"Do not roll your own random number generator for one-time pads, etc."
Working in the field of spectroscopy, I have harddisks full of randomness. It's called quantum noise and it is part of any particle detection experiment. For home use, you could rig up a Geiger counter to give you very decent randomness.
If you want maximum entropy you might want to compare particle pair distances for each bit of entropy: (time_p2 - time_p1) > (time_p4 - time_p3) ? 1 : 0. That should remove digitization artefacts (e.g., from detector dead-time, non-normal distribution of your source, ...).