1 / 23

Calorimeter Trigger for Phase I

M.Bachtis, S.Dasu, K.Flood, W.H.Smith University of Wisconsin Calorimeter Trigger Upgrade Workshop CERN Feb 19/2009. Calorimeter Trigger for Phase I. Introduction. SLHC Calorimeter trigger goals Provide High performance triggers in the SLHC environment

Download Presentation

Calorimeter Trigger for Phase I

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. M.Bachtis, S.Dasu, K.Flood, W.H.Smith University of Wisconsin Calorimeter Trigger Upgrade Workshop CERN Feb 19/2009 Calorimeter Trigger for Phase I

  2. Introduction SLHC Calorimeter trigger goals Provide High performance triggers in the SLHC environment Electrons and photons: Provide similar thresholds as in current trigger for SLHC conditions – factor of 2 rate reduction needed for same efficiency Taus: Improve background rejection so that the thresholds will be relatively low (not trying Ztt!) Jets: Provide better energy and position resolution Support communication with the tracking trigger High spatial resolution so that tracks can be mapped to Calorimeter objects Counting of tracks and track Isolation will resolve electrons from photons , Taus from jets

  3. Calorimeter signatures Electrons/Photons Spatially confined in a cluster of 2x2 trigger towers Significantly higher ECAL contribution Isolated e/γ should have low energy deposits in the surrounding area HCAL Δη x Δφ=0.087x0.087 η ECAL φ e/γ HCAL η • Taus • Confined in 2x3 Clusters • 3 prongs/1 prong + π0s have wider φ profile • Small energy leak in surrounding towers ECAL φ τ HCAL η • Jets • Most of the energy confined in a central core • For jets over 20 Gev the energy is included in a 8x8 region ECAL φ jet

  4. Algorithm Overview Particle Cluster Finder Takes Calorimeter TPG input and applies tower thresholds Creates overlapped 2x2 clusters Cluster Overlap Filter Removes overlap between clusters Identifies local maxima Prunes low energy clusters Cluster Isolation and Particle ID Applied to local maxima Calculates isolation deposits around 2x2,2x3 clusters Identifies particles Jet reconstruction Applied on filtered clusters Groups clusters to jets Particle Sorter Sorts particles and outputs the most energetic ones MET,HT,MHT Calculation Calculates Et Sums, Missing Et from clusters

  5. Particle Cluster Finder (I)‏ • Input • Each clustering block takes as input four towers (4x17 bits) • Tower • 4 ECAL Et bits 4 HCAL Et bits,1 finegrain bit • Algorithm • Apply ECAL,HCAL Et activity thresholds on the towers • For each four towers create a 2x2 cluster • If cluster pattern is not satisfied cluster is dropped • Calculate e/γcompatibility bit • Output a cluster (38 bits)‏ • 4 Towers (4x9 bits, E+H)‏ • OR of the finegrain bit • e/γCompatibility bit • Algorithm is applied with a step of one tower (overlapped clusters)‏ Pattern comparator 17 bit 1 bit Threshold 17 bit 1 bit Threshold 17 bit 1 bit Threshold 17 bit 1 bit Threshold 1 bit Pattern Decision check no Zero (38 bit)‏ yes Finegrain OR 1 bit e/γ bit calculation 1 bit Tower Energy Sums 4x9=36 bits

  6. Electron/ Photon bit ECAL energy fraction is a straightforward selection cut Looking towards a generic LUT implementation Zee QCD Electrons in the EB-EE gap

  7. Clustering examples TTbar -NoPU TTbar -NoPU TTbar -PU TTbar -PU

  8. Cluster Overlap Filter Cluster Overlap Filter is applied on the clusters produced by the Particle Cluster finder The algorithm takes as input 9 clusters A central cluster The 8 neighboring clusters Algorithm For a pair of the central cluster and any neighbor: If central Et <neighbor Et: Remove overlapping towers with the neighbor from the central cluster If no towers are pruned Cluster is characterized as central After pruning threshold is applied in cluster energy Output 11 bits of cluster energy 1 Finegrain Bit, 1 e/γ bit, 1 central bit Cluster origin (holds all cluster info)‏ Central cluster Pruned tower 38 bits per input Neighbor cluster NE E SE S SW W NW N

  9. Cluster Filter Logic E1 E2 Finegrain,e/γ 2 bits Energy Adder 11b tower bit sequence (4x9 bits)‏ E3 E4 Central 1bit Cluster Threshold E>X? Central < NE? E1 E2 E3 E4 Energy Adder 11b NE Central (1 bit)‏ 1bit Central <= E? E1 E2 E3 E4 Energy Adder 11b E 1bit Central <= SE? E1 E2 E3 E4 Energy Adder 11b Energy (11bits)‏ SE 1bit Central <= S? E1 E2 E3 E4 Energy Adder 11b S 1bit Central <= SW? E1 E2 E3 E4 Energy Adder 11b SW 1bit Central <W ? E1 E2 E3 E4 Energy Adder 11b Energy Adder E1+E2+E3+E4 W 1bit Central < NW? E1 E2 E3 E4 Energy Adder 11b NW 1bit Central < N? E1 E2 E3 E4 Energy Adder 11b N

  10. Cluster Filtering examples A large portion of the clusters are removed in the filtering step The thresholds for this test are relatively low Cluster Threshold = 3 GeV TTbar - PU TTbar - NoPU Filtered cluster

  11. Cluster Isolation Two isolation approaches Absolute Isolation Require low Energy deposits around the object of interest Relative Isolation Require insignificant energy deposits compared to the energy of the object of interest Sliding isolation cuts Isolation should be tight only in low Et region Isolation should be very loose where the rate is low (Higher Et)‏ Object of interest Isolation Annulus • Isolated Deposits • Iso2x2 = ∑E (annulus)‏ • Iso2x3 = Iso2x2 – max {Eφ-1, Eφ+1} • For 2x3 isolation we used the maximum of the neighboring φ cluster • Isolation Selection • Iso2x2 <a + b E cluster + c Ecluster2 • Iso2x3 <d + e E cluster + f Ecluster2 • Different cuts for electron / tau isolation

  12. Cluster Isolation Logic 2x2<e/γ Cut? e/γ Isolation bit EGamma Cut LUT (a+bE+cE2)‏ Energy Scale LUT Central cluster Tau Cut LUT (d+eE+fE2)‏ φ+1 Max of the {φ+1,φ-1} Subtraction 2x3<τ Cut? _ τ Isolation bit φ-1 + Adder (63 inputs)‏ Clusters In annulus …

  13. Isolation Deposits The isolation distributions look very similar This is due to the Activity thresholds and the overlap filter Activity Thresholds: (E,H) = (1,3) GeV Cluster Threshold: 3 GeV Zττ Zee No PU No PU PU PU GeV GeV

  14. e/γ, τidentification After all previous steps, each cluster is described by 16 bits Energy : 11 bits Finegrain : 1 bit e/γ: 1/bit Central (local maximum) : 1 bit 2x2 Isolation : 1 bit 2x3 Isolation : 1 bit Single particle deposits are described by clusters that are not pruned (Central bit is set) Electrons/Photons Central bit is set (Cluster is not pruned)‏ e/γbit is set, finegrain veto is not set (e/γ-like cluster)‏ Isolated Electrons/Photons Electron requirements AND (2x2 isolation bit set)‏ Taus Central bit is set (Cluster is not pruned)‏ 2x3 Isolation bit is set

  15. Jet Reconstruction Jets are reconstructed by summing energy around a local maximum cluster Require The energy weighted position is inside the central cluster Minimum jet energy Alternatives Variable size is also possible Split the jet region in zones and require energy cutoffs on each zone Reconstruct jets in 8x8 regions η φc Jet core (central bit=1)‏ φ ηc Require centered energy deposits

  16. Jet reconstruction logic 64 bit inputs η<ηc ηL η-weighted energy calculation |ηL-ηR| |ηL-ηR|≤1 Energy Adder η>ηc η-weighted energy calculation ηR Jet Check φ>φc φd φ-weighted energy calculation |φu-φd| |φu-φd|≤1 Jet E φ<φc φu φ-weighted energy calculation Esum Esum>Thr

  17. ttbar after full chain Here the pattern recognition is depicted There are a lot of reconstructed candidates in those events but we trigger only on the four most energetic ones A lot of clusters in PU event but most of them do not pass the filter TTbar – No PU TTbar – PU Clusters Filtered Clusters e/γ τ jets

  18. Electron/Tau Efficiency Preliminary results Selection cuts not optimized yet ie Isolation could relax at lower Et for taus e/γ/Zee Isolated e/γ/Zee No PU No PU PU PU τ/Zττ No PU PU

  19. Comparison with present trigger Comparison with the standard trigger without energy corrections Benchmark Lumi: 2E32 cm-2s-1 Note:Both algorithms use uncorrected Et but for taus the energy thresholds are slightly different The SLHC Trigger tau Et is ~5-10% lower than LHC Trigger tau Et Isolated e/γ Trigger Single τ Trigger LHC Trigger LHC Trigger SLHC Trigger SLHC Trigger Significant improvement for taus ! The single Tau Threshold could be <1/2 the value we have now At high Et the isolation in SLHC Trigger relaxes So the rate slightly increasesIsolation cuts need to be tuned tighter for high luminosity

  20. Layout Example I (Clustering/Filtering)‏ Clustering /Filtering algorithms have: Number of Inputs = Number of outputs 24 Rocket I/Os can be used for input One Rocket I/O can receive9 calorimeter TPGs That corresponds to 216 Calorimeter TPG input (17 bit)‏ Some bits are needed for Error correction 24 Rocket I/Os can be used for output One rocket I/O can hold 10 clusters That corresponds to maximum 240 Filtered Cluster output (14 bits)‏ Optimal segmentation: 8x16 towers / card (+ 1,2 border/overlap layers)‏ 36 cards for Clustering/Filtering XILINX VIRTEX 5 TXT -48 Rocket I/Os -6.5Gbps 24 in 24 out Clustering/Filtering layout η φ

  21. Layout Example II(Isolation/Particle ID/Jets)‏ To perform jet reconstruction(8x8) on a 8x16 lattice, a card of 14x22 is needed (308 clusters) = 4310 bits The output is smaller in this case Assume maximum output / card = 25 objects (25 bits) = 625 bits Number of I/O bits ~5000 48 Rocket I/Os can handle 7800 bits So jet reconstruction and isolation could beunified to one card 10x10 /12x12 jets are also possible Summary Card(s)‏ Take as input the most energetic objects from each card Take as input Tracker Trigger data Separate objects by mapping tracker and calorimeterdata Sort the objects and output the most energetic ones to GT Restriction: Circuit logic resources Isolation/Jets layout η φ

  22. Algorithm implementation (Firmware) M.Schulte, K.Compton, B.Buchli, A. F-Farmahani, T.Gregerson, S.Naumov (ECE Department/UW)‏ Tested some RCT ASIC algorithms in FPGA Sorters, Adders (very high performance)‏ Now Implementing and optimizing the clustering algorithm Implementation of thresholds and pattern comparator Comparison of LUT, Division implementation for electron ID LUT is slower/larger than straightforward division implementation (but can implement the most general function)‏ 20 bit LUT:~380 MHz implemented at Block RAMs / 320 MHz on FPGA LUTS @ 6ns [7% of the LUTs] Alternative LUT sizes also under study Straight forward division (E/(E+H))‏ 420 MHz [0.1% of the logic] Shruva Bhattacharya and Will Plishker, Maryland FPGA Firmware design methodology Development management tools

  23. Conclusion Simulation of a set of SLHC Calorimeter trigger Algorithms is in place More studies and improvements on the way Threshold/Cut optimization Jet reconstruction alternatives Performance of Calibration/ eta correction on cluster level Studies of SLHC (Higher) PU Need to line up with Tracking Trigger Simulation What will be the track position granularity in Calorimeter surface? At which level , should we do the Calorimeter /Tracker Trigger combination We are looking towards the Hardware implementation & constraints

More Related