Shielded vs. unshielded Ethernet and Grounding

[Jud]

2 minutes ago, mansr said:

First I'd like seeing it defined in a way that makes sense to most engineers. Talking about the impedance of a voltage or current is like asking what colour the speed of your car is.

 

So is the possible difficulty that John and jabbr were referring to a cause, while you took it as a description of a quality?  In other words, that they are referring to leakage current that occurs in high impedance conditions, while you are saying one can't describe the current *itself* as having an impedance (not what they're saying, as far as I can tell)?

[jabbr]

6 minutes ago, Jud said:

 

Not sure it was intended that way.  My take was that he was wondering how you would refer to it understandably to most folks without using the "c-word." But he can speak for himself.

 

Yes, the value we are discussing is current. Leakage refers to leakage current. What needs to be dropped is the “high impedance” — I don’t like that terminology either for the same reason as @mansr — what he means is “leakage current through high impedance pathways” which is long so ...

[mansr]

2 minutes ago, jabbr said:

Yes, the value we are discussing is current. Leakage refers to leakage current. What needs to be dropped is the “high impedance” — I don’t like that terminology either for the same reason as @mansr — what he means is “leakage current through high impedance pathways” which is long so ...

Well, currents flowing through low impedance connections are not usually described as leakage, seeing as connections tend to placed where current flow is desired.

[jabbr]

4 minutes ago, mansr said:

Well, currents flowing through low impedance connections are not usually described as leakage, seeing as connections tend to placed where current flow is desired.

 

Right. As I’ve said several times now, the really high impedance pathways have really low current so ?Relevance?

 

Im continuing discussing this not that I really care about really low valued currents but trying to demystify this idea that seems to have been thrown out there

 

again are nanoamp currents relevant? 

[Jud]

5 minutes ago, mansr said:

Well, currents flowing through low impedance connections are not usually described as leakage, seeing as connections tend to placed where current flow is desired.

 

I will eagerly await the decision of the panel on the proper terminology. ?  (Not that it isn't important, it just would be nice to get past the semantics and move on to discussing the phenomena.)

[Ralf11]

1 hour ago, diecaster said:

 

Yes, he did. All over the place. Look starting here and then throughout the thread: * * *

 

If it is not "high-impedance leakage current", what is it?

 

 

it is a leakage current thru a high-impedance - Mansr can propose a different term if he wants

 

jabbr gave an xlnt and concise description (which should be placed in the CA Archives)

 

my question is: how important is it likely to be - esp. given the recent ruminations from jabbr about noise levels...

 

OTOH, it is easy to get rid of (some of) these - so just do it

[mansr]

8 minutes ago, Ralf11 said:

it is a leakage current thru a high-impedance

Is there any other kind of leakage?

[Ralf11]

yes

[mansr]

30 minutes ago, Ralf11 said:

yes

Do explain.

[jabbr]

http://web.mit.edu/8.13/8.13d/manuals/LowLevMsHandbk.pdf

[One and a half]

3 hours ago, Ralf11 said:

OTOH, it is easy to get rid of (some of) these - so just do it

I don't know why I bother, but I can't let this one go. Please do explain how easy it is to remove leakage currents from an Ethernet transmission.

[mansr]

3 minutes ago, One and a half said:

I don't know why I bother, but I can't let this one go. Please do explain how easy it is to remove leakage currents from an Ethernet transmission.

Use optical fibre.

[Ralf11]

Yup

 

- he may mean on the receiving side, but that is why I used the word "some" ?

[One and a half]

1 hour ago, mansr said:

Use optical fibre.

Thank you. Power supplies for the FMC to be considered, but at least the leakage currents are known.

[jabbr]

6 minutes ago, One and a half said:

Thank you. Power supplies for the FMC to be considered, but at least the leakage currents are known.

Don't use an FMC then. 

[plissken]

3 hours ago, mansr said:

Use optical fibre.

 

Wifi worx

[jabbr]

On 8/24/2018 at 1:25 PM, look&listen said:

Mockery make you seem less smart  

I wasn’t mocking @Superdad, rather chiding him with an implied <sarcasm> tag  and a wink ? ... and a somewhat juvenile joke you missed ?‍♂️

 

Alex is trying to defend @JohnSwenson‘s terminology which I’ve said here, and said before, that it’s confusing ... if I’m wrong I’m willing to learn, and waiting to hear ?

[JohnSwenson]

Here it is,  I have already been over this before, but it seems nobody remembers the details, so here it is once more, I'm NOT going to write this down multiple times. I'm in the process of starting the clocking tests and spending time on this is taking away from time I could be working on getting the clock test working so this is going to be it.

 

Here is the story of my leakage tests and what I found.

 

It all started about a year ago when someone else was testing one of our USB products and noticed some very low level increase in line frequency components (I'm in the US so that is 60Hz, 120, 180, 240 etc) at the output of a DAC whose USB input was fed from our product.

 

I had recently moved so I didn't have my lab fully setup so it was difficult to try and reproduce what was happening. I attempted to get the test equipment setup as quickly as I could (not easy, the old HP stuff is VERY heavy) and finally had enough setup to get some testing done. I did manage to see the same thing but didn't know where it was coming from. I don't remember everything I did at that point, but I was trying out different hypothesis to try and figure out what was going on. After several weeks of this about the only thing left was leakage current flowing through the USB cable, through the DAC, through the audio interconnect to the preamp, through the safety ground connection of the preamp, the safety ground in the wall to the neutral/safety ground connection in the breaker panel and back to the power supply of the computer providing the USB source.

 

So at this point I needed a way to measure the leakage current. The obvious way to do this is a resistor from the ground of the USB cable to a wire connected to the safety ground of a wall outlet on the same circuit as the PS of the computer. (I plugged the computer into a power strip and a plug with just ground wire into the same strip, about as short a path as I could get) I connected a piece of coax across both leads of the resistor and plugged the other end into the input of my spectrim analyzer and saw a nice line frequency spectrum, strong 60Hz and diminishing amplitude harmonics. I wanted to use the largest value resistor I could, the higher the value of the resistor the higher the voltage level I would have to look at. The resistor forms a resistive voltage divider with the source impedance of the leakage current, if the resistor is small relative to the source impedance doubling the resistor will cause a doubling of the voltage across the resistor. When increasing the resistor value there is some point when this relationship breaks down, as the resistor gets close to the source impedance it actually decreases the current flow through the loop, preventing the voltage across the resistor from doubling with a doubling of the resistor value. As the resistor increases further the voltage across the resistor starts going down and with increasing resistance will go to almost zero.

 

Well this nice simple progression did not happen. As the resistor value went up the voltage leveled off but didn't go down again, it continued to go up but not as fast as before. WHAT? This shouldn't be happening. The only reason for this is that there were two current sources in parallel, one which had much higher source impedance than the other.  (at this point I'm not sure if it is JUST two or if there are more than two involved).

 

If this was true if I kept on increasing the resistance at SOME point it would start leveling off and go down. I then started thinking about my measuring setup, I had the top of the resistor connected to the 1 mega-ohm input of the spectrum analyzer, I was thinking that could start causing a problem, the resistance was starting to get pretty high. So I went with a 10X probe (10 mega-ohms) . So then another round of increasing the resistor and STILL it didn't level off, the ten mega-ohm was looking to be too small.

 

At this point I realized I needed a really high impedance differential probe, at this level the ground connection to the spectrum analyzer could cause all kinds or weird effects, it really needed to be high impedance differential. Such things are NOT cheap. I finally decided to build my own using an Analog Devices instrumentation amp chip, they have one that is theoretically rated for 20 giga-ohms. Of course in reality it would never get that high, so I'm guessing the real implementation is in the 2-3 giga-ohms range. It turns out that all the ADI instrumentation amps have the same pinout and they sell a development board for it an a kit with 6 of their most common chip models. So I ordered a bunch of resistors and caps for this and built one using the high impedance chip. I powered the amp with a pair of LiFePO4 battery packs. This is what I have used for all subsequent tests.

 

At this point I needed to change the test setup a bit, the voltage across the resistor was getting too high for the instrumentation amp chip, so I decided to go with a small resistor (I chose 10 ohms rather randomly) which was across the inputs of the instrumentation amp chip, with the resistor I was changing between that and the ground of the PS. At this point I came to the conclusion that this was all about the PS powering the computer and really had nothing to do with USB per se, so I just started measuring the PS, it cleared a lot of things off the bench.

 

So increasing the resistor did eventually cause the voltage across the 10 ohm resistor to go down, I had finally found the source impedance of this current component of the leakage from the PS. It was around 400 mega-ohms.

 

With this high an impedance it is very difficult to "block" it by putting a resistance in series. Given the frequencies we are interested in you can't just put a giga-ohm resistor in series with the ground of a cable or put an inductor in series. Fortunately there is another approach which is to shunt the leakage current to the safety ground before it gets to computers, DAC, preamps etc. This works quite well.

 

So this means that the leakage coming from an SMPS seems to have at least two current sources one which is around 400 mega-ohms and one much lower. I tried increasing the resistor across the instrumentation amp to try and see if I could figure out the impedance of the low source impedance component and I didn't get too much higher than 10 ohms before it started going non-linear. Hmm, this seemed awfully low, so I tried several more tests which seemed to point at something much higher. This is why I think there might be more than two components to the leakage.

 

At this point I decided is wasn't worth my time trying to dig into exactly what is happening inside an SMPS, especially since each one seems to be a bit different. At this point I had enough understanding to have a good idea how to deal with this in audio systems, which was the whole reason for doing this, trying to find out what is going on, not just for pure science, but for how to cut down on its effects on actual audio systems.

 

Oh yeah, I did over 500 tests on many different SMPS and LPS and found that LPS models do not have this very high impedance components, this seems to be the realm of the SMPS.

 

So conclusions of all this:

the leakage from an SMPS seems to be comprised of at least two components, a high source impedance one and a lower source impedance one. The high source impedance component can be shunted to safety ground right at the PS, preventing it from entering the audio system. Even with the shunt there still is a low source impedance leakage which needs to be dealt with. What you need to do varies with with where in the chain the output of the PS is connected.

 

A standard Ethernet transformer will easily block the low source impedance component, but does not block the high source impedance component. This can be a problem with a wired network connection since the high source impedance component will go right through the transformers in switches and routers etc.

 

So I did some more testing and found that SOME switches when powered by a PS with the negative output connected to safety ground would actually shunt the high impedance source component. I tested a number of switches and only a few did this. At this point I have NOT exhaustively tested all of them to find out what makes the difference, why some switches shunt and others don't. Preliminary findings seem to point at differences in the connections of the center taps of the Ethernet transformers.

 

So there you have it, a much abbreviated account of the testing I did and the conclusions drawn from that testing. The process took over six months and took up a HUGE amount of my time. I have over 500 plots from the spectrum analyzer from all the different phases of this trek.

 

So do I REALLY need to say "high impedance source component of the total leakage current" EVERY single time I'm talking about this or can I use the shorthand "high impedance leakage"? It seems like things are going to get very verbose if everybody is insisting I use the long hand every single time  I talk about it. 

 

John S.

 

[jabbr]

8 hours ago, JohnSwenson said:

So do I REALLY need to say "high impedance source component of the total leakage current" EVERY single time I'm talking about this or can I use the shorthand "high impedance leakage"? It seems like things are going to get very verbose if everybody is insisting I use the long hand every single time  I talk about it. 

 

You could say that this leakage current component is 12.5 nano-amps (assuming 5V signals) — is this important?

 

Impedances in series are additive so makes total sense that an Ethernet transformer (an impedance) won’t add much to 400 mega Ohms. 

 

When you shunt the leakage to ground ? 12 nano amps to 2 nano amps ? I’m guessing here but won’t be zero

[jabbr]

9 hours ago, JohnSwenson said:

Oh yeah, I did over 500 tests on many different SMPS and LPS and found that LPS models do not have this very high impedance components, this seems to be the realm of the SMPS.

 

As you know this all has to do with parasitic capacitances and inductances. The major parasitic in an LPS is the interwinding capacitance of the transformer ... where I get a little iffy on this terminology is that although there may not be "very high impedance" parasitics in an LPS there will be "very very very high impedance" parasitics ... the higher the impedance of the parasitic then less important (the lower the current).

 

(I'm writing this for the folks listening in, just trying to simplify the discussion)

 

Quote

 

So conclusions of all this:

the leakage from an SMPS seems to be comprised of at least two components, a high source impedance one and a lower source impedance one. The high source impedance component can be shunted to safety ground right at the PS, preventing it from entering the audio system. Even with the shunt there still is a low source impedance leakage which needs to be dealt with. What you need to do varies with with where in the chain the output of the PS is connected.

 

The SMPS has a much more complicated circuit and "low impedances" which result in leakage currents may be an actual part of the circuit because capacitances are placed in order to reduce EMI ('nother topic)-- these are actual capacitors placed in the circuit, not merely parasitics and so you will see a much lower impedance across these capacitors i.e. your "low impedance leakage" ... there are medical grade SMPS which have lower leakage currents but grounding the (-) output works by voltage division. (Thievenin's theorem)

 

I totally understand where you are coming from, and recall the exchange that resulted in your investigation -- one of the reasons that I prefer not to publish measurements of other peoples equipment -- it takes A LOT of work to do it correctly (I have a lot of that old HP stuff as well and setting up everything is real work).

 

For me, personally, a schematic with included parasitics greatly helps me to understand what currents are flowing where, and those 3D color coded models are even better -- too bad the software is so f^%$%$ expensive

 

[Ralf11]

Thevenin's theorem - for those listening in... no 'i'

 

** people could post the color coded model output from the $$$ software, or just point out how to avoid / deal with leakage currents

[Ryelands]

On 8/24/2018 at 1:00 PM, Em2016 said:

So I was kind of asking if jabbr had actually seen what John S is talking about [here] . . .

 

I haven’t seen it and have, for obvious reasons, no plans to try to measure it but I have heard it. My experience might be of wider interest.

 

I’ve been using a Phillips TDA1541A DAC chip more or less continuously for over 30 years. I recently tried powering it with three LPS-1s wired to provide –14v, -7v and +7v; the –14v line as the nominal –15v supply (seems fine) and the +/-7v lines driving fancy voltage regulators to give +/- 5v.

 

Hitherto, I’d been using a DIY’d but definitely-not-shabby C-L-C linear PSU. Hint: the circuit is, er, not unlike that in Uptone’s JS2 supply.

 

I thought the sound was already excellent but the improvement obtained by switching to the LPS-1s was very marked indeed. If someone tells me that I’m mad to spend ~$1,500 to drive a $10 chip first produced in 1985, I’d agree. Except that I’ve heard the result and someone hasn’t.

 

Whatever, after a few hours, I found I could hear a background noise when standing near the speakers. Cranking up the volume, it clearly resembled pink noise whose volume modulated in time with the load variations on the Meanwell SMPSs driving the system’s five LPS-1s, a pattern I happened to know from previous trials not relevant here.

 

To cut a long story short, I spotted that I’d forgotten to ground the output from one of the Meanwell SMPSs driving the LPS-1s. I fixed the error and the problem disappeared.

 

What was interesting was that the guilty Meanwell wasn’t driving one of the DAC’s power lines but the Network Audio Adapter. Note that all the Meanwells in my system are connected to the mains via individual isolating transformers, that  I have a LiFePO4 battery-powered Intona isolator thingie between the NAA and the DAC AND that I switch the Gnd lines in the USB cables out of circuit while playing music.

 

Dave

 

 

 

[plissken]

On 8/26/2018 at 12:43 AM, JohnSwenson said:

Here it is,  I have already been over this before, but it seems nobody remembers the details, so here it is once more, I'm NOT going to write this down multiple times. I'm in the process of starting the clocking tests and spending time on this is taking away from time I could be working on getting the clock test working so this is going to be it.

 

Here is the story of my leakage tests and what I found.

 

It all started about a year ago when someone else was testing one of our USB products and noticed some very low level increase in line frequency components (I'm in the US so that is 60Hz, 120, 180, 240 etc) at the output of a DAC whose USB input was fed from our product.

 

I had recently moved so I didn't have my lab fully setup so it was difficult to try and reproduce what was happening. I attempted to get the test equipment setup as quickly as I could (not easy, the old HP stuff is VERY heavy) and finally had enough setup to get some testing done. I did manage to see the same thing but didn't know where it was coming from. I don't remember everything I did at that point, but I was trying out different hypothesis to try and figure out what was going on. After several weeks of this about the only thing left was leakage current flowing through the USB cable, through the DAC, through the audio interconnect to the preamp, through the safety ground connection of the preamp, the safety ground in the wall to the neutral/safety ground connection in the breaker panel and back to the power supply of the computer providing the USB source.

 

So at this point I needed a way to measure the leakage current. The obvious way to do this is a resistor from the ground of the USB cable to a wire connected to the safety ground of a wall outlet on the same circuit as the PS of the computer. (I plugged the computer into a power strip and a plug with just ground wire into the same strip, about as short a path as I could get) I connected a piece of coax across both leads of the resistor and plugged the other end into the input of my spectrim analyzer and saw a nice line frequency spectrum, strong 60Hz and diminishing amplitude harmonics. I wanted to use the largest value resistor I could, the higher the value of the resistor the higher the voltage level I would have to look at. The resistor forms a resistive voltage divider with the source impedance of the leakage current, if the resistor is small relative to the source impedance doubling the resistor will cause a doubling of the voltage across the resistor. When increasing the resistor value there is some point when this relationship breaks down, as the resistor gets close to the source impedance it actually decreases the current flow through the loop, preventing the voltage across the resistor from doubling with a doubling of the resistor value. As the resistor increases further the voltage across the resistor starts going down and with increasing resistance will go to almost zero.

 

Well this nice simple progression did not happen. As the resistor value went up the voltage leveled off but didn't go down again, it continued to go up but not as fast as before. WHAT? This shouldn't be happening. The only reason for this is that there were two current sources in parallel, one which had much higher source impedance than the other.  (at this point I'm not sure if it is JUST two or if there are more than two involved).

 

If this was true if I kept on increasing the resistance at SOME point it would start leveling off and go down. I then started thinking about my measuring setup, I had the top of the resistor connected to the 1 mega-ohm input of the spectrum analyzer, I was thinking that could start causing a problem, the resistance was starting to get pretty high. So I went with a 10X probe (10 mega-ohms) . So then another round of increasing the resistor and STILL it didn't level off, the ten mega-ohm was looking to be too small.

 

At this point I realized I needed a really high impedance differential probe, at this level the ground connection to the spectrum analyzer could cause all kinds or weird effects, it really needed to be high impedance differential. Such things are NOT cheap. I finally decided to build my own using an Analog Devices instrumentation amp chip, they have one that is theoretically rated for 20 giga-ohms. Of course in reality it would never get that high, so I'm guessing the real implementation is in the 2-3 giga-ohms range. It turns out that all the ADI instrumentation amps have the same pinout and they sell a development board for it an a kit with 6 of their most common chip models. So I ordered a bunch of resistors and caps for this and built one using the high impedance chip. I powered the amp with a pair of LiFePO4 battery packs. This is what I have used for all subsequent tests.

 

At this point I needed to change the test setup a bit, the voltage across the resistor was getting too high for the instrumentation amp chip, so I decided to go with a small resistor (I chose 10 ohms rather randomly) which was across the inputs of the instrumentation amp chip, with the resistor I was changing between that and the ground of the PS. At this point I came to the conclusion that this was all about the PS powering the computer and really had nothing to do with USB per se, so I just started measuring the PS, it cleared a lot of things off the bench.

 

So increasing the resistor did eventually cause the voltage across the 10 ohm resistor to go down, I had finally found the source impedance of this current component of the leakage from the PS. It was around 400 mega-ohms.

 

With this high an impedance it is very difficult to "block" it by putting a resistance in series. Given the frequencies we are interested in you can't just put a giga-ohm resistor in series with the ground of a cable or put an inductor in series. Fortunately there is another approach which is to shunt the leakage current to the safety ground before it gets to computers, DAC, preamps etc. This works quite well.

 

So this means that the leakage coming from an SMPS seems to have at least two current sources one which is around 400 mega-ohms and one much lower. I tried increasing the resistor across the instrumentation amp to try and see if I could figure out the impedance of the low source impedance component and I didn't get too much higher than 10 ohms before it started going non-linear. Hmm, this seemed awfully low, so I tried several more tests which seemed to point at something much higher. This is why I think there might be more than two components to the leakage.

 

At this point I decided is wasn't worth my time trying to dig into exactly what is happening inside an SMPS, especially since each one seems to be a bit different. At this point I had enough understanding to have a good idea how to deal with this in audio systems, which was the whole reason for doing this, trying to find out what is going on, not just for pure science, but for how to cut down on its effects on actual audio systems.

 

Oh yeah, I did over 500 tests on many different SMPS and LPS and found that LPS models do not have this very high impedance components, this seems to be the realm of the SMPS.

 

So conclusions of all this:

the leakage from an SMPS seems to be comprised of at least two components, a high source impedance one and a lower source impedance one. The high source impedance component can be shunted to safety ground right at the PS, preventing it from entering the audio system. Even with the shunt there still is a low source impedance leakage which needs to be dealt with. What you need to do varies with with where in the chain the output of the PS is connected.

 

A standard Ethernet transformer will easily block the low source impedance component, but does not block the high source impedance component. This can be a problem with a wired network connection since the high source impedance component will go right through the transformers in switches and routers etc.

 

So I did some more testing and found that SOME switches when powered by a PS with the negative output connected to safety ground would actually shunt the high impedance source component. I tested a number of switches and only a few did this. At this point I have NOT exhaustively tested all of them to find out what makes the difference, why some switches shunt and others don't. Preliminary findings seem to point at differences in the connections of the center taps of the Ethernet transformers.

 

So there you have it, a much abbreviated account of the testing I did and the conclusions drawn from that testing. The process took over six months and took up a HUGE amount of my time. I have over 500 plots from the spectrum analyzer from all the different phases of this trek.

 

So do I REALLY need to say "high impedance source component of the total leakage current" EVERY single time I'm talking about this or can I use the shorthand "high impedance leakage"? It seems like things are going to get very verbose if everybody is insisting I use the long hand every single time  I talk about it. 

 

John S.

 

 

So, in summary, go WiFi. 

[Jking]

  Ethernet just moves data to the Dac's buffer.  Unless you are dropping packets or under running the buffer I don't see how it makes a difference.  If you needed ethernet to trigger something in real time, with great percission then the noise and jitter and the clock on the physical layer might be more relevant.  Texas Instruments makes low noise and jitter clocks for those applications.  It would just be wasted just to fill a buffer on Dac outside real time operation.

    

[Ralf11]

Some say the transformers on wired ethernet do not adequately reduce parasitic leakage currents, which carry noise to the DAC...

 

It is not too hard to use optical and solve that problem

 

... or use WiFi as per the above post

 

of course, it is always possible that the convertors on the DAC side inject noise into the DAC

 

or that noise arises inside the DAC or its chip

 

and then... there is Johnson Noise✋