Powered by Indico
v3.2.5
Jpp Track Reconstruction - Bouke
Bouke gave a 20-30 min presentation on the current track reconstruction algorithms in Jpp.
Some potential items for improving the documentation were put forward:
Bouke: Shouldn't this be exactly the same as the tangent of the Cherenkov angle?
Karel: Only if the group and phase velocities of the light are the same.
Bouke: Why for example for JPrefit do we define the quality as Q = NDF - chi2 / 4 / NDF?
Karel: We divide by the number of degrees of freedom, in order to adequately weigh fit directions with small associated hit statistics versus fit directions with very large corresponding hit statistics. The division by 4 was found using a scan and was chosen to provide sufficient separation between the qualities of the individual fits.
Some questions regarding the present status of the reconstruction and potential improvements were discussed at the end of the presentation:
Ronald: Translating time-over-threshold into an amount of photons is really hard, this keeps us from using them. There may be systematics etc which we don’t understand. Efforts to understand and model the ToT and ToT-to-charge conversion are essential for improving the reconstruction.
Dorothea: Efforts, including those of Bouke, have improved this.
Karel: Figure 22 of my Master Thesis (see materials) shows that the deviation of the ToT-equivalent charge w.r.t. the effective number of photo-electrons responsible for a PMT pulse is a lot smaller than the deviation of the expected number of photon hits w.r.t. the same quantity. So the argument that the ToT-distribution and ToT-to-charge conversion should be understood better before we use it in reconstruction may be less applicable than people think.
Jordan: Currently we use only the hit/no hit probability, but incorporating the ToT we could exploit the full Poissonian hit statistics (i.e. the not only the probability whether a PMT is hit, but also the probability that the PMT is hit a specific number of times), which may enhance the reconstruction.
Karel: We might be able to define a second, quality parameter in one or more of the reconstruction stages which include no-hit information. Any solutions with large discrepancy between the two quality parameters (i.e. the one including and the one excluding no-hit information) can then be distinguished and potentially excluded. The only thing to keep in mind is how including no-hit information affects the CPU time. After all, you will have to loop over all hits within the cylinder around your muon, not just the ones with a registered hit.
Bouke: Maarten indeed had a similar idea to include a second quality parameter.
Jordan