Difference between revisions of "Optics Meeting Sep 6 2023 1100AM ET"
From Moller Wiki
(→Agenda) |
(→Agenda) |
||
| (6 intermediate revisions by 2 users not shown) | |||
| Line 7: | Line 7: | ||
=Agenda= | =Agenda= | ||
| − | # Comparison of fit coefficients between hole center and MC angles (Tyler) | + | # Comparison of fit coefficients between hole center and MC angles (Tyler) [https://moller.jlab.org/wiki/images/Fit_sieve_hole_centers_Sep6.pdf PDF] |
| − | + | ||
| − | + | ||
| − | + | ||
# Energy dependency of the parameterizations [Zuhal] [https://moller.jlab.org/wiki/images/Optics_06Sep2023.pdf here] | # Energy dependency of the parameterizations [Zuhal] [https://moller.jlab.org/wiki/images/Optics_06Sep2023.pdf here] | ||
| + | # Full GEM modules in Geant4 [https://moller.jlab.org/wiki/index.php/File:Sep_06_update.pdf] Bill] | ||
=Attendance= | =Attendance= | ||
| + | Zuhal, David A., Ciprian, Juliette, Andrew, Paul S., Chandan, David Firak | ||
=Minutes= | =Minutes= | ||
| + | # (Tyler) Use of hole centers as truth values: results generally promising: residuals for theta and phi have only slightly larger standard deviations than with Monte Carlo truth values. Somewhat larger biases in the means, but presumably this could be corrected for. Looked at individual best-fit coefficients vs. pass (beam energy) - hard to see systematic pattern. See double-peaking in phi residuals for pass 1 (only), and rather large width for the theta residuals, also for pass 1. The double-peaking in phi at one pass seems to be a case where the fit is essentially a linear fit of phi_target to phi_gem. It was suggested that we look at the residuals for selected individual holes. I wonder (once again) whether in the case of phi whether we should be fitting to the local (sector-based) values, rather than the global ones to properly capture things that might depend on, for example, phi^2. Lots of discussion of whether additional parameters in the fit functions might be needed based on knowledge of the spectrometer optics; for example, the phi defocusing will be related the E' and theta (and hence to r_GEM and r_GEM' in a complicated way), where the present function uses only phi_GEM and phi'_GEM to extract phi. | ||
| + | # (Zuhal) Looked at "common fit" to 10 beam energies, running from 2.0 to 6.5 GeV. Strange results when attempted to do a scaler normalization of the input data (i.e. the GEM variables)... the idea is to set each of r_GEM,r_GEM',phi_GEM,phi'_GEM to a mean of 0 and rescale to +- 3 sigma of the mean. Notes added after the meeting: Zuhal corrected a bug with the scaler normalization (see posted slides) and also included very promising residual plots for theta and E' - we can reconstruct E' quite well! The effect of the "standardScaler" normalization of the input data did not seem to improve things (residuals were generally somewhat worse). Will discuss all this next week. | ||
| + | # We didn't get a change to look at Bill's posting on additions to remoll; hopefully next week. | ||
=Meeting link information= | =Meeting link information= | ||
See email invitation, or contact David Armstrong, Kate Evans or Jennifer McAllister for Zoom link | See email invitation, or contact David Armstrong, Kate Evans or Jennifer McAllister for Zoom link | ||
Latest revision as of 19:36, 8 September 2023
Back to Main Page >> Optics Meetings
Contents
[hide]Agenda
- Comparison of fit coefficients between hole center and MC angles (Tyler) PDF
- Energy dependency of the parameterizations [Zuhal] here
- Full GEM modules in Geant4 [1] Bill]
Attendance
Zuhal, David A., Ciprian, Juliette, Andrew, Paul S., Chandan, David Firak
Minutes
- (Tyler) Use of hole centers as truth values: results generally promising: residuals for theta and phi have only slightly larger standard deviations than with Monte Carlo truth values. Somewhat larger biases in the means, but presumably this could be corrected for. Looked at individual best-fit coefficients vs. pass (beam energy) - hard to see systematic pattern. See double-peaking in phi residuals for pass 1 (only), and rather large width for the theta residuals, also for pass 1. The double-peaking in phi at one pass seems to be a case where the fit is essentially a linear fit of phi_target to phi_gem. It was suggested that we look at the residuals for selected individual holes. I wonder (once again) whether in the case of phi whether we should be fitting to the local (sector-based) values, rather than the global ones to properly capture things that might depend on, for example, phi^2. Lots of discussion of whether additional parameters in the fit functions might be needed based on knowledge of the spectrometer optics; for example, the phi defocusing will be related the E' and theta (and hence to r_GEM and r_GEM' in a complicated way), where the present function uses only phi_GEM and phi'_GEM to extract phi.
- (Zuhal) Looked at "common fit" to 10 beam energies, running from 2.0 to 6.5 GeV. Strange results when attempted to do a scaler normalization of the input data (i.e. the GEM variables)... the idea is to set each of r_GEM,r_GEM',phi_GEM,phi'_GEM to a mean of 0 and rescale to +- 3 sigma of the mean. Notes added after the meeting: Zuhal corrected a bug with the scaler normalization (see posted slides) and also included very promising residual plots for theta and E' - we can reconstruct E' quite well! The effect of the "standardScaler" normalization of the input data did not seem to improve things (residuals were generally somewhat worse). Will discuss all this next week.
- We didn't get a change to look at Bill's posting on additions to remoll; hopefully next week.
Meeting link information
See email invitation, or contact David Armstrong, Kate Evans or Jennifer McAllister for Zoom link
