Most of your 0603 resistor is a dummy brick, the actual "resistance" where heat generation occurs is found in a thin layer of metal or metal oxide.
Yes exactly. And then it's a question of heat capacity of that thin layer, its thermal conductivity and thermal conductivity between it and the rest of the resistor.
Now, I already have some results. My signal generator (JDS2800) is somewhat limited and doesn't allow to adjust frequency and pulse width separately; it is only possible to set frequency and duty cycle, the minimum value for which is 0.1%. This signal generator drives, in the end, a MOSFET switch that turns on when there is an incoming pulse and allows current to flow via whatever (in this case the resistor under test) is connected to a 2-pin terminal block.
Now, the limitation of the signal generator means that the lowest possible average power can be P = 10
-3 * P
p, where P
p is peak power. At 50 A, for example, this means (50A)
2 * 0.1Ohm * 10
-3 = 0.25W, which is 2.5 times the typical nominal power of an 0603 resistor, so I can't properly test it for what I wanted :).
I still tested it at 50 A, 200 Hz, pulse width ~5 usec, rise time ~1.5 usec. It died in ~2 minutes, failing open. No magic smoke escaped.
I replaced the resistor and continued testing.
1) 15 minutes at 30A pulses (0.09W, again not counting the non-zero rise time, so probably ~15-20% less in reality): all good, no change in resistance. Well, maybe 0.2 mOhm less, if my TR1035 isn't lying, and I have no reason to assume that it does -- but I may have simply placed the test leads differently.
2) 15 minutes at 35A pulses, ~0.12W calculated, probably ~0.1W actual. Or I think even more: 4 usec at 35A + 1 usec rise time. No changes in the observed voltage drop waveform so far, the resistor became warm. I'll let it run until I get bored and then measure its final resistance and call this mode of operation safe enough :).
3) I think I need to build a pulse generator with independent controls for pulse width and frequency.
...update: having taken a closer look after 35 minutes, I see that the pulsed current has increased to ~36A. TCR in action, I guess.