breakthrough? (correcting phase error in DAC)

[Al M.]

Those of you who have not acquired this new "breakthrough...." Are your notes smearing? Really?

Tim

No, they aren't.

I have become skeptical about "technical breakthroughs" since the following experience:

Yes, I have had an eye-opening experience in this regard when, courtesy of Goodwin's High End, I was able to compare at home the highly acclaimed NAD M51 DAC to my Wadia 12 DAC. I sonically preferred the 20 year-old (!) Wadia 12 to the NAD DAC, except in the bass performance (well, the Wadia 12 had been upgraded in 1997 with the Wadia 860 opamp, but that was also 17 years ago). So much for the hype of ever higher performance of digital over the years. Don't get me wrong, the NAD is a great DAC when judged on its own, but folks, let's be real. Paul at Goodwin's High End wasn't surprised about the findings.

I ended up buying the Berkeley Alpha DAC 2. Now that one did live up to the hype of ever higher performance of digital over the years ... It was simply in a different league than both the NAD M51 and the Wadia 12, with a stunning price/performance ratio. Paul was like, I told you so... (yes, it is 2.5 x more expensive than the NAD M51, but in my view ridiculously cheap relative to its performance).

 

[Orb]

Here's a link to a Chord slide presentation on DAC design issues and technology, including their so-called Watts-Transient-Aligned (WTA) digital filter. However, it doesn't provide enough technical details to determine exactly how the WTA filter works.

http://www.chordelectronics.co.uk/files/Hugo DAC Technical Master Class (1).pptx

Thanks for the link.
I must admit I am a fan of Chord Electronic's engineering (and that also of Rob Watts).
Got to love how an impulse analysis just gives the classic sin(x)/x FR droop/roll-off for their DACs

Cheers
Orb

 

[esldude]

Here's a link to a Chord slide presentation on DAC design issues and technology, including their so-called Watts-Transient-Aligned (WTA) digital filter. However, it doesn't provide enough technical details to determine exactly how the WTA filter works.

http://www.chordelectronics.co.uk/files/Hugo DAC Technical Master Class (1).pptx

Thank you for the link. Having viewed it I have filled my misinformation quota for the rest of the week. I don't have worry about it.

 

[wisnon]

More from Chord:

The Qute HD "white Paper"
http://www.chordelectronics.co.uk/specs/QuteHD tech specs for web.pdf

Their DAC tech explained:
http://www.chordelectronics.co.uk/chord-dac-technology.asp

 

[Orb]

Thank you for the link. Having viewed it I have filled my misinformation quota for the rest of the week. I don't have worry about it.

Bear in mind Chord Electronics has been using FPGAs/DSP core and large tap lengths/gate count since late 1990s for their DAC/noise shaping functions - one of the first to do so it must be said.

Cheers
Orb

 

[Orb]

I better include this technical paper as I should had emphasised the Chord Electronics DACs are an evolution of what was done by Rob Watts with his joint owned company Deltec/DPA.
Things have moved on a fair bit from this to how implemented by Chord Electronics products:
http://www.fuse-network.com/fuse/demonstration/323/23290/23290.pdf

Enjoy reading but I will not get into the technical aspect myself, after the other thread discussing PCM.
Cheers
Orb

 

[esldude]

I better include this technical paper as I should had emphasised the Chord Electronics DACs are an evolution of what was done by Rob Watts with his joint owned company Deltec/DPA.
Things have moved on a fair bit from this to how implemented by Chord Electronics products:
http://www.fuse-network.com/fuse/demonstration/323/23290/23290.pdf

Enjoy reading but I will not get into the technical aspect myself, after the other thread discussing PCM.
Cheers
Orb

The link to the powerpoint was largely built around the following misinformation. Ears hear interaural timing differences of 4 microseconds implying the ear samples some aspects at 250 khz equivalent sample rate, and we don't know much about nerve transfer to the brain where real processing is done. The point about 4 microseconds is probably not quite true though it might be. The rest is not so. You don't need 250 khz sampling to time 4 microseconds interchannel timing shifts. Even lowly 44.1khz/16bit will do the job into the picosecond range. 10,000 to 100,000 times finer than needed to satisfy the ear. There also is some good knowledge of nerve transfer rates to the brain from the ear, and they are not at the level implied. Since all the rest about filtering, timing etc. being needed and beneficial is based upon this false premise the entire piece becomes misinformation. Chords filtering would appear to be extraordinarily good. But if it sounds better the reason isn't the one put forth in the powerpoint.

You don't need to debate that. The timing accuracy of sampled audio is measurably demonstrably far in excess of what is needed. This is one of those oft repeated pieces of deceptive ad copy that crops up over and over again. The Chord equipment may be great, but it isn't because existing digital formats are unable to time it well enough. You can take a loopback through the AD/DA of a laptop computer and show it capable of discerning the difference in time between one meter and two meters of interconnect. That amounts to around 1 nanosecond. 4 microseconds is 4000 nanoseconds.

 

[Orb]

I am talking about the actual technical link I provided that goes into a bit more detail regarding the Pulse Array DAC than provided so far
But yes I can appreciate that point you picked up on may cause a frown as it is rather wishy washy; just that you said it filled up your misinformation quote for the rest of the week so I assumed you wanted more specifics rather than that marketing-tech type presentation with regards to the engineering/DAC.
And I can say that semi-technical paper I linked (page 4 onwards in the paper) is going to get a lot on here asking questions, which is what I am staying out of after the hassle with the last technical discussion.
Cheers
Orb

 

[Orb]

Just to add,
Bob Stuart probably done the "best" article involving temporal encoding from an audio manufacturer (putting aside Harman scientific papers by Toole/Olive/Keele/etc), would had been better I agree if Chord Electronics either went into better detail or omitted that content but then it all depends upon who the presentation is for and scope *shrug*
Cheers
Orb

 

[esldude]

Just to add,
Bob Stuart probably done the "best" article involving temporal encoding from an audio manufacturer (putting aside Harman scientific papers by Toole/Olive/Keele/etc), would had been better I agree if Chord Electronics either went into better detail or omitted that content but then it all depends upon who the presentation is for and scope *shrug*
Cheers
Orb

Yes for general potential customers you don't want a full scientific white paper. But the powerpoint didn't simplify things it outright told untruths. Again, not uncommon in ad copy. But also uncalled for all the same. The other paper you linked to I read over and have no problems with. So it isn't a matter of the level of detail Chord went into, it is a matter of wrong information. And false information about what is known in regard to brain processing and nerve pulse trains.

Robert Stuart has written some nice papers with info about temporal resolution for transparent audio reproduction and on use in multi-channel sound. He lists the correct manner of accessing the temporal resolution of digital into the picosecond range. That is all anyone had to do. If they bring it up tell the known facts about it.

 

[Orb]

Yeah, which is why I mentioned Bob Stuart.
But then the discussion for Chord Electronic was regarding the DAC that some questioned earlier and why I think others posted said links; anyway for temporal timing/perception/etc members can search on Bob Stuart (or Robert Stuart as it is in those papers as you rightly mention) and the link I provided for the technical engineering side of the Pulse Array concept as implemented by Rob Watts/Chord Electronics.
Here is one I was thinking of earlier for Robert Stuart: https://www.meridian-audio.com/ara/multips3.pdf (temporal encoding/time delay/etc are a few pages in).

Cheers
Orb

 

[esldude]

https://www.meridian-audio.com/ara/coding2.pdf

This one is worth reading as well. From page three:


Regarding temporal accuracy, (ii), if the signal is processed incorrectly (i.e. truncated) it is true that the
time resolution is limited to the sampling period divided by the number of digital levels. However, when
the correct dither is used the time resolution also becomes effectively infinite.

The picosecond levels I mentioned referred to redbook standards. With dither it is less than that.

 

[Ken Newton]

I now suspect that Chord's WTA technology is some proprietary method of filter kernel windowing. An FIR filter's kernel actually defines the filter's behavior, and can be graphically revealed by an impulse response test. Chord clearly state their view that commonly used filter kernels are insufficiently long. Which, of course, is a well known fact. Truly perfect signal reconstruction, including amplitude and timing/transient placement resolution, would require an infinitely long kernel. 'Windowing' has long been applied to FIR filter kernels to greatly suppress the adverse performance effects of their non-infinite length. Even with their WTA technology, however, Chord says they found that comtinuing to increase the kernel length still benefits the sound. So, it's hard for us to know to what degree their WTA technology is responsible for their sound, versus their having increased the kernel length.

 

[esldude]

I now suspect that Chord's WTA technology is some proprietary method of filter kernel windowing. An FIR filter's kernel actually defines the filter's behavior, and can be graphically revealed by an impulse response test. Chord clearly state their view that commonly used filter kernels are insufficiently long. Which, of course, is a well known fact. Truly perfect signal reconstruction, including amplitude and timing/transient placement resolution, would require an infinitely long kernel. 'Windowing' has long been applied to FIR filter kernels to greatly suppress the adverse performance effects of their non-infinite length. Even with their WTA technology, however, Chord says they found that comtinuing to increase the kernel length still benefits the sound. So, it's hard for us to know to what degree their WTA technology is responsible for their sound, versus their having increased the kernel length.

Yeah, but so many high end companies have this idea of increasing anything that can be increased leads to increasing sound quality with no apparent end in sight. All of course at increasing expense to customers. 16 to 24 to 32 bit. 32 to 50 to 64 bit DSP, longer and longer filter kernals, higher and higher sample rates. ad infinitum. I would believe it more easily if someone had a good handle on what it is increasing all those parameters do beyond just "we tried it, we listened to it, and more is better". You know, more like scientists in the psycho-acoustic fields that study real limits and figure out what we can and can't hear. Then we can draw limits around what will be fully transparent. So when on top of that a company is taking the more is better, trust us, and we aren't telling you exactly what we are doing (actually we are attempting to hide your testing it), well color me at least reserved in my belief.

 

[wisnon]

Tap length (filter) science is proven. Infinite tap length is theoretically best. Of course there is diminishing return to this strategy, but everyone can test this by demo-ing the latest incarnation with longer tap lengths. If it sounds better than the previous version, then go for it, if not don't.

 

[Ken Newton]

Okay, gang. Soulution has just provided some important new clues about their DAC phase correction technology. The relevant excerpt from their press release follows in quotes:

"zerojtech (Zero-Phase-Technology)

Every D/A converter requires an analog low-pass-filter in its output in order to suppress high frequency noise and aliasing signals. The 560?s 3rd order bessel filter, with a cut-off frequency of 120kHz, would show a phase shift of up to 15° in the audio band. Due to the Zero-Phase-Technology the phase error of the analog output signal remains below 1°, 20Hz – 100kHz! The sonic impact of these seemingly small changes is tremendous.

The Zero-Phase-Technology does resolve one of the biggest problems related to digital-to-analog conversion. With the introduction of the upsampling technology low-pass-filters could be less steep than it was required with a sampling rate of 44.1kHz. But still the cut-off frequency is by far too low for not having any impact on the phase response in the audio band. Upsampling was a great step forward but did not resolve the issue completely. The Zero-Phase-Technology does overcome these short comings for the very first time."

What Soulution appears to be doing is using DSP to correct the in-band phase shift that stems from their implementation of a 3rd order low-pass analog output filter. While that's fine as it stands, I don't see how correcting only 15 degrees of phase shift could result in any subjective improvement. Simply using a more benign analog filter, or eliminating it completely would have the same effect. Erasing, or at least mitigating any benefit is that loudspeaker phase response will again shift the phase, and by some amount that's likely much more than 15 degrees. I had earlier speculated that Soulution may be correcting a linear phase shift of the typical signal reconstruction process (rendering the net response as zero-phase) but the press release doesn't give indication that is the case.

 

[dallasjustice]

Soulution created a problem other DACs don't have and then fixed it?

Okay, gang. Soulution has just provided some important new clues about their DAC phase correction technology. The relevant excerpt from their press release follows in quotes:

"zerojtech (Zero-Phase-Technology)

Every D/A converter requires an analog low-pass-filter in its output in order to suppress high frequency noise and aliasing signals. The 560?s 3rd order bessel filter, with a cut-off frequency of 120kHz, would show a phase shift of up to 15° in the audio band. Due to the Zero-Phase-Technology the phase error of the analog output signal remains below 1°, 20Hz – 100kHz! The sonic impact of these seemingly small changes is tremendous.

The Zero-Phase-Technology does resolve one of the biggest problems related to digital-to-analog conversion. With the introduction of the upsampling technology low-pass-filters could be less steep than it was required with a sampling rate of 44.1kHz. But still the cut-off frequency is by far too low for not having any impact on the phase response in the audio band. Upsampling was a great step forward but did not resolve the issue completely. The Zero-Phase-Technology does overcome these short comings for the very first time."

What Soulution appears to be doing is using DSP to correct the in-band phase shift that stems from their implementation of a 3rd order low-pass analog output filter. While that's fine as it stands, I don't see how correcting only 15 degrees of phase shift could result in any subjective improvement. Simply using a more benign analog filter, or eliminating it completely would have the same effect. Erasing, or at least mitigating any benefit is that loudspeaker phase response will again shift the phase, and by some amount that's likely much more than 15 degrees. I had earlier speculated that Soulution may be correcting a linear phase shift of the typical signal reconstruction process (rendering the net response as zero-phase) but the press release doesn't give indication that is the case.

 

[Ken Newton]

Soulution created a problem other DACs don't have and then fixed it?

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?

 

[dallasjustice]

Problem, reaction, soulution

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 #18) 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?

 

[ack]

but I can't help feeling that if this is all there is to their zero-phase technology, it is as you suggest above.

I would agree with your and Michael's assessment; still can't take TAS very seriously...