1. Karol Buńkowski Warsaw University The RPC based muon trigger of the CMS Experiment XI Workshop on Resistive Plate Chambers and Related Detectors, 5-10 Feb 2012, Laboratori Nazionali di Frascati dell'INFN, Frascati (Italy)

2. 2 CMS detector Karol Buńkowski, UW RPC2012, 5-10 February 2012

3. The Trigger and DAta Acquisition system at CMS Readout buffers 128 events = 3.2  s Level 1 Trigger Dedicated electronics (ASICs, FPGAs) @ 40 MHz, only logic functions Analyses every event (bunch crossing, BX)  pipeline processing; latency 3.2  s, including ~2  s for data transmission between the detector and counting room, dead time free operation Output ≤ 100 kHz High Level Trigger (HLT) Computer Farm: 1008 nodes, 9216 cores, 16 TB memory runs the software events selection algorithms A few hundreds of Hz recorded on the magnetic tapes Event Builder - switching network. Gathers the data from one event into one HLT computer Coarse data Detector keep reject DAQ: readouts the data for the selected events, the events are fragmented Karol Buńkowski, UW RPC2012, 5-10 February 2012

4. 4 Level 1 trigger system ` ` 4  4+4  4  MIP+ ISO bits L1A (trigger) 40 MHz pipeline ECAL Trigger Primitives ECAL Trigger Primitives HCAL Trigger Primitives HCAL Trigger Primitives Regional Calorimeter Trigger Regional Calorimeter Trigger Global Calorimeter Trigger Global Calorimeter Trigger RPC hits CSC hits DT hits Segment finder Track finder Track finder Pattern Comparator Pattern Comparator Segment finder Track finder Track finder Global Muon Trigger Global Trigger TTC system TTS system Detectors Frontend Status Link system 32 partitions Muon TriggerCalorimeter Trigger e/ , J, E T, H T, E T miss Karol Buńkowski, UW RPC2012, 5-10 February 2012 Trigger subsystems: identify, measure and sort the trigger objects Global Trigger apply cuts: single or multi-objects, topological correlations

5. 5 Counting roomDetector FEB RPC PAC muon trigger Karol Buńkowski, UW RPC2012, 5-10 February 2012 Trigger Board PAC Resistive Plate Chambers Up to 6 layers of detectors. 480 chambers in barrel, 504 in endcaps FEB Control & diagnostic Ghost Buster & Sorter RMB To the Global Muon Trigger Link Board Synchronization Unit & LMUX Optic Links 90 m @ 1.6 GHz 1104 fibers LVDS cables To Data Acquisition GB & Sor ter Data Concentrator Card 1232 Link Boards in 96 Boxes, Steered by Control Boards 84 Trigger Boards in 12 Trigger Crates Data transmission @ 320 MHz SYN CH. & LD MU X * Numbers of elements for the staged version of the system

6. 6 Geometry of the RPC detector and PAC trigger segmentation Karol Buńkowski, UW RPC2012, 5-10 February 2012 6 concentric layers of chambers in the barrel region, and 3 disc layers in each endcap (currently to |  | < 1.6, the endcap detector is staged, the 4 th endcap station will be added in 2013/2014) in phi plane: 1152 strips in each layer  one strip = 0.3125˚ the detector is segmented in the eta plane into the trigger towers (~0.1-0.2 eta unit each) A tower comprise from 3 to 6 chamber layers

7. 7 Trigger Algorithm: Pattern Comparator (PAC) Karol Buńkowski, UW RPC2012, 5-10 February 2012 The candidate is generated even though not all planes have hits. The minimum required number of fired planes is 3 (out of 3, 4, 5 or 6 planes available – depending on a tower). In this way the trigger efficiency is not suffering from the limited geometrical acceptance and inefficiency of the chambers. The number of fired planes defines the candidate quality. The quality is used for the candidates sorting and “ghost busting” (cancelation of duplicated candidates). 3/4 RPC layers A pattern is a set of AND gates connected to selected strips strips The chamber signals (fired strips) are compared with the predefined set of patterns. Each pattern has assigned the p T and sign (depending on the track banding by the magnetic filed). Muon candidate is recognized if the hits fit to the pattern and are in the same clock period (BX)

8. 8 Implementation of the PAC algorithm in the FPGAs The trigger algorithm is implemented in the FPGA devices - Altera Stratix 2, 300 chips are needed to cover full detector. Each PAC comprises max 576 chamber strips and contains 3 000 – 14 000 patterns (most of them low p T ). The patterns are built-in the firmware logic. The patterns are generated based on the simulated muon track. Advanced algorithms are used to create the patters from the simulated chamber hits, assign the p T, and then select optimal set of patterns. The goal is to achieve best possible trigger efficiency and purity with a patterns set that can be fit into the PAC FPGAs. Since each PAC contains different patterns, for each chip separate compilation is needed. The software framework for patterns generation and firmware compilation on the computer cluster was created. One iteration takes ~24 hours. As the PAC algorithm is implemented in the reprogrammable FPGAs, it can be easy changed, e.g. to correct bugs, improve performance, or implement new features. Karol Buńkowski, UW RPC2012, 5-10 February 2012

9. Synchronization of the trigger system (1) 4.2m = 14ns 14m = 42ns Seminarium Oddziału Fizyki i Astrofizyki Cząstek IFJ, 26 maja 2009 Karol Buńkowski, UW The time of muon flight from the interaction point to the different chambers varies from 14 to 42 ns - more than 1 BX The time of signal propagation from the chambers to the Link Boards varies from 33 to 107 ns (due to differences in the cables lengths) The chamber hits must be in the coincidence (in the same BX, i.e. 25 ns clock period) on the PAC input to produce the muon candidate  the system synchronization is crucial for its performance

10. 10 The initial position of the synchronization window winOpen i and data delay d i data was calculated based on: –muon hits timing t i hits which is a sum of the muon time of flight (know from the simulations) and signal propagation time in the cables, –known length of the fibers transmitting the clock (clock phase difference  i TTC ): winOpen i = (t i hits +  i TTC + offset) % 25 ns d i data = a – int[(t i hits + offset)/25ns] + b i - (1 * ) + c i win + (2 SM ) Then the synchronization was corrected based on the collected collision data Synchronization of the trigger system (2) Karol Buńkowski, UW 25ns collision LB1 LB2 LB3 delay Synchronized signals Time of flight propagation in cables time Synchronization window RPC2012, 5-10 February 2012 The chamber hits are synchronized to the 40 MHz LHC clock in the Link Boards. The hits are “quantized” to the full BX (i.e. the timing is measured with the 25 ns precision) with used of the “synchronization window”. Its position can be adjusts with 0.1 ns accuracy. Then the hits are aligned between the Link Bards by applying full BX delays. The goal is to have all hits of all muons from given event within 25 ns on all LBs.

11. 11 Calculation of the timing correction from hits BX distribution After the initial synchronization in most of the LBs the chamber hits were concentrated in ONE or TWO neighboring BXs: -1, 0, 1 Karol Buńkowski, CERN, UW, L1 DPG, 22 April 2010 BX =0BX =1BX = -1 window #hits Timing correction winOpen time From the data we know only the distribution of the hits in the BXes (w.r.t. the correct BX of the event). From this the value of the timing correction must be obtained. We have not measured the hits timing distribution from the collisions because it would required time consuming scanning. We utilized the hits timing distribution obtained from the simulations Assuming that the mean hits BX (from data) corresponds to the cumulative distribution function of the simulated timing of the hits, the timing correction can be calculated: Muon hits timing from simulations [ns] Hits distribution Cumulative distribution

12. 12 Synchronization of the trigger system - results Karol Buńkowski, UW RPC2012, 5-10 February 2012 Distribution of chamber hits BX w.r.t. the event BX Distribution of the hits mean timing and spread (rms) for individual Link Boards 99.98% of hits associated to the muon tracks are in the correct BX=0 Since the start of the collisions in the April 2010 the synchronization was corrected 7 times. In ??? of the 99.9??? % of hits is in the correct BX Bad timing on a few LBs due to problems in chambers or signal cables Data selectio n!!!!!!!! !!!!!!!!!

13. 13 Trigger on HSCPs Some supersymetry and models foresee Heavy Stable Charged Particles (HSCPs), e.g. stop, gluino, stau. They mass could range from ??? To ??? GeV, thus if produced at LHC they velocity would be ~0.2 – 0.9 c. In the CMS they will look like “slow muons”: the hits in the muon chambers (all or outermost) can be up to 1 BX later than the hits of the muons  –they will not produced the muon trigger at all (hits not in coincidence in one BX) or –the trigger will be 1 BX to late  the tracker hits will not be recorded (pixel detector stores only the hits from one BX/event). Karol Buńkowski, UW RPC2012, 5-10 February 2012

14. 14 Trigger on HSCPs PAC modification Karol Buńkowski, CERN, UW, Trigger Meeting, 28 June 2011 layer 6 layer 5 layer 4 Chamber hits (PAC input) extended hits (in the PAC) BX Muon candidate normal muon layer 3 layer 2 layer 1 L1A Masked by BPTX veto In the PAC trigger we found a way to trigger on the HSCPs: In the PACs the detector signal are extend to 2 BX and On the GMT input the PAC candidates delay is reduced by 1 BX (w.r.t. the DT and CSC candidates)  the hits of the “late particle” generate the trigger in the proper BX!  for in-time muons candidates in 2 BX appear - the first candidate is too early, but he second is in the proper BX. The first candidate is masked on the GT by the BPTX veto – signal synchronous to collision, but advanced 1 BX (used for all trigger to eliminate the pretriggering). layer 6 layer 5 layer 4 BX late particle layer 3 layer 2 layer 1 L1A Muon candidate Chamber hits (PAC input) extended hits (in the PAC) Significant increase of the efficiency to trigger on lower momentum, slower moving HSCPs e.g. for gluino 800 GeV from 24 to 32% Guino 800 GeV MonteCarlo

15. 15 Timing of the RPC PAC candidates results Karol Buńkowski, UW RPC2012, 5-10 February 2012 99.9??% of PAC candidates are in the correct BX=0 Data selection !!!!!!!!!!! !!!!!! The candidates corresponding to the muons from the collisions are duplicated in the BXs -1 and 0 To early or to late candidates (~10 -4 ) are mostly from the cosmic muons BX of the RPC candidates w.r.t. the L1 trigger BX

16. 16 Efficiency of the RPC detector and PAC trigger vs.  Karol Buńkowski, UW RPC2012, 5-10 February 2012 Data selection and method !!!!!!!!!!!!!!!!! The efficiency of RPC PAC trigger for identifying muons is a convolution of: ε acceptance – geometrical acceptance of the RPC detector (probability that muon crosses at least 3 chambers), ε chambers – chambers intrinsic efficiency, ε patterns – patterns efficiency i.e. probability that the chamber hits of a “triggerable” muon fit to any pattern; “triggerable” muon – hits in at least 3 RPC layers inside the eta-phi cone covered by one PAC unit and in the same BX ε triggerable muon = ε acceptance  ε chambers Detector acceptance Triggerable muons RPC PAC trigger eff.

17. 17 RPC PAC efficiency - turn on curves From tag and probe??? Karol Buńkowski, UW RPC2012, 5-10 February 2012

18. 18 RPC chambers monitoring via the PAC trigger hardware Karol Buńkowski, UW RPC2012, 5-10 February 2012 Time [s]strips Hits rate [Hz] In the Link Boards firmware the multichannel counters allowing to measure the signal rate for each strip individually are implemented: -All hits are counted: no bias from trigger (unlike in the DAQ data), big statistic, -The signals for all strips are counted, even those masked, -The data are stored for the offline analysis: neutron background, chamber noise  noisy strips masking, Front-end thresholds tuning (see talk by ?????). The basic plots (rate v.s. time for each chamber, average and maximal rate per strip) are produced in the real time by the software controlling the hardware – the chambers performance can be evaluated online, the problems (noisy strips, dead chambers) can be noticed promptly.

19. 19 2011 performance summary Only a few minor hardware failures, promptly repaired. Beside that 100% of the trigger hardware operational and working correctly. CMS down time (dead time?) due to the RPC PAC trigger during 2011 collisions only !!!% Excellent synchronization of the system: 99.98% of muon chamber hits in the correct BX  99.99??% of the PAC candidates in the correct BX. The only trigger subsystem capable to trigger on HSCP. Average RPC PAC trigger efficiency 80???%. Continuous work on the patterns optimization, according the CMS requirements (efficiency – rate tradeoff) Karol Buńkowski, UW RPC2012, 5-10 February 2012

20. 20 backup Karol Buńkowski, UW RPC2012, 5-10 February 2012

21. 21 Karol Buńkowski, UW RPC2012, 5-10 February 2012