Daq x-series cross talk and drift

[gimNPC]

Hello,

I have two problems with my X-series DAQ card - 1 is that I observe crosstalk between channels, and the other is that my Ai readings inexplicably begin to drift severely.

I've attached 3 images. Figure 0 is a schematic of my experimental setup. Figure 1 shows cross talk, while figure 2 shows drift.

Fig 0. I am outputting a sin wave from Ao0 through a resistor and back into AoGND. I have confirmed with oscilliscopes and SMUs that the correct voltage/current is flowing through the test resistor. I am then reading the voltage across a separate resistor using Ai0 in differential mode. These leads pick up the Ao signal even though they are not connected. Ai channels that don't even have leads coming out of them also pick up this signal. This cross talk is minimized by having a lower Ohm resistor.

Fig 1.  Ao is outputting a 1V, 10Hz sin wave where Ra=1Mohm, and Rb>10Mohms. The cross talk here is +/-6mV or 0.6% of my input signal.

Figure 2. This shows a signal from Ai0 where Rb is 1Mohm, Ao is outputting a 500mV 10Hz sin wave. The x axis is in seconds. The lower value of Rb causes less cross talk, but regardless of Rb the signal deteriorates badly after 80 seconds. This happens reliably after 80s.

I am hoping somebody can help me with this situation. I am in a situation where the DUT will have a variable resistance, and I require as little cross talk as possible, obviously the drift problem is unacceptable for my application so I'd like to know what it's caused by and how/if I can eliminate it.

Thanks,

Gim

[johnsold · Accepted solution]

This looks like a fairly standard case of ghosting and incorrect wiring practices.

The DAQ device has one A/D converter with the input channels multiplexed to it. The multiplexers are typically switched capacitor devices. So, when the inputs are not connected or are driven from a high source impedance, the charge stored on the switched capacitors can build up to produce both drift and crosstalk.  The crosstalk is often called ghosting.  One part of the solution is to ground all unused inputs (or if the other channels all have the same common mode voltage, wiring the unused channels to that can sometimes be helpful). The other part is to drive all inputs from low impedance source.

The wiring practices issue is shown in the manuals. For most devices when wired differentially, bias resistors to ground from each input are suggested.

The signal in figure 2 looks like 10 kHz rather than 10 Hz.  One cycle every 0.0001 s.

Lynn

[gimNPC]

Thanks, Lynn!