Stunning USB Finding

[Scroobius]

I recently built an XXHighEnd clone PC. A big advantage is that it fits in a rack immediately above NOS1. A much tidier implementation than having a PC standing on the room carpet. Definitely good WAF.

After getting all settings right with BIOS settings shown in my signature the sound quality has been really excellent.

But I was sitting down last night and to my big surprise the sound quality was very much better again. I mean really improved. More dynamic clear out in the room sound (and I thought it had that before!!!!) with better imaging and cohesion. Something fundamental had to have changed. Then I realised that I had changed out my old 3m USB cable for a 1m cable (both bog standard USB2 cable construction). I changed the cable simply to have a tidier cable run.

So then I removed the 1m USB cable and put the 3m USB cable back in. The sound really changed becoming "grey" less dynamic less clarity. This was a big surprise because always in the past when I have tried different USB cable types and lengths I could not hear any real difference. Something fundamental has changed and I have to wonder what.

No further investigations just yet though too busy listening to music.

But a 25cm USB cable is now on order!!!

Cheers

Paul

[PeterSt]

Paul,

Explanation should be easy, assuming you still use your tuned against each other 24MHz oscillators ...

Keep in mind that we are still only guessing somewhat about the tuning influence and that it is not necessarily so that a same speed does the job (I even tend to vote against that, but this is not important). But what changes for sure with the different length is the timing. Timing difference in your situation will be around 10ns (nano seconds).

Whether this really does tricks depends on something like an initial synchonization between the two oscillator ends, and I don't know whether it exists at all. But think like the one end waking up the other (which doesn't work like that, so just an example) and from there on all remains stable to your tuning; you can see yourself how mentioned 10ns relates to the oscillator cycle period and the drift between the two occurring.

If it makes sense at all (not to you I mean, but in general) then using a standard length 4.5m USB cable will shift the 10ns in the other direction (could be for the better, just try it).

Outside of everything, try to think like I just did, and not that a long cable "of course is for the worse". It doesn't work like that - not with digital (which is analog in the end, but you will get it).

Regards,
Peter

[Scroobius]

Hi Peter,

Yes the "tuned" 24 MHz clocks are still in place.

Makes sense my initial thoughts were that such a marked difference in sound quality would not likely be down to the cable length per se (i.e. RFI etc) but timing does make more sense. Also I did not get any real differences in sound quality when I tried different USB cables before and that was a long time before the Dexa's appeared.

Anyway the difference in sound quality is surprising to say the least - the best way to describe it would be that my speakers have had a "horn" injection.

To say I am pleased with the SQ would be big understatement to so the least. The bass quality is simply superb but then again so is the rest. The smallest changes anywhere in set up now make a very easily heard change in sound quality.

Cheers

Paul

[PeterSt]

Don't underestimate your own increased capabilities of hearing differences Paul. All is just a matter of having the experience for one time, and you will never get rid of it.

Peter

[Scroobius]

Ha good point  -  so I should definitely not listen to Mani's Orello's when he receives them !!!!!

[PeterSt]

Haha, maybe not ...

[Nick]

Peter, Paul hi,

Explanation should be easy, assuming you still use your tuned against each other 24MHz oscillators ...

Keep in mind that we are still only guessing somewhat about the tuning influence and that it is not necessarily so that a same speed does the job (I even tend to vote against that, but this is not important). But what changes for sure with the different length is the timing. Timing difference in your situation will be around 10ns (nano seconds).

I don't think the effect that Paul is describing can be linked to a "static" shift in USB clock or USB data phase of 10ns introduced bt the shorter USB cable.

I agree that the cable length will introduce the static 10ns change in USB signal timing, but what you may not be aware of is that there is already constant dynamic relative phase drift of the two USB clocks and that the levels of drift are of the order of the 10ns static change transmision time.

The dexa clocks when tuned go through a cycle of drift relative to one another (description in the the link below). Pauls clocks behaved identically to this when I set them up at my place.

http://www.phasure.com/index.php?topic=2888.msg30714#msg30714

A good estimate from tuning with my 'scope is that the rate of change of the phase angle between the two 24Mhz USB clocks varies between 0 degrees per second (clocks running at the same speed) to 1800 degrees per second (clocks running out of sync by 5 cycles per second). Each cycle of the 24mhz clocks is 40 nano seconds in length which means that the relative speed (and phase relationship) of tuned dexa clocks is of one clocks is moving constantly from synchronisation positions of 0ns to 200ns then back to 0ns.

The D+ D- USB carrier dynamic phase change as a result of the clock drift "could" be much greater as a result of the clock phase changes. The carrier frequency is 480Mhz so a carrier cycle time is a 1/20th of the USB clock cycle at 2 nano seconds (I know we did not agree on this when we were discussing this data speed, but take a look at the “Signalling” section of the link below).

http://en.wikipedia.org/wiki/USB#USB_2.0

“A high-speed (USB 2.0) rate of 480 Mbit/s was introduced in 2001. All hi-speed devices are capable of falling back to full-bandwidth operation if necessary; i.e., they are backward compatible with USB 1.1. Connectors are identical for USB 2.0 and USB 1.x.”

The point is that there is already constant “change” in relative phase of the clocks that is of the same order as the 10ns timing change introduces by Pauls shorter cable, so I don't think that introduction of a “static” 10ns timing change from the change in cable length is likely to be the cause the Pauls effect.

Now I'm just throwing out ideas here, but could be that the shorter length USB cable somehow happens to have better signal reflection characteristics or lower signal smearing for the USB signal ? or could the signal earth carried by the USB cable from the PC be "better" coupled to the DAC by the shorter run of cable ?

Regards,

Nick.

[PeterSt]

Hey there Nick,

I get what you mean, but it doesn't explain things. So try to think like I do and reason into the shown phenomenon instead of out of it which only leaves something which can't be right.

On the latter : Assumed the cable topology is the same, it would be a rare case of USB cables shoing s difference. So rare that it has not been reported yet. And so I assume that now nothing general can be goig on which should have been noticed before (by any random person). Thus what is the difference with the others (including me of course) ? that tuned oscillators and then especially "almost equally tuned". So when not tuned at all (like it is the situation normally) this wouldn't do a (noticable) thing.

I think you won't be able to do it, but possibly you can after all - but at least try to envision this :

That link (USB-USB) will show noise besides the signal;
If the oscillators are not tuned (hence the drift is fairly large like 5 cycles per second) then all would look quite messy (but you may still see something and with the notice that 5 cycles drift is not much in my view (hence in normal situations can be way larger)). Now :
If you tune the both to your very best and say one cycle drift per minute is possible, you should see "oscillation" in the noise. You'd especially see that because it changes over time and it is the phase angles of both running into eachother (via the impeded noise) that makes you see the oscillation as such.

Of course it is obvious that when 10ns of decreased (or increased) travel time right away cause a 90 degree "opposed" phase angle difference (when the full cycle is 40ns).
As of course this
a. won't last;
b. won't be an absolute state right from the start (although I hinted at something like that - possibly ot true at all).

Still things could be different at microscopic level because at that level it might happen that the current needed to operate the whole scheme influences the oscillators (the rates) themselves. Now I think about the slowing down or speeding up which is about the content of the stream (the more bits at 1 the more current) and how this could somehow mechanically balance out.
Side note : Remember that we looked at the USB noise and maybe you did not notice it (because I did not point it out) but "no streaming" is always still streaming but with "no content" and noise is already visible; make it stream really (put up the control panel) and there the real noise starts to show. Conclusion : the contents implies the less and more current and that may imply speed changes of the clock(s).

Far sought ? yes, very.

But *if* things would balance out somehow - and with the notice that tthis would only happen with not too much tension EQUALS the clocks needing to run at the same speed to begin with, then balancing out could happen.
This does not decently explain how the 10ns shorther cable implies a readily (90 degree) change which *also* stays stable, but if so we can expect the "oscillating noise" to be very different.
Also notice that running the clocks requires current as well which makes it more logical or acceptable that balancing points exist (like the cycle from osc-1 is up while the cycle from osc-2 is down or both up or 90 degree angled or anything that keeps them in balance because actually the noise itself).

So thinking "tension" it will be clear that when the both oscillators run sufficiently out of sequence, that force is too high against the microscopic balancing noise/=current.
And remember, it is you yourself who tuned the two osc's in sequence in the first place which took out the tension to begin with.

Now you can call me crazy.

Would I be right on this then you should be able to see it by means of having the two frequencies (of the both oscs) together in one screen (or the phase angle of them) and the drift should go in hops. Not necessrily 90 degrees or some recognizable number, but in repeating hops anyway. Notice that the hops don't need to go instantly, but that the "more stable time" is recognizable at least.

*If* you see something like this happening, then change the cable length and look again. I am sure it will behave differently then.

And yes, I now realize that you told that the load of the scope influences. But I think you can see through that.

Definitely crazy me.

[Nick]

Hi Peter,

I guess I wanted to highlight that in my and Pauls case (with Dexas) even with the clocks "tuned" there is cycle of change happening in the relative Phase of the clocks. (0 to ~5 beats per second is the best I can manage at the moment, its about bang on the 0.2ppm “precision” spec Dexa quote). The timing shift from the clocks drifting is similar in order of magnitude to the 10ns static timing change from the cable. I just wanted to make sure that there was not an assumption that the clocks did not drift with respect to one another and that the only timing change was the cable length.

So thinking "tension" it will be clear that when the both oscillators run sufficiently out of sequence, that force is too high against the microscopic balancing noise/=current.
And remember, it is you yourself who tuned the two osc's in sequence in the first place which took out the tension to begin with.

My thinking is similar here I think regards the current and noise balance, its interesting that the dac clock running much faster / slower than the PC changes sound quality in different ways. Could this be linked ? I still have the suspicion that the "lock" circuit of the receiving end (via on chip PPL ?) may find it easier to maintain its lock on the incoming data when the clocks are close or for that matter running faster or slower with respect to one another.

Would I be right on this then you should be able to see it by means of having the two frequencies (of the both oscs) together in one screen (or the phase angle of them) and the drift should go in hops. Not necessrily 90 degrees or some recognizable number, but in repeating hops anyway. Notice that the hops don't need to go instantly, but that the "more stable time" is recognizable at least.

The display of relative speed is continuous without “hops”, it can resolve <5 degrees of relative phase shift without having to try to estimate two separate clock wave forms by eye. Its not good at measuring large "beat" rates between the clocks but when the clocks are closely matched (10 beats and less per second) its easy to see the speed of relative phase shift. The neat thing is that the trace can be watched easily whist playing music from the listening position, so clock +- relative movement and speed can be listened to, really nice.

And yes, I now realize that you told that the load of the scope influences. But I think you can see through that.

This concern was just when an attempt is made to measure an Xtal output directly. I'm guessing that the 10pf or so of probe capacitance will change the Xtal resonance point and change the measured frequency. This is why I don’t think its possible to correctly measure / observe the clocks on unmodified PCIe and NOS1 boards. In the case aftermarket clockes with buffered outputs I don’t think the scope probe capacitance should matter.

Regards,

[PeterSt]

Hi again Nick,

Thank you for your extensive elaboration.

its interesting that the dac clock running much faster / slower than the PC changes sound quality in different ways. Could this be linked ?

I read this 10 times but whether a comma is missing or my English can't catch it ... Would you able to rephrase this ?

Anyway all interesting ...
Peter

[Scroobius]

.....and now I am using a 0.25m USB cable and sound quality got better again!!!

[PeterSt]

Maybe that perceived "USB correction speed" now responds faster ? (think in nano seconds). Don't ask me what that would do for net result. But also I don't see the relation with your "tuned" clocks really any more (see Nick's responses). So everybody should benefit from this ?

Anyone tried ? (me not)

Peter

[Scroobius]

A few nights on listening with the 0.25m USB cable installed (and it was very cheap at £1.86 inc delivery!) I am surprised by just how good the sound quality is now. It sounds as though my system has had a big injection of "analogue" maybe as though the amp has been changed for the biggest most expensive tube amp available. The speakers really have disappeared there are sounds popping out all over the place - there are seagulls all over the ceiling (ha ha!).

The sound is so good (well to my humble ears anyway) that I am beginning to wonder if I have done anything else at the same time. A note of caution is that after I finished building the new PC I made a lot of changes and tweaks one after the other (BIOS & OS settings etc.). So maybe this is tied in with other things I have done. But for sure this is a level of sound quality that is well above anything I have heard here before.

Tonight I will try a few things like comparing long versus short USB cables in both the dexa and non dexa positions because I cannot quite believe yet that this is just down to the USB cable.

[PeterSt]

So, Elvis Costello still works as well ?

[Scroobius]

Ha!!!! Elvis sings in person to me any night upon request!

Seriously though - how good do bad recordings sound now!!! I don't know how that works but for sure it is very welcome.

There is one thing though if I change any one of a number of things (like switch on page filing etc.) it really is noticeable that something is wrong.

So what will happen if the SM PS was replaced for linear? I understand one will be launched soon by a UK supplier - it will be interesting to find out.

Paul