Presentation is loading. Please wait.

Presentation is loading. Please wait.

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.

Similar presentations


Presentation on theme: "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."— Presentation transcript:

1 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 U)

2 geometry assume long barrel design for tracker 10/28/2009Ulrich Heintz - CMS Upgrade Workshop2 IP Layer 1 -- Inner station Layer 2 -- Middle station Layer 3 Layer 4 Layer 5 Layer 4 Projection Layer 3 Projection 21 Z modules 28 Z modules 10 Z modules Z R Layers 3, 4, 5 Outer Station r=350 cm r=550 cm r=1100 cm BUT: many ideas also apply to geometries with less layers and to different low pT hit rejection schemes (e.g. cluster width)

3 doublet formation reject hits from low p T tracks on detector cluster adjacent pixels with hits  reject clusters >2 pixels wide require coincidences between clusters in two closely spaced modules in a stack  reject single clusters  reject doublets from soft tracks 10/28/2009Ulrich Heintz - CMS Upgrade Workshop3 Doublet (2-layer coincidence) x x x x x x x x x x x x x x x x x x x x r-  view of rod structure see Ron Lipton’s talk at track trigger session tomorrow for details on technical aspects

4 rates MC simulation (Laura Fields, Cornell U, track trigger session tomorrow)  cluster rate max 2 pixels wide  doublet rate 1 mm stack separation 2 GeV threshold 10/28/2009Ulrich Heintz - CMS Upgrade Workshop4 R = 35 cmR = 55 cmR = 110 z=0 44  1.6  0.2 MHz/cm 2  10 44  0.5 /xing/module R = 35 cmR = 55 cmR = 110 z=0  0.3  0.13  MHz/cm 2  0.7  0.3  0.06 /xing/module

5 number of fibers average rate over length of rod  1 / 2 max z=0 safety factor of  10  1.5 MHz/cm 2 in inner layer 42 modules in z  4200 cm 2 20 bits/doublet  250 Gb/s per sector per station (2 stacks) bandwidth of fiber links = 3.6 Gb/s  70 GBT per sector per station   5000 GBT for entire tracker 10/28/2009Ulrich Heintz - CMS Upgrade Workshop5

6 basic idea represent every possible hit combination by a logic “equation”: S 1i  S 1o  S 2i  S 2o  S 3i  S 3o  create a table of all possible equations in FPGAs  feed all hits for an event into the FPGAs  evaluate all equations simultaneously.  the equations which are satisfied correspond to the reconstructed tracks  timing dominated by time needed to load hits into FPGAs problem  too many equations  too many inputs  need to factor problem 10/28/2009Ulrich Heintz - CMS Upgrade Workshop6

7 sector structure trigger logic handles inputs from sectors n-1, n, n+1 active area of module  95 mm (  ) x 100 mm (z) sector size determines min p T for full acceptance 10/28/2009Ulrich Heintz - CMS Upgrade Workshop7 n n-1 n+1

8 minimum p T what is the minimum p T we need to trigger on? built into the hardware design need to decide soon  here assume 15 o sectors 10/28/2009Ulrich Heintz - CMS Upgrade Workshop8 min p T sector opening angle 2 GeV18 o 2.5 GeV15 o 5 GeV7.5 o  requires more/wider modules  lots of inputs but geometry easier  less inputs – makes trigger easier

9 tracklet formation combine doublets from the two stacks in each station to form tracklets  drop in rate by about factor 4 (but need  40 bits/tracklet)  for each stack in inner layer need to process data from two adjacent stacks in z in outer layer 10/28/2009Ulrich Heintz - CMS Upgrade Workshop9 Doublet (2-layer coincidence) Tracklet (4-layer coincidence with P T validation) x x x x x x x x x x x x x x x x x x x x

10 n n-1 n+1 equation count assume azimuthal position resolution  0.1 mm number of  positions per sector:   2900 in outer station   1450 in middle station (for each outer station position) number of equations   2900 * 1450 = 4,200,000  too many equations for single chip number of fibers  210 fibers from each sector  too many inputs for single chip divide outer station into 12 sub-sectors  route tracklets from inner and middle stations to subsectors using interaction point and p T 10/28/2009Ulrich Heintz - CMS Upgrade Workshop10

11 sector processor 10/28/2009Ulrich Heintz - CMS Upgrade Workshop11 5 FPGAs inner station (doublets) 70 5 FPGAs middle station (doublets) 70 6 FPGAs outer station (doublets) 70 3 FPGAs (tracklets) 35 3 FPGAs (tracklets) FPGAs inner and middle station tracklets from sector n-1 inner and middle station tracklets from sector n tracks out 12 subsector blocks tracklet blocks sorter blocks

12 sector processor outer station  no pT information required for sorting  no need to form tracklets in outer station  robust against inefficient sensors  number of equations in each sector  4,200,000/12  350,000 middle and inner stations  loose entire station if there is an inefficient sensor solution  in parallel sort into 12 subsectors anchored in middle layer and into 12 subsectors anchored in inner layer  robust against inefficient sensors in any layer 10/28/2009Ulrich Heintz - CMS Upgrade Workshop12

13 sector processor for each trigger specify number of tracklets and doublets  accept tracks with 3 tracklets 2 tracklets and 0 or 1 doublets 1 tracklet and 0 or 1 doublets do r-  tracks satisfy track equations in z? remove duplicate tracks 10/28/2009Ulrich Heintz - CMS Upgrade Workshop13

14 summary 3 trigger stations provide full coverage for tracks with p T >2.5 GeV in 15 o sectors hits collected in real time, sent off-detector on  5000 optical fibers at 3 Gb/s all possible track equations for each sector evaluated in parallel using FPGAs in USC to produce inputs for L1 total of  2000 large FPGAs needed to build this using current technology 10/28/2009Ulrich Heintz - CMS Upgrade Workshop14

15 next steps robust against hardware failures very powerful pattern recognition  large channel count, power, mass need MC simulations implementing specific algorithms verify rate estimates with LHC data  do we really need 6 trigger layers?  what is the smallest number of trigger layers needed? 10/28/2009Ulrich Heintz - CMS Upgrade Workshop15


Download ppt "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."

Similar presentations


Ads by Google