Sorry for the "radio silence". Big storm rolled through and falling trees took out a lot of power lines and pretty much blacked-out the entire county. Fortunately our utility co-op is great they restored our power tonight.
I now understand about posting pix here. Thanks!
The caps are polarised. I used caps from AVR, p/n SCCR12E105PRB. Available from several sources. I happen to get mine from Mouser. Around $1.50 a pop.
I'll try to paint a word picture. The trick, if there is a trick, is by connectiing two equal polarized (electrolytic) capacitors in series back-to-back (pos to pos or neg to neg) you have created a non-polarised capacitor with half the capacitance of each, but the voltage is twice the voltage of each.
This is not a new idea. It's been around for many, many years but it is usually applied in AC systems that require a large capacity non-polarized caps.
For 12 volts you need two of these pairs in series and if you are not sure about your power supply voltage you might want to go with three to provide some margin. I'm using a controllable DC regulator that supplies smooth DC up to 12 volts so I know I'm safe.
If you add more of these things you will reduce the total capacitance because caps in series works the same way as resistors in parallel but even with the four I used to produce 0.25 Farads there was plenty of energy to spare, as you can see.
I also included a 1 ohm resistor between the track pickups and the caps to limit the current inrush to the caps. A choke might be a better choice.
The big unknown is "does this really work with DC?" It definitely works with AC and, so far, I have not found a lot of explanations that describe exactly why it works with AC. There is a lot of wavy arm stuff out there and I have found nothing that descibes anything like what I'm doing here with DC.
I have a sort of sketchy idea to explain why it should work with DC - something to do with limited charge transfer - but that's about it. On the other hand the limited empirical evidence I have so far suggests that it definitely does work (to be honest I was slightly surprsed).
BTW, if you are worried about "overrun" just apply a low ohm resistor across the track and it will discharge the caps PDQ. You might also build a controller with a push-pull output that giveth and also taketh away.
Hope this makes a little sense.