13 Readout Electronics
A First Look
28-Jan-2004
Requirements

Digitize charge seen by each PMT


Provide timing of signal for each PMT


Energy reconstruction
Position reconstruction
Provide trigger for DAQ

Physics triggers






Neutrinos (prompt EM energy, delayed neutron energy)
Backgrounds (to study and subtract)
Muons
Electronic calibration triggers (test pulses)
Source/laser/LED calibration triggers
Random triggers
January 28, 2004
2
J. Pilcher
Comparisons

KamLAND is important reference point




Same reaction channel
Scintillator-based detector
Recent design
But much larger target volume


~20 times larger
KamLAND resolutions

Energy


7.5% / Sqrt[E(MeV)]  2%  5.7% at 2 MeV
Position

25cm  5 cm
– timing resolution 2.0 ns RMS after charge correction
January 28, 2004
3
J. Pilcher
KamLAND Electronics

Berkeley Analog Waveform
Transient Digitizer (AWTD)


For 1325 PMTs (32%
coverage)
Sample every 1.5ns



3 gain ranges (0.5, 4, 20)
Store analog samples in
switched capacitor arrays until
trigger


For signals above 1/3 pe
128 samples deep (200 ns)
10-bit ADC


~15 bit dynamic range
Converts 128 samples in 25s.
January 28, 2004
4
J. Pilcher
Channel Response Characteristics
Charge
(cnts/pC)
Sensitivity
pe
(cnts/pe)
Energy
(cnts/KeV)
Charge
(pC)
Full Scale
pe
(pe)
Energy
(KeV)
High Gain
247
186
55.7
4.14
5.5
18.4
Medium Gain
49.4
37
11.1
20.7
27.5
91.8
Low Gain
6.18
4.6
1.39
166
220
734
0.752 pC/pe
300 pe/MeV
Readout Resolution
Percentage Resolution
100.00
10.00
1.00
0.10
0.10
1.00
10.00
100.00
1000.00
Single PMT Energy (KeV)
January 28, 2004
5
J. Pilcher
KamLAND Signals
128 samples of 1.5ns
3 gain scales
(most events just use 20X
scale)
Gain 1/2
Gain 4X
Gain 20X
January 28, 2004
6
J. Pilcher
KamLAND Vertex Reconstruction

Calibrate timing of individual
PMT channels with variable laser
pulses at center of detector



Time offsets
T vs Q
Measure performance for
physics with sources along zaxis
January 28, 2004
7
J. Pilcher
KamLAND Vertex Reconstruction

Mean reconstructed
position depends on
photon energy

Apply energy
dependent correction
January 28, 2004
8
J. Pilcher
KamLAND Energy Reconstruction


Set gains of PMTs using LEDs
Equalize 1 pe peaks to 184 counts



Must correct for variations in storage capacitors
All signals converted to equivalent photoelectrons
Convert to energy using calibration sources
January 28, 2004
9
J. Pilcher
KamLAND Energy Reconstruction
January 28, 2004
10
J. Pilcher
Fresh look at Readout Electronics

Avoid ASICs if possible (local bias)





Long development time
Not cost effective in small volume
Do not profit from evolution of chips in the
commercial sector
Main advantage size and possibly performance
and functionality
Continued performance growth in commercial
ADCs and FPGAs (PLD)

Popular building blocks for many applications
January 28, 2004
11
J. Pilcher
Fresh look at Readout Electronics

Does one need detailed pulse shape for E and t?

Pulse shape discrimination can resolve photons
from neutrons



Depends on scintillator
Some exhibit this property and some do not
May depend on light collection from target
– Reflections could obscure the effect

Much simpler if one can do shaping of input
signal


Output amplitude proportional to input charge
Can be done with passive elements (no noise
added)
January 28, 2004
12
J. Pilcher
ATLAS TileCal Approach

For ATLAS TileCal 20 ns PMT signals converted
into 50-ns-wide standard shape



Amplitude proportional to input charge
Slower signal can be handled by commercial
ADCs (+40 megasamples per second)
Analysis process fits shape to extract amplitude
and time
January 28, 2004
13
J. Pilcher
Performance of TileCal System
Time reconstruction is excellent
amplitude independent
January 28, 2004
14
J. Pilcher
Alternatives

Use LBNL AWTD



Likely if they join the collaboration
Possibly an updated version
Build a system based on a flash ADC

Eg. Maxim MAX1151





8 bit flash
750 MHz (sample every 1.3 ns)
Power 5.5W each
Need 3 per PMT for dynamic range
Use 40 MHz “system” clock à la LHC



Easy to distribute on optical fiber if LHC hardware used
Generate local vernier clock synced to system clock
Tale 16 samples for every 25 ns period of system
January 28, 2004
15
J. Pilcher
Alternatives

Build integrating system as in TileCal

The next steps



Test LHC system reading out scintillator test cell
Look at pulse shape discrimination with test cell
Continue to think about electronics

Trigger
– Can it be derived from digital data, thereby avoiding a second
signal branch

Consult with Harold
January 28, 2004
16
J. Pilcher