BTeV%20Level%201%20Muon%20Trigger%20Algorithm

1
BTeV Level 1Muon Trigger Algorithm

• Michael J. Haney
• 16 May 2001
• www.hep.uiuc.edu/engin/btev/reports/16may01.ppt
  (and .pdf)

2
Numbers

• 2 arms, 8 octants
• each octant-arm treated independently
• 16-fold (geographic) symmetry
• 3 stations (1,2,3) , 4 views each (R,U,V,R)
• 13 (12?) planks, 32 tubes each
• double-layer (1616, offset)

3
Assumptions

• complete octant-arm data delivered to board
• implies 16 boards
• can be fewer if density permits (1/N)
• may be necessary (see last slide)
• complete crossing data
• no mixing
• must finish current before beginning next

4
(soft) Assumptions

• complete view data
• must finish current before beginning next
• data delivered radially inward
• outer to inner
• These assumptions are not critical,but if they
  are easy for the Muon front end,they will
  reduce/eliminate sorting...

5
Algorithm (FPGA)

• compress adjacent tube hits
• can be done serially (on-the-fly) if the soft
  assumptions are met
• need to understand how to handlemany-adjacent
  hits...

6
Algorithm (FPGA/DSP)

• AND/OR (selectable) R,R views
• reduce to single RR view
• working radially inward,form valid RR2-U2
  pairs(fixed table size gt outer bias)
• look-up V2s from RRU2sform RRUV2s
• and RRV2s and UV2s, etc.as space/time
  permits...

U
2
V
2
3
3
4
4
6
8
R
10
7
Algorithm (DSP)

• window-search RR3sbased on RRUV2s
• look-up U3s and V3sfrom RRUV2RR3s
• look-up RR1s, U1s, V1sfrom RRUV2RRUV3s
• declare RRUV1RRUV2RRUV3s(and various lessers)
  as tracks
• calculate c2s (?)

8
Algorithm Notes

• radially-inward processing
• biased in favor of higher PT
• implications on front end readout (order)
• fixed table sizes
• accommodates wall of fire near beam pipe
• octant-arm specific
• 16-fold symmetry in hardware
• no inter-octant communications

9
Algorithm Notes (continued)

• combinatoric explosion - suppressed
• RRU2s require O(N2)
• look-up steps require O(M)
• window search? O(M) to O(M2)
• could be fully implemented inthe Pixel Trigger
  DSP Farm
• but compression/table-building may be donein
  (Pixel Trigger?) FPGAs

10
Loose Ends

• pure 16-fold geographic symmetryproduces 16
  opinions per crossingfor the Global L1 to
  consider
• may need to collapse 161 at FPGA-DSP boundary
• same as Pixel Trigger 311
• Even if we can use the same FPGAs,will Muon
  front end look likePixel front end (number of
  spigots)?

11
Highways

• The Muon Trigger has 16-fold geographic symmetry
• (8-fold per arm)
• 8-fold highways imply 128-fold symmetry
• may need need several folds per FPGA boardjust
  to keep the system size reasonable