+1 on the above. Lacking anything else, it's a pretty standard method (e.g. it's in the "Model Engineer's Handbook" by Tubal Cain), but also +1 on the point about testing some known stuff first to get a good idea of how different those sparks really look.
AES