CM noise problems with ethernet connected instruments

[temperance]

Hi,

perhaps not a metrology question but I think people in here might have some experience with this.

The problem:
all my instruments are equipped with Ethernet. Very practical. But with all the equipment connected onto a router I start to have some problems with common mode noise to the extend that even a simple scope measurement is problematic.

What I tried:
-replaced the router with a transformer power supply powered router from way back. It did help a bit it's still but not good enough.

I noticed that it must be one specific power supply which is responsible for most of the CM noise. Investigating further, it turns out to be the siglent signal generator. With the signal gen disconnected from the router the noise is much lower. But, still not good enough. Now I'm seeing random stuff from all the other instruments and the computer attached to that same router.

I'm wondering, pondering how to solve this. Anyone with some experience in here?

[jonpaul]

insert transformer isolation, depending on the CAT and rate.

See "Ethernet Transformer" from Coilcraft, Renco, Pulse Engineering


Jon

[langwadt]

insert transformer isolation, depending on the CAT and rate.

See "Ethernet Transformer" from Coilcraft, Renco, Pulse Engineering


Jon

ethernet is already transformer isolated...

[temperance]

Indeed, Ethernet is isolated with transformers. Or did you mean something different when you said to use transformers?

[dobsonr741]

You might have conducted EMI creeping along the cable. Did you try shielded eth cable? Not for permanently, but for testing a ferrite clamp-on?

[David Hess]

ethernet is already transformer isolated...

And at both ends of every cable.

[jonpaul]

The Ethernet RX have poor CM noise rejection, hence the use of the transformers.
commodity Ethernet transformers are toroids encapsulated in plastic. Optimized for lwest cost perhaps $0.20 ea

The primary- secondary capacitance can be 15 pf..100 pF.

Inserting a low capacitance shielded transformer in series with the connection at the RX end will reduce CM noise.

Jon

[temperance]

-Thanks for the explanation about the low cost Ethernet transformers. Isn't the capacitance caused by the transmission line impedance requirement? I will have a look at some more expensive transformers.

-Conducted noise creeping in: I will try some ferrite clamps. The noise is indeed high frequency. <100MHz. I will take some measurements and try to order some suitable CM ferrite clamps.

[David Hess]

The Ethernet RX have poor CM noise rejection, hence the use of the transformers.

The transformers are used because some type of galvanic isolation is required for any networking connection which could reach between ground domains, which is definitely the case with 100 meter Ethernet.

[thermistor-guy]

...
The problem:
all my instruments are equipped with Ethernet. Very practical. But with all the equipment connected onto a router I start to have some problems with common mode noise to the extend that even a simple scope measurement is problematic.
...

In my personal lab, I have standardized on ethernet for connecting test equipment. Cables are cat5e or cat6 UTP. I add serial-to-ethernet and GPIB-to-ethernet converters
to old equipment when needed.

When running a test, I set up the lab as follows:

* turn off all the lights;
* turn off any unneeded equipment (including PCs);
* remove or power down any wifi equipment and mobile phones;
* close the lab door and run the test remotely from another room.

An rpi (with wifi disabled) in the lab controls the test instruments.

Try running a minimal setup, one or two instruments powered, with as quiet an EMI environment
as you can manage, and see what you get.

Maybe the source of your problem is something else - LED or fluorescent lights for example.

[T3sl4co1l]

Yeah, you can get lower capacitance transformers, also with higher isolation voltage ratings -- not that the latter is necessarily helpful, but the increased distance between windings also tends to reduce isolation capacitance, so it may be some help.

Shielded cable won't do you any good because you're just replacing CM over twisted pair with CM over shield, plus inviting ground loop problems on any units with shielded connectors.  Shielded Ethernet is only meaningful (with respect to this concern i.e. conducted emissions within the network) when the whole system (everything on the network) is designed around it.  (...Which really isn't saying much of anything, because "designing the whole network" implies the ability to reduce emissions of any given component in the network anyway, and, when is that ever true?)

(The other thing that shielded Ethernet does, is reject external noise from the communications within, i.e. improve CMRR.  That's evidently not a problem here, so isn't a motivation to choose it.  You'd be looking at that for, like, 30V+ (or 30V/m+ radiated) noise levels, like in super nasty industrial environments.)

As for potential offenders, I've noticed a bit of EMC on my bench before.  After shuffling through a few things, I traced it to crusty connectors on the USB hub -- contact cleaner and a bit of cycling later, and it went away.  I've used the same hub for many years, so it figures the connectors are starting to dry out, or succumb to fretting or whatever.  If you can figure out if the noise seems like power supply noise (switching peaks, or modulated by line frequency or varying load current), communications (erratic, varies with activity?), video (buzzing?), etc., that might give some clues where to look.  As in my case, note that all connectors are eventually suspect -- USB especially as they're not anchored or anything (just retained with spring tension), but so too is HDMI for example, and even screw-retention plugs like VGA or [classic] serial can flex enough to wear out the mating surfaces.

Tim

[miro123]

first try to find the source of CM.
Then try to solve it. If not try to reduce it at the source side not at the end.
My setup
Fiber connection get cheaper
1. My lab is connected with fiber to the outside world.
2. I have spare SFP hub and two converters. - https://www.tp-link.com/nl/business-networking/accessory/mc220l/
3. Use high quality switches - I assume that you are talking to switch not router. Am I correct? - I use HP Procure1800-8G for 10..15years. I had many problems with cheap tp-links and d-linkhttps://support.hpe.com/hpesc/public/docDisplay?docId=c01955915&docLocale=en_US
4. Last but not least -  I have plenty of spare CM chokes on my desk

[jonpaul]

Reading  the comments and OP, seems some confusion re Ethernet transformers, CM vs DM noise and solutions.

The commodity transformers ( often embedded in an RJ-45 module) are for galvanic isolation, and rated 1.5 kV P-S insulation. They are not optimized for Cp-s nor for CMRR.

Noise on a long TP cable is mostly CM but DM is possible and  depends on the cable type, and run legnth.

Beads can be useful for NM or DM noise and CM chokes have some DM L as well.

Excellet discussion of the sources, theory and practive of noise reduction in differetial transmission systems redily avail

See Belden Cable, app notes by Steve LAMPEN (Belden.com) and 1988 version of our old friends book   Henry OTT, Noise Reduction.
https://commons.princeton.edu/motorcycledesign/wp-content/uploads/sites/70/2019/08/otto_1988_noisereduction.pdf

Also any good book on EMC.

Jon

[temperance]

thanks everyone for the useful hints and tips. Isolating some parts by means of optical cables sounds like a good idea. I didn't know about the existence of those small media converters.

I'm going to investigate step by step where the excessive noise is coming from. For now I know that the siglent function generator seems to be generating a lot of noise. Connecting shorted probe to the output GND shows a lot of noise and ringing. Disconnecting the Ethernet cable reduces the noise quite a bit. But, the source might still be somewhere else.

My routers are just some low cost tplink devices. For normal household use they are fine and don't break down (one is perhaps 12 years old now). For now it's not sure if they are the cause of the problem.

I'll get back to this topic when time permits to investigate this further.

[jonpaul]

The isolation trsf are the most likely solution, cheap and easy.

You need two of them.

Jon

[tkamiya]

All it takes is one device with poor design/construction to compromise the galvanic isolation though.  If it's the switch itself, the whole idea is shot.  We are talking about millivolts here....

I would disconnect one device at a time and see where the problem is happening.  Guessing at it will quickly get expensive.  For a complete solution, if you cannot resolve the problem, I would try media converter and insert an optical link in the connection for the line going to the cause you've isolated.

Good luck!