XXHighEnd

A quick question on USB noise...

Would the noise increase with increasing data rate?

What I mean is, would we expect the USB noise getting through to the DAC to be lower when playing a file at its native 1411200 bps (16/44.1) vs. 45158400 bps (upsampled to 32/705.6)? Or is a 32x data rate irrelevant when it comes to the amount of USB noise getting through to the DAC?

Mani.

Hi Mani,

That depends a bit how you look at it ...

It is a bit difficult to say that the noise is higher, but it is "visible" so to speak that the higher rate incurs for more processing in e.g. the USB receiver. However, that is just theory. All I can really see is that when a USB interface is too critical(ly tuned) so to speak, the errors showing are more than with the lower rate. And then again this doesn't tell all because when there's more data to pass on, there might just the same be more errors as well.

Deep down, the speed is always the same, once the USB2 high speed (480Mbit) is engaged. This is not really related to any sampling speed that I know.

What matters more for sure is whether there's real data in there. Thus, just streaming (put up the Control Panel) implies less noise than real playback. This is about the 0's and 1's switching and such ...

Peter

Thanks Peter. I wasn't really talking about errors per se, but rather noise getting through to the DAC chips. I mean, this is the only mechanism through which sound can be affected, right?

I did a little test a few weeks ago. I recorded 4 takes of the same track:
1. (digital) USB-to-spdif -> Tascam with SFS=1
2. (digital) USB-to-spdif -> Tascam with SFS=120
3. (analogue) USB-to-spdif -> DAC -> Tascam with SFS=1
4. (analogue) USB-to-spdif -> DAC -> Tascam with SFS=120

1 and 2 sound identical to each other. 3 and 4 sound different to each other.

This just got me thinking about USB noise in general. Unfortunately, my USB-to-spdif converter is only 16/44.1 capable, so I couldn't play around with different data rates to see what difference this makes. I think there'd actually be too many variables to make any conclusions anyway (e.g. the DAC I used would no doubt process different rates differently).

Thankfully, it looks like 'we' are finally getting on top of USB audio finally.

Mani.

Mani,


But it's the same thing, so I used errors as such as an argument to "see" noise. Thus wnen errors are about, generally this is about noise. I said "generally" because in my experience it can also be about too much current "draw" over the lime meant to normally carry USB (power lines or data lines) but where thus extra current again can be expressed as noise (there must be voltage to have current) and too much noise expresses in errors.

Peter

As often, I can't guess easily what - in this case - your "analogue" means and this is because you seem to take things for granted which wouldn't come to my mind (with as result that I can't know what you are doing). I will give you an example, and I don't think that anyone in the worlds approaches audio like that :

We have a complete system for a customer here and I am to set it up for this customer. Check whether all works, make the network connection work as should, remote stuff, - everything. Because this is al here, it gives the opportunity to get mad - and I knew that in advance. Why ? well, because I would never ever be able to make the connections 100% the same as from my own system. And why is that ? easy enough ... that system of my own is already there and it first needs taking it out completely. But hey ...
It even requires the devices to be physically out of the way so that new system can be in the same place and my own must be stowed (say a few meters away) because else it has is influences. Ehm, but hey ...

Point is, at this level of matters it is easy to measure one way or the other how different things are. And wasn't it so that we are actually and sadly used to not being able to measure differences but perceive them all over the place ? and no, I am explicitly NOT referring to any "xyz Musings" because that is not the way to measure at all (and I'd seriously hope that nobody is going to adopt those completely flawed methods). So no, it is about - for example - the very easily idle measured noise on the DAC output. Remember, "for example". And next one should foremost not say "hey, that difference of noise of 2dB is 140dB down to begin with - are you crazy ?". Not not not.

Getting to the point, this is what you will be implying with comparing the things you or anyone does; differences could be perceived all over, but all you might be doing is having one of the compatitors around or even on top while device A is not influenced by device B, but B is infulenced by A and B is called out as the winner.
I am not claiming that such things are always the case in a negative sense, but I do claim that such tests should not even be thought of to begin with, because such influences can be there and you won't know it. All you can do is listen and judge and have a verdict because of that activity.
I would be taken breaths for a month before I'd attempt and then still I would not like to do it.

More in the next post.

...

... But we were discussing USB noise. OK, fine ...

So I have this Music Server PC and I can use it to test Phisolators. Remember, that small USB High Speed Isolator we designed/made ourselves. We make a fifty and all test fine by means of the DAC we use to test this.

Do not underestimate this "50", because here again there's an opportunity, in this case to find inconsistencies because of the amount we make and test. This is quite different from one only, and once that works it works. So, no ...

We made the 50 and in the very first DAC we make for real, noone work. WTF ? OK, it took as TWO full days to find out that this DAC we were making for real was connected to the Analyser and that that somehow disturbs. So indeed :wtf:.
All it took was taking the analyser out of the chain.
But wait, how can an analyser be disturbing so much ? Well ... that takes us all the way to the famous black wire. You know, that one now connected to Sw#3 for the "a" of the NOS1. So :

I sure recall a post from Nick where he said he had discovered that when this switch was down, the whole lot of signal ground was connected to the DAC's case and with that to PE (Protective Earth) (the case just is connected to the PE wire), unless of course the wall outlet used does not carry PE. I myself took that for granted because I had seen that myself by accident. So indeed Sw#3 better be Up or otherwise the internal (DAC's) isolation is quite moot (you connect things again via a backdoor).

At researching these Phisolators not working, with my luck I had so many around which ALL worked while now NONE worked, one of the suspicious matters was this Sw#3 connection. But what the heck ? this did not connect signal ground to PE (the DAC's case) at all. HUH ? I saw that and Nick saw that. So how can this be ? this NOS1a really was built the same as all of the others, including my own.

Then the coincidence of noticing that the analyser caused the troubles, it started to dawn on me. What if ...

Yes, what if the analyser somehow "created" that connection from signal ground to PE ?
And so it does.
I readily checked what would happen when my amplifiers (the Orelo MKII's internal ones) were connected, was pretty sure the same would happen, but it did NOT happen.

Are we still there ? ... things seem to go way out of control !

The Music Server PC is always more problematic for testing Phisolators and I always dedicate it to the longer USB cable run in use (it is down in the basement). But wait ... it is also in an other mains ring with separate PE from where I put in the test DACs. Remember, via that connection which isolates and what we call Ethernet.

:wacko:

Let's keep in mind that I am hinting at currents which flow over a connection where we don't want it - USB. And actually it will be so that the more we "isolate" the more we will be forcing that current - which wants to be there anyway because of potential (= voltage) differences, over that one line where we don't want it. Oh well, actually we wanted that (the isolated line again) but at some stage this "breaks" ...

Last post in this sequence ...

I mentioned that complete system for this customer for a reason ...

One of the things it features is a Stealth PC for the Audio PC. Actually it is right here where the Stealth emerged (for) and only after that something has changed : I thought to be able to add a Linear PSU myself. IOW, at first this was not meant to be so.
OK, so what's up ?
This is that PC which was going to be fed by that Terradak I had here 18 months or so ago. And yes, it came back (heavily modded) and now is (still) supposed to feed something which we indeed today call Stealth. The one LPSU against the other. What could go wrong ...

:)

What about "everything" ?
I say it like this, because everything "needs to be" upside down; where my own system is not in PE anywhere, that system needs to be in PE everywhere. Where my system is best with Sw#3 Down, this one does not work at all when we talk about Phisolators. Only errors. We have again been working for 3 days in a row to find the real merits of this all, and we can't. I also can't reason out what could be happening, just because *all* seems to be different. Think of this NOS1a being completely fine and UWB error free for a day in a row if it only connects to my own Stealth, while when it is connected to his Stealth fed by the Terradak, it is errors only. Mind you, when the Phisolator is in the chain - otherwise no problem at all.

What's up as a kind of last resort is testing with an isolated Silverstone card.


:seeyou:
Regards,
Peter

OK, one little phrase more :

Just because of such misery and because of us needing to put out quantities, each and every day something is improved just because I again have the opportunity to see something wrong.
It results in a sound now being so freakin' good that on a day like yesterday I wished that each of you would be listening to it with a lot of drooling together.

It would be true that USB is the main influencer these days, as the most benefit can be obtained from it once something has been found for improvement. It happens in dimensions which are totally new, a kind of in the realm of the Intona and its dedication of "dunno, just more realistic". This now has turned into severe "shattering" of sounds which should shatter and which out of all weren't even there before. The other day we were listening to Brothers in Arms, and the first two tracks (So far away and Money for nothing) exhibit quite the same for the guitar implying the melody : there's a trick applied to the guitar which let it end its phrases in a deep down roll of distortion and a complete detune from the lead in to that end. If you envision "whoa" as a sound, then this does "whaoooouuurrr-r-r-rt-t-t". And this just was not there at all before.
We discussed it as distortion (seriously) but with all the attention we put to it, we came to the definite conclusion that it is impossible to be distortion just because there's melody and thought in it.

In the music you all don't want to listen to (my ambient "test material") I now have examples of a kind of gun shot loud exhibits which are super sharply defined and dzzingg throug the room in a fashion that does not make you worry about the realism of your playback system. It just works out realistically. Thus, when a gun would be fired you don't hear a "plop" from the direction where your speakers are. Instead there's this origin of sound which develops so fast that only in aftermath you can recognize what/where its source was. Meanwhile it is still shattering (dying out) against the walls ad if they were resonators.
It seriously came to me yesterday that the "Phasure" idea is starting to work. Or merely, that the original thought indeed can work (re-position all sounds in mid-air by means of two transducers only, but because of the many frequencies the Localisation can be done with indeed two sound sources - and then thinking of my Local Positioning System (pre-Phasure), which works the other way around).

Have a nice Sunday y'all,
Peter

Peter, I've enjoyed reading these posts and must say that what you are experiencing is something that I have wondered might be possible to happen (your USB troubles - especially those with the Terradak) ever since I played around with lifting the PCIe ground with the Silverstone card and just how unstable USB became.  My efforts were slightly different to your issue of course, but I have been considering lately reattaching the NOS1a to PE simply from a safety perspective, and also because it is DC coupled to a valve amp that potentially could send high voltage DC back into the NOS1a if ever something went catastrophically wrong with the amp.  Plus, I wonder how a signal can be transferred between the PC and NOS1a without the potential between the boxes being equalised, or should I say without one box floating at a different potential to the other (think USB isolation combined with ethernet isolation).

The thing that puzzles me though, is how PE and signal ground are connected and why the Phisolator seems like a player in connecting the two.  PE and mains neutral are connected together at the mains entry box (to the house) in Australia, and I assume they are similarly connected in your part of the world as well...is the Phisolator by effectively severing a potential current loop via USB cable forcing a current loop between the PC and NOS1a via the mains electricity active and neutral wires all the way back to where power comes into the house (i.e. where PE and neutral are connected)?  I hope this question makes sense. 

Would a chassis to chassis ground wire between the Stealth PC and the NOS1a help?  For current to flow there must be a voltage difference between the two boxes, so perhaps a wire between them may help.

Or perhaps I have misread the situation.

Regards,

Anthony

Hi Anthony - Thank you.

First a general remark :
I learned not to apply such "box connecting" wires any more, because without knowing you'd just violate isolation. Example : If galvanic isolation between PC and DAC in USB is supposed to help (isolating) then it is obviously quite daft to connect things explicitly via the back door.
So ... you may be perfectly right in that such a thing should help, but it can not be the idea really.


Although this could be, this is not at all my subject. But for that I maybe should have been more clear what actually goes wrong ... So :

First let's say that the Intona applies decent galvanic isolation. Whether this really is so, I can't know (at least I'd need so sit back to proove that). But with this as a base, how can our own Phisolator be more detrimental to anything ? IOW, it won't.
(and this with the side note that two in series clearly work (SQ) for the better than one - so there you go with the Intona's judgement of it; we just can't know to what degree the isolation works out)

What I should have explained better maybe, is that the Phisolator itself "breaks" (as in voltage breaking through) but to the extent of the active elements in there malfunctioning. The result ? just USB errors you can see on your Driver Control Panel. And what I know see/feel/think is that this is caused by currents which we more and more incur for, the better and better we isolate everything. Think a voltage hammering on some barrier, the current building up. Until "plop" it jumps over the barrier, and no active devices liking that. I can also see too much of "repeating" in a frequency = time. Like always 10 errors per one minute, or 10 per second, or 1000 per second. But in a given situation just always the same.

I will give your other ideas further thought, but with for now the notice that in no country the mains setup are the same. I always refer to the UK where they have 5 very different types and the NOS1 at first in one of them was not not able to arrange for "no DC Offset". It took months to think of something that would counterattack that. And yes, I am sure you can not imagine the situation in the first place, but it just is so. This is btw also where the  active DC metering emerged ...

Regards,
Peter

I really like to refer to this post
Re: FIFO errors with Stealth - Clairixa - Intona (http://www.phasure.com/index.php?topic=3725.msg40733#msg40733)
and the one following it, as well.
Maybe it is not a coincidence and it at least tells me myself that I am on the right track with this. Not that it solves things readily, but isolation as such is not a free lunch and things get more complicated because of it.

Peter

Yeah sorry, my earlier post was a bit cryptic. What I actually did was set up two different chains:

1. Mach II -> USB-to-spdif -> Tascam spdif input
2. Mach II -> USB-to-spdif -> DAC spdif input -> DAC analogue output -> ADC analogue input -> ADC spdif output -> Tascam spdif input

In my earlier post, I referred to chain 1 as 'digital' and chain 2 as 'analogue'. What I found was that chain 1 seemed immune to SFS changes in XX. Chain 2 reacted to SFS changes in XX. And I was just trying to understand why this might be the case.

Does this now make sense?

Mani.

Hi Mani - Yes, I expected something like that.
So my other responses left aside, it should imply that whatever it is we incur with stuff like the SFS and probably much more, is going the analogue backdoor route (I wanted to say this  right away, but waited for your explanation and implied USB impeded noise instead - which of course is your topic).

Possibly it is even a combination occurring. Like in : what can not go via USB any more, *will* go elsewhere.
Also see 2nd new post coming up ...

Anthony, to start with the latter bquestion, this is not so at all. But it seems clear to me that when the former happens, there's too much of current on the Phisolator. And in an attempt to answer the former question :

I too need sitting back in figuring out how this happens. But say it is too complicated to follow all. Some intruduction though :

That black wire (connected when Sw#3 is down for the NOS1a and permanently connected for the normal NOS1) crosses a "surface" in the DAC case, say diagonally from the USB input to the other end (somehere under the main power supply). This is to guide away radiation and what it does is bring the USB shield to that other end.
OK ...

What I next found was that the Phisolator "just" requires this connection to be differently, additionally. I know I won't be able to explain it and it was all empirically found but where the normal USB input is connected to this black wire, the Phisolator's input goes *via* this same original input and ...
Needs is own connection to the case, but this time close by.

USB's shield is now connected two times to the DAC's case - one time right at the position of the input terminal (see photo below at the "i") and one time when it again enters the case via the normal input terminal (this goes via what I call a loop-back wire - see at the "u") but then via this input terminal shield across the DAC as always. Each can be connected and this connected and the permutation of 4, each give different results on the amount of errors - hence, this all has such a strong impact that the active part of the isolation fails, or just works fine.

While the above only explains about the internals of the NOS1a with Phisolator (SUB Isiolator) inside (and the "a" part being that other galvanic isolation further downstream), things probably get bananas when that black wire hence case connects to signal ground (which I would not like in the first place) but which is realized by my analyser but maybe is normal. Look :

My NOS1a/G3 what this is about, is not connected to PE, but of course the normal power cord is. It is only that where it enters the extension block, the PE is cut in that block.
BUT
These are copper strips and all I did with this block is cut that strip right in the middle of the 10 outlets it carries. Thus I now have 5 connected to PE and 5 are not. But think carefully, because that one half which is not connected to PE *is* connected to each other. Thus when I put the power cord from the NOS1a in there and the anayser in that same groupf of 5 (which I do), I have the cases of the two connected, well, assumed that the case of the analyser is connected to PE but which I did not check.
It would not be the first time that a device connects its dignal ground to PE and how knows the analyser requires it (it is a device to measure precisely, and I actually have not much knowledge about what's best for such a device). Anyway, if it connects its PE to dignal ground, then via its inputs the signal ground goes into the NOS1a and ... bingo. Now PE is connected to signal ground from the perspective of the NOS1a.

Anyone who uses a preamp or amp that connects its signal ground to PE *and* who does similar to what I do with my extension block or who just cut the PE connection from the NOS1a's case, ends up with the same situation and the Sw#3 should be up or otherwise this connection from signal ground to PE from the perspective of the NOS1a, is a fact.

In my own situation I measured that Sw#3 must be DOWN. If I don't do that, I might have USB errors. This is also the situation (since Phisolator) that I again connected the PC to PE, including the PE Switch on it set to "On" (connect to PE indeed). Do notice that I measure this with radiation meters, which I obtained after our conversation about them, many months ago by now. So it is quite easy to see, but it is not said that more radiation imply more errors. It could be that the radiation shows that "noise" is guided away through air, God knows what the receptor it.

Dizzy ?

Then let me again add that with this Terradak situation, both NOS1a/G3 as well as PC/Terradak must be in mains earth or otherwise really nothing works (1000s of errors per second) and that this time Sw#3 must UP or again otherwise totally nothing works. But it doesn't work well anyway, because still too many errors (like 3000 in 10 minutes).
And oh, notice that the determinatin of this single little chapter of working out these permutations, cost one afternoon. And then you end up with nothing (no satisfactory result).

Now on to the Silverstone (OK, almost) ...
Peter

[Highlight mine]

OK, so when USB errors occur, they are generally about noise. No problem.

But USB noise is still there even when there are no USB errors, right? And this noise can still affect the sound of the DAC, no? I mean, it is possible to change SFS from 1 to 120 and get zero USB errors in both cases, and yet still hear a change in sound from the DAC's analogue output between the two SFS settings, correct? Or are you saying that this change in sound must be accompanied by different USB error rates, as this is the only mechanism that can be at play here?

Mani.

Aha...

Just reading your next post now...

Mani.

If we combine my last two posts a little, and now know that I also measure a.o. electro magnetic field (which is the least one under our control), then it surprised me a little how much even a loose (!!) connection emits electro magnetic (EM) field.
It actually does not even matter whether the cable is connected or not, it just radiates once it is connected at one end. This counts for your power cord, for the USB cable and for the Interlink (and maybe more, but I just want to say "for everything and all").

What I do not know is to what degree devices suffer from this. But anyway what seems to be important is that all you isolate may not isolate really at all because something wants to go through air anyway (a not connected cable is fully isolated of course).
But this also means that when we "isolate" something, we must do it at both ends. Or "around" the device we want to isolate. Now think why there is this Sw#3 *and* the PE switch at the Stealth PC. Hey, not that I could reason it out, but I sure made them at both ends because that is what it comes down to.

Now we need to have both switches engaged. Why ? well, this has to be about something going over PE (which is indeed just the same as the neutral, which is ... just your power supply signal), and this is this analogue backdoor.
My device in between those two switches (which is now the NOS1a/G3 up to its analogue outputs I'd say), is still isolated. And the more I do that, the more the influence - now just originating from the PC of its own - must go via the mains.

Theoretically this implies a "draw" on the outputs of the DAC, which impy a "reversed" impedance to the D/A section. Thus, something is hammering on that impedance from the outside now, and any impedance change might be able to change the D/A process itself. You could even call it jitter, because it may cause the D/A "gate" to open a bit sooner or later. Say that current leaks through in reversed direction. Now THAT kind of jitter would be new !

And so I think that I just found something important. The missing link. Something which shouldn't influence so largely because it should disappear in the massiveness of high voltage (like 200uV of distortion should not be able to influence devices which run at 115V/230V). But it should imply jitter via that same backdoor.

OK, if I see radiation of 0.5V/m spouting out of everything, why shouldn't that influence fragile voltage stuff like a D/A conversion ?
And I suppose some other way around : why wouldn't that radation be less when the capacity (so to speak) is fully utilized ? This is just wild guessing. But I have something like : if radation (EM) emerges because of something not being able to go anywhere (a loose-end cable), then there should be no radiation when all is 100% connected. But not only connected - also fully consumed (no losses).
This is a bit of the opposite of isolation, but alas.
:hips:

Peter

When I did the PCIe ground isolation with the Silverstone Card, and because I was able to do it sequentially (i.e. a little bit at a time) what I noticed was that as the PCIe ground become "more" isolated that external RFI/EMI influenced the USB error rates more and more up to the point that turning on or off a flourescent light anywhere in the house would cause the USB connection to be totally severed and a restart required to get it back again.

What I "think" was happening is that the induced voltage in the PC from the EMI/RFI was enough to float the voltage difference across the PCIe connection enough so that connection was no longer possible...i.e. the potential across the signal wires became too much and because there was no mecchanism to keep them the same relative to each other that a longer current loop through the main power lines had to be forced which was not able to work in this situation.

I think it is the same sort of thing that happens when we isolate the PC via the Intona/Phisolator and ethernet cable...the boxes manage to "float" relative to each other and the connection becomes unviable.  Peter, one idea that I have had about going forward with this (assuming the actual problem/s is/are correctly identified) is to get the connection working properly with a completely disconnected PC.  If you can get things working well when there is no back-door (the PE-neutral route) by battery powering the PC (temporarily, not permanently) then by my thinking it should work once the backdoor is put back in the system and the PC is plugged into the mains.  Will a completely isolated PC even be able to reliably send USB data to a dac?  That is my question!