Excessive noise with PCI6143 and hi impedance input

Question

I have a PCI6143 installed in a Windows XP computer for evaluation right now. The test I'm evaluating for is reading up to 5 DC signals from battery cells at up to 1Ks/s. For calibration and verification purposes I need to use a voltage calibrator on the inputs. When I read the cells there is no problems. When I use a calibrator I get a significant noise in the signal. It only affects the channel the calibrator is hooked up to and not the rest and it's only at sample rates of 100s/s or higher. The noise is not AC. The pattern is usually 10 samples at the right voltage and 5 at 0.5V less. It doesn't seem to change this pattern for sample rates up to 100ks/s.
Is there a problem with the input impedance mismatch here? I've used different calibrators and power supplies with the same results. I've checked their outputs on desktop DAQ devices at 100ks/s and there is no appreciable noise.
Any thoughts or possible solutions?
Thanks

Answer 1

Otis,
By 'calibrator' I mean a precision voltage source. Since we can't use the actual battery cells for the test run we normally use a source like this to check the voltage readings each day prior to testing. The output impedance of the calibrator is <10 milliohms, which is pretty close to what the cells are. When I use another power supply to check the voltages with a higher output impedance (1kohms) I still get the problem. The PCI-6143 is used to measure the voltages of the cells during the test. It measures the cells perfectly but we need to verify that once per day with a precision source during a test run (customer requirement).
Yes the cycle is pretty much independent of the sample speed above 100 samples per second. It always seems to be 10 samples correct with 5 samples low. Below 100 S/s it doesn't seem to be there at all.
It happens on any channel the calibrator is attached to. I have tried it alone on each channel and simultaneously with as many as 6 cells each on various channels and it still happens.
Another thing is that if I do the same test using a benchtop DAQ (i.e. Nicolet Vision Xe) it doesn't happen. We are trying to move toward simpler test setups and having a card in the computer is easier than hauling around an external device with its keyboard and mouse and something to set it on along with all the data collection headaches. Plus the internal card is much cheaper and easier to use for most of our testing.
Thanks,
Jack