Drift in data acquisition toolbox legacy interface

Good afternoon,

I've been having an issue with data that I collected with Matlab's legacy DAQ interface. This is with 32-bit Matlab 2008b run on a Windows 7 machine, pulling data from a National Instruments PCI-6115 board.

The issue is this: I've noticed that non-zero data will drift down to zero. If I was to give my board a steady non-zero voltage (say hooking it up to a battery with a switch that I turn on) the signal looks like a decaying exponential, with a time constant of 64.1 ms. Given a rectangle function, it's fairly clear that the zero point has moved as when the voltage is switched off the "zero" is now some point below zero. Rewriting the code with the session based interface in 2011b (64 bit) got rid of this drift, though the session based interface currently has its own issues (such as no control of DIO).

I imagine that post-processing this is not too terrible (go step by step, and for every non-zero point calculate how much it would drift down the next point and correct), but I wanted to see if anyone else has had trouble with this and if there was a good way to resolve this issue in some other fashion.

Thank you, Paul

[Walter Roberson]

Poking around, I find various references to phase drift, and polarization drift, and drifts due to temperature changes.

I do seem to find papers that say they encountered drift while using the 6115, but the ones I glanced at did not indicate anything about the cause (and to the extent I looked, drift indicated a phase drift might have possibly been for reasons completely different, especially if optics were involved.)

[Paul]

Temperature drift shouldn't be an issue. I also tried to use my post-processing system above to try and counteract the drift, and it had minimal success (a large signal spike moved the zero point down far more than it should if I'm trying to correct for a decaying exponential).

Hmmm, frustrating.