Berkeley Alpha USB still relevant?

[Qhwoeprktiyns]

http://www.crystek.com/documents/appnotes/SourcesOfPhaseNoiseAndJitterInOscillators.pdf&ved=2ahUKEwiin42ukZDcAhUDwBQKHWTUCqUQFjAFegQIARAB&usg=AOvVaw1QY9RewFxI5h-F1Wrl1oIN

 

"Introduction: The output signal of an oscillator, no 
matter how good it is, will contain 
all kinds of unwanted noises and 
signals. Some of these unwanted signals 
are spurious output frequencies, harmon-
ics and sub-harmonics, to name a few. "

 

"Conclusion: The output of an oscillator is not perfect. 
Due diligence must be conducted by the 
system engineer in specifying and validat-
ing performance of the oscillator correctly. 
Also, the system itself can easily corrupt 
an oscillator with either conducted or radi-
ated signals. As experienced RF engineers 
know, it is best NOT to create/generate 
any un-wanted signals in the first place, 
rather than try to filter them somehow 
after they’ve been produced"

 

[mansr]

7 minutes ago, barrows said:

While I agree in a sense, for instance with an ESS chip running in its (normal) asynchronous mode with its internal DPLL active, this is not the case for the same chip(s) if one wants to run them in the synchronous mode.

The resampling filters, sigma-delta modulator, and output stage are clocked by the master clock, not the I2S bit or word clock. As I said, the I2S input needs to be synchronised with the master clock, but no phase relationship is required. Here's a quote from the DSD1793 datasheet:

 

Quote

The DSD1793 requires the synchronization of PLRCK and the system clock, but does not need a specific phase relation between PLRCK and the system clock. If the relationship between PLRCK and the system clock changes more than ±6 PBCK, internal operation is initialized within 1/fS and analog outputs are forced to the bipolar zero level until resynchronization between PLRCK and the system clock is completed.

 

A little jitter on the I2S inputs is completely harmless.

[Superdad]

1 hour ago, barrows said:

Ayre does the following:  

 

1. The USB receiver circuit is powered by an independent, isolated power supply.

2. The USB receiver circuit is isolated from the rest of the DAC by optocouplers.

3. The masterclock(s) are located on the "clean" side of the isolation, close to the input of the DAC/processing stage.

4. 1 & 2 above insure the DAC and analog stage are isolated from noise produced both by the source component (USB feed) 

     and from the USB processor itself.

5. The digital feed to the DAC/processing stage is re-clocked, directly by the masterclock, right before the dAC/processing 

    stage.

 

This approach provides the highest degree of noise isolation from both the USB feed, and the USB processor itself, while keeping the masterclock(s) performing at their best (by having them on the "clean" side of the isolation).

 

And yet would you not agree that the QX-5 still benefits from a high signal integrity USB source such as the ultraRendu?  I know of several QX-5 owners who enjoy such from use of our ISO REGEN. 

 

BTW, I seem to recall that the USB board of the QX-5 is entirely 5VBUS powered.  That makes it easier to maintain its post-XMOS galvanic isolation on the I2S lines.  Such is very common.  Berkeley does the same with their Alpha USB box.

[barrows]

4 minutes ago, Superdad said:

And yet would you not agree that the QX-5 still benefits from a high signal integrity USB source such as the ultraRendu?  I know of several QX-5 owners who enjoy such from use of our ISO REGEN. 

 

BTW, I seem to recall that the USB board of the QX-5 is entirely 5VBUS powered.  That makes it easier to maintain its post-XMOS galvanic isolation on the I2S lines.  Such is very common.  Berkeley does the same with their Alpha USB box.

Alex, as you well know, no "isolation" is ever perfect.  I was referencing to the Ayre interface as an example of the "best" way to do it.  But never would I suggest that any USB is entirely immune from source quality (I believe I mentioned this in my prior post(s).  As your customers can confirm, there are often improvements from multiple layers of isolation.

As to power, The Ayre QX-5 has four transformers on board, with multiple secondary windings, I was taking a little license in suggesting that the USB receiver power supply was from the onboard source-given the four transformers.  Perhaps @Ryan Berry can confirm.  I agree that powering the USB receiver from USB power is an easy way to provide the needed power supply isolation, and that is what I do in my DIY DACs (of course that power comes from a very clean supply in my case, the Signature Rendu SE).

I do recall that Ayre made an upgrade during the life of the QB-9 DAC where they did power the (isolated) USB receiver from an independent onboard rail.

[barrows]

28 minutes ago, mansr said:

The resampling filters, sigma-delta modulator, and output stage are clocked by the master clock, not the I2S bit or word clock. As I said, the I2S input needs to be synchronised with the master clock, but no phase relationship is required. Here's a quote from the DSD1793 datasheet:

 

 

A little jitter on the I2S inputs is completely harmless.

Depends on the DAC chip and the approach used.  I have no doubt that with the ESS chips in sync mode (DPLL and async SRC off) the incoming jitter is quite audible, enough so that a 6 dB improvement in XO phase noise at 10 Hz is entirely audible.

Also with AKM chips featuring the pure/direct DSD, the designers using this approach with that chip have expressed that it is critical for the data lines to be phase aligned.  And then there are the chinless DAC conversion approaches, like DSC-1 (Holo Audio, T+A, etc), I will leave that to @Miska to comment on if he sees fit as to incoming jitter, and phase relationships between DSDR, DSDL, and bit clock.

Having participated in measured jitter levels vs. listening tests, if I were designing a DAC, i would do everything possible to insure lowest possible jitter level at the conversion stage input.  A flip flop is a small price to pay, and again, isolation chips add considerable jitter (most values I have seen reported are around 200 pS) I would not use them without re-clocking.

[barrows]

@hopkins, I am not following your logic.  Re-clocking dose not require adding a clock, you have to have a masterclock already.

 

What bothers me is the need for two masterclcoks to accommodate 44.1/48 base frequencies.  With my DAC, I am running just the 44.1 base masterclock, such that I need only a single clock on board, and then I oversample everything to DSD in software.

[Miska]

1 hour ago, barrows said:

And then there are the chinless DAC conversion approaches, like DSC-1 (Holo Audio, T+A, etc), I will leave that to @Miska to comment on if he sees fit as to incoming jitter, and phase relationships between DSDR, DSDL, and bit clock.

 

In DSC-1 the timing is entirely on BCLK as long as it transitions somewhere within stable period of the data lines, because the value goes through a latch driven by the clock. But due to design, the phase between data lines and BCLK need to stay well within stable period of the data lines. Since for this kind of design, there's no higher speed MCLK nor low speed WCLK, one can utilize lower phase-noise of 512x (22.5792/24.576M) clocks without need to frequency dividers or DPLL's.

 

Using many (32) unity-weighted elements like DSC-1 gives lowest jitter sensitivity (btw, ESS uses 64 unity-weighted elements, while dCS uses 24), but one can decide the exact details when designing the DAC on what aspects to emphasize on the design.

[Summit]

17 hours ago, barrows said:

@KingRex, no worries.  My comments refer specifically to the topic the OP posted: is a USB-SPDIF converter irrelevant with the Ayre QX-5.  I was not commenting in general on other component pairings.

 

@Summit, no, there is no need to hear the difference in this case, as what we are talking about here are technical issues which are very well understood and not "magical" in nature.  I have a very good technical understanding of how the USB interface in the QX-5 is implemented, and as such, there will not be an advantage to inserting another component to convert to SPDIF.

 

I don’t know what you mean then you say “magical” in nature. Is the improvements “from multiple layers of isolation” with the Berkeley Alpha USB somehow magical in nature? How about the Signature Rendu SE isn’t it also magical in nature by the same reasoning?

 

I disbelief your technical understanding of how the digital interfaces in the QX-5 are implemented and which effects difference upstream gear can have. Others that have actually tried them have sometimes favoured other inputs than USB and as usual it depending on which upstream gear they have, so it’s not a merely a matter of technical understanding of how the digital interfaces in the QX-5 are implemented.

 

“Playing files with Roon also revealed that the sound had more authority via the QX-5's network input than via USB. Christian McBride's solo double bass at the start of "All or Nothing at All," from Diana Krall's Love Scenes (DSD64 file, Impulse!/Acoustic Sounds), had a slightly more optimal combination of weight and definition via Ethernet than via USB; in fact, playing CD files over the network connection sounded pretty much identical to playing the original CDs in my Ayre transport and feeding the data to the QX-5 Twenty via an AES/EBU link, with much the same sense of authority. Via USB, that authority was slightly diminished.”

 

https://www.stereophile.com/content/ayre-acoustics-qx-5-twenty-da-processor-page-2#x5Pajuur5F7G9iyU.99

 

“The QX-5 has a patent pending asynchronous S/PDIF input that completely eliminates all jitter from that interface. In fact it is the best sounding input, simply because there is no computer in the system to introduce EMI and RFI. But of course then you lose all the convenience advantages of computer playback. But it's a great check to let you know how good your computer setup is. The closer your computer sounds to an S/PDIF input, the better your computer setup is.”

 

 

 

[mansr]

13 hours ago, Miska said:

In DSC-1 the timing is entirely on BCLK as long as it transitions somewhere within stable period of the data lines, because the value goes through a latch driven by the clock. But due to design, the phase between data lines and BCLK need to stay well within stable period of the data lines. Since for this kind of design, there's no higher speed MCLK nor low speed WCLK, one can utilize lower phase-noise of 512x (22.5792/24.576M) clocks without need to frequency dividers or DPLL's.

The clock that matters is the one driving the D/A conversion stage. In your DSC-1 design, this is the DSD bit clock. In PCM DACs it is typically a "system" or "master" clock at 12.288 MHz (256x) or higher. The I2S clocks are not critical. The datasheets even say so explicitly. I quoted BB/TI above. Here's AKM:

 

Quote

The external clocks, which are required to operate the AK4497, are MCLK, BICK and LRCK. MCLK, BICK and LRCK should be synchronized but the phase is not critical. The MCLK is used to operate the digital interpolation filter, the delta-sigma modulator and SCF.

Quote

The AK4497 has a DSD playback function. The external clocks that are required in DSD mode are MCLK and DCLK. MCLK should be synchronized with DCLK but the phase is not critical.

 

[Marcin_gps]

@KingRex Mojo Mistique v3 uses the JLsounds USB input board with an independent PSU just for this board. I'm pretty sure that the Berkeley USB converter will not improve the sound quality. 

[barrows]

@summit: The subjective sound quality assessment you posted of the QX-5's performance via its Ethernet input vs. USB input could very well be an indication of poorer performance from the Ethernet input vs the USB input.  Often more jitter will result in a sound which has "more weight".  And system tonal balance problems could make this higher jitter sound preferable.  Without appropriate measurements to confirm or deny this, we just do not know the answer.  I do know the Ethernet input on the QX-5 leaves a lot of performance on the table by its nature, this is not to say it is "bad", just that it is not as good as it could be.  Atkinson often listens to the USB input directly from a Mac, and as mentioned I already have said that a good USB source is still much preferred (for any DAC).  I will go back and re-read parts of the review for more details.

 

When I say "magical", I am talking about when people comment on design aspects which are actually well understood technically, but still, somehow, those people want to insist that the technically better choice is actually worse.  These audiophiles are the same ones who suggest things like: "we know nothing about digital audio", and "USB sucks for audio", both of which are patently untrue.  This leads me to believe the real culprit is likely something else in the system which is wrong.  To his credit, JA has been at the reviewing game long enough to usually catch this kind of thing, but no reviewer these days seems to have enough time to really fully evaluate every aspect of a component (especially measuring every input and correlating those measurements with sound quality).  Unfortunately, most audiophiles do not have enough technical understanding to get this, and instead end up blaming a component (or input) which is not to blame.

As for the SPDIF (and AES, which is SPDIF just balanced) input, yes, the QX-5 has very special SPDIF approach, i recently talked to Ayre engineer Ariel Brown about this, and indeed he confirmed it is a true asynchronous approach, which is very, very rare.  Such an approach allows the (usually worse from a jitter perspective) SPDIF input to perform EQUALLY WELL as the USB input (both use the same masterclock(s).  But adding an entire other component (Alpha USB) just to use an equally good input does not make a lot of sense to me.

 

Add: The measurements of jitter were made via a direct connection from a MacBook for USB, this is of course a worst case scenario, additionally, it is probably time for the ancient J-Test way of measuring jitter to be retired, as most decent DACs will pass this test easily.  I am not sure how to test this better, as it will be different for different DACs.  But plotting master clock phase noise at the input of the DAC chip would be interesting to see (for this DAC which uses an ESS 9038), vs. the different inputs.  As far as the J-Test goes this is essentially perfect: so if the different inputs do indeed sound different (as JA reports), then we can see that the J-Test is not showing us what we really need to know here. 

[KingRex]

2 hours ago, Marcin_gps said:

@KingRex Mojo Mistique v3 uses the JLsounds USB input board with an independent PSU just for this board. I'm pretty sure that the Berkeley USB converter will not improve the sound quality. 

I just don't know.  Ben wants me to Install the JCat Femto USB board with an Independant PS in the server.  That tells me there is something to gain. 

 

I have the CAT server.  Not the Deja Vu. Seems there are other upgrades as well such as a lndustrial low power mob with low latency ram and mSATA SSD operating system.  Not to mention a solid AL case with multiple voltage regulators.  Shielded compartments to hold ssd.  

 

I think people get pretty hung up on one particular this or that and loose sight of the whole.  I don't know the price of the Berkeley, but bouncing around trying this and that may in the end cost more that just getting a well made complete unit.  

[barrows]

@KingRex, Going to Ethernet distributed audio will improve performance even more if done right.  And it can simplify things (as the "server" does not need to be so tweaky at all, it can even just be a NAS, just make sure to ground it and the router).

[R1200CL]

50 minutes ago, KingRex said:

I don't know the price of the Berkeley, but bouncing around trying this and that may in the end cost more that just getting a well made complete unit.

 

Maybe 10 times more than Singxer SU-1.  Which may be one of the best USB to SPDIF converters. (Modified and with an LPS-1.2)

 

The ultraDigital is in my opinion almost (or equal)  as good.

It’s  based on the SU-1. 

 

I haven’t tried Berkeley. The two others I own. 

[Ryan Berry]

On ‎7‎/‎8‎/‎2018 at 1:00 PM, barrows said:

Alex, as you well know, no "isolation" is ever perfect.  I was referencing to the Ayre interface as an example of the "best" way to do it.  But never would I suggest that any USB is entirely immune from source quality (I believe I mentioned this in my prior post(s).  As your customers can confirm, there are often improvements from multiple layers of isolation.

As to power, The Ayre QX-5 has four transformers on board, with multiple secondary windings, I was taking a little license in suggesting that the USB receiver power supply was from the onboard source-given the four transformers.  Perhaps @Ryan Berry can confirm.  I agree that powering the USB receiver from USB power is an easy way to provide the needed power supply isolation, and that is what I do in my DIY DACs (of course that power comes from a very clean supply in my case, the Signature Rendu SE).

I do recall that Ayre made an upgrade during the life of the QB-9 DAC where they did power the (isolated) USB receiver from an independent onboard rail.

 

Hi Barrows,

As you said, we power the entire USB section, including the USB receiver internally with the DSD-upgraded QB-9s, the DSD-upgraded DX-5s, and the QX-5.  It was kind of a late addition for us as we were always happy with the interface until a customer swore that it sounded WAY better using a USB cable that had an externally powered +5V supply running off some batteries than using the PC's power to run the older QB-9s.  So we sat down and experimented with it for about 2 weeks and couldn't get it to sound any different no matter what we did with.  What we realized (we think) is that we were running it off of a MacBook with an internal battery.  Even plugged into the wall it didn't seem to make a difference  That said, we also used a Mac Mini and didn't have much better luck there either.

So we hauled in an old PC and tried that, and sure enough we could hear the difference just like the customer suggested.  So we surmised that it's really dependent on how good the PC's power supply is.  I can't say for sure if it's a battery vs. non-battery thing, since we've only tried about 10 computers in total in that room, but there was enough of a difference that we felt it could at least offer some level of consistency across computers if we internally power the USB on the higher series units. 

 

The +5V line is completely disconnected at the PC for those products and some people have cut them out of the cable entirely.  So that ultimately leaves the ground and signal from the PC interfacing with the receiver chip, which there's not a lot you can do about.  For that reason, we use opto-isolation to allow us to separate the ground, power, and signal lines from the USB portion of the unit to the rest of our circuit (we do something similar for our Ethernet as well).  However, even with all of this, the computer still seems to play a role in how the unit sounds.  PCs tend to sound different than Apples which tend to sound different than rigs running Linux.  One program sounds different than another.  A custom-built PC that I have sounded fantastic compared to a custom-built PC I made later to run more exciting games on.  I wouldn't doubt that one USB chipset "sounds" different than another even though it's supposed to only be sending 1's and 0's.  There's no such thing as complete immunity in that regard since the problem is originating earlier than in the unit itself and goes back to the computer...which are not made to prioritize digital audio.

 

So the big questions is would a converter help the USB sound better?  It shouldn't.  But then, there's a lot of things that "shouldn't" help that people all over the place swear by, so it makes it really hard to say definitely if something is going to make the difference for you or not without trying it.  Why does one USB cable make a unit sound different than another?  Why does Ethernet on the QX-5 sound better to some while USB sounds better to others?  Why does it change when I listen to it at Ayre vs. at home?  It all comes down to a ton of experimenting and figuring out a way to keep an opinion from being biased...which is nearly impossible in many of those cases.  I can't double-blind what building I'm sitting in, for example.  So the best I can suggest is to try it and see if it makes a difference to you.  If you think it does, let us know, because computer audio is one of the most fascinating and frustrating things that I think a lot of companies are still figuring out.  I really enjoy trying to recreate what people report to me to see if I can hear it here.  A lot of the time, I can't.  Some of the time I can, and then we make the gear better if it's something we have control of.

 

The next thing on my to-do list is to run a QX-5 from the TOSLINK out of the computer.  You'd have no ground issues whatsoever from the computer and get the advantage of the asynchronous S/PDIF.  For playing music, it may be the way to go...I just have to check whether my computer will let me sent more than 96k into it.  If we're lucky, we can argue about how much TOSLINK cables make the difference someday!

[Superdad]

2 hours ago, Ryan Berry said:

So that ultimately leaves the ground and signal from the PC interfacing with the receiver chip, which there's not a lot you can do about. 

 

Awesome post Ryan!  I think it is very helpful for your client-base to hear such honest remarks.

 

As for feeding your DACs a truly galvanically isolated USB signal of very high integrity (near-perfect eye-pattern), I'd be delighted to send you one of our ISO REGENs to try in your system.  We already have a lot of Ayre DAC owners using it, so it might be fun for you to get a taste of what they experience.  Just send me a PM if interested.

 

Best,

--Alex C.

[Summit]

20 hours ago, barrows said:

@summit: The subjective sound quality assessment you posted of the QX-5's performance via its Ethernet input vs. USB input could very well be an indication of poorer performance from the Ethernet input vs the USB input.  Often more jitter will result in a sound which has "more weight".  And system tonal balance problems could make this higher jitter sound preferable.  Without appropriate measurements to confirm or deny this, we just do not know the answer.  I do know the Ethernet input on the QX-5 leaves a lot of performance on the table by its nature, this is not to say it is "bad", just that it is not as good as it could be.  Atkinson often listens to the USB input directly from a Mac, and as mentioned I already have said that a good USB source is still much preferred (for any DAC).  I will go back and re-read parts of the review for more details.

 

When I say "magical", I am talking about when people comment on design aspects which are actually well understood technically, but still, somehow, those people want to insist that the technically better choice is actually worse.  These audiophiles are the same ones who suggest things like: "we know nothing about digital audio", and "USB sucks for audio", both of which are patently untrue.  This leads me to believe the real culprit is likely something else in the system which is wrong.  To his credit, JA has been at the reviewing game long enough to usually catch this kind of thing, but no reviewer these days seems to have enough time to really fully evaluate every aspect of a component (especially measuring every input and correlating those measurements with sound quality).  Unfortunately, most audiophiles do not have enough technical understanding to get this, and instead end up blaming a component (or input) which is not to blame.

As for the SPDIF (and AES, which is SPDIF just balanced) input, yes, the QX-5 has very special SPDIF approach, i recently talked to Ayre engineer Ariel Brown about this, and indeed he confirmed it is a true asynchronous approach, which is very, very rare.  Such an approach allows the (usually worse from a jitter perspective) SPDIF input to perform EQUALLY WELL as the USB input (both use the same masterclock(s).  But adding an entire other component (Alpha USB) just to use an equally good input does not make a lot of sense to me.

 

Add: The measurements of jitter were made via a direct connection from a MacBook for USB, this is of course a worst case scenario, additionally, it is probably time for the ancient J-Test way of measuring jitter to be retired, as most decent DACs will pass this test easily.  I am not sure how to test this better, as it will be different for different DACs.  But plotting master clock phase noise at the input of the DAC chip would be interesting to see (for this DAC which uses an ESS 9038), vs. the different inputs.  As far as the J-Test goes this is essentially perfect: so if the different inputs do indeed sound different (as JA reports), then we can see that the J-Test is not showing us what we really need to know here. 

 

I believe that Berkeley Alpha USB and other comparable audio gear are design on well understood technical principles and tech and aren’t "magical in nature". The people that prefer to use a Berkeley Alpha USB, Singxer SU-1, Schiit Eitr or similar devices aren’t inevitably believing in magic. To separate the DAC from the computer is one of the best thing one can do IMO, as all standard computers generates a lot of unwanted pollutions like EMI, EMC, RFI etc. To convert and reclock from one digital interface to another can be done without degrading the digital signal and instead result in less noise, jitter, leakage current, ground loops etc etc.  

 

You said “those people want to insist that the technically better choice is actually worse”. I must inform you that there is no tech that everyone agree on is better, no one. Please provide some hard evidence that USB is technically better than other digital interfaces and that sending a signal direct from a computer to a DAC is better than using a Berkeley Alpha USB as you suggested. Maybe it’s my ignorance, but to me you don’t seems to have a “very good technical understanding of how the USB interface in the QX-5 is implemented”.

 

You also said: “But adding an entire other component (Alpha USB) just to use an equally good input does not make a lot of sense to me.” It does - it’s the same reason that make one use a renderer or ISO REGEN to get (as you explained in another post) the “improvements from multiple layers of isolation”. More is better in this case and I use both a JCAT LAN card, an ultraRendu and an Off-ramp 5 (all feed by their own PSU), because it sound better than USB direct to DAC or uR direct to DAC.

 

You seems to believe that measurements can be used to show the best digital interface. Okay let’s say it does. How about presenting some measurements on the Signature Rendu SE. You must have them if you have design it.  

[Summit]

10 hours ago, Ryan Berry said:

So the big questions is would a converter help the USB sound better?  It shouldn't.  But then, there's a lot of things that "shouldn't" help that people all over the place swear by, so it makes it really hard to say definitely if something is going to make the difference for you or not without trying it.  Why does one USB cable make a unit sound different than another?  Why does Ethernet on the QX-5 sound better to some while USB sounds better to others?  Why does it change when I listen to it at Ayre vs. at home?  It all comes down to a ton of experimenting and figuring out a way to keep an opinion from being biased...which is nearly impossible in many of those cases.  I can't double-blind what building I'm sitting in, for example.  So the best I can suggest is to try it and see if it makes a difference to you.  If you think it does, let us know, because computer audio is one of the most fascinating and frustrating things that I think a lot of companies are still figuring out.  I really enjoy trying to recreate what people report to me to see if I can hear it here.  A lot of the time, I can't.  Some of the time I can, and then we make the gear better if it's something we have control of.

 

 

Exactly, it depending on which upstream gear you have, preference, mains power quality and DAC.

[Qhwoeprktiyns]

I have tried a lot of software and PCs with USB.  It is very frustrating... As Ryan Berry concluded, i suspect the recipe to a good audio source from a PC just does not exist yet. 

 

The best sound I got is using the Tonal app (see the corresponding blog on this site), which runs on a Mac. Running on the same Mac  RoonBridge or Audirvana does not give the same SQ, far from it. The Mac Mini I use does not have an optimized OS (high Sierra running without any "optimizations").

I have made these comparisons with my Mosaic UV DAC, which I believe has a very good (and original) USB implementation and is not using the power line from the USB cable. 

 

From all this I conclude that noise of the PC is not only the issue.  You can have the quietest PC and still get bad audio quality... I tend to think it is also a question of how the software is generating the USB signal, and more à question of timing (how the signal output deviates from the audio frequency) but I may be wrong...  Simple buffer settings in Alsa (for example) unfortunately don't do the trick. There is no guarantee that a USB device attached to a PC (regenerator, reclocker, convertor... Whatever) will improve the USB audio signal either... 

 

I cam across some interesting work on this recently that you can read here: https://kokkinizita.linuxaudio.org/linuxaudio/zita-ajbridge-doc/quickguide.html

The developer also has worked on a network audio solution (zita-njbridge) that seems to provide very good results as well. I plan on testing these.

 

All this, however, is probably much less important than everything that goes on within the DAC! 

[mansr]

26 minutes ago, hopkins said:

I tend to think it is also a question of how the software is generating the USB signal, and more à question of timing (how the signal output deviates from the audio frequency) but I may be wrong...

Software doesn't generate the USB signal, hardware does. The host controller handles the low-level protocol without software intervention. DACs using the asynchronous/adaptive interface (all of them these days) do not depend on the precise timing of USB data.

 

26 minutes ago, hopkins said:

Simple buffer settings in Alsa unfortunately don't do the trick.

No, those settings determine the size of the ring buffer the USB hardware reads from. The individual USB packets are much smaller, and the hardware doesn't care how often the source memory address repeats.

 

26 minutes ago, hopkins said:

There is no guarantee that a USB device attached to a PC (regenerator, reclocker, convertor... Whatever) will improve the USB audio signal.

Of course not, since there is always the possibility that the upstream port is better.

[Qhwoeprktiyns]

Well in that case something else is going on... With the exact same hardware, how can you explain differences due to the software? How can you explain that with the exact same hardware and software people hear differences between various settings such as buffering? 

 

 

[mansr]

8 minutes ago, hopkins said:

Well in that case something else is going on... With the exact same hardware, how can you explain differences due to the software? How can you explain that with the exact same hardware and software people hear differences between various settings such as buffering?

I'm not convinced they do.

[Qhwoeprktiyns]

Sometimes I wonder too...  But I have heard some differences which I wish did not exist! When you can consistently repeat them it makes it hard to dismiss them. 

 

I wish some source immunity could be achieved. 

[LowMidHigh]

7 minutes ago, hopkins said:

Sometimes I wonder too...  But I have heard some differences which I wish did not exist! When you can consistently repeat them it makes it hard to dismiss them. 

Only ABX can establish that, preferably on an unbiased N, N being the sample size. Anything else is psychoaccoustics, vainaty and/or product biase (or good ol' fashion shilling) 

[Qhwoeprktiyns]

Well then there is a lot of wasted time and effort from some people on this forum to try optimizing the source