Thanks for the pictures. I dug out the scope and did the same experiment. My waveforms look kinda like yours except that the external influence of air currents makes the curves all "wobbly". I can see when I reached across the bench to press the scope trigger button. Interesting that the waveform integrates to an extremely nice-looking smooth exponential.
We've been at this for a while. If I were to paraphrase your input, It would be something like, "What you're doing is trivial; I've done it; Here's the pix".
My response is, "What YOU are doing is trivial; thanks for the input; nice pix".
The critical difference in our results is that my signals are 60dB weaker than yours at the digitizer.
The whole concept behind this thread is/was to resolve sub-millivolts with a PIC. It's not a thread on amplification of sub-millivolt signals... OK, it has turned into that, but that is not helpful to me.
In the spirit of "fix it in software" I'd planned to do the derivatives to predict the asymptote of the exponential in software. After looking at the scope, I've decided that the integration/averaging times required are gonna be too long for that to be a solution.
As for experiments, your PIC breadboard must be much better than mine. I'm using an old Heathkit unit with built-in generators and power. The power transformer sits right under the breadboard. I'd never even try to get 100microvolt resolution on that system. I had enough trouble with 10-bits on the A/D in a 16F877a.
I'm giving up for now. I've got some chip thermistors that I'm gonna mount on opposite sides of some very thin ECB material. Measuring differential resistance is MUCH easier and the thermal time constant can be much shorter.
Thanks for the inputs.