Interesting (maybe?) question here which I am struggling to answer. Hoping someone can help.
I have a capacitor network; 5 caps in parallel and then two modules of that configuration connected in series. I am measuring the voltage of the bottom 5 caps using an Arduino Nano. The Arduino and cap network are only linked by the cap negative (gnd) connected to the Arduino GND pin. A bit crude but means I can measure negative voltages (i.e. 0V registers as 3V3 on the Nano):
I charged the circuit up for just shy of 24hrs to 2V4 from V+ to V-. I checked the voltages across each of the caps and they were reasonably balanced (couple of hundred mV at most). I.e. each cap is sat at 1V2, no current into or out of the system (bar a few uA due to cap internal leakage).
Now, I disconnect the V+ power supply (S1 goes to posn 3) and connect a load (to posn 1 of S2). As expected, the network discharges into the load, favouring discharging the caps with lower value series resistors. I.e. C1 & C6 discharge fastest, C5 & C10 slowest.
At ~75min I set the stack to open circuit.
At ~140min I re-apply the load.
The observation/question. When the caps with fast RC constants (C1 & C6) reach 0V across the cap, they are fully discharged. However, they then get pushed negative. I.e. when measuring across the cap, a negative voltage appears. It goes up to negative a few hundred mV. Why? [Note: I have subtracted 3V3 from the voltage readings to get the correct reference with respect to the stack, not the Arduino].
I've scratched my head over this for a while and cannot understand what is driving the negative voltage across that cap, while the others in parallel remain positive.
I've deliberately been vague on the precise readings because the input impedance of my multimeter, and the Nano suck. (I am in progress making a little voltage follower to solve this issue, and, also use a potentiostat rather than an Arduino!) Apologies on this front.
Interestingly, I have modelled this circuit in Matlab and it predicted this. I was somewhat hoping my model was wrong...maybe not?
Many thanks in advance,
PS: Massive thanks and credit will be given to anyone able to help if they're happy me crediting them on a research paper. If not, happy to not credit (I know some people don't like to be). PPS: I've not gone into detail about what the circuit represents. Happy to elaborate if people are interested (it's a skew-gaussian distribution BTW).