Jan. 2009Jinyuan Wu & Tiehui Liu, Visualization of FTK & Tiny Triplet Finder Jinyuan Wu and Tiehui Liu Fermilab January 2010.

Slides:



Advertisements
Similar presentations
Impact parameter studies with early data from ATLAS
Advertisements

Track Trigger Designs for Phase II Ulrich Heintz (Brown University) for U.H., M. Narain (Brown U) M. Johnson, R. Lipton (Fermilab) E. Hazen, S.X. Wu, (Boston.
Computing EVO meeting, January 15 th 2013 Status of the Tracking Code Gianluigi Boca, Pavia University.
On-Chip Processing for the Wave Union TDC Implemented in FPGA
Simulation Studies of a (DEPFET) Vertex Detector for SuperBelle Ariane Frey, Max-Planck-Institut für Physik München Contents: Software framework Simulation.
Some Thoughts on L1 Pixel Trigger Wu, Jinyuan Fermilab April 2006.
June 6 th, 2011 N. Cartiglia 1 “Measurement of the pp inelastic cross section using pile-up events with the CMS detector” How to use pile-up.
Track quality - impact on hardware of different strategies Paola FTK meeting Performances on WH and Bs   2.Now we use all the layers.
Cluster Threshold Optimization from TIF data David Stuart, UC Santa Barbara July 26, 2007.
Silicon Tracking for Forward Electron Identification at CDF David Stuart, UC Santa Barbara Oct 30, 2002 David Stuart, UC Santa Barbara Oct 30, 2002.
Simulation Tasks  Understanding Tracking  Understanding Hardware 1.Two types of tasks: a.Implementing known functions in ATLAS framework b.Understanding.
FTK poster F. Crescioli Alberto Annovi
The LiC Detector Toy M. Valentan, M. Regler, R. Frühwirth Austrian Academy of Sciences Institute of High Energy Physics, Vienna InputSimulation ReconstructionOutput.
Without hash sorting, all O(n 2 ) combinations must be checked. Hash Sorter - Firmware Implementation and an Application for the Fermilab BTeV Level 1.
Advanced Topics on FPGA Applications Screen B Wu, Jinyuan Fermilab IEEE NSS 2007 Refresher Course Supplemental Materials Oct, 2007.
Proposal by the Numbers 3 Trigger layers (plus 2 “short” layers) provide full coverage to eta=??? in 15 degree sectors Hits collected in real time, sent.
1 Tracking Reconstruction Norman A. Graf SLAC July 19, 2006.
CS 376b Introduction to Computer Vision 02 / 22 / 2008 Instructor: Michael Eckmann.
Track Reconstruction: the trf & ftf toolkits Norman Graf (SLAC) ILD Software Meeting, DESY July 6, 2010.
Resource Awareness FPGA Design Practices for Reconfigurable Computing: Principles and Examples Wu, Jinyuan Fermilab, PPD/EED April 2007.
TDC for SeaQuest Wu, Jinyuan Fermilab Jan Jan. 2011, Wu Jinyuan, Fermilab TDC for SeaQuest 2 Introduction on FPGA TDC There are.
Non-prompt Track Reconstruction with Calorimeter Assisted Tracking Dmitry Onoprienko, Eckhard von Toerne Kansas State University, Bonn University Linear.
Optimising Cuts for HLT George Talbot Supervisor: Stewart Martin-Haugh.
14/6/2000 End Cap Muon Trigger Review at CERN 1 Muon Track Trigger Processing on Slave and Hi-pT Boards C.Fukunaga/Tokyo Metropolitan University TGC electronics.
The CDF Online Silicon Vertex Tracker I. Fiori INFN & University of Padova 7th International Conference on Advanced Technologies and Particle Physics Villa.
A Pattern Recognition Scheme for Large Curvature Circular Tracks and Its FPGA Implementation Example Using Hash Sorter Jinyuan Wu and Z. Shi Fermi National.
Vienna Fast Simulation LDT Munich, Germany, 17 March 2008 M. Regler, M. Valentan Demonstration and optimization studies by the Vienna Fast Simulation Tool.
AMB HW LOW LEVEL SIMULATION VS HW OUTPUT G. Volpi, INFN Pisa.
May Wu Jinyuan, Fermilab 1 FPGA and Reconfigurable Computing Wu, Jinyuan Fermilab ICT May, 2009.
1 Th.Naumann, DESY Zeuthen, HERMES Tracking meeting, Tracking with the Silicon Detectors Th.Naumann H1 DESY Zeuthen A short collection of experiences.
G. Volpi - INFN Frascati ANIMMA Search for rare SM or predicted BSM processes push the colliders intensity to new frontiers Rare processes are overwhelmed.
1 Endcap C SCT Efficiency Calculation Update Nicholas Austin University of Liverpool Operations Meeting June 2007.
Advanced Topics on FPGA Applications Screen A Wu, Jinyuan Fermilab IEEE NSS 2007 Refresher Course Supplemental Materials Oct, 2007.
Triplet Finder Jinyuan Wu Dec Triplets Triplet Finding (1) Filling Bit Arrays Note: Flipped Bit Order Physical Planes Bit Arrays For any hit…
Fast Tracking of Strip and MAPS Detectors Joachim Gläß Computer Engineering, University of Mannheim Target application is trigger  1. do it fast  2.
Track Reconstruction: the trf toolkit Norman Graf (SLAC) ILC-ACFA Meeting, Beijing February 6, 2007.
ATLAS Trigger Development
Oct. 16, 1998Hobbs (thanks Hal)1 SMT Road Widths & Extra CFT Layers Current Situation Adding a 3rd CFT layer Using all 8 CFT layers Summary.
1 FTK AUX Design Review Functionality & Specifications M. Shochet November 11, 2014AUX design review.
The Detector Performance Study for the Barrel Section of the ATLAS Semiconductor Tracker (SCT) with Cosmic Rays Yoshikazu Nagai (Univ. of Tsukuba) For.
CMS Upgrade Workshop - Fermilab Trigger Studies Using Stacked Pixel Layers Mark Pesaresi
Tiny Triplet Finder Jinyuan Wu, Z. Shi Dec
27 March 2003RD Schaffer & C. Arnault CHEP031 Use of a Generic Identification Scheme Connecting Events and Detector Description in Atlas  Authors: C.
The SLHC CMS L1 Pixel Trigger & Detector Layout Wu, Jinyuan Fermilab April 2006.
Layer00 Efficiency Studies Stephen Levy, UChicago.
9th October 2003Danny Hindson, Oxford University1 Inner Detector Silicon Alignment Simple approach -Align in stages + rely on iteration Barrel to Barrel.
RD program on hybrids & Interconnects Background & motivation At sLHC the luminosity will increase by a factor 10 The physics requirement on the tracker.
SiD Tracking in the LOI and Future Plans Richard Partridge SLAC ALCPG 2009.
November 15, 2005M Jones1 Track Lists in Level 2 Outputs from SLAM provide lists of tracks faster than the current path to SVT and Level 2 via XTRP Sparsify.
S. Dasu, University of Wisconsin February Calorimeter Trigger for Super LHC Electrons, Photons,  -jets, Jets, Missing E T Current Algorithms.
BEACH 04J. Piedra1 SiSA Tracking Silicon stand alone (SiSA) tracking optimization SiSA validation Matthew Herndon University of Wisconsin Joint Physics.
Track Reconstruction: the ftf and trf toolkits Norman Graf (SLAC) Common Software Working Meeting CERN, January 31, 2013.
Off-Detector Processing for Phase II Track Trigger Ulrich Heintz (Brown University) for U.H., M. Narain (Brown U) M. Johnson, R. Lipton (Fermilab) E. Hazen,
Status of FTK Paola Giannetti INFN Pisa for the FTK Group ATLAS Italia November 17, 2009.
UPDATE ON HARDWARE 1 1.VERTICAL SLICE & COOLING TESTS 1.ONLY a TEST STAND or a SMALL DEMONSTRATOR ?? 2.CRATE.
Ftksim at high luminosity Monthly meeting September 22, 2008 Anton Kapliy.
Wu, Jinyuan Fermilab May. 2014
Using IP Chi-Square Probability
2018/6/15 The Fast Tracker Real Time Processor and Its Impact on the Muon Isolation, Tau & b-Jet Online Selections at ATLAS Francesco Crescioli1 1University.
Status of the Offline Pattern Recognition Code GSI, December 10th 2012
Overview of the ATLAS Fast Tracker (FTK) (daughter of the very successful CDF SVT) July 24, 2008 M. Shochet.
Update of pattern recognition
Some basic ideas (not a solution)
eXtremely Fast Tracker; An Overview
FTK variable resolution pattern banks
Status of the Offline Pattern Recognition Code Goa, March 12th 2013
The LHC collider in Geneva
road – Hough transform- c2
The LHCb VErtex LOcator
SCT Wafer Distortions (Bowing)
Presentation transcript:

Jan. 2009Jinyuan Wu & Tiehui Liu, Visualization of FTK & Tiny Triplet Finder Jinyuan Wu and Tiehui Liu Fermilab January 2010

Jan. 2009Jinyuan Wu & Tiehui Liu, Assumption Pixel and SCT barrels are considered. Total number of channels (From Fig. 2.1 of TP): –B0: 20K –B1: 16K –B2: 14K –SCT3, 4, 5, 6: 5K each. Total number of clusters (1/1.3 of above): –B0: 15K –B1: 12K –B2: 11K –SCT3, 4, 5, 6: 3.8K each.

Jan. 2009Jinyuan Wu & Tiehui Liu, The R-phi View of Pixels Pixel modules are tilted. For hits on B0 found in rows along z-axis, several “R-phi Corridors” can be defined. In each r-phi corridor, distances between pixel planes are relatively well defined.

Jan. 2009Jinyuan Wu & Tiehui Liu, The R-Phi Corridors Consider 1 mm impact parameter, the possible range of tracks passing through a 1 mm r-phi bin in B0 spread in large range. For Pt = +- 1 GeV, the range in B2 is about 14 mm. For a r-phi corridor passing a bin each in B0 and B1, 1 mm each, the range on B2 is about 4mm. More than 30 hit patterns for each bin in B0 can be counted above. There are about 30 patterns/bin x 300 bins ~ 9K patterns in r-phi view.

Jan. 2009Jinyuan Wu & Tiehui Liu, The Thin Road Bins The size of the bins in the Thin Road scheme is (1 mm x 60 mm) in (r-phi & z).

Jan. 2009Jinyuan Wu & Tiehui Liu, Random Hits in the Thin Road Bins 3.7 hits/bin 1.4 hits/bin 0.9 hits/bin Nearly all bins are on. With or without pixel will not help given this bin size.

Jan. 2009Jinyuan Wu & Tiehui Liu, Binning in Z Direction Let’s dream we can further divide in z-direction. Consider (1 mm x 1.9 mm) in r-phi & z for B1. In B2, the bin size above is (1.7 mm x 3 mm).

Jan. 2009Jinyuan Wu & Tiehui Liu, Hits with Finer Bins hits/bin hits/bin With binning in z-direction, things become not hopeless. In r-z view, there can be 60 roads (when eta = +-1) in each Pt corridor. (x20 than the Thin Road scheme) For each hit on B0, a z range of mm on B2 must be checked in order to cover eta +- 1 range. With 3 mm binning in z: –180 mm / 3 mm = 60.

Jan. 2009Jinyuan Wu & Tiehui Liu, Roads in a Pt Corridor For each hit in B0, there can be many roads in r-z view, depending on eta range (eta are shown above). The similar set of roads repeat along z.

Jan. 2009Jinyuan Wu & Tiehui Liu, The R-Z Region to Work With The 2 sigma collision point is mm and a corresponding cut can be made to reduce fake rate. The region to work with for eta = +- 1 is shown above. For each bin (3mm) at B0, a range of about 180 mm, or 60 bins on B2 is searched. For each bin in B0, there are more than 120 hit patterns. There are about 100 bins at B0. Total number of patterns: 120 x 100 = 12K.

Jan. 2009Jinyuan Wu & Tiehui Liu, Good, But Using the Thin Road scheme as a starting point, adding x20 z-binning provides finer constraints in r-z view and reducing fake tracks becomes possible. The number of required patterns in B0 alone is 12K x 9K = 108M, (hand estimate only without considering multiple scattering etc.). Increasing AM size? Maybe it is OK for 108M. But it is endless when bins become finer, when strip layers are included, etc. There exist clever approaches. Too many patterns? 1.Reuse 2.Reuse 3.Reuse Tiny Triplet Finder is a scheme of reusing hit patterns.

Jan. 2009Jinyuan Wu & Tiehui Liu, Classification of Pattern Recognition SoftwareHardware Typical Hardware Resource Saving Approaches Doublet Finding O(n 2 ) for(){ for(){…} } O(n)*O(N) Comparator Array Hash Sorter O(n)*O(N): in RAM Triplet Finding O(n 3 ) for(){ for(){…} } O(n)*O(N 2 ) CAM, Hugh Trans. Tiny Triplet Finder O(n)*O(N*logN) O(n 4 ) for(){ for(){ for() {…} }}}

Jan. 2009Jinyuan Wu & Tiehui Liu, The Tiny Triplet Finder Triplet: e.g., 3 hits in a straight line in a plane. A semi-formal definition: n items satisfying n-2 constraints. Tiny = Small silicon resource usage. Typical implementation of triplet finding with N bins: –Hough transform: O(N^2) histogram bins. –AM: O(N^2) roads. Tiny Triplet Finder: O(N*log(N)) shifting elements.

Jan. 2009Jinyuan Wu & Tiehui Liu, Tiny Triplet Finder for FTK, Step (1a) Choose a narrow row (0.5 mm) of pixels along z-axis on B0 as seeds. Divide B1 into bins in r-phi view. The width of B1 bin is chosen to be (0.5 mm)*R1/R0 = 1.1 mm. Define the r-phi corridors. An r-phi corridor contains the seed bin in B0, a bin in B1 and the projected range (3.7 mm) in B2. Note that ranges of different corridors in B2 overlap each other. Up to 7 corridors can be defined.

Jan. 2009Jinyuan Wu & Tiehui Liu, Tiny Triplet Finder for FTK, Step (1b) For each r-phi corridor, book hit patterns of B2 and B1 (not B0) to register arrays. Bin sizes (r-phi x z): –B1: (1 mm x 1.9 mm) no overlap. –B2: (3.7 mm x 3 mm) with overlap in r-phi but not z. 7 sets of the register arrays are needed corresponding to 7 r-phi corridors.

Jan. 2009Jinyuan Wu & Tiehui Liu, Tiny Triplet Finder for FTK, Step (2) For a hit in B0, align the two arrays to search for a valid road. For another hit in B0, shift the two arrays to reuse the road patterns. Only one set of the road patterns is needed. Other sets of patterns becomes available by shifting. However, shifting is not free. See next two pages.

Jan. 2009Jinyuan Wu & Tiehui Liu, Actual Hardware Implementation *R1/R3 *R2/R3 Triplet Map Output To Encoder Bit Array Logarithmic Shifter Bit Array Bit-wise Coincident Logic 1.Fill the B1 and B2 bit arrays. (n1 clock cycles) 2.Loop over B0 hits, shift bit arrays and check for coincidence. (n0 clock cycles) Logarithmic Shifter

Jan. 2009Jinyuan Wu & Tiehui Liu, The Logarithmic Shifter S1 S2 S4 # of bits: N Shift distance: L # of stages: log 2 L Total LE usage: N*log 2 L A shift of X bit of the bit pattern is done in one clock cycle rather than X cycles. Logarithmic shifter is also known as “barrel shifter”, but the term “logarithmic” reflects nature of implementation, resource usage and propagation delay better. If number of bins x2, just add one layer, rather than x4 of total silicon resource.

Jan. 2009Jinyuan Wu & Tiehui Liu, Irregular Geometry The modules are tilted and the r at different phi in a module are different. However, relative distances in each Pt corridor are well defined. The bin widths (in both r-phi and z) are adjusted according to the relative distance when the register array is booked. Bins in these corridors are slightly narrower. Bins in these corridors are slightly wider.

Jan. 2009Jinyuan Wu & Tiehui Liu, Adding Strip Layers The occupancy of the pixel layers are high. Additional constraints may be needed. It may be necessary of do coincidence for all 7 layers. (Perhaps, 6-out-of-7?).

Jan. 2009Jinyuan Wu & Tiehui Liu, Using Outer Strip Layer Hits as Seeds It is also possible to use strip layer hits as seeds.

Jan. 2009Jinyuan Wu & Tiehui Liu, Using Middle Strip Layer Hits as Seeds It is possible to choose seeding layer to for better performance.

Jan. 2009Jinyuan Wu & Tiehui Liu, Tiny Triplet Finder for More Then 3 Layers The triplet finder can also search coincidence in more than 3 layers. A coincidence for 5 layers is shown above. Seed Layer

Jan. 2009Jinyuan Wu & Tiehui Liu, Booking Bins in z for Strip Layers Use stereo and axial strips to book bins in both r-phi and z- directions. Book a bit pattern for the stereo layer. For any hit strip in axial layer, copy selected bit pattern for z-bins. Fake hits may exist but they can be eliminated after track finding.

Jan. 2009Jinyuan Wu & Tiehui Liu, How Tiny Triplet Finder Fits into FTK Project From DF (or maybe DO), output an additional copy of HITS information via S-links. Another set of crate carrying Tiny Triplet Finder or other R&D scheme can be tested. The found patterns + hits can be fed into existing track fitter for further cleanup. FTK Pattern Recognition R&D Platform FTK Pattern Recognition R&D Platform FTK Pattern Recognition R&D Platform

Jan. 2009Jinyuan Wu & Tiehui Liu, Summary Using z information in pixel helps. However, number of patterns increases a lot. Tiny Triplet Finder reuse patterns so that binning in z becomes doable. The Tiny Triplet Finder or other pattern recognition schemes can be tested in parallel with the baseline AM approach. We wish to attract colleagues to work on simulation with z- binning to understand efficiency and fake triplet rate.