Voltage Drift on an Open Circuit (and how to get rid of it)

Question

I'm trying to create a custom lick-port to detect mouse licking behavior. I am following a design published here that uses the junction potential between the mouse's tongue and the metal water spout to detect when a lick occurred, see image below:

To do this, I'm acquiring the voltage signal as an analog input to a National Instruments DAQ card (PCIe-6363) via an NI breakout box (model SCB-68). I'm reading out the voltage using Matlab R2014a, using the DAQ toolbox to read out voltage from the analog input (ai) channel on the NI DAQ card.

My problem is that when the circuit is open, the voltage is slowly drifting to a particular value over about 8 minutes (it looks to be around 2.8V):

What I really want this system to do is have a voltage of 0 until the circuit is closed, at which point I can detect a junction-potential that signifies a lick. So, why is this drift occurring? What can I do to prevent it during the open-circuit condition?

Here is a sample of what happens when I hold the ground in one hand, and touch a wetted finger to the + wire. Each sharp drop coincides with the circuit closing:

Now, I could still potentially use this, just by calculating the derivative and seeing when that exceeds some threshold, but I want to know why this isn't working as expected.

Thanks!

Answer 1

The slow drift may be due to leakage currents in your sensor. Since the input is open-circuit (theoretically infinite impedance), even a tiny leakage current can cause a significant offset voltage.

The simplest solution is to put some resistance across the sensor. 1 MΩ will likely swamp the leakage signal. 10 MΩ may work too. It depends on the magnitude of the leakage and the input impedance of whatever is receiving the raw signal.

You want to make the resistance low enough to bring the offset you are seeing down to don't-care levels, but not so low as to significantly load the mouse-tongue-on-metal battery. I suspect the latter has significantly lower impedance than 1 MΩ, in which case that's a good value to use.