i used software defined radios to make a few sets of one time pads with entropy. The randomness of proper SDR or even a webcam in a lightproof coffee can or something is demonstrable with any of the tools for "testing randomness"; sibling is correct, MEMS are notorious for "noise" and that noise is "random", one can use a GM tube to trigger interrupts and use the timing to get entropy.

I don't know how you'd prove something is truly random, though, just that it looks and acts "random" enough; fitness for use.