You are on page 1of 5

A reference test setup and comparison between

different HPEM testing schemes


Tomas Hurtig Thorsten Pusch Martin Schaarschmidt
Defence & Security Systems and Fraunhofer INT Bundeswehr Research Intitute for
Technology Euskirchen, Germany Protective Technologies and NBC-
FOI – Swedish Defence Research thorsten.pusch@int.fraunhofer.de Protection
Agency Munster, Germany
tomas.hurtig@foi.se martinschaarschmidt@bundeswehr.org

Niklas Wellander
Mattias Elfsberg Michael Suhrke
Division of Command and Control
Defence & Security Systems and Fraunhofer INT
Systems
Technology Euskirchen, Germany
FOI – Swedish Defence Research
FOI – Swedish Defence Research michael.suhrke@int.fraunhofer.de
Agency
Agency
niklas.wellander@foi.se
mattias.elfsberg@foi.se

Abstract—Schemes for probing immunity of electronic II. REFERENCE TEST SETUP


devices against high power electromagnetics have been
developed and refined over the years. Several test environments The reference test setup developed by the authors has been
are available, some not yet fully acknowledged in standard described thoroughly in a previous paper [2], here, only an
documents or laboratory practice. In order to further their overview will be given.
development and to facilitate exchange across laboratories and
A. Hardware
methods, a new reference test setup representing a modern,
generic IT system has been designed. A comparison between The reference test setup is designed around a single board
failure modes and failure levels under two different testing computer (SBC) Raspberry PI model 3 B V1.2, with
schemes is presented. Testing has been performed using direct additional peripherals; a camera, an AD/DA converter and an
radiation in a TEM waveguide and also in a reverberation optical communication board. An analogue shield with four
chamber. Overall, the two methods of testing show very similar resistive voltage dividers provides reference voltage to the
results with respect to failure levels.
AD/DA converter via a ribbon cable. The optical
Keywords—HPEM, HPM, immunity testing, test methods, communication board makes it possible to communicate with
reference setup, standard development, measurement uncertainty the reference test setup during testing. The Raspberry Pi and
the peripheral components are mounted on a Styrofoam board
I. INTRODUCTION and powered by a rechargeable LiPo battery allowing for a
During the last decade there has been an increasing use of few hours’ autarkic use, cf. Fig. 1.
complex and networked electronics for all kinds of military B. Software and Monitoring
applications, as for example in command, control,
communications, computers, and intelligence (C4I) systems, The Raspberry Pi software consists of modules, one
weapons systems and even in ammunition. This trend is communicating with the AD/DA converter that monitors the
expected to continue to escalate in the near future. As a AD values and thus allowing to identify deviations in the AD
consequence, military missions become more and more readout. Another module communicates with the camera to
dependent on operational reliability of electronic systems and detect pixel distortions, and a third module establishes a serial
therefore increasingly vulnerable to electromagnetic attacks link via the optical communication board. Fibre optic cables
(EA) in general, and High Power Microwave (HPM) weapons connect to a computer outside the EM-testing environment
in particular. Qualification testing of military equipment with running a diagnostic software that is able to display errors
respect to this kind of intentional electromagnetic interference during testing. The computer stores all data for later analysis.
(IEMI) will be an important step in protecting one’s own
electronic systems. Central to this is the capacity to generate
sufficiently strong and varied test signals, a capacity that will
depend on the specific equipment available and will vary
between different laboratories.
In an effort to compare results of high level testing
between laboratories and test methods a reference test setup is
needed. The idea is not new and there have been suggestions
for such reference setups in the recent past [1].
The purpose of this paper is to compare results obtained
when performing susceptibility testing with HPEM on a
specific reference test setup through plane wave testing and
testing in a reverberation chamber.

978-1-7281-5579-1/20/$31.00 ©2020 IEEE


40 cm III. THE TWO TESTING SCHEMES
Results from two different testing schemes and two
different laboratories will be presented and compared. Plane
wave testing in a TEM waveguide over a wide range of
frequencies was performed by Fraunhofer INT and presented
earlier [1]. Here, we present a comparison of that data to a
more recent dataset obtained using a reverberating chamber
operated by FOI.
A. TEM waveguide testing
The first set of data is the results from testing performed
in a TEM waveguide at Fraunhofer INT. This experiment and
parts of the data have been presented previously. Testing was
performed using a power ramp approach starting at the lowest
stable field level and increased up to the maximum achievable
field strength.
The test signals followed a pulse pattern of 1µs pulse
length and 1 kHz pulse repetition rate and the duration of the
power ramp was 20 s. The reference test setup was oriented
vertically as shown in Fig. 1 with the electric field polarized
Fig. 1. Schematic overview of the test board with characteristic size indicated in the vertical direction. The uncertainty in the absolute
electric field levels in the TEM testing is approximately 30 %
C. Definition of error classes or 2.3 dB while the relative error is much smaller.
To facilitate the comparison of test results across different B. Reverberating chamber testing
laboratories and methods a common definition of error classes
The reverberating chamber (RC) used for the experiments
is needed. These classes have been defined as follows:
reported here is a Siepel EOLE 1000 with a lowest usable
1. One or more digitized analogue values leave the frequency (LUF) of 1 GHz.
predefined tolerance band around their nominal
Testing is performed according to IEC 61000-4-21 with a
values captured upfront exposure.
pulse length of 1 µs and a pulse repetition rate of 1 kHz to
2. A certain percentage of pixel values in the captured coincide with the TEM waveguide testing. The testing
camera frame lies outside a tolerance band spanning standard used gives an uncertainty of 3 dB with regard to the
the reference values of an averaged image taken electric field level. One frequency at a time is tested and after
before exposure. one complete revolution of the stirrer (20 steps) at a certain
field level the level is increased by 0.18 a.u. and the
3. The system does not communicate anymore via serial experiment is repeated until a certain maximum field value is
link for the remainder of exposure. reached. The dwell time at every stirrer position is 1 s and a
4. The system does not communicate anymore via serial complete test at a given frequency and field level is completed
link and a manual restart is required. in 120 s. A picture of the reference test setup inside the RC
can be seen in Fig. 2.
All errors are directly reported from the software
implemented in the Raspberry Pi to the monitoring software
during the tests. Error 3 is defined as loss of the serial
communication between the reference test setup and
monitoring computer during a test, but with recovery
occurring without any human intervention. Error 4 is defined
as the serial communication being lost during a test and not
recovering with less than a human intervention, a reboot.

TABLE 1 Definition of error classes

Error class Description


1 Analogue values out of tolerance
2 Video pixel values out of tolerance
3 Loss of serial link only under exposure
4 Loss of serial link requiring restart
Fig. 2. Reference test setup in RC. Raspberry Pi, AD/DA board and
communication board (A). Battery for power supply (B). Power supply board
(C). Camera under Styrofoam half sphere (D). Analogue shield with voltage
dividers for the AD/DA (E). Optical link (F). The Styrofoam half sphere
prevents the camera picture from changing due to changing light conditions
in the RC when the stirrer changes position.
After a complete test at one specific frequency, the field level
in the RC is again set to a level below the field value that gave
the first error. If no error is detected at this field it is assumed
that the lowest field giving the error was in fact found. With
this procedure, any degradation of the components regarding
susceptibility, caused by the microwave radiation, can be
detected.

In order to avoid any bias in the interpretation of results


FOI had no information about actual field levels used in the
experiments at Fraunhofer INT. The results from TEM
waveguide testing were transferred in actual field values after
the RC testing performed at FOI was completed.

IV. TEST RESULTS


Testing in the TEM waveguide was performed from
150 MHz up to 3.4 GHz. However, for the comparison
between TEM waveguide and RC testing we focus on a more Fig. 3. Absorption cross section ratio as measured in the RC (blue line) the
narrow frequency span, from 1 GHz to 2 GHz. Throughout the red crosses indicate at what frequencies the reference test setup has been
detailed testing a more limited band was needed and the lower subjected to high field testing.
end was set by the LUF of the RC and the upper end was
chosen to make it manageable to take very fine frequency Results from testing to the different error levels (1-4) are
steps. presented in Figs. 4-7. TEM waveguide data is given in black
and RC data is given in blue. The different symbols denote
For destructive testing absorption of energy into the different individual reference test setups. The TEM
electronic system is crucial and in many occasions a waveguide data is the result of testing six different individuals
measurement of the absorption cross section can be used to three times each and many of the test results for the same
identify frequencies of interest. Such a measurement was individual are closely clustered. Each individual is marked by
performed and the resulting absorption cross section ratio is a different symbol. No distinction is made between the three
plotted in Fig. 3. Here absorption cross section ratio is defined different tests, which is why there may be three symbols of the
as the ratio of the squared S-parameters in the loaded and the same kind at a certain frequency and for a certain error. The
empty RC in the following way. RC data is the result of testing three different individuals, one
2 time each.
〈 21 〉
empty
= 2 (1)
〈 21 〉
loaded

For this study the frequencies to study were already set by


the previous testing performed in the TEM waveguide and
these frequencies are indicated by red crosses in Fig. 3. From
the plot it can be seen that the current testing scheme from 1
to 2 GHz in 100 MHz steps will miss the most pronounced
absorption peak at about 1.75 GHz but that absorption does
vary substantially between the different frequencies tested.
It should be noted however that since no analysis of actual
coupling paths into the reference test setup has been
performed at this stage, there is no way to refer the different
absorption peaks to the different errors defined in TABLE 1.

Fig. 4. Test results for error 1: analogue values out of tolerance. TEM
waveguide (black) and RC (blue).
Fig. 5. Test results for error 2: video pixel values out of tolerance. TEM Fig. 7. Test results for error 4: loss of serial link requiring restart. TEM
waveguide (black) and RC (blue). waveguide (black) and RC (blue).

V. DISCUSSION AND INTEPRETATION OF RESULTS


In the following we discuss each error type separately and
identify possible sources of discrepancy between the two test
methods.
A. Error 1, Analogue values out of tolerance
In general, the results for error 1 in Fig. 4 show fairly good
agreement between the different methods. However, for some
frequencies, such as 1 GHz, only the RC testing caused the
error to occur. This discrepancy can have two basic
explanations; this particular error is not as sensitive to the
vertical polarization used in the TEM waveguide and/or the
spread for the particular error is large and it will occur but at
a field level outside the tested range.

B. Error 2, Video pixel values out of tolerance


This error show rather large difference between the two
Fig. 6. Test results for error 3: loss of serial link only under exposure. TEM testing methods, see Fig. 5. The error threshold is in general
waveguide (black) and RC (blue).
lower for the tests in the RC. It is worth noting however that
it were not the same individual setups that have been tested
at the two labs. Hence, it is possible that the lowest failure
levels recorded in the RC represents an outlier (blue crosses).
C. Error 3, Loss of serial link only during exposure
This error also show the largest difference between the
two testing methods, Fig. 6. The error occurs at significantly
fewer frequencies during the TEM-waveguide tests
compared to the RC tests. The electric field levels that are
needed to provoke the error are however similar for the two
methods.

D. Error 4, Loss of serial link requring restart


The permanent loss of the serial link error thresholds
show large similarities between the two methods, although
the error occurs at fewer frequencies during the TEM
waveguide tests compared to the RC tests, see Fig. 7. Both
methods show the same basic trend where susceptibility to
error 4 seems to be higher in the centre of the frequency band that is required to report errors is broken either temporarily, as
tested. in error 3, or for the remainder of the test as in error 4.
VI. CONCLUSIONS
E. General discussion This is the first experimental comparison using the
No direct relation between the absorption cross section reference test setup and overall the TEM waveguide and the
ratio and the susceptibility of the reference test setup could be RC testing show similar susceptibility levels for most errors.
found in the tests presented here. Further testing at one or more Future campaigns will give the possibility of comparing data
absorption peaks will reveal if the measured peaks actually that is more closely spaced in frequency and for a wider
show coupling to structures that induce one of the four frequency span. It is also interesting to investigate the effect
different errors defined. of varying dwell time in the RC and ramp up time in the TEM
The agreement between the two different methods seems waveguide and the effect of directivity and polarization
best for error 1 and 4 and less good for error 3. Again, the dependence. This will enable a closer analysis of the
difference between the results for error 3 is most likely differences between the methods.
explained by the fact that only one direction and polarization REFERENCES
is tested in the TEM waveguide. The fact that the average field
that is needed to cause any specific error across all frequencies
seems to be lower for the RC tests can probably be explained [1] F. Pythoud and E. Tas, “Design of a reference device for radiated
immunity inter-laboratory comparison”, 2017 International
by the fact that more angles of incidence and polarizations are Symposium on Electromagnetic Compatibility - EMC EUROPE, 2017,
tested in the RC. It should be noted however that the TEM doi=10.1109/EMCEurope.2017.8094672
waveguide data is more complete and a true comparison of [2] T. Pusch, et al., “A reference test setup to support research and
average levels is not possible until more RC-data is available. development of HPEM testing shcemes,” 2019 International
symposium on Electromagnetic Compatibility – EMC Europe, 2019
RC testing and comparisons to other methods has been doi=10.1109/EMCEurope.2019.8872105
discussed and analysed by several authors [4][5][6][7]. The [3] N. Wellander, et al., “Destructive Testing of Electronic Components
relation between maximum to average fields is discussed in Based on Absorption Cross Section RC Measurements,” 2019
[7] and also given as a table for different numbers of stirrer International symposium on Electromagnetic Compatibility – EMC
positions, N. In the RC testing presented here N=20 and the Europe, 2019
ratio between maximum and average total field amounts to doi=10.1109/EMCEurope.2019.8872128
| | ≈ 1.6, while the ratio between maximum and average for [4] M. Höijer, M. Bäckström, “How we confused the comparison between
high level radiated susceptibility measurements in the reverberation
a component of the field is | | ≈ 2.1. This indicates that the chamber and at the open area test site,” Electromagnetic Compatibility,
maximum field level that the DUT is subjected to can be 2003. EMC '03. 2003 IEEE International Symposium Volume: 2
higher in the RC tests than in the TEM waveguide. [5] Freyer, G.J., Bäckström, Mats, “Comparison of anechoic and
reverberation chamber coupling data as a function of directivity
There are also a few procedural issues in the RC testing pattern,” in International symposium on Electromagnetic
that can give rise to differences in the results between the Compatibility, 2000
methods. During the power ramp in the TEM waveguide each doi=10.1109/ISEMC.2000.874691
error is recorded immediately when it occurs, while in the RC [6] Freyer, G.J., Bäckström, Mats, “Comparison of anechoic and reverb-
a complete revolution of the stirrer (20 steps) is performed and eration chamber coupling data as a function of directivity pattern - Part
all errors recorded at the particular field level and frequency II,” in IEEE EMC international symposium, 2001 doi=
10.1109/ISEMC.2001.950637
are summed up and reported. This means that for one test at a
[7] J M Ladbury and G Koepke, “Reverberation Chamber Relationships:
specific field level and frequency, several different errors can Corrections and Improvements, Or: Three Wrong Can (almost) Make
be recorded. However if error 4 occurs at the beginning of the a Right,” in Proceedings of IEEE 1999 International Symposium on
test, no other errors will be recorded. I.e. when testing at high Electromagnetic Compatibility, Seattle, USA, August 2-6, 1999
field levels in the RC, error 1 and 2 are likely to happen but doi=10.1109/ISEMC.1999.812857
will be obscured by the fact that connection to the serial link

You might also like