Investigation into ground loop

[Newmarket]

Firstly the error in my test for whether the mains earth affects loop impedance or not was indeed due to the mains lead of the function generator. I learned a good lesson. So, I was wrong. It does not affect the ground loop impedance. Just to clarify again that when I say ground loop impedance I am not defining ground loop. but simply the ground path that I sketched out in my diagram above.

However, before continuing with the rest of the stuff I would like to mention that I have initially amended my first post with regards to Thor. Whatever happened has happened and I would not like to go there. I am taking my share of the blame by dismissing his mains cable improvements and not concentrating well. Put it onto Covid that I had. He is welcome to comment if he wishes to. My motto has always been that there can be a good thing coming out of bad, and this has been an opportunity for me to look into something that I would normally have not done so.

Hi sahib

Look, you possibly made some mis-steps in analysis and conclusions wrt a wiring issue. Only way no to make errors is to do nothing. And it can be instructive as I read you as thinking. No one got hurt. It's all good imo. Hope you are over the COVID. Cheers.

 

[sahib]

Thank you Newmarket. That's how one learns, at least I do anyway.

I have a pretty bad dry cough and I get really tired quickly. So, I am currently working only a couple of hours a day.

However, in my view, that test is also not valid and does not represent mains hum. Here is the reason.

The signal source is the signal generator and the output signal is naturally referenced to the signal generator ground. Therefore the signal current will return to the source ground. So the mains earth path impedance is seen as open to it.

On the other hand, if the noise current (leakage or whatever), is generated by the mains then the noise current will look for a return path to neutral. The path is the safety earth as it is tied to neutral at distribution. In which case the mains earth impedance will have effect on the loop.

I am not exactly sure but I am brain storming at the moment. Also I don't want to end up re-discovering the wheel. So, tonight I am going to read Bill's paper (embarrassing that I have not read it yet). I'll also dig out Ott's and Morrison's book tomorrow and refresh myself.

 

[thor.zmt]

However, before continuing with the rest of the stuff I would like to mention that I have initially amended my first post with regards to Thor. Whatever happened has happened and I would not like to go there. I am taking my share of the blame by dismissing his mains cable improvements and not concentrating well. Put it onto Covid that I had. He is welcome to comment if he wishes to.

Thank you. You are a gentleman.

I was at fault for loosing my cool over something so trivial.

I can only blame my own own frustration for not being able to explain my point better, so it was correctly understood. After the fifth or sixth different attempt I started not to blame myself for my lack of ability to communicate clearly, but rather blamed it on prejudice on the other side.

That was not correct.

Thor

 

[thor.zmt]

I have googled hi-fi connectors etc and came across so called high-end brands like Cardas and ETI but none of them seem to have any technical data on their IEC plugs/sockets. They are very fancy looking things with all singing and dancing chrome plated covers and so on, and cost £150 - £160, yet Bulgin with less than 10milliohm contact resistance costs under £4. Same goes for the cables. Cardas has a say ground cable at 17.5AWG (around 1mm2) and I would like to know what makes it so high-performance than a good quality standard 1mm2 wire.

Hai Enten - Hai Fi (Shark Duck - it's only funny in German and travels poorly into english) is not a good place to start. A lot of things there ARE snake oil or poorly conceived.

What should be done is to define the requirements objectively and to then attempt to find suitable solution, preferably without great expense, off the shelf and from multiple vendors covered by a general industrial specification and not claims of "improved sound staging" and "blacker Backgrounds" (even IF and that is a big IF such descriptions are correct).

I am happy to pontificate upon and wax lyrically about my experiences at a later point, actually more likely I will be very dry and technical. And yes, connectors are part of the problem.

A good mains cable terminated into IEC and a high quality mains plug can easily come in at 30mOhm total.

And yes, cleaning mains contacts and pins on mains plugs is recommended (especially the brass on UK plugs, brasso is great), as is the use of protective gels coating contact surfaces (traditionally I use petroleum jelly aka Vaseline or even vaporub - the menthol in there causes no issues - but petroleum jelly is flammable, so perhaps synthetic inflammable oil, silicone gel etc.) and where possible the use of gold plated contacts (now we are back with Hai Entrn Hai Fi).

Now it's TGIF hour and Beer O'Clock and must take my leave.

Thor

 

[PermO]

Hi, cable guy here...

Interesting discussion, though a lot of it seems to go beyond my understanding.

I can give some real world examples of the systems I have build, I have not enigineered these systems, I'm the guy with the big toolbox in the orange boilersuit.

Analog radio distribution matrix Hilversum, main broadcast hub.
All signals are balanced, only pin 2 and 3 are connected using solid core twisted pair, unshielded, no sleeving, just wire.
No connectors are used !
The matrix cosists of a big metal frame that is filled with isolated "Christmas Trees" these are copper solder lugs that can hold many wires, shaped like a christmas tree, for each connection a "branch" is used, you solder the wire directly to an empty branch.
So there will be a Christmas tree that takes pin 2 connections and next to it another one that takes pin 3 connections.
Soldering irons are hanging from the ceiling ready to go.
Sorry, could not find any image of these style connections.
The rack equipment has a clean mains ground connection over IEC connector.
In each cabinet sits a big vertical copper busbar that will connect to all the equipment chasis.
XLR pin 1 is not connected.


On off shore riggs there's two ground busbars, PE and IE, mains ground and instrumentation ground.
These are connected at a single point at the seabed.
In this case mains ground is connected over IEC power connector and pin 1 XLR would be connected to instrumentation ground.
So these devices are build differently, there's a chasis ground and a isolated IE groundscrew that has no connection to chasis like you would have on common audio devices.
XLR connector could be used as they have an isolated pin 1 and a separate solder lug the connects to it's housing, in these systems these should not connect to each other exept for that single point down at the seabed.
Cabinets and chasis (metal work) should connect to mains ground.

Instrumentation ground is only used for screening low voltage signals.

Hope this helps.


Cheers !

 

[zamproject]

Sorry, could not find any image of these style connections.

From 78' Studer 289 lower frame assembly, final control room (RadioFrance) the -sapins- (christmas tree)
All IO including signalling and fader start soldered, no connectors

 

[Voyager10]

I'll re-post the link to the Bill Whitlock paper here: https://centralindianaaes.files.wordpress.com/2012/09/indy-aes-2012-seminar-w-notes-v1-0.pdf

Essential reading. The TL;DR version is that unbalanced connections = doom (but while you're here, take a look at these nice Jensen transformers!)

 

[sahib]

Thank you. You are a gentleman.

Thor

Thank you too and so are you. Glad all behind us now.

I also agree with your second post.

I have read Bill's excellent paper. I do not mean to be smug about it but the points I made are all in-line with his findings. Particularly introducing discontinuity.

However I have sketched out a new test set up to simulate mains born noise. I'll do it tomorrow and post my findings.

................
No connectors are used !
.............

John once said best capacitor is no capacitor. I'll re-phrase it as best connector is no connector.

But in the real world it is an evil necessity.

 

[thor.zmt]

Analog radio distribution matrix Hilversum, main broadcast hub.
All signals are balanced, only pin 2 and 3 are connected using solid core twisted pair, unshielded, no sleeving, just wire.

No connectors are used !

The matrix cosists of a big metal frame that is filled with isolated "Christmas Trees" these are copper solder lugs that can hold many wires, shaped like a christmas tree, for each connection a "branch" is used, you solder the wire directly to an empty branch.

So there will be a Christmas tree that takes pin 2 connections and next to it another one that takes pin 3 connections.

The rack equipment has a clean mains ground connection over IEC connector.

Yes, in East Germany we had similar setups. Of course Tuchel connector, not cannon/xlr. Ann all gear had transformer coupled out and in. A lot of broadcast gear even today is still transformer coupled.

One needs to be a bit careful with such setupa to retain electrical safety, a large technical organisation like public radio or a oil rig naturally can deal with that.

Equally, when implementing "local wiring standards", large organisations of course can test for compatibility, modify if needed and maintain such setups.

Naturally, the same is true in a small "DIY" studio where we build much ourself.

In each cabinet sits a big vertical copper busbar that will connect to all the equipment chasis.
XLR pin 1 is not connected.

This is critical. Busbar and a heavy duty wire to the chassis will minimise any differences between different chassis and minimises inter-chassis voltage differences and short circuits any interchassis currents via a very low impedance.

This parallels my suggestion of tying chassis together with tin plated copper mesh "grounding straps", one can also use very heavy gauge coaxial cables (RG213/U or RG214/U) coaxial cable, screen and center together with spade connectors and serrated washers (to get through the paint to the metal) under under a case screw.

Again, the idea is to get as close as possible to "zero ohm" between all the chassis. Doing that minimises any error voltages between the chassis and yes, then often pin1 can be unconnected.

On off shore riggs there's two ground busbars, PE and IE, mains ground and instrumentation ground.
These are connected at a single point at the seabed.

In this case mains ground is connected over IEC power connector and pin 1 XLR would be connected to instrumentation ground.

Again, similar principles. In Audio "instrumentation ground" would be "Audio ground" and naturally is separated from "power ground". I have seen similar setups in industrial setups for instrumentation (no XLR's of course, usually no connector).

So these devices are build differently, there's a chasis ground and a isolated IE groundscrew that has no connection to chasis like you would have on common audio devices.
XLR connector could be used as they have an isolated pin 1 and a separate solder lug the connects to it's housing, in these systems these should not connect to each other exept for that single point down at the seabed.
Cabinets and chasis (metal work) should connect to mains ground.

That is my suggestion, treat XLR as 4-Pin system with shell = earth, pin1 = Audio Ground, Pin 2/3 signals.

Thor

 

[thor.zmt]

I'll re-post the link to the Bill Whitlock paper here: https://centralindianaaes.files.wordpress.com/2012/09/indy-aes-2012-seminar-w-notes-v1-0.pdf

Very much worth reading, however avoid taking it as a gospel. For example BW pooh-pooh's bus-bars and shorting out inter chassis currents, when in practice it is an EXCELLENT mitigation strategy that works with off the shelf gear, as is keeping loop impedances in the mains wiring low.

In the end we usually deal with "generic" engineering that complies with electrical safety rules and is otherwise designed relatively "generic".

So we can expect single-Op-Amp balanced line receivers with 1% resistors (so 50dB typical CMRR on line in's and ~ +20dBU maximum level).



In this situation, to get 120dB SNR (8uV noise on output), noise voltages between pin 1 of the source and pin 1 of our receiver must be below 2.5mV!!!! To get to 130dB we need < 0.8mV!!!!

Class I (with 3-Wire power cable, earthed) appliances such as IT, movable and stationary equipment have a limit of 3.5mA earth leakage per device

Class II appliances (with 2-wire power cable, double insulated) – 0.25milliamps

For Class III appliances (operating on low voltage) the limit is 0.5milliamps

With a 1 Ohm loop impedance that permits less than 2.5mA (or 0.8mA) fault currents.

With a 0.1 loop impedance 25mA or 8mA respectively are permissible.

And for 0.06Ohm (which two good quality 6ft mains cables can achieve including contact resistances) ~ 42mA current are permissible for 120dB SNR and 13mA for 130dB.

Purely on resistor and Op-Amp noise dynamic range is ~132dB for this circuit @ +20dBu reference.

Where 2-wire powered devices are used or earth is lifted, the earth leakage currents flow in the signal cables where impedances are commonly relatively high and current flowing can also couple directly into signal conductors.

So under all circumstances the aim should be to avoid having noise currents circulating through the signal cable. In addition, using low "gound" resistance cables for signals can help.

So IN PRACTICE dealing with generic off the shelf equipment and standard code compliant house wiring, the easiest way to limit noise, is to limit interchassis potential, by minimising loop impedance between individual chassis.

This means bus bars, chassis grounding straps or supplementary earth connections for Class II equipment and/or low impedance mains cables. Doing so certainly can get us well into the -120dB...-130dB noise floor region using standard house wiring and unmodified generic equipment.

Thor

 

[sahib]

For the experienced this will be a bit of re-inventing the wheel. But I am hoping that it will bring further clarification to those who could not make out what we were talking about.

Again, just to clarify that, when I say ground loop I am not defining it, but meaning the loop formed by the ground paths between the two equipment.

So, the loop formed is shown in the below first diagram.

However, I have reinstated the mains safety earth impedance as part of the loop as this is what we are concerned with. In my previous test we discovered that the mains safety earth impedance had no effect and appeared as open when the noise source was the offending equipment itself.

Just to refresh our minds again that the Audio and Audio' are the audio circuit impedances. The noise voltage developed on the ground is imposed onto them and appears at the output with the actual signal. Strictly speaking, from analysis point of view we should be showing the output impedance of the circuit but am keeping it all out of this analysis now as it is irrelevant to what we are looking for.

I will not repeat what the impedances stand for as one can refer to the diagram in my first post. But briefly, Path A and A' are the internal ground path impedances of both equipment and all the others are the impedances of signal and mains earth wires and the contact resistances of the connectors.

The ground node shown between Contact B and Path A is where the signal/circuit ground of the equipment is tied to the mains safety earth, normally at a chassis stud point immediately after the IEC inlet,

The diagram on the left reduces to the diagram on the right when we add the impedances connected in series.

As mentioned before the mains safety earth is tied to neutral at the distribution point and earthed. The earthing is again irrelevant to us here. All we have to bear in mind that the safety earth wire has connection to neutral at this point.

So, I have put together a simple test set up as shown in the below picture, and a diagram to simulate the actual mains wiring.



I wired two transformers back to back and used the low voltage secondaries. TX1 represents the step-down transformer at the distribution point to step down the high voltage to consumer level (in UK). Point A is where the mains safety earth is tied to neutral and the node is earthed. Again, we are not interested in the earthing. I also used a fuse for safety.

TX 2 is the transformer of the equipment that is plugged into the mains. Point B is again where the circuit ground is tied to mains safety earth at the chassis stud point.

Assume that a mains born noise voltage is developed at point B through source impedance R1. The current will seek for a path to neutral through the mains safety earth. However, since R3, R5, R3' and R4' are all connected in series, and the total resistance is in parallel to R4, the noise current will also flow through this path, offending the other equipment too.

The circuit simplifies to the one in the below diagram.

Here the ground loop impedance is sandwiched between the noise source impedance and safety ground impedance, and forms a voltage dividing network. R1 and R2 are fixed and there is nothing we can do about them. But if we reduce R3 then the noise voltage developed at Point B, will also reduce. So, clearly what Thor had suggested holds.

However, it is clear that we are limited by the mains safety earth impedance. So, there comes a time, particularly multiple equipment are interfaced, the only option open to us is to bring discontinuity to the loop to be able to eliminate the hum. In other words, cutting the physical hard wiring. We can not disconnect the mains safety earth as it is illegal, but more importantly it is there to save our lives. I do not know what the regulations say about isolating the signal ground from the safety ground inside the equipment. So, this could be done on the signal connection between the two equipment by introducing an isolation transformer which Bill's paper explains expertly.

In terms of verifying my test, I have used 2k resistors throughout.

R3', R5, R3' and RR4' are all in parallel to R4. So, R4 = 8k || 2k = 1.6k.

The coupled noise voltage at point B is 10.8VRMS with reference to point A.

Plugging in 1.6k at the high side and 2k on the low side of voltage divider gives 6V at point C. I measured 5.9V. This agrees that the noise current is flowing through mains safety earth and into neutral.

Finally, this is just one scenario. Things can and do get much hairier than that.

 

[thor.zmt]

So, the loop formed is shown in the below first diagram.
View attachment 114083

Yes. And to be clear, the problem is ANY voltage that appears in effect between Contact C and Contact C' and thus any current flowing in "Screen".

The triangle hanging off the whole is meaningless. If it is a second "Earth" - it should connect to either where Contact A and Contact A' join (the correct way).

Alternatively it should connect somewhere ultimately to mains earth with the possibility of quite large fault currents flowing between "Mains Earth" and "Triangle". However that is the obviously WRONG way anyway. And hopefully everyone knows that.

What I am missing in the whole set of diagrams are the sources of noise currents and their reference to mains earth. Current flow in loops. Your diagram misses in effect both the loops and noise sources. May I suggest we clarify as follows:



Only current flowing in "Screen" is a problem for audio. I labelled this current in Orange Iaudionoise.

Anything appearing anywhere else of zero consequence to the audio signal (but may be an electrical safety issue). OF course, there is still current flowing.

But if that current does not flow in "Screen" and thus there is no voltage difference - there is no problem.

We can have 10A each flowing in Path A & Path B (Inoise * Inoise') and have a 10 Ohm mains safety earth impedance (which is distinctly unsafe), HOWEVER if Path A to Contact A and Path A' to Contact A' have "zero ohm" all current will flow there and there will be no possible current flow in "Screen" and no possible voltage between Contact C & Contact C'.

Moreover, it can be demonstrated that current flowing in "Screen" and thus the voltage between Contact C and Contact C' are the result of the relative impedances of Path A to Contact A, Path A' to Contact A' and Contact C to Contact C' AND the difference in amplitude and phase of the noise current injected into the system at (for argument's sake) where Path A & Contact C connect (Noise Current I) and Path A' & Contact C' connect (Noise Current I').

The sum of Noise Current I and Noise Current I' will flow through Mains safety earth impedance and thus drop a voltage that creates a voltage at the point where Contact A and Contact A' join, however varying "Mains safety earth impedance" across a relative wide range of values will not alter the current flowing in "Screen" and thus the voltage between Contact C & Contact C'.

HOWEVER, varying the impedance, respectively of the whole chain between Path A and Contact A or (and) the whole chain between Path A' and Contact A' will alter the current flowing in "Screen" and thus the voltage between Contact C and Contact C'.

Can we agree on that?

Thor

 

[sahib]

Firstly all agreed.

May be I should have gone into the detail you have, but since you have explained it well, the job is complete.

The triangle hanging off the whole is meaningless. If it is a second "Earth" - it should connect to either where Contact A and Contact A' join (the correct way).

Indeed. Lousy of me. Although I had pointed out that the ground node shown between Contact B and Path A is where the signal/circuit ground of the equipment is tied to the mains safety earth, normally at a chassis stud point immediately after the IEC inlet, I should have been clearer on the diagram.

But if that current does not flow in "Screen" and thus there is no voltage difference - there is no problem.

Indeed. The other way of looking at is that, it is a condition when both the left and right ends (or sides) of the screen are at the same potential. But I am not sure how possible could this be in real life. Even if we managed identical impedances at Contact B / Contact B' and Contact C / Contact C', as they are external and within our control (to a great degree), Path A and Path A' impedances which are internal to both equipment will be different, hence introduce imbalance. Would that not be correct?

 

[sahib]

Not sure what you would think of this but after I posted above, below occurred to me.


Connecting the screen directly to the common ground point (chassis stud point) isolates Path A and Path A', and seems to merit to something at least on paper (though, this again constitutes a modification to an existing equipment).

 

[thor.zmt]

Not sure what you would think of this but after I posted above, below occurred to me.

View attachment 114113
Connecting the screen directly to the common ground point (chassis stud point) isolates Path A and Path A', and seems to merit to something at least on paper (though, this again constitutes a modification to an existing equipment).

What you are showing is in effect the AES recommendation, AES48-2005.

However, given that the screen is not very low impedance, it is still much easier to provide a low impedance path external to the cable. Like so:



The busbar or grounding straps link chassis with very low impedance, they usually work well at RF as well as audio frequencies. The busbar / grounding straps then also link hard to the incoming mains Earth, mains cables are now less critical, but using low impedance ones can still lower noise more.

Grounding straps:


Thor

 

[sahib]

Thank you for that. I am not a member of AES. So good to have that paper.

Once I finish the job I am working on I have a couple of things in mind that I'll try.

Great stuff.