I have a PCB I designed for a project and I'm testing it. In general, it has some analog channels and a MCU to handle all the logic and the communications. I used an isolated DC-DC converter since this will be connected to a rather noisy power supply.

Anyway, I have left it connected working and communicating to the PC for long periods of time to test its stability and sometimes it can work for days and sometimes it stops working. It's being difficult for me to see what's going on because is more the time that it works correctly than the times it just stops working. I think that external EM noise could be the problem (nearby radios, motors, other switching supplies, etc) but I'm not sure.

My specific question is: What can I use to generate high amounts of noise and test my device to see if it can survive?

I thought of just putting a radio antenna near to it and hit the transmit button but I'm not sure if this would be as bad as possible. I need a way of generating some serious noise, the more the better.

UPDATE: Also I'm using galvanically isolated RS2232 IC from maxim and galvanically isolated SPI IC to comunicate with the ADCs (since I'm not using the MCU ADC) and I have 4 layer PCB with 4 different isolated power and ground planes (one for each stage) without slots or current return interruptions. The only common ground points I have is at the voltage regulators because nevertheless I have separated LDOs for every stage (4 in total), they all have to have a common ground point at the main supply. I'm still not completely sure if I'm completely noise-free. As per the comments before I think the may be a coding issue but I still want to noise-test the system to be completely sure is not a hardware issue.

Any suggestions?


  • 2
    \$\begingroup\$ Stick it in a GTEM. \$\endgroup\$
    – Matt Young
    Feb 22, 2016 at 16:24
  • \$\begingroup\$ Connect a microphone and scream at it? (that was actually a serious suggestion). TBH it's unlikely to be noise and more likely to be a programming issue. \$\endgroup\$ Feb 22, 2016 at 16:24
  • \$\begingroup\$ Tape it to your Tesla coil. \$\endgroup\$
    – uint128_t
    Feb 22, 2016 at 16:34
  • \$\begingroup\$ If an electric motor (and any wiring connected to it) were suspect at say, 20 feet away, then relocating your PCB 1 foot away would not receive 20x more EMI as might be expected, but 400x as EM radiation follows the Inverse Square Law. \$\endgroup\$
    – rdtsc
    Feb 22, 2016 at 16:39
  • \$\begingroup\$ @TomCarpenter that's what I'm thinking, that it could be a programming issue, but I just want to be sure about the noise. \$\endgroup\$
    – m4l490n
    Feb 22, 2016 at 16:43

3 Answers 3


Hook it up to a signal generator and inject different levels of noise until you find the failure point. Then you know exactly what it can tolerate. This works better with an arbitrary waveform generator or one of many devices that can input specific noise types/levels. I say sig gen though because they're cheap.


What can I use to generate high amounts of noise and test my device to see if it can survive?

An ESD spark generator is a pretty good device for upsetting electronics that are a bit suceptible to EMI. I've used one successfully to narrow down areas of susceptibility on a PCB, fixed them to add resiliance and then ultimately tested in an EMC lab with just one failure that was easily fixed in the lab.

I've also used an ignition spark generator with the spark plug lead as the emitter of energy. It was run from a car battery. This pretty much generated good old fashioned broadband EMI for hours and hours. To make it simple I wired up a relay so that it self-oscillated by breaking it's own feed to the coil. I used a capacitor across the coil so it was going at maybe 10 pulses per second or less. Feed the on/off contact voltage to an ignition coil but don't forget the condenser to keep the relay contact from wearing out too quickly. I made vast improvements to a system on the cheap using this method and apart from one lab modification to achieve 10V/m compliance it passed H field testing at 100A/m completely.

  • \$\begingroup\$ Cool that sounds good. Do you have an schematic for this? \$\endgroup\$
    – m4l490n
    Feb 23, 2016 at 14:06
  • 1
    \$\begingroup\$ Not off hand but you can make a relay oscillate by wiring its contacts and coil together. There has to be circuits using google: google.co.uk/… \$\endgroup\$
    – Andy aka
    Feb 23, 2016 at 15:42

First suspects for malfunctions due to noise are inside!

Assuming you have some kind of Data acquisition system interfaced to PC (via USB / UART), here are some possibilities.

With MCU, ADCs, amps, powersupply and any additional logic on a single board, the malfunction may be due to something called "Ground Return Noise". Ground return noise seriously effects digital inputs of FPGAs, logic ICs, MCUs etc.

Isolating the analog section, power supply section and digital sections with independent grounds will reduce this problem. Isolate them using a series inductors and parallel capacitors. Basically forming a low-pass filter sections. One can refine it by a T or Pi section. One may want to cascade them based on the severity.

Also because the equipment is interfaced to laptop/pc which, pronounces more ground return noise. If that laptop/pc is connected to mains power then the noise level is even higher. One can use a Ferrite-bead on the USB/UART cable to suppress high-frequency power-supply noise.

If these two don't reduce the noise, try shielding the analog section and power supply sections. They won't receive or contribute noise. (As long as circuitry inside is not certain type RF circuit)

However if you want to strictly find the problem is due to noise susceptibility from 'outside', keep a cell-phone nearby that device and give a ring from another phone!

Also if the earlier earlier mentioned problems are addressed, this cell-phone noise can validate the immunity to some level!

  • \$\begingroup\$ This doesn't seem responsive. The original poster want to noise-test his system, not noise-harden his system. \$\endgroup\$ Feb 22, 2016 at 22:19

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.