snr advantage at higher sample rates

[Dan Lavry]

Originally Posted by Alexey Lukin

No. The overall power of harmonics is limited by the power of quantization distortion (which is independent of the sampling rate). But since they can spread to a wider frequency range, the audible power is reduced.

P.S. Dan, I hear you. I think that the original post mentioned a different type of noise: not the one from analog components, but the one theoretically achievable in a given digital audio format.

Theoertical noise? OK. That is the easy part, just math analysis and not that high level math. But it has little practical value. You do not have 24 bits converters, the analog noise gets into the lower bits. Real life gear quantization noise is DICTATED by analog components and circuits, not by truncation of a theoreticaly perfect converter, and more bandwidth means more noise and more errors.

Indeed a theoretical converter of say only 16 bits operating at 10000 fs can yield 3dB X 100 = 300dB better noise.... All you need to do is decimate it down by 10000... That is not a higher conversion rate, this is oversampling, because that only works when the signal bandwidth is 1/20000 of the sample rate. That is all easy to understand, and I do not see much value in it. A better approach is sigma delta where the noise shaping comes into play, enabling much less then 10000 fs oversampling. I already explained about the design tradoff of a noise shaper in a previous post.

Regards
Dan Lavry

[Jon Hodgson]

Originally Posted by Dan Lavry

The thread title is about higher sample rate and SNR. I do not see how dither got in here.

Hi Dan,

When you enter into discussions like this it's a good idea to remember that the other people are often less knowledgable than you, so they may be mssing stuff out or getting terminology wrong, as such a little effort to "read between the lines" to work out what is really being talked about can really help things go more smoothly.

It doesn't take a genius to work out that whatever termnology is used what Bob Katz and the Original Poster are talking about is oversampling, even if perhaps Bob isn't fully aware of it himself. What anyone who's either read a little on the subject, or thought about the DC level scenario I gave above will know is that for oversampling to give the 6dB resolution improvement with 4 x oversampling, dither is essential.

Since the OP categorically and incorrectly stated that the effect was not limited to a dithered system, that's raised a couple of questions from other people here... so that's where dither got in here.

Of course as you've pointed out the other glaring mistake made by the OP is to reject the real world issues of the input signal and the noise you get in analogue circuitry which makes this theoretical advantage completely useless when choosing a sample rate for your DAW.

If dither and noise shaping is the subject of interest, then I recommend reading my paper "Do you need 20 bits" on my company site.

To everyone else : Dan's white papers are very good, I recommend them.

[Waltz Mastering]

To simplify some stuff being said in this thread, is it safe to say that, if the final destination is to be 16/44.1:

With regards to "analog noise".

High sample rates (over 48k) do not provide a signal-to-noise advantage when working at a 24 bit depth.

High sample rates (over 48k) provides a signal-to-noise advantage when working at a 16 bit depth.

[rogerbrain]

[QUOTE=
With an undithered 12-bit audio source, there's no doubt



yeah, well, there ARE no perfect filters so higher rates seem to work better, at least to my ears.

All that aside, I read some studies last year that talked about a scientific study with audio recordings. They recorded an instrument to DSD, full bandwidth, directly off of the mic preamp. They put a 4th order crossover at 20KHz on the playback system and gave 100 subjects PET scans while listenig to the playback. So no matter how they played back the recording, there would always be the same filters at work in the audio so it was simply a matter of bandwidth. At first, they'd take a baseline reading with nothing being played. Then they'd randomly playback the full bandwidth audio, the same recording with the HF channel disabled and the same recording with the LF channel disabled. The results were frighteningly similar. The HF only playback was always the same as no recording at all. Now here's where it gets fun. The hearing center of the subjects' brains was the same whether using the LF only or full bandwidth audio. So that shows the subjects did not hear any notable difference between the recordings. Now, when listening to the full bandwidth playback, ALL SUBJECTS showed greater activity in the pleasure centers of the brain. So even though they can't consciously hear the difference, they can feel it. Now that's a 4th order filter which is way less steep than anything used in conventional digital audio and even that showed some pretty suprising differences in the subjects. I saw the normalized PET scans and even an amateur like myself could see the difference. I'll try and see if the papers are online.[/QUOTE]

NOW we are getting to the real point..sound and interaction with the sound (hopefully music).

Mr Neve has several talks about the need for extended, anomaly free frequency response. his notes on his new biz's page are pretty interesting. as are his talks at AES etc. I find it quite refreshing that he speaks more about enjoying sound and the parameters associated.

having worked 24 track 2" @ 30IPS with a good console, good room and good monitors I will say that the sound made me happy, and I if I got a good recording I had few issues mixing. this is something that did not occur using a DAW until the most recent incarnation which has the hardware and horse power to use high rez from start to finish.

(thanks to this thread) my plan is to record identical (close anyway) takes with the same mic and pre at both 192 and 96 and see if one actually sounds better.

would love to see the study you alluded to.

[wado1942]

(thanks to this thread) my plan is to record identical (close anyway) takes with the same mic and pre at both 192 and 96 and see if one actually sounds better.

See that's the thing, 192K DOESN'T sound better. It might FEEL better but nobody I know says they can hear the difference. That doesn't mean it can't noticably sound better but there's some technical issues that have to be solved first.

[It'sJoeAgain]

Originally Posted by Alexey Lukin

P.S. Dan, I hear you. I think that the original post mentioned a different type of noise: not the one from analog components, but the one theoretically achievable in a given digital audio format.

Not only that post read circuitry "noise" out loud to me but i always thought the initials SNR related specifically to *signal to noise floor ratio* not signal to quantize noise ratio, so WTF?? Hey Yo, this isn't the first time that a discussion gets heatedly debated and insults are thrown or flyin 'cause of some vague/ambiguous term used on the definitions man,....ppl need to be more patient and respectful and give the opposing users a chance to explain themselves especially when such parties sport real names or have verifiable credentials out there...

[Alexey Lukin]

Originally Posted by It'sJoeAgain

Not only that post read circuitry "noise" out loud to me but i always thought the initials SNR related specifically to *signal to noise floor ratio* not signal to quantize noise ratio, so WTF??

Whether SNR is dominated by the analog noise floor or by quantization distortion - is a question of the recording bit depth to me.

[It'sJoeAgain]

Originally Posted by Waltz Mastering

To simplify some stuff being said in this thread, is it safe to say that, if the final destination is to be 16/44.1:

With regards to "analog noise".

High sample rates (over 48k) do not provide a signal-to-noise advantage when working at a 24 bit depth.

High sample rates (over 48k) provides a signal-to-noise advantage when working at a 16 bit depth.

Not that i disagree with ur conclusions Tom, but i would say...that even 44.1k would do the same trick. Keep in mind that most young (or at heart) healthy ppl hear up to 18khz, ppl over 50 usually up to 15k...most ME's 17-19k and up() some blind ppl can reach 19-20k (i'll buy that)...some space cadets/weirdos swear they hear upto 22kHz().

So...44.1k/2= 22.05kHz has the whole world covered. regardless of any theoretical advantages of processing at higher rates...working at 24 bit gives u plenty of dynamic range to deal with *noise*.....

[Jon Hodgson]

Originally Posted by wado1942

All that aside, I read some studies last year that talked about a scientific study with audio recordings. They recorded an instrument to DSD, full bandwidth, directly off of the mic preamp. They put a 4th order crossover at 20KHz on the playback system and gave 100 subjects PET scans while listenig to the playback. So no matter how they played back the recording, there would always be the same filters at work in the audio so it was simply a matter of bandwidth. At first, they'd take a baseline reading with nothing being played. Then they'd randomly playback the full bandwidth audio, the same recording with the HF channel disabled and the same recording with the LF channel disabled. The results were frighteningly similar. The HF only playback was always the same as no recording at all. Now here's where it gets fun. The hearing center of the subjects' brains was the same whether using the LF only or full bandwidth audio. So that shows the subjects did not hear any notable difference between the recordings. Now, when listening to the full bandwidth playback, ALL SUBJECTS showed greater activity in the pleasure centers of the brain. So even though they can't consciously hear the difference, they can feel it. Now that's a 4th order filter which is way less steep than anything used in conventional digital audio and even that showed some pretty suprising differences in the subjects. I saw the normalized PET scans and even an amateur like myself could see the difference. I'll try and see if the papers are online.

Engineers should probably work hand in hand with some psychologists before comming out with the next revolution in digital audio.

I haven't seen the specific study (would be interested to), but I immediately see a major problem with the inference, which makes me wonder about the test.

The problem is... why should imperfect filters in an "inaudible" band lead consistently to a reduction in pleasure?

In the context of this test "imperfect filter" is going to be imperfect in one of three ways that I can think of, it affects frequency levels it shouldn't, it is calculated with too little precision and/or dithered incorrectly leading to noise/distortion, or it messes the phase up.

Now it is possible to get those errors to be very small, to the point where most (if not all) people wouldn't notice them if they were in the conciously audible band. Of those people that did notice I would expect that if you were to test for subjective preference over multiple people, the results would be pretty evenly distributed, one person's bright is another person's harsh and remember that we're talking differences in the order of blowing on the EQ controls.

But you move these differences up to an area where humans undoubtedly have considerably less sensitivity, and then suddenly everyone enjoys the same differences?

That just doesn't seem likely to me, assuming I've understood your description of the test correctly, I would be more convinced if the change in the pleasure centre attributed to the presence or absense of the imperfections of the filter were distributed above and below zero.

Edit : sorry, just re-read it and noticed that you only mentioned greater activity, not necessarily more pleasure? Or is one the same as the other?

How much of a difference did you see? Anything that wasn't really subtle would still strike me as very questionable, because we're still talking about tiny differences in areas where humans are not very sensitive, it makes little sense that they should have much effect relative to everything else going on.

[Waltz Mastering]

Originally Posted by It'sJoeAgain

i would say...that even 44.1k would do the same trick.

I was going to say 44.1, but since Dan L. sited 48k as a reference in one of the post concerning this, I thought that would be a safe bet. But yes , saying 44.1?

[It'sJoeAgain]

Originally Posted by Waltz Mastering

I was going to say 44.1, but since Dan L. sited 48k as a reference in one of the post concerning this, I thought that would be a safe bet. But yes , saying 44.1?

that's imo to make his filter render the math/processing a little easier (48,96, 192kHz) than filters internally operating at 88.2kHz. That's why i assume working with a great converter is more important than which high sampling rate u choose.....maybe i am wrong...

[Bob Olhsson]

Any filter problems are going to be artifacts generated down in the audible band.

The reason I advocate working at double sample rate is better sounding DSP. On the fly upsampling doesn't sound as good to me as simply working at the higher sample rate whenever it's practical. In many cases the low frequencies seem to thin out with on the fly upsampling. I have no idea why.

[Dan Lavry]

I re read the first post, but I am an engineer and a gear designer an I live in the real world, as most of you guys.

The first post is about dither and noise shaping. That kind of a process is there for word length reduction (to overcome low level signal quantization distortions and noise modulation). The CD format is 16 bits, so there are times when we need reduce word length. When one adds dither (with or without noise shaping), the noise level goes up, so it is not an improvment in SNR, it is a trade off between
more noise and less distortions and noise modulation of low level signals.

But the process of archiving or music and the released in 88.2-96KHz few "real world" formats does not require dither and noise shaping. Why trade off if you do not have to?

But let me try an go along with the theoretical concept here for a moment. If one has a system that is virtually noiseless, and the sample rate is say 88.2KHz, one can noise shape the signal perfectly (in theory). With an 88.2KHz system, the bandwidth is 44.1KHz, so take the noise and move it from under say 25KHz to the upper frequencies say 25Khz-44KHz. That is a huge range for noise shaping.

In the case of noise shaping a 44.1KHz, we move the noise from the frequency hearing sensitive range (such as 1-3KHz) into less sensitive range such as 14-22KHz. That upper range is still within the audible range. But at 88.2KHz-96KHz, we have a lot of extra non audible range to shape the noise into - plenty of room in fact.

So why go to higher sample rate then 88.2KHz? The noise shaping argument does not call for 100MHz range. Yes, in theory, one can have a real easy time moving all the noise if the sample rate was approaching infinity.

In fact, in theory, there is no down side to sampling at 1GHz, there is no theoretical limit to storage, to DSP speed, there is no real world noise and distortions.

One more time, we are faced with decisions that are best left to real world considerations, and much of it is about compromises that steer us toward optimal points.

Given that 88.2-96KHz yields enough theoretical and even practical range to noise shape outside the audio range, we can conclude that higher and higher sampling is not needed, thus the argument that higher sampling for better SNR should be put away.

Instead, we can say that high enough sampling such as 88.2-96KHz yields enough range to design a noise shaping capable of moving virtually all the noise from audible frquencies to higher inaudible frequencies.

I never thought I will see the day where Dither and SNR will become a factor in a discussion about the trade off between sample rate and SNR. Not even a theoretical discussion.

I view this as just one more attempt to support and rationalize higher sampling rate, and like the other various arguments, it ends up at 60-70KHz being the optimal rate. 60-70KHz yields enough range to park the unwanted truncated noise away to non audible range (say 22- 30KHz for a 60KHz sample rate).

regards

Dan Lavry
Lavry Engineering

[It'sJoeAgain]

Originally Posted by Dan Lavry

I view this as just one more attempt to support and rationalize higher sampling rate, and like the other various arguments, it ends up at 60-70KHz being the optimal rate. 60-70KHz yields enough range to park the unwanted truncated noise away to non audible range (say 22- 30KHz for a 60KHz sample rate).

thanks Dan, that makes a lot of sense on paper, but how much can a mortal discern these quality differences in the real world, using as an example, speech or even classical orchestrations?? How obvious these differences are and can be heard (between the *standard* 44.1k and the 60kHz or that extra 8 kHz of bandwidth)

[Jon Hodgson]

Dan,

thanks for that, perhaps you can give people here some more real world perspective on another aspect of conversion.

A repeating mantra of many people who either think that digital can never be good, or insist that sample rates should be as high as possible, is the one "there are no such things as perfect filters".

Now aside from the irony that this line so often comes from people who absolutely love analogue recorders, which are far from perfect in themselves, there seems to be little discussion as to how much things deviate from the theoretical ideal and how perceptible that is likely to be.

So, typically speaking, how imperfect are these filters? In what ways are they imperfect? How much are they affecting the passband? How much aliasing noise does imperfect attenuation in the stop band let through? How audible is any of this likely to be?

I think it would be good for people to get some perspective on the performance of real world anti-aliasing filters. I could try working out the numbers for myself, I know, but I'm sure you already did that, and of course you have the experience of how they actually translate to audibility in converters.

[Dan Lavry]

Originally Posted by Jon Hodgson

Dan,

thanks for that, perhaps you can give people here some more real world perspective on another aspect of conversion.

A repeating mantra of many people who either think that digital can never be good, or insist that sample rates should be as high as possible, is the one "there are no such things as perfect filters"....

So, typically speaking, how imperfect are these filters?

I think it would be good for people to get some perspective on the performance of real world anti-aliasing filters....

OK. I will respond, and it is going to be long and slightly technical and some folks may want to skip it. If it is too long, please read the last 2 paragraphs.

In the good old days (such as before 1990), filters were a real limiting issue. When and AD recorded at 44.1KHz, one needed to make sure and filter out the energy above 22.05KHz that, but at the same time, the goal was to pass all the energy below say 20KHz intact. That task is about analog filtering. It is really impractical and perhaps not even possible to do a good job of such a filter.

The transition band is only around 2KHz – pass 20KHz but block 22KHz, and that is a real problem. All analog filters are circuits based on a concept of poles and zeros. To simplify, a pole or zero means an attenuation slop of 6dB per octave (20dB per decade). So a single pole filter that passes 20KHz, would yield only 6dB attenuation at 40KHz. So take 10 poles, and you have 60dB attenuation at 40KHz. So with 60dB over 20Khz, you end up with only 6dB over 2KHz transition band. That is “almost doing nothing”.

So how about 20 poles? That would yield only 12 db at 22KHz. So lets try for 100 poles, and on paper we will have 60dB attenuation at 22KHz. That too is not good enough. Now, no one designed 100 poles filter. 10 pole was closer to reality. I never heard of a 20 pole design. There are reasons for that. Each pole calls for a precision cap, and for a precision resistor. In fact, most analog filters have at least 2 caps and 2 resistors in an op amp circuit, for a pair of poles. So 100 pole design would call for 50 op amps, 100 precision caps and 100 precision resistors.

You really do not want to send your audio through 50 op amps. The precision requirement of high order R an C values gets impractical, the power dissipation, the cost and so on is just too much.

Also, an ANALOG filter that is very near 20KHz, brings about some serious phase non linearity issues, which can be somewhat compensated for when the number of poles is not too high, but with more poles, the phase curve is also an issue.
Well, what we did is “our best”. One way to help matters was to move the pass band to say 17-18KHz, making the transition band wider (22-18=4KHz, twice that of 22-20=2KHz). We also tended to rationalize that instrument level at around 20-22KHz is lower then full scale, so the filter requirement is relaxed by a few dB.

To summarize, things were far from idea or desirable.

Around 1990, the new concept of up sampling and oversampling came about, and that virtually solved the AD anti aliasing filters and the DA anti imaging filter problems. Say I over sample by only a factor of 2, (88.2KHz sampling) then my requirement is to pass 20KHz and below but block 44.1KHz and above. The transition band is now 24.1KHz, and that is 12 times easier filtering (12 times less poles). But way stop there? Lets go for say 64fs or 128fs or 256fs or 512 fs or even 1024fs.

Lets try for a 256fs AD. The sample rate is around 11.2896MHz (Nyquist is at around 5.6MHz). Say we want to pass everything up to 50KHz (no longer restricted to a tight 20KHz). The transition band is now 5.6MHz – 40KHz = 5.56Mhz (the 50Khz pass band is almost un importent- almost out of the calculation). You now have 5 MHz range to drop from full low pass at 50KH to blocking the energy above 5MHz.

A pole has 20dB per decade slope, and we have 2 decades (50KH, 500Khz, 5MHz is 2 decade). So a single pole yield 40dB. Take 3 poles and you have 120dB attenuation of any energy above 5MHz. The filter is low order (3 poles) so it is practical, and one can do it all with a single or a pair of Opamps. Low order filers have more linear phase, and we placed the poles at 50KHz not at around 20KHz so with a good design, the phase issue are moved to frequencies above hearing…

Folks get confused because they point out that 192KHz as having higher transition band then say 96Khz. And here I am talking about 11.2896MHz. So what is going on? Well, the FRONT END of a modern AD, as well as the BACK END of a modern DA operate at very fast rates (in the many MHZ). It is true that the end result is at many more bits at much lower rates, but from a filter design stand point, form what is needed to avoide aliasing and imaging, it does not matter. The filter of an up sampling or oversampling device needs to reject the energy above the modulator rate not the “final” AD data rate or the “Initial” DA data rate.

But this is audio. Folks heard something 50 years ago, and they hang on to the false notions for the rest of their lives. The fact is that with modern AD and DA’s the filter for a 44.1KHz is very similar to a filter for a 96KHz.

Please note that I am answering you about the anti aliasing filters which are analog filters. I included anti imaging filters which are also analog and the situation is very similar to anti aliasing (one is for an AD the other for DA). There are 2 more filter types that are digital, one is in the AD decimation, the other is in DA up sampling.

The digital filters at a 44.1KHz are somewhat tight, but one can do a real fine job, if one is willing to put the needed processing resources in, especially if one stays away from real low latency conversion, which is usually not a needed feature. At 88.2KHz there is plenty of room to do a real fine job. One can make a perfectly linear phase digital filter, but for low latency, people do what they can to reduce time delay, and they “cut corners”, so the phase becomes less then linear. Some argue that it does not matter that much. I think it needs to be better quantified, it may not matter to 1 degree of phase, but it may matter at 10 degrees…

I am not saying that filters do not matter. Everything matters to some degree. But the analog filters of 20 years ago mattered a lot. In fact, the analog filters were the bottle knack in performance, and they imposed very difficult compromises. But in today’s world, analog filters issues are not the overwhelming factors, and it is not just due to slow and constant evolution, it is due to the introduction of newer concepts (up sampling and oversampling). The improvement was a “sudden large step up”.

It is long past due that audio professionals stop dragging concepts from 2 dozen years ago into discussions about today’s technology. So much of the chatter about audio is about things that mattered long ago and matter so much less now. Technology does move forwards as some of you folks realize. We now have air in the tires, and gears in the cars, some cars have automatic transmission :-)

Regards
Dan Lavry

[Dan Lavry]

Originally Posted by It'sJoeAgain

thanks Dan, that makes a lot of sense on paper, but how much can a mortal discern these quality differences in the real world, using as an example, speech or even classical orchestrations?? How obvious these differences are and can be heard (between the *standard* 44.1k and the 60kHz or that extra 8 kHz of bandwidth)

I am not one to advocate going to real high sampling rate. In fact I am against it.
Most mics, speakers and ears do not operate above 20KHz. The only reason to extend the range is to yield some margin so you can do a better job at around 20Khz.

I am not doing much talking about 24 bit audio either, because there is no such a thing, and we really do not even need 144dB dynamic range. In virtually all cases the noise out of a mic pre is already a bottle knack, making 120dB a very rare case (that is around 20 bits), and yes it is analog noise.

I am still trying to figure out why folks are trying so hard to focus on what they can not hear (higher frequencies and unreal noise floors) instead of focusing on what they do hear.

Question:
Would you trade off 48KHz frequency range for say 1KHz frequency range?

I would not if we are talking about trading the 0-1KHz audible range, where so much of the music is, for the useless 48-96KHz non audible extra range offered by goind from 96KHz sampling to 192KHz sampling....

44.1KHz may be a bit tight for people with high frequency hearing, because with a 20KHz mic, 20KHz speaker, 20KHz AD and 20KHz DA, you have 4 factors hammering the 20KHz down. Each one contributes 3dB loss at 20KHz so you end up with -12dB at 20KHz, and the real bandwidth moves down to say 18KHz (?)

If you get the AD and DA up to say 60KHz, you end up with only mic and speaker limitations thus -6Db at 20KHz. That is more of an issue for young people, and less so for my 64 years old ears.

Regards
Dan Lavry

[Waltz Mastering]

^^^^^
For the vast majority of audio engineer's, beyond pressing buttons, and trying to make things sound good, there are some very complicated and ever changing technologies going on in the background that can be a little hard to get a handle on, so thanks to you guys who are "in the know" for helping to explain some of the cutting edge stuff logically.

[wado1942]

That's some great info DL. I do have to take issue with something though.

In virtually all cases the noise out of a mic pre is already a bottle knack, making 120dB a very rare case (that is around 20 bits), and yes it is analog noise.

The noise floor has little to do with usable dynamic range. A guitar running through an amp usually has a 40-50dB S/N ratio but you better believe it sounds way better going to a 16-bit recording than an 8-bit recording, which according to your specs is all you would need in that case.

It's also been proven already that most people can't even hear above 15KHz, much less 20KHz. But research that's spanned decades, both official and accidental, shows that people are aware of when there's a super-sonic issue. I don't know if their minds are registering an unexpected aurel issue or if it's something more to do with feeling through bone conduction or whatever. Perhaps it's just HF stuff interfering with LF information. Now at this stage in the game, I agree that 192KHz converters should be avoided but that's another issue alltogether.

Anyways, why not just leave out the decimation process and store the raw sampled data? You leave out the digital filters, simplify the reconstruction etc. I know storage was the issue at one point but when you can buy a 250GB USB HDD for $60 at the local electronics store...... I know there's a lot of debate over the digital filters being audible, especially on top quality converters but still.... If you can avoid a potentially lossy process, shouldn't you? I always learned that the shorter the signal chain can be to achieve a desired sound, the better.

Edit : sorry, just re-read it and noticed that you only mentioned greater activity, not necessarily more pleasure? Or is one the same as the other?

Greater activity in areas associated with pleasure indicates the test subjects enjoyed the listening experience more with the full bandwidth material. Remember, the playback was always running through the crossover so the influence of the filters were always in place. They simply shut off the feed to the HF amps for the 20-20K tests.

How much of a difference did you see? Anything that wasn't really subtle would still strike me as very questionable, because we're still talking about tiny differences in areas where humans are not very sensitive, it makes little sense that they should have much effect relative to everything else going on.

It was a normalized PET so it's hard to say what's subtle or major. But compared to the other activity, I'd say it was moderate. Quite a few pro musicians can tell the difference between 44.1K sampled stuff and 96K blind folded and you know a lot of them can't even hear up to 10KHz. There's more at work here than people think which is why I'm pushing to combine the sciences. I was a psych major in college BTW. Physiology and to a lesser degree, psychology is in its infancy, but they're learning about all sorts of things in the human condition that seem impossible all the time. I say, if people seem to know the difference, why fight it?

[Waltz Mastering]

Originally Posted by wado1942

Quite a few pro musicians can tell the difference between 44.1K sampled stuff and 96K blind folded and you know a lot of them can't even hear up to 10KHz.


I say, if people seem to know the difference, why fight it?

It's a coin flip...?
Interesting and controversial article about that here:
The Emperor’s New Sampling Rate -- Are CDs Actually Good Enough?

[Dan Lavry]

Originally Posted by wado1942

That's some great info DL. I do have to take issue with something though.

The noise floor has little to do with usable dynamic range. A guitar running through an amp usually has a 40-50dB S/N ratio but you better believe it sounds way better going to a 16-bit recording than an 8-bit recording, which according to your specs is all you would need in that case.

It's also been proven already that most people can't even hear above 15KHz, much less 20KHz. But research that's spanned decades, both official and accidental, shows that people are aware of when there's a super-sonic issue. t it?

First, the discussion was about dynamic range and sample rate so I responded about dynamic range. I did not talk about distortions, but that does not mean that distortions do not count or that more bits are not better (to a point). That is a whole other topic, and there is a lot to be said about it, as well as about other topics.

You said: "The noise floor has little to do with usable dynamic range". Sorry, but that does not make sense. I think you are using the term "usable dynamic range" as a mixture of noise and distortion (you pointed out that an 8 bit did not sound good and it was not just the about dynamic range). Fell free to clarify, if you wish.

BTW, a guitar has only 40dB-50 dynamic range? That doe not registers. You play a loud note all the way to the peak, then you stop playing, the room is quite and you will have much more then 40-50dB.

When you state that "it has been proven that people hear supersonics" you should provide some reference or some links. To my knowledge there was one study by Pioneer, around the mid 1990's that suggested that folks can react to up to 27KHz or so. That is when I decided to go for 88.2-96KHz, which is more then enough to cover that range. There where two later experiments trying to get the same results, and in both cases the experiment failed to show sensitivity to 27KHz.

Also, what I said about mics and speakers is true in most setups. They limit the audio to around 20KHz, and that is because the ear does not need higher. There are a few recorded cases of folks hearing in the low 20KHz range such as 22-23KHz. I assume you know that the lowest bandwidth device in the audio chain determines the bandwidth. Say your mic is 20KHz but your speaker is 30KHz, then you end up with the lower frequency - 20KHz.

Regards
Dan Lavry

[It'sJoeAgain]

Originally Posted by Waltz Mastering

It's a coin flip...?
Interesting and controversial article about that here:
The Emperor’s New Sampling Rate -- Are CDs Actually Good Enough?

Tom, that's a very interestin article i gotta say...and it's also waht i always say: too much self-interest/self-promotion/personal agenda/placebo bullsh*t goin round on forums, man. It tears at the notion that we are daily visited by BatMEn, CatMEn, MiceMEn, u know?? every human creature that hears super high frequencies or claim they do better masterin 'cause they hear well above 20k

"If you assume the equipment, the listening environment and the listeners' critical faculties are all at least good, then what's most amazing about their findings is that the results were always the same, no matter what equipment they used or who was listening to it or what they were listening to. Not one listener, under any circumstances, could consistently distinguish between high-resolution audio that was passed through the 44.1kHz/16-bit CD “bottleneck” and audio that wasn't."

and imo if u want to have a *theoretical* advantage then be workin/processing/listening at 192k/24 bit...otherwise 44.1k/24 is all the ice on the cake u'll ever need....

[rogerbrain]

Originally Posted by wado1942

Now at this stage in the game, I agree that 192KHz converters should be avoided but that's another issue alltogether.

Hi again.. I am probably the least technical guy logging in here. This is fascinating and VERY enlightening and I really appreciate the time you all are spending sharing real world knowledge.

Why should 192 sample rate be avoided? is there a universal quantifiable disadvantage?


Are these correct statments?

In a given converter:

1)192 'can' give more audio bandwidth (audible or not)

2)If the filters, being an analog componant and fixed, are designed for 96, 192 sample rate would yield no extention of audio range over using the 96 sampel rate.

3)There will be more noise (or is that better expressed SRN) at 192 than at 96

4)more data will be recorded using the same bit depth at 192 than at 96.

Also, what specs should be looked at when deciding what converter and resolution to use?

Is the data even available (to me) to make proper choices?

thanks

[Bob Olhsson]

Originally Posted by rogerbrain

...Is the data even available (to me) to make proper choices?

No it isn't! Folks are grabbing at numbers out of thin air while totally ignoring the engineering trade-offs involved.

I know some folks I respect who consider a certain converter run at 192 to be the very best they've heard. Does this mean 192 is better? Not at all, just that some factor in the engineering design compromises of that particular converter made it sound better. Maybe it's something about the clock being better isolated from the analog stage at 192. Who knows but you can't generalize from any number.

I'm reminded of tube mike preamps. They always sounded quieter than early generation solid state preamps despite measuring 10 or 15 dB. noisier. Why was this? They had a much greater dynamic range between acceptable noise and an acceptable level of audible distortion. If you measured output at 5% harmonic distortion, they had 20 dB. more headroom than their transistor brethren. If you measured at .5%, the solid state crap looked vastly superior on paper. A friend of mine built a prototype for Deane Jensen in the mid '80s that was the first solid state preamp I'd ever heard that was as good as a late '50s tube mike pre.

Numbers have always been used to lie about audio gear. Dan refuses to play that game. I can't admire him enough for that.

[It'sJoeAgain]

Originally Posted by rogerbrain

Are these correct statments?

In a given converter:

1)192 'can' give more audio bandwidth (audible or not)

Correct and nyquist frequency of 96kHz which is way beyond what earthly creatures can ever hear...

2)If the filters, being an analog componant and fixed, are designed for 96, 192 sample rate would yield no extention of audio range over using the 96 sampel rate.

Correct...the converter doesn't extend the source's bandwidth

3)There will be more noise (or is that better expressed SRN) at 192 than at 96

Correct, more *theoretical* noise...

4)more data will be recorded using the same bit depth at 192 than at 96.

Correct again, more memory requirements.......
I beleive u are now an expert in converters...

[wado1942]

You said: "The noise floor has little to do with usable dynamic range". Sorry, but that does not make sense. I think you are using the term "usable dynamic range" as a mixture of noise and distortion (you pointed out that an 8 bit did not sound good and it was not just the about dynamic range). Fell free to clarify, if you wish.

You're correct in your assumption. I think distortion and noise should be measured together, particularly since a lot of people don't know the difference, especialy where digital technology is involved. For instance, if I recorded a guitar to 12-bit digital, I'd have to keep the level pretty high in order to sound acceptible. That's due to the distortion being so high at lower levels. However, on 1/4" tape, I could keep it comparatively low and while there may be more noise, the overall sound is much less irritating. This means the usable dynamic range of tape is wider than 12-bit digital even though the 12-bit medium may have lower noise.

I assume you know that the lowest bandwidth device in the audio chain determines the bandwidth.

Of course, but there inlies a problem because the definitions are a bit blurry where this comes to play. For instance, I have some microphones that are +-3dB 10Hz to 25KHz or so. Now if you record those though my preamp with 50KHz +-1dB, that bandwidth is preserved. Comming back off of tape, the playback maybe -3dB at 20KHz but I can see on an FFT that there's valid information going clear up to the limits of my testing ability on 96KHz converters. What I'm saying is, even though our ears may be -45dB or something (just an uneducated guess) at 25KHz, we're still aware of and expecting information up there at least on a subconscious level. When you put a 50th order filter on the signal at 20KHz, that expectation is not met. As subtle as it may be, the natural world does not behave this way. So it's no surprise to me in that test I mentioned that even though the subjects did not HEAR the difference, they FELT it. Be aware, also, that your brain ramps up the gain and fills in blanks so to speak with what we percieve. Like with our eyes. The image is very blurry, distorted and missing all sorts of information. But we work in conjunction with memory, combining experiences from 10ms ago with 10 years ago, we can give a glance at something and see a sharp, detailed image. Case in point, people with cochlear implants who had hearing recently say that they can identify and follow songs they knew before they went deaf. However, a new song on the radio just sounds like noise to them. Granted, that's probably in part due to how music is made now, but all jokes aside, it's more evidence that no EE or MD or whoever else knows even a fraction of what there is to know about perception. BTW, did you know that you can actually hear sound directly with your brain?

Numbers have always been used to lie about audio gear. Dan refuses to play that game. I can't admire him enough for that.

No kidding. I have a live console that claims to be a +4 device but it hard clips at +8. I actually keep the average output around .38V because it starts to get a little fuzzy if I go too much above that. Then there's that little Behringer that I use for routing signals to my talkback monitors which claims to be 10-channel when in fact it's 4 because they're counting the tape & AUX returns, and the stereo channels are counted double.

Correct and nyquist frequency of 96kHz which is way beyond what earthly creatures can ever hear...

Well, humans anyway.

Why should 192 sample rate be avoided? is there a universal quantifiable disadvantage?

No, I said "at this stage" because there just aren't any 192KHz converters that outperform 96KHz converters. But that's part of the argument I've been making. Instead of upping the clock to get higher sample rates and continue decimating, we should just store the raw sampled data. Then we'd have all the sample rate we'd ever want. When advancements in filtering & such are made, it can be applied in software offline. If you ever wanted to use DSP on your recorded audio, you can just add in some empty less significant bits, maybe an extra 12 or so that can be dithered back to 4 or whatever and process to your heart's content without upsampling. We'd need to rewrite the books on how DSP is handled because we'd be dealing with Delta slopes instead of hard, set voltages but, that's why I say "this stage".
On playback, you'd only need analogue LPFs for anti imaging purposes.

[Dan Lavry]

Hello Wado,

You said:

"For instance, if I recorded a guitar to 12-bit digital, I'd have to keep the level pretty high in order to sound acceptible. That's due to the distortion being so high at lower levels".

I say:

Yes of course. If you drop the level of a 12 bit converter to around 70dB below FS, you end up with almost square waves. The errors are no longer just random noise, they are signal dependent and are rather bad sounding. That is because the steps are too big in relation to a low level signal.

But there is some relationship between dynamic range and the distortions you are talking about. In order to get a huge dynamic range, the steps must be very small, so one can go for lower level signals before the quantization errors will sound that bad. With 8 bits, you drop the level to 1/256 of full scale (around -50dBFS) and the signal is either between 2 quantization levels (you hear nothing) or is crossing a single quantization up and down (horrible sounding).

But with say 20 bits, you drop the signal by 50dB from full scale, and you still have around 4096 quantization levels (12 bits) to play with.

I agree that there is more to the story then the size of a quantization step, but dynamic range is importent, to a point. Personally, I would not touch a converter with less then 100dB noise floor.

You said:

"Of course, but there inlies a problem because the definitions are a bit blurry where this comes to play. For instance, I have some microphones that are +-3dB 10Hz to 25KHz or so. Now if you record those though my preamp with 50KHz +-1dB, that bandwidth is preserved. Comming back off of tape, the playback maybe -3dB at 20KHz but I can see on an FFT that there's valid information going clear up to the limits of my testing ability on 96KHz converters... "

I say:

The definition of bandwidth is the points where the power drops to 1/2 of the usable range. The ear is not an FFT. I can have an FFT show me energy spikes at 1MHz, and I will never hear 1MHz.

Devices are not "brick wall" so the frequency response has a slop to it. But you can add the responses of various device (for a dB scale vs, frequency plot), and the sum is the outcome. So if your mic is at say -3dB at 20KHz, and your tape is at -.001dB at 20KHz, the overall is -3.001dB. Say your 20KHz mic has a slop such that at 25Khz you are down by 10dB, and your tape perfect at 20KHz (0dB at that point). The outcome is -(10+0) = -10dB. You can not get better response then the lowest device.

Given that we do not hear much above 20KHz, and certainly not anywhere near 44.1KHz, using 88.2KHz sample rate is more then enough for all times. Given that we can reconstruct EVERYTHING we can hear by sampling at 88.2KHz and even lower, we will never need higher sampling rates.

People with cameras do not care about capturing non visible light, because such wavelength do not impact what we see. I do not know why folks that insist on using 192KHz are not insisting on video gear and camera that can record and display infra red, utlra violate, and why not X rays.... Making such a camera would bring about all sorts of compromises, performance trade offs and costs for no good reason. Same with 192KHz for audio.

We found out long ago that humans require around 20KHz. That is why the old hi fidelity was specified for 20-20KHz. It was an ear driven decision, and it was based on research. People that claim that we hear higher, should have an iron clad, repeatable proof that such is the case. Once such claim is proven, and the hearing range is QUANTIFIED (such as 20KH, 22KHz or 30KHz or whatever), then it is time to consider recording such a range.

An argument that one day we may find out that we can hear higher, does not cut it. Why not argue that one day we may figure out that we can see X rays, and all cameras should accomodate such a range? We do have research about hearing range and visable range and the research stands until such time that it is found to be wrong, which is not particularly likely.

I do not understand the fascination with what we can not hear, other then marketing forces that try to sell you new gear, which WAS the driving force behind 192KHz! I am glad it mostly died down. I did what I could to stir audio back to a more sensible range 44.1-96KHz, and it did take some serious effort to make it go away.

Regards
Dan Lavry

Regards
Dan Lavry

[wado1942]

I do not know why folks that insist on using 192KHz are not insisting on video gear and camera that can record and display infra red, utlra violate, and why not X rays

All CCD and CMOS cameras DO RESPOND to infrared. They put in filters to stop it. I have a security camera that simple doesn't have one of those filters, and if you put an IR pass filter in front of the camera, you can see what's in people's pockets. But I digress.

People that claim that we hear higher, should have an iron clad, repeatable proof that such is the case.

Right, I'm agreeing that we don't need 192K converters, at least at the moment. I'm also agreeing that we do not hear above 20KHz, at least not consciously. However, what we do not hear can influence us. Just like how we can feel the influence of IR and UV on and in our bodies, I think we can also feel the influence of 5Hz vibrations and 30KHz vibrations. Now, I actually think 88.2-96KHz is ideal with the current state of technology. I do happen to have a set of 192K converters but the only time I use them as such is when I'm grotesquely slowing down audio for special effects since I mix movies and videos as part of my profession. What I'm saying, in effect, is that we should eventually drop the decimation process, not to have higher bandwidth, but to eliminate other stages in the process that I think are working against the purity of the sound.

[Dan Lavry]

Hi Wado,

Ok, I did not know that CCD records IR, but that is not the point. You want the camera to do a fine job on the wavelength people see. You see, you need to remove the infrared, for a reason. The same with audio.

There is a certain upper frequency range that will include all that we respond to, be it direct sound, mystery brain waves and what not. At some point, there is a limit. You would agree, I hope, that 1GHz is a bit much. So how about 20KHz?, 30KHz?

Well, 40KHz is very high to include even the most psychic vibrations, and 88.2KHz will accommodate that.

Now, say we somehow get to respond to some 100KHz energy, because it finds its way into the audible range. If that is the case, I would expect that we will hear that "effect" in the performance space. If I do not hear it there, I do not want to hear it later either. Why record it?

So say you have extra bandwidth, and you record the 100KHz energy, and then you play it back. This time, you have DOUBLED the impact of the 100KHz energy. First it took place at the performance space (energy into the audio band), then it happens again at the playback space.

The way the high frrequency (100KHz in the example) gets down to the audibble range is by a distortion mechanism, typically intermodulation. But if you do not record that high 100KHz energy, the audio will not have a 100KHz energy to intermodulate against (the sound of one hand clapping).

So here is the reason for NOT including what you do not hear. It will not help, it may hurt, and if it is not there it can not hurt. Going for much more bandwidth then you need is not a good practice.

I do appreciate much of what you said. Be sure I am not arguing with you. I have taken the opportunity to express some of my observations.

Regards
Dan Lavry

[It'sJoeAgain]

Originally Posted by wado1942

Well, humans anyway.

i guess i meant to say ground mammals...here...since u like to get *technical* about things i gathered this info for u (noticed that no ground mammal reached 96k):

dolphins allegedly can reach up to 220 kHz with their sonar!
crickets can do ultrasonic chirps by rubbing their fore wings together of up to 130k
frogs emit ultrasounds around 128 kHz.
bats calls vary from 14 to over 100 kHz and emit from 10 and upto 250 clicks per second!
cats 64k (meow )
rats make high frequency calls of up to 50 kHz.
dog (me ) 45kHz...........................