CIS Tech Update

1

Chicago Weekly Meeting, 3 January 2022

Dawit Belayneh and Katie Hughes

Outline

2

• TileCal Electronics
• Charge Injection System (CIS)
• CIS and recalibration overview
• Summary of Sept., Oct., Nov. CIS updates
• Cooling Valve Installations
• Motivation
• Installation Procedure
• Status
• LAr Simulation Work
• Pictures

​

TileCal Electronics

ATLAS Tile Calorimeter

4

• 4 Partitions
• Long Barrel: LBA, LBC
• Extended Barrel: EBA, EBC
• 64 modules per partition
• Scintillating tiles in the modules emit photons when they are excited
• Fibers deliver light to PMTs

Detector Status

5

Charge Injection System (CIS)

Overview

7

• Calibration system for front-end readout electronics
• Simulates physics signals by injecting a known charge and measuring the electronic response
• Characterizes each readout channel with a CIS constant

Charge Injection

8

• 3-in-1 cards contain 5 pF and 100 pF capacitors
• Vary charge of injected pulse
• 64x gain amplifiers shape and amplify pulse
• Pass pulse to analog-to-digital converter (ADC)

Calibration Constants

9

• 60 injections of a set charge sampled 7 times
• 15 steps varying phase between injection and sampling
• 4 capacitor discharges each
• Mean amplitude of pulse measured for different charges
• Amplitude (ADC counts) vs charge (pC) fit gives CIS constant
• For each of ~19,700 channels!

CIS Constant Recalibration

10

• CIS runs taken twice per week during LS2
• Goal: verify maintenance and identify DQ (data quality) problems
• TileCal Unified Calibration Software (TUCS) used to track detector performance and diagnose problematic channels
• Calculate new constants within given interval of validity
• Exclude outliers and stable outliers, identify DQ flags
• Update databases approximately monthly

Procedure

11

• Run a general TUCS script that generates plots of CIS constants vs time
• Time range: last CIS update to present
• Go through plots by hand to check validity of constants
• Good CIS plots are very stable over time and close to the detector mean
• Possible changes that need to be made manually:
• Recalibration (run over a shorter time range)
• Flag update (add BadCIS / remove BadCIS)
• Exclude problematic runs for certain partition+gain combinations (bad timing, bad amplitude/charge ratio)
• Add manual changes and upload constants to database

Recalibration

12

• Often there is a constant shift in the CIS constant at some date, or 1-2 outlier points that skew the mean
• Solution: run over a shorter time period, so long as the newer value appears stable

Flag Update

13

• Sometimes channels are incorrectly flagged as BadCIS, or vice versa

Timing and Amplitude/Charge Ratio

14

• Amplitude/charge ratio for this CIS run in LBC highgain is unstable, so we should not use it.

• Timing should be consistent across all valid runs, typically in the range of (-15, 10) ns. Here, it’s shifted by +25 ns.

CIS Update Summaries

15

September

• CIS runs from Aug 23 - Sept 24
• 153 channels updated
• 6 with >5% change

October

• CIS runs from Oct 12 - Oct 25
• 159 channels updated
• 20 with >5% change

November

• CIS runs from Oct 28 - Nov 27
• 505 channels updated
• 15 with >5% change

CIS Constant Distributions (Low Gain)

16

Should be Gaussian and centered around 1.29 ADC counts/pC

CIS Constant Distributions (High Gain)

17

Should be Gaussian and centered around 81 ADC counts/pC

CIS Constant Stability

18

TUCS Quality Flags

19

We used advanced run selection for Sept. CIS update

TUCS Quality Flags

20

CIS Constant RMS/Mean

21

Detector Stability

22

Runs in early October were not used to calculate CIS constants

Next Steps

23

• Starting CIS update for December (?)
• Continue figuring out ways to streamline the update process
• Look into how the TUCS script assigns/removes bad CIS
• Create a new TUCS script / wrapper that allows us to specify different valid runs for different modules in one command
• Can sort of accomplish this with a bash script, but we think there is a more concise way to do it
• Would allow us to produce accurate cumulative plot

Cooling Valve Installations

Motivation

25

• Currently, a single cooling loop passes through 8-12 drawers
• 6 loops per partition
• A leak in one drawer means all other drawers in the loop must be shut down to fix it
• Valves will now isolate leaks to a single drawer

Cooling Loop Distribution

Procedure

26

1. Build and test valves
2. Mount brackets with valves on drawers
3. Stop cooling loops, connect hoses to valves, restart loops and check for leaks
4. Connect air hoses and verify they actually shut off the valves (commissioning)

Building and Testing Valves

27

• Add non-magnetic and radiation-resistant parts
• Attach nozzles with adhesive, let valves sit for >24 hours so that seal has time to dry
• Test seal quality under pressure. If valves pass, they are ready to be installed
• Since the last report we assembled over 120 with 0 failures

Mounting Valves

28

• Have a variety of different shapes of mounting plates
• Orientation is critical!
• Hoses must be able to reach valve nozzles without bending too much
• Valve nozzles must not be too close to walls/other wires
• Ideally plates should be secured with two bolts, but sometimes that is not possible

Initial Hose Connection

29

Pipes

Method

30

Disconnect hoses here -- don’t touch the hoses inside the drawer. Cut new hoses and fit their end to the in/out pipes.

Method

31

This is a much quicker method as we don’t have to open the drawers! However, sometimes it can be awkward to reach the in/out pipes as often they are far away from the scaffolding.

Examples

32

Examples (more complicated)

33

Middle Beam

Cs source box

Cs electronics box

Air Hoses Preparation

34

• We measured out all the hoses to be installed on EBC (as well as a decent number of those on EBA)
• They were all labeled and bundled

Connecting air hoses

35

• Hoses were run around the outside of the detector to the patch panel by another team
• We went around and plugged each one into the appropriate valve on EBA and EBC (LB already done)

Commissioning

36

• Hoses are gathered near a patch panel where they can be connected to a pressurized system
• We check the in and out hoses for each module separately, and confirm they both increase the module’s temperature using DCS monitoring
• Did on both EB and LB

​

Final Touches

37

• All measurements given for the hoses on EBA and EBC seemed to be about 1-2 m too long once run around the detector, so we trimmed and relabeled them all.

Identifying Leaks

38

• Check PPV cycle
• How often cooling system must be pumped down to a lower pressure
• Now we can use the air hoses to shut off valves we suspect have leaks, then check PPV cycle to see if it improves

Current Status

39

• All 4 partitions have valves on every module! They are all connected to air hoses and have been commissioned.
• There are about 20 extra valves if any need to be replaced in the future.

LAr Simulation Work

Energy Correction due to Accordion Structure

41

Goal: To obtain a statistical model of energy modulation as a function of phi arising from accordion structure of LAr Calorimeter.

• Final product: A functional form (TGraph) for the energy modulation in each EM layer of LAr Calo
• Strategy: Use muon events generated at the center of detector to sample layer-energy dependency on phi
• Fit these sample points to obtain a TGraph for each EM layer of LAr Calo

We will use these models to implement energy correction in FastCaloSim code.

Accordion Structure & FastCaloSim

42

• FastCaloSim is a framework that provides fast simulation of particle energy response of ATLAS detector.

​

• Key simplification for fast simulation in LAr EM layers: idealize accordion wiggles by prismatic cells in 3D space

​

• Run physics on idealized detector geometry

​

• Correct for accordion effect on energy deposition at the end

​

• To implement correction, we need to understand the phi dependence of energy deposition arising from the accordion structure.

Input Statistics

​

43

Muons (|Pdg = 13|) generated at the center of detector with

• CME of 100 Gev
• 0 < phi < 2𝝅 & |eta| < 3
• Eta range chosen to cover all of EM layers

Advantages of Muons: Heavier mass means more localized energy depositions (compared to electrons or photons). This gives us better phi precision during our scan.

Disadvantage of Muons: Heavier mass means muons deposit on average only 3-4 Gev with minimal ionization. This, in-turn, increases our energy uncertainty, requiring larger size of input data.

• Muon event selection: We are only interested in Muon events that mainly deposit energy through MIPs. Other forms of deposition, for example, photon-radiation contribute to uncertainty in energy deposited at a particular phi value.
• Therefore, we only keep muons with high enough final energies (E > 92 Gev).

Final Truth-E of Muons

44

Region of large non-MIP energy deposition

Selected events

E (Mev)

Computing Energy Correction

​

45

The sample points (Ideal_phi, layer_energy) to be fitted are produced from our input data as follows:

for each EM layer in LAr Calo:

• obtain total energy deposited by muon in layer (layer_energy)
• obtain phi offset between full geometry and FastCaloSim (ideal) geometry
• obtain muon truth eta/phi as: phi = newTTC_mid_phi[0][layer]
• get impactCell of muon in layer (code courtesy of Michele)
• compute phi offset as: phi_correction = impactCell_phi - impactCell_phi_raw
• translate phi value from full geometry to ideal geometry accounting for periodicity
• ideal_phi = (|phi - phi_correction|) mod (𝜋/512.)

Result

​

46

• Result for Layer 2 of LAr EM layers. Most important for this work as it is the layer with the most phi partitions.

​

• Result shows, as expected, phi periodicity of energy deposition.

​

• The final-energy cut, eliminating non-MIP energy deposition, significantly improved our energy uncertainty

​

• Currently at about 2-3% for initial data size of 150k events.

Remaining Work

​

47

• Submit data request for large input data
• Still need to understand what the LAr-Sim team considers to be an acceptable error range which determines the size of the input data required
• If input data size is too big, my code will be quite slow.
• Might need to re-organize code to make it suitable for condor submission for fast execution

​

• Once we have enough statistics, fit a curve to our points to obtain the functional dependency: E(phi)

​

• Implement the energy correction in the FastCaloSim code to account for accordion structure

Pictures:)

The CIS techs go to Barcelona!

When we had to evacuate the detector :)

Thank You!