We have a PCI-6733 card that is connected to our experiment through a BNC-2110. We've recently tracked down one of our day to day fluctuation to the change of the output voltage of these boards caused by grounding. We simplified our setup and we could reproduce the issue with the following procedure.
When the ground (BNC shield) of the output is not connected to anything, the output we measure (using a battery powered multimeter) is stable and accurate. However, as long as we connect the output ground to any real ground (including the one the computer is plugged in), the output starts to change by as much as a few hundreds of mV (though, more often between +- 30mV).
Almost all of the online document we've find are about grounding when using the analog input function and not analog output. We did make sure the BNC-2110 switches are all at FS position though switching to GS doesn't seem to make any difference. It's not really an option for us to make all the devices driven by this board to be floating either (we don't even have control over all of the devices).
What's the right way to solve this problem/is this a known issue? Do we have to have some output buffer for all the analog output channels?
To me, it sounds like there is a ground loop since you don't get issues as long as the shield is isolated.
Did you read this white paper? I know this talks about Analog Inputs a lot, but the issue still sounds like a ground loop like the described in the white paper.
Yes I've read that and yes I do think it's a ground loop and that's why I said it should be fixable with a buffer. However, it seems to me that unlike analog input the analog output doesn't support load that's grounded. It would really surprise me if that's the case.