Proton Calorimetry/Meetings/2021/07/02: Difference between revisions

From PBTWiki
Jump to navigation Jump to search
(Created page with "== Minutes for UCL Proton Calorimetry Meetings, 23rd June == === Present === '''Simon Jolly''', '''Saad Shaikh''', '''Fern Pannell''' === Saad Shaikh === ...")
 
 
(7 intermediate revisions by the same user not shown)
Line 1: Line 1:
== Minutes for UCL Proton Calorimetry Meetings, 23rd June ==
== Minutes for UCL Proton Calorimetry Meetings, 2nd July ==


=== Present ===
=== Present ===
Line 6: Line 6:


=== [[ELogs/SaadShaikh|Saad Shaikh]] ===
=== [[ELogs/SaadShaikh|Saad Shaikh]] ===
*Painted 2 pairs of scintillator sheets with Derek.
*Painted 4 scintillator sheet pairs:
**Will collect and measure painted pairs tomorrow.
**Sheets 48 & 23 with primer – one layer to glue pair.
*Optimised ROOT fitting speed: using Legendre-Gauss integration algorithm and Fumili minimisation.
**Sheets 54 & 1 with appliance – one layer to glue pair.
**32 data points: 12Hz with ROOT display and 35-40Hz without. Fitting smooth QB function.
**Sheets 52 & 11 with appliance and then primer to glue pair – 10-15 mins for appliance layer to dry.
**Will consolidate new version of fitting into rest of fitting scripts.  
**Sheets 53 & 2 with appliance and then primer to glue pair – 30-40 mins for appliance layer to dry.
**Will test differences in fit results using Legendre-Gauss/Fumili compared to default GSL/Minuit.
*Neither combination gives acceptable results: lots of gaps in paint layer due to either inconsistencies in sheet thickness, placement misalignment or paint reaction.
**Will add regular Bortfeld curve to live fit and output file for GUI.
**Placement of sheets when glueing is difficult: not much room to adjust once placed as this ruins paint layer.
**Painted pairs will be stripped of paint again.
*Will paint 32 sheets for beam test with single coat of primer to be ready for beam tests.
*New fitting routine using Gauss–Legendre algorithm and Fumili minimiser implemented into photodiode scripts.
**New routine gives reasonable results where default options either fail or give strange fit results.
**With a given set of fit results, Gauss–Legendre draws Bortfeld curve slightly differently compared to default Adaptive Singular algorithm.  
**Might require some optimisation of numerical parameters to correct this. Will investigate, but might be that "slow, accurate, non-live" fit uses slower Adaptive Singular algorithm to draw curves.
*Organised new inventory and beam test equipment into new DeWalt tower – tower kept in D109.
*Installed Argon One V2 case on Raspberry Pi.
*FTDI script uses 30-35% of CPU, probably maxing out 1 processor core. Performance issues likely due to lack of power.
*Adding "-O3" optimisation flag when compiling FTDI script gives significant performance improvements on both Pi and Mac.
**With 4 modules, Mac now gives average time per read of 172us rather than 188us, with an integration time of 170us.
**With 2 modules, Pi down to 200us from 750us. Still observe some freezes during acquisitions, but far fewer than before.
**Also observe improvement in frame rate of replay script with this flag. Fitting and plotting scripts are largely the same.
*Will install ROOT on Pi and compile fitting and plotting scripts.


=== [[ELogs/FernPannell|Fern Pannell]] ===
=== [[ELogs/FernPannell|Fern Pannell]] ===
*
*

Latest revision as of 13:18, 26 July 2021

Minutes for UCL Proton Calorimetry Meetings, 2nd July

Present

Simon Jolly, Saad Shaikh, Fern Pannell

Saad Shaikh

  • Painted 4 scintillator sheet pairs:
    • Sheets 48 & 23 with primer – one layer to glue pair.
    • Sheets 54 & 1 with appliance – one layer to glue pair.
    • Sheets 52 & 11 with appliance and then primer to glue pair – 10-15 mins for appliance layer to dry.
    • Sheets 53 & 2 with appliance and then primer to glue pair – 30-40 mins for appliance layer to dry.
  • Neither combination gives acceptable results: lots of gaps in paint layer due to either inconsistencies in sheet thickness, placement misalignment or paint reaction.
    • Placement of sheets when glueing is difficult: not much room to adjust once placed as this ruins paint layer.
    • Painted pairs will be stripped of paint again.
  • Will paint 32 sheets for beam test with single coat of primer to be ready for beam tests.
  • New fitting routine using Gauss–Legendre algorithm and Fumili minimiser implemented into photodiode scripts.
    • New routine gives reasonable results where default options either fail or give strange fit results.
    • With a given set of fit results, Gauss–Legendre draws Bortfeld curve slightly differently compared to default Adaptive Singular algorithm.
    • Might require some optimisation of numerical parameters to correct this. Will investigate, but might be that "slow, accurate, non-live" fit uses slower Adaptive Singular algorithm to draw curves.
  • Organised new inventory and beam test equipment into new DeWalt tower – tower kept in D109.
  • Installed Argon One V2 case on Raspberry Pi.
  • FTDI script uses 30-35% of CPU, probably maxing out 1 processor core. Performance issues likely due to lack of power.
  • Adding "-O3" optimisation flag when compiling FTDI script gives significant performance improvements on both Pi and Mac.
    • With 4 modules, Mac now gives average time per read of 172us rather than 188us, with an integration time of 170us.
    • With 2 modules, Pi down to 200us from 750us. Still observe some freezes during acquisitions, but far fewer than before.
    • Also observe improvement in frame rate of replay script with this flag. Fitting and plotting scripts are largely the same.
  • Will install ROOT on Pi and compile fitting and plotting scripts.

Fern Pannell

Retrieved from "https://www.hep.ucl.ac.uk/pbt/pbtWiki/index.php?title=Proton_Calorimetry/Meetings/2021/07/02&oldid=4152"