So I decided to do an extremely rudimentary circuit analysis to see if I could gain a better understanding of what's happening.



I've made some pretty major simplifications - the transistors have been swapped out for simple switches, the motor windings are shown as just resistors, and there's probably all sorts of other things I should technically add, like the windings probably have some amount of capacitance between them, or a resistor to represent the IR of the battery, etc. But for the most part I was looking to get the basic gist of what's going on and I think this helps in that regard so hopefully it's not too gross of an oversimplification.

The graph of the voltage between points A and B on the circuit over time isn't necessarily to any kind of scale - what I don't know without some serious calculating/simulating is 1) how far the voltage actually drops when the switches are closed and 2) how long the voltage actually takes to drop. Thinking about it now, if I go with the numbers I used in my last post, an 8.4V battery with a 0.02 ohm IR should have a 0.6V drop when outputting 30A. That seems pretty realistic from what I remember seeing on battery discharge curves.

Intuitively, a cap with more capacitance will take longer to discharge so you'll have a higher average voltage over the duration of the switches being closed. And capacitors with a lower ESR will have less of a voltage drop which should also equate to a slightly higher total average voltage over the timespan the switches are closed. So in that regard, I could see how it's theoretically possible that a larger cap could provide performance gains by keeping the average line voltage slightly higher over the duration of a PWM pulse.

Some ESCs have adjustable PWM frequencies, but I wouldn't look at that as a way to increase the average voltage over the duration of a PWM pulse - windings are inductors and they have higher impedance at higher frequencies so you'll just end up heating up your motor more.

Anyway, all that said, my best guess is the real purpose of the capacitor is to prevent the switching transistors from seeing instantaneous voltage changes. Without a cap the voltage drop in my graph when the switches close would be instant, and voltage impulses contain lots of high frequency noise that probably isn't good for the transistor.