Choose timezone
Your profile timezone:
CIDeR-ML General Meeting
→
America/Los_Angeles
Description
https://u-tokyo-ac-jp.zoom.us/j/83932834349
Recording
Minutes:
Quick recap
The team reviewed Ryotaro's work on LED data and PMT calibration, including his plans to prepare calibration data and test SIREN calibration frameworks. They discussed LED system and PMT data analysis, with the team agreeing to establish a centralized data format and improve Monte Carlo simulations for cosmic event reconstruction. Ka Ming and Patrick explored implementing angular-dependent efficiency in MDT, considering the need for efficient sampling along emission lines and mirroring the traditional pipeline's per-photon information process for SIREN models.
Next steps
- Ryotaro: Look up the coordinates of the LED or ask someone about the LED position ID
- Ryotaro: Look for Jakob's prepared LED data after various trigger types
- Ryotaro: Estimate the variety of the number of photons received by each PMT when a signal LED fires using methods such as simple integration
-
Ryotaro: Learn how to use the geometry package and extract coordinates from it
-
Ryotaro: Try the WCSim code that Ka Ming provided to simulate LED events
- Ryotaro: Start WCSim simulations for LED data using the container version 1.4.2
- Ryotaro: Use all LED data instead of limiting to self-trigger for calibration analysis
- Kieren: Have a discussion with Ka Ming about getting started with tuning Monte Carlo for cosmics and PMT charge overflow issues
- Kieren and Nick: Figure out how to generate more events around and outside detector boundaries
- Ka Ming: Implement angular-dependent efficiency changes in the MDT package based on extracted efficiency ratios from cosmic analysis
- All: Think about how to extend SIREN pipeline to include PMT response
Summary
LED Data and PMT Calibration
Ryotaro presented a status report on his work with LED data and PMT calibration. He explained his goal to adjust calibration data and test SIREN calibration frameworks, noting that timing calibration using LED data is currently the main use. Ryotaro outlined his next steps, which include 1PE gain calibration in standard analyses and converting data to a format usable by the SIREN framework. He also discussed the LED setup, including the different types of PMTs and their corresponding LED configurations. Ryotaro mentioned his focus on analyzing QDC distributions for specific PMTs during LED flashes and his plan to investigate LED position IDs and coordinates.
LED System and Trigger Analysis
The team discussed LED system and PMT data analysis, with Ka Ming providing guidance on using the geometry package to extract accurate coordinates and suggesting Ryotaro try the simulation codes in WCSim. Patrick inquired about trigger systems, and Ka Ming explained the differences between self-trigger, software trigger, and hardware trigger, recommending the use of all LED data for calibration purposes. The team agreed that a centralized data format should be established, with Jacob potentially serving as the expert for triggering and readout data format.
Cosmic Event Simulation Improvements
The team discussed Monte Carlo simulations and cosmic event reconstruction, focusing on improving accuracy by expanding the generation volume to include events near detector boundaries. Kieren and Nick are working on implementing a new tool that removes events starting in PMTs and directed outward, with plans to update simulations to exclude events in PMT areas. Ka Ming presented results from cosmic muon reconstruction studies, showing resolution peaks at 10-15 centimeters and efficiency variations across different PMTs, with a target accuracy of 5% for SIREN calibration. The team agreed to continue refining the simulations and implementing efficiency changes in the MDT package for reprocessing data.
Angular Efficiency in WCSim Models
Ka Ming and Patrick discussed implementing angular-dependent efficiency in MDT, noting that the current response function is flat. They explored the possibility of tuning SIREN models through forward propagation and batching, considering the need for efficient sampling along emission lines. Patrick suggested mirroring the traditional pipeline's per-photon information process for SIREN models, which Ka Ming acknowledged would require post-processing to extract angular response curves. They agreed to revisit discussions on chaining SIREN together with/without intermediate sampling.
