1. Cesium calibration for ATLAS Tile Calorimeter
Alexander Solodkov
for IHEP+UTA Cesium team
TileCal DP Workshop, Sep-2013

2. 137Cs calibration principles
The powerfull 10 mCi 137Cs -source, embedded in a capsule, moves with a constant speed of ~30 cm/s inside the stainless steel tubes through all the calorimeter volume exiting all the scintillating tiles.
The system is composed of three independent parts having closed circuits with three separate sources.
The movable source system is the main tool to equalize the calorimeter cells responses, to transport the energy scale, and to monitor detector performance over time together with other calibration systems.
TileCal DP Workshop, Sep-2013

3. Data treatment
Results of Cesium scan: signal in every PMT measures with 90Hz frequency
Signal vs time plot provides an X-ray picture of the detector and can be visualized immediately
Useful tool during maintenance
Raw data are processed to obtain so-called “integrals” or “amplitudes”
3 different methods exists, which complement each other
An example of barrel module (LBA20) misalignment (top) in the large D-cell.
maintenance.
TileCal DP Workshop, Sep-2013

4. Calculation of Cs response: integral method
Mean period of the peak grid is calculated. Left/right boundaries of the cell are taken as the position of the first/last peak -/+ half of the period.
Integral within cell boundaries as well as integrals below left and right tails , are calculated.
If cell is in the middle of the calorimeter, both tails are good ones and Cs response is
If one of the tails has an abnormal shape, the integral under another tail with an appropriate correction is used
When average cell response is calculated, 22% leakage from one tile row to another tile row is taken into account
Stability of the method ~0.2% (in several runs one after the other)
TileCal DP Workshop, Sep-2013

5. Calculation of Cs response: amplitude method
Raw data
Amplitude method allows to calculate individual tile response
In this method response is fit by sum of Gaussian + exp. tails for every tile
22% leakage signal to the next tile row is subtracted before fit
Accuracy of single tile response is about 2%, average cell response is known with 0.3% precision
Short-term stability of both integral and amplitude methods is better than the stability of the system (PMT gain? HV?) over a few days
Run-to-run variations of responses at the testbeam when measurements were done after a week pause were about 0.5%
Estimated leakage signal
Signal after leakage subtraction
TileCal DP Workshop, Sep-2013

6. New approach to calculate integral for C10 developed by UTA in 2010
Cell C10
New approach to calculate integral for C10 developed by UTA in 2010
TileCal DP Workshop, Sep-2013

7. Calculation of integral in E1, E2 similar to calculations for C10
Developed by UTA in 2010
TileCal DP Workshop, Sep-2013

8. Comparison of the methods
Integral method
Stability – RMS of the subsequent measurements 0.2%
Doesn’t work for C10, E1,E2
Amplitude method
Stability 0.5%
Works well for C10 and special cells
Comparable with integral method for normal cells
UTA method
Stability ~1%
Designed to work for C10, E1, E2
Sometimes results are not reliable, because program doesn’t find position of the signal correctly
TileCal DP Workshop, Sep-2013

9. Calibration chain from raw data to constants in COOL
Integral program provides
ROOT file with integrals
ASCII file with diagnostics & summary => loaded to mySQL DB
Additional ROOT macro provides
ASCII file with preliminary integrals for E1, E2, C10
Shell script is used
to rescale E1.E2,C10
TUCS macros
Read integrals from ROOT & ASCII files
Read run quality flags from mySQL, HV values from raw data
Read integrator gains from COOL
Do quality checks (comparison with previous values in COOL)
Provide final sqlite file with constants
TileCal DP Workshop, Sep-2013

10. Experience with existing system
Calculation of Cesium calibration constants is quite fast – no CPU and human resources are required
The most difficult task is to validate the constants and to make sure that all they make sense
One trivial quality check – constants measured with a months interval can not change by more than XX%
Deviation level is adjusted depending on the time interval between two cesium runs
Order of 5-10 outliers are checked in every run
There is no automatic way to classify the problem
Variation of response due to HV variation
Data corruption or capsule is stuck or … problem in a run
There is no automatic way to compare behavior of similar PMTs (e.g. if 2 PMTs of a cell have similar problem)
TileCal DP Workshop, Sep-2013

11. Monitoring and public plots
ROOT files with calculated integrals are used to produce monitoring plots
Signal variation in every channel or in group of channels
A bit heavy procedure if every channel should be checked
Not integrated with TUCS
TileCal DP Workshop, Sep-2013

12. Ongoing work (offline)
2 main activity in Cesium calibration this year
Qualification task for Zoya Karpova (JINR, Dubna)
Integrate program for E1/E2/C10 cells into main chain
Validate this program and make it robust against possible problems in the data
Second half of qualification task for Andrey Kamenschikov (IHEP, Protvino)
Make quality check scripts in TUCS more modular and configurable
Provide automatically various validation plots, so it’ll not be needed to look at raw data to make decision
Provide summary plots for Tile-In-One
TileCal DP Workshop, Sep-2013

13. Ongoing work (online) Online software is being modified as well
Updates due to hardware modifications (new water stations)
More updates are in pipeline – to automatize some decisions during data taking
Confirm that system is ready to take data before the run (e.g. pedestals are within expected limits)
Detect abnormal situations and fix them (e.g. capsule is stuck)
Provide short report about data quality for every module during the run, so it’s possible to rescan problematic modules
TileCal DP Workshop, Sep-2013

14. Future work (demonstrator)
Cs system would like to send sensor data via slow control path from the drawers to RODs
This requires the usage of GBT interface in the drawer electronics
Cs system would like to reserve a port/connection to the drawer interface FPGA for Cs communication
The port could be CAN-bus and/or SPI
CAN-bus messages will be encapsulated inside GBT packets and unpacked by the RODs
Non-negligible amount of work is needed to update online and offline software for that
TileCal DP Workshop, Sep-2013

15. Conclusions
Cesium calibration system works reliably and is able to provide calibration constants for all the cells promptly
Over next year efforts of experts will be concentrated on
Integration of E1/E2/C10 analysis into main calibration chain
Automatic visualization of the results, with emphasize on problematic cases
Integration into Tile-in-One
Online software will be updated as well
Protvino team will be responsible for Cesium calibration constants in TileCal during RUN II
TileCal DP Workshop, Sep-2013

16. BACKUP
TileCal DP Workshop, Sep-2013

17. Source drive & control Cs hydro drive Cs garage
In order to transport the radioactive source in a safe and controllable way along the 10 km of tubes inside the calorimeter, an elaborate source drive and monitoring system are needed.
The hydraulic drive, which pumps the liquid to move the source is equipped with electronically operated pump and valves, and placed in the experimental cavern.
It is controlled via CAN interface.
Calibration tubes sequences in each parts are divided into a number of contours with corresponding number of supply tubes.
Contour system requires active control and monitoring of the source position to switch the valves according to the capsule movement.
Due to readout limitations, one has to switch from one module to another, this also requires the knowledge of source position
Between the scans the radioactive sources are stored inside the lead containers equipped with Geiger counter, capsule sensor and remotely controllable locks
Cs hydro drive
Cs garage
TileCal DP Workshop, Sep-2013

18. TileCs control software
The control software is a distributed system of parallel running processes in multiple machines/crates in USA15
Use of TDAQ libraries (CORBA, IS, etc.)
Python scripting
Multifunctional QT GUI
Audible alarms
History playback
Cs control architecture
TileCal DP Workshop, Sep-2013

19. Cs run
During Cs run the source is going from one garage to another one with the help of liquid flow
During the run the integrated currents from concerned modules PMTs are read out
Data are stored in a ROOT tree for further analysis
One run takes up to 1.5 hours (24 LB modules)
One Cs scan consists of at least 3 runs for LB and 6 runs for EB.
Cs GUI during the scan
TileCal DP Workshop, Sep-2013

20. TileCal DP Workshop, Sep-2013