Anthony,
A bit difficult to answer;
If this is measured "rawly" on the USB cable your 0.5mV will probably show something but you won't be able to recognize anything of it. And, maybe in relation to another USB connection in the PC you may find a difference, but you still won't be knowing what you are looking at and what's the reference. Additionally it won't tell much because it is all how the NOS1 "rejects" it and that now is related to so many things that I can't predict it. For example, your PPAStudio should not behave differently from any other USB connection (but still it can) while I expect the Dexa solution to behave differently in the base. This is all about how the PSU relates and how new groundloops (which can be for the better !) emerge and that kind of stuff.
If you'd measure at the end and how (outside of jitter) the analogue signal comes out (the NOS!) your 0.5mV (per vertical section that is) will never show a thing and if it does, things will be not the best.
And then still you can't do much because it really needs an FFT to observe what is needed (frequencies) and if *that* shows up at the NOS1 outputs you are dealing with something like 300uV (normal is 8uV RMS).
It will be a bit vague, but if things are all OK any USB noise (internally) will be under that 8uV, while still sound is able to change drastically already by software.
Summarized : Only when we (me too) are able to dig out some frequency which normally is not there, it will be audible for sure and for the worst. I can't see a thing changing with any type of USB connection, while for example USB2 and USB3 already make a very noticeable difference.
I hope that this isn't going offtopic, but if it were for me these SQ changes are ALL because of software behavior and are otherwise (or together with) incurred for by the general current draw (say power usage) which changes drastically.
With this new knowledge I was able to create yesterday a "sound" which can plainly be seen as an I-don't-know-what-kind-off equalizer/filter which again makes your albums unrecognizable, shows the most sweet highs with a fairly high degree of lacking snap - is close up to zooming into standing waves (bass blooms more) and which yesterday played at 90dBSPL for 3 hours before finally someone told me "isn't this loud ?".
What I tried to tell with the above is that at this moment I am rather explicitly working on the electrical influence on current spikes and spreading them which in my view is similar to anything which is injecting noise (more frequencies) which spreads the peaks just the same. One difference : with the latter the "RMS" (read : average) noise level will go up, while my software means brings it down. So what I am doing (trying to do) is with some sort of Ohms law (but better : "energy never gets lost") utilizing the current peaks for the gaps in between them. Ok, this should not bring down RMS noise, but at least it won't go up as well (and spikes vanish).
That something like this is working out, is audible by the Q5=1 setting and that this really can be overdone prooves my super equalizer from yesterday.
So to quote a bit what Coen was suggesting "we now need hardware solutions" ... no. I am far from finished. Actually just started in this new dimension although it was already sneaked in in 0.9z-9 (Q5).
I am also now regularly thinking how crazy it is that a version from half a year old can today bring so drastical SQ changes if only someone tries "that" combination.
Yes, this was offtopic. But I really must add that my first USB3 PCIe card was destroyed with a 24.xxx oscillator and I never listened to the merits of it (always used the MoBo USB3) and that the second (way more expensive) card never found its way to the PCIe bus because I personally and 100% think that *I* am not up to that at all. Thus, as long as software needs to be explored way more and can lead to so much better suddenly, why would I (anyone ?) try to improve with hardware. And I am serious : this hardware is more out of (my) control than the software plus the software can influence the hardware (mind you, also how your disk behaves, just saying).
Yes, this is quite another post than my first in this topic.
So I must be honest :
My previous post was merely and finally filfilling up to recognizing a two guys needs on very long and hard and $$ work on improvement. This post more honestly shows what I am really thinking and that -for example- there is no single way that I am able to judge in my current new means how all the *other* software parameters can influence this particular setting. Arjan pointed out the 5ms ClockRes instead of my 0.5ms. I tried his 5ms (with another DAC of course) and it didn't work out. Still it brought me back to 1ms which made me scratch my head at thinking "yea, start all over - takes a year".
With yesterday's new application I found the sound very similar to W8.1 and that again made me think that I may have found some ugly W8.1 base to possibly undo and proceed with that again.
Last notice and to be on-topic again :
If it were for me, not any low jitter USB clock can improve on jitter with asynchronous USB. But and as I said elsewhere, maybe it can after all. It is, however, a 100 times more likely that it is just that other noise pattern doing this (what about very low oscillation because of two now very closely running oscillators at both ends). So, it sure all is about jitter, but the means to it (better pattern up to lower jitter) should be very indirect and should not be about low jitter oscillators in the first place (but mind the low oscillation because then it is after all - hence it is more direct).
But if it works, it works. Sadly also : if it works, what happens with one of the drastical changing software settings.
All together (and offtopic again) I don't think I am up to hardware changes - if that was not clear already. And you know, for me it is really difficult to even stuff in a PCIe USB3 card when I don't see any need anywhere. Not as long as I know how much just putting in one additonal disk makes a difference and my perception of how the last one to remove should do jobs. Then I rather work for a year on the latter than spend 3 minutes on the former, in the end also knowing that all of it (whatever it is) in the end hammers on this interface.
Wow, I talk too long / too much.
Peter
PS: Paul, Nick - still invited. But know what you're up to.
