The company I am working for is developing a patient-connected device intended to deliver Functional Electrical Stimulation to a connected user. We are currently struggling to pass the radiated emissions standard.

We have located the source of our problem to our isolated DC/DC power supply by shutting off the isolated power



emissions results DCDC enabled

DCDC enabled

emissions results DCDC disabled

DCDC disabled

We also notice that when we disconnect the external patient leads, our radiated emissions are reduced significantly. The signals that we are sending down the cable are biphasic pulses with the following characteristics

  • Pulse Width = 500us
  • Pulse Amplitude = 150V
  • InterPhase Interval = 100us
  • Frequency = 60hz biphasic pulse We are in the process of designing a new board with the following features

  • Get rid of internal patient lead cable in favor of a right angle connector

  • Y capacitors across the isolation barrier of the DCDC converter
  • Pi filter and Common mode choke on the input of the DCDC converter
  • Pi filter at the output of the DCDC converter
  • LC filters at the patient connector

However I am still concerned that the external patient cables will radiate excess EMI. Since they are patient cables we are limited on the filtering that we can do and we are not able to shield the cables to the enclosure because it is made of plastic. Does anyone have any suggestions for what we can do to make our device more compliant, specifically as it relates to the patient cables?

  • 1
    \$\begingroup\$ Others better than me can discuss the patient cabling ideas. But I'm wondering if you've considered replacing your isolated DC/DC converter with a spread spectrum version. If you also provide a clock to a micro, and if it has EMI as well, there are also spread spectrum chips for that, too. \$\endgroup\$
    – jonk
    Nov 21, 2016 at 21:24
  • 1
    \$\begingroup\$ You might want to trim that photo of the device some more. The name of your company shows in the paperwork underneath it. \$\endgroup\$
    – JRE
    Nov 21, 2016 at 21:28
  • \$\begingroup\$ @jonk I will certainly look into that, however it has been hard to find a variety of DC/DC converters that meet our 4Kv isolation requirements as well as our 5W, 12V power requirements. \$\endgroup\$ Nov 21, 2016 at 21:32
  • 1
    \$\begingroup\$ When you shut off the DC/DC, did you power the patient side with something else like a battery? If not, the source of the noise may not be the DC/DC but the electronics that was not operational? \$\endgroup\$
    – owg60
    Nov 21, 2016 at 23:16
  • 1
    \$\begingroup\$ We did not do that in the actual test chamber, but we were able to confirm with a spectrum analyzer that the processors and other power supplies on the isolated side of the DC/DC didn't contribute to the noise nearly as much as the DC/DC. \$\endgroup\$ Nov 22, 2016 at 0:40

1 Answer 1


So the rule is: Current will follow the path of lowest Impedance There is a radiating source on your board, you've isolated that section board and you've created a nice antenna by attaching the cable to it. The preferred pathway (and lowest impedance pathway) is out the cable and into the air.

One way to control the current would be to put caps and attenuate the signal where the cable connects to the PCB, this would sink the current out at the cable.

The better (and easier way) would be to increase the impedance of the cable. I can't speak for your signals because you haven't defined what they are, if you have high frequency signals in your cable you will attenuate those too so use caution. A ferrite is a good way to increase the inductive impedance of the cable without interfering with your PCB. You could put it on the inside of the box around the cable that goes to the outside of the box (the internal patient lead, I hope your not putting that into the patient) .

Another way to increase the inductance would be to put filtering on the PCB at the connector to allow your frequency of interest to pass through and to attenuate the RF.

Redesigning the board and getting clocks and converters away from the the cable may also help. If the main source is that DC to DC isolation converter, you may want to put a heavy filter after the converter on the V+ Side and stop the problem before it gets to the rest of the PCB/Design.

Get Electromagnetic Compatibility Engineering by Henry W Ott


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ Thanks for your input. I have added information on the signals that we expect out of the patient cables. We attempted a fix by adding ferrite beads in series (1.5kohm @ 100Mhz) and a shunting capacitor to each line (10nF) between the internal patient lead cable and the external patient lead cable on the PCB mounted to the enclosure. Unfortunately this did not seem to help at all. In our new design we plan on removing the internal patient cable in favor of a right angle connector and also plan to add filtering to the input and output of the DCDC. \$\endgroup\$ Nov 21, 2016 at 22:26
  • \$\begingroup\$ Did you try a clamp on ferrite? \$\endgroup\$
    – Voltage Spike
    Nov 21, 2016 at 22:40
  • \$\begingroup\$ We did not, unfortunately. \$\endgroup\$ Nov 21, 2016 at 22:43
  • \$\begingroup\$ Actually, looking at your signal that might interfere too much just on the cable itself. Another thing I'd suggest is shielding from end to end on the cable with ferrites on the shield make sure your shielding inductance is low as any parasitic inductance between connectors back to your board will create problems. The idea with shields is you need a conductor around your cable 'antenna' to capacitively shunt the currents back to your board. \$\endgroup\$
    – Voltage Spike
    Nov 21, 2016 at 22:44
  • \$\begingroup\$ Do you have any suggestions on how to ground the shield since we don't have a grounded enclosure? \$\endgroup\$ Nov 22, 2016 at 0:41

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.