What is the best way to filter spad data without losing the faint signal?

[AbigailBM]

I’ve been prototyping a small device that uses a single-photon avalanche diode array, and I’m stuck on a practical problem. The raw data stream is incredibly noisy, and my attempts at filtering seem to either lose the faint signal I need or require more processing power than my embedded system has. Has anyone else hit this wall with SPAD sensor data?

[Jeffrey23]

Yep, I’ve been there. The raw SPAD stream looks like static until a real photon shows up, and every filter I tried rinsed away the faint signal or left me with hundreds of thousands of timestamps to crunch. I started with thresholds, then a moving average, then a Wiener-ish filter, and still nothing reliable without crashing the CPU.

[StevenH]

A small win: cooling helped a bit. The dark counts dropped noticeably, and the baseline noise shrank, but the faint event rate didn’t rebound. I measured the baseline drift over a minute and it still drifted with ambient light.

[Penelope_T]

Could the real problem be timing jitter or afterpulsing, not just the noise floor? I keep wondering if the detector’s timing resolution is the bottleneck and the filter is trying to fix the wrong problem.

[Aria_W]

I wandered off topic for a moment thinking maybe gating hardware or binning in hardware would help, but I kept circling back to filtering. Ended up feeling like I was spinning wheels and didn’t get a clean path.