Before start, this is going to be a long question. My apologies in advance!

At my company, we are trying to design an automated engineering validation test (Auto EVT) tool for power supply board (PSUs). To achieve that we are going to use these commercially available CPLD and FMC cards:

FMCDAQ2 from AD for data acquisition UltraITX+ Baseboard + UltraSOM+ MPSoC Module with Zynq UltraScale+ for processing and connectivity

There are a few slight problems with this setup which are that we have to do DC coupled measurement in PSU EVT.

Transitor measurement

For example this one of the measurements on switching transistor in PSU. We need DC couple measurements to evaluate this.

And also we need 16 measurement points where FMC card only has 2 ADC input. Moreover, lastly, we will measure alive part of the PSU also, therefore the measured voltage difference will go as high as 300Vs.

To overcome this problem we thought of designing an extra relay network to add to in front of the ADC and DAC outputs of the FMC card.

This relay network board will provide selection of a single probe (out of 16 probe) to selected input channel, selecting the voltage scale (if the measured voltage difference is 300V select 100x if 3V select 1x etc.). Also, it will provide galvanic isolated RS232 drivers to send commands to other test equipments. But nevermind that part, that part is easy actually. The hardest part is the floating ground of the PSU.


Measured PSUs are usually don't have Earth connection. Also other parts, relay network board and FMC+BaseBoard are also powered separately without earth connection, with isolated dc.

The biggest problem is, FMC card has a common ground for everything,

adc input


And since we are going to turn ac couple into dc couple by bypassing the ac components and making them short circuit as below:

dc couple adc input

when we close the relay that connects the channel to alive part of the PSU (maybe 400V above in reference to the GND of FMC card) we don't know whats going to happen and these are really expensive instruments. We want the secure them and take every action for safe operation before starting experiments.

In relay network card, I have to provide a differential amplifier with common voltage equal to 2.05V according to the datasheet of ADC in the FMC card(AD9680):

adc datasheet 1

I wanted to build a simulation to see the effects. You can see and download ltspice files here.

If I have to briefly explain the simulation, I connected the GNDs of the relay network board and FMC board with 2m ohm resistors which is more than reasonable. SMA connectors will have at least 5 miliohms serial resistance.


I charged a 10 nF capacitor up to 200Vs, that refers to a floating ground reference in the alive part of PSU. I put a voltage source to provide differential voltage between the tips of the probe. And put the relay accordingly. Relays have the same Ron and Roff resistance that we are going to use in the Relay network board, TLP3107.


When I run the simulation, I expect that it will simulate the close of a relay that is connected to a high floating voltage in alive part of the PSU.

It didn't work. When the relay closes, even though there are only 2 mohms resistance between the grounds.

gnd spikes

Diff opamp inputs go up to 30Vs,

opamp inputs

Tried putting ferrite beads, TVS/ESD/Zener diodes between the grounds and it didn't work.

So my question is how can I overcome this problem? I believe the same problem must have been in oscilloscopes too. I want to know how do they overcome this problem?

  • 3
    \$\begingroup\$ I wanna help, but I gotta tell ya - I cringe when seeing the word 'gonna' in serious writing. Any chance you could change it? \$\endgroup\$ – Bruce Abbott Dec 14 '19 at 14:56
  • \$\begingroup\$ @BruceAbbott There were two 'gonna' and I have turned them into 'going to'. I am taught in elementary school that both 'gonna' and 'going to' can be used in formal writing though. \$\endgroup\$ – Alper91 Dec 14 '19 at 15:24
  • 2
    \$\begingroup\$ I'm with Bruce on "gonna". It is unfortunate if you were taught that it is acceptable formal English. \$\endgroup\$ – Elliot Alderson Dec 14 '19 at 15:36
  • \$\begingroup\$ @Alper91 your simulation is basically just dumping a capacitor to ground (FLG on left side of R3 is attached to ground through your 2mohm resistors. \$\endgroup\$ – Analog Arsonist Dec 20 '19 at 3:15
  • \$\begingroup\$ I wasn't done.. 1. your simulation is basically just dumping the 10nF capacitor to ground (FLG on left side of R3 is attached to ground through your 2mohm resistors). 2. What is V5 source doing? Please show everything. And use net labels in your plots.. we don't know what V(n001) is for example. But... 3. Why don't you just connect all the grounds together, including the PSU? Then it won't be floating and your 300V signal will be the relative to all other boards. \$\endgroup\$ – Analog Arsonist Dec 20 '19 at 3:25

Your Answer

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

Browse other questions tagged or ask your own question.