Clustering4D: timing cut default too small
In [Clustering4D]
the default time cut is 3*time_resolution
of a detector.
I think this default should be increased, up to discussion, which value is more reasonable. But I think the Timepix3 illustrates quite well why that's too small:
Looking at W5_E2 (the one we used at DESY as a timing) plane has a time resolution of about 1.5ns (without calibration). If we put this value into the geometry file, we (implicitly) use a time cut of 4.5 ns.
However, this value does not describe the behaviour on pixel level, which can be much slower due to:
- timewalk for small "secondary" pixels
- slow front-end responds for small signal
- ...
The plot below shows the time difference of all pixels (excluding the seed) minus the seed pixel timestamp (i.e. the cluster timestamp), both for using the earliest pixel and for using the pixel with largest ToT as a seed. Here we see that a time cut of 100-200ns is more reasonable.
What's your opinion? Generally, a too large time cut on the clustering is only bad at a really large occupancy (or very very long frames). A too small time cut can chop your clusters apart and lead to wrong cluster sizes and bad tracking.
Of course, the user should always understand what they're doing, but I think in most situations taking a too large cut is less harmful than taking a too small cut in the first iteration of the analysis.