A setup for in-situ measurements of potential and UV
induced degradation of PV modules inside a climatic
Stefan Mitterhofer, Marko Jankovec, Marko Topič
Faculty of Electrical Engineering
University of Ljubljana
Stefan.Mitterhofer@fe.uni-lj.si, Marko.Jankovec@fe.uni-lj.si, Marko.Topic@fe.uni-lj.si
Abstract An experimental setup inside a climatic chamber for in -situ measure ments of PV modules under simulated
outdoor conditions including UV irradiation, temperature, humidity and high voltage bias is proposed. The purpose
is the determination of the interaction between those factors and their influence on the degradation of PV module
components. Different components are reviewed for the proposed setup, with special focus on the possible light
sources. The choice for an array of UV LEDs is made. The main advantages are good scalability of the array size,
high temperature stability and long lifetime.
A methodology is presented to calculate the irradiance distribution over the size of the target area. An iterative
optimization method for a better homogeneity of the irradiance is devised. Each step, the radiance of single LEDs is
changed to approximate a function, which converges to an average irradiance value at certain points at the target.
The results of the calculations are verified via comparison to measurement dat a.
As early as 1978, the possible degradation of
photovoltaic (PV) module components caused by the
combination of environmental parameters and a high
voltage between single cells in a string and the module
frame was identified . Further work by the Jet
Propulsion Laboratory found a good correlation
between integrated cell-to-frame leakage current and
observed corrosion of the transparent conductive
oxide (TCO) in amorphous silicon (a-Si), and between
integrated leakage current and corrosion of the solar
cell metallization in crystalline silicon (c-Si) modules
. Temperature (T) and humidity (RH) were
identified as the main environmental factors of the
Other degradation modes observed in the field and
in laboratory tests could also be connected to those
parameters. In 2005, Sunpower observed the reduction
of the effectiveness of the silicon oxide surface
passivation layer in their high-efficiency c-Si solar
cells, caused by an accumulated charge on the surface
. In the same year, the Florida Solar Energy Center
observed increasing delamination in a-Si thin-film
modules with increasing cell-to-frame voltage, T and
In 2010, Solon introduced the term “potential
induced degradation” (PID) for an observed reduction
of the power output of their commercial c-Si modules
[6, 7] caused by said parameters. They observed a
strong reduction of the shunt resistance. Stacking
faults in the crystalline structure [8, 9] were found to
play a major role in potential induced shunting (PID-
s) by Fraunhofer CSP. Sodium ingress into the silicon
along the faults is proposedly causing PID-s . This
is observed in negatively-biased p-type c-Si cells.
Nevertheless, Q-Cells also observed PID-s in sodium-
free modules, indicating the possibility of other ions
causing the degradation . Fraunhofer ISE found T
and RH to have an impact on this degradation mode
UV light can slow down the PID-s type degradation
. However, real-field analysis has shown it to be a
main factor of backsheet and encapsulant degradation
, in addition to T and to a lesser extent RH. Light
intensity, duration of illumination and wavelength
each have an impact on yellowing and changes of the
UV transmission of the encapsulant .
Still, the exact mechanisms of PID and the
interactions between different technological factors,
on module and system level, and environmental
factors are not yet fully understood, emphasizing the
need for further experiments . For accelerated
module qualification and testing new combined cycles
(for example [17, 18]) or sequential tests (for example
) are thus being developed.
In this paper, an experimental set up for in-situ
measurements in a climatic chamber is proposed. In
section 2, an overview of the proposed setup and the
components used is presented. Their choice is backed
by examples in literature. In section 3, a review of
different available light sources is given. In section 4,
a methodology is described to calculate and optimize
the irradiance of the light sour ce of choice at the target
area. The results are verified by comparison with
experimentally obtained data of a specific example,
presented in .
2 EXPERIMENTAL SETUP
The climatic chamber used is a Kambič KK-2310
CHLT. Temperature control between -70°C and
180°C is possible, with an uncertainty ranging
between ± 0.1°C and ± 0.5°C. RH control is possible
between 10% and 98%, with an uncertainty of ± 3%.
However, RH control is available only for
temperatures between 10°C and 95°C. This range
encompasses all current IEC standards for PV
modules, for example the general qualification
procedures in IEC 61215 [20, 21], but also the
technical specifications specifically for PID in c-Si
IEC TS 62804 . A voltage source (Keithley 237
Source Measure Unit) is able to provide a bias between
cells and frame of up to 1100 V, which includes the
value required in .
The modules used will be c-Si mini-modules.
Leakage current from cell to ground will be measured.
Dark I-V curve measurements will monitor the
characteristics of the modules . An initial
characterization under Standard Testing Conditions
(STC) will be done using a Newport Oriel Class A
solar simulator, model 93194A. Because changes of
the short circuit current cannot be seen in dark I-V
measurements, the modules still have to be taken out
of the setup and measured under STC. A similar setup
has already been used to successfully show different
degradation modes in positively-biased p-type c-Si
cells, including conduction lines corrosion, EVA
evaporation and cell degradation .
Further measurements will include determination of
T and RH at the backsheets of the modules.
Encapsulating miniature sensors in the modules to
monitor humidity ingress has shown promising results
 and is under consideration.
3 COMPARISON OF DIFFERENT LIGHT
A light source will be added to the chamber. It has to
withstand high T and RH in the chamber without
strong degradation. Three choices will be qualitatively
evaluated here: Xenon-arc lamps, with corresponding
filters commonly used in solar simulators,
fluorescence tubes, and LEDs, which are due to their
modularity currently being researched for PV module
testing (for example ).
3.1 Xenon-arc lamps
Xe-arc lamps, correctly filtered, provide a good
spectral radiance fit over a big range of the solar
spectrum. An advantage for the proposed setup is the
possibility of measuring the illuminated IV curves of
the modules without the need to remove them from the
chamber. In addition, light over the whole spectrum
induces other degradation modes seen in real-field
applications, commonly referred to as light-induced
degradation (LID) .
However, simulating light over the whole solar
spectrum can lead to problems with temperature
control. Module temperatures of up to 120°C are
expected with such a light source inside the chamber
when using the T and RH values of the IEC standards
. In addition, Xe-arc lamps degrade quickly
compared to other light sources and need to be
changed every 1500 hours .
3.2 Fluorescence tubes
UV fluorescence tubes have previously been used to
design a light source for a climatic chamber . They
can emit UV light down to the solar cutoff wavelength
at ~295 nm. It is possible to manufacture long tubes,
allowing illumination of big areas, for example full-
sized PV modules.
Their resistance to high T and RH is adequate. They
lose approximately 50% of the radiance at 85°C
compared to 25°C at STC. The high RH causes
degradation products accumulating on the surface of
the tubes, emphasizing the need to clean them
regularly. Doing this, their output is relatively constant
over a long time of usage.
LEDs have several advantages: They can be
mounted in various constellations on a printed circuit
board (PCB), which results in more freedom of the
size and form of the setup. LEDs also have a very long
lifetime compared to other light sources. Furthermore,
the output of individual LEDs can be varied by
changing their current, allowing optimizing the
irradiance on different areas of the target for a specific
purpose. The current can be varied by a constant or
pulsed current source. The latter would greatly
simplify the control of bigger arrays. However, an
effect of the pulsed light intensity to the degradation
of PV module components and materials has to be
LEDs’ light spectra exhibit low half width of the
spectral radiance compared to the other light sources.
Additional intensity peaks at different wavelengths,
observed in other light sources, are not common in
LEDs. This enables a comparison of the impact of
different wavelengths by choosing different LEDs.
Availability of high-power LEDs in the UV
spectrum is limited. There are multiple choices on the
market with a peak irradiance at 365 nm from various
manufacturers, for example Nichia, LED Engin and
LiteOn, which is well above the solar cutoff
wavelength. LEDs below 365 nm lack the high
radiance needed and are comparably very expensive.
The step to full-sized modules can be expensive due
to the huge number of LEDs required to cover such a
3.4 Choice of the light source
We chose to focus on the UV spectrum, ruling out
Xenon-arc lamps. The advantages and disadvantages
of LEDs and fluorescence tubes are compared. The
choice falls to LEDs, due to better scalability of the
array size, better temperature resistance and longer
4 LED ARRAY CALCULATIONS
4.1 Irradiance of single LEDs
A nonperfect Lambertian emitter  is a good
approximation of the radiance distribution of most
Here, θ is the viewing angle, the on-axis
irradiance at distance r and m a parameter describing
the radiance distribution of a LED.
is defined as the viewing angle at
which the irradiance is at half of the maximal value.
of UV LEDs available on the market
are given in table 1. The corresponding irradiance
patterns are shown in figure 1.
LED Engin LZ1-00UV00-0000
LED Engin LZ4-04UV00-0000
of various UV LEDs
Figure 1: Approximated radiance of different LEDs
To calculate the irradiance at any point (x, y) of a flat
target surface at the distance z from a single LED,
additional factors have to be considered. An inverse-
square law with increasing distance r between light
source, approximated as a point source, and target
reduces the irradiance. In addition, it is reduced by the
cosine of the angle between the axis normal to the
target and the incident light beam. Thus, the
Here, is the irradiance at the target directly in front
of the LED, i.e. θ = 0°, and at distance r = 1 m. If we
consider a LED at an arbitrary point , the
distance r is:
Thus follows a more generalized formula:
4.2 Irradiance of a square LED array
To determine the irradiance of an N × M square array
of LEDs, the irradiances of the single LEDs can be
summed up. Their positions are, with a
distance between the LEDs:
Note that in this coordinate system, the point ( x = 0,
y = 0) on the target area is directly in front of the LED
(1, 1) in the corner of the array at distance z.
If not all LEDs provide the same irradiance,
changes to .
To obtain an ideal irradiance uniformity at the target
area, an optimization method for the calculation of the
distance between the LEDs is used . It yields
-120 -90 -60 -30 0 30 60 90 120
accurate results for high values of m. However, small
values of m (~1), which are common in high-power
UV LEDs and equivalent to a wide angle
result in non-uniformity of the irradiance. It is much
stronger above the center of the array than at the edges.
Thus, the LEDs in the center of the array have to be
dimmed by varying the power for better uniformity.
4.3 Radiance variations of the LEDs
In the following, an optimization method is described
to vary the output of the LEDs for better homogeneity
of the irradiance at the target area. Each light source is
connected to a single optimization point (OP) at the
target. Those OPs are homogeneously spread across
that area and form an N × M array. Their positions
on the target are:
and are factors introduced to vary the size of
the target area, while keeping the size of the LED area
constant, introduced for practical purposes.
Based on this geometry, an iterative optimization
method is devised to recalculate the radiance needed
of single LEDs in the array. At first, the irradiance at
the OPs and their average are calculated
according to equation (9). Afterwards, the irradiance
value of an OP is brought closer to the average at the
next iteration step by changing only the corresponding
LED, without considering changes of the other LEDs.
This calculation is being done for all optimization
Hereby, is the part of the irradiance at an OP only
from the corresponding LED, the part from all
other LEDs. o is the number of the iteration. is
the target irradiance at the OP and changes with each
iteration. The final goal is to obtain . However,
using as a target in equation (11) can result in
unrealistic values, for example negative radiance of
the central LEDs at some iteration steps for big arrays,
i.e. N × M > 10 × 10. Thus, an arbitrary value is used
for, which is recalculated each iteration step:
This function converges as the irradiance at the OP
approaches the average:
From the results of equation (11), the needed
variations of the power input to the LEDs can be
calculated to dim them accordingly. It is normalized to
the highest power needed, which can be found in the
corners of the array.
Then, the calculations are repeated. They converge
towards a solution, describing the needed radiance of
every single LED in the array. Figure 2 shows an
example of the convergence of the process. The
irradiance at 4 different optimization points for a 10 ×
10 array during the first 20 iterations is plotted. It is
normalized to the solution found after 100 iterations,
when the standard deviation of the irradiance at the
OPs is below 0.01%. The distance between array and
target is 10 cm, the angle
60°, resulting in
an ideal distance 7.1 cm between the
LEDs using the method described in . The size of
the array is thus (63.9 × 63.9) cm. The area to optimize
for is set to 90% of the size of the LED array, i.e.
The time needed by the method for this array size is
approximately 1 second for 10 iterations.
Figure 2: The first 20 iterations of the optimization
In figure 3 (top), the irradiance over the size of the
optimized area before varying the single LEDs is
shown. The same values are used as in the example
described before. is set to 1 W m2
. In figure 3
(bottom), the same pattern is visualized after 100
iterations of the optimization method. Note the
difference of the irradiance axis’ scale.
Over this area, the calculated irradiance values range
between 392.3 and 588.0 W m2
resulting in a 50% difference between minimum and
maximum. After, they range between 245.7 and 251.3
, i.e. a 2% difference between minimum and
maximum. While this is a huge improvement, the
variations are still larger than those only at the OPs.
This can be explained by higher or lower irradiance
between the OPs, which can also be seen in figure 3
(bottom), and is a limitation of the method. A
possibility for improvement is to use multiple OPs
0 5 10 15 20
connected to each LED, spread homogeneously in the
area around a central OP described in equation (10).
Figure 3: Irradiance of an LED array at a flat target
area before (top) and after (bottom) optimization
4.4 Comparison with experimental results
The optimization method presented in chapter 4.3 is
used to calculate the idealized radiance for a specific
example, presented in . It is a 6 × 3 LED array.
The LEDs used are LiteOn LTPC-C034UVH365,
mounted in a distance of 3.1 cm and
3.4 cm. The distance to the target area is 3 cm.
is set to 64° for the calculations, which yields the
best fit for the experimentally obtained angular
radiance distribution of the used LEDs. The area
optimized for is 0.95 of the LED array size.
Table 2 shows the difference between the manually
optimized, taken from , and the calculated
radiance values, normalized to the maximum at the
corner LED (1,1).
Table 2: Comparison of the power input to LEDs
The irradiance calculations, presented in chapter 4.2,
are compared to irradiance measurements. The
calculated values result in 100.6 % of the measured
results over a 15 × 6.8 cm sized area with a standard
deviation of 0.9 %, shown in figure 4.
Figure 4: Comparison between measured and
calculated irradiance of a 6x3 LED array
The proposed setup for in-situ measurements of PV
modules under simulated outdoor conditions has been
presented. The choice of several components has been
explained and backed up with already published
results and comparison with international standards
and technical specifications of the IEC. A qualitative
analysis of different light sources has been presented,
resulting in the choice to use an array of UV LEDs in
Calculations of the spatial irradiance homogeneity of
the LED array have been presented. Based on them, an
optimization method has been derived to vary the
radiance of single LEDs for better irradiance
uniformity at the target area. A comparison with
experimental results yields a good fit of both
irradiance calculation and optimization method.
This project has received funding from the European
Union’s Horizon 2020 programme under GA. No.
721452. The authors thank Matija Pirc, Faculty of
Electrical Engineering, University of Ljubljana, for
useful discussion and the measurement data used in
 A.R. Hoffman, R.G. Ross, “Environmental
qualification testing of terrestrial solar cell
modules”, 13th IEEE PVSC, pp. 835-842, 1978.
 G.R. Mon, R.G. Ross, “Electrochemical
degradation of amorphous-silicon photovoltaic
modules”, 18th IEEE PVSC, pp. 1142-1149, 1985.
-28.7 -14.4 0.0 14.4 28.7
-28.7 -14.4 0.0 14.4 28.7
calculated / measured
 G.R. Mon, L. Wen, R.G. Ross, D. Adent, “Effects
of temperature and moisture on module leakage
currents”, 18th IEEE PVSC, pp. 1179-1185, 1985.
 R. Swanson, et al, “The surface polarization effect
in high-efficiency silicon solar cells”. 15th PVSEC,
 N.G. Dhere, V.V. Hadagali, K. Jansen,
„Performance degradation analysis of high-voltage
biased thin-film PV modules in hot and humid
conditions“, 31st IEEE PVSC, pp. 507-510, 2005.
 S. Pingel, et al. “Potential induced degradation of
solar cells and panels”, 35th IEEE PVSC, pp.
 J. Berghold, et al. “Potential induced degradation
of solar cells and panels”, 25th EUPVSEC, pp.
 V. Naumann, et al. “The role of stacking faults for
the formation of shunts during potential-induced
degradation of crystalline Si solar cells”, physica
status solidi (RRL)-Rapid Research Letters, vol. 7,
no. 5, pp. 315 – 318, 2013.
 V. Naumann, et al. “Microstructural analysis of
crystal defects leading to potential-induced
degradation (PID) of Si solar cells”, Energy
Procedia, vol. 33, pp. 76-83, 2013.
V. Naumann, et al. “Explanation of PID of the
shunting type by Na decoration of stacking faults
in Si solar cells”, Solar Energy Materials and Solar
Cells, vol. 120, pp. 383-389, 2014.
M. Schuetze, et al. “Investigations of potential
induced degradation of silicon photovoltaic
modules”, 26th EUPVSEC, 2011.
S. Hoffmann, M. Koehl, “Effect of humidity and
temperature on the potential-induced degradation”,
Progress in Photovoltaics: Research and
Applications, vol. 22, no. 2, pp. 173-179, 2014.
P. Hacke, et al. “Interlaboratory study to
determine repeatability of the damp -heat test
method for potential-induced degradation and
polarization in crystalline silicon photovoltaic
modules”, IEEE Journal of Photovoltaics, vol. 5,
no. 1, pp. 94-101, 2015.
T.J. Trout, “Module durability – Connecting field
results, accelerated testing, and materials”, 44th
IEEE PVSC, Washington, 2017.
X. Gu, “Effects of UV light intensity and
wavelength on degradation of PV backsheets and
laminates”, 7th SPOHIA workshop, Freiburg, 2017.
W. Luo, et al. “PID in photovoltaic modules: a
critical review”, Energy and Environmental
Science, vol. 10, no. 1, pp. 43-68, 2017.
A. Borne, “New combined cycles provide new
options and challenges for durability assessment”,
7th SOPHIA workshop, Freiburg, 2017.
T. Sample, “New combined cycles for future test
standards in IEC TC82 WG2”, 7th SOPHIA
workshop, Freiburg, 2017.
M. Pirc, M. Topič, „Design of a homogeneous
LED based portable UV lamp for medical
research”, MIDEM Conference 2017, Ljubljana,
IEC TG 82, “IEC 61215-1:2016; Terrestrial PV
modules – Design qualification and type approval
– Part 1: Test requirements”, International
Standard, edition 1.0, 2016.
IEC TG 82, “IEC 61215-2:2016; Terrestrial PV
modules – Design qualification and type approval
– Part 2: Test procedures”, International Standard,
edition 1.0, 2016.
IEC TG 82, “IEC TS 62804-1:2015 – PV modules
– test methods for the detection of potential-
induced degradation – Part 1: Crystalline silicon”,
Technical Specification, edition 1.0, 2015.
D.L. King et al. “Dark current-voltage
measurements on photovoltaic modules as a
diagnostic or manufacturing tool”, 26 th IEEE
PVSC, pp. 1125-1128, 1997.
K. Brecl, M. Bokalič, M. Topič, “Examination of
photovoltaic silicon module degradation under
high-voltage bias and damp heat by
electroluminescence”, Journal of Solar Energy
Engineering, vol. 139, no. 3, pp. 031011, 2017.
M. Jankovec, et al. “In-Situ monitoring of
moisture ingress in PV modules using digital
humidity sensors”, IEEE Journal of Photovoltaics,
vol. 6, no. 5, pp. 1152 -1159, 2016.
P. Fuss-Kailuweit, “LEDs instead of broad-band
irradiation”, 7th SOPHIA workshop, Freiburg,
J. Lindroos, H. Savin, “Review of light -induced
degradation in crystalline silicon solar cells”, Solar
Energy Materials and Solar Cells, vol. 147, pp.
M. Koehl, D. Philipp, N. Lenck, M. Zundel,
„Development and application of a UV light source
for PV-module testing“, Proc. SPIE: Reliability of
Photovoltaic Cells, Modules, Components, and
Systems II, pp. 7412021-7412027, 2009.
on the 22.08.2017.
D. Wood, “Optoelectronic semiconductor
devices”, Prentice-Hall International, 1994.
I. Moreno, M. Avedano-Alejo, R.I. Tzonchev,
“Designing light-emitting diode arrays for uniform
near-field irradiance”, Applied optics, vol. 45, no.
10, pp. 2265-2272, 2006.