Devialet USB input

[Confused]

This exchange has been 'stolen' from Computer Audiophile:


Originally Posted by JohnSwenson
The large increase in perceived bass is likely due to leakage loops moving around due to the addition of another power supply in the system. Without seeing a map of your system (power supply and interconnects) it's hard to tell exactly what might be happening.

I keep on meaning to write up an article on this and how assembling the whole power network and interconnect network can radically change your system with out much change in your equipment.

I'm also working on some products specifically designed to alleviate issues caused by these leakage, loops, but they are not out yet.

John S.


Originally Posted by Confused
This interested me. Without going into too much detail, running my mR via direct USB into a Devialet appears to provide more (but less accurate) bass than if I run mR, to a Mutec MC3+USB to the AES/EBU Devialet input. However, via the Mutec AES/EBU route the bass is far more balanced and a LOT more accurate and realistic. For a while it has baffled me why the Devialet USB input should sound so different in this particular aspect than the AES/EBU input, the same one's and zero's are being sent. Could this be related to the 'leakage loops' you mention? With the loops being broken by the Mutec? Or maybe I am simply clutching at straws here!
It very well could be a major factor in the difference. It is quite easy to galvanically isolate an SPDIF/AES3 interface and very hard to do this with USB. For quite some time I have been suspecting that the recent fondness for SPDIF over USB is due to leakage loops rather than actual performance of the interface.

Originally Posted by JohnSwenson

Because the microRendu is connected to the computer system with wired Ethernet which is already galvanically isolated, there are two different ways to break a leakage loop in the microRendu to DAC path, one is to galvanically isolate the USB connection, the other is to galvanically isolate the power supply to the microRendu. There are couple of devices that can isolate the USB connection, and a couple ways to do it in the power supply. You can run the microRendu off batteries, or use an LPS-1, an isolated supply I have designed that will be going on sale very soon now.

Either of these approaches will work, you do not need both at the same time.

John S.

[Confused]

OK - Regular readers will know that I have found some issues with bass performance with USB sources, including both Melco and microRendu. It does puzzle me why bass performance should be fundamentally different from a source feeding via USB and one by AES/EBU. What is wrong with the USB input exactly? It's the same 1's and 0's being fed to the same amp, very puzzling! Then I read the above from John Swenson re 'leakage loops', and this got me thinking....

OK, I will be 100% honest here, I do not know what a leakage loop is or why it may influence audio performance. The reason I am interested is the 'leakage loops' issue as described does appear to exactly match what I have observed via USB. Plenty of bass, that is more bass than via AES/EBU, but bass with less definition and accuracy. Quite notably, this was noted by a number of people at an OAC demonstration of the Aurender N10, where we listened to one track via USB, then again via AES/EBU. Also, it had been niggling me a bit that I had not particularly noticed this issue with the CAD CAT via USB. OK, I listened to the CAD CAT for only about 45 minutes, but bass issues I had noted with the Melco via USB seamed OK via the CAD CAT. However, the CAD CAT was running with their new electronic grounding box device. Could this be the reason the CAD CAT appeared better via USB, that maybe only this device with it's grounding box and external LPSU was somehow breaking the problematic 'leakage loop'.

This is currently pure speculation on my part, I do not (yet) know enough about 'leakage loops' to know if this makes any sense. What I can say is that the evidence does appear to fit the theory. Coincidence? Maybe, maybe not?

EDIT: and another thought..... I recall Guillaume rated the Melco N1Z a lot higher than I did. However, I did think some aspects of the Melco's mid-range performance was absolutely outstanding, it was just the bass accuracy I had a bit of an issue with. However, maybe when Guillaume tried the N1Z in his system, with decent mains isolation kit, the 'leakage loop' issue was fully mitigated, so Guillaume would never have heard the bass issue I did? (more wild speculation that kind of fits the theory!)

[Hifi_swlon]

I felt the Intona made a useful (but subtle) difference with the mR. Johns explanation was also possible ground loop from the ifi.

John S has a vested interest in USB enhancers he designs so despite enjoying his posts I can never convince myself I 100% trust what he's saying (besides the fact I don't understand much of it). He has made some claims surrounding the regen for example that he backtracks on a bit on when discussing the rendu and power supplies (like the LPS-).

But given that the rendu was meant to be the ultimate USB source, it does kind of beg the question why not incorporate intona type technology and galvanically isolate the USB, and why sell it with the ifi which does appear to have some sort of noise issue?

It will be interesting to see what swapping ifi for Paul Hynes does. But that's a fairly long way off as Paul does operate on quite a long order time (youd think the stock SR3 would be a good enough seller to not need to be built to order for every customer but to get ahead of the game a bit!).

Still want to try an AES connection comparison vs Intona and rendu to see whether it's purely the USB input weakness as a generic thing, or specific to the Devialet. My hunch is a bit of both but more the latter since other people get seemingly transformational experiences using the rendu with other DACs.

[Antoine]

The Devialet USB input has no real isolation built in as far as I can judge. The Mutec has.

My chain currently is:
HQ 4 rail LPSU (Paul Hynes) with galvanically isolated/floating powerrails
fiber optic ethernet (converter fed clean power from PH) for complete noise isolation from the main network
high quality ethernet cable
microRendu (fed clean power from PH) with galvanic isolation on its ethernet input and providing a USB output with high quality/integrity USB signal
high quality USB cable
Mutec (fed clean power from PH) for reclocking, USB galvanic isolation and USB->AES3 conversion
high quality AES/EBU interlink

See a pattern here? Isolation, isolation, isolation plus high quality signal and power to keep noise and distortion in both the digital and analog domains as low as I can.

[Hifi_swlon]

My chain currently is:
HQ 4 rail LPSU (Paul Hynes) with galvanically isolated/floating powerrails

Antoine, was the galvanic isolation a special request just on your one, or that model, or part of his generic design?
I only ask as I could only justify an SR3 (still £400 when you add silver lead and shipping), but in all our communications I don't remember isolation mentioned specifically for this model....

[Hifi_swlon]

EDIT:  and another thought.....   I recall Guillaume rated the Melco N1Z a lot higher than I did.  However, I did think some aspects of the Melco's mid-range performance was absolutely outstanding, it was just the bass accuracy I had a bit of an issue with.  However, maybe when Guillaume tried the N1Z in his system, with decent mains isolation kit, the 'leakage loop' issue was fully mitigated, so Guillaume would never have heard the bass issue I did?  (more wild speculation that kind of fits the theory!)

You'd think on kit costing what the N1Z does, that bearing in mind their intended use, these things would be covered in the design of the kit rather than relying on external configuration?

[Antoine]

Hi Hifi_swlon, the SR3 is a single rail PSU so it's not applicable for that model. The isolation of the different rails in multi rail models of the SR5 and 7 are part of his generic design.

This is how Paul describes it: "An SR7EHD-MR4 four rail Multirail power supply in an SR7EHD-MR4 silver chassis. All four rails are galvanically isolated from each other and have totally floating outputs so there will be no issues with ground return inter-modulation from common ground loops." I'm not sure if this also tackles possible leakage loops John Swenson describes but I don't think so. Hopefully the additional isolation I provide does.

He goes into more detail abou the different "ground loop" fenomena here:
http://www.computeraudiophile.com/f27-up...post573900
and here:
http://www.computeraudiophile.com/f27-up...post580313

[Antoine]

EDIT:  and another thought.....   I recall Guillaume rated the Melco N1Z a lot higher than I did.  However, I did think some aspects of the Melco's mid-range performance was absolutely outstanding, it was just the bass accuracy I had a bit of an issue with.  However, maybe when Guillaume tried the N1Z in his system, with decent mains isolation kit, the 'leakage loop' issue was fully mitigated, so Guillaume would never have heard the bass issue I did?  (more wild speculation that kind of fits the theory!)

You'd think on kit costing what the N1Z does, that bearing in mind their intended use, these things would be covered in the design of the kit rather than relying on external configuration?

Intona was the first to succesfully implement real USB galvanic isolation for high speed USB (480Mbit/s). The technology didn't exist yet when the Regen was designed and probably the same is true for the Melco's as well. IIRC Uptone is working (or has said they once were) on a new Regen that has galvanic isolation built in. Fortunately we're seeing more and more products employing the galvanic isolation for USB so in time it'll probably become a standard like it has for AES/EBU and S/PDIF. I'm curious if Devialet will implement it on their new streamer board, but i'm pessimistic to be honest.

[Confused]

If all this is true and hangs together, I have to say it is fascinating. OK, I could see the logic of galvanic isolation and similar, but for it to have such a profound effect on something fundamental such as bass accuracy is fascinating.

So is there an explanation here as to why the CAD CAT performed better than the Melco, maybe galvanic isolation via the power supply?

The CAD CAT spec refers to 'Linear power supply with 4 independent power rails.' Would this be enough to make such a fundamental difference?

Fascinating stuff!

[Hifi_swlon]

Intona was the first to succesfully implement real USB galvanic isolation for high speed USB (480Mbit/s). The technology didn't exist yet when the Regen was designed ....

It did at the point of the microRendu design though.
That said having now looked inside the rendu and the Intona, it's obvious who are leagues ahead in circuit design so maybe it's far from trivial for anyone outside that specific field - the Intona board looks pretty complex.

I'm too much of a layman but I can't quite see why the same technology used on an Ethernet connector can't be used on USB? The Intona seems so complex by comparison.