breakthrough? (correcting phase error in DAC)

[Jazzhead]

sounds like it may be like the Zanden phase switch...I don't think so. that may relate to recordings made with inverted phase. I think this phase discussion in this thread has to do with phase integrity through the d/a process...I think some digital filters apparently are not phase coherent or something...I am no techie. I know Zanden was super super focused on phase coherence throughout the D/A process and that was one of his chief goals. I think Mike Stahl shares many similar design philosophies with Yamada San...many Audio Exotic fans were long-time Zanden digiphiles...though his work is designed differently/ Nevertheless, mike also makes a point about talking about phase coherence.

Thanks for pointing it out , will keep you posted on how it fares .

 

[Ken Newton]

from soulution's facebook page:

Alan Taffel highlights soulution's new DAC technology:

...Based on what I heard, the Soulution phase shift cancellation circuit could be a watershed development in digital sound evolution...

http://www.theabsolutesound.com/articles/ces-2014-digital-components/

In the linked story, Taffel matter-of-factly states that a phase shift is endemic to the D/A process. I don't know what endemic phase shift he is talking about. The only phase shift I might imagine would be linear (so, not a phase distortion), and during the A/D conversion process. The fact is, that the original band-limited waveform can be sampled, recorded, and then accurately reconstructed using a linear-phase brickwall digital filter. Which also means, without phase distortion.

While I may be missing something here, this story strikes me more as a manufactured problem needing a manufactured solution, with some cheerleading by the industry press, than some true new innovation. However, I'm not an expert on data conversion, so, as I said, I may be missing something. Perhaps, the possible linear phase shift during A/D sampling which I mentioned earlier could be more of problem than it would seem. I'll have to think through this a bit more deeply.

 

[cat6man]

i'm not an expert in the conversion process either, but basic signal processing theory and nyquist say that the signal can be perfectly reconstructed, assuming perfect conversion/sampling and perfect reconstruction. there may be many places where the real world rears its ugly head and introduces degradations and deviations form ideal.

what i would like to see is the change to the impulse response with this filter.
there are an infinite number of filters with a specific amplitude spectrum but different phase characteristics.
does this new 'phase correction filter' make the filter closer to the Nyquist ideal?
a filter meets the nyquist crteria in the amplitude domain but not in the phase domain is not a nyquist filter.

an example of phase correction is the use, over the last few years, of reconstruction filters with suppressed pre-ringing (at the expense of larger post-sample ringing) in the impulse response. a rough consensus seems to exist that pre-ringing is 'unnatural' sounding, though both can be 'nyquist'.

all of this is complex, most importantly (imho) due to jitter. the amount of ringing and slope of the impulse response at the zero crossing determines jitter sensitivity. the more ringing, the larger the amplitude errors due to jitter (a given timing error causes a larger amplitude error since the zero crossing slope is larger/higher). that is one reason why a single jitter spec (X ps rms) is not the entire story as the jitter interacts in complex ways with the reconstruction filter.

i believe one of the reasons that the original Wadia DACs were so enjoyable (I had a 27ix until very recently) was the relative absence of ringing in the impulse response. i remember seeing photos of the impulse response in the first magazine reviews that showed how little pre- and post- sample ringing their filter (spline interpolation if my memory isn't faulty?). in graduate school, a professor told us a story about how a company came to him with a problem that turned out to be an attempt to build a nyquist reconstruction filter with a brick wall at the nyquist frequency............the filter of course rings forever..............simply using a gentler nyquist filter (i.e. accepting some amplitude rolloff before the nyquist frequency) easily solved the problem..............and of course, if the sampling rate for CD had been set higher in the first place, these tradeoffs would not have been needed and sensitivity to jitter would have been lessened from the start (the two are related, which is again my main point here).

lastly, let me also bring up one more related approach. i'm working from memory so i may have some of the details wrong, but i believe that emm labs is doing something different as they claim an impulse response with no ringing. now that is clearly not a nyquist reconstruction filter but the dac2x sounds pretty good to my ears..............so what are they doing with their proprietary (forget the name) techique?
i think it may be the following:
all of the discussions around nyquist reconstruction filters has assumed that the reconstruction filter is both linear and time invariant.
there is nothing in theory or practice that says a non-linear or time-variant filter cannot sound better!.............so for example, imagine a group of filters are available and the DAC chooses to use a specific one based on either the amplitude (different approaches for high and low volume) or based on rate of change (high versus low versus mixed frequency content). these are all possible tweaks to the linear time invariant filter approach and while i know of no theory to suggest a specific nonlinear or time-varying filter is best, there certainly would seem to be room for innovation. to suppose another possible path, what if wadia's proprietary spline interpolation filter were extended to a family of spline interpolation filters and the best one was selected based on some intelligent (i have no ideas here, this is theoretical/hypothetical) criteria and dynamically changed over time as the music changed.

 

[Orb]

If they have implemented a minimum phase filter then there will be some group delay distortion/phase shift (although this should be at higher FR).
So might be some way implement benefit of minimum phase filter without the disadvantages *shrug*.
Or maybe they are talking about something totally different

Cheers
Orb

 

[Whatmore]

is phase error the new jitter?

 

[LL21]

This is the little bits I could pick up from Zanden's views on phase shifts from the filters typically used in D/A work:

"6 Moons 2003. Explains Yamada-San that standard digital low-pass filters start to phase shift at 1kHz and exceed 30-degree rotation at 20kHz, contributing to the somewhat harsh, brittle high end of digital when compared to analogue. His 'bridged-T fixed-impedance filter' is used predominantly in deep-sea telephone cables but here uniquely implemented for audio use."

"...DVD-A and SACD. According to this Japanese engineer, the latter suffers "very significant" high-frequency phase shift problems intrinsic to the format per se..."

"...I now believe that crimes pinned on resolution are really sins committed in the high frequencies. Take an adjustable time-aligned speaker like a Green Mountain Audio. Set it up deliberately wrong such that the tweeter precedes the output of the midrange. Sibilants will be emphasized, transients crispified - hello digital etch and annoyance. Even apparently insignificant timing errors in the treble cause undue sharpness and etchiness. This is easily demonstrated by moving a tweeter module forward or back by a quarter inch when the filter network is of the minimum phase kind. What do we call timing errors in audio? Phase shift. Could the bad we conventionally accuse the hyper in resolution of simply be due to treble phase errors?..."

 

[JackD201]

Unfortunately, Phase Linearity is a spec not often published. Not that it can tell the whole story but every little bit of info can help.

 

[LL21]

Unfortunately, Phase Linearity is a spec not often published. Not that it can tell the whole story but every little bit of info can help.

If I can find any more snipits, I will post them here.

 

[JackD201]

I tried reading some of the research papers (audiology). Most of it went over my head. To me right now, it is an abstract concept. I wouldn't know what to listen for much less identify its effects.

 

[LL21]

I tried reading some of the research papers (audiology). Most of it went over my head. To me right now, it is an abstract concept. I wouldn't know what to listen for much less identify its effects.

The guys who interviewed Yamada San of Zanden seemed to suggest it was treble-related shrill, hardness. I cannot say either. I just know I like what he does in his designs.

 

[JackD201]

As do I

 

[LL21]

As do I

Yes, that's right. I remember we spoke off-line about that...thanks again for your advice way back then. I am extremely happy.

 

[JackD201]

Big thumbs up Lloyd

 

[Ken Newton]

...Explains Yamada-San that standard digital low-pass filters start to phase shift at 1kHz and exceed 30-degree rotation at 20kHz, contributing to the somewhat harsh, brittle high end of digital when compared to analogue.

This statement is misleading, if unintentionally so. A linear-phase digital filter is the only sharp slope filter (digital or analog, of which I'm aware) that does not distort phase. Since Bridged-T filters are typically used for delay compensation, I suspect the phase shift Zanden is concerned with is linear. Which may be the same linear phase-shift Souloution is touting a correction for. Interestingly, if you look up Anthony (Tony) Taddeo's patent relating to improving the sound of DACs, you'll find mention of his conclusion that there is being a phase lead error inherent to A/D conversion.

 

[LL21]

This statement is misleading, if unintentionally so. A linear-phase digital filter is the only sharp slope filter (digital or analog, of which I'm aware) that does not distort phase. Since Bridged-T filters are typically used for delay compensation, I suspect the phase shift Zanden is concerned with is linear.

Hi Ken...thank you. I have always said I am no techie...so just trying to be helpful by adding what little I could find out about phase-related comments and digital.

 

[Ken Newton]

Please, no worries, LL21. My observation was directed toward Yamada-san's statement.

 

[LL21]

Please, no worries, LL21. My observation was directed toward Yamada-san's statement.

Thank you...I really enjoy 'learning' a bit more each day about digital...I have spent countless hours listening to different digital sources, and remain perfectly happy with the Zanden, but I have also found great sound from Vivaldi and Metronome Kalista Ref/C2A.

So I look forward to learning more about Soulution's phase adjustment.

 

[esldude]

This statement is misleading, if unintentionally so. A linear-phase digital filter is the only sharp slope filter (digital or analog, of which I'm aware) that does not distort phase. Since Bridged-T filters are typically used for delay compensation, I suspect the phase shift Zanden is concerned with is linear. Which may be the same linear phase-shift Souloution is touting a correction for. Interestingly, if you look up Anthony (Tony) Taddeo's patent relating to improving the sound of DACs, you'll find mention of his conclusion that there is being a phase lead error inherent to A/D conversion.

I don't think Tony Taddeo actually said there was a phase lead. Merely that his fix was a lagging phase signal mixed with the original. If of equal level what it did was create a steep notch filter at half the sample rate. He mentioned it shelves high frequencies a bit, -1 db at 10 khz and 2.5 db at 20 khz. What this did was reduce the "ringing" on impulse and square waves and presumably any transients in the recording. Of course listening one won't be able to separate that from the audible change in frequency response.

If you would like to experience it yourself it is quite simple actually. Take a music file and duplicate it. Chop off the first bit of the duplicate file. Reduce both files by -6db, and mix them together. You now will have reduced transient ringing at least as they visually appear in the resulting waveform with a notch filter at half the sample rate (22.05 khz for 44.1). It is a little bit of smoke and mirrors and a little bit of truth. Easy to try and see if you like the resulting sound.

An easier way to get most of the effect is to simply create a digital filter notch with the appropriate shape. Apply it to file and listen. Not quite exactly the same as the Digital Antidote, but very similar result.

 

[LL21]

This statement is misleading, if unintentionally so. A linear-phase digital filter is the only sharp slope filter (digital or analog, of which I'm aware) that does not distort phase. Since Bridged-T filters are typically used for delay compensation, I suspect the phase shift Zanden is concerned with is linear. Which may be the same linear phase-shift Souloution is touting a correction for. Interestingly, if you look up Anthony (Tony) Taddeo's patent relating to improving the sound of DACs, you'll find mention of his conclusion that there is being a phase lead error inherent to A/D conversion.

Thanks for continuing to update your posts, Ken. Thank you. I will say, as a non-techie, I am pleased to see that Yamada San was identifying and seeking to address digital issues as far back as 2004-2007 that continue to be a focus of much later digital playback technology from serious firms like Soulution.

 

[cat6man]

I don't think Tony Taddeo actually said there was a phase lead. Merely that his fix was a lagging phase signal mixed with the original. If of equal level what it did was create a steep notch filter at half the sample rate. He mentioned it shelves high frequencies a bit, -1 db at 10 khz and 2.5 db at 20 khz. What this did was reduce the "ringing" on impulse and square waves and presumably any transients in the recording. Of course listening one won't be able to separate that from the audible change in frequency response.

If you would like to experience it yourself it is quite simple actually. Take a music file and duplicate it. Chop off the first bit of the duplicate file. Reduce both files by -6db, and mix them together. You now will have reduced transient ringing at least as they visually appear in the resulting waveform with a notch filter at half the sample rate (22.05 khz for 44.1). It is a little bit of smoke and mirrors and a little bit of truth. Easy to try and see if you like the resulting sound.

An easier way to get most of the effect is to simply create a digital filter notch with the appropriate shape. Apply it to file and listen. Not quite exactly the same as the Digital Antidote, but very similar result.

a filter with a notch at 1/2 the sample rate..............now where have i heard that before?
this sounds like another implementation of the miraculous apodizing filter, if i'm not mistaken.