How to avoid cross-talk in high sampling rate?

[Youngkey]

I got large cross-talk when I tried NI6024E muntifunction DAQ board through Labwindow/CVI. The setup parameters were as follows:
sampling rate: 4000 ~ 200000 Hz (I tried several rates)
scan rate: 1000 ~ 10000Hz
channel : 4 ~ 8
signal source: RION(japanese) function generator (octave band random noise, center 63~4000 Hz)

In discussion forum, I read recomendation to adjust interchannel delay. However, I cannot understand what the interchannel delay is. Is it related to the combination of sample rate and scan rate? By any chance, I tried several combination of sample rate and scan rate, but I got the same cross-talk.

If I increase the frequency of the input signal (function genrator signal) the cross-talk decrease.
But, it is still large.

What can be the solution? Thank you very much.

[PdB]

Before answering your question it may help explaining how most multi-channel DAQ-board work.
They are build around 1 ADC and that design-choice was made in a time when ADC's were big and expensive.
When adding an extra digital switch just in front of the ADC, you can easily increase the channel-count of the ADC (making it a more cost-effective board).
This is what most board-manufactorers did and still do.
There are some dis-advantages to this principle.
A)Delay-time between samples on different channels.
B)Possible crosstalk.
If the ADC reads in the signal too fast after switching, it may measure some of the old voltage (last channel) it measured.
This problem is related to the output impedance (resistance) of the signal you are measuring.
Anot
her problem you may have is wiring cross-talk.
Signals running close to each other are likely to pick-up signals from the other wires.
The solution to both problems is to lower the impedance of the signals or at leased making them in the same order and rerouting the signals to decrease the pick-up. Adding extra C's may help.
Another solution that was suggested is decreasing the interchannel delay, the time in between the switching and the starting of the ADC conversion process, but this will not get rid of the problem entirely.


Good luck, Patrick de Boevere

[pelesl]

I think somewhere in the manuals they also recommend adding a channel "before" the channel in question with the input grounded. That's provided you know the scan order, and NI seems to have a nack for hiding such information (like finding the corresponding differential channel pair...).

In my application I am manually measuring an unamplified thermocouple signal and performing my own CJC. I can clearly see that as the sample rate increases, the voltage on that channel keeps increasing. Further, I've noticed that if you set it to the lowest voltage range (highest gain) for the channel in question, the effect gets worse. So I can actually get a better signal if I set a lower gain because of this problem. In the end, this makes it impossible to do small signal stuff directly without pre-amplification if sample rates wlil be high. I know that's not recommended either, but I had to do it at the time.

This is on a PCI6289.