Dummy Channel Ghosting Solution Problems

[aanthis]

I am using the PXIe-6363 card to read 8 analog input signals in my experimental setup (AI channels 0-7). I am using eddy current sensors which I assume have a high impedance because I see ghosting in the signals. A swing from 0 to 10 of AI 0 will cause AI 1 to increase by about 0.4 V. Each signal affects the signal which is sampled after it. I have been going through the troubleshooting steps in this link: How Do I Eliminate Ghosting from My Measurements?

• If I only use 2 channels instead of 8 to maximize inter-channel delay (my code uses the minimum convert clock rate possible) and lower the clock rate from 10 kHz to 1 kHz (convert clock rate from 80 kHz to 2 kHz) then ghosting disappears. However, I need all 8 channels and a faster clock rate.

• If I physically space channels out (i.e. 2 channel test with only AI0, AI7), ghosting still occurs.

• If I add global virtual 'dummy channels' so that I sample a channel twice before recording its value (throw the first measurement away), nothing changes and I still get ghosting. Does this dummy channel solution actually add more time between channels or does it just double the channels while halving the inter-channel delay? Is there a different way to do this dummy channel solution?

I do not want to switch to a different non-multiplexing card as this is the card that I have. I also do not want to have to create 8 buffer circuits for each of my 8 signals. Is there any solution that I can try that I am missing?

[Henrik_Volkers · Accepted solution]

Well, you have a settle time error (fig3 in specs) ..

This settle time is due to the the capacity in the sample/hold circuit.

OK... You know the previous voltage of the S&H cap, assuming linearity, stable and known source impedance of the next channel, fixed timing , one might apply (a hopefully good enough model and) a correction.... 

(If it would be easy, I'll bett someone had done it before)

However, implementing and qualifying that is more work/time/money than building buffers or  buy a suitable DAQ. 

[aanthis]

I do not know what specs you are referencing when you mention figure 3. I also do not know what you mean by 'capacity' of the sample/hold circuit. Do you mean capacitance?

It is too bad that my >$2000 DAQ device isn't suitable for this application...

I ended up borrowing a simultaneous sampling device from someone and eliminated the ghosting that way. The PXIe-6363 does multiplexing. 

[Henrik_Volkers]

It's always a good idea to have a look into the specifications, even better before buying.

here: https://www.ni.com/pdf/manuals/377776a.pdf

Fig.3 declaire the settle time errors for various source impedances.

We try to avoid multiplexed inputs for that reason...

And yes, capacitance would be a better wording... not my native language 😉

[aanthis]

Thanks for your help