Title: Status report of the L2Cal upgrade for the L2Cal team Chicago, Fermilab, Madrid, Padova, Penn, Pisa, Purdue
1Status report of theL2Cal upgradefor the L2Cal
team Chicago, Fermilab, Madrid, Padova, Penn,
Pisa, Purdue
2Motivation
- Observe large increase of jet/met based trigger
cross sections - Reach limit of available bandwidth
- Efficiency of multi-jet triggers decreases
- Cause ?
- (True/fake) jets merged together
- Low ET jets promoted to high ET jets
- Multi-jet events ? single jet events
- Temporary solution
- PS or DPS
- Long term solution
- Modification of the L2 calorimeter trigger
Project approved on September 14th 2006 As vital
for preserving and improving the sensitivity for
Higgs searches
3System Overview
- Current Pacman
- uses 8-bit resolution for tower energies
- finds a seed (gt 3 GeV) from lowest ??
- attaches any shoulder (gt 1 GeV)
- defines cluster centroid as lowest ??
- How can we improve ?
- Calorimeter trigger tower information available
to the L2 Decision with 10 bit resolution - New algorithm in L2 CPU
- Jet clustering (cone based)
- Met calculation
- Install parallel L2CAL hardware path
- to be tested in fully parasitic mode
4Hardware Configuration
Muon XTRP L1 SVT XCES
Global L2 Decision
New Hardware path
- Set of Pulsar boards to
- receive,
- preprocess
- merge
- send to L2 decision CPU
8-bits
10-bits
Calorimeter
5Pulsar Setup
- SERIALIZER
- Filar boards are interface between SLINK and PC
- RECEIVERS
- Mezzanine cards are interface with LVDS cable
7 ?s (132 towers X2)
144 cables
Pulsar Merger
144 cables
12 ?s (312 towers)
SLINK 32-bit _at_ 40 MHz
6Clustering Software
- Ouline of clustering
- Orders trigger towers in ET (using min ET 0.125
GeV) - Cluster ET in code around the seed ET gt ET,MIN
GeV ( ?R 0.7) - Flag towers as used
- Sort clusters in ET (max 20)
- Cluster ? and ? are weighted by ET
7Clustering Pass
Pass 0 Seed 2 GeV
Pass 0 Shoulder 1 GeV
Pass 0 One tower per cluster One tower per cluster
Pass 1 Seed 8 GeV
Pass 1 Shoulder 7.5 GeV
Pass 2 Seed 3 GeV
Pass 2 Shoulder 1 GeV
Pass 4 (alpha) Seed From Pass 1 Had/Em lt 0.125
Pass 4 (alpha) Energy from Pass 0 Had/Em lt 0.125 Central 1? wedge ?1 ? wedge Plug 3 x 3 box No crossing in ? Energy from Pass 0 Had/Em lt 0.125 Central 1? wedge ?1 ? wedge Plug 3 x 3 box No crossing in ?
- New Clustering Algorithm
- Pass 0 not changed
- Pass 1 replaced by Pass 4
- Pass 2 cone based NEW
- Pass 4 not changed
- New Isolation Algorithm
- Starts from the seed position
- Finds iso 3x3 and (4)3x4
- Uses the smallest of the 5 sums
- Calculated for Pass 1
8FER Monitoring put
- Modification of the FER code
- Needed for backward compatibility
- Energy of towers retrieved from Dirac and print
out in TC2D format - New temporary bank TUPD
- Will replace the TC2D from DCAS when DCAS removed
- Monitoring of the cable occupancy and real E vs
simulated one
G.Cortiana
9Timing performances
?t between L1A and the arrival of first fragment
to the L2TS
- Optimization
- Hardware
- Pulsar setup
- Data volume reduced by sending 3 20-bit towers in
2 32-bit words - Exploit LUT for tower position
- Other possibilities, zero suppression
- Software
- New PCs (2-dual core AMD opteron processors
Gentoo 2006.1 gcc 4.1.1) - Threaded software
- Clustering 5 ?s (on demand)
D.Krop
Low luminosity data Unbiased (previous) events
10Commissioning
- Hardware
- All in place and tested
- Interface Software
- Code for decision interface, finalized
- ready to build trigger tables for L2Cal
- Clustering Software
- Finalized
- FER Monitoring
- Almost finalized
- System successfully run in parasitic mode
- at the end of store and in between stores
- Next, compare decision between main and upgraded
system - (Pass 0 and Pass 4 expected to be identical)
11Summary
- The L2Cal upgrade is close to completion
- (scheduled by end of May)
- Cone based clustering is performed at L2
- MET is calculated with 10 bit full resolution
- With the new L2 Calorimeter,
- better estimate of MET
- better estimate of the cluster E location
- improve efficiency ? improve acceptance
- rate reduction in jet/met based trigger
- Could implement dijet mass cut, angular cuts, SVT
matching, isolation, etc.
M.Schmidt