• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

MQA creator Bob Stuart answers questions.

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,563
Likes
10,119
Location
North-East
Again, what about this shows us some momentum transfer above 20 khz is going on?
View attachment 28417
What is with the high frequency noise above 60 khz running at about a 17 hz rate?
View attachment 28418

There's definitely content above 20kHz. But audible it's not at -68dB @ 20kHz. It extends to higher frequencies, but at levels below -100dB. I can post a WAV file with the frequencies shifted down, if you want to hear what this sounds like.

1561635100673.png


My hearing threshold is down by about -65dB at 20kHz compared to the high @720Hz (blue curve below). So a total of more than -130dB down from 0dBFS. Somehow I don't think I'll hear this at normal listening levels, but maybe if I crank up the gain by 240dB, like those lab rats, I might actually hear it for a split second. Until my hearing is destroyed.

index.php
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,441
Likes
36,872
There's definitely content above 20kHz. But audible it's not at -68dB @ 20kHz. It extends to higher frequencies, but at levels below -100dB. I can post a WAV file with the frequencies shifted down, if you want to hear what this sounds like.

View attachment 28426

My hearing threshold is down by about -65dB at 20kHz compared to the high @720Hz (blue curve below). So a total of more than -130dB down from 0dBFS. Somehow I don't think I'll hear this at normal listening levels, but maybe if I crank up the gain by 240dB, like those lab rats, I might actually hear it for a split second. Until my hearing is destroyed.

index.php
Yeah, I've looked at that too.

I've also filtered out the 20 khz and below then done the shift to lower frequencies so I can surely hear all of it. There is nothing to hear. If you boost the volume some there is still very little to hear. People keep promising how this stuff matters. There just isn't anything up there much, and we SIMPLY CANNOT HEAR those frequencies. Funny how the idea more is better, and digital was missing something vs analog so we can fix it with hugely extended sample rates when none of the old analog stuff had response equal to basic digital. Myth-making you'll never get rid of I don't suppose.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,407
Likes
4,004
Location
Pacific Northwest
Here is how I see our discussion:
original signal --> ADC --> digital link/media --> DAC --> output signal
I thought we were comparing the original analog signal that enters input ADC connectors vs output analog signal on the output DAC connectors. In that terms original signal is what has been recorded. While I'm aware that AD process involves input lowpass filter, as well as DA process involves it on the output, I don't really see them as part of our discussion about Nyquist theorem.
The filters are an essential part of theory and intrinsically tied to Nyquist theorem. If the original analog signal has frequencies higher than twice your sampling rate, and you don't apply a lowpass filter to remove them before sampling, then you will encode aliased artifacts. Then the reconstructed wave won't match the encoded wave. Theory says reconstruction is perfect only when the original wave was lowpass filtered before sampling.

The relevance is this: Sergei seems to be claiming that removing high frequencies causes audible changes to the wave, even when the frequency threshold is above human hearing. This would violate theory, so the only way that could happen is if the encoding or reconstruction is not done properly. Or, if human hearing somehow has more acuity in the time domain, than in the frequency domain. If that were the case, after decades of research on human hearing I'd expect somebody to have discovered it by now.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,064
Location
Zg, Cro
The filters are an essential part of theory and intrinsically tied to Nyquist theorem. If the original analog signal has frequencies higher than twice your sampling rate, and you don't apply a lowpass filter to remove them before sampling, then you will encode aliased artifacts. Then the reconstructed wave won't match the encoded wave. Theory says reconstruction is perfect only when the original wave was lowpass filtered before sampling.

The relevance is this: Sergei seems to be claiming that removing high frequencies causes audible changes to the wave, even when the frequency threshold is above human hearing. This would violate theory, so the only way that could happen is if the encoding or reconstruction is not done properly. Or, if human hearing somehow has more acuity in the time domain, than in the frequency domain. If that were the case, after decades of research on human hearing I'd expect somebody to have discovered it by now.

Nyquist theorem doesn't mention filters at all - I merely corrected your wording in the context of the theorem, nothing more.

I agree with you, of course, that filters are necessary and I recognise the relevance. I do, however doubt, that anything we say will make Sergei change his mind. :D
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
@Sergei

Which of these two would you advocate that the would-be audiophile should attend to first?
  1. Eliminate phase and timing anomalies in their speakers
  2. Ensure that most of their recordings from now on are 192/24 or MQA
If there is any notion of speaker timing anomalies being 'constant' and therefore nulled out by human hearing, I would dispute it. If a percussionist hits a drum twice in a row, both strikes will be different. Different proportions of the sound will make their way to the woofer versus the tweeter, for example. If there's a physical timing anomaly in the speakers (such as both drivers being mounted on a vertical flat baffle without delay correction), the timing variations thus produced will swamp the supposed inaccuracy of CD by a factor of 'many'.

And yet the MQA people don't mention this in their publicity material...
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,697
Location
Hampshire
Nyquist theorem doesn't mention filters at all - I merely corrected your wording in the context of the theorem, nothing more.
You guys are using the word frequency to denote different things, hence the confusion. Krunok seems to mean the fundamental frequency of an arbitrary periodic signal. In the context of sampling theory, this isn't particularly interesting, which is why the rest of us are talking about the highest frequency component in a Fourier decomposition. Perhaps referring to it as bandwidth would be less confusing. To all but Sergei, that is.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
3,064
Location
Zg, Cro
You guys are using the word frequency to denote different things, hence the confusion. Krunok seems to mean the fundamental frequency of an arbitrary periodic signal. In the context of sampling theory, this isn't particularly interesting, which is why the rest of us are talking about the highest frequency component in a Fourier decomposition. Perhaps referring to it as bandwidth would be less confusing. To all but Sergei, that is.

Well, you are correct, and that settles our differences. But I'm affraid it won't be so easy with Sergei.. :D
 

Sergei

Senior Member
Forum Donor
Joined
Nov 20, 2018
Messages
361
Likes
272
Location
Palo Alto, CA, USA
I believe I isolated the transients. The graph below shows one with amplitude of about 5% of the S192 signal. Going to recheck and publish details over the weekend, if my family doesn't drag me away from computer :) Please stay tuned.

S192 Audio Transient 2019-06-28.png
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,839
Likes
35,808
Location
The Neitherlands
Instead of only showing sample values you should also show the resulting wave form.
And, if you really want to make a point, you should also overlay the exact same down sampled waveforms.
How do you know what the original waveform looked like before it was recorded.
How do you know this is relevant to what you hear. I mean one can record in DXD format and look at the final bit values but can draw no conclusions on what bit levels sound like as the sample points are not the same as the waveform it represents.
Most certainly not R2R as that 'converts' sample points to sample and hold which is very different from the original. Even more so than DSD which at least makes 'smoother' transitions between sample points instead of relying on steep analog post filtering.
 

Sergei

Senior Member
Forum Donor
Joined
Nov 20, 2018
Messages
361
Likes
272
Location
Palo Alto, CA, USA
Instead of only showing sample values you should also show the resulting wave form.
And, if you really want to make a point, you should also overlay the exact same down sampled waveforms.
How do you know what the original waveform looked like before it was recorded.
How do you know this is relevant to what you hear. I mean one can record in DXD format and look at the final bit values but can draw no conclusions on what bit levels sound like as the sample points are not the same as the waveform it represents.
Most certainly not R2R as that 'converts' sample points to sample and hold which is very different from the original. Even more so than DSD which at least makes 'smoother' transitions between sample points instead of relying on steep analog post filtering.

Moved this topic over to https://www.audiosciencereview.com/...apturing-audio-at-higher-sampling-rates.7939/.

You are right, what I attempted to do with Sox in the first iteration may be of historic interest - there are albums made two-three decades ago that involved the Sample And Hold ADCs, and perhaps less accurate first generation Delta-Sigma ADCs - yet my first attempt is not indicative of the last decade and today.

In the v4 comparison, which I propose discussing in the new thread, I made my best time-bound effort to model the real-time downsampling behavior of the actual ADC I was using. The differences are less drastic than that exhibited by the ancient ADCs, yet they are still there. Qualitatively, that's in line with what I remember about our experiments involving 192/24 vs 48/24 captures.
 
Top Bottom