You are on page 1of 9

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/326606105

An investigation into the failure mechanisms of Schottky barrier diodes, as


commonly implemented as bypass diodes in photovoltaic modules

Conference Paper · June 2018

CITATIONS READS

2 2,549

3 authors:

Kurt Michael Coetzer Pieter Gideon Wiid


Stellenbosch University Cape Peninsula University of Technology
7 PUBLICATIONS   16 CITATIONS    48 PUBLICATIONS   133 CITATIONS   

SEE PROFILE SEE PROFILE

Arnold Johan Rix


Stellenbosch University
86 PUBLICATIONS   367 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Helping South African schools reduce their water and electricity bills View project

Characterising the EM Environment of MeerKAT View project

All content following this page was uploaded by Kurt Michael Coetzer on 25 July 2018.

The user has requested enhancement of the downloaded file.


AN INVESTIGATION INTO THE FAILURE MECHANISMS OF
SCHOTTKY BARRIER DIODES, AS COMMONLY IMPLEMENTED AS
BYPASS DIODES IN PHOTOVOLTAIC MODULES

KM Coetzer1, PG Wiid2, and AJ Rix3


Stellenbosch University, Department of Electrical and Electronic Engineering, Stellenbosch, South Africa

E-Mail: 117511771@sun.ac.za; 2wiidg@sun.a.za; 3rix@sun.ac.za

lightning strikes near to the service or structure – with the


Abstract
interested case being utility-scale photovoltaic installations.
The purpose of this paper is to identify cases in which the
Schottky barrier diodes, used as bypass diodes in photovoltaic The bypass diode in photovoltaic modules has been identified as
modules, are susceptible to damage as a result of induced a common point of failure – sometimes resulting in the
currents and voltages caused by nearby lightning strikes. replacement of an entire photovoltaic module.

In this paper, a modular voltage impulse generator is used to The current South African standard (the adoption of IEC
induce failures within Schottky barrier diodes. The design of this 61215:2005) for the design qualification and type approval of
impulse generator was based on that of a small-scale solid-state crystalline silicon terrestrial photovoltaic modules is presented
Marx generator, the design of which is documented in [1]. The in [2]. The sole mention in [2] relating to bypass diodes is the
effects of voltages and currents larger than the diodes’ maximum “bypass diode thermal test”, which is a test of the long-term
ratings, both in reverse and forward biases, are noted. Any reliability of the bypass diodes used in photovoltaic modules.
degradation of the diode’s current-voltage curves in differing test The interest of this paper specifically relates to the effects of
conditions are also noted. The waveforms produced by the impulse voltages and currents on these diodes (with the aim of
generator resemble those used in the imitation of lightning minimizing the damaging effects of these waveforms), therefore
waveforms. it is suggested that mention should be made of diode impulse
operation in future standards literature.
The study concluded that the type of Schottky barrier diodes used
in photovoltaic modules are particularly resilient to impulses This paper aims to investigate the manners in which these diodes
when placed in forward bias but are rather frail when subjected are prone to failure, in order to provide insight into the larger
to impulses in reverse bias – often resulting in a short-circuit problem of failing photovoltaic modules as a result of nearby
failure. lightning strikes.

Keywords: Schottky diode; bypass diode; photovoltaic; solar; 2. Test Setup


lightning; impulse generator
2.1. South African National Standards (SANS)
1. Introduction requirements
The current South African standard (the adoption of IEC 60060-
Solar photovoltaic installations of any size can be susceptible to 1:2010) which details the general definitions and test
damage as a result of lightning storms. This damage can be requirements for high-voltage test techniques is presented in [3].
caused by lightning strikes either directly to, or near to, a A lightning-impulse voltage is defined in [3] as an intentionally
structure or service connected to that structure. applied aperiodic transient voltage, which usually rises rapidly
Solutions exist for the mitigation of damage as a result of direct to a peak value and then falls more slowly to zero, with a front
strikes to a structure or service in the form of lightning protection time of less than 20 µs. A full lightning-impulse voltage is
systems (made up of air termination systems, down conductors, defined in [3] as a lightning-impulse voltage which is not
and earth termination systems). Solutions also exist for the interrupted by a disruptive discharge.
handling of large current surges within power systems, such as Fig. 1 shows a full impulse voltage waveform with time
surge protective devices. Even with these systems in place, little parameters, as defined in [3]. T1 represents a virtual parameter
to no protection is offered against the electromagnetic effects of labelled the front time and represents 1/0.6 times the interval T
between the instants when the impulse is 30% and 90% of the 2.3. Simulation
peak value on the test voltage curve [3]. Resistors R1, R2, and R3 in Fig. 4 have values of 4.35 Ω, 17.4 Ω,
O1 represents the virtual origin, which is defined as the point on and 1.31 Ω, respectively. These resistors were chosen for their
the time axis preceding the projection of point A onto the time- ability to handle the large voltages and currents present in the
axis by a time 0.3 T1. pulse-shaping network. Inductor L1 is a hand-wound 7.1 µH air-
core inductor. These component values were determined by
T2 represents the time to half-value, which is a virtual parameter means of simulation in the Simulink modelling environment.
defined as the time interval between the virtual origin, O1, and
the instant when the test voltage curve has decreased to half the
test voltage value.
A waveform is generally defined in the format T1/T2 – for
example, a waveform with a front time T1 of 1.2 µs and a time to
half-value T2 of 50 µs is represented by 1.2/50. Ue is defined as
the maximum value of the recorded curve. The tolerances
Fig. 2. The 1.2/50-8/20 µs impulse generator circuit
described in [3] allow for up to a ±3% on the test voltage value described in [4] in the open-circuit voltage condition.
Ue, ± 30% on the front time T1, and ± 20% on the time to half-
value T2.

Fig. 3. The 1.2/50-8/20 µs impulse generator circuit


described in [4] in the short-circuit current condition.

In this study, capacitor C1 and switch S1 in Fig. 2. were


replaced with the Marx generator, as shown in Fig. 4.

Fig. 1. A full impulse voltage waveform with time


parameters as defined in [reference]
Fig. 4. Block diagram of the impulse generator used in this
2.2. Pulse-shaping network
study.
A pulse-shaping network was required to shape the voltage 2.3.1. Open-circuit simulation
impulse provided by the solid-state Marx generator in order to
conform to the above-mentioned specification. The chosen A six-stage Marx generator was charged to 64 V per stage,
pulse-shaping network was based on the 1.2/50-8/20 impulse producing a total of 377 V at its output (after internal losses were
generator circuit described in [4], shown in Fig. 2 and Fig. 3. taken into account). This was supplied at the input to the pulse-
This generator can produce a 1.2/50 voltage impulse in an open- shaping network, with the output of the pulse-shaping network
circuit voltage condition and an 8/20 current impulse in a short- in the open-circuit position (no device under test attached). The
circuit condition. waveform recorded at the output of the pulse-shaping network
can be seen in Fig. 5.
the 30% of Ue and 90% of Ue points was calculated using the
following equation for a straight-line

𝑦 = 𝑚𝑥 + 𝑏
where 𝑥 and 𝑦 represent the specific x and y coordinate at a point
365.9 V on the curve, 𝑚 represents the gradient of the line, and 𝑏
represents the y-axis intercept of the line. Since two points were
known, the y-axis intercept could be calculated by substituting
the known values in, as follows:
329.31 − 109.77
329.31 = ( ) (8.8 × 10−7 ) + 𝑏
720 n
3659
→𝑏=
Fig. 5. The simulated open-circuit voltage waveform 60
recorded at the output of the pulse-shaping network in Secondly, the x-intercept (the location of O1) was subsequently
simulation. found.
For a front time of 1.2 µs, the required time between 30% of Ue 329.31 − 109.77 3659
to 90% of Ue would be: 0=( ) (𝑥) +
720 n 60
0.6 → 𝑥 = −0.2 μs
1.2 μs × = 720 ns
1
Adding the value of T2 to this gave the following location.
As the front-time is determined by the time-constant L1/R2, once
an appropriate value of R2 was chosen L1 was then adjusted until −0.2 μs + 50 μs = 49.8 μs
the desired 30% of Ue - 90% of Ue time of 720 ns was obtained,
It was found that with a value of R1 of 4.35 Ω, the waveform
as can be seen in Fig. 6.
reached a value of 182.95 V (50% of Ue) at a time of 47.18 µs
(instead of the ideal 49.8 µs), giving a T2 of 47.38 µs (instead of
the intended 50 µs), as seen in Fig. 7. Since this simulated value
of T2 was only 5.24% too short, it was deemed as acceptable as
90% Ue it fell well within the allowable ± 20% tolerance for T2.

30% Ue

Intended
Recorded

Fig. 6. A figure showing the simulated open-circuit voltage


rise times.
The intended location of 50% of Ue on the falling edge of the
waveform was then calculated as described below. Fig. 7. The simulated intersection of the open-circuit voltage
waveform with the 50% value of Ue, along with the intended
Firstly, the y-axis intercept of the straight line projected through intersection point.
2.3.2 Short-circuit simulation The front time of the short-circuit current impulse was then
According to [3], the complementary impulse to a 1.2/50 open- examined. For a front time of 8 µs, the required 30% to 90% peak
circuit voltage impulse is an 8/20 short-circuit current impulse rise time would need to be 4.8 µs, which was what was obtained
(as seen in Table 1). (as can be seen in Fig. 9).

Definitions Front time (µs) Time to half-


value (µs)
90%
Open-circuit voltage 1.2 ± 30% 50 ± 20%

Short-circuit current 8 ± 20% 20 ± 20%

Table 1. Definitions of waveform parameters, according to


[3].

According to [5], a relationship should exist between the


measured open-circuit peak voltages and the measured short- 30%
circuit peak currents. This relationship is shown in Table 2.

Open-circuit peak voltage Short-circuit peak current


± 10% ± 10%
0.5 kV 0.25 kA
1.0 kV 0.5 kA Fig. 9. A figure showing the simulated short-circuit current
2.0 kV 1.0 kA rise times.
4.0 kV 2.0 kA The simulated value of T2 from the simulation of the short-circuit
Table 2. The intended relationship between peak open- current impulse was 29.13 µs (45.65% longer than intended), as
circuit voltage and peak short-circuit current, as specified can be seen in Fig. 10.
by [5].

In order to ensure that the peak short-circuit current conformed


to the relationship specified by [5], each capacitor in the six-
stage Marx generator was charged to 89 V. This resulted in a
1.2/50 open-circuit peak voltage of 501.3 V. The output of the Intended
pulse-shaping network was then short-circuited and R3 was Recorded
adjusted until the simulated peak short-circuit current was 250.1
A. The value of R3 in this case was 1.31 Ω. The simulated short-
circuit current impulse waveform can be seen in Fig. 8. Fig. 10. The simulated value of the intersection of the short-
circuit current waveform with the 50 % of max value point,
along with the intended intersection point.
2.4. Evaluation of constructed impulse generator
performance
250.1 A The performance of the constructed impulse generator was
analysed with the output terminals in the open-circuit and short-
circuit positions.

2.4.1 Open-circuit voltage test


Fig. 11 shows the measured voltage waveform at the output of
the impulse generator when the output terminals were in the
Fig. 8. The short-circuit current waveform simulated at the open-circuit condition. The waveform reached a peak value of
output of the pulse-shaping network. 511.04 V, with a front-time of 3,3 µs and a time to half-value of
46.35 µs. The front-time falls outside of the 30% tolerance of the
IEC front-time requirement for a 1,2/50 waveform. The peak
value and time to half-value fall within the 3% and 20%
tolerances, respectively.

Fig. 12. The current waveform measured at the output


terminals of the impulse generator when in the short-circuit
condition.

Fig. 11. The voltage waveform measured at the output 3. Pictures of test-setup
terminals of the impulse generator when in the open-circuit
condition. The test-setup consisted of a six-stage solid-state Marx generator
2.4.2. Short-circuit current test (as can be seen in Fig. 13).

Fig. 12 shows the measured current waveform at the output of


the impulse generator when the output terminals were in the
short-circuit condition. The waveform reached a peak value of
251.5 A, with a front-time of 14.5 µs and a time to half-value of
38.11 µs. The front-time and time to half-value both fall outside
the 20% tolerance of the IEC requirements for a 8/20 waveform.
The peak value fell within the 3% tolerance. The slower than
expected waveform can be attributed to the effects of the
inductances found within the power resistors which are found in
the shaping network. These excessive inductances should be
removed or compensated for in order for the impulse generator
to fall within IEC tolerances. As the impulse generator was still
seen as capable of providing useful insight, testing was
continued.
Fig. 13. The six-stage Marx generator used for the testing in
this article.
The complete test-setup involved the Marx generator, the pulse- current impulses of peak values 70 A, 140 A, 200 A, 230 A, 270
shaping network, a 12 V DC source to power all electronic A, 340 A, 390 A, and 450 A. Ten impulses were given to each
components, a 330 V adjustable DC source to charge the Marx diode at their respective current impulse level. Fig. 15 shows the
generator to the desired voltage, a Tektronix digital-storage 10th 450 A current impulse waveform supplied to diode 8. All 8
oscilloscope with Tektronix TCP303 current probe for impulse diodes were tested after each impulse and found to function
current measurements, a 10x voltage probe for impulse voltage correctly, with the expected forward voltage.
measurements, and a laptop to run the software which automated
the process of executing the impulses. A Perspex blast-screen
provided a level of safety for the operator. The complete test-
setup used can be seen in Fig. 14.

Fig. 14. The complete test-setup used for the gathering of Fig. 15. The measurement of the 10th 450 A current
results in this article. impulse waveform supplied to diode 8.
4.3.2. Reverse bias
4. Testing
Thirty diodes were tested in the reverse-biased position with
4.1. The device under test voltage impulses of peak values ranging from 160 V to 250 V.
The device under test in this case is a HY 10SQ045 Schottky
Around 200 V (± 5 V) it was found that one could not estimate
barrier diode. This diode is typical of one used as a bypass diode
the result of a test. Above this level diodes would almost
in photovoltaic modules due to its high current capacity (10 A)
certainly fail in a short-circuit position, and below this level
and low forward voltage (0.55 V at 10 A) [6].
diodes would almost certainly not exhibit any signs of damage.
4.2 Test methodology When a diode did not fail in reverse bias it was found to initially
Firstly, diodes were tested in the forward-biased position with experience a sharp voltage rise which stopped near 70 V. After
current impulses of increasing magnitude in order to determine this, the voltage would rise much more slowly to approximately
the peak current impulse at which they would repeatedly fail. 75 V over the next 20-30 µs. During this time, a reverse current
would begin to flow through the diode, peaking at near the 20 µs
Secondly, diodes were tested in the reverse-biased position with
mark, before gradually decaying to zero between 40 µs and
voltage impulses of increasing magnitude in order to determine
60 µs. After this reverse current had stopped flowing, the voltage
a point at which they would repeatedly fail. A fresh diode would
across the diode would begin to decay exponentially before
be subject to an initial impulse. Should this impulse result in a
eventually reaching zero. An example of this can be seen in Fig.
failure, the mode of the failure was noted (short or open-circuit).
16, which was the testing of diode 25 with a voltage impulse
Should this initial impulse not result in a failure, the diode would
which when open-circuit was found to be 209 V.
be subjected to further impulses of the same magnitude in order
to ascertain whether subsequent impulses would have a When a diode did fail in reverse bias it was found to experience
degrading effect on the electrical characteristics of the diode. the same initial voltage rise (followed by the same current rise)
as described above. The trend changes from that described above
4.3 Test results
if the reverse current reaches a critical value (around 60 A), after
4.3.1. Forward bias which the current begins to rise again (despite the rapid drop-off
Eight diodes were tested in the forward-biased position with in voltage) until it reaches a global maximum, before falling to
zero. This phenomenon can be seen in Fig. 17, Fig. 18, and Fig.
19. The test for diodes 15-17 used an impulse with a peak open-
circuit voltage of 240 V.

Fig. 18. The current and voltage waveforms measured


during the failure of diode 16 (reverse bias).

Fig. 16. The measured current and voltage waveforms from


impulse 1 over diode 25 (reverse bias).

Fig. 19. The current and voltage waveforms measured


during the failure of diode 17 (reverse bias).

Fig. 17. The current and voltage waveforms measured


during the failure of diode 15 (reverse bias).
It was of particular interest to determine whether a diode would
experience a degrading effect as it was subjected to subsequent
impulses of identical magnitude. Diode 38 was subjected to
multiple impulses (with an open circuit voltage of 160 V). Fig.
20 shows a plot of the voltage and current waveforms measured
and overlaid over each other after the first, tenth, fiftieth, one-
hundredth, and two-hundredth impulse. It can be seen that after
200 impulses there was no noticeable degradation. The diode
exhibited an unchanging forward voltage when tested after
subjection to the impulses. Fig. 20. An overlayed plot of the current and voltage
waveforms from impulses 1, 10, 50, 100, and 200 over diode
38 (tested in reverse bias).
5. Conclusions and recommendations 2. How altering the trace layout affects the induced
voltages and currents within the module.
5.1. Conclusions
3. How altering the module interconnections affects the
The investigation concludes that the Schottky diodes typically
induced voltages and currents within a string of
used in photovoltaic modules are quite robust when subject to
modules.
impulses in the forward bias direction.
6.2 High-voltage coupling testing
The diodes are, however, especially sensitive to impulses in the
reverse bias – a single impulse voltage as low as 200 V could The first and third experiments mentioned in Section 5.1 will be
potentially damage a diode, warranting replacement. recreated within a high-voltage test facility in order to validate
the simulation results. From this, useful recommendations will
If a diode survives a single impulse of a certain magnitude in
be made to manufacturers and plant constructors in order for
reverse bias, it tends be able to survive multiple impulses of that
them to minimize the damaging effects of nearby lightning
magnitude. There was no noticeable degradation of a diode’s
strikes on their photovoltaic modules and installations.
performance when subjecting it to up to 200 subsequent impulses
in reverse bias.
References
5.2. Recommendations
As the impulse generator began approaching its current
[1] KM Coetzer, AJ Rix, and PG Wiid, “Impulse Generator
limitations when testing diodes in forward bias before causing a Design to Investigate Indirect Lightning Strike Effects on
diode to fail, further testing should be performed with an impulse Utility-Scale Photovoltaic Installations,” in South African
generator of larger current capacity in order to investigate the Universities Power Engineering Conference 2018,
limitations of these diodes when subjected to large impulse Johannesburg, 2018.
currents in forward bias. [2] South African National Standards, SANS61215:2015
Crystalline silicon terrestrial photovoltaic (PV) modules -
Design qualification and type approval, Pretoria, 2015.
6. Further work
[3] South African National Standards, “SANS 60060-1:2011
6.1. Coupling simulations High-voltage test techniques Part 1: General definitions
and test requirements,” Pretoria, 2011.
In order to investigate the extent to which the magnetic field
[4] M. Maytum, “Impulse generators used for testing low-
emanating from a nearby lightning strike is able to couple into a
voltage equipment,” IEEE Power & Energy Society -
photovoltaic module (or string of modules), a high-resolution Surge Protective Devices Committee, New Jersey, 2012.
(< 1 mm accuracy) model of a photovoltaic module (including [5] South African National Standards, “SANS 61000-4-5:
the frame) has been generated. The intention is to use this model Testing and measuremnet techniques - Surge immunity
within an electromagnetics simulation package to evaluate: test,” Pretoria, 2006.
1. The degree to which lightning is able to couple into a [6] “Schottky Barrier Rectifiers - 10SQ030 thru 10SQ100,”
HY Electronic (Cayman) Limited, New Taipei City.
photovoltaic module as it is currently designed.

View publication stats

You might also like