Cooking eggs on a mosfet

[ronv]

Motors hate micros.

Alec_t has the most likely problem. You don't want the high current of the motor going thru the same ground as the micro ground.
You might want a big cap right across the bridge in any case, especially if the leads are long to the power supply.
Sometimes the brush noise can get it as well. Some small ceramic caps - one across the motor and one from each lead to the frame of the motor may help this. If possible tie the frame of the motor to AC ground. And then just a decoupling cap at the micro itself.
One last thing to watch out for is that if you reverse the motor before it has stopped the back emf can "pump up" the big supply. You can probably measure this with a meter as the motor probably takes a while to stop.

 

[mindoverflow]

alec_t well, yes an no, at the beginning the intention is to make something like this:

But squeezing a power regulation, rs232 communication, a uC, the hbridge circuit, current limiting, 10 leds and about 12 inputs and outputs in a 9x15cm perfboard you end up with a messy circuit (specially when you find you made 3-4 mistakes that you'll fix).
Then if you consider the above diagram a star grounding, then I think yes, this is pretty much how it looks like.
I tried triggering the h-bridge with a 555 circuit (the uC is separated from the hbridge but still sharing the ground) the uC is still being reset then that way I guess an optocoupler wouldn't help much, I don't have one, but i think I'll try an optical switch instead just to be sure. Note that the first hbridge circuit (no transistors) didn't reset the uC working in the same conditions.
ronv that looks to be quiet complicated, I used some big capacitors between the supplies' leads and a small ceramic capacitor between the motor leads, but nothing improves.