gubu

If that statement proves your point, then I have no idea what your point is because the statement supports what everyone else has been saying. Here is a summary of facts to hopefully end this confusion: -For a sampling rate of x, frequencies can be reproduced up to x/2 (ignoring antialias filters) -Antialiasing filters can distort audible frequencies for 44.1 kHz -Sampling rates above 44.1 cannot capture any additional audible frequencies (assuming humans can't hear above 22.05 kHz) -Despite the inability to capture additional audible frequencies, there *is* an audible difference (to some) for higher sampling rates because the filters do not affect audible frequencies. -Ignoring filters, all sampling rates above 44.1 kHz can reproduce signals with frequencies below (22.05 kHz) Are there any of these that you disagree with? I don't disagree with any point but the final one needs clarification - You cannot ignore filters. If you are not reproducing harmonics above 22.05kHz, then you are not accurately reproducing waveforms at 22.05kHz because the harmonic information that gives these waveforms their strange shapes is absent due to the filtering. That statement you quoted above supports my argument because it's saying that frequencies at the upper limit of human hearing ARE NOT faithfully reproduced at 44.1 or 48kHz. If you input a square wave and output a sinusoid, that is NOT an accurate and faithful reproduction. Lavry himself states that the optimal sample rate for recording audio using current technology is about 62-63kHz

Only if you're recording a lot of 20kHz square waves. Well, no one can argue with the fact that spatial and depth information is contained in that part of the signal approaching the upper limit of human hearing and if these frequencies can be accurately reproduced, and they can't be accurately reproduces at 44.1 or 48kHz, you have a much more detailed and faithful representation at a higher sample rate. Lavry himself states that the optimal sample rate using currently available technology for recording audio is 62-63kHz

Did you read the last paragraph of the article? I haven't read the rest yet, but it sounds to me like he pretty strongly opposed 192 kHz. "The optimal sample rate should be largely based on the required signal bandwidth. Audio industry salesman have been promoting faster than optimal rates. The promotion of such ideas is based on the fallacy that faster rates yield more accuracy and/or more detail. Weather motivated by profit or ignorance, the promoters, leading the industry in the wrong direction, are stating the opposite of what is true." He states in the same article that the optimal sampling frequency for recording audio is around 62-63kHz and that the reason it's a fallacy that 192 yields more clarity and detail is because machine speeds are not yet fast enough to capture enough data per sample at that speed. If machine speeds were fast enought to capture the same amount of data per sample at 192 as at 44.1, 192 would still give you a more accurate representation of the audio by applying the very math that Lavry outlines in his paper. Even with the very concise math involved, any digital representation is still an approximation, albeit an extremely fine approximation, and would be a finer approximation at a higher sample rate. I repeat:- Lavry says himself that the current optimal sample rate is about 62-63kHz

A 20kHz square wave requires frequencies higher than 20Khz, so it gets rounded into a sine wave by the input filter of the converter. Square waves are often used to test frequency response in just that manner. But you knew that. A 20kHz square wave will come out sinusoidal from any DAC that has a input filter set for audio, i.e. ~ 20kHz cutoff - regardless of what the sample rate is thereafter. Once you have at least 44.1 sampling, it's all about the filters. Terry D. This actually proves my point

Guys, I've had enough of this. You are not getting an accurate reproduction of waveforms at the upper limit of human hearing when sampling at 44.1 or 48kHz and it DOES make a difference to the quality of the audio. This is backed up by Lavry's paper above and is the one point that I've consistently made that none of you have been able to successfully rebut. Lavry himself says that the optimum sample rate for audio is around 62-63kHz taking into account that the only reason that sample rates higher than this are not optimal is due to machine speeds not yet being fast enough so clearly there is a benefit from recording with a higher sample rate. That's all I've been saying for the last 2 days. I've read the math and it backs up what I'm saying so take it or leave it people, higher sample rates = better quality audio

Where does he say 192kHz would be the most desirable sampling rate? Unless I missed something in the linked PDF (which is very informative, by the way, and thanks to rlindsey0 for posting the link - it's a good refresher on stuff I haven't looked at much since college) he says even 88.1kHz is overkill. Right on page 1 - "The above suggests that 88.2 or 96KHz would be overkill." That's not what I said, pardon me if it wasn't clear enough for you. He says that 192 is not an optimum sample rate only because the processors can't encode enough information per sample at that speed.

By the way, I didn't see anyone say that math can give you better time domain resolution so you seem to be arguing with your self. In fact, it seems like everyone has simply said that math defines the frequencies that can be reproduced for a given time domain resolution, which cannot be argued with (unless you want to ignore mathematical proofs). If you think that reproducing what most consider to be inaudible frequencies sounds better, then go ahead with that. But, to my knowledge, I have not seen any scientific tests that support this. Of course, there are other factors involved as others have pointed out (plugins, etc.). Yes but math doesn't give you an accurate representation of those frequencies at the upper limit of human hearing. Indeed the math proves that it is impossible to get an accurate and faithful reproduction of these frequencies at sample rates of 44.1 or 48kHz. and this is what I've been saying all along. Lavry himself states that the optimum sample rate for audio is about 62kHz, so you do need a higher sample rate to get an accurate map of the waveform. That's all I've been saying. The only reason 192kHz is not the best and most desirable sample rate right now is because processor clock speeds are not fast enough to encode enough information per sample, and that's per Lavry himself.

This may be of interest. www.lavryengineering.com/documents/Sampling_Theory.pdf I'll get my coat

here's my visual for viewers: (ignore the obvious sub par drawing and binary representation errors) That's all well and good but you're showing 11 or 12 samples per cycle there for a 20kHz waveform in 44.1kHz and that's just not correct. Could you redraw it showing what happens to a 20kHz square waveform when you sample it at 44.1kHz? According to other posters who were arguing with me, it should come out sinusoidal. I had it all arse-backwards earlier on in the thread but now I know I'm 100% correct.

This discussion has been a lot of fun. A lot of Nyquist and wave theory that I'd learned before was way back in the grey matter before this all started and it was a great way to dig it all out again. I may have talked garbage on a few points, been belligerent and insulting on others and downright silly at times so sorry for that everyone. Sorry too to the OP for hijacking your thread The grey matter can't be doing too bad because even after being corrected on a number of points, I still believe I'm 100% correct in saying:-- MATH CANNOT GIVE YOU BETTER TIME DOMAIN RESOLUTION, FOR THAT YOU NEED A HIGHER SAMPLE RATE :D:D:D:D:D:D:D:D

I track at 44.1kHz/24-bit. Through Apogee, Focusrite, and Digidesign converters. I've tried 48 and 96kHz and did several songs that way, and just couldn't tell any difference (audible difference, I mean - the computer was certainly getting loaded down a lot faster!) so I went back to 44.1. Fair enough, what about the old square 20kHz wave being reproduced sinusoidal at 44.1 then?

Believe what you want to believe, but I implore you to do some more reading up on converter design and sampling theory. I think you'll be surprised what you might find out, and it could go a long way toward helping you avoid pseudo-scientific sales pitches in the future. I have no vested interest in promoting 96 or 192 over any other sample rate, nor do I listen to sales pitches, ever. I decide what I need for the job and go and get it. Indeed, if you've been following this thread, you'll have read that I use 48kHz and I think it's just bloody great. You might also have read that I prefer 48 over 44.1 because it DOES sound less grainy after you apply DSP. As a matter of interest, what sample rate do you track with? I'll take a guess that it's higher than 44.1 anyway

Because, as we all know, musicians never buy into hype. Or maybe they'd convince themselves that the recordings sounded better, despite the laws of physics. The results would sound better. You can't argue with the fact that mixing at 96 will produce better results, even if only for the fact that you'll run into far less 'graininess' problems when adding DSP at a higher sample rate. And that is fully supported by the laws of physics

good point, but those higher harmonics aren't there to begin with. No microphone will produce them. Good condensers will produce anything up to 50kHz or more. And that's per spec sheets that I've just Googled

nope it comes out as a square wave, because a square wave is a bunch of sine waves added together. Not if the higher harmonics that make it a square wave are being filtered out by the LPF that would be active when converting at 44.1

You don't need to. Their clients are. whatever. If they continued recording in 44.1, their clients would go down the road to another studio running 96 because the results would sound better

You don't need something like a subjective listening test to prove this. If you don't accept the math, you can simply input a 20kHz sine wave into your A/D and then output it through your D/A and put that on a scope. It's a sine wave, not a square wave or any other odd shape. But if I input a 20kHz square or other oddly shaped waveform, it will always come out as a sine wave?

If I put two points on paper, tell you they're taken .00002268 seconds apart, and ask you to draw a curve through them, any shape you draw that isn't a sine wave then must have component frequencies higher than 22kHz that wouldn't be there after the cutoff filter. ANY function can be represented by a sum of sine waves. Ok, now we are into inaudible frequencies affecting audible waveforms, which I was trying to stay away from and had dismissed as nonsense earlier. That 20kHz violin harmonic mentioned above in this case cannot be reproduced faithfully by a convertor running at 48kHz but can be reproduced faithfully by on running at 96kHz. I realise that we are at the limits of every other device in the recording/monitoring chain but surely this does make a difference to what we hear?

And I've spent fifteen minutes tweaking an EQ in bypass. That's your problem pal, the 96kHz recording sounded better every time, like taking a baffle away is the best description.

P.S. By the way, your idea of sampling an 84Hz bass string at 168Hz would indeed not sound much like a bass string - because bass strings have harmonics, not because Nyquist was wrong. Read down the thread, I asked later if an 88hz fundamental by itself could be accurately reproduced by a system running say 256hz. And indeed in the original post I did say to disregard harmonics altogether

There's no way to test that, obviously. However, if identical sources at different rates dither down identically, it would be a real stretch to claim that combining higher rate sources and then dithering would have a detectable improvement over mixing at the destination rate, even if there were an advantage to using the higher rate in the first place, which pretty much the entire field of audio engineering rejects. Why do so many top end studios opt for 96 or 192 then? And don't tell me it's to attract clients, if there was no appreciable difference in audio quality they'd have stayed at 44.1. And also, as pointed out by WRGKMC above, things get grainier much faster in lower sample rates than in higher ones when you start adding processing and this is audible to me even between 44.1 and 48, that's why I use 48. Regarding the tests you can find online where the same audio was sampled at different rates, dithered down to 44.1 and combined 180' out of phase to produce a null, if they were only using noise or simple sinusoids or squares or whatever, I'm not interested. What struck us about the test we did is the level of detail revealed at 96 on an acoustic guitar, which obviously produces a far more complex waveform than a signal generator or suchlike. And again, I'm only telling you what I heard with my own ears, the difference was remarkable.

Why would I duplicate a hopelessly flawed test? If you're not double-blind, you are susceptible to placebo effect, regardless of your room, your monitors, the phase of the moon, and who wins American Idol. Knowledge of the tests by the test subjects invalidates any results. OTOH, if I sniff around, I can find links where people have converted the same audio at a lower and higher rate, dithered down to 44.1, summed one track with the other with one inverted 180 degrees, and ended up with a complete null. Well it worked fine for us, the difference in sound quality was as clear as the difference between day and night and there was no moon up that night . Hey, I still use 48kHz myself and I'm quite happy with it but if I had the money I'd spend it on convertors, storage and processing that could handle 96kHz or even 192kHz. Those other tests your talking about do not prove that if you record and mix a project in 48kHz, it will be identical sounding to the same project recorded and mixed with a higher sample rate, all they prove is that single sources recorded in different sample rates are identical dithered down.

You might want to Google "double blind test" and "placebo effect." You might want to try running a similar test to our one yourself. We weren't mucking about, the 96kHz recordings had more space, resolution and depth on Genelec monitors in an acoustically treated studio listening room. I don't take placebos, I use my ears. Math does not give you better time domain resolution. Only a higher sample rate can do that.

Point us to an ABX test that proves that anyone can reliably tell the difference. Not "me and my buddy sitting in a room" but a solid test, like the one I linked to. The funny part about that test was that they were using prosumer converters in the HHB. And folks STILL couldn't tell the difference. This reminds me of the audiophile gear arguments - everyone arguing their oints w/o a shred of definitive, scientific evidence of any audible difference. MG I've been talking about the benefits to recording all along and the test you linked to has nothing to do with recording whatsoever. As I've said before, I've no doubt it's very difficult to discern between SACD and CD in a blind test but that has nothing to do with what I'm talking about. And 'me and my buddy sitting in a room' repeated that test running the different sample rates on either machine and running either sample rate from either side of the XLR splitter numerous times and every time without fail, the 96k recordings had more space, definition and depth so you can sneer all you like. Math does not give you better time domain resolution. Only a higher sample rate can do that.

Is it because there's electricity in them wires or because of my coffee beans? Now you're just being a dick. Time domain resolution my friend. Apply the math that you say gives you perfect resolution at 44.1 or 48 to 96 or 192 and whaddya know, even more perfect resolution!