Notes for Using an LED as a SPAD
[Please read the student handout before going over these notes.]
Brief background:
In this experiment students build, test, and use a circuit with a very inefficient single photon avalanche
diode (SPAD) – an inexpensive GaP light emitting diode (LED). With this device, students can explore the
probabilistic nature of the arrival of photons at a detector, without using delicate and easily damaged SPADs or
photomultiplier tubes, and without the need for any high voltages. The experiments compliment traditional
experiments using radioactive decay.
A SPAD uses a reverse biased p-n junction as the photon detector – in this case an LED. At the
appropriate bias voltage, no current will flow ‘backward’ through the LED when no light is incident on the
junction, but the absorption of a photon will produce an electron-hole pair that is quickly separated by the large
electric field in the junction. The electrons are accelerated by the field and collide with other atoms, which
produce additional excited electrons. The process continues (see figure), building up an ‘avalanche’ of electrons
that produces a current/voltage pulse that is large enough to be measured. This process is analogous to the
behavior of a Geiger-Müller tube.
Avalanche Photodiodes – Geiger Mode
• Reverse biased
p-n junction
Electron energy diagram
E gap
n
photon
p
Vbias ~ 15 – 150V
conduction band
valence band
n-doped
p-doped
Figure 1. Electron energy diagram in a reverse biased p-n junction. An absorbed photon creates an electron (solid circle) and a hole (empty
circle). The large reverse bias voltage produces an electric field in the junction that accelerates the electrons toward the n-doped side.
When an electron collides with an atom, its kinetic energy can be used to excite an additional electron into the conduction band. Inset:
diagram of a reverse biased p-n junction.
The voltage pulse produced may only be a few hundred millivolts, so it can be helpful to use a
comparator to condition the pulses up to the common 5V digital level. (A comparator is basically an op-amp
without any feedback, so that the output of the op-amp is bistable – ‘high’ when the voltage at the + input
terminal is greater than the voltage at the – input terminal, and ‘low’ when the reverse is true.)
The LED used in this experiment is obviously not designed to detect light, and so its quantum efficiency is
extremely low. We would probably never want to use this as an actual light sensor, unless the light sources were
quite bright. However, this means that students can do the experiments in ambient light, without worrying
about damaging the detector.
The data collected are the times that the electrical pulses occur and how frequently they occur.
Measurements of the number of the pulses (‘events’ or ‘counts’) during a specific time interval ought to be
described by the Poisson distribution, if the pulses occur randomly. Similarly, the distribution of the times
between the pulses ought to be described by an exponential distribution. Students can collect and analyze this
data in exactly the same way that they might for radioactive decay measurements.
Counts per Interval
700
80
Gaussian 
2
red
500
= 57.8
Gaussian 2red = 1.50
60
Frequency
Frequency
Poisson 2red = 2.02
Poisson 2red = 2.57
600
400
300
200
40
20
100
0
0
1
2
3
4
5
0
0
6
5
10
15
20
25
cts per 0.15s
cts per 0.015s
50
Poisson 2red = 3.01
40
Frequency
• Compare to Poisson and Gaussian
predictions for same mean and s.
• Gaussian approximates Poisson for
large mean values
• s2 > mean due to ‘afterpulsing’
Gaussian 2red = 0.997
30
20
10
0
80 90 100 110 120 130 140 150
cts per 1.5s
Figure 2. The distribution of the number of photons (counts) detected in intervals of 0.015s, 0.15s, and 1.5s. The data visually appears to
be described by a Poisson distribution, but the reduced 2 value indicates that there is not quantitative agreement. The disagreement
results from the variance of the distribution being greater than the mean due partly because of afterpulsing.
Afterpulsing
One crucial aspect of the behavior of SPADs is afterpulsing. After a photon initiates an avalanche, it is
possible for an electron in the p-n junction to be left in an excited state, and if that electron gains enough energy
(usually thermal energy) to obtain the conductance band, a second avalanche will occur, without a photon
initiating it. So, a single photon can cause not one, but a sequence of two or more pulses that are not
independent and that occur at very short time intervals.
Afterpulsing can show up in the data as non-Poisson distributions (since some of the pulses are no longer
random), and as a large spike at short times when examining the time between pulses.
Figure 3. The distribution of the time between pulses (events). The data shows the expected exponential behavior, but with a significant
deviation at short times - the result of afterpulsing.
Dead time and pulse widths
By varying the value of the resistor in series with the LED/SPAD, the RC time constant of the circuit can be
controlled. This effectively controls the dead time of the detector (the time during which the detector cannot
register the arrival of a photon).
Figure 4. Left: Pulses produced by the comparator for four different values of the resistance in series with the LED, showing the increase in
the time constant. Right: The distribution of the time between events (data shown for short times only). As the pulse width increase, the
apparent dead time of the circuit (the range where the number of counts is too low) increases also – showing that photons arriving during a
pulse cannot be counted.
Vary Pulse Width to
Determine Capacitance
Vbias
80
 = (62.3
1.3 s)
3.2pF) R+ (-0.8
R
Pulse width (s)
60
40
20
0
0
200
400
600
800
Resistance (k)
1000
Pulse width is
determined by the time
constant of RC circuit:
 = R*C
The slope gives the
capacitance of the
circuit.
Figure 5. Data showing how the average pulse width varies linearly with the resistance in series with the LED. The slope of the line gives
the capacitance of the circuit, which is mainly due to the LED.
Additional exploration:
Students can extend this experiment to look at temperature effects, spectral response, material
variation, etc. There is a huge parameter space for students to explore with this experiment.
Equipment list:
Andor AND-114R LED ($0.279 for quantities less than 100, Newark.com )
LM311N/NOPB comparator ($0.91 for qty less than 10, Mouser.com)
10k potentiometer (T63YB103KT20, $2.86 for qty less than 10, Mouser.com)
5V, 15V, and 0-30V DC power supplies
Electronic breadboard/protoboard
Two channel storage oscilloscope (>100 MHz if possible)
Pulse counter/interval timer, LabView &DAQ, LabQuest, etc.
Building and adjusting the circuit
The construction of the circuit is quite straight forward, but students will need to know/learn/discover a few
things in order to be successful:
-
-
-
-
The proper orientation of the LED (I encourage them to determine the forward bias direction of the
Andor AND114R LED by applying a voltage to see light emission)
The proper choice of the reverse bias voltage. For these LEDs, this voltage is ~24V. While this is
much lower than the ~100V bias voltage used in some SPADs, students are often reluctant to raise
the voltage above 5V for fear of damaging something. Students can use an oscilloscope to monitor
the voltage signal sent to the positive input of the comparator as they raise the bias voltage. A power
supply with a fine voltage control knob is best. Too high of a bias voltage is also a problem - students
need to be sure that when no light is incident on the LED, no (or very few) pulses are produced.
The proper choice of the passive quenching resistor (100-900 k), which determines the RC time
constant of the circuit and therefore the duration of the pulse produced by the LED. I encourage
students to explore how this resistance determines the dead time of the detection circuit – which can
be directly measured by collecting data on the time between the pulses.
How to properly choose the resistance of the 10 k potentiometer. I’ve found this to be a valuable
place for students to struggle – because of the direct correlation between this comparator circuit and
the ‘trigger level’ of an oscilloscope. Students can monitor the pulses from the LED on one channel of
an oscilloscope, and the voltage level at the negative input of the comparator on a second channel of
the oscilloscope. They can then visually bring the comparison voltage down to below the peak value
of the pulses. Observant students will quickly understand that the choice of the ‘trigger level’ has an
impact on how many pulses are missed and therefore not counted.
On the output of the comparator, the choice of the ‘pull-up’ resistor is important, particularly if the
device used to measure the output pulses has a low internal resistance.
Using a Vernier LabQuest to collect data
The signal from the op-amp can be sent to the digital input of a Vernier LabQuest to be detected. Telling the
LoggerPro software that it is connected to a Radiation Monitor allows students to quickly and easily collect
‘events per time interval’ data that can be compared to Poisson and Gaussian distributions. To obtain
information regarding the times of the photon arrival events, the LabQuest can be set as if it were connected to a
Photogate. LoggerPro provides several functions that let students easily calculate the time between the events
that they can then compare to an exponential distribution and examine the dead time of the circuit.
Figure 6. A picture of the SPAD circuit. The decade resistor is used to vary the resistance in series with the LED.
Using LabView to collect data
A simple LabView VI, along with a ‘fast enough’ DAQ can be used to collect both the number of pulses from the
LED and the times that the pulses occur. An example is shown in the Figures 7 and 8. We’ve used a USB 6251
(1.25 MS/s) DAQ to collect data both from the output of the op-amp as well as directly from the LED itself. The
crucial part of the LabView program is the “Threshold Detector.vi” that is used to determine when a peak occurs.
Figure 7. The front panel of a LabView VI to count the number of pulses and the times at which the pulses occur.
Figure 8. The Block Diagram of the LabView VI shown in Figure 7.
Useful references about SPADs
S. Cova , M. Ghioni , A. Lotito , I. Rech & F. Zappa “Evolution and prospects for single-photon
avalanche diodes and quenching circuits”, J. Mod. Opt., 51 (9-10), 1267-1288 (2004).
S. Cova, M. Ghioni, A. Lacaita, C. Samori, and F. Zappa, “Avalanche photodiodes and quenching
circuits for single-photon detection”, Appl. Opt. 35 (12), 1956 – 1976 (1996).
V. A. Krasnov, S. V. Shutov, Yu. M. Shwarts, and S. Yu. Yerochin, “Note: Determination of temperature
dependence of GaP bandgap energyfrom diode temperature response characteristics” Rev. Sci. Instrum.
82, 086109 (2011)
A. Gallivanoni, I. Rech, and M. Ghioni, “Progress in Quenching Circuits for Single Photon Avalanche
Diodes”, IEEE Trans. Nuc. Sci., 57 (6), 3815-3826 (2010)
F. K. Reinhart, “Reverse-Biased Gallium Phosphide Diodes as High-Frequency Light Modulators,” J.
Appl. Phys. 39 (7), 3426 – 3434 (1968).