How to improve Jitter issue via USB/ETHERNET ?

[PeppaPig]

Hi All

I am planning to use a Mac MINI connected to my Devialet using USB or Ethernet port.

I read the spec of Devialet, it seems all input goes into Devialet will be first digitized and then controlled by the Devialet internal master clock. So I assume it will has no jitter issue at all or very minimum right?

[thumb5]

Both USB and Ethernet inputs are already digital so there is no digitization as such in the Devialet.  If you're referring to analog inputs, then you might want to ask Devialet about the level of jitter added during digitization; I haven't seen any measurements.

Both USB and Ethernet are packetised, asynchronous protocols so the audio samples are buffered and clocked into the output path (DAC) by the Devialet's internal clock.  In other words I would argue that the jitter will be entirely limited by the Devialet's clock.  Others may disagree although I haven't seen a coherent explanation as to what might cause additional jitter.

Although the signal is "digital" it is represented as analog voltage levels that are compared against thresholds at the receiver (Devialet).  In principle a digital transmission can be affected by noise that might cause bits to change value; in any normal (to-specification) USB or Ethernet cable in a home environment this should be a very rare event if it happens at all.  In the case of both USB and Ethernet this would cause a whole packet to be dropped because the receiver (Devialet) can detect that the packet is errored.  Normally this would result in easily audible drop-outs.

I don't think this is what people are arguing about when they say that one or other kind of USB or Ethernet cable affects sound quality - rather I think they're talking about more subtle things like (for example) how much noise can leak through the cable into the sensitive parts of the DAC.  Now we're into subjective territory -- in other words, if you are concerned about this you have to try a number of cables and see which you prefer.

[Confused]

Both USB and Ethernet inputs are already digital so there is no digitization as such in the Devialet.  If you're referring to analog inputs, then you might want to ask Devialet about the level of jitter added during digitization; I haven't seen any measurements.

Both USB and Ethernet are packetised, asynchronous protocols so the audio samples are buffered and clocked into the output path (DAC) by the Devialet's internal clock.  In other words I would argue that the jitter will be entirely limited by the Devialet's clock.  Others may disagree although I haven't seen a coherent explanation as to what might cause additional jitter.

Although the signal is "digital" it is represented as analog voltage levels that are compared against thresholds at the receiver (Devialet).  In principle a digital transmission can be affected by noise that might cause bits to change value; in any normal (to-specification) USB or Ethernet cable in a home environment this should be a very rare event if it happens at all.  In the case of both USB and Ethernet this would cause a whole packet to be dropped because the receiver (Devialet) can detect that the packet is errored.  Normally this would result in easily audible drop-outs.

I don't think this is what people are arguing about when they say that one or other kind of USB or Ethernet cable affects sound quality - rather I think they're talking about more subtle things like (for example) how much noise can leak through the cable into the sensitive parts of the DAC.  Now we're into subjective territory -- in other words, if you are concerned about this you have to try a number of cables and see which you prefer.

I cannot fault your logic Mr Thumb.  That said, I remain baffled by digital audio.  As an example, last weekend I was comparing the microRendu to the SOtM sMS-200Ultra.  OK, this is a subjective comparison with all the perils that this can entail, expectation bias, poor aural memory, and all the rest of it.  There were, subjectively, many differences between the microRendu and SOtM kit, but one thing that appeared to be very clear is that the SOtM kit was brighter than the microRendu, it quite obviously had a little more treble, something equivalent to tweaking the treble setting on the Devialet by about 2 or 3 dB.

Why this should be utterly baffles me, we are talking about a comparison of two devices that feed 1's and 0's to another device (MC3+USB) that then re-clocks and feeds the same 1's and 0's to a Devialet.  The SOtM kit has a very good reputation for filtering noise, the microRendu less so.  So could having more noise via the microRendu lead to a 'darker' sound, or less noise make the SOtM kit brighter?  Plus, this noise needs to make it though the MC3+USB, which itself has clever isolation on it's USB input.  As I said, utterly baffling!

[ogs]

@Confused I am sure this is some sort of bias on your end! Digital is digital and bits are bits....

[Confused]

@Confused I am sure this is some sort of bias on your end! Digital is digital and bits are bits....

One possibility is that I have gone completely insane.  I would not rule it out.

[thumb5]

Bits are bits: there, I've said it.  But I will wager that has nothing to do with the kind of sound quality differences that @Confused has been talking about recently: I trust @Confused would agree they are subtle and not caused by drop-outs due to bit errors.  Over USB or Ethernet, any errored bit will cause drop-out that should be clearly audible, unless Devialet have deliberately engineered their USB and Ethernet hardware and drivers to do something different.  So if there is a sound quality difference it is not caused by "bits not being bits".  There are plenty of proposed mechanisms for such differences, but I'd say the most plausible are in the analog domain.

For example, could it not be that the sound quality differences are caused by subtle differences in grounding, noise injection, etc. that would vary with the number and type of boxes connected, and the cables between them (more or less independently of what the boxes are doing in the digital domain)?  That would explain, without resort to expectation bias, why there's always a change to be heard when changing equipment.  Just because it's "digital" audio doesn't mean these kind of analogue effects are ruled out -- far from it, they may become relatively more significant because they are not masked by higher levels of noise and distortion that would be present in an analogue signal.

[ogs]

@thumb5 Joking aside, I am sure the feed from both mR and SOtM is bit perfect. The Mutec clearly has no problem receiving from either, but there is still a difference in sound. Maybe it is noise related, maybe not. At some point in the future this will be explained.
I think @Confused is an Audiophile!

[Confused]

Bits are bits: there, I've said it.  But I will wager that has nothing to do with the kind of sound quality differences that @Confused has been talking about recently: I trust @Confused would agree they are subtle and not caused by drop-outs due to bit errors.  Over USB or Ethernet, any errored bit will cause drop-out that should be clearly audible, unless Devialet have deliberately engineered their USB and Ethernet hardware and drivers to do something different.  So if there is a sound quality difference it is not caused by "bits not being bits".  There are plenty of proposed mechanisms for such differences, but I'd say the most plausible are in the analog domain.

For example, could it not be that the sound quality differences are caused by subtle differences in grounding, noise injection, etc. that would vary with the number and type of boxes connected, and the cables between them (more or less independently of what the boxes are doing in the digital domain)?  That would explain, without resort to expectation bias, why there's always a change to be heard when changing equipment.  Just because it's "digital" audio doesn't mean these kind of analogue effects are ruled out -- far from it, they may become relatively more significant because they are not masked by higher levels of noise and distortion that would be present in an analogue signal.

Yes, I would agree that the differences I am referring to are not caused by drop-outs due to bit errors.  Noise, grounding and 'leakage loops' are certainly factors.  Although I would say the difference in treble I mentioned is somewhere in the rage between subtle and easily audible, this is what really fascinates me.  I would say it is at a level that could be measured with REW or similar.  I would try this, but the only way to run REW is via AIR or USB.  If anyone out there knows a straightforward way to run REW via a microRendu, then I would be happy to try this objectively, but it is not possible as far as I am aware.   

So, back to the OP's question about jitter.  The first point I would make is that I do not think 'zero jitter' is possible, even the very best audio clocks in the world produce jitter, albeit at fantastically low levels.  Many would argue that modern well engineered digital audio systems have now reduced jitter to levels that are not audible, so it is no longer an issue.  The Devialet's Ethernet and USB inputs use asynchronous protocols, so in simple terms the Devialet's clock is in charge, it is a very low jitter clock, so this is a non issue.

Others are going beyond this, claiming that decent clocks in audio systems reduce noise, (or bad clocks increase it) and that improving the 'clocking' on a USB feed can improve sound quality.  This is a very controversial area, John Swenson (UptoneAudio / Sonore designer) is currently doing some testing in this area, apparently he has put together some unique test equipment to do this.  He has been doing this for a few months now with no sign of him producing any results.  It should also be remembered that digital audio systems need multiple clocks.

This subject has been debated extensively elsewhere, and often includes the statement that there are effects in digital audio due to mechanisms 'not yet fully understood'.  I tend to think this is true, although it is a rather week statement, after all, the 1's and 0's appear to be more than capable of running aircraft autopilot systems, industrial control systems, the internet, space missions, medical scanners, and this kind of stuff.  it's just digital audio that has not yet fully understood these 1' and 0's.

[thumb5]

Nice posts @ogs and @Confused; I think we're busy agreeing with ourselves even though we might express things differently.

[Confused]

I think @Confused is an Audiophile!

Worry not, I've been called worse....