Simulation Tasks Understanding Tracking Understanding Hardware 1 Two

Simulation Tasks Understanding Tracking Understanding Hardware 1. Two types of tasks: a. Implementing known functions in ATLAS framework b. Understanding their limits new algorithms understanding hardware 2. We need a new training: a) Too many pairs of close-hits software problem? good to go to V 12 b) Too many pairs of near-hits 2 nd pixel layer will help. Now is missing c) Too many pairs of near-hits clustering/space point problem? Stop space point use? Dump raw hits, now they are missing 3. The more tracking users we have the more we understand about detector/FTK. Meena work is a good example, very useful work, weak points found. Area (b) could be shared by CMS-ATLAS.

FTKsim/FTK STATUS summary 1. FTKsim efficiency/resolution is similar to Ipatrec efficiency/resolution BUT… 2. Long story about reducing RW-Ghosts (~Done in Pisa): a) RW-Ghosts due to overlap regions solved with new digitization of training samples in Pisa. Detector inefficiency and noise turned off as much as possible (to our knowledge). b) Selective RW balanced by a track fitter that performs all the 5/6 fits if the 6/6 fit doesn’t pass the c 2 cut. 3. Long story about reducing duplicated tracks, generic ghosts (in progress in Pisa): a) Many tracks too near (used z, phi, eta, Pt to judge the distance - not IP) b) How to replace the Ghost Buster function inside SVT (duplicate clean up)? c) Hit Warrior under test least c 2 if many tracks share 10 -8 meas. over 12 d) HW works inside a road, should work on roads that differ for 1 SS ?

WHAT COULD BE DONE IN PARALLEL-1 FTKsim V 12 GO TO V 12 – New Ftk. Sim. Wrap – New training (yellow box) Pattgen: creates pattern bank Corrgen: generates fit costants Ftksim – simulation training Francesco Crescioli 3

WHAT COULD BE DONE IN PARALLEL-2 FTKsim V 12 (a) has the highest priority!! a) Update Ftk. Sim. Wrap to produce from Athena V 12 FTKsim input file format 1. Activate the 2 nd pixel layer and dump all the layers 2. Dump raw hits (R-PHI Layers and stereo Layers) in addition to space points 3. Ipatrec Dump: is it possible to run Ipat. Rec without TRT information? b) Study again FTK performances on fakes using raw hits: Ø Not use any clustering and use only clean up after TF. Processing time? Track resolution? Ø Develop & include in FTKsim an FTK clustering algorithm thinking to the hardware implementation (DATA-FORMATTER in the next slide)? Ø Which is the pattern bank (road) size ? Inside the proposal: Which layers (& how many)? We know it is better to reject the central ones. Which is the impact on fakes and processing time?

Inside Fast-Track ~75 9 U VME boards – 4 types Pixels & SCT 50~100 KHz event rate S-links NEW Data Formatter (DF) cluster finding split by layer EVENT # N HITS DO-board DATA ORGANIZER SUPER BINS ROADS Giga. Fitter ROADS + HITS ex. GB: Hit Warrior? Raw data ROBs EVENT # 1 AM-board RODs PIPELINED AM + RW overlap regions Track data ROB 2 nd step: track fitting ~Offline quality Track parameters

WHAT COULD BE DONE IN PARALLEL-3 use FTKsim We planned to study the signal of each physics case on the full simulation. This means to see FTKSIM performances for: (a) Bs mm (done for V 10 -Francesco-Guido – redo it for V 12) (b) uu, bb jets from SM WH (done for V 10 -Meena-Guido – redo it for V 12) (c) taus ? (d) energetic u-b-jets? Supersymmetric Higgs at high masses? (e) other samples? (f) Different samples could require different FTK features ! (g) Taus or energetic b-jets can be more demanding than WH events ! (h)Produce new Physics samples and simulate FTK !

CONCLUSIONs: work in parallel with Pisa that continues to optimize algorithms 1. HIGH PRIORITY: Link FTKsim with Athena V 12 2. Study FTKsim performances on different samples and give back informations to Pisa. 3. Developing FTK private clustering? 4. Study FTKsim as a function of possible options (which/how many layers…, clustering y/n…, which algorithms…). Give back informations to Pisa.

Ghost handling Plan Problems now: 1. Deletion of 5/6 by RW when a 6/6 is available produces small inefficiencies if the 5/6 was the real track reject all 5/6 and do 7 fits in parallel (6/6 and all 5/6 combination) to choose the best c 2 2. Deletion of 5/6 by RW not possible if the empty strips belong to different sectors 3. Training tracks passing through the overlap region producing many Ghost patterns not identified by the RW. Reduce the ghosts of type 2 due to overlap regions, generating a single pattern,