What's the point of higher sampling rates in audio?, Such as 96 and 192kHz. Options

[stranhoROX]

Could someone explain to me why there are some recordings with sample rates as high as 192kHz? If most of us hear up to 20kHz, wouldn't 44.1kHz or even 48kHz be enough? Or is there other practical aspects besides boosting maximum frequency in higher sampling rates I am not aware of?

[2Bdecided]

@benski,

...but if you're simply relying on the high sample rate to keep processing aliases out of the audible band, you need to go into the MHz for some operations. Or do them in a smarter way. I agree that keeping the base Nyquist limit a little higher when you're resampling through each of many different DSPs can help in an objective and measurable sense: it can mitigate some of the effects of careless processing and/or resampling in each DSP module. This can be audible is the processing is bad enough - e.g. really trashy resampling. But is the improvement enough to turn trash into perfection? No.

C.R.Helmrich's response is right IMO - people do it because they can, and they believe it sounds better. Despite an almost total lack of any corroborating evidence. Plus a few people will pay more for a recording at a higher sample rate, for the same reason.

EDIT Examples:
http://www.soundkeeperrecordings.com/purchase.htm
http://www.paulmccartney.com/bandontherun/gbp.html

I guess the question becomes: why wouldn't you do it? If it costs nothing, and sells one more copy, or lets you charge some people more, or gets you bragging rights in some circles, then it's "worth it" - even if the technical benefit is zero.

Cheers,
David.

This post has been edited by 2Bdecided: Oct 17 2011, 10:11

[benski]

@benski,

...but if you're simply relying on the high sample rate to keep processing aliases out of the audible band, you need to go into the MHz for some operations. Or do them in a smarter way. I agree that keeping the base Nyquist limit a little higher when you're resampling through each of many different DSPs can help in an objective and measurable sense: it can mitigate some of the effects of careless processing and/or resampling in each DSP module. This can be audible is the processing is bad enough - e.g. really trashy resampling. But is the improvement enough to turn trash into perfection? No.
David.

Sure, but for polynomial equations (or polynomial approximations of transcendental functions), the aliasing is predictable. Because multiplying two signals M and N will produce a sideband at M+N, each polynomial order will require an equivalent increase in sampling frequency. Certainly there are other aliasing-reducing techniques such as using an all-pass for fractional delays and minBLEP for waveform generation, but for resonant IIR filters, especially, the higher sample rate can really make a difference (and admittedly most plugins probably upsample/downsample internally)

Note that I'm not at all trying to imply that high sample rate in final, delivered, consumer audio is justified. I'm just pointing out that doing the mixing and production at high sample rate is worthwhile. And if you already have a 192kHz master, why not try to sell it for a few bucks more?

[Northpack]

And if you already have a 192kHz master, why not try to sell it for a few bucks more?

Because you don't want to make your money by fooling people?

[2Bdecided]

And if you already have a 192kHz master, why not try to sell it for a few bucks more?

Because you don't want to make your money by fooling people?

It's retailing. It's based on fooling people, and people want to be fooled. They want to believe expensive food tastes better, for example. They want to read about all the reasons it tastes wonderful. The sunshine in the fields. The beautiful maidens who picked each crop by hand. Reading about all those reasons will make the food taste better to them - even though none of those reasons changes the actual taste of the food at all.

The biggest problem IMO is when there's only fooling, and real progress disappears where it might otherwise have been possible and beneficial. Also, where outright lies are told.

I don't mind a free market where several quality levels are offered; I can try them, and pay for the one I find acceptable.

I know certain people will claim that they hear differences I cannot. However, with all these quality levels available, I can set up rigorous ABX testing


Anyway, back to reality: if you are making high quality recordings, and some of the people purchasing your high quality recordings want to pay you $10 extra for a 192kHz version, why on earth wouldn't you make one available? As long as it doesn't make the quality worse, and doesn't cost you more than the financial return, it's really not a problem if people want to pay more for no tangible benefit.

I think it should be quite clear to anyone here that it's of no audible benefit what-so-ever, but it may create an excuse (that the accountants will accept) to create better (re-)masters. Which will then be used for the 44.1kHz version that can now be bought for a bargain price. Everyone's a winner.

Does this explain why parts of the industry are heading down this route? And those who should speak out, don't? Granted, it could be The Emperor's New Clothes all over again, but I suspect many people know exactly what they're doing.

The downside is that we don't get proper surround. Though some people are still quietly working on that too.

Cheers,
David.

[Dirk95100]

Anyway, back to reality: if you are making high quality recordings, and some of the people purchasing your high quality recordings want to pay you $10 extra for a 192kHz version, why on earth wouldn't you make one available? As long as it doesn't make the quality worse, and doesn't cost you more than the financial return, it's really not a problem if people want to pay more for no tangible benefit.

I think it should be quite clear to anyone here that it's of no audible benefit what-so-ever, but it may create an excuse (that the accountants will accept) to create better (re-)masters. Which will then be used for the 44.1kHz version that can now be bought for a bargain price. Everyone's a winner.

Not necessaraly so....

A general rule of measurements is that accuracy and measurement time are related. Low measuring times means low accuracy and high accuracy means long measuring times.
So there comes a time that while you think you increase the amount of information (increase sampling rates), you are actualy decreacing the amount of information. Again information theory explains this all.

[2Bdecided]

Anyway, back to reality: if you are making high quality recordings, and some of the people purchasing your high quality recordings want to pay you $10 extra for a 192kHz version, why on earth wouldn't you make one available? As long as it doesn't make the quality worse, and doesn't cost you more than the financial return, it's really not a problem if people want to pay more for no tangible benefit.

I think it should be quite clear to anyone here that it's of no audible benefit what-so-ever, but it may create an excuse (that the accountants will accept) to create better (re-)masters. Which will then be used for the 44.1kHz version that can now be bought for a bargain price. Everyone's a winner.

Not necessaraly so....

A general rule of measurements is that accuracy and measurement time are related. Low measuring times means low accuracy and high accuracy means long measuring times.
So there comes a time that while you think you increase the amount of information (increase sampling rates), you are actualy decreacing the amount of information. Again information theory explains this all.

There are two ways to realise this is misleading...

1) It's the RMS noise that typically increases as sampling frequency increases. The dB/Hz noise doesn't. So the noise level within the audio band remains roughly constant as sample rate is increased. When we didn't have enough bits to out-do the real world, the in-band noise fell as the sample rate increased.

2) No ADCs or DACs run at 44.1kHz natively. They run at a higher rate internally. The 96kHz output is no less accurate than the 48kHz output - both are usually derived from the same higher rate version internally. The 96kHz version can't be worse than the 48kHz version.

Cheers,
David.

[Dirk95100]

Anyway, back to reality: if you are making high quality recordings, and some of the people purchasing your high quality recordings want to pay you $10 extra for a 192kHz version, why on earth wouldn't you make one available? As long as it doesn't make the quality worse, and doesn't cost you more than the financial return, it's really not a problem if people want to pay more for no tangible benefit.

I think it should be quite clear to anyone here that it's of no audible benefit what-so-ever, but it may create an excuse (that the accountants will accept) to create better (re-)masters. Which will then be used for the 44.1kHz version that can now be bought for a bargain price. Everyone's a winner.

Not necessaraly so....

A general rule of measurements is that accuracy and measurement time are related. Low measuring times means low accuracy and high accuracy means long measuring times.
So there comes a time that while you think you increase the amount of information (increase sampling rates), you are actualy decreacing the amount of information. Again information theory explains this all.

There are two ways to realise this is misleading...

1) It's the RMS noise that typically increases as sampling frequency increases. The dB/Hz noise doesn't. So the noise level within the audio band remains roughly constant as sample rate is increased. When we didn't have enough bits to out-do the real world, the in-band noise fell as the sample rate increased.

2) No ADCs or DACs run at 44.1kHz natively. They run at a higher rate internally. The 96kHz output is no less accurate than the 48kHz output - both are usually derived from the same higher rate version internally. The 96kHz version can't be worse than the 48kHz version.

Cheers,
David.

http://www.lavryengineering.com/documents/...ling_Theory.pdf

[2Bdecided]

http://www.lavryengineering.com/documents/...ling_Theory.pdf

As others have pointed out, that's contradictory - it makes the point that virtually all sampling now uses hundreds of kHz or MHz...

For example, most front ends of modern AD
(the modulator section) work at rates between 64 and 512 faster than a basic 44.1 or 48KHz
system. This is 16 to 128 times faster than 192KHz. Such speedy operation yields only a few
bits. Following such high speed low bits intermediary outcome is a process called decimation,
slowing down the speed for more bits. There is a tradeoff between speed and accuracy. The
localized converter circuit (few bits at MHz speeds) is followed by a decimation circuit, yielding
the required bits at the final sample rate.

...but then claims...

Sampling audio signals at 192KHz is about 3 times faster than the optimal rate.
It compromises the accuracy which ends up as audio distortions.

So there's a point in the ADC where the input is a many kHz few bits signal. This can be converted to 48kHz, 96kHz, or 192kHz. In each case, the quality in the audio band is the same. Downsampling to 192kHz rather than 48kHz cannot introduce more audio distortions.


There is no technical point to using 192kHz - I agree with the paper in this respect - but this specific point "It compromises the accuracy which ends up as audio distortions" - or the idea that there's more noise in the audio band when sampling at 192kHz vs 48kHz, is wrong.

Cheers,
David.

P.S.

There is also a tradeoff between speed and accuracy. Conversion at
100MHz yield around 8 bits, conversion at 1MHz may yield near 16 bits and as we approach
50-60Hz we get near 24 bits. Speed related inaccuracies are due to real circuit considerations,
such as charging capacitors, amplifier settling and more. Slowing down improves accuracy.

This is true, but
a) it doesn't tell you anything about the noise in the audio band, and
b) in the context of 48kHz vs 192kHz, we're talking about the exact same analogue electronics (capacitors etc) - it's only the digital (mathemataical!) downsampling that's adjusted.

I am not denying information theory - I'm saying it's being misapplied.