Sampling / time resolution (“sample time”) and timing behaviour of the CosmicWatch muon detector.

Here’s what I found on the sampling / time resolution (“sample time”) and timing behaviour of the CosmicWatch muon detector:

The SiPM output pulse is very fast: it has a rise time of a few nanoseconds and decays over a few hundred nanoseconds. (ar5iv)
Because that pulse is too short for the Arduino to sample reliably, the output is passed through a peak-detector circuit which stretches / holds the pulse to a width of ~100 µs. (Hackaday)
The Arduino’s ADC reads the peak detector output at discrete intervals. The sampling (i.e. successive ADC measurements) occurs roughly every 5.8 µs. (SETI)
Because of that discrete sampling, the uncertainty in the timestamp of the trigger (i.e. when the muon-induced pulse actually occurred) is on the order of ±4 µs. (SETI)
In addition to that, when data are sent to the computer via serial (USB), there’s an extra timing uncertainty (latency) of a few milliseconds from the serial interface. (Scribd)

So, in practical terms:

The “stretched” pulse is ~100 µs in duration, which gives the Arduino enough time to sample it.
The time resolution (precision) of the recorded event time is limited by the 5.8 µs sampling interval, giving ~4 µs trigger uncertainty.
There is additional jitter / latency introduced when transferring the data to the PC.