Crosstalk characterization in gmap arrays

218 IEEE ELECTRON DEVICE LETTERS, VOL. 29, NO. 3, MARCH 2008
Crosstalk Characterization in Geiger-Mode
Avalanche Photodiode Arrays
Emilio Sciacca, G. Condorelli, S. Aurite, S. Lombardo, M. Mazzillo, D. Sanfilippo, G. Fallica, and E. Rimini
Abstract—Following our work on Geiger-mode avalanche pho-
todiode arrays, we have recently been dealing with the crosstalk
issue in newly developed dense arrays with a minimum dis-
tance between pixel centers of 84 µm. In this paper, we present
our crosstalk measurement approach, including the experimental
setup and the offline calculation methods. Different characteriza-
tions of the crosstalk probability PCT versus capacitance have
been performed to extrapolate the PCT when no measurement
setup loads the pixels. We also present results regarding the
crosstalk probability versus pixel distance and bias. Moreover, by
adopting a slightly different approach, the probability density de-
cay time has been measured to investigate about crosstalk origin.
Index Terms—Crosstalk probability, Geiger-mode avalanche
photodiode (GMAP) arrays, photon counting.
I. INTRODUCTION
IN THE last few years, there has been a great interest for
the possibility to realize solid-state photodetectors behav-
ing like photomultiplier tubes (PMT), starting from silicon
photomultipliers (SiPMs) fabricated by standard planar silicon
production processing.
Due to their sensitivities down to the single-photon regime,
gains ranging from 105
to 107
, timing performances better than
most PMTs, insensitivity to magnetic fields up to 15 T, power
consumption and bias voltages lower than standard PMTs,
ruggedness, and compactness, SiPM is currently promising
to find widespread use in many applications such as nuclear
medical imaging [1].
Even if all the previous arguments would be enough to
explore this alternative to PMT, it is the lower production cost
of silicon planar technology that attracts the most and may
lead to the push that ultimately enables the realization of this
device. Unfortunately, a few problems have to be first solved;
the most critical one is the crosstalk effect. SiPM linearity can
be worsened when a statistically fluctuating number of pixels is
triggered because of crosstalk coupling [2].
The results presented in our previous paper [3], [4] lead us to
fabricate large-area SiPMs with a high fill factor by shrinking
dead regions around the pixel active area.
Manuscript received October 24, 2007; revised November 28, 2007. The
review of this letter was arranged by Editor P. Yu.
E. Sciacca, S. Aurite, S. Lombardo, and E. Rimini are with the L’Istituto
per la Microelettronica e Microsistemi, Italian National Council of Research
(CNR), 95121 Catania, Italy (e-mail: emilio.sciacca@imm.cnr.it).
G. Condorelli, M Mazzillo, D. Sanfilippo, and G. Fallica are with the
Research and Development Department, Industrial and Multisegment Sector,
STMicroelectronics, 95121 Catania, Italy (e-mail: massimo.mazzillo@st.com).
Digital Object Identifier 10.1109/LED.2007.915373
Fig. 1. Layout and pitch values of a 6 × 6, 20-µm-diameter active-area
GMAP array.
Because of this dense integration, important coupling effects
are known to be introduced [5]–[8]. To quantify them, our
research group is currently facing the crosstalk issue, carry-
ing out measurements on particular Geiger-mode avalanche
photodiode (GMAP) arrays in which different pitches are
available.
In this paper, we first present our measurement scheme to
evaluate the crosstalk probability. The second and last sections
are dedicated to the results of our measurements, showing the
calculated PCT versus capacitance, distance, and bias. The
crosstalk decay time is also presented.
II. CROSSTALK MEASUREMENT SCHEME
Before describing the measurement scheme, we first intro-
duce the crosstalk probability definition. We relate the intensity
of the crosstalk effect to the probability that a pixel called R is
triggered by an avalanche occurring in another neighbor pixel
called S, and we define it as the crosstalk probability PCT.
Our approach to measure this probability is based on the
following idea: First, we measured the dark noise distribution
(the sum of dark count and afterpulsing events) of both pixels R
and S with only one pixel active each time. Then, the same dis-
tributions were measured when R and S were simultaneously
active. When later, through an offline analysis, we compare
these distributions, it is possible to evaluate how the crosstalk
effect has changed the dark noise of each pixel. The crosstalk
probability is then calculated from this change. Later in this
section, it is explained how to extract this probability in our
measurement conditions.
We have characterized arrays fabricated in the same technol-
ogy of those described in [3] with pixel distances between the
centers of 84, 126, 151.5, and 168 µm, as illustrated in Fig. 1.
0741-3106/$25.00 © 2008 IEEE
Authorized licensed use limited to: CNR AREA RICERCA DI BOLOGNA. Downloaded on July 9, 2009 at 02:59 from IEEE Xplore. Restrictions apply.

SCIACCA et al.: CROSSTALK CHARACTERIZATION IN GEIGER-MODE AVALANCHE PHOTODIODE ARRAYS 219
These pixels have an average dark noise rate of about
1000 s−1
. To measure the dark noise distributions, each pixel
is biased above breakdown and quenched by a 200-kΩ passive
quenching resistor.
A digital oscilloscope Tektronix DPO 7104 acquires the
voltage waveform at the quenching resistor terminals.
To measure crosstalk events, the same setup is used to
simultaneously acquire the two waveforms on two different
channels of the oscilloscope when both pixels are biased above
breakdown and quenched by two different resistors. Since the
rising and quenching of the avalanche are mechanisms that
last no more than few tens of nanoseconds, the oscilloscope
resolution should be set at a minimum of 10 ns if cause–effect
relations are to be measured and the exact distribution of the
crosstalk probability density is to be calculated.
Having evaluated the maximum duration of all the possible
crosstalk mechanisms, either electrical or optical, to be in the
range of few microseconds after the avalanche of S, we restrict
our investigation to a time window dt of 2 µs. Measuring the
number of times nmeas when R is triggered within the first 2 µs
after an avalanche in S, and subtracting the normal dark counts,
we calculate the crosstalk probability PCT. In fact, if dt is much
smaller than the reciprocal dark count rate τR, the probability
that R has a dark count in the time window dt can be simply
calculated as
PDK_10 ns =
dt
τR
(1)
and thus, the crosstalk probability is
PCT_10 ns =
nmeas_10 ns ±
√
nmeas_10 ns
N10 ns
− PDK_10 ns (2)
where N10 ns is the total number of times when S is triggered
after R. This method allowed us to evaluate the time density
distribution of the crosstalk probability, but it is very time
consuming if the integral crosstalk probability is to be quickly
found. In fact, a tradeoff between the resolution and number
of samples has to be found because of the limited oscilloscope
memory buffer.
To quickly estimate the integral crosstalk probability when
its value is as low as 10−4
or less, the oscilloscope resolution
must be in the microsecond range. In this last case, using a
resolution of 1 µs, 20-s-long waveforms are taken, and thus,
about 20 000 samples can be evaluated. The drawback of this
resolution is that the cause–effect relation is not observable, and
thus, we need to extract the crosstalk probability in a slightly
different way.
Because of the impossibility to distinguish which pixel
causes the crosstalk, we simply search for the coincidence
counts in a window of 2dt exceeding the coincidences due to
random dark counts in the same window.
The probability that two pixels have a coincidence because
of the dark count in a 2dt time window is
PDKR−DKS =
(2 · dt)2
τRτS
. (3)
Fig. 2. By studying the PCT dependence from capacitance, it is possible to
evaluate by means of extrapolation the PCT at 100 fF.
In a complete waveform long T, the number of coincidences
due to the dark count will be
nDK =
(2 · dt)2
τRτS
·
T
2 · dt
=
2 · dt · T
τRτS
. (4)
The number of crosstalk events can be obtained by subtract-
ing the number of measured coincidences to the number of dark
count coincidences, i.e.,
nCT = nmeas − nDK. (5)
The crosstalk probability is then
PCT =
nmeas ±
√
nmeas − nDK
N
(6)
where N is the total number of counts in T, and
√
nmeas is
introduced as the Poissonian fluctuation error.
III. EXPERIMENTAL RESULTS
Measurements performed at a 1-µs oscilloscope resolution
allowed us to determine the PCT dependence on the capac-
itance, on the pitch, and on the applied bias. Moreover, to
estimate the crosstalk decay time, PCT density versus time
has been calculated due to the 10-ns oscilloscope resolution
acquisitions.
PCT is known to be strictly related to the amount of charge
flowing through the single pixel [2], [9], i.e., to the capacitance
connected at its terminals.
The single pixel capacitance has been calculated to be ap-
proximately of 100 fF. In total, 18-pF capacitance CMS has
been calculated for the measurement system by taking into
account pixel, probes, cables, and pad capacitances. To study
the dependence on the capacitance and to extrapolate the pixel
PCT without the parasitic effects, a parallel connection of
different capacitances, equal to 2.5, 5, 7.5, and 10 times CMS,
has been used. PCT dependence from capacitance is reported in
Fig. 2. Data shown in this figure refer to pixels with the mini-
mal distance between centers (84 µm), where higher crosstalk
effects are expected to be present. Here, extrapolation at 100 fF
indicates a value of 5 · 10−2
%.

220 IEEE ELECTRON DEVICE LETTERS, VOL. 29, NO. 3, MARCH 2008
Fig. 3. PCT versus distance dependence (at 15% OV). Higher values are
found as capacitance increases.
Fig. 4. Increasing bias values applied to the pixel of the 6 × 6 array result in
a higher PCT.
Fig. 3 shows that PCT decreases as pitch increases [2], [7]. In
a semilogarithmic scale, PCT variations when different values
of capacitance are added to the system are also reported.
Larger values of capacitance imply a higher coupling effect
probability. By taking into account a breakdown voltage VBD
of 26.6 V for the single cell of the array, measurements were
carried out by applying three different bias levels of 30.6, 35,
and 38 V, corresponding to about 15%, 30%, and 43% of
overvoltage (OV), respectively.
As expected [6], PCT increases with bias (Fig. 4). This
behavior can be explained by the fact that an increasing bias
results in an increasing charge flowing in the avalanche. Pixels
with the same distance between centers are superimposed, thus
showing a very similar behavior.
To estimate the crosstalk decay time, we calculated the
PCT density versus time by analyzing the 10-ns resolution
measurements. As shown in Fig. 5, a decay time of 65 ns has
been found, which referred to the minimal pitch at 30% OV.
This slow decay time not only indicates that the mechanism
responsible for the crosstalk effect could be electrical but also
indicates that the optical contribution cannot be excluded since
photons are emitted during all the avalanche time, which we
Fig. 5. PCT density versus time. A decay time of 65 ns has been calculated
at 30% OV for the crosstalk.
measured to be a few tens of nanoseconds. This is due to the
high-value parasitic capacitances. A better understanding of
these mechanisms is currently the focus of our study.
IV. CONCLUSION
Dense GMAP arrays have been fabricated in the STMi-
croelectronics Catania facility. The crosstalk coupling results
obtained and extrapolated by the measurement scheme pre-
sented in this paper indicate that a probability as low as 10−2
%
is found. At present, we are repeating a similar study on
denser SiPM arrays where heavier crosstalk coupling effects
are expected, since all the outputs of the pixels are connected
together, and the pitch between adjacent cells has to be reduced
as much as possible to increase the geometrical fill factor and
the photodetection efficiency of the whole photosensor.
REFERENCES
[1] B. Dolgoshein et al., “Status report on silicon photomultiplier development
and its applications,” Nucl. Instrum. Methods A, vol. 563, no. 2, pp. 368–
376, Jul. 2006.
[2] P. Buzhan et al., “Large area silicon photomultipliers: Performance and
applications,” Nucl. Instrum. Methods A, vol. 567, no. 1, pp. 78–82,
Nov. 2006.
[3] E. Sciacca et al., “Arrays of Geiger mode avalanche photodiodes,” IEEE
Photon. Technol. Lett., vol. 18, no. 15, pp. 1633–1635, Aug. 2006.
[4] E. Sciacca et al., “Silicon planar technology for single-photon optical
detectors,” IEEE Trans. Electron Devices, vol. 50, no. 4, pp. 918–925,
Apr. 2003.
[5] F. Zappa et al., “Integrated array of avalanche photodiodes for single-
photon counting,” in Proc. 27th ESSDERC, Stuttgart, Germany, 1997,
pp. 600–603.
[6] W. J. Kindt et al., “Optical cross talk in Geiger mode avalanche photodiode
arrays: Modeling, prevention and measurement,” in Proc. 28th ESSDERC,
Bordeaux, France, 1998, pp. 192–195.
[7] J. C. Jackson et al., “Toward integrated single-photon-counting microar-
rays,” Opt. Eng., vol. 42, no. 1, pp. 112–118, Jan. 2003.
[8] C. Niclass et al., “Design and characterization of a CMOS 3-D image sen-
sor based on single photon avalanche diodes,” IEEE J. Solid-State Circuits,
vol. 40, no. 9, pp. 1847–1854, Sep. 2005.
[9] A. Rochas et al., “First fully integrated 2-D array of single-photon detectors
in standard CMOS technology,” IEEE Photon. Technol. Lett., vol. 15, no. 7,
pp. 963–965, Jul. 2003.

Crosstalk characterization in gmap arrays

More Related Content

What's hot (19)

Similar to Crosstalk characterization in gmap arrays (20)

Recently uploaded (20)

Crosstalk characterization in gmap arrays