Async-USB 24/768 Phasure NOS1 DAC

[barrows]

"I remember when I heard for the first time Gidon Kremer's "Hommage a Piazzola" in CD resolution through the Octave and I was blown away by the additional transient information and the sheer physicality of the images of the acoustic instruments."

 

you were blown away by the sound of additional artifacts folded into the audio band caused by the lack of filtering.

There are a lot of audiophiles who seem to love the sound of NOS/no filtering DACs even playing at 16/44.1 rates, even though this approach clearly has tons of digital artifacts, at high levels, well within the audible range. There is nothing wrong with that, but it probably is a good idea to at least know that you are hearing additions to what is in the music.

 

There is a reason that DAC manufacturers avoid making these kind of DACs for the most part. It is not hard to make a NOS DAC capable of accepting up to 24/768 or even 32/768 input sample rates-just about any TI DAC chip can be used this way (their internal 8x filters can be turned of entirely), but most DAC manufacturers do not want to put out a product with the which produces so much distortion in its output.

 

 

[PeterSt]

Hi Clemens,

 

But of course I am not sure if digital upsampling to 756/24 with the Phasure Nos would even better that significantly

 

Well, I sure am. I have been using the NOS1 for two months or so after the USB interface was ready, but the 768 driver was not yet (384 was). When 768 was ready, I got a whole new DAC again. And new music. Just unrecognizeable (you said similar I think).

So it really matters.

 

At going from 176.4 to 352.8 it was the same story, which is why I hunted for 768 (705.6). If a previous step matters, a next may matter again. Sadly, going from 352.8 to 705.6 matters (even more, and it can be measured why) ... "Sadly" because I can't go higher than this, and I would like to ...

 

Peter

 

[manisandher]

"It is not hard to make a NOS DAC capable of accepting up to 24/768 or even 32/768 input sample rates-just about any TI DAC chip can be used this way (their internal 8x filters can be turned of entirely)..."

 

Hey barrows, I'm not so sure about this. Yes, their internal filters may be bypassable, but I doubt you could run them as NOS DACs - they're all sigma-delta types and need oversampling in order to work. There is a reason why Peter chose the PCM1704U-K for his NOS1 and why Metrum use the chips they use...

 

Mani.

 

[vrao]

@ Barrows

 

Hi,

 

"There are a lot of audiophiles who seem to love the sound of NOS/no filtering DACs even playing at 16/44.1 rates, even though this approach clearly has tons of digital artifacts, at high levels, well within the audible range. There is nothing wrong with that, but it probably is a good idea to at least know that you are hearing additions to what is in the music."

 

This does not make sense to me. Artifacts meaning that the recording engineers have a habit of letting digital artifacts onto the music? And the other DAC manufacturers are making sure these are filtered out? Where does one draw the line?

 

As far as I know the $$$$ or even $$$$$ CD players are trying to get the extra information from the CD - ambient, subtle very last of the details. The more the better it is, apart from other factors, of course.

 

This is an interesting subject, I would like to learn more.

 

p.s this is also my first post on CA, but found your response quite interesting.

 

Kind Regards,

 

 

 

 

 

[Miska]

This does not make sense to me. Artifacts meaning that the recording engineers have a habit of letting digital artifacts onto the music? And the other DAC manufacturers are making sure these are filtered out? Where does one draw the line?

 

No, because the samples are representation of an analog waveform, but they are not the waveform itself. Just as the name says, they are just samples, just like in the kids' connect-the-dots drawing exercises. Filters are used to re-construct the arcs between the dots (now guess where Peter got the name for his algorithm). Upsampling/oversampling filters create new dots between the old ones. Then in the conversion and analog stages the dots are finally connected. Preferably with small correct arcs instead of straight horizontal and vertical lines.

 

If you input sine wave to ADC and then don't have any filtering anywhere before/after DAC you have "square waves" output. Listening to square waves is listening to digital artifacts.

 

Here's one example, original at 44.1 kHz, zoomed-in on Audacity:

 

 

Here's the same position, upsampled to 96 kHz:

 

 

[barrows]

"Hey barrows, I'm not so sure about this. Yes, their internal filters may be bypassable, but I doubt you could run them as NOS DACs - they're all sigma-delta types and need oversampling in order to work. There is a reason why Peter chose the PCM1704U-K for his NOS1 and why Metrum use the chips they use...

 

Mani."

 

My point was, that chips like the TI 1792 (for example) can be run straight, with their internal OSF turned off. This is how Ayre and Aesthetix use the chip, then they do all filtering and oversampling in an FPGA, running their own, proprietary code. In Ayre's case, they run their MP filter, and oversample to 1.4112/1.536 mHz (as I recall), and then that signal goes straight in to the DAC section. So, these chips can be run the same way as using XXHighend with the Phasure NOS. I guess it is really a matter of semantics. Now I do not know what would happen if you tried to feed one of these chips straight 16/44.1, that may not work at all. But, of course, the 1704 does not like running this way either, and will have tons of distortion on its output running in this fashion.

 

Of course, this does not take into consideration the difference in sonics of R2R multi bit DACs, and sigma delta DACs-that is another discussion, and a highly contentious one at that.

 

 

 

[Miska]

chips like the TI 1792 (for example) can be run straight, with their internal OSF turned off. This is how Ayre and Aesthetix use the chip, then they do all filtering and oversampling in an FPGA, running their own, proprietary code. In Ayre's case, they run their MP filter, and oversample to 1.4112/1.536 mHz (as I recall), and then that signal goes straight in to the DAC section.

 

You can bypass the 8x oversampling interpolation filter, but not the 64x oversampling delta-sigma modulator. Those BB chips have a special mixed PCM-SDM converter architecture, where MSBs go through a 8x PCM and LSBs go through 64x SDM. There's no way you can input this dual format natively to the chip because it has two different sampling rates in use at parallel. At the output it merges the two outputs through a special integrator.

 

 

[Clemens Wergin]

@barrows What I just don't get is the following: the whole idea of a NOS is not to oversample or upsample. The idea being that this will cause problems in the time domain. So what is the difference if you upsample 44/16 to 756 and then send it through the Nos from what usual CD players are doing, albeit to lower upsampling rates? Well you could say, you don't do it on the fly like a CD player: you upsample the file first and then send the new file through the Nos. That doesn't convince me though. Because that process in an ordinary computer is usually faster than playing a file. So it is difficult to imagine that upsampling done on the fly is worse in quality than when you convert the file beforehand. I am definitely not into the technical aspects of these things, I just try to apply my simple kind of logic to it...

 

[barrows]

First, if one uses a true NOS DAC, with no filtering, you get huge amounts of distortion (artifacts) in the output. Some audiophiles seem to like that sound, not me! It can sound nice on very simple music, but becomes cloudy and veiled with any music which is complex at all.

 

Peter's approach is based on the theory that he feels he can do the oversampling/filtering much better than anyone else can when he does it in software. To find out the details of how his "Predictive Arc" oversampling works, you will need to ask him, or go to his own forums and read up on it (XXHighend). In any case, it is different than what goes on in the typical DAC chips oversampling/filtering stage. Just as companies like Ayre, Aesthetix, and (I think) Berkeley Audio Designs run their own, proprietary, oversampling/filtering (except) on FPGAs inside their DACs.

 

If Peter's "Predictive Arc" approach really is so revolutionary, I still wish he would license it for use inside DACs. Now, there are certainly DSP chips with enough power to run his approach, and this way more audiophiles could take advantage of his work (and he might be able to make a little money too). Not everyone wants to have to use XXHighend and his DAC.

 

BTW, PeterSt (probably asleep now), is all about time domain performance, so I am pretty sure he will respond to your assertion that his approach may have time domain problems!

 

[PeterSt]

My point was, that chips like the TI 1792 (for example) can be run straight, with their internal OSF turned off.

 

Hi Barrows, if that only were true; we would have switched long ago to this chip. But Miska is right.

 

Regards,

Peter

 

[PeterSt]

So what is the difference if you upsample 44/16 to 756 and then send it through the Nos from what usual CD players are doing, albeit to lower upsampling rates?

 

Clemens,

 

Here Barrows is right. Maybe not because I think I can do it better via software instead of hardware, but because it is much more easy to test and upgrade (think about this latter too).

Besides this Arc Prediction is a genuine interpolating algorithm. And I know, all these "filters" are called Interpolating Filters, but they really are not. So :

 

Filters are used to re-construct the arcs between the dots (now guess where Peter got the name for his algorithm). Upsampling/oversampling filters create new dots between the old ones.

 

This really is not the complete truth. Theoretically yes, but practically these filters move the original samples to somewhere else. Completely somewhere else ...

Therefore what Miska showed counts for such a genuine interpolating filter; not for a filter we are used to use ...

(I don't think it will be possible for such a filter to ever find the same "position" because of the totally different output, unless right at the start of the track where the filter can't work well yet).

 

... but which you may like anyway.

 

Regards,

Peter

 

 

 

 

 

PS: Btw, it is "to 705.6", not 756.

 

 

 

[PeterSt]

As far as I know the $$$$ or even $$$$$ CD players are trying to get the extra information from the CD - ambient, subtle very last of the details. The more the better it is, apart from other factors, of course

 

Hi vrao,

 

That may be what they is said (based on good THD figures), but I really don't see how it can be done with all the problems in the time domain (!) this gives. Well, delta-sigma assumed of course, and sure not every $$$$$ CDPlayer uses that. So, the whole point is : keep the time domain in-tact (no smear) but keep the amplitude domain well under control.

At least this is what the NOS1 allows for (no filtering in there).

And the remainder is up to the creativity of someone like me.

 

Peter

 

PS: Never make a mistake with fake details. They sound interesting, may not occur with certain types of music, but in the mean time they determine the "sound" of the DAC. Well, don't DACs all have a sound ?

 

[Guest Claude]

The NOS1 is something very special. It already sounds very good with audirvana but with XXHighend it is something else.

I am still finding out, how to find the best setting in XXHighend. It is hard to describe so far. The change in sound is so profound, that almost nothing sounds the same anymore.

Right now still a bit too harsh in the hights, but if this goes away (and all NOS1 owners say so), this will be nothing but unbelievable.

What I learned in the first 5 days:

XXHighend and Phasure is all about getting the best sound possible and Peter found a way to tailor the sound through his player. Don't judge this book by its cover.

The NOS1 really needs break in.

You can use the NOS1 with mac and get excellent sound (the up sampling settings in Audirvana are very important, Pure Music does not sound good at all).

You have to try XXHighend. The saying Anything matters becomes a whole different meaning with this player. BUT: do everything exactly as written in the install notes and browse phasure.com. Don't expect fancy optics, easy installation... and what we all usually call UX. This is a bit like in the good old windows days.

The settings in XXHighend really matter a lot (you will not know what this really means until you tried it).

PETER,

why don't you do this for mac users! (No need to answer that one)

It is 2:36 in the morning and I am still sitting in front of my stereo.

Greetings from Berlin

Claudius

 

[Pedal]

I have been a happy owner of the Phasure NOS1 DAC for a year. Upgraded to the 768 USB version a few months ago.

 

First a few words about my previous Buffalo II DAC, build to a high standard, taking it almost as far as can be done.

 

-I think with modern quality DACs like the Buffalo II, we have reached such a high level of fidelity, that we can describe it as “faultless”. In the sense that a trained listener cannot hear anything wrong with its sound. It’s very resolved and transparent, treble is pure and detailed without any edge, and the bass is deep and tight. I remember a year ago, I questioned myself if - and how - it could possibly be improved?

 

Well, that “mystery” was solved when I purchased the NOS1 mk. I.

What I heard was a small improvement across the board. The treble becomes more alive, more real. Not because it emphasizes detail, but because it better reproduces the micro dynamics in the treble. In hindsight I realized that the Buffalo II was slightly too kind/soft in this region.

The same improvement takes place in the midrange, and even more (interestingly) in the bass. With NOS1 you get a more natural sounding bass. Indeed it can rock and shake the earth, but listening to acoustic bass, it digs out more overtones and nuances.

 

The point is that the many small improvements across the frequency range ads up to a significantly more realistic listening experience. The upgrade from mk I to mk II (NOS1 USB) pushed it one step further.

To use an old cliché; The NOS1 USB gives you the impression of listening directly to the mastertapes. Medicore recordings sound more interesting and good recordings transform my listening room into a bandstand.

 

Keep in mind that we are not speaking about an ordinary stand alone DAC. It’s a complete concept of player software + interface + DAC. The player software even takes command of your PC/Mac, making it perform better, sound wise.

 

Also, the player (XXHighend) improves your CDs through the proprietary ArcPredict algorithm which reconstructs the square waves in the treble, turning your old CDs into hi-rez quality. With the Buffalo I heard a marked difference between the CD and the hi-rez version. With XX/NOS1 the gap is almost closed.

 

The only concern is that XXHighend still is in beta version and that the setup of the NOS1 DAC and the PC takes some patience. It’s not a plug&play solution. It needs a little bit of care and attention, like a high end turntable combo does. (I trust PeterSt to improve on these minor quibbles when time allows him). After all, you listen to the sound, not the Graphic User Interface...

 

On the other hand you can expect very quick and extensive service from the Phasure.com user forum. I have seen PeterSt helping out overseas users again and again, for free, even if they only use the program in free demo mode.

 

The reward is a state of the art sound quality. If there are DACs of equal sound quality, I am afraid you have to look at competitors costing 10 times the price of a Phasure NOS1!

 

 

[manisandher]

"Right now still a bit too harsh in the hights, but if this goes away (and all NOS1 owners say so), this will be nothing but unbelievable."

 

Claude, if you're still finding this to be the case, then you may want to give Miska's HQPlayer a go. I'm getting very nice results with the following settings:

 

- 16/44.1 upsampled to 352800

- poly-sinc-short-mp (minimum phase polyphase sinc filter, no pre-echo and short post-echo)

- noise-shaping: NS5 (fifth-order noise-shaping, especially good with PCM1704 chips)

 

The NOS1 and HQPlayer make a very nice match...

 

Mani.

 

[Guest Claude]

Thanks Mani,

I will give it a try!

Claudius

 

[Telstar]

Barrows,

If Peter's "Predictive Arc" approach really is so revolutionary, I still wish he would license it for use inside DACs. Now, there are certainly DSP chips with enough power to run his approach, and this way more audiophiles could take advantage of his work (and he might be able to make a little money too). Not everyone wants to have to use XXHighend and his DAC.

 

I dont think there currently are DSP chips with enough power. On windows,xxhighend player 8X AP uses about the power of a dualcore 2ghz intel C2D. And it's completely multicore/multithreading. Mani uses a 6-core HT xeon IIRC.

Maybe in 3-4 years.

 

[stereotaipei]

Chris,

I believed that you were supposed to receive one NOS1 for review more than 8mths ago.. what happened? is it still planned to review it?

thanks

[Julf]

Also, the player (XXHighend) improves your CDs through the proprietary ArcPredict algorithm which reconstructs the square waves in the treble, turning your old CDs into hi-rez quality.

 

So it works a bit like MP3 and other perceptual codecs?

[barrows]

Julf, I think the answer is "no", and hopefully Peter will weigh in. My understanding is that his "predictive arc" oversampling algorthms are designed to result in a more accurate waveform than is possible with ordinary DAC oversampling filters, and they are not designed "by ear" to be euphonically "pleasing". As I recall, Peter claims no ringing, and full scale transient response which, AFAIK, is unheard of with typical OSFs and redbook playback.

[Julf]

My understanding is that his "predictive arc" oversampling algorthms are designed to result in a more accurate waveform than is possible with ordinary DAC oversampling filters, and they are not designed "by ear" to be euphonically "pleasing".

 

OK! The description, by pedal, that "(XXHighend) improves your CDs through the proprietary ArcPredict algorithm which reconstructs the square waves in the treble, turning your old CDs into hi-rez quality" sounded like it was adding some harmonics/distortion (the only way you can "recreate the square waves"). And we know you can never turn Cd material into true hi-rez after the fact - when information is gone, it is gone.

[eternaloptimist]

OK! The description, by pedal, that "(XXHighend) improves your CDs through the proprietary ArcPredict algorithm which reconstructs the square waves in the treble, turning your old CDs into hi-rez quality" sounded like it was adding some harmonics/distortion (the only way you can "recreate the square waves"). And we know you can never turn Cd material into true hi-rez after the fact - when information is gone, it is gone.

 

Where's the "Subscribe to Julf's posts" button?? Gold.

[manisandher]

I'm not sure what PeterSt's 'Arc Prediction' algorithm does exactly, but what I do know is that 99.9% of XXHighEnd users use it and that 100% of his NOS1 DAC owners use it. I doubt we're all into "harmonics/distortion". Oh and his Arc Prediction scheme also substantially improved the performance of my Pacific Microsonics Model Two when I had it, with its already supposedly very fine, albeit quite old, filters.

 

On a final note, I find XXHighEnd's 'Arc Prediction' filter to be substantially superior to HQPlayer's non-ringing filters, which actually do make the music sound distorted and quite painful to listen to, just as Miska warns. (Fortunately, the minimum phase filter offerings in HQPlayer are totally superb.) There is something that PeterSt has applied in his 'Arc Prediction' scheme that totally mitigates the ill-effects of other non-ringing filters.

 

Oh and the NOS1 still remains easily the best DAC I've ever heard...

 

Mani.

[Julf]

I'm not sure what PeterSt's 'Arc Prediction' algorithm does exactly, but what I do know is that 99.9% of XXHighEnd users use it and that 100% of his NOS1 DAC owners use it. I doubt we're all into "harmonics/distortion".

 

OK, do you know of any way you can "reconstruct the square waves in the treble" without adding harmonics? And isn't that the definition of distortion?

 

The problem is that as you state, nobody knows what the "Arc Prediction" algorithm does exactly. If it is really such a wonderful thing, my suggestion to PeterSt is to patent it as soon as possible, and license it to all major manufacturers. But I understand he might not care for wealth and fortune...

[earflappin]

I'm not sure what PeterSt's 'Arc Prediction' algorithm does exactly, but what I do know is that 99.9% of XXHighEnd users use it and that 100% of his NOS1 DAC owners use it. I doubt we're all into "harmonics/distortion". Oh and his Arc Prediction scheme also substantially improved the performance of my Pacific Microsonics Model Two when I had it, with its already supposedly very fine, albeit quite old, filters.

 

On a final note, I find XXHighEnd's 'Arc Prediction' filter to be substantially superior to HQPlayer's non-ringing filters, which actually do make the music sound distorted and quite painful to listen to, just as Miska warns. (Fortunately, the minimum phase filter offerings in HQPlayer are totally superb.) There is something that PeterSt has applied in his 'Arc Prediction' scheme that totally mitigates the ill-effects of other non-ringing filters.

 

Oh and the NOS1 still remains easily the best DAC I've ever heard...

 

Mani.

 

Hi Mani, would you mind sharing what DACs you have compared your NOS1 to in addition to the PM2? Thanks.