forward trigger upgrade and auau pattern recognition v. cianciolo, d. silvermyr forward upgrade...
DESCRIPTION
3 AuAu Efficiency Loss J/ efficiency in central AuAu events is ~20 (30)% for the North (South) Arm Reconstruction code still under development, but ~50% of clusters have contributions from >1 track. The statistical improvement cannot be much better than x3. –For most measurements it is hard to justify large expenditures for a “marginal” improvement. –For low-statistics, low S/N signals (J/ in central AuAu collisions) this improvement can be crucial (e.g., by reducing a 3-year measurement to a 1-year measurement). Also a systematic effect. –When the efficiency is low any error on our calculation of that efficiency (due to incomplete simulation of real- life effects) is magnified in the cross section determination. North Arm J/ EfficiencyTRANSCRIPT
Forward Trigger Upgrade and AuAu Pattern Recognition
V. Cianciolo, D. SilvermyrForward Upgrade Meeting
August 18-19, 2004
2
Motivation• This group is proposing to upgrade the
Muon Arms to increase their trigger rejection power in pp collisions.
• We suffer from significant inefficiencies in central AuAu events.
• Can we solve both problems at once?
3
AuAu Efficiency Loss• J/ efficiency in central AuAu
events is ~20 (30)% for the North (South) Arm
• Reconstruction code still under development, but ~50% of clusters have contributions from >1 track.
• The statistical improvement cannot be much better than x3.
– For most measurements it is hard to justify large expenditures for a “marginal” improvement.
– For low-statistics, low S/N signals (J/ in central AuAu collisions) this improvement can be crucial (e.g., by reducing a 3-year measurement to a 1-year measurement).
• Also a systematic effect.– When the efficiency is low any error
on our calculation of that efficiency (due to incomplete simulation of real-life effects) is magnified in the cross section determination.
North
Arm
J/
Effi
cienc
y
4
How Can The Trigger Upgrade Help? Reduce Occupancy
• Assume 1-degree -slices (based on discussions w/ Wei) • We set a goal of 0.5% occupancy and find we need 28 (20) -pads
per -slice for station 1 (station 3).• Plots show occupancy (vs. ) South Arm MuTR Stations 1 and 3 for
200 AuAu “central” events.– Minimum bias events scaled up by x4. – 24 equal- pads per -slice.– Note: for station-1 equal- pads are probably not the optimum choice,
but that’s not important for now.
(degrees) (degrees)
Occ
upan
cy O
ccup
anc
y
5
Would 2D info really help with AuAu reconstruction?
• We can’t make a iron-clad case now.– Someone would need to incorporate a hypothetical detector into PISA, run some simulations and
re-tool the reconstruction algorithm. – Hugo, Melynda, etc. could comment on the feasibility of this.
• However, consider all hits on station-3 in MB AuAu events. We determine the number of possible partner hits on station-1 based on two cuts:
– () < 23 - 9/15()– () < 23 - 9/15() && < 1– These two cuts are appropriate for 2.5 GeV muons (pmin of interest to heavy flavor)
• We then count the number of combinations per station-3 hit (left) and per event (right).
• The basic contention motivating these plots is that high-strip occupancy leads to stereoscopic ambiguity, thus eliminating much of the segmentation in the orthogonal direction at the pattern recognition stage.
(degrees) (degrees)
Red – -cut onlyBlack – and -cuts
Red – -cut onlyBlack – and -cuts
# C
ombi
natio
ns p
er S
t3
Hit # C
ombi
natio
ns p
er
Eve
nt
6
“Known” Benefits of 2D Info• You really know where the hits are.
– You know how many hits contributed to a cluster.– This must help w/ fitting cluster position.– Deprecate the weight of those clusters?
• Most hits have no potential high-momentum partner – why do any cluster fitting for them?
• You have (many) fewer combinations that need to undergo detailed reconstruction.
• Both seem like great potential for significantly reducing execution time.
7
Downstream Chamber Location?• Collision-related hit density highest in “gap-5” due to splash off DX.• Tight correlation should be lost in “gap-5” due to multiple scattering. For the trigger, go in between station-3 and gap-0.
• Having 2D info in back of the MUID may help with heavy flavor measurements in AuAu and may also be important for W physics, but not from a trigger point of view. They would make it significantly more difficult to falsely extend a track by combining it with a hit from behind.
Gap
0
Gap
1 Gap
2 G
ap
3 Gap
4
“Gap 5”
MuI
D G
ap H
its (2
00 A
uAu
MB
Eve
nts)
8
PC FEE• David points out that there are 90k channels
of PC electronics currently on the shelf (~half in the form of chips, half assembled into Readout Cards (ROCs)).
• Each ROC has 48 channels and 4 trigger bits.– With suggested segmentation a ROC would cover
a 2 degree slice, and each 1 degree slice would have 2 trigger bits (one for low-, one for high-).
• Each FEM reads out 45 ROCs, or 90 degrees.– Need 4 FEMs/chamber; 8 per arm.
9
Connecting it to MUIDLL1• PC not a part of the trigger, but 4 bits per ROC are available.• From previous slide we have 45*4=180 bits per FEM. • This fits on two 6XBCLK fibers.
– Need a New Trigger Board (one per FEM) but the pieces are all literally copies of what is on a MUID ROC.
• An entire arm (16 fibers) would fit onto one Generic LL1 Board
PC FEM8/arm
New Trigger Board8/arm
6XBCLKGeneration
MuMUX6 FPGA
G-Link Daughterboard
BCLK
Data (1)45*2 bits MuMUX6
FPGAG-Link
Daughterboard
Data (2)45*2 bits
Generic LL1 Board
Accepts all fibers for one arm
10
Timing?
• Up to now this talk has had no mention of using timing info.
• If it is really needed I don’t think the PC FEE option works.
• One possibility may be to use MUID FEE with a modified ROC.
• The signals would go through some (new) appropriate preamp and discriminator.
• They would be gated on a signal derived from BCLK with width and delay programmed to gate in only collision-related signals.
• The simplest solution would be to use 96 channels/ROC. Then all the rest of the system could be used unchanged.
48 Channels per PC DMU
BCLK w/ programmable width, delay
Signal i
Signal i+1
Signal i+2
Signal i+3
Signal i+4
Data FIFO
BCLK w/ programmable width, delay
Signal i
Signal i+1
Signal i+2
Signal i+3
Signal i+4
BCLK w/ programmable width, delay
Signal i
Signal i+1
Signal i+2
Signal i+3
Signal i+4
BCLK w/ programmable width, delay
Signal i
Signal i+1
Signal i+2
Signal i+3
Signal i+4
11
Modified MUID FEE• The solution on the previous slide represents 9
FEMs/arm, which is probably somewhat expensive, and is overkill for two reasons:– Bandwidth limits (conversion time and DCM
communicaiton) allow us to have many more signals/ROC.– We don’t need to send all the accepted event bits to the
trigger – only 1/12 with the suggested numerology.• Various limitations suggest a maximum of 288
channels/ROC.– This would reduce the requirements to 2 FEMs/arm.
• Necessary development for this option:– A scheme to get all those signals onto a ROC.
• New transition cards and passthrough backplane.– A plan for processing all those signals on a ROC.
• Board real estate may be a problem, but reduced complexity (no CFD, no complicated timing adjustment circuitry) may solve this.
– A trigger interface board, as described for PC FEE option. • It could live in the MUID FEM crate.
– Relatively minor mods to the FEM firmware.