I've been doing some stuff with supercapacitors recently, mostly using them to store power from solar panels. One of their shortcomings is their low voltage rating, typically <3V, so using several in series may be necessary. Like any polarized capacitor, they have to be balanced when connected in series, to ensure the voltage is shared equally. There are some off-the-shelf parts for this, but all the ones I have seen don't actually balance until the full voltage is reached. This is no good for me, as the unpredictable nature of solar means that they may
never reach full charge.
Thus I put together this circuit, which will start balancing as long as there is enough voltage for the op-amp to function - the MCP6041 is specified down to 1.4V, and it still functions well enough to provide some balancing down to 1V per capacitor.
The schematic shows one section. To balance a string of capacitors, one section is needed per capacitor, with the three marked nodes connected together between sections. With n sections in series like that, the resistors labelled R5 form a voltage divider with exactly 1/n of the total voltage across each resistor, which is the target voltage per capacitor for perfect balance. The op-amp is simply connected as a differential amplifier, comparing the target voltage with the actual voltage, and turning on the transistor to drain charge if the actual voltage is too high. Since perfect balance is impossible with real components, the feedback values are chosen so that it doesn't drain charge if the actual voltage is only slightly high, otherwise it would constantly try to drain a little bit. The diodes are to prevent reverse bias, since that's still possible if the capacitors are drained much below the point where the op-amps stop working.
With SMT parts, it easily fits inside the footprint of a big supercapacitor.
I put four of them in series for 10.8V total, suitable for being charged by a nominally 12V solar panel. Note that I designed it to use tiny SMT reverse polarity protection diodes, but I have a drawer full of big schottky diodes that I'm never going to use up otherwise, so I soldered some of those on instead.
The green text on the capacitors is the capacitance that I measured. I bought a bunch of these "500F" capacitors, and although some are within 500F ±20%, some of them are lower. The are all within 400F ±20% though, so I suspect that they are really meant to be 400F and have been mislabelled.
Here's a thermal image of it being charged at 5A. I have no experience with thermal imaging, so the temperatures might not be right, but it shows that the two capacitors on the right, the smallest ones, increase in voltage the fastest, and so need charge to be drained from them.
And the reverse, being discharged at 5A. As expected, it's the ones on the left that get hottest.
And an image of it after it has been sat fully charged for a while, showing nothing since it's all balanced.
After all that, I'm not entirely sure how necessary this really is. There's not a lot of hard data about the effects of reverse charging supercapascitors. There's a paper by NASA showing that it reduces capacitance, but they only tested reverse charging to full voltage. If small reverse voltages are tolerable, then a simple diode may be sufficient, in combination with the voltage clamping type of balancer that you can find on eBay. however, my design does have the advantage of dissipating power over the whole charging cycle instead of only at full charge, which should mean the peak temperature is lower. It's also just as simple.