The 48KHz vs 96/192/etc Khz debate

Manning

Manning

Experienced
Even Fractal land has been touched by the outrageously heated debate going on in the recording world over what frequency digital audio should be sampled at.

On one side of the debate are the audio hardware (including hifi hardware) manufacturers claiming "more is better". Their argument seems to make sense - we are sampling an analog waveform, so surely the higher the rate we sample at, the more accurate we will get, right?

On the other side are a group of scientists and engineers who assert that the above argument is complete and utter bollocks, and betrays a complete lack of understanding on how sampling actually works. Here's an excellent presentation by one of them:

24/192 Music Downloads are very silly indeed.

Make sure you seriously get your geek on before reading the above link.

For the record your Axe runs at 48KHz. Coldplay just released their most recent album after recording it at 48KHz. Lots of people (like myself) refuse to record at anything above 48K.
 
I'm in the camp that thinks the race for ever-higher audio sampling rates is serving little useful purpose, other than to line the pockets of certain equipment manufacturers.

Audiophools and "golden ears" aside, the only reason for oversampling an audio signal beyond the Nyquist rate is to deal with "foldback" aliasing effects that result from less-than-ideal ("real world") filters. Some amount of oversampling is good, to deal with these problems. For example, this is why CD Audio is at 44.1 KHz, instead of at 40 KHz, for audio program material with bandwidth of 20 Hz - 20 KHz. The Axe-Fx is clocked at 48 KHz, another common audio sampling rate. I think both 44.1 KHz and 48 KHz represent reasonable oversampling rates, providing frequency headroom, to deal with foldback aliasing.

In my opinion, 96 KHz sampling (a doubling of 48 KHz) is overkill, and 192 KHz (a doubling of 96 KHz) is just a money-grab.
 
I regularly record and mix at 96k, though most often I do 44.1. I can hear a clear and obvious improvement at 96/88. But I'd rather keep it simple most times and save disk space. Although I have the capability to do 192 I don't and rarely have.
 
@xristo - absolutely.

To record program material at a max freq of 20 KHz, we need a Nyquist frequency of 40 KHz, plus a low pass filter to prevent foldback aliasing of ultrasonic noise into the audible spectrum. We'll set the LPF to 22Khz and give it a 2Khz bandwidth to allow a margin of error in the LPFs accuracy. (This is excessive, but whatever.)

So I've now (more or less) successfully argued the case for 48 KHz. (You could use the same argument to prove 48KHz is excessive and the 44.1KHz of CD players is just fine... and I would have no real counter to that.)

A sampling rate beyond 48KHz requires the audio quality to become "better" as we increase the sampling rate. However this essentially requires the Nyquist formula to "not work", despite it being a proven mathematical fact. It also requires increasingly broader LPFs to filter out the ultrasonic noise that is otherwise available to be aliased back into the audible spectrum.

To quote Johnnie Cochran in the Chewbacca defence - "It does not make sense!"
 
henryrobinett said:
I regularly record and mix at 96k, though most often I do 44.1. I can hear a clear and obvious improvement at 96/88. But I'd rather keep it simple most times and save disk space. Although I have the capability to do 192 I don't and rarely have.
Click to expand...

Henry - I'd invite you to do a double blind ABX study on this and report back if you can still hear this difference. My opinion would be that you can't, and that the "clear and obvious improvement" you perceive is entirely due to confirmation bias. (Of course, opinions are nothing and evidence is everything, so I'll gladly retract if proven wrong).

For reference, the above article cites comprehensive study does by the Boston audio society which established that there is no discernible difference. I'll quote from the article:

"This paper presented listeners with a choice between high-rate DVD-A/SACD content, chosen by high-definition audio advocates to show off high-def's superiority, and that same content resampled on the spot down to 16-bit / 44.1kHz Compact Disc rate. The listeners were challenged to identify any difference whatsoever between the two using an ABX methodology. BAS conducted the test using high-end professional equipment in noise-isolated studio listening environments with both amateur and trained professional listeners.

In 554 trials, listeners chose correctly 49.8% of the time. In other words, they were guessing. Not one listener throughout the entire test was able to identify which was 16/44.1 and which was high rate."
 
We were just talking about this in another thread the other night. I've had instances where I could hear a huge difference between say 48 and 96kHz, but looking back on it I think that it came down to the native sample rate of the device in question and by changing it went through an onboard SRC to downsample it. The key here is that I was taking the digital output of the unit. It sounded pretty muddy and lifeless at anything less than 96kHz.

But just last night I recorded the analog outputs of the AxeFXII into my firewire audio interface at 48 and then at 96kHz. It is indistinguishable. I mean I tried to play the same thing (just mucking around) and I got confused a couple of times thinking that I hadn't switched or I had started the wrong one. There wasn't any additional noise, no change in response, depth, texture, flavor, color, sex, religion or race.
 
I just realized you had a link in your thread. :) That's a really good article even if it's a bit on the argumentative side, there is a ton of great information in there.
 
Shasha - glad you enjoyed it.

@Henry, I just wanted to ease up the tone of my earlier comment, on re-reading it seems a little confrontational and that was definitely NOT my intent - I have nothing but the highest respect for you. My comment was directly inspired by the section on confirmation bias in the above article. However without this context my earlier comment may have seemed a little derogatory, and so if any offence was taken I sincerely apologise.

:)
 
There are a couple of different scenarios here:

For reporoduction/recording, the experts are right: Any audio frequency below half the sampling frequency can be reproduced exactly (not close, exactly). Quality of A/D and D/A converters matters more.

For digital processing, other experts (such as Cliff) know what they're doing and I won't pretend to understand it fully. But for calcualting non-linearities, there are good reasons to use much higher sampling frequencies for greater accuracy of the end audible result. For example, IIRC the "high resolution" amp mode runs at a very high sampling rate (> 700KHz ???).
 
@ nikki - that prosoundweb link is awesome, many thanks. I'll have to get that book by Nika Aldrich too.

EDIT - Holy crap, that prosoundweb discussion has Dan Lavry and George Massenburg on it. (In Pro Audio terms this would be like having Vai and Satriani posting here).
 
Last edited:
Stratoblaster said:
Here is an interesting read on the subject:

http://www.lavryengineering.com/documents/Sampling_Theory.pdf
Click to expand...
Yeah I downloaded that about a year or two ago and have yet to get to reading it. Thanks for reminding me.

Oh and thanks for all the great links in here. I've got an upper respiratory infection so I've been home the last few days and I was up till frickin 4AM reading this crap. :) There's some stuff in there that's over my head, but there were a lot of explanations to things that I had the smallest idea of that were really driven home.

I like getting smarter.
 
Manning- I just saw this. I had done an inadvertent blind test already. I was sitting in on a mix of a cd of mine many years ago. I mean I was mixing it with an engineer. For three days we were mixing at 88.2. Then we started mastering. While he was setting it up I went to the bathroom I came back and nearly died. What had sounded so three dimensional was suddenly flat. I asked him what the hell happened he said we just converted to 44.1.

I've since become an engineer and mastering engineer. I've been around these debates for as many years. I think they're silly. I'm not really a nerd when it comes to reading white papers. I believe in my ears. It doesn't mean anything to me if I can't hear it. If I can it means the world. I record and mix at 96, 88.2, 44.1 with high quality converters. Arguments that I can't hear the difference are ludicrous. And most people have stopped that debate. People who HAVE to do null tests to tell them whether what they hear is what they hear is silly.

You can waste a lot of time getting lost in the minuscule minutia of random pointless detail. And what you lose is the time to write, play and record great music. What I hear is what I hear. Someone can try and invalidate what I hear. That's fine. But it's still what I hear. Tell someone what a banana tastes like or what the color green is.

No offense taken. :)
 
Someday, science will give us cool little DSPs that can do everything we want -- at 96k and higher.
Multiple Axe-Fx emulators will run on one little box.
 
Thanks Manning for helping re-affirm what my ears always figured anyway. I never imagined I heard any difference above 48Khz. I've never bothered getting too geeked up about the reasons behind that.. I just went with my ears.

My first interface that could do 96Khz was a Presonus Firestudio Tube, I remember doing some non-scientific A/B between 48 and 96 and thought the results were pretty much identical. Later on when I bought a Presonus Studiolive I didn't care in the least that it was a 'downgrade'.. maxing out at only 48 Khz.
 
The Nyquist theorem states that if you have the correct time and amplitude for each sample of a waveform, (and you are sampling at or above the Nyquist frequency of double the highest audible frequency) then you can reconstruct the waveform perfectly. This is mathematically proven, and it is indisputable.

The caveat is "correct time and amplitude". Hence the converters are what really matter.

To sample sound perfectly, you need (a) to sample at the correct frequency and (b) to correct measure the amplitude for the sample. (I've omitted anti-aliasing for simplicity, and anyway, if you don't get these first two elements right, the anti-aliasing doesn't really matter.)

So 48K means 48000 samples per second, not 48004 or 47995. Ditto for the correct amplitude - it has to be as accurate as possible, not "within 10%".

This is the real difference between the 26 channel, 8 mic preamp M-Audio 2626 which retails at $500, and the 2 channel, zero preamps, Weiss ADC2 which retails at $9000.

It's also why a studio might use an $8000 Antelope Isochrone which does nothing other than provide a clock signal for A/D converters. (For the record,I own neither of these devices, but oh boy, I wish I did).

While I still vehemently disagree with those believe there is an audible benefit to recording at 96K or higher, we can all agree that quality converters are what matters. Nyquist represents a theoretical ideal, and implementing that ideal as best as technically possible (at 44.1, 48 or whatever) is where the quality is found.
 
Great convertors is all that matters. If a person can't hear the difference between 48 and 96k I would suggest he needs better converters to teats this out.

But it's possible you didn't read what I said. PROVING stuff theoretically by theorems etc, means more than absolutely zero to me. Science has been wrong too many times. If one can't trust what one sees and hears as their first line of defense, god help us all.
 
henryrobinett said:
Great convertors is all that matters. If a person can't hear the difference between 48 and 96k I would suggest he needs better converters to teats this out.
Click to expand...

Again - all scientific testing done on this (using the same high-end equipment but with different sampling rates) indicates that it is impossible to tell the difference between 48K and higher sampling rates. If you agreed to do a rigorous, double-blind test using your own converters, I think you would confirm this also.

henryrobinett said:
But it's possible you didn't read what I said. PROVING stuff theoretically by theorems etc, means more than absolutely zero to me. Science has been wrong too many times.
Click to expand...

No I definitely read it. But your final statement looks like intellectual cheating to me. "Science has been wrong too many times"... Really? When? Give a concrete example relevant to this discussion.

Sure, science doesn't know everything, otherwise it would stop. But that fact doesn't invalidate what science DOES know, nor does it mean we can plug up the gaps with whatever wild notion takes our fancy. There is simply no evidence that we can hear ultrasonics and there is overwhelming evidence that we can't. Believing you can still hear them is no different to believing in ghosts, auras and unicorns. (And I've met people who absolutely believe in these things as well.) Of course if you want to keep believing in them I can't/won't stop you.

henryrobinett said:
If one can't trust what one sees and hears as their first line of defense, god help us all.
Click to expand...

Here's the heart of the matter. I completely disagree, our senses frequently cannot be trusted - this has been repeatedly proven beyond doubt. Our senses lie to us constantly, via confirmation bias and other mechanisms. I once got really sick and experienced both auditory and visual hallucinations which seemed extremely real to me at the time (despite me knowing they were illusions).

This doesn't mean that our subjectivity has no value, of course it does. But it does mean we should be skeptical about our senses as much as anything else, and to blindly follow our "feelings" at the expense of reason is not a wise course. To quote um... (I forget who, maybe Gibran) "...reason and passion must be as the wings of a bird, without both the bird cannot fly".

LOL, looking back this whole discussion is becoming a bit "Zen and The Art of Motorcycle Maintenance" - ie. the conflict between classical and romantic thinking. And it certainly won't get resolved here, you can't make people change how they think.

Anyway, keep doing whatever makes you happy.:D
 
You must log in or register to reply here.
Back
Top Bottom