USB Cables... again

[juanpmar]

Hi Mani, with balance connections you don´t have that noise but any difference in SQ with the Ridge Street Enopias USB cable versus the stock cable using in both cases xlr connections?

This is more difficult to test as it would require taking the Berning out and replacing with one of my other amps, and I don't really want to do this just now. But in my office system, I use a fully balanced setup - NOS1 going into active Genelec satellites/sub. From what I remember, the Enopias reduced some of the brightness of the printer cable even in this balanced setup. Could be a good thing, or a bad thing. Will need to try again and get back to you.

Mani.

Ok, thanks Mani

Juan

[manisandher]

1. Are the yellow lines actual wires?

2. Similarly do you really have two earth wires (3 & 4) connected to NOS1?

Hi Paul. Yep, and yep.

USB is just a nasty thing, and I myself don't understand it. This is about how it is organized in the PC which I think is "a standard" and how it runs its gnd over PE as far as I can see.

Just an observation it seems from the Berning cct diagram that the electronic earth is connected to protective earth.

It's really because of your observation that I'm doing all this. In the PC, the ground (electronic earth) is connected to PE. In the Berning, the ground is connected to PE. My feeling is that my connections 2 and 3 would reduce ground loops in the USB cable and in the interconnects between the NOS1 and Berning, as there would be very little potential difference (depending on how 'rigid' my star arrangement is) between each ends of these cables.

Again, pure speculation.

Mani.

[CoenP]

This is where the refrencing the usb at two points goes astray: the groundrefs 2 and 3 are able to and will carry a signal.

Coen, my feeling is that this is a good thing. What we want to do is reduce the return current in the USB cable, as this will only increase the ground plane noise. With 2 and 3 connected, the return current has a much lower impedance return path via the thick (6mm) earth wires.

But hey, this is pure speculation on my part.

Mani.

If that is true then:

You don't have to run through the central ground point. A thick wire parallell will do the same thing, maybe better because it will be shorter (less impedance and much shorter loop).

Regards, Coen

[CoenP]

My hunch is that the noise on the usb 5V is able to enter your NOS1 circuitry as a result of damage.

Well, this is easily tested - I have two NOS1s! I'll take the one from my office and try it in my main room.

Mani.

That would definitely exclude the NOS from the picture!

My memory about this issue starts coming back:
I managed to somehow create a short circuit on the 5V in the NOS ESD protection chip. This immediately shuts off the usb ports on the pc and renderes the NOS1 unusable. Awaiting the NOS1 repair I played some tunes through the mobo soundcard, where the PC noises also where very apparent. Same thing happened  when I swapped the PSU to my other completely different internet pc. Using another spare PSU, noises where geatly reduced. The PSU causing the noise was decommissioned.

The thing is I didn't start out with these noise issues. They appeared likely as result of some failure somewhere, the PSU being prime suspect. I cannot exclude the possibility that the fault also partially distroyed the ESD chip making the NOS1 transparent for the noises. The ESD chip could have failed later completely by something else.

This remains speculation, because I never internally examined the suspect PSU nor the NOS1 on this matter.

Regards, Coen

[PeterSt]

But this has now done exactly what Peter wanted to avoid - use the PE for ground in the NOS1. What puzzles me though is that this is unavoidable isn't it? I mean, where does that ground get it's reference from anyway? The PC's PE, no?

The NOS1 doesn't need any ground. It creates it herself. Same with your Sauermann.
And this right away is the danger as well;

Any self-created ground (also named Virtual Ground) will have an exact 0V reference. This is because this 0V is just in the middle of the voltage rails used. So, put e.g. an exact +5V and -5V opposed to eachother, and in the middle is half of that - 0V. Now :

While such a device has a real 0V reference, any other device not doing that will be based on the what it's provided through the mains. This is more difficult to explain, but take it that the 0V reference which is inherently there, is not that once devices are connected. The reference is pulled up somewhat, and the difference in DC you measure will be the difference between NOS1/Sauermann and the device connected. In the mean time a current flows through the connection.

This is where I need to stop, because all further is speculation. But one example, relating to my post day before yesterday about J.S. :
When we would be able to isolate the potential difference, for example by means of explicit isolators, you can see the difference in (DC Offset) voltage between both parts. In my case this showed 56V at some moment, and what it comes down to is that this is the created difference on PE by other devices (which goes into the neutral as well).
I, very personally, think that it is not allowed to have an isolated means on one side only. So think of it :
There's the left part with a reference of 56V. There's a right part (behind the isolation) of 0V. Now we connect that to our amplifier which again is related to the same 56V as our left part. The 0V will fight the 56V and the 56V should drop (never think to 28 because it is current related (this is also about your equal length wires which don't tell a thing for this reason)). So, now our left part references to that lower Voltage ? hmm ... something is looping somewhere, but think oscillate.

The above is not official but is derived from empirical finding to some extend. I had a couple of PCBs made to experiment, and they are able to go into the NOS1 and prove to me that any isolation is not allowed. No no, this is nothing about an upgrade, just proving something in an actually working environment. In the end this is about a practicle case where isolation works all right, but I could find no way to eliminate the mains noise I otherwise can easily do. If I can't so this anymore with the isolation, by case is proven.

If you have two PE references anyway (Mani, you seem to have that) measure the DC voltage between the two. AC Voltage, also OK. I think we have been through this before, but re-do that in your current (pun) situation. You could be surprised.

I should add : Whatever you provide to prevent such loops which in my view *have* to be there (my 56V example) will only make it worse because out of control; it will now go through air to ... something.

Once you can detect voltage between two PE rings, shut *everyting* off. Now there should be nothing (and if not, something is still on or wrong (like transformers not creating 0V). But now switch all on again. Chances are very fair that you won't be able to reach the same voltages you saw before. Or just 0V now.
All you implied was a groundloop to flow through a different path which is related to the sequence of switching on. Or, that you somewhere in the past weeks implied a loop somewhere, which can't be broken anymore (tried a new TV for example).

Peter

[manisandher]

If you have two PE references anyway (Mani, you seem to have that) measure the DC voltage between the two. AC Voltage, also OK. I think we have been through this before, but re-do that in your current (pun) situation. You could be surprised.

OK, I've just taken a few measurements:

1) DC between beginning of 1m Enopias cable shield and NOS1 USB connector shield (my earth 3 connected) = 6mV

2) DC between beginning of 1m Enopias cable shield and NOS1 USB connector shield (my earth 3 disconnected) = 6mV

3) DC between NOS1 USB connector shield and Berning RCA -ve input (my earth 3 connected) = 0mV

4) DC between NOS1 USB connector shield and Berning RCA -ve input (my earth 3 disconnected) = 0mV

Readings 1) and 2) are surprising. Why is there a 6mV offset between the two ends of the Enopias cable's shield? (It's only 1m in length after all.) And does this mean a ~36mV offset between the PC in the basement and the NOS1?

Readings 3) and 4) suggest to me is that my extra earth connection, 3, is not introducing an extra ground loop. But rather it might be channeling ground plane noise away from the USB and interconnect cables. Of course, the converse could be true and it could be adding noise. But based on what I'm hearing, I suspect the former rather than the latter.

I have a voltmeter here and a basic oscilloscope. Anything else I should try to get to the bottom of why the sound is changing by connecting my earth connection 3?

Mani.

EDIT: All the above readings were with my connection 2 in place. Disconnecting this changes things. Readings 1) and 2) jump to 22mV. Again, extrapolating this to the whole length of the USB cable, this suggests a 130mV offset between the PC and NOS1.

So if nothing else, my connection 2 is at least greatly reducing the PC/NOS1 offset. My connection 3 seems to have no affect on this offset.

[CoenP]

A few remarks from my side:

- how about the monitor? Where do you plug it in and how is the earth connected?
I have best results if it is plugged into a different powerreceptable from the pc. The monitors switching psu remains allways connected....

- as I went om inspection in my own house it turned out that I don't have a real earthing since they replaced the incoming water with plastic pipes and a new meter but left the connection to the fusepanel in place. That left me under the impression that I had earthing but it was actually the safety earthingh of the water system  .
Safetywise I have two Ground Fault Circuit Interrupters so no lethal situation arises, however all groundsh&t is shared between devices that are connected to the PE. I just requested a few quotes to get a proper groundbar in place.

Now both the PC PSU and NOS are connected with a common PE on the powerstrip but to no real earth. Measuring voltages across the usb is a little hard, but between pc usb and amp-in they match the NOS1 offsetreadings.

I still am under the impression that most noise is generated by the poweramp and that all other stuff is below that.

Regards, Coen

[Nick]

Hi,

A few notes from measuring round NOS1 and PC chassis ground voltages measured at the USB shield connection points of the PC and NOS1. I currently have standard PE arangemnts (eg no star PE eathing of NOS PC AMPs etc).

Measurements are taken at:
a) USB shield connection at the PC.
b) USB shield connection at the NOS.

Both of these points have low resistance connection to the PC PE connection and the NOS PE connection respectively.

Case 1
With PC turned off, NOS1 turned off, NO USB lead connection between PC and NOS1.
Voltage a) to b) = 0mv.

As expected 0v as both chassis grounds are connected to PE via the mains leads to the components.

Case 2
PC turned on and booted, NOS1 turned off, NO USB lead connection between PC and NOS1
Voltage a) to b) = 10.6mv (eg PC is is now 10.6mv higher than NOS1 PE).

Almost all PCs that I have owned elevate their chassis ground potential with respect to mains PE when the PC is booted.

The oscilloscope trace below is taken between point a) and b) under Case 2 test conditions. This gives some idea of the level of noise that is set up between the PC and NOS1 PE / chassis grounds when the PC is booted.

Mani mentioned being able to hear mouse movement through his speakers. I found that moving the mouse did visibly excite the noise shown on the scope trace, particualy the spikes above the main noise level (in real time these spikes appear relatively random and frequent).


Case 3
PC turned on and booted, NOS1 turned off, USB lead now connecting PC and NOS1.

Voltage 6.4mv (eg PC is is now 6.4mv higher than NOS1 PE).
The NOS Chassis ground connection to the PC via the USB shield is pulling down the PC chassis voltage from 10.6mv to this new value of 6/4mv.

Thoughts.

The PC sets up noise on its chassis ground when it boots and elevates its chassis ground by an average of 10mv but a detailed view shows the type noise that is present below. This must set up a ground loop with some complex noise current patterns circulating. The shield of the USB cable in actually connected to NOS signal earth via a 104kR and 4.7nf capacitor which form a parrallel connection between chassis ground and signal ground (this seems to be normal practice in USB interfaces not just for music devices). This means that the noise that is in the trace generated by the PC can in theory transfer in to the NOS signal ground (I must take some more traces to see it this is actually happening).

Mani’s connection “3” may be presenting a low impedance route for this noise (in the trace below and the associated loop currents) to be lead back to PE before it transfers into the NOS Chassis ground / NOS signal ground.
As Coen points out there is the possibility of EMI pick up in the PE wiring loops which but I could not really guess the effects of these.

Hope this helps.


Nick.

[Nick]

If you have two PE references anyway (Mani, you seem to have that) measure the DC voltage between the two. AC Voltage, also OK. I think we have been through this before, but re-do that in your current (pun) situation. You could be surprised.

OK, I've just taken a few measurements:

1) DC between beginning of 1m Enopias cable shield and NOS1 USB connector shield (my earth 3 connected) = 6mV

2) DC between beginning of 1m Enopias cable shield and NOS1 USB connector shield (my earth 3 disconnected) = 6mV

3) DC between NOS1 USB connector shield and Berning RCA -ve input (my earth 3 connected) = 0mV

4) DC between NOS1 USB connector shield and Berning RCA -ve input (my earth 3 disconnected) = 0mV

Readings 1) and 2) are surprising. Why is there a 6mV offset between the two ends of the Enopias cable's shield? (It's only 1m in length after all.) And does this mean a ~36mV offset between the PC in the basement and the NOS1?

Mani hi,

I just read this again and your question about why there is 6mv over a 1m shield and would this meant that you get 36mv from your basement PC with longer USB lead.

It just clicked looking at the trace I posted above. This is not a steady state offset we are looking at across the cable (or even audio frequency). If this were just DC offset then the resistance (proportional to the length of the shield) would determine the voltage drop. What we are seeing is a ver high frequency noise profile (a wave length of 1 horizontal division in the trace is a 2mhz frequency and as you see the noise is much higher frequency than that). What our volt meters are measuring is the rms value of the noise which in my case is 10mv with out the USB shield connection between the pc and nos1 and 6mv with the USB connection. The USB connection dose allow the NOS1 to effectively "pull down" the amplitude of the noise by connecting it to PE but the impedance of this route to PE is not low enough (being meters long) to completely drain the noise away to the actual protective earth. However your additional connection 3 probably lowers the impedance to PE and helps to reduce the noise amplitude. If you want to see what causes that hard top end sound, I think that scope trace could be the culprit.

If I am right your other PC in the cellar will have a similar USB shield voltage (provided that the ATx psu and mobo produce the same offset (as the PC you have measured) above chassis ground when the PC is booted.

I think things are starting to make more sense.

Nick.

EDIT - a little more thinking and I think this now links into Paul's experience using a star PE setup. Im thinking that just as your additional connection "3" may reduce impedance to ground for this noise, Paul's star earth is possibly doing the same thing at the PC and generally in the NOS1 ?

What would happen if a choke was placed in line with the shield connection between the PC and NOS could the noise be filtered ??? Of course this may help with the noise transmitted over the USB shield but since the PC's signal earth, is linked to the chassis earth, this noise is likely also to be in the USB signal ground wire and possibly the d+ and d- data USB wires. Still it should be easy enough to check these other connections for noise.

[manisandher]

Hi Nick, thanks so much for taking the time to share your thoughts.

I've repeated your readings here. But a couple of points first:

1) On the Ridge Street website it clearly states that the Enopias has its shield connected. But I've found that it does NOT - the type B connector shield is connected neither to the type A data shield nor type A power shield on the other side. I'm not sure if this is intentional, to help remove ground loops between the PC and DAC, but it seems strange to me. In any event, for these readings I've used a standard 1.5m USB cable.

2) I have a 5m USB extension lead between my PC and the 1.5m USB cable. I have ignored the 5m extension cable in these measurements and have taken point a) as the USB type A shield of the 1.5m USB cable and point b) as the NOS1 shield.

Anyway, to my measurements using your cases:

Case 1: 0mV
Case 2: 22mV (22mV at PC)
Case 3: 3mV (22mV at PC)

If I connect my earth connection 2, these become:

Case 1: 0mV
Case 2: 6mV (6mV at PC)
Case 3: 1mV (6mV at PC)

So, my connection 2 seems to be pulling the PC chassis voltage down quite a bit, in much the same way that the NOS1 does. However, in both configurations of Case 3, the NOS1's 'voltage pull down effect' does not extend all the way back along the 5m extension cable to the PC.

Interesting exercise...

Mani.

[manisandher]

The shield of the USB cable in actually connected to NOS signal earth via a 104kR and 4.7nf capacitor which form a parrallel connection between chassis ground and signal ground (this seems to be normal practice in USB interfaces not just for music devices). This means that the noise that is in the trace generated by the PC can in theory transfer in to the NOS signal ground (I must take some more traces to see it this is actually happening).

This may well be why the Enopias cable is having such an affect on the sound as the shield is not connected. What is now not clear to me is how the Enopias is eliminating the noise coming through the USB cable. Is it a) because the shield is broken, or b) because the power line is broken?

Mani’s connection “3” may be presenting a low impedance route for this noise (in the trace below and the associated loop currents) to be lead back to PE before it transfers into the NOS Chassis ground / NOS signal ground.
As Coen points out there is the possibility of EMI pick up in the PE wiring loops which but I could not really guess the effects of these.

I strongly suspect that the differences I'm hearing having my connections 2 and 3 in place have nothing to do with channeling noise away from the NOS1. I mean, they have zero affect on the general 'crackling' noise and the more specific peripheral noises I can hear coming through the speakers. Rather, I believe the real benefits of connections 2 and 3 are in providing a close-to-0V reference across the whole system. Ordinarily, using something like the Sauermann amp, this just wouldn't be an issue. But with the design of the Berning amp, there's an inevitable ground loop across the whole system from the PC to the amp. This cannot be good, with currents running through the USB cable and the RCA interconnects. As I've said before, my connections 2 and 3 should be forcing these inevitable currents through the additional earth wires, and not the USB and RCA cables.

That's how I imagine things working in my head at least. But again, pure speculation really. What isn't speculation at all is that they affect the sound... an in a very predictable and repeatable way.

Mani.

[Nick]

Mani hi,

I think there is some real progress here in getting to an understanding PC generated noise. The debate is really helping to draw together snippets of information and the wider picture is coming into view.

Thanks for posting your measurements of Case 1, 2, and 3 tests.

Was your higher value of 22mv for case 2 measured on you Asus mobo PC system by any chance ?

Interesting about the Entopias cable having a break in the shield ground. I had also tried lifting the USB shield connection at the NOS end some time ago when it became clear that there is coupling within the NOS of PE and signal grounds. I was in mail discussion with Peter and Paul about it, but we all seemed to get slightly different results so I did not post at the time. Now I tend to move back and fourth between lifted and connected shield at the NOS but generally I also experience a cleaner top end with less hash when the shield is lifted which seems consistent with your findings.

The light that has gone on for me is that the USB transmitted noise is not a DC offset or even audio frequency. For these noise types we could expect connections to PE to drain the noise resonably effectively. The noise is however in the MHz range touching the 100s MHz range which means that although it can be attenuated to some extent by grounding at the NOS (about 3db improvment in my case, see our case 2 and case 3 measurements), the impedance of the ground is too high at these frequencies to reduce the noise level with any great effectiveness. I agree with you about your connections 2 and 3, and your measurements really show their effect very well. At these noise frequencies even these low resistance grounds can only help reduce and not eliminate the noise, but there is good correlation between managing the noise down and improved sound !

Of course the SQ changes could still be due to better ground referencing at lower frequencies (audio down to DC) as you mention, so this must be considered. Looking at the noise in the scope trace though, you can really see how jitter in the DAC and resends of USB data could be caused, this would defiantly leading to the hash and hardness in SQ.

Seeing the the problem in this way is leading to all sorts of thoughs on what to test next and how it might be managed which is great. I'v got stuff in mind to look at over the weekend and will post back but I'm wondering now if this could really lead to something.

The USB shield is a supper kick off point. Understanding what is going off on the USB d+ d- and signal earth is next to help understand the overall connection (we should be able to forget the +5 volt USB wire as its not needed by the NOS).

Best,

Nick.

[PeterSt]

May it help ...

You guys - most of the time if not always - ever talk about less harshness or less bright etc., while you will not have seen such a post from me ever. This could mean I am deaf of course ...

But with the greatest emphasis you HAVE to understand that - as far as I can get it through forum posts - you same guys have other problems in the first place. So, with no exception I see you talk about noise through the speakers of whatever kind. And if not, tell me now that this is not so, but prove it please by means of how you solved it.

Of course, you are just hunting for the elimination of "noise" and this is always a good thing. But it really is of no use to eliminate noise in whatever MHz reason as long as the in-band noise is there and possibly at even higher levels. And I don't see how the cause can be the same.

I am not saying that this is all a waste of time, but I do tend to think that it is a wrong subject. It is, as long as you as you perceive this audible noise (which Nick's as a special situation of course).
I should add that in the end I am only glad with these experiments and analysis, so please keep this in mind. However, I see you become obsessed and whatever you will solve, it will not solve the issues you *will* be having. But, might you still think that the noise which is coming through your speakers, and which will be under 16KHz without me listening, is not harmful to the sound, well, then I said nothing. 

Best regards and happy hunting,
(and don't get obsessed)
Peter

[Scroobius]

Peter,

Best regards and happy hunting,
(and don't get obsessed)

Hey come on we all know that the only reason we are all here in the first place is because we are "obsessed" 

Maybe I will start a new thread "HiFi Obsessives Anonymous" and each post must start with "Hello my name is XXXX and I am a Hi Fi Obsessive".

P

[AlainGr]

Ha ah! Finally I know who I am 

Alain