Mozaic trigger system for transverse momentum physics

1
Mozaic trigger system for transverse momentum
physics

• G.Vesztergombi, A.Agocs, B.Bozsogi, A.Fulop
• CBM Collaboration Meeting
• GSI Dubna
• 13-18 Oct, 2008

2
(No Transcript)
3
Beier (1978)
NA49 (CERN) results at 158 FODS (IHEP) at 70 GeV
4
Y production
Due to high mass ( 9.5 GeV/c2) two high pT
particle in leptonic decay
pT gt 3 GeV/c
High selectivity for high pT pair even without PID
5
(No Transcript)
6
Special requirements for Y-gt ee- and high pT
Extremely high intensity -
Pile-up Segmented multi-target -
Relaxed vertex precision Straight tracks
- High momentum tracks
DREAM 109 interactions/sec
7
High transverse momentum means high 3-momentum
Illustration for mid-rapidity at sqrt(s) 7 and
14 GeV
( )
( )
( )
E
g gb 0
Lab
p

long
gb g 0
0
0 0 1

• 3.5 1.5
  5.25
• 2.0
  7.0
• 3.0
  10.5

• 90 6.6 1.5
  9.9
• 2.0
  13.2
• 3.0
  19.8

8
High ( gt 5 GeV/c ) momentum ? Straight track
x,y cm
PxPy 3 GeV/c Pz 10 GeV/c
PxPy 1 GeV/c Pz 5 GeV/c
z cm
9
Straight tracks from main vertex
Tracks with pxz gt pmin remains within the
df(pmin) wedge
1-dim Hough transform f - histogram
f , df
N( fi) in df-bins
Correct for bin boundary crossing
M(fi) N(fi) N(fi1 ) in 2df-bins
SELECTION on pxz
10
3 dimensional scheme
k3
i1
k2
i
k1
j
j1
Mosaic cells in plane k M(i,j,k)
(i,j) Corridor contains M(i,j,k), M(i,j1,k),
M(i1,j,k), M(i1,j1,k) k1,2,3
11
MAPS vs Hybrid
Vertex resolution dz 1 mm, dx,dy 0.05
mm High intensity radiation hard
Practical 4 2 3 9 planes ( 4 Hybrids 5
strips) Selectivity depends on the availability
of TOF information
12
Silicon planes
4 hybrids 2
3
strips
(XX,YY,ZZ)
a
delta









s sqrt(XXXXYYYY) - delta
2
4
6
Basic planes
Sagitta
10-20 cm track sections are practically
straight fractals
13
New algorithm
Parallel processing CORRIDOR corNum
Basic planes 2 (x2,y2,z2) pixel ,
4 (x4,y4,z4) pixel, 6 (x6,z6) strip
Straight tracking in 2 and 4 planes in space
gt (mx,bx) and (my,by) Approximation
starting direction is given by (mx,my)
TUBE definition x-tube xi mx(zi-z2) bx
parabol(x6,z6,zi) /- deltaxi y-tube yi
my(zi-z2) by /- deltayi
14
4-5 GeV/c pT gt 1.0
15
7-8 GeV/c pT gt 1.0
16
9-10 GeV/c pT gt 1.0
17
15-17 GeV/c pT gt 2.5
18
20-40 GeV/c pT gt 2.5
19
Acceptance
-

pT
Npoint9
pxz
20
Selecting only tracks with pTgt2.5 GeV/c
Npoint9
21
Pileup
Fixed pT-cut at 1.8GeV/c No pileup
Tracks with ptin gt 1.8 1136
ptin lt 1.8 but ptrec gt 1.8 430 Npileup 10
Tracks with ptin gt 1.8 1136
FAKE and ptrec error 1 430 Npileup
100 Tracks with ptin gt 1.8 1136
FAKE and ptrec error 28
430 Npileup 1000 Tracks with ptin gt
1.8 1136 FAKE and ptrec error
464 430
No LOSS of GOOD tracks due to pileup (exhaustive
search!!!)
Number of FAKE triggers even in 1000-fold pileup
is lt 50
22
Pileup cont.

• pT dependence No pileup
  1000-fold pileup
• 1.8 GeV/c 430/1136 464430/1136
• 2.0 GeV/c 312/704 363312/704
• 2.2 GeV/c 208/453 306208/453
• 2.4 GeV/c 151/301 265151/301
• 2.6 GeV/c 103/213 205103/213
• 3.0 GeV/c 52/154 168 52/154

The FAKE/GOOD ratio is moderately increasing
with pT
23
Deviations within the tube
dx
Chargedx
24
Y-deviations
25
Difference between exact direction and mx
pxz
pT gt 2.5 GeV/c
pT gt 1.0 GeV/c
26
DAQ scheme
27
Mozaic DAQ system
Two separate systems PRETRACKING network
Pixel 2 , 4 Strip 6x
TRACK-QUALITY TUBE network Pixel 1, 3
Strip5x, 5y, 6y, 7x, 7y, 8x, 8y, 9x,
9y In each network parallel CORRIDOR
processors CorID corNUM Number of CORRIDOR
processors ndxndy Data select their routes
according to plane number and corNUM In plane
zi track-hit xi,yi calculates its corridor
address corNum idxndy idy
28
Corridor processors
OLD system consecutive cycling on all
planes If only 2 points per plane number of
cycles 2(425) 214 16384 NEW system
cycling only on 3 planes (for pixels x and y
has common cycle) If only 2 points per plane
number of cycles 2(21) 23 8 The
PRETRACKING is producing a list containing
corNUM, x1,x3,x5,x7,x8,x9,
y1,y3,y5,y6,y7,y8,y9 There is NO PROCESSING TIME
in the TRACK-QUALITY TUBE network because It is
only an ASSOCIATIVE memory which provides
YES/NO. The gain in processing time (if only 2
points per plane) 211 2048-fold
29
Mozaic trigger for low pT
30
(No Transcript)
31
END
32
For discussion
33
Physical mosaic cells can be different from
logical cells. Hybrid (pixel) logical cells
may be created by software Strip planes
hardware should be harmonized Corridors can be
filled by hardware or software Number of
processor can be less than number of
corridors Corridors processing speed can be
very fast if they are narrow Corridors can be
arranged hierarchically for processing order
34
Plong GeV/c
40
50
10
20
30
0
0 0 0 1 0 0 0 0
1 0 0 1 0 0 1 0 0
2 0 0 0 3 22 5 0 0
0 0 0 0 0 0 0 1
0 0 0 0 0 0 4 0 48
73 44 7 1 0 0 0 1 1
0 0 0 0 1 0 0 0 0
0 34 81 90 81 35 0 0 0
0 0 0 0 1 0 1 0
0 0 0 2 28 83 92 87 89 81
25 0 1 0 2 0 1 0
0 0 1 0 0 0 17 69 93 89
91 86 87 78 16 1 1 0 0
1 0 0 0 0 0 0 4 40
86 90 90 85 91 87 86 63 13 1
0 1 0 0 1 1 0 1 1
18 72 94 91 90 89 85 87 87 83
42 9 0 1 0 1 0 0 0
0 9 52 88 90 88 85 86 93
93 91 89 77 50 8 0 0 3 0
0 1 0 28 72 88 90 90 91
85 90 86 91 90 84 80 37 0
0 0 0 0 0 8 51 74 92 92
87 91 86 91 95 90 83 93 88 75
26 0 0 0 1 1 37 62 89
87 87 88 84 85 92 96 92 91 90
90 89
2
Ptrans GeV/c
4
6
72 Points efficiency