Preventing ground loops on shielded twisted pair (STP) cabling

[Zeyneb]

Dear people,

I think I have to try my luck here to find an answer. As I have asked about the following on Network Engineering Stack Exchange but my question was discarded because they don't allow topics on home networks.

Maybe for my home network I don't really need STP. But still I would like to know how to properly build such a network. And I'm actually implementing this with STP cable.

Regarding grounding of STP cables I was facing with contrasting recommendations. I believe this topic is much more general than Ethernet networking. One camp claims you should only ground on one end of the cable to prevent a ground loop. The other camp claims to ground both ends in order to have an effective way to prevent EMI entering the cable. To be sure with a ground loop for my case I mean a loop in the ground/earth wire of my residential mains electricity installation. And I think it is useful for this case to state that I live in the Netherlands.

After searching around the internet I found the following question on Network Engineering Stack Exchange valuable:
https://networkengineering.stackexchange.com/questions/6773/shielded-twisted-pair-stp-termination-and-grounding?newreg=87f74fa1868f434eb44dd79f275e5602

And especially the answer from Mike Pennington there. I've also read his linked powerpoint: http://www.pennington.net/archives/cabling/BICSI_Mythbusting_Shielded_Cabling.pdf

Based on this I was drawn to the conclusion to ground the STP shield only on one end. Which means for my case that only the patch panel side will have connections from the STP shields to the breaker box ground which in turn is connected to the ground electrode of the apartment building I live in. I hope the majority of the Eevblog audience agrees with this plan.

Now leading to my question. Regarding the wall plate side where you should interrupt the ground connection. What products should you use for that? Obviously you do buy an STP cable and practically this is F/UTP or SF/UTP. But I'm worried that I will make an accidental ground loop with the shielded keystone jacks I currently have. You know, you can connect a shielded patch cable from the wall plate to a computer and for the computers I have they all make electrical continuity with the network port shield and the power plug ground. For the assembly of the cable to the shielded keystone I did cut away the foil, braided shield wires and the drain wire. Now the shielded cover is only attached to cable jacket. But I believe I'm better off buying the more regular plastic UTP keystones (for the wall plate side) even for this STP network. However a proper strain relief is still a good thing to have in my opinion.

I'm curious what other people think of this.

[David Hess]

You pretty much covered it.  Analog Devices makes the same recommendation not to connect the shield at both ends.

If you did connect the shield at both ends, then the high frequency common mode current from various switching power supplies which is coupled through their input RFI filters into the ground connection would be applied across every cable shield.

I would use an unshielded plug on the side where the shield is not connected to indicate how it is actually wired.

[Zeyneb]

Hi David,

Good to know reputable sources share the approach to only ground on one end.

In your last sentence you talk about an unshielded plug. I think you actually mean an unshielded keystone jack. The one that is put in a wallplate. As it is common for my nanny state that on the living space side you can deal with the most naive people. Who don't care to plug in an shielded or unshielded patch cable.

[David Hess]

In your last sentence you talk about an unshielded plug. I think you actually mean an unshielded keystone jack. The one that is put in a wallplate. As it is common for my nanny state that on the living space side you can deal with the most naive people. Who don't care to plug in an shielded or unshielded patch cable.

I am not used to thinking in terms of keystone jacks but that seems right.  I usually terminate the cable in a plug and use a wall plate with a socket on each side.

[Zeyneb]

Hi there!

So I was looking around at Digikey for a suitable unshielded keystone jack and found the following:



It's from Assmann WSW Components, part A-KEY-8-EG-GIB-WIW2.
Digikey product page:
https://www.digikey.com/product-detail/en/assmann-wsw-components/A-KEY-8-EG-GIB-WIW2/123-A-KEY-8-EG-GIB-WIW2-ND/10392871

The thing is specified for Power over Ethernet PoE. However I want to use it for normal Ethernet, either 100BASE-TX or 1000BASE-T.

So what is this PoE thing doing to my Keystone? Is it just that voltage rating and/or dielectric strength is higher than usual?

[reboots]

So I was looking around at Digikey for a suitable unshielded keystone jack and found the following:

The thing is specified for Power over Ethernet PoE. However I want to use it for normal Ethernet, either 100BASE-TX or 1000BASE-T.

So what is this PoE thing doing to my Keystone? Is it just that voltage rating and/or dielectric strength is higher than usual?

I believe this is a standard CAT 6 jack which happens to be designed and rated suitable for PoE use, but is not intended exclusively for that application.

[T3sl4co1l]

Regarding grounding of STP cables I was facing with contrasting recommendations. I believe this topic is much more general than Ethernet networking. One camp claims you should only ground on one end of the cable to prevent a ground loop. The other camp claims to ground both ends in order to have an effective way to prevent EMI entering the cable. To be sure with a ground loop for my case I mean a loop in the ground/earth wire of my residential mains electricity installation. And I think it is useful for this case to state that I live in the Netherlands.

Note that neither camp is wrong.  But their actions are exclusive.

If you are leaving one end open, you are letting in 100% of the EMI on the cable.

So why even bother getting the shield in the first place?  It's literally not doing anything.

The only thing a single-grounded shield can protect against is magnetic induction between the signal lines, which the twisted pair already deals with reasonably well, and electric induction, which the differential lines don't care about.  It cannot deal with electromagnetic (wave) induction, i.e. where the offending frequency is higher than the electrical length of the cable.

To avoid ground loop, you might compromise by using bypass capacitors around the ground.  This retains galvanic isolation (low frequency -- mains and DC -- ground loop voltages are allowed) while shunting EMI.  But this isn't a connector-level thing you can do, and I'm not aware of anything other than board-level design, that handles this.


Points against the ground loop camp are:
1. Who cares if the EMI is doing ground loops?  If the shield is complete, EMI isn't getting into your signals!  It's fine, end of story.  (Resonances are common -- the low impedance connections allow cables to act as resonators -- but this is easily tamed with a ferrite bead or two.)
2. If you have ground loops that are so bad they're actually causing problems, you have bigger concerns than mere network wiring -- namely, your building wiring is fucked and needs to be repaired immediately before the defective grounding starts a fire or something!

Note that #1 is predicated on what is "fine".  Ground loops are offensive in audio applications because the standard signaling method is single-ended, and the signal is low frequency where shields are ineffective.  That is, the shield's resistance dominates, and ground loop and signal currents are allowed to mix.  The resulting voltage drop is additive to the signal.

This goes away with a differential signaling method, like XLR audio jacks, or Ethernet which is differential and high frequency (AC, transformer coupled).

Tim

[ve7xen]

If you are leaving one end open, you are letting in 100% of the EMI on the cable.

For an explanation:

https://www.emcstandards.co.uk/cable-shield-grounded-at-one-end-only

It's not really doing nothing, but it's not doing much for the interference you care about, as T3sl4co1l points out.

Maybe for my home network I don't really need STP.

Almost no networks need STP. Ethernet over UTP is very robust, and STP would only be necessary in a truly horrific EMI environment. And it creates a lot of complications, you need special patch panels, special keystones, special cables, all of which are much more expensive... Then you need to actually think about your ground system and make sure everything is bonded correctly. Really not worth it. The discussion here just proves this point. If you are working in an environment where STP is actually used/required, you should get domain specific training.

[Ranayna]

Hi there!

So I was looking around at Digikey for a suitable unshielded keystone jack and found the following:

(Attachment Link)

It's from Assmann WSW Components, part A-KEY-8-EG-GIB-WIW2.
Digikey product page:
https://www.digikey.com/product-detail/en/assmann-wsw-components/A-KEY-8-EG-GIB-WIW2/123-A-KEY-8-EG-GIB-WIW2-ND/10392871

The thing is specified for Power over Ethernet PoE. However I want to use it for normal Ethernet, either 100BASE-TX or 1000BASE-T.

So what is this PoE thing doing to my Keystone? Is it just that voltage rating and/or dielectric strength is higher than usual?

If that thing is explicitly rated for PoE, it might (or rather should) have a couple of "flaps" on the pins. Those are supposed to help reduce arcing when you unplug, which can actually be an issue, especially with the high powered new standards where you can supply up to 90 watts.
In any case, it will work for normal, non PoE network.

Regarding STP: In my experience (and i am a network admin), connecting the shield on both ends should not give you any issues. In our network, we exclusively use CAT 6a rated cabling (actually CAT7 S/FTP cables, with CAT6a keystones), which have the shield connected on both ends. In our server network, we also use exclusively CAT6a S/FTP patch cables. So our shields are connected, and can utilize 10GBaseT without issues.
For the clients we use a mixture of STP und UTP patch cables.

[Qmavam]

Since this seems to have run it's course, I'll add my question.
 I'm trying to decide whether to use UTP or STP Cat6.
But, I have an off label use, I'm using it as a feedline for
a directional MW/HF antenna, specifically because of the
twisted pair rejection of external RF.
My feedline is 230 ft long. And my question comes down to
If I use STP am I possibly adding a source of noise with the shield
by grounding one end, both ends?
                           Mikek

[PixieDust]

Almost no networks need STP. Ethernet over UTP is very robust, and STP would only be necessary in a truly horrific EMI environment. And it creates a lot of complications, you need special patch panels, special keystones, special cables, all of which are much more expensive... Then you need to actually think about your ground system and make sure everything is bonded correctly. Really not worth it. The discussion here just proves this point. If you are working in an environment where STP is actually used/required, you should get domain specific training.

I would like to run an ethernet cable into my workshop. I've got one of those 2.2kW Chinese spindles and VFD (Not me, just a video that shows the equipment in question):



Haven't been able to find much information in terms of how much radiated EMI these things likely make. According to:

https://www.researchgate.net/publication/392206796_Investigation_of_Radiative_Interference_in_Variable-Frequency_Drives_VFD

it looks like cables running from the VFD and the spindle produce next to no EMI if shielded. So the culprit for EMI is the VFD itself. It seems that even a simple wire mesh can reduce the EMI quite respectably. And in the YouTube video I linked @3:16 - 5:06, he shoved the whole VFD into a metal box, so that solution will I would imagine get rid of most of EMI. All this leads me to believe that it shouldn't be too difficult to avoid any EMI issues when running ethernet into the workshop as long as common sense is used i.e. simply routing the ethernet cables as far away as possible from the VFD as possible?

[Halcyon]

Speaking from 25+ years in the IT industry (in one capacity or another), best practice is to ensure continuity of the shield from device all the way through the ground bonding at the patch panel end. If the shield is only connected to ground at one end, it's not doing its job. Professional installers will typically ensure ground continuity.

"Traditionally" the advice was to connect only one end, but I don't really understand the purpose behind this as you might as well just use UTP. Some people cite EMI reasons, but I don't buy it in most installations. Ground loops are another thing that is commonly referenced, but it's not really relevant, we aren't talking about unbalanced analog audio here.

If you aren't going to bother installing STP properly, you're just wasting money.

This explains it better than I can: https://www.ampcom.com/blogs/industry-insights/stp-grounding-best-practices

[David Hess]

"Traditionally" the advice was to connect only one end, but I don't really understand the purpose behind this as you might as well just use UTP. Some people cite EMI reasons, but I don't buy it in most installations. Ground loops are another thing that is commonly referenced, but it's not really relevant, we aren't talking about unbalanced analog audio here.

Your link mentions the exception:

For Ethernet STP, the preferred practice is proper shield termination at both ends within an equipotential bonding network.

Industrial applications between buildings can lack an equipotential bonding network, and sometimes within the same building.  Fiber solves this problem, but used to be much more expensive.

The audio exception includes many types of low noise instrumentation.  Twisted pair Ethernet uses transformer isolation on both sides reducing common mode susceptibility, but without this, current noise through the shield couples inductively to the conductors and can be a major noise source; connecting the shield only at the receiver end solves this while still reducing electrostaticly coupled noise.

[Halcyon]

"Traditionally" the advice was to connect only one end, but I don't really understand the purpose behind this as you might as well just use UTP. Some people cite EMI reasons, but I don't buy it in most installations. Ground loops are another thing that is commonly referenced, but it's not really relevant, we aren't talking about unbalanced analog audio here.

Your link mentions the exception:

For Ethernet STP, the preferred practice is proper shield termination at both ends within an equipotential bonding network.

Industrial applications between buildings can lack an equipotential bonding network, and sometimes within the same building.  Fiber solves this problem, but used to be much more expensive.

The audio exception includes many types of low noise instrumentation.  Twisted pair Ethernet uses transformer isolation on both sides reducing common mode susceptibility, but without this, current noise through the shield couples inductively to the conductors and can be a major noise source; connecting the shield only at the receiver end solves this while still reducing electrostaticly coupled noise.

Absolutely true!

There will be some edge cases where environmental issues could cause issues, but it's also important to remember that the cabling standards for twisted pair Ethernet is rated to 100 metres. In corporate and industrial environments, network closets servicing end-user devices and nodes are commonly seperate to the main server room because of this cabling length limitation. These kinds of environments (along with STP installations) should be properly planned, taking into account the environment, ground points etc...

Fibre and SFP/SFP+ modules these days are cheap as chips, even for the "homelab" user.

[ejeffrey]

Your link mentions the exception:

For Ethernet STP, the preferred practice is proper shield termination at both ends within an equipotential bonding network.

I don't understand this point, at least in the context of Ethernet.

What is the advantage in that situation of running STP without grounding vs just using UTP?  The point is that for runs longer than lambda/4 an ungrounded shield isn't a shield it's an antenna.  Ethernet is almost always going to be in that limit.

I've seen designs for non Ethernet applications where the shield is capacitively coupled to ground to prevent large DC and 60 Hz current while still providing rf shielding  that makes sense to me but leaving one end disconencted entirely does not.

The audio exception includes many types of low noise instrumentation.  Twisted pair Ethernet uses transformer isolation on both sides reducing common mode susceptibility, but without this, current noise through the shield couples inductively to the conductors and can be a major noise source; connecting the shield only at the receiver end solves this while still reducing electrostaticly coupled noise.

Sure, for short runs of balanced audio frequency signals a shield connected at one end can provide electrostatic shielding.  It's best if the signal path has rf termination and low pass filtering in the band where the shield won't be effective.  But that doesn't really ever apply to Ethernet.

[tatel]

Some 15 years ago, our ancient (built in the 1930s) neighborhood decided to install a wifi network to share internet connection, using ubiquiti gear. Ubiquiti manuals advise having good ground -protective earth, should I say. We used S/FTP cat 5e cable from antenna's POE to the antenna; UTP for almost anything else. S/FTP cables got ground connected to both sides.

Since it's a POE thing, I guess probability of having ground loops is pretty low.

Since in reality this matter was more about protective earth, it turned out we needed to check if the plug for the POE really had good PE, otherwise antennas could suffer sudden death. In the worst cases -buildings where an electrician lived, curiously- we had to change like an antenna every month, until the electrician was threatened to check the electrical installation or else.

Of course, these old buildings, particularly in the space -not used- directly under the roof, had no protective earth in the very dubious electrical installation (usually just a switch, some plug and a bulb) made by some people dwelling in that building, god knows how many decades ago, didn't have any PE up there, but we found some cases where an entire building had dodgy PE. In the first three months, antenna sudden death ratio was about 10%

After making sure PE was good everywhere, sudden antenna deaths became pretty rare.

Since then, I usually connect to ground, both sides of any S/FTP cable. No ground loop problem whatsoever was ever detected.

[David Hess]

Your link mentions the exception:

For Ethernet STP, the preferred practice is proper shield termination at both ends within an equipotential bonding network.

I don't understand this point, at least in the context of Ethernet.

What is the advantage in that situation of running STP without grounding vs just using UTP?  The point is that for runs longer than lambda/4 an ungrounded shield isn't a shield it's an antenna.  Ethernet is almost always going to be in that limit.

The problem comes about where "ground", even within the equipotential bonding network as they call it, has different potentials.  If the STP is grounded at both ends, then considerable ground currents can flow through the shield and inductively couple to the signal wires.  In context they are referring to situations where equipment at both ends is powered by separate distribution panels which have separate grounds.

High power AC loads can couple huge amounts of noise into ground through their y-class EMI suppression capacitors.  In the past, I have encountered office equipment which managed it, especially laser printers.

I've seen designs for non Ethernet applications where the shield is capacitively coupled to ground to prevent large DC and 60 Hz current while still providing rf shielding  that makes sense to me but leaving one end disconencted entirely does not.

I have used capacitive coupling that way in specific applications.  In other cases, I connected the shield at only one end, or did something even wilder.

By design Ethernet has considerable noise immunity, but in extreme cases, even it needs help.

Some 15 years ago, our ancient (built in the 1930s) neighborhood decided to install a wifi network to share internet connection, using ubiquiti gear. Ubiquiti manuals advise having good ground -protective earth, should I say. We used S/FTP cat 5e cable from antenna's POE to the antenna; UTP for almost anything else. S/FTP cables got ground connected to both sides.

Since it's a POE thing, I guess probability of having ground loops is pretty low.

Since in reality this matter was more about protective earth, it turned out we needed to check if the plug for the POE really had good PE, otherwise antennas could suffer sudden death. In the worst cases -buildings where an electrician lived, curiously- we had to change like an antenna every month, until the electrician was threatened to check the electrical installation or else.

Of course, these old buildings, particularly in the space -not used- directly under the roof, had no protective earth in the very dubious electrical installation (usually just a switch, some plug and a bulb) made by some people dwelling in that building, god knows how many decades ago, didn't have any PE up there, but we found some cases where an entire building had dodgy PE. In the first three months, antenna sudden death ratio was about 10%

After making sure PE was good everywhere, sudden antenna deaths became pretty rare.

Since then, I usually connect to ground, both sides of any S/FTP cable. No ground loop problem whatsoever was ever detected.

A protective ground is required for antennas and especially antennas integrated with network equipment to sink static buildup, which can be considerable during storms or even wind.  For Ethernet connected RF equipment without a protective ground connection, high voltage breakdown of the isolation transformers at both ends of each twisted pair will break things.  STP cable can provide this protective ground.

[Ranayna]

Some 15 years ago, our ancient (built in the 1930s) neighborhood decided to install a wifi network to share internet connection, using ubiquiti gear. Ubiquiti manuals advise having good ground -protective earth, should I say. We used S/FTP cat 5e cable from antenna's POE to the antenna; UTP for almost anything else. S/FTP cables got ground connected to both sides.
[...]

I think it was Ubiquiti that once had a "small" wiring issue with one of their access point lines
If you used STP cables with the shield fully connected, the AP would reset...
https://help.ui.com/hc/en-us/articles/8823742725015-UniFi-6-Access-Point-Professional-U6-Pro-Advisory