USB-6363 Analog input SINKS CURRENT when another AI channel is enabled. Wha???

[FunGussy]

The pic below shows a very simple configuration.   And using NI-MAX software:

1.  I sample AI2 alone, the ammeter reads 0.00uA (as you'd expect;  10GΩ input impedance and all, dontcha know).

2.  I then sample AI2 and AI4,  the ammeter reads 85uA.   So right away, something is fishy:  the ammeter should never ever ever read above zero.  

3.  I still sample AI2 and AI4,  and disconnect the blue wire, and the ammeter drops back down to 0.00uA, and the disconnected (and now floating) blue channel perfectly tracks (as displayed on the NI software) the 1.8v of AI2.

I'm flummoxed.   Can anyone explain this??

Speculation:   I smell a bug in the analog IO multiplexer which sits in front of the ADC.   It's as if the mux is a make-before-break,   and for a tiny fraction of the total cycle time, channels get shorted together.  Here's more evidence for this:

At 1MHz sampling rate:  85uA

At 500khz:   42.5uA

at 250khz:   21.25uA

If the short-circuit time is fixed at, say, 10ns,  the above dependency on sampling rate makes perfect sense:  slower sampling means fewer shorts-circuits-per-second, and lower uA.

Comments welcome.

[santo_13]

I guess it is current required to charge the sample and hold circuit of the multiplexed ADC.

So, when the ADC is sampling AI4 at 0V, the sample and hold circuit is at 0V, next, when it measures AI3 at 1.8V, the sample and hold circuit at 0V need to be charged up to 1.8V

[ft_06]

And are the ADC measures correct ? Only ammeter is "wrong" ?

We have the 6353 and struggling with some noise that seems cyclic when measuring 2A (noise is up to 1A) with a 3ohm shunt (thus measuring 6mV and dividing by shunt value of 3), while other setups are not noisy.

But it is DAQ measure itself so probably not same topic than yours

[santo_13]

These articles explain ghosting in multiplexed ADCs
https://knowledge.ni.com/KnowledgeArticleDetails?id=kA00Z000000P84RSAS&l=en-US

https://knowledge.ni.com/KnowledgeArticleDetails?id=kA03q000000YHy6CAG&l=en-US

https://knowledge.ni.com/KnowledgeArticleDetails?id=kA00Z000000P9mKSAS&l=en-US

https://www.ni.com/en/shop/data-acquisition/measurement-fundamentals/analog-fundamentals/is-your-dat...

[ft_06]

Well, with your articles, it reminds me that we have seen the same noise when using only 1 AI and low acquisition rates (and also with or without grounding, with or without resistors on differential pair, on windows, on linux, within a PC or in PCIe external chassis on Thunderbolt port,...)

So my issue is not a multiplexed ADC issue, it won't help, let me not hack this topic 😉 and we have submitted a ticket to our vendor