Study of
TEM errors
Giovanni Petri
University of Pisa/INFN Pisa/SLAC
29 August 2005
Lost in Translation
Introduction
1. Can the LAT produce errors during data acquisition?
–
–
Choice of Error Type
How often does it happen?
–
Under what condition does it happen?
2. What is the impact of these Errors on-orbit?
How does it happen???
A particle hits the tower
Trigger Primitives fired
by subsystems
Readout starts
from the
BOTTOM!!!
TEM collects them, checks 3-in-arow and (if TKR triggers) sends it to
GEM
Calorimeter
GEM opens window on first
trigger type and waits for others to
arrive (Coincidence Window)
After few ticks CW closes and
TAM is sent back to the readout
controller to start readout
Tracker
Glt
Electronics
Module
Tower
Electronic
Module
Tower Subsystems
Overview
TEM CC FIFO Error
– 1 cable stores up to 128 hits
– 8 cables per Tower
– 128 x 8 = 1024 hits per tower
Si Planes
Other Buffer Limits:
– Readout controller: max of 64 hits allowed
– Plane: max of 128 hits allowed (64x2)
– How: Cosmic showers e.g.
– Let’s call an event with FIFO error
“BAD”
FIFO Definition
Cables
Tracker Sketch
How Often?
Statistics Summary
•
•
2 Towers Runs:
• # Runs: 48 (all!!)
• # Register Config: 19
• # Events: 9,564,116
• # Bad Events: 915
4 Towers Runs:
• # Runs: 31 (B type)
• # Register Config: 2
• # Events: 1,084,655
• # Bad Events: 93
•
6 Towers Runs:
• # Runs: 62 (B type)
• # Register Config: 3
• # Events: 15,346,394
• # Bad Events: 1760
•
8 Towers Runs:
• # Runs: 81 (B type)
• # Register Config: 3
• # Events: 22,391,000
• # Bad Events: 2900
Bad Events Rates
• Error Rates for B type Runs:
• 2 Towers: 7.4 · 10-5
» B2: 8.3 · 10-5
» B10: 7.1 · 10-5
» B13: 7.5 · 10-5
• 4 Towers: 8.66 · 10-5
» B10: 8.40 · 10-5
» B13: 8.79 · 10-5
• 6 Towers: 1.13 · 10-4
» B2: 1,15 · 10-4
» B10: 1.09 · 10-4
» B13: 1.20 · 10-4
B2:
Flight Settings
How Often?
• Among 2 Towers runs (not B type)
– 135002057-2103
– Single RC Right/Left
» ER 2 · 10-4
– 135002166-2168
– Single RC R/L + Overlay 10 KHz
» 2,1/2,7 · 10-4
– 135002107
– Only Cal Trigger
» 5.9· 10-3
What’s that?
2 orders of magnitude?
B10: Cal 4 range
B13: Zero Suppression OFF
Can we explain this?
Single RC ER Anomaly
30
40
• RC on one side only:
• Max hits per plane is 64
• RC try to read its entire plane!
Planes
– An event that had 70 hits on a
plane now saturates the plane!
– It’s easier to have more hits on the
same Cable!!
• The factor 2-3 of difference is not so
strange!!
»This can be a first order
explanation!!
Cables
70
Do you believe me?
1.
2.
3.
CAL LE has more probability
to be triggered by high energy
events.
Energetic events have more
probably high hits occupancy
Is it enough to explain the big
difference?
•
No other runs to compare
rates!!
BUT...
Cal Only Trigger Anomaly
ER = 5.9· 10-3
# Events: 3048
# Bad Events:18
The run has few events:
3000 instead of 300,000!
NOT STRANGE!!
We can REPRODUCE that!!!
Cut on CAL LE triggered Events!!
What comes out is ER ~ 3.6 · 10-3!!!
When do Bad Events happen to good people?
Trigger
Hits Occupancy
• Which primitive triggers are
there?
• How are hits distributed?
• Are there odd configurations?
» GemConditionsWord
• You understand “odd” later
• When do they arrive?
» Stay tuned…
» Are there temporal
patterns?
• Is a Bad Event influenced by
the previous one?
» GemDeltaEventTime
You clearly recognize
Eduardo when he’s been
working late in the night
• GemConditionsWord:
– Tells which primitive
triggers arrived in the CW
– Possible combinations:
•
•
•
•
•
•
•
TKR (2)
CAL LE (4)
CAL HE (8)
TKR + CAL LE (6)
TKR + CAL HE (10)
CAL LE + CAL HE (12)
TKR + both CAL (14)
Trigger Topology
B10 runs 2 towers
Bad
Good
•No 8s, 10s, 12s:
•This was expected!!
•Bad Events are “big”!
–High Multiplicity
Trigger Types
We expect that the TKR often
arrives first!
–TKR is big: high probability to
trigger first
Trigger Topology
Trigger Primitives Arrival Times
Chained B10 runs 2 towers
~80%
TKR arrives soon!!!
This is no surprise!
The number of
events goes down
very rapidly!!
Ticks
1 tick = 50 ns
Trigger Primitives Timing
Trigger Topology
• CAL LE should open when
TKR is not the first:
»
CAL LE is faster than CAL HE!
»
# Times CAL LE opens CW is
consistent!!
3 events
Ticks
This is odd!!
Explanation???
Just a case? Remember Log
Scale
Ticks
Trigger Topology
Temporal Correlations
The time between a Bad Event and the
The minimum Delta Time is longer than
previous one is long!!
for Good events
Just low statistics
probably
Good
1000 ticks
Bad
2000 ticks
Event Display
“Recognizable track”
Hits Occupancy
Salt and Pepper
50%
25%
Drittoni
(big straight)
Hits
everywhere
Cal lit up where
“track” arrives
Hits only in
upper layers
No Cal
Cliffhanger
20%
Hits Occupancy
Evt 135002052-268476
Puffettae
1. Qualitatively: you can distinguish the
single layers, one by one, from the
other.
Readout Controllers
2. Hits are only on the borders and are
uniformely distributed.
Characteristic signature: everything’s FULL
Cables
Puffettae
• There is a Puffetta in 6 Towers
Runs too!!
• None found in 4 and 8 towers runs
• It looks exactly the same as the 2
towers one.
135004119-115100
LDF Dump file
Puffettae Dumps
• Looking at Dumps you
find this:
8
1
2
• 0040 (hex) = 64 (dec)
• For every single RC above 1
on every CC.
EventID
CalEneSum
(MeV)
Gem
Word
Tkr
Cal Le
Cal He
Delta
Time
Gem
Discarded
268475
0
2
0
31
31
46600
1367
268476
87766
14
2
0
9
65535
1367
268477
0
2
0
31
31
65535
1367
115099
117
2
2
31
31
10221
3607
115100
351303
14
3
0
2
47043
3607
115101
26
2
0
31
31
56818
3615
2 Towers
6 Towers
Are these energies
consistent with
showers that big?
Puffettae Data
There is no obvious hint of electronics
gone wild!! • Delta Times are long
• GemDiscarded seems reasonable (TOT very long)
Energy (GeV)
From Russia with… CAL!
Asimmetry
Layers
Too good to be true!!!!
CAL says:
“Everything normal pal!! Just big shower!!”
350 GeV measured
All strips hit ⇒ 104 particles in 6 towers
For Ep=105 GeV (for consistency with
the observed rate)
p
350GeV/10000 = 35 MeV per particle
6 towers
LAT
1. At sea level we see 10% of Ep (Tune)
2. From graph 106 e- at sealevel (Tune)
High multiplicity shower of 10 MeV e-
10 MeV particles don’t go through the TKR!!
What happens on-orbit?
• There will be high-energy photons!!
– Will these be high multiplicity events in the TKR?
• Let’s see what MC has to say about this!!
•
•
•
•
Used photons coming from 45°-60° from the vertical axis
Energy of 300 GeV
Searching for behaviours like those observed in FIFO errors
Backsplash?
NO PUFFETTAE!!
300 GeV
Total # MC photons 105
# Triggered Events 13006
#1 Towers Saturated: 38
Conclusions
1. The TKR works (poor me…) too well!!
• FIFO errors are no mistery anymore!!
• We know how to characterize AND understand them!
– Rates are low:
• 1 (lonely) bad guy every 100 thousands!!!!
– No influence on other events!!!
2. Bad Events are consistent with showers
3. High energy photons MC needs to be further
studied
– What about reducing lower layers buffer
bandwidth to improve recon??
TACK
Anders Borgland
Favorite saying:
“I told you NOT my
Camaro!! Now I’m
angry…”
High energy
Muon shower
HIM
ME
Eduardo do Couto
e Silva
…trying to sneak home
early… 1 AM…
Favorite saying: THIS IS TOO COOL!!!!
BACKUP SLIDES
…hic sunt leones…
EM Showers
107 Gev
•
Need ~ 104 particles
•
Total Energy ~ 350 GeV
•
<Ep> ~ 35 MeV
•
Let’s say initial total energy was 105-106 GeV
•
We get at sealevel ~ 106 particles
•
Assume for such initial energy, Freq ~ 2 x 10-4/s
•
The 6 tower data acquisition lasted ~ 1 day
108 Gev
To be in the core area 3.14x4202=5.5x105 m2
Freq = 2 x 10-2/s (~4-8 e/ m2)
To be in a 10 times denser area
Freq ~ 2 x 10-3/s (~40-80 e/ m2)
To be in a 100 times denser area
Freq ~ 2 x 10-4/s (~400-800 e/ m2)
To be in the core area 3.14x4202=5.5x105 m2
Freq = 2 x 10-4/s (~40-80 e/ m2)
To be in a 10 times denser area
Freq ~ 2 x 10-5/s (~400-800 e/ m2)
To be in a 100 times denser area
Freq ~ 2 x 10-6/s (~4000-8000 e/ m2)
~ 16 Puffettae (or like)
Is this consistent with Showers?
Saturated tower:
> 64 hits on each of the
4 bottom planes
(both on x and y)
# Saturated Towers
1
2
3
4
5
6
7
8
6 Towers
Data
78
6
3
1
1
2
/
/
8 Towers
Data
145 11
91
=
3
3
2
1
0
4
169
V.H.E. Cosmic Rays and Air Shower Profile
Step 1) Read off the flux of 107GeV proton rate
proton
Take a proton with Ep=107GeV=1016eV
Observation
Flux is 6.8/E1.75 per cm2, second, steradian
and bin-width of E where E= 107GeV.
We then get,
Flux(Ep=107GeV)=3.8x10-12 /cm2/s/sr
for a bin-width of 107GeV
Step 2) Estimate the lateral distr. of particles
Normalized density of 10-2 /X2 = 10-2 /(30m)2
X=Rad. length of atmosphere=36g/cm2=30m
Distance from the core is about 14X = 420m
Electron Component in Hadronic Shower
Step 3) Estimate the number of electrons in a 107GeV air shower at sea level.
These are shower measured profiles for
105GeV proton. Since there is no
measurement for 107GeV, we assume one
sample profile from these.
This gives highest number of electrons at
sea level: use as an upper limit.
For a 107GeV proton we get
Number of e = 0.2x107 = 2x106
Electrons and Gammas within an EM Shower
Step 4) Calculate electron density per m2
From Step 3) Total number of electrons = 2x106 electrons
From Step 2) Assume they are distributed uniformly in r=14X=420m of the core.
Electron density is then 1.8x10-6 (1/m2) times 2x106 = 3.6/m2
Uncertainty:
a) Fluctuation: Trade-off with frequency. Can give a factor of 10-100?
b) Low critical energy for LAT? Ec=10MeV > 1.5MeV: a factor of 2? (see Figure)
Step 5) Calculate frequency:
From Step 1) 3.8x10-12 /cm2/s/sr
From Step 2) core radius=14X=420m
To be in the core area 3.14x4202=5.5x105 m2
Freq = 2 x 10-2/s (~4-8 e/ m2)
To be in a 10 times denser area
Freq ~ 2 x 10-3/s (~40-80 e/ m2)
To be in a 100 times denser area
Freq ~ 2 x 10-4/s (~400-800 e/ m2)
Horizontal Air Shower? (1/2)
Ref: S. Mikamo et al., ICR-Report-100-82-3 (1982) [Spires]; Lett. Nuovo Cimento 34 (1982) 273
Ordinary air shower initiated by protons and nuclei lose
~all energy for zenith ang. >50 deg.
A different population “hirizontal” shower has been
detected. If LAT is hit horizontally the electron multiplicity
can be much lower.
Step 6) Take horizontal air showers with Ne>104
Intensity = 2 x 10-13 /cm2/s/ster (see the right fig.)
Likely zenith angle = 65, 75, 85 deg. (see the fig. in
the next slide.)
Overburden=1kg/cos(65,75,85deg)=2.4, 3.9, 11.5 kg
= 67, 108, 319 X (the shower hist. may be shorter.)
Typical lateral size: assume to be half the detector
size of the Akeno exp. > radius=20m
Horizontal Air Shower? (2/2)
Ref: S. Mikamo et al., ICR-Report-100-82-3 (1982) [Spires]; Lett. Nuovo Cimento 34 (1982) 273
Step 7) Calculate frequency and electron density
Core area = 1250m2
Electron density = >104/1250 = >8/m2
Freq = 2.5x10-6/s/sr
To be in the core area (1250m2)
Freq ~ 2.5 x 10-6/s (>8 e/ m2)
To be in a 10 times denser area
Freq ~ Prob. of 10 fold fluct. x
2.5 x 10-7/s (>80 e/ m2)
Conclusion
1) Frequency of LAT being within the core radius (~420m for 107GeV) is
high (~1/min) but average electron density is only ~4-8/m2.
2) Electron density probably fluctuate as much as 100 times, but
the product of frequency and multiplicity remains constant for
a given shower energy.
Freq ~ 2 x 10-2/s (~4-8 e/m2)
Freq (x 10) ~ prob. of 10 fold fluct. x 2 x 10-3/s (~40-80 e/m2)
Freq (x 100) ~ prob. of 100 fold fluct. x 2 x 10-4/s (~400-800 e/m2)
3) Guestimate for 108GeV protons:
Frequency 1/100, multi. is 10 times.
Freq ~ 2 x 10-3/s 2 x 10-4/s (~40-80 e/m2)
Freq ~ prob. of 10 fold fluct. x 2 x 10-5/s (~400-800 e/m2)
Freq ~ prob. of 100 fold fluct. x 2 x 10-6/s (~4000-8000 e/m2)
4) Horizontal showers are likely to produce high multiplicity events than
normal showers.