DAC NOS vs DELTA SIGMA, Who is the real winner...

[mocenigo]

Can we train kittens to do YouTube HI-FI reviews. It would be an improvement; and more entertaining.

I am trying. The little guy seems only interested in climbing to the top of my speakers and sleep there.

 

[mocenigo]

That's not an answer to my question : why "Yup, but the reconstruction filter implies oversampling. Thefore NOS if flawed."

And I told you what I meant. I was a bit lacking in details in the first message, I admit.

(However, DIGITAL reconstruction filters are by definition used with oversampling, and if you claim that without a reconstruction filter a system is defective, THEN NOS is defective.)

 

[widemediaphotography]

In a practical sense. Most music is at 44.1khz sampling rate.

— according to some studies the latter could audibly affect transients.

"Could" .... A doubt devoid of all certainty!

 

[Verig]

Hifiman EF400 is completely broken. See measurements, I'm not nitpicking here. EF600 on the other hand is pretty nice with functional form.

 

[mocenigo]

"Could" .... A doubt devoid of all certainty!

That is, properly performed double blind studies reported in peer reviewed papers. Which means, they give a likelihood that the data is statistically significant. There is no need to be zealots.

 

[widemediaphotography]

Claims are never proof, do you have any measurements? Are measurements a religion of fanaticism?

 

[solderdude]

The only DAC where one can find a NOS with a steep (analog) filter are some 1st gen CDP that were not based on Philips tech in those days.
It is practically impossible to create an external NOS DAC with a sharp filter for all different sample rates it must be able to do.
It is possible to create a sample and hold filter for DS DACs which use oversampling but in this case emulate a filterless R2R DAC.

 

[MaxwellsEq]

And I told you what I meant. I was a bit lacking in details in the first message, I admit.

(However, DIGITAL reconstruction filters are by definition used with oversampling, and if you claim that without a reconstruction filter a system is defective, THEN NOS is defective.)

Ahh - I get what you are saying. Although you didn't state this in several posts above, you are apparently exclusively talking about digital reconstruction filters and their application in oversampling DACs.

BUT, to be clear, all DACs, no matter how implemented require reconstruction filters. I worked with PCM systems before CD appeared and they did not do oversampling. They had analogue reconstruction filters - because you must use a reconstruction filter. The earliest CD players came in two flavours: Philips oversampling and everything else (principally Sony) that did not oversample. The latter had brickwall filters. These were like the filters I encountered in non-domestic PCM systems.

It's worth stating that these non-domestic PCM devices were DACs that had to operate only at the sample clock frequency. Similarly the non-oversampling CD Players also only had to to work with one sample frequency: 44.1kHz. These were not like modern standalone DACs that could change their behaviour to accommodate multiple sample frequencies!

BUT, I will state again - if the non-domestic PCM systems and the first CD players had not been fitted with brickwall reconstruction filters they would have been broken.

A low pass filter would therefore have to remove the images at around 22.1khz and the filter would either attenuate frequencies in the audible range or be if such a high order that there would be significant phase rotations — according to some studies the latter could audibly affect transients.

You are correct. Analogue brickwall reconstruction filters are difficult to build, they can have ripple and phase errors in the audio band and incomplete filtering above the stopband. At 44.1kHz these artefacts may be audible.

BUT - these potential artefacts are a minor issue compared to the mess created by a DAC without a reconstruction filter. Such a device would be broken.

 

[jed1]

I suggest to read Nyquist-Shannon sampling theorem for those who would like to dig deeper into this subject covering over-sampling and anti-aliasing filter. (https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem)

 

[mocenigo]

Claims are never proof, do you have any measurements? Are measurements a religion of fanaticism?

You are trying to argue with the wrong person. And you are showing an extraordinary lack of tact, manners, and understanding. I am an objectivist and I am talking about properly conducted double blind tests published in reputable academic journals. I have read the papers years ago and i just know you can find them. Go to the AES site. I am not talking about subjectivist hi-fi magazines that publish what sponsors want.

Positive and statistically significant ABX differentiations are scientific data, with a specific “sigma”. Look at Archimago’s recent DAC test. It is a listening test. And it is scientific.

For instance, measurements are also data. But not the only type of data available.

So, no, I do not consider “measurements a religion of fanaticism” and you completely missed the mark. I consider measurements a fundamental way of evaluating audio equipment.

You, on the other hand, not only are rude, but display the utmost inability to understand both the English language and how science works.

Now, apologise.

 

[mocenigo]

Ahh - I get what you are saying. Although you didn't state this in several posts above, you are apparently exclusively talking about digital reconstruction filters and their application in oversampling DACs.

BUT, to be clear, all DACs, no matter how implemented require reconstruction filters. I worked with PCM systems before CD appeared and they did not do oversampling. They had analogue reconstruction filters - because you must use a reconstruction filter. The earliest CD players came in two flavours: Philips oversampling and everything else (principally Sony) that did not oversample. The latter had brickwall filters. These were like the filters I encountered in non-domestic PCM systems.

It's worth stating that these non-domestic PCM devices were DACs that had to operate only at the sample clock frequency. Similarly the non-oversampling CD Players also only had to to work with one sample frequency: 44.1kHz. These were not like modern standalone DACs that could change their behaviour to accommodate multiple sample frequencies!

BUT, I will state again - if the non-domestic PCM systems and the first CD players had not been fitted with brickwall reconstruction filters they would have been broken.

You are correct. Analogue brickwall reconstruction filters are difficult to build, they can have ripple and phase errors in the audio band and incomplete filtering above the stopband. At 44.1kHz these artefacts may be audible.

BUT - these potential artefacts are a minor issue compared to the mess created by a DAC without a reconstruction filter. Such a device would be broken.

We agree 100%. Not 99%. Not 99.9%. 100%.

(When I mention the analog reconstruction filters, they are explicitly mentioned as such.)

 

[Blumlein 88]

So the votes are in. THE WINNER IS: Delta Sigma by a landslide.


A tutorial on delta sigma here. Which several participants in this thread don't need, but some do.

https://ccrma.stanford.edu/~dattorro/BobAdamsAD1879.pdf

https://www.analog.com/media/en/training-seminars/tutorials/MT-017.pdf

And these are just on the ADC side.

https://www.ti.com/lit/an/slyt423a/slyt423a.pdf

https://www.analog.com/en/resources/technical-articles/sigmadelta-adcs-tutorial.html

 

[ajmxco]

anyone claiming their Grand Seiko is more accurate than some G-Shock Quartz watch is a clown in my book.

You buy them for their novelty or because you appreciate the craftsmanship that goes into them. Not because they're better at telling the time.

100%

 

[mocenigo]

Already responded several times above : Delta Sigma.

The end.

Let us make instead a thought experiment.
The question was regardless of budget.

So, suppose you can manufacture resistors with the highest possible precision.
Then, you parallel SEVERAL (16? 32? 64?) sign-magnitude R2R DACs, or, even better, you actually consider them as differential, you have 2^n positive contributions and 2^n negative contribution - and each R2R ladder has a different offset to balance every imprecision, it is important that 1) the offsets sum to zero and 2) they are randomised with each sample. The latter decreases distortion, and the heavy parallelism decreases noise.

The final result could easily reach or surpass the performance of a good implementation of a delta sigma DAC, with the advantage of a shorter settling time.

Of course the cost would be horrible.

(And I am aware that you can to the same with Delta Sigma DACs, whether discrete or parallel — this was not my point.)

 

[MaxwellsEq]

Let us make instead a thought experiment.
The question was regardless of budget.

So, suppose you can manufacture resistors with the highest possible precision.
Then, you parallel SEVERAL (16? 32? 64?) sign-magnitude R2R DACs, or, even better, you actually consider them as differential, you have 2^n positive contributions and 2^n negative contribution - and each R2R ladder has a different offset to balance every imprecision, it is important that 1) the offsets sum to zero and 2) they are randomised with each sample. The latter decreases distortion, and the heavy parallelism decreases noise.

The final result could easily reach or surpass the performance of a good implementation of a delta sigma DAC, with the advantage of a shorter settling time.

Of course the cost would be horrible.

(And I am aware that you can to the same with Delta Sigma DACs, whether discrete or parallel — this was not my point.)

And operate it at a single fixed temperature.

 

[mocenigo]

And operate it at a single fixed temperature.

It would be a cost no object system, so any expense to keep, for instance, temperature constant everywhere would be fair game.

 

[Blumlein 88]

It would be a cost no object system, so any expense to keep, for instance, temperature constant everywhere would be fair game.

You've already said you could do this to sigma delta designs as well. So where is the advantage? There is none. What part of your approach provides an advantage against sigma delta designs?

 

[mocenigo]

You've already said you could do this to sigma delta designs as well. So where is the advantage? There is none. What part of your approach provides an advantage against sigma delta designs?

The point is that no technology would have an advantage over the other.

 

[majingotan]

Let us make instead a thought experiment.
The question was regardless of budget.

So, suppose you can manufacture resistors with the highest possible precision.
Then, you parallel SEVERAL (16? 32? 64?) sign-magnitude R2R DACs, or, even better, you actually consider them as differential, you have 2^n positive contributions and 2^n negative contribution - and each R2R ladder has a different offset to balance every imprecision, it is important that 1) the offsets sum to zero and 2) they are randomised with each sample. The latter decreases distortion, and the heavy parallelism decreases noise.

The final result could easily reach or surpass the performance of a good implementation of a delta sigma DAC, with the advantage of a shorter settling time.

Of course the cost would be horrible.

(And I am aware that you can to the same with Delta Sigma DACs, whether discrete or parallel — this was not my point.)

Only the snake oil manufacturers would implement such parallel or discrete DS DACs. Reconstruction on far beyond the limits of audibility from DAC with a single chip are already a solved puzzle

 

[mocenigo]

Only the snake oil manufacturers would implement such parallel or discrete DS DACs. Reconstruction on far beyond the limits of audibility from DAC with a single chip are already a solved puzzle

You are missing the point of my thought experiment. Please read it in its context (there are little arrows that allow you to jump to parent messages).

Of course modern delta sigma DAC chips are relatively inexpensive and perfect. Any difference between the complete DACs originates in other components.