How comparable are PXI_Gnd between DSA boards

[peabody124]

I have 7 PXI-4498 boards in my chassis, and I'm trying record a large number of signals with the best common mode noise rejection.  I was wondering about sending out the signal from the back of the PXI chassis to set my ground for all the signals, but then recording them all as well as a reference version of ground at the positive inputs.  Now I'm not sure if I should short all the negative inputs to that reference ground, which would make another 50 ohm/96 channels connection to ground - which would seem to be a ground loop - or leave all of them floating and let the internal pulldown ground to the PXI_Gnd and then digitally subtract out my ground reference recorded on a positive input.  If I do that, how comparable will the PXI_Gnd signals for each board be when they are at opposite ends of the chassis. 

Thanks,

James

[HeatherBaker]

Do you have this document?  It is worth saving as a reference for later use.  Can you please elaborate more on what you are measuring and your set up?  All of the PXI_Gnd signals will be the same regardless of where they are in the chassis as they are on the same backplane.

[peabody124]

I've read it, but I'm not sure it answers my question because it doesn't cover pseudodifferential inputs.  I'm basically recording ~100 ground referenced signals that are some distances from my chassis so have ~30 ohms impedance.  I have an interface board between my sources and the chassis, and on that I short all the shield channels and the cable shield to the ground plane and send it out to ground my source.  All of my sources come back to the positive inputs, and a copy of the ground comes back for all of the negative inputs. 

The problem is (I think) that creates essentially figure 9 in that diagram because with 100 negative inputs connected to that ground reference their impedance to the PXI_Gnd there is .5 ohms.  I want to switch to approximating an NRSE input by recording the ground copy on a positive input (while the source reference is still driven to PXI_Gnd), but I'm not sure if I should just let all the negative inputs float and have the 50 ohm pull down to PXI_Gnd internally, or if I need to have a bus shorting them all together to make them the same.

[peabody124]

So I've done some tests using just the BNC cables on two cards on opposite sides of the chassis (slot 5 and 11) that seem to indicate their internal grounds are fairly different.  I've pasted these below, just ask me if there are any questions about what I mean.  Given this I want to make sure that shorting them all together is a good idea, and ask if what they are shorted to should be allowed to float (thus becoming close to their average, probably with noise from the ground loops) or should be explicitly grounded to something.

 1. [Done] Short shield from one PXI-4498 to positive input of another.  This should allow us to compare their grounds and see if they have relative noise. ==> See large amount of 60 Hz (10 uV RMS).  This implies the PXI_Grounds are not the same.  If we do this within a board there is no problem.
 2. [Done] Short positive and negative input for one channel, and connect to shield of another board.  This might let us see how the ground loops look. ==> Not much noise.
 3. [Done] Short positive and negative of inputs from two boards.  Connect their negatives together ==> No noise.
 4. [Done] Short positive and negative input for one channel, and connect to negative input of another board.  This will replicate a little what we will make in the circuit board if we make a big negative bus ==> Does not cause an increase in noise
 5. [Done] Short two positives together (I know, biasing issue...) measure noise in difference.  Repeat with negatives also shorted. ==>  Without negatives noise is ~10 uV RMS < 200 Hz.  With negatives shorted is 700 nV.
 6. [Done] Connect positive input to PXI_Gnd on back of chassis ==> Lots of noise, looks mostly like 60 Hz modulated by another slower frequency (~10 Hz).

[HeatherBaker]

Are you committing a channel to find the ground reference source?  From the your description, you will want to let the PXI ground internally else you will end up in the not recommended RSE configuration from the document.  If you are still having any grounding discrepancies, then please attach a schematic of your set up.

Have a nice day!

[peabody124]

Sure, here is a schematic of how my system looks with all the inputs shorted (i see a lot of 60 hz noise that is the same on the groups of inputs).  However, all the tests I described in the previous post were with two of the Infiband to BNC converters from NI attached to a board on each side of the chassis.  The easiest way to see the noise is just wire the negative of one board to the positive of the another.  You essentially have 50 ohm between them and see all the noise.  Keeping all of them floating did not prove to be an adequate solution because of the different ground between boards.

I blue-wired a board together to have all the negatives bussed together but floating, then added another positive input for each of the cables that measures the kelvin connection to ground.  This seems to give me a good reference I can digitally subtract out afterwards.  As you said I have to treat it like an RSE since that many pseudodifferential inputs are shorted to ground.  However, I do have to make sure all the negative inputs are shorted together.

[ManLongAtNI]

Can you provide a better image of your schematic? It appears that two of the channels on the bottom don't have any input to the positive lead. Are the channels on the left hand side representative of the measurement source or the 4498?

Regards,

Glenn