1. L1Calo Intro Cambridge Group, Dec 2008 Norman Gee

2. Trigger strategy RoIB High-level Trigger ROD Other detectors Calorimeters Muons CALO trigger MUONS trigger CTP Level-1 L1A. 40 MHz 75 kHz <2.5  s ROB Det. readout DAQ Event Builder ROS Level-2 <40 ms L2S V EFN L2A. 2 kHz RoI requests RoI Data Regions of Interest (RoI) Event Filter ~1 s L2N EF Acc. 200 Hz Full events hardware Software 300 Mb/s 120 Gb/s 3 Gb/s L1 Dedicated hardware (ASICS & FPGAs) Calorimeters & muons Latency < 2.5  s L1A 75 kHz (Max 100kHz) L2 ~500 dual CPUs Full granularity Regions of Interest (~2%) Latency ~40 ms L2A 2kHz Event Filter (L3) ~1600 dual CPUs Access to full event & calibration constants More detailed reconstruction Use Offline algorithms Latency ~1s 200Hz 2

3. Level-1 trigger system (L1) 3 sub-systems L1 - Calorimeters (L1Calo) L1 – Muons (L1Muon) Central Trigger Processor (CTP ) Signature identification e/  /hjets  e/ ,  /h, jets,  Multiplicities per E T threshold Isolation criterion Missing E T, total E T, jetE E T Preprocessor Cluster Processor e/  /h e/ ,  /h Jet/Energy Processor jetsE T jets, E T Muon Barrel Trigger Muon Endcap Trigger Muon-CTP Interface Central Trigger Processor LAr TileRPCTGC RoIB L2 supervisorDetector readout CTP Receive & synchronize trigger information Generate level-1 trigger decision (L1A) Deliver L1A to other sub-detectors 3

4. L1 Calorimeter - Architecture L1Calo partitioned into 3 sub-systems  Real time transmission to CTP  DAQ + RoIs at each L1A ( 75kHz ) Pre-Processor (PPr) Receive & sample signal from calorimeters Trigger Towers Coarser granularity (Trigger Towers) Noise filter Bunch crossing identification (BCID) E T Determine final E T value Processors JEP & CP Physics algorithms Search for and identify:  isolated leptons, taus  jets Compute E T total, missing,… 4

5. 5 Electromagnetic calorimeter - Endcap Cryostat Electromagnetic calorimeter - Barrel Tile Calorimeter Hadronic Calorimeter – End cap Forward Calorimeter Calorimeters

6. Trigger towers (TT) LArTiles (semi-projective segmentation ) trigger towers map 0.1x0.2 3.1 <|  |< 3.2 0.4x0.4125 3.2 <|  |< 4.9 0.2x0.2 2.5 <|  |< 3.1 0.1x0.1 |  |< 2.5  x  Position Analogue summation of calorimeter cells 3584 x 2 (EM+HAD) trigger towers 6

7. 7 Overview of the trigger front-end Long cables: Cavern to USA15 Typically 40–70m long Detector responsibility 256 TileCal, 360 LAr Short cables: Internal to trigger racks Typically 5–8m long L1Calo responsibility 776 in total All analogue cables: 16 individually shielded pairs F/R denote front/rear of modules All connectors D37 except TCPP inputs, which are D50

8. 8 Cable Installation

9. 10 b Sampling 40 Mhz, Flash-ADC 10 bits 1 FADC count  250 MeV Pedestal 40 FADC counts PPr Receivers (Rx) Input signal conditioning to L1 (2,5V  250GeV) Variable gain amplifier (VGA) E  E T Conversion (Hadronic layers only, LAr already E T ) Local signal monitoring Pre-Processors – Energy reconstruction Transmission to processors & DAQ 10 b E T calibration Look-Up Table Look-Up Table (LUT) Pedestal subtraction, noise suppression FADCGeV FADC (10b)  GeV (8b) conversion 1023 255 E thres 8b Bunch crossing identification (BCID) Finite impulse response filter (FIR) Peak finder (linear/saturated) Assign E T to the ‘correct’ bunch crossing a0a0 a1a1 a2a2 a3a3 a4a4 + FIR filter

10. 10 PreProcessor Module (PPM) [Heidelberg] RAW FADC values and LUT outputs are Read out when L1A fires Multi-chip module (MCM) with FADCs, ASIC and LVDS serialisers Clock for FADC individually adjusted to sample on pulse peak. Then digital data synchronised (coarse & fine adjustment) to identical times. Towers for CP multiplexed, then serialised to 400 Mb/s and sent as LVDS on cables to CP For JEP, 2 x 2 em or Had towers summed, then sent at 400 MB/s to JEP

11. Algorithms and Regions of Interest (RoI) Processor input is a matrix of tower energies Algorithms look for physics signatures (sliding window) RoI data sent to Level-2 trigger following L1A Cluster Processor Jet/Energy-sum Processor Criteria for e/  or  /h candidate: EM or Had. cluster > E threshold Total E T in EM Isolation Ring  EM isolation thresh. Total E T in Had. Isolation Ring  Had. isolation thresh. Local E T Maximum compared to neighbor windows. e/  only: Had. core  core isolation threshold Jet candidate Coarser granularity 0.2x0.2 (jet element) Digital summation EM + Had. Sliding, overlapping windows (3 sizes) Missing energy ECAL+HCAL 11

12. 12 Cluster Processor Module Serialisers CP FPGAs Readout

13. Backplane Traffic 2 columns of fan-out towers to adjacent CPM on ‘left’ (- η) 1 column of fan-out towers to adjacent CPM on ‘right’ (+η) Backplane Fan- in/out runs at 160 Mb/s to reduce pin count Output: hit counts. 3 bits for each of 8 threshold sets, + parity. 25 e/  bits go left, 25  /h bits go right

14. Backplane 14

15. Jet-Energy module [Mainz/Stockholm] 15

16. Common Merger Module

17. 17 Central Trigger Processor (CTP) Receivesynchronizealign Receive, synchronize and align trigger information Other signals: Random trigger Calibration Minimum biais events (MBTS) Generate the level-1 trigger decision (L1A) Programmable trigger menu Latency 100 ms (4BC) Deliver the L1A to the other sub-detectors

18. 18 -2 0 +1 +2 Bunch crossing Average offset from L1A timing in EM layer Average offset from L1A timing in EM layer

19. 19 Energy correlation with calorimeters HadronicElectromagnetic L1Calo reconstructed E T vs calorimeter precision readouts Cosmic muons Reasonable correlation achieved Very crude calibration applied, still room for improvement L1Calo

20. 20 LAr EMB timing from pulser run Coarse timing ok Focusing on fine tuning Timing calibration Internal timing of L1Calo trigger is achieved Input timing realized in pre-processors Not a trivial task Several strategies depending on signal origin: calibration cosmic rays collisions Different setup & automatic procedures Setup coarse/fine timing Ensure signals are correctly sampled (3 rd sample at maximum) Signal shape with 1 ns sampling step

21. 21 Pedestal & Noise Pedestals set to 40 ADC counts Sensible RMS ~ 400 MeV Nearly all channels behaving correctly (>99%) HadronicElectromagnetic Mean RMS ADC counts