Analog Frequency Response

[Lawrence]

I'm on a bit of a crusade for an answer that nobody seems to have. When something irritating gets in my brain I just can't let it go until I have an answer... I'm anal like that ... so here I am.

I'm asking people here to go look in the manual or on the mfg website or anywhere for the high sample rate converter or digital recording device you own and tell me if there is a published analog frequency response measurement. I'm looking for frequency response measurements of their analog circuits only as in...

20hz - 20khz +/- 3db @ 1k or whatever ...

High end, low end whatever. Digital consoles with 96k converters or more (X-Bus?), Roland tabletop daws with converters, MOTU, RME etc.

Then come back to this thread and tell me if you can find a published spec of analog frequency response for the converter (the analog circuit feeding the converter) or the high sample rate device you own that comes anywhere near 1/2 or even 1/3rd of the higher sample rates your device can record at.

If you can't find it and/or it's not in the manual please post that also.

I think the possible implications of my request are pretty obvious. Please post results and links to results if they're on the web.

Thanks in advance.

[Lawrence]

I'll kick it off with the API A2D - which has a 192k capable converter built into it and the 192 I/O which many HD users have.

A2D 'Analog Section': +0, -5, 10hz to 20khz (-.5 at 10hz)

192 I/O: ±0.05 dB @ +2 dBu, 20 Hz – 20 kHz

[Kiwiburger]

Good luck with that.

Why would you be "asking people here to go look in the manual or on the mfg website " ... are your own hands only painted on?

I expect the reason manufacturers don't publish this stuff because it's completely freakin' irrelevant. Who cares what the specs of the analog section of a digital converter are? You can't use them seperately, so that would be meaningless.

They obviously publish the frequency curve for the whole device - and obviously the analog path is better than that. What more do you need to know?

Would it change anything if you knew?

You question is so non-specific. If you are looking for a good sounding converter, these are plenty of opinions about that.

Somehow, I think you have missed the point about high sample rates ... I could be wrong, but surprise me.

Anyhow - it's very possible to make analog circuits that can handle DC through to radio frequencies if necessary. It's not that hard. But in a digital audio converter, the analog circuit has a job to do - filtering is absolutely necessary.

[Kiwiburger]

API A2D's rock - I have two. And they sound so much better at 96 than at 44.1. If that helps.

[Lawrence]

Originally Posted by Kiwiburger

Why would you be "asking people here to go look in the manual or on the mfg website " ... are your own hands only painted on?

I thought that manuals may have technical info that was not on the website. Or that registered users would have access to information on a mfg web site that casual browers would not.

I expect the reason manufacturers don't publish this stuff because it's completely freakin' irrelevant. Who cares what the specs of the analog section of a digital converter are? You can't use them seperately, so that would be meaningless.

Irrelevant? So if the analog circuit of your converter box only passed frequencies up to say, 20khz, then tell me... exactly what is the point of a 192k converter? 96k I get... gentle slope away from the audible band and all that. 192? Explain that to me. It would be over 4 times the rate of the highest frequency being recorded.

And the point is ???? Sell more HD Accell cards?

They obviously publish the frequency curve for the whole device - and obviously the analog path is better than that. What more do you need to know?

So the 192 I/O 20-20 spec is for "the whole box"? It can only record up to 20k? Hmmm.... really? PTHD owners might find that amusing. Or not.

Again, a published spec up to 20k doesn't tell you what it's doing to (or even if it's passing) anything above that. If it's not (and I don't know since they don't publish them) how do I know if I'm even recording those "amazing frequencies we can't hear" that have such an effect on my sound. Or if I'm just wasting disk space recording emptiness.

Let me see if I got his right...

Some pro engineers say there is energy in the higher frequencies that can have an audible (and/or unknown) effect on what we do hear or feel or whatever (we're talking 192k not 96k), but the frequency response curve of those frequencies is irrelvant to the capturing of them? They could be down -90db/100db (or not be captured at all?) and you'd still somehos get the "inaudible and/or unknown pleasing effect" ?

That would be quite the trick. .

Would it change anything if you knew?

Yes. If my 192k converter's analog circuit before the AD stage sloped off and didn't pass anything over 21k into the actual converter then recording at 192k is (subjective arguments for it aside for moment...) 100% pointless.

You question is so non-specific. If you are looking for a good sounding converter, these are plenty of opinions about that.

True. I could have done a better job. Sorry about that. I apologize.

Somehow, I think you have missed the point about high sample rates ... I could be wrong, but surprise me.

I fully understand and could rationally justify using 96k, even though I don't. It makes perfect sense if you understand converter science and the current market's offerings. 96k should give, for those with good hearing, the best sound... even though there is still a little 'waste' in that rate.

There actually is no logical or technical - as far as I can tell - any reason for recording at 192k. It makes no - logical or technical - sense although people do many things that aren't exactly logical that they like or choose to do.

I have no issue with that. Anybody who likes something better is free to do that, but that doesn't make it (the supposed effect) real.

I do have an issue with a mfg trying to sell me a unit that supposedly captures frequencies I can't hear, but won't publish data that tells me how the unit responds to and captures those frequencies ... if they even capture them at all.

Have you actually passed 30-40-50-60-70-80-96k tones through your converters at 192k and recorded the output to daw at 192k? Can you record a 96k tone? If you can then fine, it's working as advertised even if you can't hear the tone.

At least that (which I plan to do tommorrow at another studio - I don't have a 192k converter) will tell me that those "mysterious things we can't hear that affect the sound we do hear" are at a bare minimum - actually making it to the daw.

If not then.. ????

For now I have no proof of that... since the people who make them don't publish frequency response specs that go that high. We just assume they do because the converter speed is capable of it.

So while a Presonus ADL will pass audio up to 45k (freq responses they publish by the way) wil it ever get to your daw? I don't know ... I'll find out soon.

Anyhow - it's very possible to make analog circuits that can handle DC through to radio frequencies if necessary. It's not that hard. But in a digital audio converter, the analog circuit has a job to do - filtering is absolutely necessary.

Absolutely. I ask again though... if that's the case - and this is pro audio gear - why do the published specs for that stop at 20k? Why design a box that can theoretically record up to 96k (192) when the highest frequency getting to it is - for instance - 25k? What is the rest of the bandwidth for?

Now if someone with a 16x comes back to this thread and tells me it will pass up to 80k to the daw, I'll say fine... if those unknown frequencies that you can't hear that you subjectivey think are the cause for the better sound get through well...

...at least now you know you are actually capturing them. That's a start. I don't know that now... which was the original point....

...because they don't publish the frequency response graphs that would tell me that. I find it very strange that amongst all the arguments for 192k I can't find anyone who actually personally knows (or will say) if these units are even capable of physically handing off audio frequencies that high to their AD converter chips.

If they can't (and perhaps they can, but do you know?) then what exactly is the point of 192k? Or 384k? Or whatever they would sell us next year?

[Lawrence]

Originally Posted by Kiwiburger

API A2D's rock - I have two. And they sound so much better at 96 than at 44.1. If that helps.

I believe you, and I want one. 96k (and it's effect) is real and can be proven to be real, even for those who can't really hear the differences. Even if your converter box filtered everything over 20k.

192k? For now, to me, until I see quantifyable non-subjective proof... is not. It simply makes no sense.

[tINY]

I think you may need to limit your intake of stimulants, Lawry....

If you can be bothered to look up some white papers and product decriptions before picking a fight, you will notice that most converters are actually single-bit converters with decimation filters. The faster you run them, the better frequency performance you get from them. The anti-aliasing filters are mostly handled in the digital section.

So it would be stupid to put a brick-wall analog filter in a modern converter.

The overall performance is a system problem - the digital and analog filters both have an effect on the "specs"

Pay attention for a few more years and you may learn just what "specs" are worth.....





-tINY

[Lawrence]

Nice insults.. 9+ on a 10-point scale.

Interestingly enough you didn't directly address the direct point of the question in the thread.

What is the point of 192k?

Frequency response graphs (whatever they show for any unit) in this context were only mentioned for the curious nature of them all stopping at 20k. Which is because we can only hear to about 20k. Which is why I mentioned analog frequency response. And which is why I wonder what happens to the rest if anything. Which led me to question why they stop there.

I would like to know what happens from 20k up to 96k in a 192k converter. Tell me. +/- 90db from 40-80k? Filtered at 32k? What? As an engineer who might day be pressured by a client to record at 192k I really would like to know.

Which brings me back to...

What is the point of 192k? I know the point of 96k. Does the white paper address 192k and it benefits directly? Please educate me on the reasons why recording at 192k (from your white paper) should sound better or be more beneficial to the great unwashed (like me) than 96k.

Please do.

[Kiwiburger]

Sorry for my aggressive post - but now I understand more about what you are trying to find out.

Personally - I agree; I don't think there is any need to go higher than 96 kHz, and I can't hear any gain for the cost. (That's with a Benchmark DAC-1, testing API A2D - both with can run at 192). My Lucid's only go to 96 kHZ, so I don't have any other gear to compare at 192.

So I will concede that there might be some high end gear out there, and some golden ears that can hear an audible difference.

But why waste the diskspace, when 99.99999% of the target market will never appreciate it.

[Kiwiburger]

Here is a cut & past from the Benchmark site for their ADC-1 (possibly my next converter purchase ...)

Frequency Response at Fs=192 kHz
-3 dB, +0 dB, 2 Hz to 92 kHz
+/- 0.01 dB, 20 Hz to 20 kHz
-0.06 dB at 10 Hz
-0.01 dB at 20 Hz
-0.00 dB at 20 kHz
-0.18 dB at 88 kHz
-3 dB at 92 kHz
-100 dB at 108 kHz


I think you can rest easy now about the maths. But I still don't see the need.

[Kiwiburger]

Although - this reminds me of a local newspaper article the other day, about Hayley Westernra (our young hometown pop diva, protege of George Martin) ...

When she sings the high notes, cats and dogs freak out. Apparantly her voice contains ultrasonic energy.

So I guess if you want the total Hayley Westernra effect, you might need 192 kHz. But i'm not exactly sure the cats and dogs really appreciated this ...

You would need a high class analog chain, including speakers.

I wonder how many people think they are deaf above 15kHz because their headphones simply can't reproduce up there - even if they are sold as 20/20 ...

[Lawrence]

Originally Posted by Kiwiburger

Here is a cut & past from the Benchmark site for their ADC-1 (possibly my next converter purchase ...)

Frequency Response at Fs=192 kHz
-3 dB, +0 dB, 2 Hz to 92 kHz
+/- 0.01 dB, 20 Hz to 20 kHz
-0.06 dB at 10 Hz
-0.01 dB at 20 Hz
-0.00 dB at 20 kHz
-0.18 dB at 88 kHz
-3 dB at 92 kHz
-100 dB at 108 kHz


I think you can rest easy now about the maths. But I still don't see the need.

That's the first I've seen that way. Thanks.

The need would seem to be driven by the existance of the device itself. What I mean by that is that if I as a professional consumer decide that I buy into the "ultrasonic having an effect on the rest" theory ( I don't) and wanted to record at 192k for that specific reason (which frankly is the only reason that, while a longshot, has any remote chance of making any sense at all to me), I would want to have some accuracy in those ranges, not have the response curve be all over the place or dropping off suddenly somewhere well before the capability of the converter rate. I want to know for sure the device makes it *possible*.

Since you can't actually hear that high you can look at the specs and trust that, at the very least, those frequencies (if present in the signal) will remain in a relatively consistent balance with the signals you can hear. Otherwise, the reality of the "ultrasonics" and how they (if they) interact on playback are all skewed.

Normally you use your ears to make those judgements. Obviously you can't do that here so Benchmark has (rightfully) published them for you. If it were down -30db @ 50k and -80db @ 92 it could (if the effect even exists) be a non-factor on playback through wideband speakers and useless for capturing and reproducing those frequencies... and would render the use of 192k for that specific purpose pointless.

This AD along with a wideband mic and pre (Earthworks / Grace?) may be a good starting point for some definitive blind testing of that "ultrasonic" supposition and whether people can actually detect/feel/sense them in the context of listening to music.

I wonder how many people who claim they hear (detect/feel) that effect have actually taken the time to test their systems to see if they are even recording any ultrasonic frequencies or even if their gear is capable of it?

Thanks. At least now I know it is *possible* to test the theory accurately using real world devices and real music. Earthworks -> ADC-1 -> Grace pre -> daw. That is a definitive chain that will get ultrasonic frequencies at 192k to the daw relatively accurately, along with the rest of the music we can hear.

Finding someone to actually do it is another story... the quest continues.

We're gonna need some very good speakers... A.D.A.M.?

[Ravian]

Originally Posted by Kiwiburger

Here is a cut & past from the Benchmark site for their ADC-1 (possibly my next converter purchase ...)

Frequency Response at Fs=192 kHz
-3 dB, +0 dB, 2 Hz to 92 kHz
+/- 0.01 dB, 20 Hz to 20 kHz
-0.06 dB at 10 Hz
-0.01 dB at 20 Hz
-0.00 dB at 20 kHz
-0.18 dB at 88 kHz
-3 dB at 92 kHz
-100 dB at 108 kHz


I think you can rest easy now about the maths. But I still don't see the need.

mmmmmhhhh 192KHz

-100 dB at 108 KHz ????

and what above 108 KHz

so every thing 108KHz and above you cant hear ??????

can some 1 explaine ??????

[Lawrence]

Originally Posted by Ravian

mmmmmhhhh 192KHz

-100 dB at 108 KHz ????

and what above 108 KHz

so every thing 108KHz and above you cant hear ??????

can some 1 explaine ??????

If you're an adult and can hear anything over 17k you're doing pretty darn good.

A 192k converter will only actually record frequencies up to 96k. The 108k slope was designed out there to keep most of the drop-off out of the capture range.

They sloped it off at a point where it has much less of an effect near the end of the range that is capable of being captured... 96k.

Are you Walters?

[Ravian]

Originally Posted by Lawrence

If you're an adult and can hear anything over 17k you're doing pretty darn good.

A 192k converter will only actually record frequencies up to 96k. The 108k slope was designed out there to keep most of the drop-off out of the capture range.

They sloped it off at a point where it has much less of an effect near the end of the range that is capable of being captured... 96k.

i know a bout the adult hearing .
i got 18K on my test.thumbsup

ooooohhh you mean the nyquist freq. Right

[Tim Farrant]

Digital filters have a very steep roll off, hence -3dB @ 92k and -100 @ 108k.

That's a roll off of greater than 100dB per octave -phheeewww!!

Most analog circuits however are more gentle and natural, typically 6dB per octave.

[Jim Kerr]

FWIK 96k and 192k sample rates are just that. They really don't have any thing to do with analog sound except for the fact that the frequency of a sound wave can't exceed one half of it.

Sample rates are the rate at which the samples are computed.

Kiwiburger - You said "Who cares what the analog specs of a digital coverter are?"

Well lets say you have some very very nice converters, part of the reason why they are nice is because of the analog section!?!

Would you want some great performing chip blasting into poor (analog) components just to muddy up your HD rig?

I would say that the analog part is just as important. A chain is only as strong as the weakest link!

[tINY]

I was mistaken. Regardless of the title, this is a thread about recording high frequency content.

I guess I would be more concerned about aliasing than extended bandwidth.

Since processing power is now really cheap and digital filters are built into converters, running higher sample rates on music recording is relatively painless. It's not common because 96k is good enough. One more octave doesn't improve the sound quality enough to worry about. A 96k sample rate is high enough that your AA filters can be pretty gentle and spurious signals have to get up to the 180kHz region before they fold back into the audible range.

When making measurements (especially if there is an FFT involved somewhere) the higher sample rate is nice to have.

As far as your assumption that we can only hear to about 17kHz - there is a question about why a majority of your high-frequency cilia connect to a bundle of nerves that control mussles near your ear cannal instead of sending signals to your brain. Mr Blackmer's theory is that these cause autonomic reasponses that influence how you preceive transients. The implication is that high frequency sounds aren't audible as tones, but do affect your hearing of transients.




-tINY

[Kiwiburger]

Originally Posted by Jim Kerr

FWIK 96k and 192k sample rates are just that. They really don't have any thing to do with analog sound except for the fact that the frequency of a sound wave can't exceed one half of it.

Sample rates are the rate at which the samples are computed.

Kiwiburger - You said "Who cares what the analog specs of a digital coverter are?"

Well lets say you have some very very nice converters, part of the reason why they are nice is because of the analog section!?!

Would you want some great performing chip blasting into poor (analog) components just to muddy up your HD rig?

I would say that the analog part is just as important. A chain is only as strong as the weakest link!

I totally agree. I consider digital audio converters to be primarily an analog device, and need to be treated like an analog device with respect to gain staging.

I wasn't for one second trying to argue that the analog section isn't critically important. I was trying to understand why anyone would want (what I percieved as) boring, arcane, academic knowledge about the analog section only. In other words - the unit must be used as a whole, and if the specs and performance as a whole are excellent - then who cares how it was achieved? Apart from the designer and his competitors.

Rhetorical bullshit, basically.

[Mike Caffrey]

Originally Posted by Lawrence

If you're an adult and can hear anything over 17k you're doing pretty darn good.

Hear or perceive?

The whole premise of the discussion of the effect of hypersonics as based on the premis that there is a difference.

You probably mean perceive.

[Mike Caffrey]

Heh, I thought this was going to be about the sonics of the analog circuitry, forget range, but how true is the analog path? Is it adding 3k or cutting 400hz?


I had a random clocking problem with Radar last week. Initially I had to monitor out of my 192 as a workaround and then a little while later, track through it too. At each change I could hear more tone disappear. I had never expereinced it quite like that. It was pretty shocking.

It would be interesting to hear comparisons of the different analog paths in converters to see how much difference they make.


Maye this has change, but someone once explained to me how you have to have a brickwall low pass filter infront of the converter to prvent it from getting hit with frequencies that it can't handle. Presumably a converter that does a higher sampling rate can handle more. Does that mean there are two separate low pass filters? Or do they change with the sampling rate?

[O.F.F.]

A quick google ('ultrasonic perception') shows that there is plenty of scientific evidence that humans perceive sound of 60-108kHz. The only question is how we do it...
This makes perfect sense to me as it is easy to distiguish between a 16k sine and a 16k square wave were the difference is that the square contains all harmonics at the same level starting at the first harmonic at 32k.
May be thats why cymbals tend to sound so rubbish coming of cds since they radiate most of their acoustic energy well above 20k.
Either way the analogue FR of of high sample rate convertors seems to be very important!
The 20k upper limit was 'established' by Fletcher/Munson in the 1920s and just repeated ad nauseam reinforced by the fact that we actually cannot hear sines above that. On the other hand prior to the introduction of digital audio it was well known that an extended HF response was always preferred in blind tests. Lets not forget that analogue kit easily reaches >35k (reel-to-reels, phono cartridges and cutting heads) in sine sweep tests and can actually record waveforms other then sines at this level which do include even higher harmonics.

[synthoid]

There is no reasonable way to isolate the analog portion of most modern A/D converters. They are fully integrated circuits, most often, with the analog portion being fabricated in the same technology as the digital portion. You could analyze the circuitry outside the A/D converter, but that would not tell you much. Even if the circuit were built out of discrete components, the signal does not flow in a simple way through the analog part into the digital part -- there is a feedback loop between them. So I don't think it would be easy to give a meaning to "the frequency response" of the analog part.

-synthoid

[Mike Derrick]

Originally Posted by O.F.F.

A quick google ('ultrasonic perception') shows that there is plenty of scientific evidence that humans perceive sound of 60-108kHz. The only question is how we do it...
This makes perfect sense to me as it is easy to distiguish between a 16k sine and a 16k square wave were the difference is that the square contains all harmonics at the same level starting at the first harmonic at 32k.
May be thats why cymbals tend to sound so rubbish coming of cds since they radiate most of their acoustic energy well above 20k.
Either way the analogue FR of of high sample rate convertors seems to be very important!
The 20k upper limit was 'established' by Fletcher/Munson in the 1920s and just repeated ad nauseam reinforced by the fact that we actually cannot hear sines above that. On the other hand prior to the introduction of digital audio it was well known that an extended HF response was always preferred in blind tests. Lets not forget that analogue kit easily reaches >35k (reel-to-reels, phono cartridges and cutting heads) in sine sweep tests and can actually record waveforms other then sines at this level which do include even higher harmonics.

OFF, As well as googling "ultrasonic perception," just wondering if you could provide links or references for some of the points you make. thanks.

[Mike Caffrey]

Originally Posted by O.F.F.

A quick google ('ultrasonic perception') shows that there is plenty of scientific evidence that humans perceive sound of 60-108kHz. The only question is how we do it...
This makes perfect sense to me as it is easy to distiguish between a 16k sine and a 16k square wave were the difference is that the square contains all harmonics at the same level starting at the first harmonic at 32k.
May be thats why cymbals tend to sound so rubbish coming of cds since they radiate most of their acoustic energy well above 20k.
Either way the analogue FR of of high sample rate convertors seems to be very important!
The 20k upper limit was 'established' by Fletcher/Munson in the 1920s and just repeated ad nauseam reinforced by the fact that we actually cannot hear sines above that. On the other hand prior to the introduction of digital audio it was well known that an extended HF response was always preferred in blind tests. Lets not forget that analogue kit easily reaches >35k (reel-to-reels, phono cartridges and cutting heads) in sine sweep tests and can actually record waveforms other then sines at this level which do include even higher harmonics.

Semantics, I know, but I think it's the other way around, we can hear frequencies that we can't perceive.

[bcgood]

Originally Posted by O.F.F.

May be thats why cymbals tend to sound so rubbish coming of cds since they radiate most of their acoustic energy well above 20k.

Wow, I thought I was a sound snob. I have to disagree here. I have many cd's with cymbals and they sound great to me, not like rubbish my good chap! Chip Chip!

bcgood

[nosebleedaudio]

If I really wanted to know the specs I would check them my self. This is one of the simpliest test on equipment.
Get a frequency gen that goes to at least 150Khz

[O.F.F.]

For references try Wikipedia and search for 'ultasonic perception'. Its only a stub but they do give you the main scientific papers under 'references'.

As for semantics: Since the research scientists refer to 'perception' rather then 'hearing'
without exception I think I'll stick with that.

Whenever I had the chance to compare an analogue source to a digital one of the same material the cymbals of the analogue source sounded substantially more realistic.