USB audio cracked... finally!

[jabbr]

1 hour ago, manisandher said:

By design, the Lush does not follow the USB 2.0 spec, it follows Peter's own 'USB audio spec'. And this 'USB audio spec' was created not in an ad-hoc manner, but from working from first principles. It is my strong belief that it is the adherence of the Lush to Peter's new 'USB audio spec' that gives it its unique musicality. [And I think 'unique' is the correct word here, because I doubt any other USB cable sounds like it.]

 

Now that I've had this level of musicality in my system for the very first time since switching to USB audio, I couldn't live with anything less musical. And I suspect all USB cables/connectors adhering to the USB 2.0 spec would be.

 

Hi Mani, really interesting. What is the "USB audio spec?" Is 90 Ohm impedance not optimal? Or is it something else?

 

[jabbr]

1 hour ago, manisandher said:

It's funny, only 3 weeks ago I wrote the following on the Phasure forum: "I'm growing a real hatred for all things USB. It's like we're all being held hostage by its whims. And unfortunately, it has us by the balls..."

 

It is perplexing that despite purported improvements in USB isolation there are still ?dramatic? differences in the "sound" of USB cables? I'd think that the goal of USB isolation/reclocking/regeneration is to eliminate the vagarities of cable impedance etc. @PeterSt : I would really like a technical explanation if feasible.

[jabbr]

57 minutes ago, Dev said:

 

Good question but even with external isolators & reclockers, like Intona and ISO-Regen, the USB cable matters. Folks have gone as far as saying that the uspcb before and after makes dramatic differences. If it matters for external boxes, why won't it matter when these functions are implemented inside the DAC ?

Again, I'd like to hear the details but I am concerned that this implies that the USB protocol stack is not optimal for audio as designed. We now are talking about a protocol that looks like USB but with a different PHY layer. 

 

My understanding of USPCB is that great care is taken to conform to the USB specification.

[jabbr]

1 hour ago, PeterSt said:

No, why or how ? What would be true though is that an isolation device may act as a buffer (as an active device) between the start of the cable (say at the PC port) and the USB receiver and that cable reflections may be consistent over DAC implementations (read : must always be the same DAC). So if you think like this, another such a device at the PC end could make the reflections consistent over isolation devices (beginning and end of the cable provided by the same manufacturer). All it now further takes is a cable also provided by that manufacturer and the interface between the isolators is under control. Assumed that the manufacturer can measure reflections of course and that he is able to control your DAC for received amount of data.

 

As you know the reason to impedance match the cable as well as to properly terminate the receiver is to prevent reflections. Indeed that's how you know your impedance is matched!

 

Regarding skew, indeed different than phase noise but generally lumped in with "jitter". There are also different types of skew. eg speed of bit/clock transition vs different time of signal arrival.

 

Very interesting... if you look at the clock  optimized version of the DSC1 you note that 1) clock distribution chip (potato semi has low pin-pin skew) as well as 2) equal trace lengths to the shift registers.

 

Balancing tends to reduce ground plane switching noise. Also careful bypass cap networks are essential -- any one bypass cap will have impedance peaks.

 

Also dealing with these issues in FPGA or chip is black magic and so ther are undoubtedly many receiver circuit dependencies.

 

But yes there are a class of situations where introduction of skew on purpose leads to less noise!

 

I think these are relevant and hopefully readable: 

1) http://web.stanford.edu/class/archive/ee/ee371/ee371.1066/lectures/lect_07.2up.pdf

2)  http://pages.hmc.edu/harris/class/e158/04/lect19.pdf

 

See the discussion of pros & cons of removing "hard edges"

[jabbr]

1 hour ago, marce said:

90 ohm differential impedance does not equate to 45 ohm single ended... sorry just dosen't work out that way. I have Eric Bogatins books, Howard Johnsons books, even the full tome by Lee Ritchley

Yes. as you say, detailed in Howard Johnson's "Black Magic" and "Advanced Black Magic" books on high speed digital design  There are two issues:

1) differential pair impedance

2) termination  impedance -- (can be to ground)

 

https:\\www.micron.com\~\media\documents\products\technical-note\dram\tn4606_point_to_point-termination.pdf

Note Figure 20, as the rise time of the transition slows the SI improves ... this is an example of removing "hard edges".

 

https://fenix.tecnico.ulisboa.pt/downloadFile/3779579943588/[Bus Terminations] Ethirajan and Nemec - Termination Techniques for High Speed Buses.pdf -- see Figure 4 for use of capacitance along with resistance.

 

http://www.ti.com/lit/an/snla043/snla043.pdf gives an extensive discussion of the effects of and use of termination capacitance.

 

I wouldn't at all be surprised if audio designers are not using high speed digital design techniques -- who knows what termination is being used, but I wouldn't be surprised if at best resistive and as demonstrated in the Micron paper I referenced above, it is very possible that this has deleterious effects on the USB SI at the receiver.

 

I have no idea how the "Lush" actually works, but it might include its own termination, or as also referenced in the above papers, the transmission line properties may have similar effects.

 

@PeterSt -- the above papers detail what I was saying regarding capacitive termination -- all of this would need to take into account the cable impedance (and transmission line model) itself.

[jabbr]

5 minutes ago, mmerrill99 said:

But seeing as you are averse to what he says in that article (or didn't read it), maybe this one, again from Lee Ritchie, will more directly address your issues "Differential Signaling Doesn’t Require Differential Impedance"

 

 

Single ended impedance on a differential pair requires a common ground between transmitter and receiver  

[jabbr]

17 minutes ago, mmerrill99 said:

As far as I can see there is common ground between transmitter & receiver ala the grounding of the circuit, not necessarily a specific ground wire.

 

Did you read the Lee Ricthie articles I linked to & looked at the schematics in them?

 

That's why I linked to them so that these questions could be seen to be answered in the article without bringing such an esoteric topic here which will bore the pants off most  readers!

He is discussing PCB and agreed that it is often desirable not to route differential signals together e.g the + and - of a balanced signal have go to different physical areas to be processed.

 

The 1/2 impedance desires where a "twinlead" transmission line e.g. Ethernet meets a PCB. Each line is terminated and then go on to PCB traces. You could have twin coax but that's not how Ethernet (nor USB I presume) is specified. The spec is differential impedance 100 (and 90) Ohm.

[jabbr]

5 minutes ago, Speed Racer said:

 

I still don't get this. Look here:

 

 

How does the noise get past the magnetic transformer?

 

Those little transformers have parasitic capacitance > 0. a.k.a. "leakage current"

[jabbr]

32 minutes ago, Jud said:

Trafo isolation sounds like Ethernet.  Except that (Ethernet built-in transformer isolation) doesn’t always work completely to stop noise, as I found out in my system.

 

There's also GMR which really is very similar to transformer: https://www.nve.com/Downloads/galv.pdf

[jabbr]

2 hours ago, The Computer Audiophile said:

Yeah, should be interesting. I thought it was appropriate to post in this thread because it's right up the alley of this discussion. 

 

Looks similar to the DIYINHK USB interface which is well regarded: https://hifiduino.wordpress.com/2013/04/03/c-media-cm6631a-based-usb-i2s-interface/

 

Not sure anything earth shattering here ... the C-Media CM6631a USB interface has been out for awhile.

[jabbr]

8 minutes ago, The Computer Audiophile said:

 

Good thing that components in the hands of good designers can turn into products that are much better than the sum of their parts :~)

 

That's what they claim  In any case the DIYINHK version is $40 but doesn't have the isolation. I hope that we will continue to see more effort given to DAC interface isolation and good engineering practices (none of this is too esoteric). Schiit does a great job of keeping their costs reasonable.

 

You'd be surprised at how many high-end DACs use $40-50 USB interface boards ... time for some value add, particularly at the higher end

[jabbr]

7 hours ago, PeterSt said:

 

Then you are looking at the wrong DIYINHK version. 

The one costing $109 sure has isolation. And much more than the Schiit offers. Like 1024 DSD support. Hahaha. Oops.

 

But it is still USB. Who wants that. 

 

Yes that's correct. But of course $60 for a bare board translates to ?3x ?10x to the end customer?

 

DIHINHK has very good prices.

 

Ha, USB, well maybe if you've figured out the issues, then Ethernet won't be better ... although Fiber does get rid of the capacitive coupling issues to some extent

 

[jabbr]

5 minutes ago, marce said:

We use signal integrity software (as well as physical checking of waveforms) and only for the hard interfaces, if USB audio is that hard why do I see NO evidence of proper signal integrity engineering...

 

Most DAC manufacturers outsource their $50 interface or depend on $5-10 SoC to do everything...

[jabbr]

1 hour ago, marce said:

Yes because its a well understood and used serial interface and the devices are there where its all on chip so better for everyone.

Some of this is tickling me, I really do have some views on why some like to promote USB audio as terrible and have solutions for this...

Over and out.

 

Ground loops aren't an issue between your mouse and keyboard and hard drive  ... maybe 4K monitor? ... I got started down this track when I installed a new GPU card and 4K monitor on my workstation and my Sound Devices USBPRE2 ... which I use for recording ... let out a howl... tried a few things that didn't work ... very easiest solution was to remote the music to another room over Ethernet.

 

So I guess the issue with noise and USB DAC wasn't solved after all despite all these decades of understanding

 

The question is why? I am suggesting that its not USB per se, rather the common implementations... but nonetheless I've personally experienced noise on the USB interface, and other people hear differences depending on a number of factors and there is an industry of devices that provide USB isolation/regeneration etc. I am suggesting that these are properly placed in the DAC itself.

 

Oh and the USBPRE2 works just great with my MacBook so... hmm...

[jabbr]

18 minutes ago, marce said:

That is noise not the interface though and I agree there is a lot of it about these days.

 

The interface, by requiring a 5V and ground at least for handshaking, goes out of its way to beg for ground loops/leakage currents -- a.k.a capacitive coupling  and that's how the noise gets into the system... trying to decouple the power/ground starts people down a rathole of solutions... let's see the first useless product I got was the Aqvox (didn't work for me) -- at the time I didn't understand why this was actually much worse ... so I've been reading  ... and so folks can do trial and error of multitudes of "solutions" which is what we are seeing ...

 

As you note, the common isolation techniques have issues with capacitive coupling. Breaking the ground loop/leakage current issues are not entirely obvious. Other interfaces have their own issues.

 

I do think that if proper design and techniques which may be obvious to you, were widely implemented, that these problems can be mitigated.

[jabbr]

14 minutes ago, marce said:

Thinking about your last comment I revisited this and some other notes...

http://www.x2y.com/filters/TechDay09kr_hpa_Track2_1_Precision_Analog_Designs_Demand_GoodPCBLayouts _JohnWu.pdf

 

 

Classic. Try this: how many circuit boards use X2Y caps, or low impedance caps, or bypass capacitor networks?

 

ala: http://www.intersil.com/content/dam/Intersil/documents/an13/an1325.pdf

http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf

 

If there's just a 0.1 uF, the decoupling is suboptimal...

[jabbr]

4 hours ago, PeterSt said:

And nothing sounds the same wherever you can change elements or settings like buffer sizes etc.

 

Past experience does not always predict future experience  

 

I view these types of idiosynchronies (as well as digital cable dependencies) as symptoms rather than necessities -- these behaviors are not intrinsic to digital design nor DAC.

 

One solution is to alter buffer size, another is to track down and eliminate buffer size dependency

 

Digital switching noise and precise sequence of digital transitions are a perfect example. Careful attention to digital ground & power planes is needed. If we are switching in the 40-480 MHz range, we need to design in the Ghz range in order to tame the "harmonics".

 

On topic, the "Lush" cable I am going to predict, not knowing anything about the design nor having seen a measurement I predict will reduce Ghz harmonics and increase the USB transition time to > 1 nS (the signal will look more sine wavish).

 

Bypass caps like: https://www.johansontechnology.com/S42E and or networks of bypass caps properly placed on the ground plane should help to ameliorate such issues. (the digital gates can properly operate at higher rates)

[jabbr]

3 hours ago, PeterSt said:

The equipment needed for this is quite unaffordable for us poor guys. Maybe you disagree ?

 

Nah ... nowadays with eBay it's possible to pick up some really fantastic old HP gear for hardly any money. If you are willing to repair an old LPS you can get e.g HP 8662a frequency generator for $200 (and $100 shipping) or working for a few hundred more. Or HP 3561a/3562a spectrum analyzer for a few hundred, and then there is phase measurement interfaces. 

[jabbr]

19 minutes ago, PeterSt said:

 

Now you are too funny.

So what I design in, you design out ?

If you order a Lush I will send you a Clairixa. Saves you the trouble !

 

haha

 

You are designing in the real world, and as I referenced way above, tuning the rise time on the bit transition can go a long way to improve SI (I am assuming this is one of the things Lush does).

 

What I am saying is that a different USB interface design may not have this dependence (also referenced above). Also what I've said is based on the theory that this is digital switching noise, or that the bit transitions don't have time to "recover" as you said above... digital switching noise can certainly be tamed (bypass cap network on ground and power planes).

 

In any case since I don't know how each cable is exactly designed this is all guesswork. If you give me the exact specs, I'll know which one to order

 

[jabbr]

40 minutes ago, jabbr said:

If you are willing to repair an old LPS you can get e.g HP 8662a frequency generator

e.g. http://www.simonsdialogs.com/2016/03/hp-8663a-signal-generator-another-power-supply-repair/

 

the thing about this "old school" equipment is that it is repairable...

[jabbr]

2 hours ago, marce said:

Do you mean rise time of the signal... cant be increased only reduced.

 

 

"Rise time" as synonymous with "Transition time" -- time taken for voltage to rise from 0v to Vcc or time taken to transition from low to high -- we are saying the same thing (I think)

 

2 hours ago, marce said:

NP0's are too large for most decoupling duties, X7R's are a better solution, especially when used in decade groups.

 

I was using that for the picture, it's a low ESR geometry where the pad width and inter pad spacing are inverted -- this can be replaced with several smaller eg 0402s or 0603s in series. My point was that if you see this geometry or X2Y or bypass cap networks then there's the idea that the designer is attempting to deal with digital switching noise in a more proper fashion. ... otoh if only single 0402 is being used then the designer doesn't appear to be as concerned about high frequency switching noise.

 

This was just a suggestion at simple ways to look at a PCB and see where design effort is being made. @marce I'm sure would simply eyeball the board and come to these conclusions in a second or two  

[jabbr]

3 minutes ago, manisandher said:

Back to the technical discussion...

haha well there have been numerous discussions about how USB cables sound but this could be the first which discusses an actual mechanism by which cable design might affect improve SI -- kudos to @PeterSt 

 

I don't think decreasing the slope of the transition aka transition time is outside of the USB spec, so moreover a physical change in a cable which is within spec and changes SI -- this could be historic  

[jabbr]

1 hour ago, marce said:

If you have ringing on your signal line then there is too much energy going into the line, to much drive strength, this can be solved by either reducing the rise time to limit the signals bandwidth (Knee Frequency= 0.35/Rt where Rt is the 10-90 rise time of the signal) with a series resistor as close to the driver gate pin out as possible, or parallel termination at the end of the line to absorb the excess energy.

 

Thanks, you've got waay more experience in this than me... I'm still (always) learning. Would using parallel termination with a capacitor or including a capacitor be an alternate? I've seen that so wondering.

[jabbr]

7 minutes ago, Jud said:

 

@PeterSt didn’t confirm that it was within spec, necessarily, did he?  Or did I miss it?

 

@PeterSt hasn't confirmed anything we are guessing from little breadcrumbs  

 

I'm focussing on isolating the DSD/I2S/PCM signal after the interface and as @marce says, FPGA design gets knee deep into these issues. The FPGA basically can accept a USB in or Ethernet in on its I/O lines, so similar issues. 

 

[jabbr]

15 minutes ago, marce said:

Jabbr, if you terminate with just a resistor only you reduce the level the wave will ultimately reach due to the potential divider effect of the resistor... usually to reduce ringing you use a resistor down to ground. I prefer ac termination, where you use a resistor in series with a cap to ground,  this keeps the levels intact, but will reduce the ringing still. The best source of info for this is chapter 6 of this:

https://www.amazon.co.uk/d/cka/High-Speed-Digital-Design-Prentice-Modern-Semiconductor/0133957241

 

I have both books, yes. Up closer to the beginning of the thread I had suggested that using some capacitance in the USB termination would have the effect of increasing the rise time (I've used Transition time because people often replace increasing rise time with increasing rise speed and it's the opposite  

 

@PeterSt -- I think that if a designer has a choice of the interface , one could fold many transmission line effects of a cable into the receiver itself eg you might have altered the Phisolator's termination rather than introduce a cable -- obviously for the majority of situations one doesn't have the choice of altering the receiver so I can see justification for building specific transmission line characteristics into the cable.