Dear All,

Under the supervision of Cameron, Alic has developed the channel-by-channel constants for the shapes used to fit the APV25 pulses from the SVT, where those fits are used to determine the amplitude and time of the hits in individual strips. As has been discussed, the reconstruction has previously used a single fit shape of unknown origin, which obviously does not provide a good fit to the data.  While it is quite obvious that the fit shapes Alic has extracted from the data are a better fit, it was also obviously important to sanity check things by ensuring that the tuned fit shapes don’t affect the key performance parameters of the hit data: the efficiency and purity, and the amplitude and time resolutions. On Tuesday, we saw direct comparisons of the amplitude and time distributions between data reconstructed with the ad hoc fit shapes and Alic’s improved fit shapes, and it is clear that - as expected - none of these key performance parameters is degraded by using the shapes that provide a better fit.  Meanwhile, we are already aware of other mistakes in the downstream steps of hit reconstruction (especially in thresholds for clustering individual strip hits into the hits that provide our measurements) where those will need to be resolved to get a better handle on how much improvement the new shapes provide.

On Tuesday, it was proposed to deploy the new constants for the APV fits into the database for 2019 and 2021 reconstruction. The only reservations expressed were focused on two points:
1) Our hit making is working, so what motivates changing it? (if it ain’t broke, don’t fix it)
2) We should more carefully explore this before making it the default for our reconstruction, simply because it is a change in how we have been doing things

I would push back on both of these:
1) is a generic argument against improvement, and simply a non-starter, especially for an experiment striving for best sensitivity
2) supposes that there is some existing baseline for the reconstruction that we don’t want to correct, even where we know it is wrong, because we don’t want to have anything moving underneath active analysis efforts.  However, we have not even processed a Pass 0 on 2019 and 2021, so there is no baseline reconstruction, and no active analysis.  Waiting to make such changes until we DO have active analysis only ensures that we will have to grapple with that issue in the future.

In short, I think it is clearly misguided to choose to treat our data using randomly chosen calibrations, when the work has been done to derive sensible ones.  Continuing to pick apart the minutiae of the effects of this change, rather than spending our limited effort attacking other important improvements in our hit reconstruction does not seem like a wise course.  This is especially true because those other improvements are necessary to illuminate the impact of the change to pulse fitting.

Therefore, I would propose to get the new pulse fit shapes into the database for reconstruction, and to shift our efforts to the improvement of clustering, and estimating hit position and time, which can feed directly into our tracking performance.  Certainly, if there are arguments substantively different than (1) or (2) above, I would like to hear about them.  If there are not, and a consensus cannot be reached here, the EC should take up the question rather than continue a deadlock.

Best Regards,
Tim


########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the HPS-SOFTWARE list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=HPS-SOFTWARE&A=1