I have an isolated switching power supply (120v in, 12v out) that gives an output shown in Figure 1 when referenced to the output ground (I'm not TOO worried about that noise). The bigger issue I have is the noise I have when referencing the output voltage to the input ground shown in Figure 2. To me it looks like the differential mode noise is acceptable but there is a ton of common mode noise but I am not 100% sure how to go about fixing this. I read a lot about common mode chokes but I am not exactly sure how to choose the correct one.

EDIT: So to give the full story to clearify. I have a touchscreen that is picking up random touches only when powered by the power supply described. I have already tried full power and signal isolation and power regulation which was able to get rid of nearly all differential noise. The only noise still getting through is all common-mode.

Also, I only have the option to fix this issue on the touchscreen circuitry with no ability to alter the power supply. A simple fix would be to bridge the input and output grounds with either a jumper or cap but unfortunatly, this cannot be done.

enter image description here

                          Figure 1

enter image description here

                          Figure 2
  • \$\begingroup\$ A large CLC-filter on the output should rectify that situation. \$\endgroup\$
    – winny
    Nov 18, 2016 at 13:36
  • \$\begingroup\$ Common mode noise at mains frequency is pretty common with isolated switch mode supplies. You should find that it has a very high source impedance, i.e. it goes away if you connect either output rail to ground, even via a big resistor. If so, it is not dangerous, despite being relatively high voltage. If you still want rid of it, do you have a protective earth in the equipment you are designing? \$\endgroup\$
    – Jack B
    Nov 18, 2016 at 13:46
  • \$\begingroup\$ Are you measuing differential or common mode ripple? Your question switches between the two. \$\endgroup\$
    – winny
    Nov 18, 2016 at 14:41
  • \$\begingroup\$ @Tyler Fontaine , rather than make us guess, show layout, schematic and calibrate two probes to test square wave then use Ch1-2 to get flat line to use for differential measurements with short ground leads and measure output noise. without a low Z out gnd, floating noise is "common" LOL \$\endgroup\$ Nov 18, 2016 at 14:43
  • \$\begingroup\$ The reason I am worried about common-mode noise is due to the fact that one of the items running off of this supply is a touch screen. When running off this supply I get false touches on the screen when the screen is touched and also randomly when it is sitting untouched. I have tried other supplies with worse/better differential mode noise and common mode noise and it is only on supply's with worse common mode noise that cause the issue. This is the reason why I am trying to lower the common mode noise. \$\endgroup\$ Nov 18, 2016 at 15:52

2 Answers 2


The core of an isolated switch mode power supply is still, though cleverly switched, a transformer.

Thus, your problem is the same as this circuit:


simulate this circuit – Schematic created using CircuitLab

The whole point of an isolated supply is that there's not a single conductor from the primary to the secondary side.

Therefore, you can imagine the pair "OUTPUT+" / "OUTPUT GND" to have a defined voltage between them, but they float on a "random" offset relative to the primary side's Ground. That offset doesn't actually contain much energy, and you can break it down very easily, for example by connecting "OUTPUT GND" to primary ground (thus eliminating the isolation!).

  • \$\begingroup\$ your simple model does not show inductive loop effects, CM noise, coupling capacitance, ground inductance or cap ESR. This is similar to how a chopper amp passes low frequency signals. \$\endgroup\$ Nov 18, 2016 at 14:38
  • 2
    \$\begingroup\$ True, but showing that in a simplifying, explanatory example would definitely defeat the purpose of having that example. Showing of knowledge on all the effects that happen in complex, inductively-coupled systems wasn't the point of my answer :) \$\endgroup\$ Nov 18, 2016 at 15:05
  • \$\begingroup\$ @TonyStewart.EEsince'75 it's a "simple" model hence does not need to show that stuff (even you refer to it as "simple" in your comment) and the relationship between this question and a chopper amp is anyone's guess. \$\endgroup\$
    – Andy aka
    Nov 18, 2016 at 15:37
  • \$\begingroup\$ Impedance of probe and impedance of signal defines how big is the result. That's what stray transformer coupling capacitance will do to this signal. Adding a small cap to input Gnd much bigger than stray cap will attenuate result. This is the fault of the model, thus reduce CM noise. \$\endgroup\$ Nov 18, 2016 at 15:44
  • \$\begingroup\$ I have updated the question to try and clarify. \$\endgroup\$ Nov 18, 2016 at 16:02

I have an isolated switching power supply (120v in, 12v out)


The bigger issue I have is the noise I have when referencing the output voltage to the input ground

It's not an issue at all - who cares what it is - it's the output noise that is the only concern when using an isolated power supply.

  • \$\begingroup\$ I have updated the question to try and clarify. \$\endgroup\$ Nov 18, 2016 at 16:01
  • \$\begingroup\$ If you can't use a capacitor to reduce the common mode voltage to ground I believe you are going to be stuck. \$\endgroup\$
    – Andy aka
    Nov 18, 2016 at 16:20
  • \$\begingroup\$ Would a common-mode choke not do anything? \$\endgroup\$ Nov 18, 2016 at 16:30
  • 1
    \$\begingroup\$ Seriously, how many times did Andy, Tony and I say that now? \$\endgroup\$ Nov 18, 2016 at 16:39
  • 1
    \$\begingroup\$ You had said an example to correct this would be to connect output ground to primary ground. You didn't say there were no other solutions. All I asked was that if you are unable to tie the output ground to the primary ground would there be any other solutions. \$\endgroup\$ Nov 18, 2016 at 16:43

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.