A technician was suggesting we try connecting an isolated ground (described below) to a non-isolated ground suggesting it could reduce EMI noise and possibly fix an issue we're seeing (which we don't know the root cause of yet). This confused me so I thought I would bring it up to discuss here.

Is this a valid theory?

What we have is two isolated potentiometer amplifiers (Turck IM36-11EX-I/24VDC). Originally their isolated grounds were not connected but we've ending up connecting them as a side effect of another isolated circuit we introduced. The two grounds are still isolated from both output's ground, but not from each other.

We've been having some unexplained issues, which may very well be completely unrelated to the new circuitry, but one theory is noise on the inputs of the amplifiers is causing the issue. We can try removing our circuitry to see if that resolves the issue we're seeing (thus once again isolating the two amplfiers from each other), but while discussing a technician suggested our issues might be resolved by connecting the ground from the output side of the amplifiers to the input's ground (thus removing the isolation). Besides ignoring the reason the isolation is there in the first place, would that actually have any merit in solving a noise problem?

As far as I understand, disabling the isolation could just introduce more noise, but I'm looking for more insight.

  • \$\begingroup\$ As clear as mud. \$\endgroup\$
    – Andy aka
    Feb 14, 2019 at 14:31
  • \$\begingroup\$ If it is easy to do try it. It may give you a clue as to what's happening. If it is difficult it is probably not worth it \$\endgroup\$ Feb 14, 2019 at 14:35
  • \$\begingroup\$ It's not really clear what's going on. Can you provide a diagram? Isolation can help with EMI, but has a lot of potential do make things much worse when implented poorly. Rules of thumb in EMC often become harmful myths. It's probably worth a try if it's safe. \$\endgroup\$
    – lnowak
    Feb 14, 2019 at 14:40
  • \$\begingroup\$ Draw a block diagram if you want multiple choice solutions in my answer, otherwise your question lacks these details. \$\endgroup\$ Feb 14, 2019 at 14:51
  • \$\begingroup\$ Yes it seems like a waste to short circuit isolation when you have paid $500 or an isolated pot. But we do not know your SNR and signal, noise BW nor any idea of your environment. \$\endgroup\$ Feb 14, 2019 at 15:30

1 Answer 1



Many solutions exist that may either divert EMI away from (shield,shunt or block) or improve isolation between unintended noise and higher impedance sensor or reset logic signals with custom designed Hybrid transformers or Baluns ( e.g. ethernet)

Your case

$500 potentiometer with 0~20mA output and LED indicator. Current mode is in theory infinite impedance but in practice is 1uH of Common mode inductance per meter. Then cable capacitance can add impedance to CM earth to improve isolation but degrade if there are circulating ground currents. The CC isolated Pot has a recommended load R=500 Ohms which is differential only.

Since we do not know your SMPS noise or other interference E or H field size or distance, Measure your input with 2 two 10:1 probes first calibrated on 1 signal with Ch1-2 mode to give a flat line on 50mV/div.

  • Then use as a differential probe across the 500 Ohm R, with probe gnd on 0V and no earth gnd connection. Record and report results of DC and AC noise.
  • then add caps across each signal CM, to 0V nearby start with matched values from 1nF to 0.1uF ceramic and record/report same values of DC and noise with worst case surge currents in cables nearby as in application and compare with ADC std deviation of results.
  • then only put 1 CM cap across 500 Ohm R without other caps and compare.
    • this maintains high CM impedance but shunts R with impedance of C across R in DM mode only.

If you don’t have STP cables with shield terminated ONLY at 0V of 500R load.

  • repeat tests without caps again under worst case noise conditions using STP cable
    • add DM cap and record results
    • remove the DM cap, add 2 CM caps and repeat recording results

The above is for you to gain experience,

  • not what I would do or similar EMC expert with 40 yrs EM design experience.

EMC is about impedance and spectral management to achieve desired egress and ingress.

Next compare

Every situation is different for impedance , spectrum, unintended noise BW, signal BW and CM impedance, DM impedance, resonant cable lengths, etc so EMI solutions are often complex in analysis, yet easy to implement with many variations.

Details in general

Stray EMI either inductive or capacitively coupled in the MHz band is fairly high impedance since it is a much longer wavelength.

When you have an isolated supply, it is also even higher impedance with some stray capacitive coupling from primary to secondary. So RF and impulse noise dV/dt and surge currents, dI/dt can cause CM or common mode noise. but your your signals are not low impedance or balanced then CM current can create a differential mode voltage (DM) being added to the signal relative to 0V which is floating.

Thus connecting 0V to earth ground being low impedance , either with a cap (10nF or more) or short wire connection ( low inductance) it can divert noise or attenuate CM noise with a very low CM shunt to earth ground. This results in better immunity.

If one knows the exact noise source and susceptible inputs , there can be many solutions from raising the CM impedance with a BALUN of ferrite or iron, shielding, hybrid transformer, twisted pair, ground plane, STP wire, Vcc cap to gnd, gnd cap to earth ground, VGA connection from floating laptop with noisy charger to earth bonded monitor for a few examples.

. Motors and their drivers are notorious offenders of surge currents, so, STP or twisted pairs, decoupling caps near delivers, snubbers, etc are also considered. Also putting cables apart or at right angles is often done to improve mutual inductance isolation of motor currents.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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