RE: What is the pupose of ESc capacitors?
These days most new ESCs don't have large power caps inside for a good reason. Whether they are in a surface mount package or not it still takes about the same volume of package to get the same capacitance. Tantalum capacitors aren't very good for high ripple current applications (they get pretty hot when you demand high current pulses out of them) and even surface mount electrolytics have their drawbacks. The main concern is that they too get hot and they transfer that heat to the circuit board they are on. Using an external capacitor gets that heat out of the ESC package. Even if you don't notice a difference in the performance of the ESC with the addition of an external capacitor you have still dropped the instantaneous peak current demanded from the ESC at each PWM (Pulse Width Modulation) turn on by more than ten times which will make the ESC run cooler and extend it's life.
Access suggested earlier in this post that an external capacitor had far higher resistances and was therefore ineffective but there were some inaccurate assumptions used in the rationale. Sorry to be picky but I wanted to get the facts out here.
First it was suggested that you could use 3 1000 uF surface mount caps in parallel in the ESC. The highest SMT aluminum electrolytic capacitor value that is available 'off the shelf' is 470uF for a 16V rating and they are slightly more than 3/8" in diameter and almost 1/2" tall so you have increased the size of the ESC by about 1.5" surface area. Also the best ripple current rating you can get in an SMT package is about 670 mA and they cost about $0.30 each if you buy them by the reel. By comparison you can get an external 1500uF capacitor (1/2" diameter X 1" long) with a 2.2A ripple current rating for $0.33 each in the same purchasing volume. Also, the ESR (Effective Series Resistance) of the 470u SMT caps is 0.15 ohms (3 in parallel is therefore 0.05 ohms) where the external 1500u is at 0.03 ohms for about 1/3 of the cost .
The resistance of the wire used in the example was 0.1 ohms for the wires. Assuming the wires are a reasonable length (I will use 2" leads as a reference (so 4" overall wire length to and from the cap)) the actual resistance even if you use 18 gauge wires is actually 0.0021 ohms. This is straight from the AWG tables (18 AWG wire has 6.385 ohms of resistance per 1000 feet). (To get the 0.1 ohms suggested for a 4" wire run you'd have to use 35 AWG wire.) This makes the resistance of the path 0.0321 ohms counting the wires and the capacitor. At a 10 amp draw this means a voltage loss of 0.321 volts - better than the 0.5 volt loss for the SMT caps with the added plus of a higher ripple current rating. Traces on the PC board are not going to be enough of an improvement from 2.1 milliOhms of the wire to offset the difference in ESR and with the combination of radiated and conducted heat inside the ESC from the capacitors it makes it a bad tradeoff.
It was also suggested that soldering the capacitor leads directly to the ESC could help. Since the capacitor leads actually have a smaller cross-sectional area than the wire you would likely use to hook them up you will drop more voltage across the leads than across the wires (assuming your solder joints are good) so it's actually better to use the wires. In either case the actual lead or wire resistance is less than 10% of the loop loss and is in effect swamped out by the internal resistance of the capacitor. Given 18 AWG wires the actual voltage drop across the wires is only 0.021 volts compared to 0.3 volts dropped across the cap.
That is also sort of a mistaken impression. When the voltage on the ESC output is above the voltage in the capacitor the capacitor is being charged. When the voltage drops below the level in the capacitor the current flows out of the capacitor in an effort to oppose the change in voltage. The ESR specification is, then, really a rating of the impedance of the current source the capacitor represents and therefore the maximum current that it can source. Using the 1500u capacitor above as an example again since it has an ESR of 0.03 ohms then using Ohm's Law (Current = Voltage / Resistance) and using a standard 6 cell battery voltage (7.2V) as an example it can source 240 amps on an instantaneous basis if it drives into a 0 ohm load (in other words a dead short). Now you can see why a capacitor can get hot when it is filtering a PWM output. That is why you see a ripple current rating - no capacitor is going to survive a 240 amp ripple current so the cap I used as an example is only really good for a ripple curent equal to about 1% of it's total capacity on an average basis before it overheats and blows up and it's one of the better ones.