I'm working on a PLC cabinet, and I'm wondering whether I should hook cable shields to earth ground, or to the power supply negative.

I checked and the power supply DC output is isolated. There's no conductivity between earth ground and the power supply negative.

Would the answer change if the supply were not isolated?

The cables go out to sensors, and they carry 24v power, 4-20ma signals, and 0-10v signals.

Side question: Is there anything wrong with leaving the dc supply isolated? The chassis is grounded to earth.

  • \$\begingroup\$ is the shield there for safety, emi ? \$\endgroup\$ – JonRB Jun 13 '18 at 21:00
  • \$\begingroup\$ Is it coax or shielded pairs? How long? \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 13 '18 at 21:16
  • \$\begingroup\$ The shield is there for EMI. The cables are probably about 10 meters max with 2-8 conductors per cable. \$\endgroup\$ – Drew Jun 13 '18 at 21:23

What you have here is this:


simulate this circuit – Schematic created using CircuitLab

It may be most advantageous to tie the shield to the case, the goal of a shielded cable is to take electric fields that might capacitively couple into the signals you wish to shield and shunt them someplace else. Since the isolated supply is most likely capacitively coupled (isolated) to the box tying the shield to the isolated supply is not going to provide a good pathway to shunt currents away from your signal, it could also create common mode problems if a large current is moving down the shield. If you tie the shield to the chassis ground, the currents that impinge across the will flow down to ground away from the signals you want to protect.

It really depends on which is the lesser of two evils, there are circumstances where one would be better than the other, I'm talking about the general case. Cable shielding is about controlling currents on the shield. Typically in my experience tying the shield to the case is best, but I've also seen a few cases where it isn't.

If your requirement is also having the shield isolated then tie it to the powersupply ground.

A second thing to note is what is tied on the other end of the shield, because if the shield is tied on both ends, there can be currents that flow from ground loops and if you get that, the current can inductively couple and create noise in your signals inside of the cable. In a general sense, tying the shield on both ends to ground is bad for this reason.

| improve this answer | |
  • \$\begingroup\$ "create common mode problems if a large current is moving down the shield." is true for one example but not true in general to say CM current is high. CM noise can be and most often due to a high CM impedance (floating DC) with poor balance eg stray C2 or saturating input CM voltage creating a low SNR. (FYI) \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 14 '18 at 14:39
  • \$\begingroup\$ Thats what will happen if you tie the shield to the isolated supply, and a large current will bump everything up if the supply really is floating. So in general, tie the shield to the case. Since I don't know the values of C1 and C2, I can't speak for which case will be better \$\endgroup\$ – Voltage Spike Jun 14 '18 at 15:13
  • \$\begingroup\$ Ok but the ungrounded shield is the worst case . Ethernet often uses 1nF to receiver case 0V then puts it thru a CM choke then an isolation transformer then if EMI on an external mic is bad with charger connected , an earth ground or VGA monitor (which is earth bonded) is needed. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 14 '18 at 16:58
  • \$\begingroup\$ An ethernet cable doesn't have a shield and if it does, the shield is not connected to the transformer. The shield is most often connected to the PCB ground, which is usually tied to ground in some way. \$\endgroup\$ – Voltage Spike Jun 14 '18 at 17:42
  • \$\begingroup\$ I should have said CAT6 Ethernet and yes they do use a cap from shield to case \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 14 '18 at 19:29

You should connect 0V of the DC supply to ground to prevent it floating. The cable shields should also be connected to ground.

| improve this answer | |
  • \$\begingroup\$ I've heard this before, but what's the advantage? As far as I can tell it just makes ground loops more likely. \$\endgroup\$ – Drew Jun 14 '18 at 2:26

Isolated supplies and cables is the worst case for CM suppression on unbalanced sensors.

In order to divert high impedance stray E and H field noise on signals you want to attenuate CM noise from DM signal to avoid the nonlinear conversion of CM noise currents into mismatched input impedance or exceeding the CM input range of high impedance inputs.

Classical EME solutions or EMI reduction methods include ;

  • lowering differential mode (DM) impedance of link end-to-end while raising CM impedance ( useful in telephony , ethernet , HDMI and commutation noise in motors , etc ) or using CLLC line filters.

    • Useful for line noise to microwave
  • Lowering impedance of Rx to HF noise with shunt RF caps (like Y caps on a line CM choke). Or both of above (best case)

  • Raising impedance of series RF with CM choke and ferrite beads and absorbing loss into lower impedance load ( useful for RS-422 , and HDD low impedance signals etc)
  • The path of diverted ground currents should be the shield ground path and not the signal differential path.
  • terminated shield with low impedance earth bonded ground at source ( if possible) or load ( if not) but not both if risk of ground difference noise voltages
  • back drive shield from load with low impedance CM signal derived from Instrument Amp IC (common for EEG uV signals)
    • this shunts external stray E and B field noise from reaching the internally shielded pair and satisfies medical specs for high SNR and low AC line leakage current

However 20mA current sources have high impedance to cable losses and improved immunity to voltage drop in cables they may not be perfectly balanced to Shane the Rx bandwidth or diode protection rectifiers If CM range is exceeded by RF modulation from say nearby mobile phones. Thus common mode LLC line filters may be needed.


One cannot predict performance without knowing the EMC spectral noise. I had a couple career embarrassing moments when high power Radios or arc welders nearby caused untimely errors during operation.

| improve this answer | |
  • 1
    \$\begingroup\$ Who is the -1 W5V0? DT!? \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 13 '18 at 21:12
  • 1
    \$\begingroup\$ I don't know who but I am seeing really somebody is frustrating you for sure.. With downvotes.. Why don't you ask in meta if possible to track the silent protestor? \$\endgroup\$ – User323693 Jun 15 '18 at 16:58
  • 1
    \$\begingroup\$ This answer has a lot of information, but I do not see a clear answer to any of the questions asked. \$\endgroup\$ – Joe S Jun 15 '18 at 17:17
  • \$\begingroup\$ OK I agree but the question lacks even more details on EMI impedance spectrum nearby , impedance of cable and length. available earth bonds and spectral difference in current if shunted. etc etc. Any speculation on other answers is purely by guesswork. with many assumptions. If anyone disagrees they have alot to prove and even more to learn. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 15 '18 at 18:34
  • \$\begingroup\$ The 1st line is correct. Floating shields are the worst case as they induce ground shift noise if connected to signal ground. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jun 15 '18 at 18:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.