sPHENIX tracking discussion

["Huang, Jin" <jhuang AT bnl.gov>]

Also available on the agenda page: https://indico.bnl.gov/conferenceDisplay.py?confId=1930

 

Summary on plan of tracking software development:

·       Pile up capability:

o   Action Item [Mike]: merge pile up support in event generation and digitization after cross check momentum resolution with Tony and clean up of vertex constraint

·       Improve detector simulation:

o   Action Item [Tony, Jin]: pull request to merge ladder geometry on MAPS. Geometry file -> calibration database

o   Action Item [Gaku]: Build INTT ladder in Geant4 based on the existing chain of PHG4SiliconTrackerSubsystem -> PHG4SiliconTrackerCellReco -> common reco code

o   Action Item [TPC team]: Assign updated TPC size and gas material in the setup macros. Update distortion map with simulation of event fluctuation.

·       Pattern recognition:

o   Carlos is testing OLUMPUS pattern recognition code

o   Action Item [Volunteer needed]: test performance with an extended vertex z-window in the current Hough transform pattern recognition, as an attempt to allow tracking in multi-vertex

o   Other pattern recognition packages under survey: STAR/ALICE Cellular Automata, STAR/Alice STV progressive finders

·       Track and vertex fitting:

o   Action Item [Haiwang, Jin]: pull request to replace track fitting with GenFit2.

o   Action Item [Haiwang, Tony]: pass the MAPS ladder normal direction to GetFit2

o   Action Item [Haiwang, Sanghoon]: verify quality of RAVE reconstructed event vertex by double checking the pull distribution. If good, pull request to replace silicon vertex fit with RAVE

o   Action Item [Sanghoon]: pull request to add new module processing a new vertex map to store secondary vertex found for tracks in given jet as used in B-jet tagging analysis

 

 

STAR tracking overview - Gene Van Buren

 

Many thanks to the STAR software team to give an nice overview of STAR track finding and fitting algorithms

 

Slide3: STI tracker

Dave: How the geometry is udpted in STI

Gene: Automatic generated Geant to STI import

Chris: what is the maintenance load of geometry imports?

Gene: few configuration changes per year

 

Gene: HFT self-tracking for claibration purpose only

Jin: was Vertex used as part of pattern recognition steps?

Gene: Vertex was not used

 

Dave: what was the source of field map

Gene: full field map used as input to tracking

 

Jin: how TPC distortion applied?

Gene: applied before patter reco. Pattern allow hit off pad row by a small amount.

 

Slide 6: Cellular Automata used in HLT for max penalization

Jin: can CA be used in silicon tracking too?

Gene: Not used by silicon, but not used in STAR yet

 

Slide 8: KNN algorithm

Victor Perevoztchikov: The plot is using R-Theta projection

Gene: software team has been careful on the treatment for the low-pT tracks

 

Summary:

Dave: what is the role of STI in tracking?

Gene: STI contains both seed finder and fitter

 

Dave: what is the portion of computing time in seed finding?

STAR team: 10% time in seed finding and 90% in fitting

 

Chris: Concern on reproducibility for cross platform when using local tracking algorithms.

Gene: small changes can be observed when migrating on compiler. Global property was checked to be consistent between different production environments.

 

Chris: what is the percentage of TPC clusters assigned to the tracks

Victor Perevoztchikov: about half

Dave: what is the feature of the clusters that was not used in tracks?

STAR team: For example delta electrons looping in active volume. Short tracks are not used in physics. Efforts was made to allow track finding to jump through known dead areas.

 

Chris: are merged clusters get separated in tracking?

Gene: Larger clusters are flagged and not used. Not using deconvolution of merged clusters.

 

Chris: what’s the evaluation of global tracking algorithm like hough transform VS local finder like STI?

STAR team: tried hough transform, not pursued.

 

STAR team: Useful to have framework that allow swapping tracking algorithms

 

 

 

 

Evaluating Olympus tracking - Carlos Perez

 

Carlos: Fit for complex magnetic field

 

Jin: Does this algorithm limit to primary tracks only?

Carlos: not for V0 decays.

 

Jin: Concern scaling of the dictionary size VS number of Layer of detector.

Chris: Concern on track density

Carlos: testing for both

 

 

Jason Webb: how does pre-generated dictionary handle TPC distortions?

Discuss: apply distortion before patter recognition, allow some migration of hits off pad row. 

Gene: STAR monitor via scaler rate. Map distortion to luminosity. For high rate distortion (1Hz), used TPC hit history to predict instant distortion.

 

 

 

Evaluating Olympus tracking - Carlos Perez

 

Tracking algorithm: Localized tracking algorithm.

 

Victor Perevoztchikov: how vertex found

Carlos: using first few layers of ITS. Try to suppress of vertex dependence of pattern recognition by dropping vertex constraint early on

 

Carlos: V0 track finding with rest of clusters after primary tracks was found

Tony: how to deal with gaps in track

Carlos: dead area was observed in data. Interesting to find out how they were handled.

 

Tony: speed for Kalman filtering based track finding

Carlos: couple months in ALICE data production of Run1.

 

Chris: why momentum resolution is worse in AuAu

Gene: suspect loss of hit in pattern recognition that reduced resolution in AuAu

 

Pile up correction - Mike McCumber

 

Page 4:

Jin: was BBC used to identify an in-time vertex?

Mike: Using tracking detector only in this study

 

Page 7:

Mike: the drift timing was implemented in TPC cell reco

 

Page 10:

Jin: For MAPS, is it plausible that hardware integrates the same 2us timing window for all ladders?

Mike: hardware readout should be synchronized

 

Page 13:

Chris: Input manager has repeat option that automatically looping input files

 

Page 14:

Tony: observed better resolution in TPC in central events in the tracking review tests

Mike: will confirm

 

 

Mike: condor fail at very large collision rate likely from index out of range

Mike: need tracking code with multi-vertex tracking code.

Mike: Alan’s code does allow lower layers Hough searching for further optimization.

 

 

Silicon ladder - Tony Frawley

 

Dave: is it possible optimize threshold for simulation only?

Mike/Tony: initial threshold set to remove hits near zero energy loss

 

Dave: suggest Geant material scan to quantify thickness of full ladder

Mike: part of the performance degradation for laddered silicon ladder when compared with simple cylinder tracker is that multiple scattering in realistic material distribution was not counted with the current Kalman filter

Haiwang: Will work with Tony to use GenFit2 for Kalman filter, which counts realistic geometry

 

Mike: Cautious HoughTransformTPC had assumption on hardcoded layers of silicon detector

Tony: it was now fixed.

 

 

TPC simulation - Sourav Tarafdar

 

Carlos: distortion maps under study taking into account of event fluctuations. Result distortion map to be use as input to G4 simulations  

 

Slide 5:

Klaus: the ANSYS + GARFIELD analysis is used to estimation charge cloud distortion beyond the current parametrization model used in G4 simulation.

 

Slide 6:

Dave: why was the inner radius of TPC adjusted to 24 cm?

Sourav/Carlos: The inner radius is still under study. Will update macro with final number

 

 

Generic Kalman filter GenFit2 - Haiwang Yu

 

Slide 6: error estimated from both Kalman filters is larger than reality, lead to a narrow pull distribution

Jin: how cluster error assigned for the MAPS tracker?

Mike: pixel size/Sqrt(12) for silicon detector

 

Slide 8: testing GenFit2 on MAPS ladders

Dave: which part of the software take care of error transformation

Haiwang: GenFit2 handle that. Need to get normal direction to input to GenFit2

 

 

 

Generic Vertex finder RAVE- Sanghoon Lim

 

Jin: Suggest use second vertex map container and new version of vertex

 

Tony: how to handle multi-collision in displaced vertex measurement

Jin: cut on secondary vertex not far along z-direction and displaced along the pT direction of the jet.

Mike: Use of INTT to reject off-time tracks/vertexes.

Tony: could be the demonstrating case for INTT.

 

Need simulation that support pile up.

 

 

 

 

 

 

______________________________

 

Jin HUANG

 

Brookhaven National Laboratory

Physics Department, Bldg 510 C

Upton, NY 11973-5000

 

Office: 631-344-5898

Cell:   757-604-9946

______________________________

 

From: Huang, Jin
Sent: Monday, October 3, 2016 12:27 PM
To: 'sphenix-software-l AT lists.bnl.gov' <sphenix-software-l AT lists.bnl.gov>; 'sphenix-tracking-l AT lists.bnl.gov' <sphenix-tracking-l AT lists.bnl.gov>
Subject: [Reminder] Tracking-focused sPHENIX simulation and software meeting, Tue Oct 4 1-4PM EST

 

Hello Everyone

 

This is a reminder for the tracking-focused sPHENIX simulation and software meeting, Tue Oct 4 1-4PM EST.

 

As in the announcement in the last week, the goal of the meeting is to coordinate tracking software development after the encouraging Sept-tracking detector review. The talks will include survey of tracking software developed by other collaborations, as well as status on our tracking software components. We will try to conclude with plan and mile stones for the next stage of software development.

 

The agenda page is at https://indico.bnl.gov/conferenceDisplay.py?confId=1930 , for which the common modification password is "1008"

•            For residents at BNL, we will meet at the 2-219 conference room

•            To join the Meeting via phone bridge:  https://bluejeans.com/619960505 

•            To join via Phone:

1) Dial:

        +1.408.740.7256

        +1.888.240.2560(US Toll Free)

        +1.408.317.9253(Alternate number)

        (see all numbers - http://bluejeans.com/numbers)

2) Enter Conference ID: 619960505

 

Cheers,

 

Jin

 

 

 

______________________________

 

Jin HUANG

 

Brookhaven National Laboratory

Physics Department, Bldg 510 C

Upton, NY 11973-5000

 

Office: 631-344-5898

Cell:   757-604-9946

______________________________

 

From: Huang, Jin
Sent: Wednesday, September 28, 2016 12:25 PM
To: 'sphenix-software-l AT lists.bnl.gov' <sphenix-software-l AT lists.bnl.gov>; sphenix-tracking-l AT lists.bnl.gov
Subject: Tracking-focused sPHENIX simulation and software meeting, Tue Oct 4 1-4PM EST

 

Dear tracking and software friends

 

As mentioned in Tony’s email, we are organizing a sPHENIX simulation and software meeting dedicated to tracking software development, Tue Oct 4 1-4PM EST. A draft agenda is here

https://indico.bnl.gov/conferenceDisplay.py?confId=1930

 

The goal of the meeting is to coordinate tracking software development after the encouraging Sept-tracking detector review. This meeting will begin with overview talks on STAR, ALICE and Olympus tracking software to help survey existing tracking packages. Then we plan to have summary on status and plan for the components of our tracking software: simulation, digitization, Kalman filter and vertex finder. We have made many progress on new capabilities in these components even before the tracking review. Now, it is time for us to coordinate and integrate them into the default simulation and reconstruction.

 

Hope to see many of you at the meeting. Suggestion welcomed too.

 

Jin

Also for Tony, Mike, Chris

 

______________________________

 

Jin HUANG

 

Brookhaven National Laboratory

Physics Department, Bldg 510 C

Upton, NY 11973-5000

 

Office: 631-344-5898

Cell:   757-604-9946

______________________________