Ghosting Resolved (Crosstalk)

[Foreshadow20]

This is a post as a reference to others on how I resolved my ghosting issues. Well mostly resolved, I still have a 1mV per 1V adjustment. Works fine with my hardware in between. You may want to tweak the values around to get none at all. Feel free to post your values and how you fixed ghosting issues. I only added Crosstalk as a reference in subject for those that are unaware of ghosting.

Same Topic on LAVAG:

https://lavag.org/topic/19243-ghosting-resolved-crosstalk/

My Offline Setup:

Direct connection of a NI PXI-6723 to a NI PXI-6255. I tweaked around with the capacitors, but not so much with the resistors.

Ghosting was resolved having a 100K resistor and 25nF capacitor on each Analog Input Signal to GND.

A dummy channel (in my case I just added an extra AI reads to DAQmx AI) between each Analog Input Signal. These dummy channels were connected to GND using a 100K resistor and 47nF capacitor.

I had to reduce either my rate or the AIConv.Rate proprety node.

My Live setup with the minor hardware in between:

I had to remove the 100K resistor and 25nF capacitor on the Analog Input Signals (This caused too much voltage drop on the system).

Everything else remained. No ghosting at all.

Other References:

How Do I Eliminate Ghosting from My Measurements?

M Series User Manual (Chapter 4)

[Hooovahh]

Yeah this can be an issue and it is hard to explain to non technical people why it is happening.  In the past I was sampling lots of PWM signals with an analog input (they wanted to see the full wave).  So naturally if I sample one PWM which is at 50Hz and another at 75Hz in the same DAQ task, that there might be ghosting where maybe I read the 50Hz signal, then the 75 and the 50 was high, and 75 was low.  The value read for the 75 may not be low, it might be in the middle causing odd looking signals.  

For us we added reading a dummy ground signal between each signal, which effectively cut our maximum sampling rate in half.  Reading a 75Hz signals should be done with a 750Hz sample rate, meaning two signals and I need a 1.5Khz sampling rate card, or with two dummy signals, a 3Khz sampling rate card.  For us this was fine, but if we had been trying to read fast signals, or if we were planning on using all the AI channels, we would need to redesign the system.

We tried the impedance matching technique at one point but didn't ever get it working right.

[Henrik_Volkers]

just by the nature of the cause, it should be better to use lower sample rate or read one channel twice.  With the ground channel you just know you will always read to small values.

The RC filter lower the dynamic range but also lower the dynamic impedance.

I don't buy multiplexed DAQ cards anymore .... they all get blended by the low $/CH but missed the settle time error spec 😛

Well NI could spend individual e input buffers ... would be +1$ per channel?

And the high input impedance is also only valid for a single channel DC measurement 😉