I've got new results from the capacitor based supply. Same basic procedure: repeated 50J pulses into a 1mOhm weld spot resistance. That resistance was a bit high this time (1.3mOhm), and the resulting pulse duration is approx 40ms.
During this test, the 5xBCAP0310 were kept topped at 12.5V by the HP server supply. There is no current control mechanism yet, so I assume that the PSU went to its current limit of 100A during recharge. When I pulse faster than maybe once a second, it shuts down after a few pulses.
The caps receive approx 1050A^2 * (5*1.5mOhm) * 40ms = 330J during each pulse, so I decided to fan-cool them and limit the pulse rate to once every 4 seconds. The results are as follows (ambient=25°C, I hadn't let them cool down from a previous test):
There is some more headroom (max 70°C), but this shows me that a 5S configuration is not capable enough. On the other hand, I could not see any degradation or damage from this abuse because the current indicates that their ESR is not rising:
I did some calculations, and a 3S2P configuration will reduce the energy loss per 50J pulse to 113J, which is only one third of what it is now. This will allow my targeted pulse rate of once per second. But how to charge them? I don't really want to go for the additional cost of a 70A step down converter...