Вы находитесь на странице: 1из 6

876 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 53, NO.

4, NOVEMBER 2011
Maximum Radiated Power Density From Electrically
Large SourcesComparing Probability Theory
and Measurements
Magnus H oijer, Member, IEEE
AbstractThe unintentionally radiatedpower density fromelec-
trically large electronic equipment shows a strong directional
dependence. The polarization of the emitted eld is also a pri-
ori unknown. Time and economy makes it prohibitive to measure
all orientations of the equipment. A simple probability model has
been proposed to quantify the magnitude of the effect: the power
received by a measurement antenna is exponentially distributed.
We verify that simple model by measurements.
Index TermsDirectivity, electrically large source, electromag-
netic statistics, emission testing, maximum value statistics.
I. INTRODUCTION
E
MISSIONmeasurements are important to ensure that elec-
tronic equipment does not radiate more than prescribed
regulations, which have been introduced as an important con-
stituent in ensuring electromagnetic compatibility. A method
often used is to measure the radiated electric eld at an open-
area test site or in a semianechoic chamber [1], [2]. Standard
methods to be used in a fully anechoic chamber are also devel-
oped [3]. For electrically large objects, i.e., the size of the object
is larger than the wavelength in issue, the radiated power den-
sity from the equipment under test (EUT) does show a strong
directional dependence [4][6]. This issue is further discussed
in the introduction of [5], and in [7] the issue of correlating
different test facilities is addressed. Here, we will not further
address this issue, but simply cite Wilson et al. [5], Even with
the automated help provided by turntables and motorized masts,
the effort required to check all possible source congurations is
prohibitive.
We recognize that the reverberation chamber can be used
to measure the total radiated power
1
[5], [8], [9]. Probability
methods developed by the National Institute of Standards and
Manuscript received February 23, 2011; revised April 14, 2011 and May
20, 2011; accepted July 11, 2011. Date of publication October 6, 2011; date of
current version November 18, 2011. This work was supported by the Swedish
Armed Forces.
The author is with the Swedish Defence Research Agency FOI, SE-581 11
Link oping, Sweden (e-mail: magnus@ieee.org).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TEMC.2011.2167514
1
Often the term total radiated power is used. The reason for including total
is probably to emphasize that the radiated power in all directions are to be
summed. However, the term total is superuous, as radiated power is only
one quantity, all power radiated from the EUT, in difference to electric eld or
power density which have a directional dependence.
Technology (NIST) [5], [9], [10], [11] can be used to estimate the
magnitude of the effect that the radiated power is not uniformly
distributed over all directions.
In this paper, we go a step further by explicitly giving the
distribution functions for howmuch higher than the average, the
radiated power density is in the direction and polarization with
highest power density. We also develop measurement theory,
perform measurements, and compare the measurements to the
developed theory.
II. PROBABILITY THEORY
The quantity directivity D is the radiated power density in one
specic direction r compared to the average taken over all di-
rections. We assume that we measure the radiated power density
from the EUT by a linear receiving antenna. The polarization
efciency p is a measure of the efciency of the coupling be-
tween the polarization e
t
of the radiated eld and the receiving
polarization e
r
of the antenna. If we assume that e
r
is xed, p
is a function of the rotation angle of the EUT around r. The
product of the directivity and the polarization efciency, named
partial directivity [12]
D
p
= D( r)p() (1)
does completely describe how the received power in the receiv-
ing antenna varies with the orientation of our EUT.
For an intentional radiator, e.g., an antenna, the maximum
value of the partial directivity is engineered to a specic value. A
typical value for, e.g., a horn antenna is 20 dBand for a parabolic
reector antenna 40 dB, but even higher values can be reached.
It is the possible size of the antenna relative to the wavelength of
the eld which put an upper bound on the maximum value. Our
EUTs are typically unintentional radiators and their maximum
partial directivity is lower, but also for unintentional radiators,
the partial directivity will increase with their size relative to the
wavelength of the emitted eld.
NIST [5], [9], [10], [11] has developed distribution functions
to estimate the magnitude of the partial directivity for an unin-
tentional radiator. Though they do not explicitly say it, the result
is astonishingly simple; the partial directivity is exponentially
distributed. The exponential distribution has only one parame-
ter, the expectation value. The expectation value of the partial
directivity is
ED
p
= ED Ep =
1
2
(2)
0018-9375/$26.00 2011 IEEE
H

OIJER: MAXIMUM RADIATED POWER DENSITY FROM ELECTRICALLY LARGE SOURCES 877
where the expectation value of the polarization efciency
Ep =
1
2
(3)
follows directly from assuming that the polarization of the ra-
diated eld from the EUT is completely random. We dene the
normalized partial directivity
2
as
X

=
D
p
ED
p

= 2D
p
(4)
and as stated earlier, it is exponentially distributed, and hence
its probability density function (pdf) is,
f
X
(x) = e
x
(5)
and cumulative distribution function (cdf) is
F
X
(x) =

x
0
f
X
(y)dy = 1 e
x
. (6)
It is to be observed that (5) and (6) are completely independent
of the physical properties of the EUT. The reason is that in [5]
and [9][11], it is assumed that the EUT radiates randomly.
However, as described at the beginning of this section, the par-
tial directivity varies with the radiation direction r and the rota-
tion angle of the EUT around r. Hence, the normalized partial
directivity can be seen as a random process
3
X = X( r, ). By
introducing a combined direction and rotation vector
v = [ r, ] (7)
the random process can be written as
X = X(v). (8)
Aspecic value of v we denote v
n
. To every v
n
corresponds
a random variable X
n
= X(v
n
) and to N different values of v
belongs the random variables X
1
, X
2
, . . . , X
N
. We are inter-
ested in the maximum value
Z(N)

= maxX
n

N
n=1
. (9)
We now assume that X is an ergodic stationary process. That
assumption implies that all samples in X
n

N
n=1
are indepen-
dent and equally distributed if the differences between the vec-
tors in the set v
n

N
n=1
are big enough. If these assumptions are
fullled, the cdf of Z is
F
Z
(z) = P(Z z) = P(All X
n
z) = (1 e
z
)
N
(10)
and the pdf
f
Z
(z)
dF
Z
(z)
dz
= N(1 e
z
)
N1
e
z
. (11)
The condition that the differences between the vectors in the
set v
n

N
n=1
need to be big enough is vague. However, it gives
an indication that the number of independent random variables
is upward limited. The derivation in [5] and [9][11] is based on
expanding the eld in spherical mode functions. For an EUT of
limited size, there are a limited number of non cutoff spherical
2
The sign

= denote dened as.
3
The concept of random processes is described in some text books on proba-
bility theory (see, e.g., [13]).
mode functions. Hence, there is a limited number of degrees of
freedom N
s
and we assume that number is also the supremum
(upward limit) for the number of possible independent random
variables. We also assume that the supremum of the maximum
normalized partial directivity, the true maximum normalized
partial directivity is

Z = Z(N
s
) . (12)
In [10], an explicit expression for the number of degrees of
freedom is given as
N
s
= 4ka|(ka| + 2) (13)
where ka| denotes the largest integer equal or smaller than ka,
where a is the radius of the minimum sphere circumscribing the
whole EUT,
k =
2

=
2
c
(14)
is the wave number, and , , and c denotes the wavelength,
frequency, and speed of light, respectively. Hence, once we
know the size of the EUT relative to the wavelength, we know
the distribution function for the maximum normalized partial
directivity.
III. MEASUREMENT THEORY
In a measurement, the measured maximumnormalized partial
directivity increases with the number of samples of the normal-
ized partial directivity. In Section II, we stated that the number of
independent samples is limited, and (13) gave us the maximum
number of independent samples N
s
. However, even if we make
N
s
independent measurements of the normalized partial direc-
tivity, we are unlikely to have measured the supremum value

Z
given by (12). We will now look further into this issue.
We will start by explicitly pointing out that the measured
maximum normalized partial directivity
Z
s
= Z
s
(N, N
s
) (15)
depends both on the number of measurement samples N as well
as the number of degrees of freedomN
s
in the modal expansion
of the radiated eld fromthe EUT. To emphasize the dependence
on N
s
, we have added the index s to Z.
We will assume for a while that the maximum value

Z has a
xed value z. It follows that Z
s
z. For all values z
s
z, we
assume that their relative likelihood remains the same, and by
renormalizing, so that the total probability is one [14, p. 17], we
get the cdf
F
Z
s
[

Z= z
(z
s
) =

0, z
s
0

1 e
z
s
1 e
z

N
, 0 < z
s
< z
1, z
s
z
(16)
878 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 53, NO. 4, NOVEMBER 2011
and pdf
f
Z
s
[

Z= z
(z
s
) =

0, z
s
0
N
e
z
s
1

1 e
z
s
1 e
z

N
, 0 < z
s
< z
0, z
s
z
.
(17)
We now release the condition that the EUTs maximum nor-
malized partial directivity has a xed value z, and instead recog-
nize that it is a randomvariable (

Z = Z(N
s
)) with pdf according
to (11). It follows from the law of total probability [15, p. 34]
that the joint pdf of Z
s
and

Z is
f
Z
s
,

Z
(z
s
, z) = f
Z
s
[

Z= z
(z
s
)f

Z
( z) (18)
and the pdf for Z
s
follows as the marginal probability
f
Z
s
(z
s
) =


0
f
Z
s
[

Z= z
(z
s
)f

Z
( z)d z . (19)
The calculation of the integral is straightforward and the pdf for
Z
s
is
f
Z
s
(z
s
)=

NN
s
NN
s
e
z
s

1 e
z
s

N
s
1

1 e
z
s

N1

N ,= N
s
N
2
e
z
s
(1 e
z
s
)
N1
ln (1 e
z
s
)
N = N
s
(20)
and the cdf for Z
s
is
F
Z
s
(z
s
)=

N
N N
s

1 e
z
s

N
s
+
N
s
N
s
N

1 e
z
s

N
N ,= N
s
(1 e
z
s
)
N
[1 Nln (1 e
z
s
)]
N = N
s
.
(21)
If N N
s
, Z
s
(N, N
s
) Z(N), which means that the mea-
sured maximum value depends on the number of measurement
samples, and if N N
s
, Z
s
(N, N
s
) Z(N
s
), which means
that the measured maximumvalue depends on the number given
by (13).
IV. MEASUREMENTS
A. Test Object
To perform emission measurements, we have used a
generic electronic device, called GENEC, as test object (see
Fig. 1). GENEC has been put at our disposal as part of a
GermanSwedish cooperation. GENEC is built as a metal-
lic cylinder with wings. The total length is 815.5 mm and
the diameter is 105 mm. At each wing, there is a wing slot,
size approximately 5 88 mm. The size of each wing is
50 120 mm. On the inside of the fuselage, the wing axes
are connected to the structure. The wings are attached in such
Fig. 1. Generic test object GENEC, reprint from [16] with permission of
DIEHL.
Fig. 2. GENEC: the wings, one wing slot, and the location of the two eld
probes. Probe A, located close to the wing slots, is not connected in this gure,
but Probe B, located closer to the electronics, is connected.
Fig. 3. Closeup of the eld probe and its SMA connector. The part under the
metal plate is placed inside GENEC.
a way that they do not have any electric connection to the fuse-
lage. Inside GENEC, there is a generic electronic device, which
incorporates the functional behavior of real missile electronics
with a minimum number of components. The electronic device
is kept in place inside GENEC with help of a plastic construc-
tion. Two eld probes, A and B, were installed inside GENEC
(see Fig. 2). Each probe consists of an SMA connector in which
a center conductor, of 20-mm length, is inserted (see Fig. 3).
The probes are regarded to be representative for typical cables
and wires.
B. Description of Measurements
To be able to get controllable and reliable measurement re-
sults, to automate the measurements, and to be able to span
H

OIJER: MAXIMUM RADIATED POWER DENSITY FROM ELECTRICALLY LARGE SOURCES 879
Fig. 4. To measure the radiated eld in many different directions, the test
object GENEC has been rotated in three planes.
the whole frequency domain from 500 MHz to 18 GHz, we
have performed the measurements in form of coupling mea-
surements. The probes mentioned in Section IV-A have been
excited, and the radiated eld from GENEC has been measured
by a linearly polarized horn antenna. The frequency domain
500 MHz to 18 GHz has been swept with 1401 frequency points.
GENEC has been rotated in three orthogonal planes, in accor-
dance to Fig. 4. For every plane, GENEC has been rotated
one revolution with an increment of 1

. For every position of


GENEC, the radiated eld has been measured in two orthogonal
polarizations: one parallel to the plane and one orthogonal to
the plane. Three planes, two orthogonal polarizations and two
probes, give us a total of 12 measurement series, and a total of
12 360 1401 = 6 052 320 measurement values.
We did not perform sampling over a full sphere, but over
planar cuts. That affects the number of degrees of freedom,
and the value given by (13) is no longer valid, but Koepke
et al. [10] also give an explicit expression for the number of
degrees freedoms taken over a planar cut
N
c
= 2(2ka| + 1) . (22)
In the results shown in Section V, N
c
has replaced N
s
in (21).
V. RESULTS
In Fig. 5, the cdf of the measured normalized partial directivity
is compared to the suggested cdf in (6). The agreement between
theory and measurements is very good, and it gives us a good
condence in assuming that the partial directivity of the EUT is
exponentially distributed.
As discussed earlier, we are interested in the maximumpartial
directivity, corresponding to the highest radiated power density
and hence the most severe case. In Figs. 6 and 7, the cdf of the
maximum normalized partial directivity is plotted. The dashed
lines show the results from our measurements, and the solid
lines are the cdf of (21). The curves are plotted for three different
frequencies, taken fromthe left, 0.5, 2, and 18 GHz. The number
of degrees of freedom, as given by (22), are 18, 70, and 618,
respectively. In Fig. 6, the maximum value is taken over 12
samples, and in Fig. 7, the maximum value is taken over 36
samples.
The agreement between experiments and the simple expres-
sions in (21) is good. It is tempting to try to describe the dif-
ference between measurements and theory in Figs. 6 and 7, but
we refrain from that. We have a limited amount of measurement
data, and hence, there is an uncertainty in the measurement re-
Fig. 5. Measurements of the cdf of the normalized partial directivity (X)
(dashed line) is in excellent agreement with the exponential distribution (solid
line). (The difference is actually so small that the measurement curve is hard to
observe.) In the measurement curve, all radiation directions, polarizations, and
frequencies described in Section IV-B have been included.
Fig. 6. Measurements of the cdf of the measured maximumnormalized partial
directivity Z
s
(dashed line) is in agreement with the theoretical prediction (solid
curve) according to Section III. The maximum value has been taken over 12
samples. The three sets of curves show the result for 0.5 GHz (left blue curves),
2 GHz (middle green curves), and 18 GHz (right red curves).
sults. The measurement results in Fig. 5 includes all 6 052 320
measurement values, but the maximum partial normalized di-
rectivity has a frequency dependence, and hence, we get in
principle a different curve for every new frequency. In Figs. 6
and 7, we show the result for three different frequencies, the
two extremes of our measurement, 0.5 and 18 GHz, and a fre-
quency in between, 2 GHz. Equation (21) is derived under the
assumption that all measurement values are independent. As we
assumed ergodicity in Section II, the measurement values are
independent if they are taken at radiation directions not too close
to each other. For the lowest frequency (0.5 GHz), (22) gives
us 18 degrees of freedom. By dividing 360

by 18, we get 20

,
and we take 20

as a measure of the smallest angle between two


independent measurement values.
To increase the number of measurement values to the experi-
mental curves in Figs. 6 and 7, we have included a few different
frequencies. It is justied by that the radiation pattern of X
varies faster with frequency than N
c
in (22) does, and that we
880 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 53, NO. 4, NOVEMBER 2011
Fig. 7. Similar graphs as in Fig. 6, but the maximum value is here taken over
36 samples.
assume that X is an ergodic process also as function of
frequency. In Figs. 6 and 7, we have in 0.5 GHz in-
cluded the frequencies, 500, 567.5 MHz, in 2 GHz,
1.9925, 2.0192, 2.0492, 2.0892 GHz, and in 18 GHz,
17.910, 17.940, 17.970, 18.000 GHz. All frequencies in each
set corresponds to 18, 70, and 618, respectively, degrees of
freedom. For the lowest frequencies, X varies so slowly with
frequency that measurement values for only two frequencies
could be included.
VI. CONCLUSION
The good agreement between theory and experiments gives
us a good condence in assuming that the partial directivity is
exponentially distributed. The maximum value distribution is
given by (11) and the belonging number of degrees of freedom
is given by (13).
The theory in Section III and the expression in (22) is solely
developed to be able to compare measurement with theory, but
the practical implication of this study is that two simple equa-
tions, (11) and (13), can be used to calculate the maximum
partial directivity of an unintentional radiator. No extensive ro-
tations of the EUT to check all possible source congurations
need to be performed.
By knowing the maximum partial directivity, we know the
maximum radiated power density relative to the average. The
maximum radiated power density is radiated in one specic
direction and polarization. The average value is the radiated
power divided by the whole solid angle (4) and multiplied
by the average polarization efciency (
1
2
). Hence, to be able
to quantify the maximum value, we need to know the radiated
power but, e.g., the reverberation chamber is an excellent facility
to measure the radiated power [8].
VII. DISCUSSION AND FUTURE OUTLOOK
The measurements in this paper are performed on one single
object. Further results from measurements on objects of dif-
ferent forms and complexities would be interesting. GENEC is
a rather simple object; it is a metal cylinder with two wings,
two wing slots, and a small circular aperture. One dimension is
substantially larger than the other two. Despite that, the theory,
developed by NIST [5], [9], [10], [11] and in this paper, seems
to work surprisingly well.
The ergodicity assumption has not been studied in detail in
this paper. We think that a further study of this topic will be
benecial.
ACKNOWLEDGMENT
The author would like to acknowledge the effort done by the
reviewers to improve the contents of this paper. Some measure-
ments have been performed by J. Lor en.
REFERENCES
[1] Specication for Radio Disturbance and Immunity Measuring Appara-
tus and Methods, Part 1: Radio Disturbance and Immunity Measuring
Apparatus, CISPR Standard 16-1, 1993.
[2] Specication for Radio Disturbance and Immunity Measuring Appara-
tus and Methods, Part 2: Radio Disturbance and Immunity Measuring
Apparatus, CISPR Standard 16-2, 1996.
[3] Electromagneic Compatibility (EMC) Part 4-22: Testing and Measure-
ment Techniques: Radiated Emissions and Immunity Measurements in
Fully Anechoic Rooms (FARs), IEC Standard 61 000-4-22, 2010.
[4] P. G. Landgren, Some directivity properties of test objects in the
microwave region, in Proc. IEEE Int. Symp. Electromagn. Compat.,
Montr eal, Canada, Aug. 1317, 2001, pp. 887891.
[5] P. F. Wilson, D. A. Hill, and C. L. Holloway, On determing the maximum
emissions from electrically large sources, IEEE Trans. Electromagn.
Compat., vol. 44, no. 1, pp. 7986, Feb. 2002.
[6] M. H oijer, M. B ackstr om, and J. Lor en, Angular patterns in low level
coupling measurements and in high level radiated susceptibility testing,
in Proc. Int. Zurich Symp. Tech. Exhib. Electromagn. Compat., Z urich,
Switzerland, Aug. 1820, 2003, pp. 347352.
[7] H. G. Krauth auser, Statistical analysis of correlation of emission limits for
established and alternative test sites, IEEE Trans. Electromagn. Compat.,
vol. 53. Accepted for publication.
[8] D. A. Hill, Electromagnetic theory of reverberation chambers, Natl. Inst.
Standards Technol. (NIST), U.S. Dept. of Commerce, Boulder, CO, NIST
Tech. Note 1506, Dec. 1998.
[9] P. Wilson, G. Koepke, J. Ladbury, and C. L. Holloway, Emission and im-
munity standards: Replacing eld-at-a-distance measurements with total-
radiated-power measurements, in Proc. IEEE Int. Symp. Electromagn.
Compat., Montr eal, Canada, Aug. 1317, 2001, pp. 964969.
[10] G. Koepke, D. Hill, and J. Ladburry, Directivity of the test device in
EMC measurements, in Proc. IEEE Int. Symp. Electromagn. Compat.,
Washington, DC, Aug. 2125, 2000, pp. 535539.
[11] P. F. Wilson, Advances in radiated EMC measurements techniques,
Radio Sci. Bull., vol. 311, pp. 6578, Dec. 2004.
[12] IEEE Standard Denitions of Terms for Antennas, IEEE Standard 145-
1983, 1983.
[13] G. R. Grimmett and D. R. Stirzaker, Probability and Random Processes.
Oxford, UK: Oxford Univ. Press, 1990.
[14] M. H oijer. (2006, Jun.). Radiated susceptibility test in reverberation
chamber, Swedish Defence Research Agency (FOI), Scientic Report
FOI-R2007SE, [Online]. Available: http://www.foi.se/
[15] G. Blom, Sannolikhetsteori och statistikteori med till ampningar. Lund,
Sweden: Studentlitteratur, 1980.
[16] T. Ehlen, Generic electronicGENECsystem description. DIEHL
GmbH & Co., R othenbach a d Pegnitz, Germany, DIEHL Rep.:
MM-ESI/EME/00/02-02.
H

OIJER: MAXIMUM RADIATED POWER DENSITY FROM ELECTRICALLY LARGE SOURCES 881
Magnus H oijer (M01) received the Ph.D. degree
in photonics and the M.Sc. degree in engineering
physics from the Royal Institute of Technology,
Stockholm, Sweden, in 1998 and 1992, respectively.
He is currently a Senior Scientist at the Swedish
Defence Research Agency, Link oping, Sweden. In
19992001, he was with Ericsson Saab Avionics,
Link oping. The work included numerical descrip-
tions of the Reverberation Chamber, and electro-
magnetic coupling to critical electronic equipment
in aircrafts. In 1991 and 1992, he was with Seiko
Instruments, Chiba, Japan, and Siemens Halbleiter, Munich, Germany, respec-
tively, investigating the ESD-endurance of transistors and constructing dynamic
random-access memory. His current research interests include polarization and
directivity dependences in radiated susceptibility testing, electromagnetic cou-
pling to electronics, and the reverberation chamber. Electromagnetic statistics
is his favorite interest.
Dr. H oijer has actively contributed to the international community as a
member of the International Electrotechnical Commission working group on
Reverberation Chamber, the IEEE EMC and the Swedish International Scien-
tic Radio Union Commission E. He has been the Swedish POC for EU FP 7
projects, and has been reviewer to national projects within the EU.

Вам также может понравиться