Ethernet Cable Shield Grounding

Hi Dick,

I agree that grounding both sides of the cable is the probable cause of poor performance. A rule of thumb is usually to only ground one side for this very reason (preventing ground loops). 

Have you tried only grounding one side? That's probably the first thing to do in order to troubleshoot the problem. If you have tried this and are still experiencing the problem, let us know so we can dive in further. 

Thanks,
Jason Blackman 

Jason - Thanks very much for the response.

The (umbilical) cable in question is 100 feet long, and goes between a portable "cart" containing laptops, test equipment, etc. and an aircraft on the tarmac.The cart is earth grounded to a ground stud in the tarmac and the aircraft has its own tarmac ground stud. Note that the cart is powered by a UPS.

Given the above, is it best to ground the Ethernet shields at the cart end, or the aircraft end, or doesn't it matter? Or should it be determined via testing?

The overall umbilical shield needs to be grounded at both ends for EMI/EMC purposes. However, the internal Ethernet cable shields can be grounded as required for optimum performance.

It's worthwhile noting that sometimes the Ethernet performance is very good, and sometimes it's very poor. I believe that a ground loop can result in erratic performance. Do you concur?

To date, we haven't tried just grounding one end of the Ethernet shields. To do this would require some major surgery on the cable.

Can you suggest any documentation and/or papers, either from TI or elsewhere, stating that grounding the shield at one end is the proper technique?

Before I "go to war" with the cabling group, I want to have done my research, have evidence that a change is needed, and be prepared to support this position, as I expect some reluctance to make any modifications to the cable.

Thanks again for your support,

Dick Berg

I remember very well an installation I did for a beverage bottling company with many controllers interconnected with RS485. The technician did connect the cable to ground at each closet and I simply DIDN'T get any communication at all. Only when I grounded at the star center ONLY (where the central computer was), avoiding multiple ground connections, we got excellent results.

I think that you could make a small box or jig with two back to back LAN connectors that can disconnect the ground in any one of the sides of the cable and experiment, since I am not sure what configuration is best (airplane or cart). That way you could test and show your coworkers if you get good results, without making major surgery to the cables themselves.

Dick, I like the idea provided by Albert_83 with inserting back-to-back connectors. if you ethernet net cable is not isolated, perhaps also insert isolation transformers in the back-to-back connection. As you already stated the existence of ground loops due to grounding at both ends can cause faulty data transmissions. The reason for that is, if you have large ground potential differences between airplane and cart, grounding the shields at both ends can cause high loop currents as you shorten differenet ground potentials. Any motor noise from tools or other which connect to this ground potential ,can modulate the groundloop current  which shows up in your differential data.  Hence isolating ground and signal lines get rid of the common-mode problem.

Now you only need to isolate the cable ends at the airplane. At the cart you don't need isolation. Here grounding of all shields at the cart will actually help you prevent static charge build-up on the cable.

If you were to isolate signal and ground at both cable ends (so cart and airplane) you still would have to provide some controlled R-C discharge for the ground wires to the cart ground to avoid static build-up.

Hope this helps. - Thomas

Wire and Cable Shields Grounded at One End Only

The myth of actually shielding with the “shield” grounded at one end only (SPG) has spread throughout the EE design community.  Pseudo-technical argument instills fear of ground loops formed by cable shields grounded at both ends.  See Analog Devices, Analog Dialog 17-1, 1983, for one of the most asinine drawn out of rationales.  The ultimate insanity is when coaxial signal line shields are grounded at one end only.

A cable shield grounded at one end (SPG) behaves like a low pass filter to magnetic fields and a high pass filter to electric fields.  A shield it is not.  The SPG shield actually amplifies EMI at the ¼-wave resonance of the shield with Q ≤ 10.

In fifty years’ experience, this author has trouble shot too many failing systems where the coax signal line shields were left open at one end.  Grounding the coax shield to the line driver and receiver signal return pins with no break in continuity in between killed the noise problem every time.

Two points: (1) A shield installed for shielding purposes only must be grounded at both ends preferably to the box connectors.  (2) A coax signal line uses the shield for signal return therefore must be grounded to the line driver and receiver signal ground/common pins with no breaks in continuity.  If there is an EMI noise problem with this, the cable should be replaced with a triaxial cable with the outer shield grounded to chassis at both ends or a twinax.

There is a real problem grounding any conductor including shields between two units operating off of two power mains in or between buildings.  Ground potential differences can be kilovolts.  Ethernet has common mode in-line isolation and capacitors to chassis ground rated at 1-2kV.  If a shield is to be installed on such wiring, one unit or the other should be floating with respect to its local ground and referenced back to the other unit or the signal lines should have isolation installed such as optical, inductive, capacitive, or resistive so as to reduce the induced common mode EMI.  The shield is only one issue.

With respect to the ground loop induced noise, the ratio of the induced current to the current in the shield ground loop is I_load/I_{gnd loop} = Z’_T·l/2·Z_load, exactly what a shield with transfer impedance, Z’_T·l, is supposed to do, reduce the induced current below the current on the shield.  The ratio of the induced load voltage to what it would be without the shield due to a magnetic field, ω·B·A, is V_load/ ω·B·A = Z’_T·l/(Z’_T·l+ Z_gnd) which boils down to L’_T/L’_gnd.  The shield’s transfer inductance, L’_T, ranges from 200pH/m to 1nH/m, over an order of magnitude lower than any external cable inductance, L’_cable ≈ 200-400nH/m in the shield-ground loop (shielded cable ~3” off of the groundplane). [L’_cable = (μ/2·π)·ln(h/2·r)]  Without the shield, the loads are in the ground loop and are excited accordingly.  Load impedances so low that the shield is ineffective are unreal except when common mode TVS fire.

Now to the real problems with coax signal lines even when properly grounded at both ends: unless they are routed close to a structure/chassis ground/cable tray, they emit and receive more EMI than we like.  Large work spaces are too often polluted with noise from PC controller coax lines not routed close to structure and not grounded through bulkhead fitting at the source and load ends.  Running them through bulkhead connectors on chassis ground at their source and end also helps.  The problem is no different than STP cables carrying too much common mode current on the twisted pairs like a TTL signal without common mode suppression; they too radiate too much, often causing noise related problems up into the UHF band.  If noise problems persist with coax signal lines, they should be replaced with triaxial lines with the inner shield grounded at the line driver and receiver signal grounds and the outer shield chassis-grounded at both ends.

The only time where cable shields cannot be grounded at both ends is when connected to sensors that are themselves not grounded.  There, the shield should be grounded to the sensor case if there is one or left floating with the sensor.  If the sensor has too much capacitance to its chassis location, then ground the shield to the chassis nearest the sensor.

The myth of actually shielding with the “shield” grounded at one end only (SPG) has spread throughout the EE design community.  Pseudo-technical argument instills fear of ground loops formed by cable shields grounded at both ends.  See Analog Devices, Analog Dialog 17-1, 1983, for one of the most asinine drawn out of rationales.  The ultimate insanity is when coaxial signal line shields are grounded at one end only.

A cable shield grounded at one end (SPG) behaves like a low pass filter to magnetic fields and a high pass filter to electric fields.  A shield it is not.  The SPG shield actually amplifies EMI at the ¼-wave resonance of the shield with Q ≤ 10.

In fifty years’ experience, this author has trouble-shot too many failed systems where the coax signal line shields were left open at one end.  Grounding the coax shield to the line driver and receiver signal return pins with no break in continuity in between killed the noise problem every time.

Two points: (1) A shield installed for shielding purposes only must be grounded at both ends preferably to the box connectors; (2) A coax signal line uses the shield for signal return therefore must be grounded to the line driver and receiver signal ground/common pins with no breaks in continuity.  If there is an EMI noise problem with this, the cable should be replaced with a triaxial cable with the outer shield grounded to chassis at both ends or a twinax.

There is a real problem grounding any conductor including shields between two units operating off of two power mains in or between buildings.  Ground potential differences can be kilovolts.  Ethernet has common mode in-line isolation rated at 1.5kV rms per IEEE 802.3.  The shield is not the issue; engineers’ knowledge of ground potentials, isolation, and shielding is the issue.  Floating one end of the shield only makes the problem worse.  Isolation units can be inserted to solve most problems.  Fiber optics solves the most difficult problems

With respect to the ground loop induced noise, the ratio of the induced load current to the current in the shield ground loop is I_load/I_{gnd loop} = Z’_T·l/2·Z_load, exactly what a shield with transfer impedance, Z’_T·l, is supposed to do, reduce the induced current below the current on the shield.  The ratio of the induced load voltage to what it would be without the shield in a magnetic field, ω·B·A, is V_load/ω·B·A = Z’_T·l/(Z’_T·l+ Z_gnd) which boils down to L’_T/L’_gnd.  The shield’s transfer inductance, L’_T, ranges from 200pH/m to 1nH/m, over an order of magnitude lower than any external cable inductance, L’_cable ≈ 200-400nH/m in the shield-ground loop (shielded cable ~3” off of the groundplane). [L’_cable = (μ/2·π)·ln(h/2·r)]  Without the shield or with the shield floating at one end, the loads are in the ground loop and are excited directly.  The grounded shield only reduces the induced noise.

Now to the real problems with coax signal lines properly grounded at both ends: unless they are routed close to a structure/chassis ground/cable tray, they emit and receive more EMI than we like, particularly in the broadband mode.  Floating one end of the shield makes it worse.  Routing cables along grounded cable trays helps reduce radiated EMI problems.  If noise problems persist with coax signal lines, they should be replaced with triax or a twinax.

Cable shields should be grounded to the boxes their wires are connected.  No separate connections should be made to the telecommunications grounding busbar (TGB) or telecommunications main grounding busbar (TMGB).  “Grounding” is a misnomer here when we really mean ensure the continuity of the box shields through the cable shield.  ANSI-J-STD-607-A-2002 guides box grounding. 

When shielding cables connected to floating loads like sensors, ground the shield to the sensor case.

Bottom line: Disassociate ground loops from cable shielding.  It is a false argument devoid of correct physics and engineering of shielding and of ground loops, themselves.  If a cable shield doesn’t form a ground loop, it is not a shield!!  Ground loop problems and high ground potentials need to be solved by other means like isolation.  Ground loops involve signal and power return connections.