur.
So Dr. Kunkur was never contacted by arny nor j_j so he could reply to the charges arny and J_J made against him (and his papers). Instead they go behind his back.
When was the last time a real professional resorted to such tatics. I was the one who voluntarily contacted Dr. Kunkur some days ago.
4) While Dr. Kunkur was discussing what the ear could detect and relate it to bandwidth, j_j manipulated the subject from music and the need for high bandwidth to first, 5us tests. When that failed, J_J and lately axon attempted to use special "stretched" 5us tests, and other manipulations etc; anything to discredit Dr. Kunkur. This, of course, all without Dr. Kunkur's knowledge and of course, ability to respond.
5) And not only are they attacking Dr. Kunkur (and his papers) without his knowledge, but J_J, Xenophanes, Axon, Krabapple etc will not sign their legal names to their own posts. We have no idea who Xenophanes, axon, Krabapple, and ncdrawal are. But the no names are quick to attack another, behind his back, without notice.
Is this the way Hydrogen audio forum and the Audio Engineering society function?
Take care gents.
My proof? This time it's Arny's proof. It's that simple.
My proof appeared pages ago, and you have yet to understand the simple, irrefutable nature of it.
Pages