How much difference in sound quality will you hear between qutest and tone board ?

[edechamps]

Microphones are better at this than humans.

I wouldn't be so sure. In particular, humans can perceive side reflections better than a single microphone can because we have two ears and our auditory system is capable of interpreting the delay and level difference between them. Of course, it should be theoretically possible to do the same using two microphones, but it's non-trivial from a signal processing perspective and I don't know anyone who does measurements that way.

This is the main reason why people like @Floyd Toole warn against reading too much in "in-situ" frequency response measurements done with a microphone in a room, especially in medium and high frequencies.

I do agree that for most other kinds of measurements, an appropriate microphone and/or audio analyzer can be way more accurate and sensitive than any human "golden ear" could ever hope to be.

 

[andreasmaaan]

In particular, humans can perceive side reflections better than a single microphone can because we have two ears and our auditory system is capable of interpreting the delay and level difference between them. Of course, it should be theoretically possible to do the same using two microphones, but it's non-trivial from a signal processing perspective and I don't know anyone who does measurements that way.

I thought it was implicit in my comments that I was talking about 2 microphones, a preamp, an ADC and a DSP vs two human ears and a brain, since neither an eardrum nor a microphone on their own can measure anything

 

[March Audio]

Well the video shows in visual form exactly what i said.
In a perfect situation and all parameters known, one can reconstruct(=synthesize) .
However there is no such thing as perfect situation.

No it shows precisely the opposite in crystal clear terms.

Only one waveform fits the sample points in a bandwidth limited system. The exact original waveform.

 

[Frank Dernie]

Some believes no difference some will notice differences.
Some don't care about the differences because of the great specs or just they listen and don't care.
Some have preferences and don't care the specs.
It would be wonderful if the members on this forum that have a lot of dacs could do a/b tests and give the results and to compare them with the specs.
To my mind scientist listeners are as much afraid (if not more) of a/b testing than audiophiles as they want to believe that there is no difference.

Several years ago, probably 2001, I decided it was about time I changed my DAC since it only had SPDIF inputs and the maximum sampling rate it catered for (iirc) was 48kHz.
It was a Goldmund Mimesis 22 bought in 1997ish.
I was quite sure a high end DAC would be much better so the top DAC was a Linn Klimax DS and the least expensive was the DAC section of an ADC/DAC recorder.
In level matched comparison I couldn’t tell any difference on the music I tried. That was when I realised I had been wasting money on high end digital in the past.
I still own the Goldmund and also have a dCS P8i (which sounds different using some of the built in filter choices, but not with the standard one), A RME ADI-2 PRO (in my study) a Resolution Audio Cantata (in my bedroom) and others in storage, so I have plenty of experience of in depth comparisons I didn’t have time for years ago when I believed it mattered.
I absolutely did NOT believe there would be no difference before doing the level matched compare, I was just wondering whether the sound quality improvement justified the difference in price.
It did not, since there was none.
Having written that I bought the Cantata because I adore the styling and the functionality, not for the SQ.

 

[Blumlein 88]

I agree
It only illustrates some aspects of sampling fft noise shaping effect of filtering... I don't know what the video prooves.
You cannot prove a theory is good by using the theory.

Did you pay attention and understand the video? He shows he can put an analog signal from a purely analog source into an ADC, send it to a DAC and the output is a high quality reconstruction of that analog signal. He is not proving the theory using the theory, he is demonstrating via analog means that the theory does work. Those old big HP machines he uses are all analog in their functioning.

The theory says as long as you sample at little more than twice the highest frequency you can reconstruct the signal. That you don't lose anything between samples. Then the demonstration does just that. What more could you ask for?

You keep wanting some 'complex' signal that you don't believe will be recreated. You could get into the details of how that works, but it has been demonstrated how even complex signals are correctly recreated.

What more would it require to convince you this all works?

 

[SIY]

The theory says as long as you sample at little more than twice the highest frequency you can reconstruct the signal. That you don't lose anything between samples. Then the demonstration does just that. What more could you ask for?

It's important to note that the Shannon-Nyquist theorem is a THEOREM, not a theory. Big difference!

If one chooses not to believe it, one can also choose not to believe the Pythagorean Theorem that all them fancy-pants math'meticians use to make us plain folks seem iggerant.

 

[Blumlein 88]

It's important to note that the Shannon-Nyquist theorem is a THEOREM, not a theory. Big difference!

If one chooses not to believe it, one can also choose not to believe the Pythagorean Theorem that all them fancy-pants math'meticians use to make us plain folks seem iggerant.

Yes, that is an important point. Glad you pointed it out.

Then again, are you sure about that old theorem? How does it work if one side of a triangle is 152.665 cm and the other is 69.82169 cm. Not like they are real even numbers every time. I mean sure it works with 3-4-5, but that is hardly the same is it?

 

[Frank Dernie]

It's important to note that the Shannon-Nyquist theorem is a THEOREM, not a theory. Big difference!

If one chooses not to believe it, one can also choose not to believe the Pythagorean Theorem that all them fancy-pants math'meticians use to make us plain folks seem iggerant.

Beat me to it! I was about to point out the same, The theory is already proven.

 

[Frank Dernie]

Yes, that is an important point. Glad you pointed it out.

Then again, are you sure about that old theorem? How does it work if one side of a triangle is 152.665 cm and the other is 69.82169 cm. Not like they are real even numbers every time. I mean sure it works with 3-4-5, but that is hardly the same is it?

Ha-ha

 

[SIY]

Then again, are you sure about that old theorem?

I mean, like, that Pythagorean stuff is like a hundred years old and it was done by, like, some dead white guy. So, like, it's just your opinion, man.

 

[Calexico]

Did you pay attention and understand the video? He shows he can put an analog signal from a purely analog source into an ADC, send it to a DAC and the output is a high quality reconstruction of that analog signal. He is not proving the theory using the theory, he is demonstrating via analog means that the theory does work. Those old big HP machines he uses are all analog in their functioning.

The theory says as long as you sample at little more than twice the highest frequency you can reconstruct the signal. That you don't lose anything between samples. Then the demonstration does just that. What more could you ask for?

You keep wanting some 'complex' signal that you don't believe will be recreated. You could get into the details of how that works, but it has been demonstrated how even complex signals are correctly recreated.

What more would it require to convince you this all works?

But he doesn't send complex signals. We all now the theory is correct for pure sine and that square wave show some ringing and that noise shaping lower the noose floor.
But he doesn't talk about real life signals and that's what people ask i think.
All theorists talk about theory but seems afraid to experiment with more real life signals to see the limit of approximations effect on actual dacs. although it could convince people.
Theory talk about infinite steps of calculation and no dacs have it. There is always rounded calculation on the practice. (Because limited cpu power of dac).
The video don't talk about it.
When we see the square renderered he said it's perfect without arguing on the ringing.

What i found interesting is that the digital filters ringing didn't cumulate when converting multiple times.

Goal is to see the limits of dacs and how it evolved with new tech no?

 

[Blumlein 88]

But he doesn't send complex signals. We all now the theory is correct for pure sine and that square wave show some ringing and that noise shaping lower the noose floor.
But he doesn't talk about real life signals and that's what people ask i think.
All theorists talk about theory but seems afraid to experiment with more real life signals to see the limit of approximations effect on actual dacs. although it could convince people.
Theory talk about infinite steps of calculation and no dacs have it. There is always rounded calculation on the practice. (Because limited cpu power of dac).
The video don't talk about it.
When we see the square renderered he said it's perfect without arguing on the ringing.

What i found interesting is that the digital filters ringing didn't cumulate when converting multiple times.

Let us start with your last statement. Ringing filters. He said they showed, exactly what they should show. A square wave can only be reproduced with infinite bandwidth. If you begin limiting bandwidth, you get what you called ringing. It really isn't ringing. And there is a clue. If it were ringing, each pass would add more and more ringing. However, it isn't ringing which is why once you band limit a signal once, passing thru the same band limiting doesn't change the shape of the band limited square wave.

This animation might make it clear. It also might help you understand how any signal no matter how complex can be broken down into equivalent combinations of sine waves.

Or maybe the animation of at the bottom right of this will help.
https://en.wikipedia.org/wiki/Square_wave

A square wave is a fundamental and all of its odd harmonics. If you limit the bandwidth the upper level harmonics get filtered out and reduce the squareness of it. Of course you can't hear infinitely high frequencies so you'd not hear a difference. So what looks like ringing is really just limited bandwidth of a squarewave.

This page explains it a little more and shows adding more and more harmonics.
https://www.mathworks.com/help/matlab/examples/square-wave-from-sine-waves.html

The theorem works for complex signals too. And yes people have investigated those. There is no fear of them. That it works can surely be demonstrated.

 

[Calexico]

Beat me to it! I was about to point out the same, The theory is already proven.

Theory is proven for what?
It's also proven that when more steps are calculated more precise is the reconstruction filters. No dacs have infinite steps of calculation so they don't fullfeel the theory.
Also having a mix of multi bit and delta sigma in dacs doesn't respect the theory. There are some math tricks that are not purely shanon theory.

Don't forget that to respect purely the cd format a dac should be able to deal 16 true bits at the frequency of 44.1khz.
No r2r and no delta sigma dac do it perfectly.

The most linear r2r dacs that have the higher rate oversampling filters are the most close to the pure cd decoding theory.

Delta sigma dacs truely respects the theory only with dsd format.
Converting pcm to dsd in dacs is an approximation and cannot be considered as perfect in the theory.

So theory is not fullfilled by dacs that's why we asks questions.

 

[Calexico]

@Blumlein 88
To me a dac is lot of math tricks and not only purely respected theory.
The math tricks are developped to measure good on classics test.
But was cannot be sure that these math tricks have no effect on others parameters.
Theory proves something but as dac do lot of math tricks that are not taken into account in the theory we cannot always say they will act as the theory suggests.
For dsd i agree theory is always true as there's no math trick and decoding theory is respected.
For multibit delta sigma it's more complex and we should evaluate the quality and what we loose in the pcm to dsd multibit convertion.

 

[Blumlein 88]

Theory is proven for what?
It's also proven that when more steps are calculated more precise is the reconstruction filters. No dacs have infinite steps of calculation so they don't fullfeel the theory.
Also having a mix of multi bit and delta sigma in dacs doesn't respect the theory. There are some math tricks that are not purely shanon theory.

Don't forget that to respect purely the cd format a dac should be able to deal 16 true bits at the frequency of 44.1khz.
No r2r and no delta sigma dac do it perfectly.

The most linear r2r dacs that have the higher rate oversampling filters are the most close to the pure cd decoding theory.

Delta sigma dacs truely respects the theory only with dsd format.
Converting pcm to dsd in dacs is an approximation and cannot be considered as perfect in the theory.

So theory is not fullfilled by dacs that's why we asks questions.

https://www.audiosciencereview.com/...on-dac-linearity-measurement.3754/post-158703

Here is a post where I measured an inexpensive DAC. You say no delta sigma can do 16 bits. Yet I don't even bother much with 16 bits as it is so easy to get that at the perfect level. This post above shows this DAC getting all the way down to the 24th bit at the correct levels. In this case 48 khz, but it works at 44.1 khz just as well. So regardless of how it does it, this DAC fed a PCM signal can correctly output the values all the way down to the 23 or 24th bit. BTW, the DAC in this measurement used a Khadas Tone board.

 

[Calexico]

https://www.audiosciencereview.com/...on-dac-linearity-measurement.3754/post-158703

Here is a post where I measured an inexpensive DAC. You say no delta sigma can do 16 bits. Yet I don't even bother much with 16 bits as it is so easy to get that at the perfect level. This post above shows this DAC getting all the way down to the 24th bit at the correct levels. In this case 48 khz, but it works at 44.1 khz just as well. So regardless of how it does it, this DAC fed a PCM signal can correctly output the values all the way down to the 23 or 24th bit.

Ok but it's not correct to say theory is always applicable whatever the level or whatever the frequency of dac because dacs make tricks on the theory.

 

[SIY]

To me a dac is lot of math tricks and not only purely respected theory. That is a problem with you rather than with the math.
The math tricks are developped to measure good on classics test. This is nonsense.
But was cannot be sure that these math tricks have no effect on others parameters. This is even more arrant nonsense.
Theory proves something but as dac do lot of math tricks that are not taken into account in the theory we cannot always say they will act as the theory suggests. Except for the inconvenient fact that this is exactly what they do.
For dsd i agree theory is always true as there's no math trick and decoding theory is respected.

I would again urge you to demand a refund from whatever school failed to teach you basic science and engineering.

 

[Calexico]

I would again urge you to demand a refund from whatever school failed to teach you basic science and engineering.

I would again urge you to demand a refund from whatever school failed to teach you basic science and engineering.

The theory don't take into account the pcm to multibit delta sigma convertion of most dacs.
As each brand has it's method of converting the pcm to multibit delta sigma you cannot use the same theory if they don't calculate the convertion the same way. As these convertions are made to measure good you think it's 100÷ fidel to the pure theory. You don't know what's behind the hood.
I agree that it should be more and more close to the theory.

 

[Blumlein 88]

Ok but it's not correct to say theory is always applicable whatever the level or whatever the frequency of dac because dacs make tricks on the theory.

How do you know a theory works? You test it to see if things act as the theory says. The testing has been done.

A theorem, do you realize what a theorem is, and how it differs from a theory?
a general proposition not self-evident but proved by a chain of reasoning; a truth established by means of accepted truths. In math such as the Shannon-Nyquist sampling theorem, it has all been shown to work based upon established prior mathematics. The same is true for how oversampling works. It isn't a math trick that only works with test signals. It is math that works period.

The hardware has been built upon these theorems and tested to prove the results are what you'd expect to within a very small margin of error. How small well some fraction of a db at 1/8th of a microvolt in the test I linked to above. And this from an affordable consumer DAC. Again what would convince you?

 

[Calexico]

How do you know a theory works? You test it to see if things act as the theory says. The testing has been done.

A theorem, do you realize what a theorem is, and how it differs from a theory?
a general proposition not self-evident but proved by a chain of reasoning; a truth established by means of accepted truths. In math such as the Shannon-Nyquist sampling theorem, it has all been shown to work based upon established prior mathematics. The same is true for how oversampling works. It isn't a math trick that only works with test signals. It is math that works period.

The hardware has been built upon these theorems and tested to prove the results are what you'd expect to within a very small margin of error. How small well some fraction of a db at 1/8th of a microvolt in the test I linked to above. And this from an affordable consumer DAC. Again what would convince you?

Yes you prove that theory is met for one frequency.
As dacs are made of compromise to approximate the theory we should consider testing more exhaustively.
For exemple what happens for ds multibit when the level is lower than the multibit part of the dac and is in the delta sigma part.
Is it the same with all frequencies,?
It's proven that dacs reacts differently under a certain level transition below the multibit ability of the dac.
We should find tests that show the difficulties of dac at the limit of their approximations and then see if it's important or not
What is sad is we only see tests that dac pass easily. That's why i ask for tests to see limits of the maths used in dac
Testing difficulties is more revealing than testing facilities no?