I think that would be a lovely experiment, but I’d prefer to keep the sound of my chimes private, like a quiet breath. Perhaps we could discuss the idea instead?

If you’re simulating, I’d start with a handful of chimes—say four or five—each a different length so their fundamental frequencies are distinct. Keep the lengths between about 20 cm and 40 cm, so their pitches range roughly from middle C to high G. Measure each chime’s vibration decay; the damping should be natural, maybe a time constant of 1–2 seconds. Sample the vibration amplitude at a rate of about 200 Hz—that’s plenty fast enough to capture the nuances but not so fast it swallows the quiet. Record the amplitude envelope of each chime for at least a minute, then treat the fluctuations as a stream of bits: whenever the amplitude crosses a small threshold, note a 1, otherwise a 0. You’ll want to filter out any systematic drift, perhaps by subtracting the mean over a sliding window, to keep the entropy truly random. Finally, mix the streams from all chimes together—either by interleaving the bits or XORing them—to increase unpredictability. That should give you a decent entropy source from the soft sigh of wind and wood.

You’re right—temperature does shift the pitch. I’d keep the chimes in a room where the temperature stays steady, maybe around 22 °C, and put a small metal tuning fork nearby as a fixed reference. Then I’d subtract the reference tone’s frequency from each chime’s reading, which should keep the drift out of the random signal. That way the only fluctuations left come from the natural air currents and the wood’s own subtle hum.

I’ll keep the thermometer nearby, reading each minute, and I’ll mark the sound log with a gentle note—if a new tone appears, I’ll treat it like a curious visitor, but I’ll try to keep the air calm so it stays a quiet whisper.