Content uploaded by P. T. Smith
Author content
All content in this area was uploaded by P. T. Smith on Dec 10, 2018
Content may be subject to copyright.
Nuclear Instruments and Methods in Physics Research A 414 (1998) 466—476
A Cockcroft—Walton base for the FEU84-3 photomultiplier tube
A. Brunner, R.R. Crittenden, A.R. Dzierba, J. Gunter, R.W. Gardner, C. Hamm,
R. Lindenbusch, D.R. Rust, E. Scott, P.T. Smith*, C. Steffen, T. Sulanke, S. Teige
Department of Physics, Indiana University, Swain Hall West 117, Bloomington, IN 47405, USA
Received 23 January 1998; received in revised form 5 May 1998
Abstract
The design and construction of a Cockcroft—Walton-type base for the Russian FEU84-3 photomultiplier are described.
Several unique features are emphasized. Results from the tests of prototypes are presented. (1998 Elsevier Science
B.V. All rights reserved.
PACS: 29.40.Vj; 84.30.-r; 84.70.#P; 85.60.Ha
Keywords: Photomultiplier; Calorimeter; Cockcroft—Walton; Voltage multiplier
*Corresponding author. Tel.: #1 812 855 4554; fax:
#1 812 855 0440; e-mail: ptsmith@indiana.edu.
1. Introduction
Several photomultiplier tube (PMT) base designs
have been implemented which use the so-called
Cockcroft—Walton voltage multiplier circuit to
generate a graded series of voltages [1—3]. This type
of circuit has the advantage of dissipating much less
power than a resistive voltage divider. Minimal
power dissipation is critical in the detectors in
which PMTs are densely packed.
We have designed a PMT base for an electro-
magnetic calorimeter to be used in an experiment
(Radphi) to measure rare radiative decays of the
/meson at the Thomas Jefferson National Acceler-
ator Facility. We have produced 328 of these bases,
whichwereusedinanengineeringrunofthatexperi-
ment. The main features of these PMT bases are:
zThe only input power required is a single 15 V
DC supply.
zThe voltage setting of each base can be controlled
remotely via a daisy-chained serial data line.
zThe voltages can be set individually, in groups, or
all together.
zThe base can produce a self-test pulse propor-
tional to its generated high voltage to make cer-
tain the base is operating correctly.
zThe cathode to first dynode voltage is maintained
at a relatively constant voltage independent of
the cathode to anode voltage.
The main parameters of the design are listed in
Table 1.
0168-9002/98/$19.00 (1998 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 9 0 0 2 ( 9 8 ) 0 0 6 5 1 - 2
Fig. 1. Block diagram of the photomultiplier base.
Table 1
Main parameters of the photomultiplier base
Photocathode voltage 1000 V$2%P2023 V $2%
adjustable in 1 V$3% steps
CathodePfirst dynode voltage 250—300 V
Maximum average anode
current
100 lA
Supply voltage 15—16 V
Switching noise at anode ;0.25 pC rms in a 250 ns gate
Typical power consumption 60 mW
(at photocathode setting of
2 kV)
Maximum power consumption 150 mW
(when ramping to a higher
setting)
Time to send a base command 3.2 ms
Self-test pulse amplitude 250 pCP500 pC
2. Physical description
The base circuit is entirely contained in a cylin-
drical delrin housing, 3.8 cm in diameter by 15.5 cm
in length. A 3.9 cm square aluminum cap serves to
affix the base and photomultiplier to a wall with
3.9 cm holes on 4 cm centers. This wall is located
just downstream of a stack of lead glass blocks,
each 45 cm long with a 4]4cm
2cross section.
A calorimeter of similar construction built by our
group is described in Ref. [4].
The end cap of the housing has holes for two
connectors. A lemo connector outputs the photo-
multiplier anode signal. A 10-pin mass termination
header connects to a daisy-chained ribbon cable.
The clock and data are each transmitted differen-
tially on two pins. Three pins carry the #15 V
supply, and three pins are grounded. Four hexa-
decimal switches, accessible through holes in the
housing, set a 16 bit address for the base.
3. Circuit description
Refer to the block diagram shown in Fig. 1.
A step down regulator converts the 15V to 5 V for
the use by the logic circuitry. A dual comparator
receives the clock and data signals. A programm-
able gate array (PGA) compares the incoming ser-
ial address to the address set on the four switches to
determine if the base should respond to the trans-
mitted address. Eight bits are allotted to address
a row in the stack of detector modules, and eight
A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476 467
Fig. 2. Pulse frequency modulation voltage regulator.
Fig. 3. Conventional voltage multiplier.
468 A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476
Fig. 4. Interstage voltage distribution for conventional voltage multiplier.
bits are for a column address. Single bases,
rows, columns, and all bases can be selected by
means of two additional enable bits in the serial
data stream. An LED flashes when the base is
addressed. Once a base is addressed, a digital—
to—analog converter (DAC) can be updated with
a new high-voltage setting of 10 bits. Alternatively,
the base can be commanded to produce the self-test
pulse output on the bases anode output connector.
This pulse checks that the voltage has been proper-
ly set and can also be used to verify the address
switch setting, the signal cabling, and the data ac-
quisition ADC.
The first step in producing a suitable voltage
distribution for the PMT is to generate 175 V from
the 15 V input power with a 14 stage Cock-
croft—Walton chain. This voltage is regulated by
a pulse frequency modulation scheme, shown in
Fig. 2. A fraction of the output voltage is compared
to a reference voltage. If the output voltage is too
low, the clock pulse drives the input of the multi-
plier. If the output voltage is correct or too high, the
pulse is skipped. This type of regulation is very
efficient since it simply coasts once the desired
output voltage is reached and then needs to pulse
only at a rate sufficient to supply any leakage and
output currents.
The 175 V supply is used as the input to a second
Cockcroft—Walton chain, which feeds the dynodes,
focus electrode, and cathode of the photomultiplier.
Previously reported circuits have driven the multi-
plier chain from the anode end alone, as shown in
Fig. 3. This results in an interstage voltage distribu-
tion that decreases with each stage. A typical distri-
bution for such a multiplier with pulse frequency
regulation of the cathode voltage is shown in Fig. 4.
A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476 469
Fig. 5. Voltage multiplier with split drive.
470 A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476
Fig. 6. Interstage voltage distribution for multiplier with split drive.
The stage-to-stage voltages are lowest at the cath-
ode end of the photomultiplier leading to a low
photoelectron collection efficiency, especially in
the presence of magnetic fields. To remedy this, the
drive capacitor string is split into two parts,
as shown in Fig. 5, which keeps the interstage
voltages high at both the anode and cathode ends,
as shown in Fig. 6. (A similar effect is achieved
in the more common resistive divider bases by
including zener diodes at each end.) The cathode
and first dynode voltages are fed back to the pulse
frequency modulators that produce the drive sig-
nals for the two ends of the Cockcroft—Walton
multiplier string. The sixth dynode voltage is
divided to produce a voltage level for the self-test
pulser circuit connected in parallel with the anode
output signal.
4. Construction
Most of the circuitry is contained on a four layer
printed circuit. All components are surface moun-
ted except the 10-pin header and lemo connector.
Bypass capacitors for the last two dynodes and the
selftest pulser circuit are on a printed circuit moun-
ted on the back of the tube socket. Fig. 7 shows the
various assembly stages of the base.
One side of the printed circuit contains the pro-
grammable gate array, address switches, oscillator,
DAC, and 175 V multiplier. This side of the board
will operate by itself and is tested before the second
side is assembled.
The other side of the printed circuit contains the
drivers and high-voltage multiplier. The drivers are
assembled first and tested. Next, the high-voltage
A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476 471
Fig. 7. Photograph of photomultiplier base at various assembly stages.
multiplier string is assembled without the feed-
back resistors. This assembly is tested again and
typically produces 2200 V. If this works, the feed-
back resistors are installed, and the computer con-
trol of the cathode voltage is tested. Finally, the
socket assembly is attached, and the test pulser is
checked.
Since surface leakage current is an exponential
function of humidity, with a slope of one decade per
20% humidity change [5], it is important to test the
high-voltage multiplier under low-humidity condi-
tions. We find that the circuit will operate properly
if the ambient humidity is less than 40%. Higher
humidity levels often cause sparks. Cleanliness of
the assembly is also important. Solder with a water
soluble flux is used. Before each testing step, the
board is cleaned with water in an ultrasonic
cleaner, rinsed with ethyl alcohol, and baked at
150
"
C for 1 h. Once the assembly is working, the
socket assembly is coated with a silicone RTV, and
472 A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476
Fig. 8. Photomultiplier anode output current linearity. Fig. 10. Self-test pulse amplitude versus cathode voltage.
Fig. 9. Photomultiplier gain versus cathode voltage.
the main circuit board is coated with an acrylic
paint. Once coated, the base will operate reliably in
relative humidity levels over 60%.
5. Performance
5.1. Linearity test
Any PMT base will have a maximum anode out-
put current after which the response becomes nonlin-
ear due to changing potentials at the last few dynodes.
Light from a green LED was measured by a photo-
diode and compared to the PMT output current. The
results are shown in Fig. 8, and are linear up to
'100 lA. In our application detector channels near
the beam axis experience a rate of &10 KHz, with
typical signal pulses containing a charge of &100 pC,
yielding an average output current of 1 lA.
If a higher average output current were desired,
the Cockcroft—Walton multiplier chain could be
split into three sections and the voltages on the last
few dynodes could be regulated in addition to the
cathode and first dynode voltages. It might also be
necessary, in that case, to use larger capacitors
and/or a higher switching frequency.
5.2. Pulsed laser tests
Prototype bases were tested in a 25 element lead
glass calorimeter. A 5]5 array of lead glass blocks
was wrapped in an aluminized mylar that optically
A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476 473
Fig. 11. Typical pedestal event distribution (ADC response"
0.25 pC/count).
Fig. 12. Pedestal width versus cathode voltage.
isolated each block. A nitrogen laser excited a scin-
tillator producing light pulses which were sent
through a fiber optic cable attached to a plexiglass
sheet that illuminated the entire front face of the
glass [4].
The bases were exercised in the above setup
under computer control. Tests were conducted at
six different voltage settings within the optimum
operating range (between 1500 and 2000 V) of each
base. The gain of each channel was determined by
estimating the number of photoelectrons from the
widths of the laser response distributions. A typical
gain versus high-voltage plot is shown in Fig. 9.
The selftest pulse height of each base was examined,
as were the pedestal widths. The behavior seen in
these tests is shown in Figs. 10—12.
During long-term runs, light pulse events, base
self-test events, and pedestal events were recorded
every 20 min. The variations of these characteristics
during a long run are shown in Fig. 13.
5.3. Beam test
The first engineering runs of the Radphi experi-
ment used the Cockcroft—Walton bases and took
place at the Thomas Jefferson National Accelerator
Facility’s Hall B. A tagged Brehmstrahlung photon
beam was generated from a 4.045 GeV electron
beam.
The photon beam interacted in an 8% radiation
length berylium target. A double row of scintil-
lators detected the recoil proton for reactions of the
sort cpPX/%653!-p, positioned to optimize for Xbe-
ing a /. The electromagnetic calorimeter was
located 1.1 m downstream of the target. A veto wall
of scintillator paddles just in front of the lead glass
vetoed all events with a charged particle.
The trigger required a recoil proton, no charged
particles in the veto wall, a coincidence from the
photon tagger, and a minimum energy deposited in
the calorimeter. Fig. 14 shows a plot of the two-
photon effective mass for events in which two clus-
ters were reconstructed. Additional software cuts
were applied: '1 GeV total energy was required
in the glass, each cluster was required to be
'0.15 GeV, and the events with clusters within
474 A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476
Fig. 13. Long-term stability of photomultiplier base.
A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476 475
Fig. 14. n0mass plot.
10 cm of the beam hole were eliminated to reduce
the effect of beam halo. The peak centered at the
mass of the n0was used to provide the absolute
energy calibration of the lead glass photon calori-
meter. The n0mass resolution (p), after calibration,
is 12 MeV.
6. Conclusion
We have successfully constructed and used
Cockcroft—Walton multiplier PMT bases in an
electromagnetic calorimeter. Currently, we are
building 400 more of these bases for use in the next
phase of the Radphi experiment. This work was
funded by Indiana University and the U.S. Depart-
ment of Energy.
References
[1] B. Lu, L.W. Mo, T.A. Nunamaker, Nucl. Instr. and Meth.
A 313 (1992) 135.
[2] S. Neumaier et al., Nucl. Instr. and Meth. A 360 (1995) 593.
[3] V. Astakhov et al., Multichannel high-voltage system for
photomultiplier tube arrays, http://www.tsl.uu.se/suk-
hanov/HVSys/Astakhov/welcome.htm.
[4] R.R. Crittenden et al., Nucl. Instr. and Meth. A 387 (1997)
377.
[5] C. Coombs, Printed Circuits Handbook, 4th ed., 1996,
McGraw-Hill, New York.
476 A. Brunner et al. /Nucl. Instr. and Meth. in Phys. Res. A 414 (1998) 466—476