USB Clock Upgrade My Experiences (so far)

[acg]

Hi Peter,

I've been trying to figure out how to make a good post about digital jitter and attenuation but I am having trouble getting things done on this iPad (am travelling).  Here is a link to an article instead.  

My key take home point is that the higher frequencies attenuate more easily and are more prone to jitter, which means that using a faster bus (eg. Sata 3 compared to sata 2) must run a higher frequency and is therefore more prone to jitter (but not the kind of jitter that is not totally eliminated before the dac).  How much extra noise is created by digital jitter through resends and the the like along with the size and frequencies of this noise are the pertinent questions.  Can this noise actually become audible?

Sorry for the lazy reply, but I have been trying to get this reply done for nearly an hour...damn iPad.

Anthony

Hey Anthony, thank you for that effort;

It is a bit tough for me to decently reply to it, because it seems that you are on the right track with your perception of necessary recends (because of a higher transfer frequency ? <- I don't think so) while using an in my view unrelated article. All I can do is repeat my post about this, which doesn't seem much useful.

That article discusses jitter in a different fashion, because it compares the integrity of a signal (with itself) through an analyser which can be connected the wrong way. So, envsion the original signal (virtually on that scope) while it is compared with a degraded signal because of cable reflections and all mentioned there; It would show a false result only because of how the analyser is hooked up.
"Jitter" there is just used as delay, which in the end is a form of jitter.

Only when jitter gets that bad that not only the sample is captured too late (like for audio this will matter audibly) but that it is completely missed (crucial for your data send over the internet etc,), you can call jitter important up to devistating (but resends can recover the good signal assumed the errorneous receipt can be trapped); So this "digital jitter" (I'd like to call our audio phenomonen analog jitter) is quite crucial when very long distances are in order and it becomes easy miss complete "samples" because the jitter gets too bad.
Missed sample in THIS case : "Sampler" just reads signal, but signal is low again because too late while when it would have been in time it would have read a high (this is not 100% correct but gives the idea, especially emphasizing that such a thing NEVER will happen with audio <- haha  So, people often think that plain errors slip through and therefore wrong valued samples and such, but for audio it is not about that; only timing (when is this always good value put out up to the femto second). This not to be confused with my little subject of skipped samples because the OS thought it better cook an egg first.)

Peter

Hi Peter,

I have a little more time now.  For our purposes digital jitter is the jitter in any one of a number of locations _within the pc_ and does not have anything _directly_ to do with jitter in the dac.  As summarised in that paper that I linked to, jitter and attenuation in digital signals is influenced by the length and characteristics of the conduit that the signal must pass, the shield properties of the conduit, and the frequency at which the bus operates.  In the example that I gave a sata cable is able to alter the length of time that a computer takes to do a task such as transfer a file or boot.  The better the cable is shielded the less rfi influences jitter, the shorter the cable the less attenuation influences jitter, the more suitable the construction of the wires the less impact on jitter. 

Now the sata cable is just one part of one chain of asynchronous communication in the computer...there are other things in each of those chains that may be able to influence digital jitter such as the clocks that have been Nick and Paul's endeavour.  Nick and Paul's clocks are in the final stage (USB card) of the computer as well as in the dac USB input which is probably the link that would have the most potential for improvement in sq IF there is the room for an audible improvement.  Now, we have digital jitter occurring in a number of places within the computer and this jitter is solved in each of those chains before the data is passed to the next chain, so the jitter itself is never going to be a problem for audio BUT the processing overhead of the rate of resends may possibly be a problem and may become audible through rfi (airborne) and noise on the grounds IF the rate of resends can be substantially improved.  Like I said in the earlier post...maybe...if the digital jitter and attenuation is fine tuned to a level that is significantly lower than the default situation (which may not even be problematic in the first place), then perhaps there is some scope for an improved sq.  Maybe.

Anthony

[PeterSt]

Ok Anthony, I can't see any "duality" anymore in your last post.


...


Hey, I had a longer post here, but I feel out of shape today. So all I left in was above sentence, meaning that when you look at it from that consistent angle (your post), you most probably will be correct that it will be helpful.
But also out of control and one of my scratched lines said that I personally will be able to find a needle in a haystack if only it is guaranteed that the needle is in there or otherwise I won't even start looking.

I'll leave it to this, waiting for a better day.
Peter

[PeterSt]

Nick,

Because to me all feels more like data loss than jitter influence ...

I don't know whether you have the equipment to do this :

Hook up that equipment to anywhere in the DAC hence at the end of the USB interface;
Capture a played track which probably needs to be very short so all can be captured in a one file to examine later (but if your capture can be infinitely large, a larger track is also good of course - even better).

- Start that playback and let the track die out;
- Trim it to the first visible music data at both ends (so no silence data is in there anymore);
- Make two runs of this, one with your sound "1" and one with your sound "3" (or anything, if only well audibly different).
- Compare the two captures for their length.

The thing you could be dealing with for false alarms is retransmissions, so I think the best would be to capture the i2s Data line (just one of the two).

In the mean time you could also try capture USB data itself, in front of the receiver chip, so you could proove retransmissions. Notice that this can be tricky for optimal results because you might tend to just observe raw USB data which comes in packets of fixed format and don't ask me at this time whether such a packed can be half full with real data, a header denoting that (and if so you won't be able to deal with it, unless a real USB logic analyser - and still).
So at first try to avoid this path (I think).

All obviously meaning :
When you can proove that especially first mentioned setup (capture i2s data) leads to the exact same length of data while the sound is totally different, then at least we can put that idea to the garbage can. So you know, this at least *is* something which can be physically checked for; how different patterns of noise influence jitter is quite something else for reliability testing.

No obligations of course, but the thing is damn important. So just look at this, when samples are discarded :

If so, ANYTHING done in the PC will matter; we can go as wild as we like. That the real solution is to get ourselves another means of playback is something else.

All 'n all this should have the priority because to me it is totally obvious that when data loss is in order all the remainder is completely moot.

Regards,
Peter

Do I sense it correctly that nothing will be happening with this ? I mean, silence is not a good sign in this case.

Peter

[Nick]

Hi Peter,

Im in a push to sign agreements before Christmas at work so iv not had chance to read through this weeks posts in detail.

It would be great to compare i2s data to wav and I can proberbly capture something. The issue may be that my data analyser can only reliably capture to 4x upsampling rates. For the usb connection this may be fine as that will still run at 480mhz and errors may still occur. To understand transfer problems in the pc it would proberbly need ability to capture at 16 times. This is mased on your post above where you mention that data could be dropped at points in the os due to resources of the pc not being ready.

Ill think about how to set the test up for x4 upsampling rates.

Regards.
Nick     

[PeterSt]

Ah, that makes me happy Nick (also see next post which I almost finished by now).

If what I describe in there is audible at 4x just the same, or if your own means of trimming is audible through 4x only, 4x should be enough, right ?

Super ...
Peter

[PeterSt]

Nick, there is another crucial thing which I suddenly think of :

You will now about my so many times being the most doubtful how in the world these vast changes in SQ (it really goes mad by now) can be bit perfect, right ? (and that digital loopback to me would not prooce that because it takes (or at least could take) another more direct path);

Well, assumed that at some stage you will be able to trim off the leads-in and leads-out of the capture, save that back into a file ... why not compare the both once you find them of equal length. They should be totally equal, not only for length ...

Which btw also makes me think about the USB data, which will easily be able to be different of length because you won't be in control really of which USB data goes in which packets and which relates to the timings of both DAC clock and OS. So I shouldn't even attempt that I think. i2s yes.

Regards,
Peter

[PeterSt]

In the mean time, day before yesterday I was as far as being sure to have the "fastest system" ever. How to explain it ...

In Dutch I would have said " so enourmously 'rete snel' " and it will point out exactly what I mean to say. In English I don't know how to put that. But something like "beyond warp speed". But what would that be, hence how does it express ?

For starters, I tried to apply what was merely Nick's finding about "trimming" with Paul's trimming of the USB clocks as the base (for this thinking). Notice that in my case this is all about software only, BUT I used the Silverstone for the "SQ base" because for me (here) this is crucial.
The actual trimming is in there, but I can use it to the extreme only. A bit difficult to explain, but think like a theoretical best setting which indeed is theory only, and with that setting I may receive the best SQ. May, because in between that and a next too rough step should be more settings, which need further development (enabling those steps to be smaller) with the knowledgde that this is theoretically not possible. But maybe I can find something after all.

So, how does music sound which is subject to warp speed ? I'll give it a try :

It is explicitly NOT about super transients and this is the first to occur. What it is about is a "glide up" towards any transient which occurs within say less than 1 ms, while previously that "glide" comprised of maybe 5 steps. Now try to compare these 5 steps with a perceived unlimited with mentioned "glide".
The 5 steps sound rough and hard (I did not say harsh per se) while the infinite number of steps (hence glide) just is not able to show and hardness. This, while the transient is still the same transient, were it about the time needed for its development (from zero to max). Mind you, I talk about a development (or envelope) of 1ms, indicating how super short this is anyway; maybe it is even shorter, maybe it is way longer (like 10ms). But anyway very short.

The sort of sadness which is in my mind, is that this highly depends on the Silverstone card to begin with. So, that firstly allows for the "separation" I talked about (on/off working faster hence better), while by that this seems to imply that in the gaps again "music" is missing. Remember by best example of the "gaps" : a tambourine like we perceive it in live fashion. So first the Silverstone (by accident ?) allows for this separation, while next we need more resolution which I could enable by software.

If I now listen to Fireball or Who do we think we are (Deep Purple) I just can not imagine/envision how any real live sound can improve that (son performing is drumming practice upstairs right through the music).

Another description which keeps floating throug my mind during listening : a total freshness while "fresh" now is a totally wrong phenomenon. Fresh (in my English) is equal to snappy and a overal glare (positively meant) of highs. One problem : today the glare has gone, which lets excel the snappyness which now also can sing. Just take that tambourine again, and try to envision that the hit on a snare is now able to make "bell sound" (bell - sing). And I don't mean false harmonics or other distortion, but if now so many more over tones are shown that wherever the bell sound is, it will now let hear that (and NONE of that before).

Maybe we can envision that a rougher sounding sound is undistorted fresh, while the same sound with added resolution will make a tone instead of sounding fresh. So, freshness gone, reality instead.

Another means of explaining it comes to my mind :
I think we all know about the ride cymbal, and that at least that can be nicely audible because it's the only cymbal which definitely sounds for real and singing when used as "ride" (which means hitting it in the middle and not at the edges). It becomes profound hence tunable (for the right sound) in tracks like No No No (Fireball). So, we know what I mean. But what changed here ? That Ian Pace appearently likes to hit that ride cymbal right in its middle all over the place throughout any more wild drumming. So what happens ? the freshness of the cymbals which never was any real good (because not real enough) has vanished, and the singing of the ride cymbal now gets profound.
This is how I mean that fresh actually has become more fresh without the hardness of it. Things start to sing all over and in the mean time there an additional dimension because you suddenly "see" things (like Ian Pace moving around as a wild man to each time find that ride cymbal).

The major contribution of latter description will be the Silverstone card because without the separation nothing will work. The injected resolution becomes profound in the real fast transients like synthesizers can do that; it is easily audible how "hard" the music gets when my little trimpot is in another position. This is undistorted by itself (blame the NOS1) but it is hard. Or cold if you want. Now put my trimpot in that best found position and actually all highs disappear. But what comes back is that warp speed now transient with colour and which has become meaningful. Not a tick, but a sound. I could also say : nothing is squared anymore at the somewhat lower frequency level; it has all become rounded and square for square you hear that and what you get from it is the warp speed.

I know, this will come across as contradictionary, but when the squares are there I'd say "hey, what a super system that it can follow that so well !", while the injected resolution tells me "warp speed - wow, how is it done !?".
So, squares are not squares (they really are not because they can't exist in Redbook at all), but will sound like that when not enough samples are present to render the (thus) somewhat more sine. Skipped samples theory again ?

Bass sounds completely different too. More rounded and with more varying colour. No standing waves, no zooming, so should not be wrong. Must listen more for the real merits of this.


Right. I think I have prooven that this trimming on the inside of the hardware indeed does a couple of things, but the more I explicitly work on things like this with the goal of what Nick has done litteraly with hardware, the more I should be focusing on what the real thing is we influence because it will make it better again. This time my focus has been "skipped samples" and all I do leads to indeed that being the culprit.
*If* it could be so that it is "just" about this, then creating the best SQ will be a piece of cake; just turn some dials and look for the most samples to arrive at the other end.
But I'm afraid it is noise we're dealing with after all and which very well can be because my approach about the skipped samples is coincidentally or a super high noise creator or a super low one (but it should be the latter).

Peter

[Nick]

Peter hi.

Your post on transient response above is really interesting, I get the sense that we are hearing very similar probably the same effects.

In my post on consistently tuning the PC  (moved to this thread above) in order to kept  things simple i gave only basic descriptions of the "sharp", "smooth", and "sweet spot" sound characteristics. This was to place emphasis in the post on the consistency of results obtained from all of the tuned points in the pc.

Your post brilliantly describes some of the qualities of the "sweet spot " sound and the "sharp" sound I hear here. Particularly you descriptions of transients seeming to be less "square" in nature but actually much more true to life. The fresh / sharp sound appears to have speed and attack but its a "hifi" sound, not a real and true to life. At the sweet spot you can hear, almost see, the "nature" material the instrument from and way for instance cymbals are being played. Also your  descriptions of bass reproduction is ringing bells (haha). The heavenly trimmer post over in the "hunting for noise" thread was also trying (less articulately  ) to describe similar sound changes.

For me it took my resent chance to listen to live church organ music in a cathedral for me to really be confident that the presentation of the PC and usb trimmed sound, with its different portrayals of transients etc, is truer to life.

This is great that similar changes are giving what appear to be consistent  results and I was smiling reading your post because you  nailed the analysis and description of the sound so well (and much better than I could put it across).  There is a question of the size of the overall change to sound though now. Are you planning to put high quality tunable clocks in at both end of the usb link soon ? I think the major improvements on offer from the usb link might come from this.

Regards.

Nick.

[Nick]

Peter,

I just reread you post again. You absolutely nailed describing what is happening here 

Are you doing this in software   ! ! ?

N

[PeterSt]

Hi Nick, super stuff again.

Yes, software only, but mind the Silverstone card which is needed for the good base.

I guess I have a very strange disorder in being able to mimic sound from description (or hearing) in software and then by this (still asumed) bit perfect means.
I think I recently explained somewhere how a Jan Garbarek sax sounded nasty when a first version of XXHighEnd was out and really not a single SQ control was in there yet. So in an evening I thought of "how to" and Q1 was born. Worked next day and Garbarek became listenable (which is a miracle to begin with). Similar happened dozens of times with Bert, him applying something to a speaker filter, telling about it, me thinking of how to mimic that by means of software. An the other way around btw (so credit to Bert there) and which is how we both now ended up with the ultimate playback chain (the whole lot created by our co-operation).
I tried the same in the context we are talking about now and possibily that suceeded. Of course it takes some knowlegde of the OS + PC and what actually can happen, and that is what you (and Paul) described.
That's really all.


All I now need further (I'm sure) is something like fine tuning at maximum the frequency of the processor. Quite undoable. At least the last 3 hours brought me nothing yet.
Anyway there's a new Q4 here so you can focus on that one later ...

Peter

[Arjan]

Hi Peter,
What Silverstone card are you using? I have one laying around.
In the meantime I did some changes on the bios BLCK setting! Tested different settings, 93, 100, 99.8, 101, 101.5 etc Still more to try. And yes they make a difference. Currently my best setting is 94.6.
My daughter is helping me, she is 13 and hears the highs better than me, I am 50.
She complains about aer hurting highs with most of the setting, but the 94.6 gives the best results so far. Maybe there are better settings.
Feel free to delete this post if off topic......
Regards

[PeterSt]

No way OffTopic Arjan. My Silverstone is a SST-EC04-P. 6 or so of them exist, and I never searched for the difference.

Hey, the most nice that you (can) join this dicussion; the more the better.

And don't underestimate the value of someone around who just complains without you understanding why; maybe over here we must perform some test about who hears to up what frequency (me 55 here by now).

Regards,
Peter

[PeterSt]

Nick, here's a nice other one for you;

Remember sound stalling under W8 after 10:50 after a reboot - completely repeatable ? Sound continuing by moving the mouse ? Nothing to detect anywhere by whatever log files ?
Gone.

That issue not being there in W8.1 that sounding so different ?

Sit back and think what this tells (not really referring to W8.1).
Peter

[Nick]

Nick, here's a nice other one for you;

Remember sound stalling under W8 after 10:50 after a reboot - completely repeatable ? Sound continuing by moving the mouse ? Nothing to detect anywhere by whatever log files ?
Gone.

That issue not being there in W8.1 that sounding so different ?

Sit back and think what this tells (not really referring to W8.1).
Peter

Hummmm there is a riddle   

[Arjan]

Hi Peter,
My Silverstone usb is the same! I replaced it by the PP. What is the trimpot? The round thing underneath the clock next to the molex connection? And how does it work? I can do some tests with it.
Regards