Update on 50 um Sensor Calibration using LNLS Data Mathieu Benoit

Outline Reminder of previous results – XAFS – Fluorescence measurements Improved Calibration procedure Calibration of A06-W0110 and results

Energy Measurements with monochromator (XAFS1) X-Ray incident beam goes to Crystal Monochromator and is then delivered to the experimental area. Energy can be selected with 0.1keV accuracy along the 5-15 keV range. 2 sensors (C04-W0110,D09-W0126) were tested in this beamline to obtain single pixel calibration for a subset of pixel (beam 2.5x3mm2) Analysis on one of the sample will be presented by Mateus Vincente for UFRJ

Global Surrogate Fit After disscusion with LNLS experts, it seems there is a problem with the XAFS1 bealine energy calibration, a new calibration will be performed and hopefully correction to the dialed energy will be provided

Fluorescence measurements (MX1) TargetCoCrCuFeMnNiTiVZn E (kα) in keV X-Ray beam (7.9,15.8 keV) 1e12 photons/ s Target Fluorescent photons field ~ 1000 single pixel clusters per pixel / 50k frames (1ms integration) -> ~30-45 min per fluo Only tested with one assembly (A06-W0110) but possibility to run multiple assemblies at the same time (Large Fluorescence radiation field ) Timepix+FITPix

Single Pixel TOT Spectrum First Harmonic scattered

Calibration Curve

Single Pixel Energy Spectrum First Harmonic scattered

Improved Calibration procedure Clustering (>15 keV) Extraction of 65k Histogram to ROOT Format (TH1) Clustering (>15 keV) Extraction of 65k Histogram to ROOT Format (TH1) Raw Data 65k Histograms per source Peak Value extraction (Tspectrum +1.5 sigma gaussian fit) N*65k peak values per source Iterative surrogate fit: Bad point taken out from fit Iterative surrogate fit: Bad point taken out from fit The calibration code has been optimized to reduce execution time : Take advantage of large memory -> Parallel processing of 65k pixels Skip clustering for low energy (low charge sharing probability for thin sensors) ROOT6 -> TreeReader C++ Subroutine implemented for big data parsing A factor 1000 has been gained in calibration speed !! (4-5 hours on a single CPU vs a 15-20h on 256 CPUS )

Peak TOT Distribution

Surrogate Fit example

Threshold dispersion electrons

Shift between Fe55 and the rest of dataset

Chi2 Distribution

Calibrated results Pixel per pixel calibration, E= Global calibration, E= Clustered Pixel per pixel calibration, E= Fe 55 E=5.89 keV

Calibrated results ---- fit results : Cd Pixel per pixel calibration, E= Global calibration, E= Clustered Pixel per pixel calibration, E= Cd109 E=22-25 keV

Calibrated results ---- fit results : Am Pixel per pixel calibration, E= Global calibration, E= Clustered Pixel per pixel calibration, E= Am241 E=13.9,26.3,59.5 keV

Conclusion A06-W0110 has been fully calibrated using LNLS Data + Am241 + Cd109 Analysis time has been reduced by a factor 1000! A few more points can be improved for the surrogate fit (better selection of points, Chi2 cuts etc...) Calibration results have limited effects compared to global calibration, however, calibration is still needed for comparison to simulation LNLS propose to perform XRF measurements on more assemblies for us if needed, UFRJ Rio student ready to go to Campinas to take measurements for us