Content uploaded by Christopher Scott Olsen
Author content
All content in this area was uploaded by Christopher Scott Olsen
Content may be subject to copyright.
Abstract—In this paper a low-noise preamplifier for liquid-
xenon ionization detectors is presented. The preamplifier has a
cold front-end to be operated in liquid xenon at ~160 K and a
warm part installed 15-cm apart and working at room
temperature. It features an unusually high sensitivity of
~70 mV/fC. This specification is required because the net charge
delivered in liquid xenon per unit energy is one order of
magnitude lower than in silicon or germanium detectors. The
preamplifier consists of a JFET-input charge-sensing stage with a
feedback capacitance as low as 0.2 pF and a low-noise gain stage
with differential output stage, able to drive 50Ω terminated
cables. With a detector capacitance of 33 pF the preamplifier
features an Equivalent Noise Charge of ~110 electrons r.m.s. at a
shaping time of 6 µs, and a risetime of ~75 ns.
I. I
NTRODUCTION
In the context of R&D for a Liquid Xenon Time Projection
Chamber (LXeTPC) as a Compton telescope for MeV gamma-
rays, we have developed a low noise preamplifier with cold
front-end inside liquid xenon. The LXeTPC approach shown in
Fig. 1 exploits the excellent scintillation and ionization
properties of ultrapure liquid xenon, in a temperature range of
about 165-180K [1-2]. With a pair-creation energy of 15.6 eV,
LXe is an efficient ionization medium and provides more
charge than other noble liquids. Yet, the charge signals of
~1 fC/100 keV are also significantly smaller than in
semiconductor detectors. Operation as a Compton telescope
requires both excellent energy resolution and low thresholds.
Since the LXeTPC uses a cryogenic liquid as detector medium,
it is usually made of a detector vessel and a cryostat. As a
result, any measured charge signals must pass feed-throughs
and typically substantial cable length before reaching a charge-
sensitive amplifier (CSA). Placing a first amplification stage
inside the xenon liquid has several potential benefits: it places
the amplification stage in immediate proximity to the sensing
electrodes, removing the input capacitance of long leads, and
reducing the potential for noise pick-up. Furthermore, the
operating temperature of the LXeTPC is near the noise
optimum for Si-based electronics, hence reducing thermal
noise in the preamplification without otherwise impeding the
A. Pullia and F. Zocca are with the University of Milan, Department of
Physics and Istituto Nazionale di Fisica Nucleare, Milano, Italy.
U. Oberlack, C. Olsen, and P. Shagin are with the Dept. of Physics &
Astronomy, Rice University, Houston, TX, USA.
device performance. On the other hand, the location inside an
LXeTPC also imposes significant constraints on this first
amplification stage: insertion of electronegative impurities into
the xenon liquid must not exceed the ppb level; the electronics
and connections must withstand thermal cycling from a mild
“baking” temperature of ~400 K under vacuum to the low
operating temperature inside LXe; and the introduction of
excess heat must be reduced to a minimum, in order to avoid
local formation of gas bubbles and to reduce heat load in
general.
We present an approach that places the first-stage Junction
Field Effect Transistor (JFET, Philips BF862) of an ultra-low
noise charge sensitive amplifier on a board next to the sensing
electrode inside LXe, together with the surface-mount resistor
and capacitor feedback elements, as well as a capacitor for
charge calibration with a testpulse. As shown in Fig. 2 this
"cold stage" is connected with the remainder of the CSA
through twisted pair cables: one pair each for feed-back,
source/drain of the JFET, and the testpulse input. The main
CSA stage is placed in a box directly connected to a detector
feed-through. The Milan group developed a custom CSA to
match the requirements of the LXe detector. In particular, the
A Cold Low Noise Preamplifier for Use in
Liquid Xenon
A
. Pullia, F. Zocca, C. Olse
n
, P. Shagi
n
, U. Oberlac
k
Fig. 1. Scheme of a LXeTPC detector. The ionization charge drifts through
liquid xenon towards the anode, passing by a Frisch grid. The photomultipliers
PMT 1 and PMT 2 on top and bottom are used to detect xenon scintillation
light.
gain was brought to a level of ~70 mV/fC into a differential
(100 Ω) output, to match the ±1 V differential input range of
an FADC board sampling the signal. The Rice group designed
the cold stage.
In this paper we discuss the preamplifier circuit structure
and performance. A complete testbench characterization is
presented. The integration of this system in the Rice XeSpec
detector is in progress and will be discussed elsewhere.
II. PREAMPLIFIER STRUCTURE
In Fig. 3 the schematic diagram of the realized circuit is
shown, and the components on the “cold” and “warm” boards
are put into evidence. It basically consists of a negative-
feedback charge-sensing stage, using C
F
, R
F1
, R
F2
, R
F3
, as
feedback network, followed by a single-ended to differential
gain-stage built around operational amplifiers U
1
and U
2
,
configured as an “instrumentation amplifier”. The charge
sensing stage is in practice an active integrator with a
continuous reset time-constant given by τ
F
=C
F
×
(R
F1
+R
F2
+R
F3
), or 600 µs nominally. The forward path of the
integrator has a simple folded-cascode structure (Q
1
and Q
2
)
with an active load (Q
3
).
Input “HV” provides low-pass filtering for the detector
high-voltage bias by means of R
H2
and C
H
. Input
“DETECTOR” is therefore biased to the high voltage provided
by “HV”. The detector signal current entering pin
“DETECTOR” is AC coupled to the preamplifier and reaches
the preamplifier’s input, or the JFET’s gate, through CR
network C
B
, R
H1
. Input “TEST” is used to inject a detector-like
calibration testpulse to the preamplifier’s input pin.
The response of the circuit to a detector charge signal Q is in
first approximation a step function with amplitude
+×−=
9
8
1
1
R
R
C
QV
F
OUT
, (1)
9
10
/
1
R
R
C
QV
F
OUT
×=
. (2)
Note that as the detector electrode collects negative charge
V
OUT
is positive and V
/OUT
is negative. In a better
approximation the step shows a slow exponential decay with
time constant τ
F
due to the continuous-time discharge of C
F
through the series of resistors R
F1
, R
F2
, and R
F3
. The reason
why a series of three resistors was used rather than an
individual one will be discussed in the next section. The
preamplifier overall sensitivity is derived from (1) and (2), or
Q1
BF862
Q2
BFT92
Q3
BFR92
Q4
BFR92
R1
1K
R2
12K
R3
15K
R4
12K
R5
15K
C4
1u
C5
1u
R6
12K
C6
2.2u
R7
22K
R11
22K
R8 7.5K
R9
470
R10 8.2K
R11
47
R12
47
CF
0.2p
RF1
1G
RF2
1G
RF3
1G
CT 0.2p
RT
47
12 V
-12 V
U2
AD829
U1
AD829
CB
470p
CH
1n
RH2 10M
RH1
400M
HV
DETECTOR
TEST
Components on
the “cold” board
“drain”
“feedback”
Components on
the “warm board
OUT
/OUT
Fig. 3. Simplified schematic diagram of the charge sensitive preamplifier.
“cold” part of the
preamplifier
“warm” part of the
preamplifier
twisted pair cable
interconnection
Fig. 2. Photograph of the “cold” and “warm” parts of the preamplifier, as
linked by twisted pair cables.
++=
−
9
10
9
8/
1
1
R
R
R
R
CQ
VV
F
OUTOUT
(3)
If the differential output signal is transmitted to a remote
FADC through a terminated cable, the sensitivity (3) becomes
2
1
1
1
9
10
9
8
++=
∆
R
R
R
R
CQ
V
F
FADC
(4)
as the signal gets splitted by half across the termination
resistors.
In order to implement the large required sensitivity of
~70 mV/fC a low-value feedback capacitance C
F
of 0.2 pF has
been used, and a relatively large gain of ~17 has been
implemented in each branch of the differential output stage.
We used the AD829 opamps in the differential output stage
because it has an extremely low noise, a large output voltage
swing, and a relatively large bandwidth even at high gains.
Table I shows some relevant parameters of the AD829. Note
that the effective value of C
F
including the stray components,
is expected to be in the 2.5 to 3 pF range.
The so-obtained high sensitivity poses a few design issues,
some of which are evident while some others are more subtle,
as shown in the following section.
III. DESIGN ISSUES AND EXPERIMENTAL RESULTS
A first issue related to the high circuit gain is the potential
high DC offset at the preamplifier output, enhanced by the
presence of a negative DC bias voltage at the input JFET’s
gate. As shown in Fig. 3 we addressed this issue by AC
coupling the output stage by means of C
6
, R
7
and by using
resistor R
11
for compensating the bias currents of the Analog
Devices AD829 operational amplifiers. We so obtained a DC
offset for the differential signal V
OUT
−V
/OUT
of the order of
-50 mV, as shown in Fig. 4. Fig. 4 also shows the exponential-
decay transient with an observed time constant of ~800 µs.
Fig. 5 is a blow up of Fig. 4 in the microsecond time scale
showing the ~75 ns rise time of the preamplifier, which is to be
regarded as a good performance when using a detector
capacitance as large as 33 pF, a feedback capacitance as low as
of 0.2 pF and while driving terminated 50 Ω output cables.
Such a rise time is in any event adequate for this application.
A second more subtle issue related to the high circuit gain is
its extreme sensitivity to any voltage ringing directly or
indirectly coupled to a few preamplifier’s sensitive points in
the charge sensing stage. In particular the circuit shows a
tendency to pick up any ringing of the positive power-supply
and of a few imperfect ground points. Even if the power
supply is strongly filtered it inevitably rings slightly when the
output signal swiftly rises. The ringing becomes apparent when
the preamplifier outputs are connected to terminated cables,
i.e. to low-value equivalent resistances, and so a remarkable
current has to flow from the power supply to ground in a short
while. In our case the power-supply ringing is mostly picked
up through resistance R
1
, and the ground ringing is mostly
picked up by capacitance C
4
. As a result the output waveform
appeared distorted when using the terminated cables. The
distortion disappeared by disconnecting the terminated cables.
We could greatly reduce this undesired effect by: (a) using
separate filters for the +12V power supply used in the charge
Differential signal
Single ended positive
Single ended negative
Fig. 4. Oscilloscope screenshot showing the preamplifier response to a test
signal in the ms time scale. X-axis: 2ms/div, Y-axis: 200mV/div. A 50Ω load
is present on both oscilloscope channels. The DC offset of the differential
signal is <50mV. The exponential-decay time constant is of ~800 µs.
Differential signal
Single ended positive
Single ended negative
Fig. 5. Oscilloscope screenshot showing the preamplifier response to a test
signal in the µs time scale. X-axis: 200ns/div, Y-axis: 200mV/div. A 50Ω load
is present on both oscilloscope channels. A risetime of ~75 ns is obtained on
the differential signal.
T
ABLE I
R
ELEVANT PARAMETERS OF AD829
Input voltage
noise @ 1 kHz
Input current
noise @ 1 kHz
Bandwidth of inverting or non
inverting amplifiers (1kΩ load)
Output swing into
a 150Ω load
1.7 nV/√Hz
1.5 pA/√Hz
65 MHz @ G=-9 or G=+10
55 MHz @ G=-19 or G=+20
39 MHz @ G=-24 or G=+25
±3 V
sensing stage and that used for the operational amplifiers, and
(b) re-laying out the ground via of C
4
in such a way to separate
it physically as much as possible from the ground vias used for
the opamps power-supply filters.
Special care has also been devoted to the connection
between the “cold” and “warm” preamplifier boards. As shown
in Fig. 3 such connection consists basically of the “drain” and
the “feedback” wires plus an implicit ground link. Each of
these links was realized with a twisted pair cable. The two
wires of the first pair are “drain” and “ground”, the two wires
of the second pair are “feedback” and “ground”. The “ground”
wires provide an easy path for the high-frequency return
currents, which greatly helps stabilize the feedback and reduce
the Electro Magnetic Interferences (EMI) [3].
We are now to discuss a few other issues concerning the
cold preamplifier board.
As can be seen in Fig. 3 and Fig. 6 the feedback resistor is
realized in practice as the series of three 1 GΩ resistors. This
approach yields a reduced intrinsic stray capacitance. In fact
the stray capacitance of a chip resistor depends mostly on its
geometry and is of the order of 150 fF for the 0805 size. Use of
three resistors in series makes the intrinsic stray capacitance
three times as low, i.e. of the order of 50 fF. Minimization of
all stray capacitances is evidently important in our circuit,
which needs a feedback capacitance as low as 0.2 pF to
implement a high charge sensitivity.
Finally, we carefully studied and optimized the grounding of
the “cold” preamplifier board. On the one hand a ground plane
under the JFET and feedback devices greatly helps reduce the
low-frequency disturbances, or microphonism, as it works as
an electro-magnetic shield against all stray capacitances which
could enhance and modulate the effective feedback
capacitance value. For example the stray capacitance between
the two traces/pins used for the feedback capacitor depends in
part on the physical distance of the “cold” preamplifier board
from the grounded mechanical parts in immediate proximity,
and is sensitive to any mechanical vibrations. It should
therefore be shielded. On the other hand a ground plane under
the three feedback resistors R
1
, R
2
and R
3
creates two new
capacitive paths from the intermediate connection points
between these resistors to ground. These stray capacitances
yield an apparent distortion of the exponential-decay shape of
the preamplifier response, as can be shown analytically and
experimentally. So, we found that the optimal solution consists
of laying out a ground plane under all input/feedback devices
but the feedback resistors.
IV. NOISE MEASUREMENTS
We eventually installed the “warm” part of the preamplifier
in the hermetic shielded box shown in Fig. 7. We used a
capacitance of 33 pF to simulate the detector and arranged the
“cold” part of the preamplifier with or without a ground plane
underneath. In these testbench measurements the “cold”
preamplifier could not be cooled to cryogenic temperatures,
and was operated at room temperature. For signal-to-noise
optimization we used a Gaussian shaper, Ortec 572, with
selectable shaping time in the 0.5 to 10 µs range. In the first
tests we used 13-cm long twisted pair cables to connect the
“cold” and “warm” parts of the preamplifier and housed both
parts in the same shielded box. The “cold” board had no
ground plane underneath. The CSA showed a sensitivity of
~62 mV/fC and an excellent noise performance with a
minimum of ~110 electrons achieved for a shaping time of
6 µs, as shown in Fig. 8, which is adequate for the foreseen
applications. Cooling of the cold part of the preamplifier will
help reduce the noise further. No significant difference in the
noise performance was observed using a grounded “cold”
preamplifier board. However as expected the presence of a
ground plane definitely made the system less sensitive to
microphonism. In Fig. 8 the ENC values have been fitted and
the principal noise contributions have been disentangled using
the procedure described in [4,5]. The dominant contribution to
the ENC in the 2 to 10 µs shaping-time range is the flat one,
due to the 1/f noise of the JFET and the dielectric noise of the
FR4 board where the JFET is mounted [6]. This noise
contribution could be reduced using a higher quality dielectric,
like teflon or alumina [7]. Cryogenic cooling will also help
reduce this noise component.
Fig. 7. Photograph of the “warm” part of the preamplifier as mounted in its
hermetic shielded box. The two twisted pair cables used to connect the
“warm” to the “cold” part pass through the big cylindrical connector.
Fig. 6. Photograph of the “cold” preamplifier board. The feedback resistor can
be seen in the center-left region of the board as realized as the series of three
chip resistors.
We eventually placed the grounded “cold” part of the
preamplifier in a separated box using the hermetic feed-
through shown in Fig. 7. In this case the overall length of the
connection between the “cold” and “warm” parts of the
preamplifier was 15 cm. The CSA showed a sensitivity of
~75 mV/fC and again an excellent noise performance with a
minimum of ~110 electrons achieved for a shaping time of
6 µs, as shown in Fig. 9. As can be seen in this case the charge
sensitivity is higher and the noise is just a little bit higher. The
increase in the charge sensitivity depends on the shielding
effect of the ground plane, which remarkably reduces the stray
capacitance between the feedback capacitance traces. The little
noise increase could depend on the absence of an effective
shielding along the connection between the “cold” and “warm”
parts. Also in this case the dominant noise contribution is
related to the 1/f noise of the JFET and to the dielectric noise
of the FR4 board where the JFET is installed.
V. CONCLUSION
We designed, realized, and testbench characterized a low-
noise preamplifier for use in Liquid Xenon Time Projection
Chamber for MeV gamma-rays detection. The Charge
Sensitive Amplifier showed a high sensitivity of ~75 mV/fC, a
risetime of ~75 ns and an excellent noise performance with a
minimum of ~110 electrons achieved for a shaping time of
6 µs. Work is in progress to integrate this electronic system in
the Rice XeSpec detector.
VI. ACKNOWLEDGEMENTS
The authors acknowledge R. Bassini and C. Boiano for
technical assistance. This work was supported in part under
NASA grant NNG05WC24G.
VII. REFERENCES
[1] E. Aprile, A. Curioni, V. Egorov, K.L. Giboni, U. Oberlack, S. Ventura,
T. Doke, K. Takizawa, E.L. Chupp, P.P. Dunphy, “A Liquid Xenon Time
Projection Chamber for Gamma-Ray Imaging in Astrophysics: Present
Status and Future Directions”, Nucl. Instr. Meth., vol. A461, pp. 256–
261, 2001.
[2]
U. Oberlack, E. Aprile, “A Study of Liquid Xenon Detectors with
Enhanced Spectroscopy and Time-of-Flight Background Rejection for an
Advanced Compton Telescope”, NASA proposal APRA04-0084-0051,
2004. URL: http://astroparticlelab.rice.edu/
[3] See e.g. H. W. Ott, “Noise Reduction Techniques in Electronic
Systems”, Wiley and Sons, p. 140, New York, 1976.
[4] G. Bertuccio and A. Pullia, “A Method for the Determination of the
Noise Parameters in Preamplifying Systems for Semiconductor Radiation
Detectors”, Rev. Sci. Instr., vol. 64, no. 11, pp. 3294-3298, 1993.
[5] E. Gatti, P.F. Manfredi, M. Sampietro, V. Speziali, "Suboptimal filtering
of 1/f noise in detector charge measurements", Nucl. Instr. Meth.,
vol. A297, pp. 467-478, 1990.
[6] V. Radeka, “Field Effect Transistors for Charge Amplifiers”, IEEE
Trans. Nucl. Sci., vol. NS-20, no. 1, pp. 182-189, 1973.
[7] A. Pullia, G. Bertuccio, “Resolution limits of silicon detectors and
electronics for soft X-ray spectroscopy at non cryogenic temperatures”,
Nucl. Instrum. and Meth., vol. A380, pp. 1-5, 1996.
0.1 1 10 100
10
100
1000
ENC [el]
186.6
152.9
126.5
117.2
111.0
111.1
τ [µs]
0.5
1
2
3
6
10
C
DET
= 33 pF
Shaper ORTEC 572
13-cm twisted pair cables
ENC [ el. r.m.s. ]
Shaping time [ µs ]
Fig. 8. ENC vs. shaping time as measured placing the “cold” and “warm” parts
of the preamplifier in the same shielded box, and using a “cold” preamplifier
board with no ground plane.
0.1 1 10 100
10
100
1000
ENC [el]
193.9
152.4
125.2
119.5
111.7
113.9
τ [µs]
0.5
1
2
3
6
10
C
DET
= 33 pF
Shaper ORTEC 572
15-cm twisted pair cables
ENC [ el. r.m.s. ]
Shaping time [ µs ]
Fig. 9. ENC vs. shaping time as measured using a “cold” preamplifier board
with a ground plane underneath.
