breakthrough? (correcting phase error in DAC)

[dallasjustice]

Does Soulution even design their own digital gear? I thought they outsourced all digital to another firm in die Schweiz?

Michael.

 

[Ken Newton]

I don't know.

 

[Anatta]

http://www.head-fi.org/t/702787/chord-hugo/1875#post_10460853

On the WTA filter, yes the WTA algorithm does give roughly the same sound as a conventional half band filter that has ten times the tap length. But I have always said that increasing tap size improves performance,even with using the WTA, and I am certain that 26k is not the limit. You can hear big changes from 16k to 18k; and there was a big change from 18k to 26k. Now there has to be a limit to when the tap size will no longer improve the sound, and if I were to put a number on it (and this is entirely a guess) I would say 1M taps for an 8 times OS filter. That would guarantee accuracy to better than 16 bits, as the sinx/x coefficients are now well below the 16 bit level. But this is a guess. It could be 100K, it might be 10M. Nobody knows. I will find out - that's what makes this so interesting at the moment.

http://www.head-fi.org/t/702787/chord-hugo/1905#post_10462953

1. PCM Pro: excellent resolution of small signals, very small signals do not disappear into the dithered noise floor.
Cons: Timing. Ear/brain can resolve 4uS, CD innately is at 22uS.

2. DSD Pro: Samples at 0.34 uS, albeit at not very good resolution, so has much better timing innately
Cons: resolution. Noise shaper noise is not the same as dithered noise, any signal below noise shaper noise floor is lost.


Now the timing issue can be resolved by the DAC interpolation filter, and with an infinite tap length filter, timing is completely reconstructed. So red-book is capable of very much better performance, if you improve the interpolation filter. But with DSD, the encoding means that low-level details are lost in the noise shaper noise floor, and they are lost forever. So DSD has a compression in depth and instrument separation, due to poor resolution, but does not innately have timing problems. Check out 2L website, and compare the DXD recording to the DSD64 or DSD128 - to my ears, the loss in transparency of DSD is not small.

Getting back to design of the DAC. Now I run my DAC's with a very simple single stage active analogue section, with only 2 caps and 2 resistors in the direct signal path - and I do this for transparency. But this means the digital RF noise in the 100k to 1M band must be very low, so the digital source must be filtered - and DSD is at -20dBFS at 100kHz. So I can't put raw DSD into the DAC, or it will sound very hard. So the DSD is filtered, which converts it to regular PCM.

If I were doing a DAC for only DSD would I do it this way? Yes, I think I would, as simple analogue is always the best.

Incidentally, the DSD filters on Hugo has been improved - they are much smoother than with Qute.

 

[cat6man]

Soulution appear to have an excellent industry reputation, but I can't help feeling that if this is all there is to their zero-phase technology, it is as you suggest above. Adding to the hype, is the press report by Alan Taffel (see thread post #15) where, after a demonstration of the technology, he wrote the following: "Based on what I heard, the Soulution phase shift cancellation circuit could be a watershed development in digital sound evolution..."

Perhaps, there is more to the technology than is indicated in the Soulution press release?

For a reconstruction filter with no ISI (inter-symbol interference), Nyquist also derived conditions that the reconstruction must meet.
Over-simplifying a bit, it requires a symmetry around the nyquist frequency so that in the frequency domain, the sum of shifted versions of the filter add to a constant value. This requirement applies both to the amplitude and to the phase.

Does the phase of soulution's solution create a better nyquist reconstruction filter? I don't know.
The phase can certainly impact the zero crossings of the impulse response and an amplitude-centric design for a nyquist filter may not capture important phase information that can distort the impulse response.

We have clearly seen how minimum phase filters (less ringing before main sample) sound different than filters that result in symmetric impulse responses (equal ringing before and after). It seems feasible to me that phase optimization has not received the same level of attention as the amplitude response.

It would be very cool if this turns out to be real (like jitter and minimum phase filtering).

 

[Ken Newton]

Except, that Soulution do not address the Nyquist reconstruction filter in describing their phase correction technology. They clearly state that they are correcting the phase shift due to an analog 3rd order Bessel filter having a turnover frequency of 120KHz. While this analog filter technically contributes to the total reconstruction process when used in conjunction with a digital interpolation Nyquist reconstruction filter, it mostly serves to prevent high ultrasonic and RF content from being sent to the next amplification stage.

Compared to the more severe phase shift produced by most loudspeakers, or the phase shift produced by a minimum phase digital interpolation filter (if one is utilized), or the phase shift produced by analog vinyl mastering and playback, or the phase shift produced by many transformer coupled tube amplifiers, correcting the 15 degrees of in-band phase shift stemming from a DAC's analog output clean-up filter would seem highly unlikely to be the key to unlocking significantly improved subjective playback quality.

 

[cat6man]

I agree with all of your points, including distortions due to other causes.
Perhaps they are doing more than the bessel correction?

In any case, I was referring more to the potential of optimizing an impulse response in the phase domain.
If we accept that more taps (lots and lots, 10k-100k or more) can improve the sound, I would have no difficulty accepting that phase correction of small errors could be equally important. The 100,000th tap will have a pretty small coefficient and a correspondingly small impact on phase, so if the former is a feasible/audible source of degradation, I'd guess the latter is also.

 

[Werner]

the latest absolute sound mentions a new Soulution DAC with 'phase correction' that is the new BREAKTHROUGH paradigm for digital playback.
supposedly, on/off tests for the soulution DAC with and without 'phase correction' is "insert your favorite term for &*! unbelievable here".

I read the whole thread. I understand exactly what Soulution are doing.

It is exactly the same as what Philips did in their first generation of true 16-bit CD-players in 1985.

Exactly.