You are on page 1of 14

76 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO.

1, JANUARY 2015

Modeling of Noisy EM Field Propagation Using


Correlation Information
Johannes A. Russer, Member, IEEE, and Peter Russer, Life Fellow, IEEE

Abstract—In this paper, an efficient method for the numerical The importance of an electromagnetic theory-based in-depth
simulation of near- and far-field propagation of stochastic elec- analysis of wireless communication systems, taking impedance
tromagnetic (EM) fields is presented. The method is based on the matching, antenna mutual coupling, and different sources of
transformation of field correlation dyadics using Green's functions
or the field transfer functions computed for deterministic fields. noise into consideration, has already been stressed in [1] and [2].
The method accounts for arbitrary correlations between the noise Mutual coupling of antenna elements in array antennas intro-
radiation sources and allows to compute the spatial distribution of duces limitations on element efficiency and influences the noise
the spectral energy density of noisy electromagnetic sources. The penalty in multiantenna communication systems [3]. Integrated
introduced methodology can be combined with available electro-
circuits are frequently the main source of radiated emissions in
magnetic modeling tools. It is shown that the method of moments
can be applied to solve noisy electromagnetic field problems by net- electronic systems. Near-field characterization of radiating elec-
work methods applying correlation matrix techniques. Examples tromagnetic structures allows for a modeling of the electromag-
demonstrating the strong influence of the correlation between the netic field generated by these structures when embedded into a
sources on the spatial distribution of the radiated noise field are circuit or system environment, and therefore can be an efficient
presented.
tool for computer-aided manufacturing of electric and electronic
Index Terms—Electromagnetic interference, near-field scan- systems.
ning, noisy electromagnetic fields, stochastic electromagnetic The use of narrow-band random noise as carrier for infor-
fields.
mation transmission has been discussed for [4] and [5] modu-
lation systems. With stochastic carriers, a high signal-to-noise
I. INTRODUCTION ratio can only be achieved when the modulation bandwidth is
small compared with the carrier bandwidth. In the micrometer

N OISE in electromagnetic fields and radiated electromag- wavelength region with large available bandwidths, such ap-
netic interference (EMI) are ubiquitous limiting factors plications may be of interest. Infrared detectors for the long-
for the performance of wireless communication and electro- wave infrared range based on nanoelectronic thermocouples in-
magnetic sensor systems. The origins of electromagnetic noise tegrated in nanoantenna arrays may allow, due to their ultra-
may be thermal background noise, terrestrial noise, and elec- short response time in the picosecond regime, detection of the
tromagnetic interference radiated by electric and electronic sys- carrier phase after envelop demodulation [6], [7]. This plays a
tems. Electromagnetic noise and radiated electromagnetic in- role also for chaotic signals since they exhibit transversal co-
terference may either superimpose the wanted RF signal at the herence. Such detectors are attractive candidates for various
antenna or may interfere with electronic circuitry and by this applications including energy harvesting, target tracking, and
way degrade the signal-to-noise ratio. This degrades the signal thermal imaging. Since such systems will operate with chaotic
quality and requires an increase of signal power for achieving carriers, design methodologies accounting for stochastic elec-
the required signal quality. Therefore, the design of modern tromagnetic fields will be required.
electronic systems requires an accurate computer-aided design Due to the equivalence principle [8, p. 106], [9, p. 121], an
also with respect to optimization of the signal-to-noise ratios in equivalent source distribution determined by amplitude and
critical parts of the circuits. phase scanning of the tangential electric or magnetic field on
a surface enclosing the radiating structure is equivalent to the
internal sources and allows the modeling of the environmental
Manuscript received July 15, 2014; revised October 12, 2014, November 24, field. Numerous research groups have investigated the char-
2014, and November 25, 2014; accepted November 25, 2014. Date of publi- acterization of the radiated electromagnetic interference of
cation December 18, 2014; date of current version December 31, 2014. The devices by near-field measurements [10], [11]. The goal has
research discussed in this paper was supported by the German Research Foun-
dation (DFG) under the project RU314/45. This paper is an expanded version been to establish a model of the device under test that allows
from the IEEE MTT-S International Conference on Numerical Electromagnetic modeling the EMI radiated by the device under test into the
Modeling and Optimization for RF, Microwave, and Terahertz Applications, environment.
Pavia, Italy, May 14–16, 2014.
The authors are with the Institute for Nanoelectronics, Technische Universität So far, numerical tools for the modeling of complex electro-
München, 80333 München, Germany (e-mail: jrusser@tum.de; russer@tum. magnetic structures under stochastic excitation are not avail-
de). able. Therefore, these problems usually are treated using elec-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. tromagnetic full-wave simulators for deterministic electromag-
Digital Object Identifier 10.1109/TMTT.2014.2376962 netic fields. However, deterministic field modeling tools cannot

0018-9480 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 77

account for stochastic electromagnetic sources, which may be describing linear circuits and the corresponding equations
partially correlated or uncorrelated. In this paper, a methodology governing the relations between the correlation spectra of the
for the analytic and numerical modeling of stochastic electro- noise signals. In Section III, we introduce the scalar Green's
magnetic fields is presented. Arbitrary correlation of the sto- function method for the computation of a scalar field excited
chastic sources will be accounted for. Throughout this paper, by a scalar stochastic source field. We show that we have to
we assume that the considered stochastic signals and stochastic consider the correlations between the exciting field amplitudes
fields are Gaussian random processes [12, p. 154]. This guaran- at each pair of points. In Section IV, the description of vectorial
tees that auto- and cross-correlation functions and spectra, re- stochastic fields by correlation dyadics is introduced and the
spectively, yield a complete description of the stochastic signals. Green's function method is extended to vectorial stochastic
The scanning of the near-field amplitude distribution only is fields. In Section V, we extend the method of moments (MoM)
possible if phase and amplitude of the near-field signals are well to treat stochastic electromagnetic fields. The MoM allows to
defined, i.e., when phase and amplitude signals are fully corre- formulate the field problem as an algebraic network problem
lated for any pair of sample points in the near field. In general, and to apply the correlation matrix methods from network
this requirement is not fulfilled for the EMI radiated by elec- theory for the treatment of noisy electromagnetic fields. We
tronic circuits. Even if the signals flowing in an IC or in a circuit have applied the MoM for the formulation of the theory since
board are deterministic, the complexity of the interdependence the MoM is a versatile framework for the general formulation
of signals flowing in different parts of the circuit areas cannot be and algebraization of numerical electromagnetic methods as
revealed by the near-field measurement. Therefore, due to this we have already shown for the FDTD and TLM methods [23],
inherent lack of information, the near-field measurement has to [24]. In Sections VI and VII, we present numerical examples of
be performed, accounting for the statistical nature of the near the computation of near- and far-field distributions of electro-
field. magnetic fields excited by stochastic sources. In Section VIII,
Due to the lack of information about the sources of the ra- we apply the method of correlation dyadics to the numerical
diated EMI, the near field to be measured has to be treated as a computation of a two-dimensional image of stochastic noise
stochastic field. Stochastic electromagnetic fields with Gaussian sources. We discuss in Section IX a method for full character-
probability distribution can be described completely by the au- ization of stochastic fields by tangential two-point correlation
tocorrelation spectrum of each field variable and the cross-cor- near-field scanning in a plane of reference. Finally, we present
relation spectra of field variables at distinct points of observa- an example for object reconstruction from measured near-field
tion [13]–[19]. Characterization of a stochastic electromagnetic data in Section X.
field requires the sampling of the EM field in pairs of observa-
tion points and the determination of the cross-correlation func- II. NETWORK-ORIENTED NOISE MODELING
tions for all pairs of field samples. Numerical values of noise amplitudes cannot be specified for
It already has been shown for deterministic fields that net- stochastic signals. For numerical modeling of noisy circuits, one
work methods applied to electromagnetic field problems can re- has to deal with energy and power spectra [12], [25]. Hence,
duce the computational effort substantially [9], [20]. Network methods based on correlation matrices have been established
methods for the efficient solution of stochastic electromagnetic to cope with the numerical simulation of noisy linear circuits
field problems were introduced in [13]–[16]. A method for char- [26]–[31].
acterizing and modeling the noisy electromagnetic fields ra- In the superposition of random signals, their correlation is
diated from complex circuit boards using a simplified repre- playing a decisive role. If a signal is superimposed with
sentation of correlated dipoles has been described in [21]. De- a time-delayed signal , where the time-delay of the
pending on the number of statistically independent field sources, second signal is given by , we are interested in the correlation
the eigenvalue decomposition of the correlation matrix yields a of and . We define the correlation function
compact description of the measured EM noise field [22]. of the stationary random signals and by
In this work, an efficient method for the numerical simulation
(1)
of near- and far-field propagation of stochastic electromagnetic
fields is presented. The basic idea was introduced in [17], where the brackets denote the ensemble average [32].
and this manuscript expands by completing the theoretical For stationary random signals, the correlation function
framework in deriving the formulas for the electric as well as is independent of . Furthermore, for ergodic random signals we
the magnetic fields computed in both cases, either from electric can substitute the ensemble mean value by the time mean value,
or magnetic excitation currents. Additional theoretic and ex- and vice versa. We can write
perimental examples are given showing the application of the
methodology for the near- and far-field modeling of stochastic
(2)
electromagnetic fields. The method is based on the transfor-
mation of field correlation dyadics using Green's functions or
the field transfer functions computed for deterministic fields. For , we obtain the autocorrelation function
In Section II, we recall the correlation-matrix-based methods
for network-oriented noise modeling. We show the simple and
general correspondence between the linear network equations (3)

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
78 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

Let us first have a look at the noise analysis of linear net- The power spectra of the signals can be calculated from the cor-
works and start by investigating which statements can be made relation matrix. The correlation spectra of noise signals may
concerning the spectrum of a random signal. Since a stationary be summarized in the quadratic correlation matrix
noise signal is not decaying for , the noise signal is not
square integrable or absolute integrable, and therefore an ampli-
tude spectrum of a stationary noise signal does not exist. How-
ever, we can consider a sample of a noise signal with finite ex- (12)
.. ..
tension in time. We consider the segment of the stationary . .
noise signal , given by
for
(4) which may be also obtained from
for

We call the signal the cut-time signal or time-windowed (13)


signal. In general, the Fourier transform of the time-win-
dowed signal exists and is given by The superscript denotes the Hermitian conjugate.
The equations describing linear networks have the form
(5)
(14)

where is the signal transfer matrix relating the input and


The corresponding inverse Fourier transform is
output signal vectors and of the linear network
under consideration. Usually, is an impedance-, admittance-,
(6) or scattering matrix. This yields the relation between the corre-
lation matrices

We can also represent the correlation function by (15)

The correlation matrix is derived from the amplitude


(7) vector of the time-windowed signal spectra . The gen-
eral relation between (14) and (15) is the rule for deriving the
bilinear transformation equations for the correlation matrices
The Fourier transform of the correlation function is the correla- from the linear equations for the complex signal amplitude.
tion spectrum
III. SCALAR STOCHASTIC FIELDS
(8)
We already have presented in [14] a method for the numer-
ical simulation of stochastic electromagnetic fields accounting
for the correlations of sources and applied this method for far-
(9)
field computation of the spectral electric-field energy density. In
this paper, we present the general methodology, including also
is called the autocorrelation spectrum, whereas near-field computation and presenting the complete formulas
with is called the cross-correlation spectrum. The cross- for electric- and magnetic-field computation, both in the case
correlation spectrum also is given by of excitation by electric, as well as magnetic, currents of arbi-
trary mutual correlation.
(10) We consider the tangential field in an aperture de-
picted in Fig. 1. This aperture field is the source yielding a far
An ensemble average is taken, indicated by the brackets . field . Let the far field be related to the near field
The ensemble average has to be taken before performing the via
limiting process . The autocorrelation function
represents the spectral energy density of the signal . We (16)
note that .
From this time-windowed signal, we can compute the spec-
trum . We summarize the spectra of time-windowed where the scalar Green's function is given by
signals flowing in a network in the vector
(17)

.. (11)
. Whereas amplitudes cannot be defined for stochastic quantities,
we can introduce power and energy spectra for their description.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 79

The function describes the scalar field in time domain.


The Fourier transform for a deterministic field would be

(18)

A correlation function can be defined for the field


function by

(19) Fig. 1. Near and far field.

More specifically, this function is termed the autocorrelation


Hence, we can obtain from the correlation spectrum of
function of the field for and cross-correlation function
the source field the field correlation spectrum
for [12], [14].
for the field amplitudes at the observation points
Now, let us look at the Fourier transform of a sample of the
and . Knowledge of the spatial distribution of the spectral
noise field, which is of finite length in time . This Fourier
energy density of the source is not sufficient in order to compute
transform exists and is given by
the field excited by a distribution of stochastic sources; also full
information regarding the cross correlation of the source field
(20) amplitudes at any pair of points and is necessary.
Consider the correlation functions defined in (23) and (24),
respectively. If the field spectrum describes a de-
where the subscript denotes the time-windowed signal, de- terministic signal, the ensemble averaging described by the
fined by (4). We can express (19) also by brackets can be omitted. In that case, the correlation spec-
trum can be factorized into the factors
and , i.e., it can be represented by a product of
(21)
functions with a space dependence on and only. In the
general case of a stochastic field, such a factorization is not
possible.
The Fourier transform of is the correlation spec-
trum IV. VECTORIAL STOCHASTIC FIELDS
(22) In the case of EMI problems, we have to deal with stochastic
electromagnetic fields. If a stochastic electromagnetic field
originates from a sufficiently large number of statistically inde-
Alternatively, we can take the spectra of the time-
pendent processes, the field amplitudes will exhibit a Gaussian
windowed fields and find the correlation spectra as
probability distribution due to the central limit theorem. A
Gaussian process can be described completely by its mean value
(23) and its second-order moments. These second-order moments
are represented by the respective auto- and cross-correlation
functions.
where an ensemble average is taken indicated by the brackets.
The stochastic electric and magnetic fields can be described
To express the correlation spectrum of the far field
by the dyadic
as a function of the correlation spectrum of the
near field, given by
(26a)

(26b)
(24)

where and are the Fourier transforms of


we insert (16) into (23) and obtain the time-windowed electric- and magnetic-field vectors, and the
symbol denotes the Hermitian conjugate. The components of
the dyadic are summarized in the matrix

(27)
(25)

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
80 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

The correlation field of stochastic source currents at the points where is the free-space wave impedance, and
and can be described by the dyadic the denotes the transpose of the vector.
Consider a magnetic current density vector de-
(28) scribing the source of the electromagnetic field. The magnetic
current is the source of the dual vector potential , given
Consider a current density vector describing the source by
of the electromagnetic field. The magnetic vector potential
is given by (38)
(29)
With
With
(39)
(30)
we obtain the magnetic field excited from as
we obtain the magnetic field excited from as
(40)
(31)

where is the Green's dyadic relating the excited


where is the Green's dyadic relating the excited electric field to the vector potential , and the in-
magnetic field to the source current density , tegration is extended over the whole volume , where
and the integration is extended over the whole volume , where is nonvanishing. is given as
is nonvanishing [33, p. 306]. Accounting also for the
near-field contributions, the Green's dyadic is given by [34, pp. (41)
282–284]

(32) From (40), we obtain with

(42)
with the identity matrix , , the phase
coefficient , the free-space light velocity , and
the magnetic field excited from as

(33) (43)

From (31), we obtain with


where is the Green's dyadic relating the ex-
(34) cited magnetic field to the vector potential ,
and the integration is extended over the whole volume , where
is nonvanishing. is given as
the electric field excited from as

(35) (44)

A similar approach was used in [35] for representing the EMI


where is the Green's dyadic relating the excited sources in a printed circuit board by magnetic dipoles.
electric field to the source current density , and Neglecting in the terms going to zero with higher
the integration is extended over the whole volume , where order than , we obtain the far-field approximation
is nonvanishing.
(45)
(36)

with of the Green's function.


From (26a), (28), and (35), we obtain
(37a)

(37b)
(46)

(37c) Using the relation above, the correlation dyadic of the electric
field can be derived from the correlation dyadic

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 81

of the source currents. The spectral electric en- and can write (51) as
ergy density is obtained as
(55)

(47) We expand the correlation dyadics and


, introduced in (28) and (26a), into basis func-
where is the permittivity of the medium. tions. We obtain the matrix elements and
of the correlation matrices and , given by

V. APPLICATION OF THE METHOD OF MOMENTS (56a)


In the following, we use the MoM [36] to convert the electro-
magnetic-field problem formulated by integral equations into a
(56b)
network problem represented by algebraic equations. We apply
the MoM to compute the integral (35). At first, we expand the
current density and the electric field into or- These matrix elements can be summarized in the matrices
thonormal vectorial basis functions . This yields

(48a) .. .. .. (57a)
. . .

(48b)
.. .. .. (57b)
. . .
The expansion coefficients and can be considered
as generalized currents and voltages, respectively. The vectorial
basis functions fulfill the orthonormality relation We can also obtain these correlation matrices from the time-
windowed current and voltage amplitudes and
(49) via

(58a)
Inserting the series expansions (48a) and (48b) into the integral
equation (35) yields (58b)

From (55), (58a), and (58b), we obtain


(50) (59)
Multiplying both sides of this equation from the left with the This is the fundamental equation describing linear noisy net-
Hermitian conjugate of the vectorial basis function and works, already given in [28]–[31] and extended to the descrip-
integrating under consideration of (49) over and , we obtain tion of noisy EM fields in [13]–[15]. With this equation, we can
the linear system of equations compute the correlation matrix of the series expansion
coefficients of the electric-field dyadic from the se-
(51)
ries expansion coefficients of the source current den-
sity. We note that the transformation matrix , which describes
The matrix elements are given by the electric-field response to the excitation current, is the same
as that occurring in the deterministic problem. Therefore, the re-
(52) sponse functions computed for deterministic field problems can
also be applied to stochastic field problems.
Interpreting as the matrix elements of an impedance VI. EXAMPLE: NEAR-FIELD DISTRIBUTION EXCITED BY
matrix, the MoM gives a network description of the field SOURCES IN A PLANE
problem. For a chosen dimension of the series expansions
Due to the increasing operating frequencies and bandwidths
(48a) and (48b), we introduce the generalized current and
and decreasing signal amplitudes in modern electronic systems,
voltage vectors
the EMI radiated from integrated circuits and circuit boards of
(53a) the electronic systems may affect the whole system. Internal
(53b) shielding increases weight, volume, and costs of the system and
may also yield signal degradation. Therefore, an accurate mod-
and the impedance matrix eling of the expected EMI spectral energy distribution inside
a piece of electronic equipment under design is a requirement
.. .. for modern computer-aided design and fabrication of electronic
.. (54)
. . . systems. The EMI data of the components for modeling the EMI
distribution inside an electronic apparatus may be taken either

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
82 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

Fig. 2. Planar array of stochastic sources sampled at a sampling plane


parallel to the source plane. Fig. 3. Three-dimensional logarithmic plot of the spectral electric energy den-
sity for 100 mm, 100 mm 1 mm and
0.1 mm .

from theoretical knowledge of the components or from mea-


with the correlation matrix elements of the dipole currents
surement.
In the following, we present a numerical example of the sto- (63)
chastic near-field radiated by a number of stochastic sources.
We consider the near-field distribution of stochastic Hertzian Inserting (62) into (46), we obtain the correlation field of the
dipole sources with arbitrary orientation arranged in a plane. stochastic electric-field vectors at the points and , as fol-
Fig. 2 shows the sources arranged in the plane lows:
at positions with , and where the
superscript denotes the transpose. Assuming for the th dipole
an impressed current , an infinitesimal length and an
angular orientation characterized by the azimuth and pole an-
gles and , respectively, the current density of all (64)
dipoles is given by In the following, we compute numerically the correlation
dyadic of the field excited by the set of dipoles in pairs of points
and in the scan plane, as depicted in Fig. 2. We assume
(60) all current sources to be mutually uncorrelated, i.e., statistically
independent and to exhibit equal normalized amplitude 1. This
yields
where is the three-dimensional Dirac delta distribution,
and the normalized orientation vector , given by (65)

For this case, (64) reduces to

(61)

(66)
describes the angular orientation of the th dipole. We assume Fig. 3 shows the spectral electric energy density magnitude
the dipoles to be arranged above a perfectly conducting layer logarithmic plot computed for a scanning plane at a distance
PEC positioned at 1 mm, as shown in Fig. 2. In the 1 mm from the source plane for 0.5 mm, and
computation, this layer will be considered by mirror dipoles. 0.1 mm , i.e., 6.28 cm. To collect these data by
The electric field is scanned in the plane with vec- measurement, only a single probe would be required. The plot
torial electric-field probes sensing the - and -components of shows the spectral electric energy density distribution without
the electric field. To measure the electric-field correlation in the containing any information concerning the correlation between
plane , two probes and positioned at and are used the field at different scan points. The spectral electric energy
simultaneously. In Section IX, a near-field scanning system for density distribution exhibits strong maxima when the probe is
performing these measurements will be discussed. positioned opposite to one of the dipoles.
Inserting (60) into (28) yields the correlation field of sto- Fig. 4 shows the 3D logarithmic plot of the magnitude
chastic source currents of electric-field correlation density, where
probe 1 with the coordinates scans over the area
100 mm 100 mm , whereas probe 2 is positioned as a
reference at the fixed point 20 30 1 mm just
over the source at . This yields a strong maximum of the
(62) scan at . The magnitude of electric-field correlation density

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 83

Fig. 4. Three-dimensional logarithmic plot of the magnitude of the electric- Fig. 6. Three-dimensional logarithmic plot of the magnitude of the electric-
field correlation density for 100 mm 100 mm , field correlation density for and are varied independently,
1 mm, fixed reference point 20 30 1 mm, and 1 mm, and 0.1 mm .
0.1 mm .

Fig. 5. Three-dimensional logarithmic plot of the magnitude of the electric- Fig. 7. Three-dimensional logarithmic plot of the magnitude of the electric-
field correlation density for 100 mm 100 mm , field correlation density for fully correlated in-phase sources
1 mm, fixed reference point 100 100 1 mm, and of equal amplitude and for 100 mm 100 mm , 1 mm, fixed
0.1 mm . reference point 20 30 1 mm, and 0.1 mm .

exhibits a strong primary peak when the position of probe 1 For this case, (64) yields the electric-field correlation dyadic
coincides with that of probe 2 and smaller peaks when probe 1
is positioned opposite to one of the other dipoles.
In Fig. 5, the 3D logarithmic plot of the magnitude of
electric-field correlation density is shown (68)
for the case where the reference is taken at the fixed point Computing the 3D logarithmic plot of the magnitude of elec-
100 100 1 mm, which is remote from all sources. tric-field correlation density for the reference
This reference is correlated with all sources and therefore pro- taken at the fixed point 20 30 1 mm over the
vides a suitable reference for the correlation field measurement. source at now yields the result shown in Fig. 7. Although
Due to the position of probe 2, remote from all the dipoles, the the reference probe is positioned opposite the source dipole at
fixed probe 2 may be a suitable reference for the correlation 20 30 0 mm, as in the case of Fig. 4, the maxima of
measurements. the magnitude of electric-field correlation density are of equal
Fig. 6 shows the magnitude of the electric-field corre- height, as in the case of the remote probe in Fig. 5. Due to the
lation density , where and are varied full correlation of all sources, any position of the reference
independently and 30 mm corresponding to the probe would give a suitable reference. In this case, the reference
-coordinates of the dipoles at 20 30 0 mm and probe 2 could be in a fixed position, and scanning with probe 1
at 30 30 0 mm. This plot shows strong main maxima only would yield the complete information.
when both probes are above either or and smaller relative Fig. 8 shows, like Fig. 6, the magnitude of the electric-field
maxima when only one probe is above or . correlation density , where and are varied
Now, let us assume full correlation of all sources and further- independently and 30 mm corresponding to the
more all sources to be in phase and to exhibit same amplitude. -coordinates of the dipoles at 20 30 0 mm and at
In this case, the correlation matrix of source currents 30 30 0 mm. However, Fig. 8 shows the electric-field cor-
is given by relation density for in-phase fully correlated sources and ,
whereas Fig. 6 shows the case where and are statistically
(67) independent. The plot in Fig. 8 shows strong main maxima not

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
84 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

Fig. 10. Nonuniform linear array of Hertzian dipoles positioned at


.

In the far field, the term on the right-hand side of (69) can be
approximated in the denominators by . Introducing the vec-
tors

Fig. 8. Three-dimensional logarithmic plot of the magnitude of the electric-


field correlation density for fully correlated in-phase sources of (71a)
equal amplitude and for and are varied independently, 1 mm, and
0.1 mm . (71b)

we obtain with (69) the electric field in the far field of the an-
tenna array

(72)

To describe a stochastic excitation of the antenna elements, we


Fig. 9. Linear array of 9 near-field cells. introduce the correlation matrix describing the antenna feed cur-
rents

(73)
only when both probes are above either or but also when
one probe is above and the other probe is above . The correlation function describing the spectral electrical en-
The graphic representations of this section gave only a weak ergy density in the far field on a sphere of given is
impression of the information contained in the field correlation
dyadics. The field correlation dyadics contain full amplitude
and relative phase information, allowing the application of the (74)
equivalence principle [8], [9] for computing the environmental Inserting (72) and (73) into (74) yields
stochastic field on the basis of known tangential electric or mag-
netic field on the boundary surface of a volume enclosing the (75)
stochastic EM sources.
We note that in the equatorial plane , we obtain
. Inserting (75) into (47) yields the
VII. EXAMPLE: FAR-FIELD DEPENDENCE ON THE spectral electric energy density of the far field
NEAR-FIELD CORRELATION
(76)
As an example, let us consider a linear antenna array formed
by a number of linear dipoles of uniform shape arranged in Let us consider a nonuniform linear array of parallel Hertzian
parallel with equal mutual distance, as shown in Fig. 9 [9, p. dipoles positioned at , as
575]. The far field exhibits only the -component of the depicted in Fig. 10. Fig. 11 shows the polar plot of the angular
electrical field and a -component of the magnetic field, where distribution of the magnitude of the electric energy
and are the polar and azimuthal angles [9, p. 575]. For ex- density in the plane and at a given distance from the origin
citation of the th dipole with the current , is given by for the following five cases.
1) All Hertzian dipoles are excited by fully correlated currents
of equal amplitude and equal phase. The correlation matrix
(69) of the excitation currents is given by

is the wave impedance of the free space, and is


the radiation characteristics of the single dipole. The distances (77)
between the center of the th dipole and the far-field observation
is
The curve (1) in Fig. 11 shows the angular far-field distri-
(70) bution of the radiated electric energy density. The far-field

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 85

Fig. 11. Polar plot of the angular distribution of the magnitude


electric-field spectral energy density in the far field.

Fig. 12. Magnitude of the correlation spectrum of the electric-


field amplitudes measured in the far field at angular positions and for
distribution exhibits a strong maximum in direction uncorrelated excitation currents.
.
2) All excitation currents are mutually uncorrelated but ex-
hibit equal rms values. The correlation matrix of the exci-
tation currents is

(78)

The angular far-field distribution of the electric energy den-


sity shown in curve (2) is uniform.
3) The excitation currents are a weighted superposition of cor-
Fig. 13. Magnitude of the correlation spectrum of the electric-
related currents represented by and uncorrelated cur- field amplitudes measured in the far field at angular positions and for
rents . The correlation matrix of the excitation currents partially correlated excitation currents.
is
For the case of two antennas only, separated by a distance
(79) , the magnitude of the correlation spectrum
is plotted in Fig. 12 for the case of uncorrelated sources, com-
In this case, the far-field distribution shown in curve (4) parable to case 2) above. For the case of partially correlated
exhibits a narrow maximum in direction . How- sources, as considered in case 3) above, the magnitude of the
ever, incoherent radiation with broad angular distribution correlation spectrum is plotted in Fig. 13.
is superimposed.
4) In this example, the Hertzian dipoles again are excited with VIII. EXAMPLE: IMAGING OF STOCHASTIC
mutually correlated currents of equal amplitude. However, SOURCES OF RADIATION
the phases of the currents are given by , ,
We already have discussed the imaging of stationary noise
. This yields the correlation matrix
and EMI scenarios by two-point sampling in [37] and [38]. Sam-
pling of the two-point correlations in a Fourier plane of a virtual
imaging system yields the information required for the digital
reconstruction of the image of the sources. In the paraxial ap-
proximation, a thin lens with focal length , thickness , and
refractive index performs a phase delay
(80)
5) In this case, the Hertzian dipoles are excited with a super- (82)
position of correlated currents with nonuniform phase ac-
cording to (4) and uncorrelated currents according to (2), and, by this, transforms the field in front of the lens into the
as follows: field behind the lens [39, p. 80]. The image is computed from
the two-point scans in the Fourier plane by compensating this
(81) phase delay and Fourier transformation of the scanned correla-
tion dyadics [37], [38].
In this case, the angular distribution (5) has a tilted max- Fig. 14 shows the intensity distribution of five uncorrelated
imum with a strong superimposed noise floor with broad stochastic sources in the object plane. Due to the incoherence
angular distribution. of the sources, the far-field intensity distribution in the Fourier

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
86 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

Fig. 14. Intensity distribution in the object plane [37].

Fig. 16. Schematic drawing of the near-field scanning system.

meet the requirements for high sensitivity and high spatial res-
olution [43]–[45]. The measurement time could be reduced by
using of a switched matrix array of antenna elements for scan-
ning. Such an array of antenna elements already has been de-
scribed for one scan channel in [46].
For a number of sampling points chosen on the scan plane,
the probes are positioned so that the fields in all possible pairs
of sampling points are sensed. Since every vectorial probe has
Fig. 15. Reconstruction of the intensity distribution in the object plane [37].
two outputs sampling points yield output signals. Consid-
ering these signals to be stochastic signals, we can form au-
plane is constant, and we need only to sample the field correla- tocorrelation functions and cross-correlation func-
tions between the chosen set of sampling points and one selected tions from these sensed signals. The positioning of the probes is
point of reference. controlled by the central control and processing unit. The switch
We have chosen a set of 10 10 sampling points. A two-di- selects the polarization of each probe. Furthermore, the switch
mensional rectangular windowing in the Fourier plane was per- also allows to feed both output signals of the same probe into
formed. The reconstructed image of the source distribution is channels 1 and 2 of the system. By this way also the cross-cor-
depicted in Fig. 15. Using only 10 10 sampling points resulted relation function of the two polarization amplitudes at one sam-
in a broadening of the source width. However, source localiza- pling point can be determined. The two channels are sampled
tion and separation is still possible. and digitized in the analog-to-digital converters (ADC) 1 and
2. The digitized signals are further processed in the central pro-
cessing and control unit. In the central processing and control
IX. NEAR-FIELD SCANNING
unit, the sampled signals are processed by fast Fourier trans-
If we want to fully characterize the near field, we have to scan form. Autocorrelation and cross-correlation spectra are com-
the tangential electric- or magnetic-field components on a sur- puted and stored in the output and display unit. From these auto-
face enclosing the device under test. We have to determine the and cross-correlation functions, the angular distribution of the
autocorrelation functions of the field in all sampling points, and radiated spectral power density and the spatial distribution of
the cross-correlation functions of the field amplitudes have to the spectral energy density can be computed.
be determined for each pair of sampling points. We also have to Time-domain electromagnetic interference measurement sys-
determine the cross-correlation functions of both polarizations tems are up to 64 000 times faster than conventional hetero-
in one sampling point. dyne receiver-based systems [47]–[49]. Two-channel time-do-
Fig. 16 shows the schematic of a near-field scanner, allowing main EMI measurement systems already have been realized for
full characterization of the stochastic electric or magnetic near ambient noise cancellation [50]–[52]. Such two-channel mea-
field. Its two vectorial electric- or magnetic-field probes are po- surement systems can be easily adapted for correlation mea-
sitioned by the measurement system. The signal processing unit surements and allow to handle the large amount of data. Tan-
comprises a two-channel system, which processes the signals gential vectorial scanning of a near-field array in two polariza-
delivered by the field probes. From the sampled and digitized tions with 10 10 sampling points would require the acquisi-
data, autocorrelation and cross-correlation functions of the spa- tion of 200 autocorrelation functions and 19 900 cross-correla-
tially sampled field probe signals are computed. tion functions. Due to the short measurement times of modern
Probes for measuring EMI radiated on integrated circuit or time-domain EMI measurement systems and powerful FPGA
miniature package and board level are described in [40]–[42]. postprocessing in the systems, such a high amount of data ac-
Submillimeter-size active integrated sensors with on-chip gain quisition will be feasible.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 87

deterministic field Green's functions. The spatial and temporal


correlations of the stochastic source distributions are described
by the correlation dyadics.
Converting the field problem to a network problem allows
the modeling of stochastic electromagnetic fields by applying
correlation matrix methods. One general approach to do this is
the application of the method of moments (MoM). Near-field
scanning of the radiated electromagnetic interference (EMI) is
motivated by the desire to analyze and predict the EMI radiated
from electric or electronic components or systems into space or
a given environment, as well as to compute images from the
radiating objects.
Fig. 17. Measured near-field magnitude distribution of the correlation spec- As examples, we have investigated the correlation spectra
trum [53]. of electric-field samples in the far-field radiated by arrays of
Hertzian dipoles. In the case of incoherent excitation, the ar-
rays' Hertzian dipoles radiate with uniform azimuthal distribu-
tion of the spectral energy density. The cross-correlation spec-
trum of the amplitudes in different angular positions, however,
exhibit a strong angular dependence. This shows that the char-
acterization of incoherent electromagnetic field requires the in-
formation about the correlation between the field values at dif-
ferent points. Furthermore, we have shown the application of
two-point scanning for the construction of images of stochastic
source demonstration and demonstrated the image reconstruc-
tion from scanned data. Finally, we have shown an experimental
example for the reconstruction of the object distribution from
scanned near-field correlation data.
Considering the power of modern time-domain EMI
Fig. 18. Reconstructed magnitude distribution of the field in the object area measurement systems in connection with future switched
[53].
two-channel antenna array scanners, we expect the develop-
ment of near-field stochastic field measurement systems that
X. EXAMPLE: OBJECT RECONSTRUCTION FROM MEASURED will facilitate an accurate modeling of the EMI radiated into
NEAR-FIELD DATA complex environments.
Measurements were performed with a near-field scanning ex- ACKNOWLEDGMENT
perimental setup as shown in Fig. 16 [53]. The radiated EMI of
two small electromotors placed at 10 cm and 10 cm The authors gratefully acknowledge the German Research
and both at 0 cm was scanned with two electric dipole Foundation (DFG) for the support of this research.
probes in one dimension with five equidistant sampling points
REFERENCES
at 0 cm and 18 cm along the -axis in 5-cm intervals
[1] M. Ivrlac and J. Nossek, “Toward a circuit theory of communication,”
by measuring the -component of the electric field only. The IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 57, no. 7, pp. 1663–1683,
signals of the two dipole probes were recorded by a Tektronix Jul. 2010.
two-channel digital oscilloscope TDS 7704B and further pro- [2] M. Ivrlac and J. Nossek, “The multiport communication theory,” IEEE
cessed on a PC using Matlab routines to compute the auto- and Circuits Syst. Mag., vol. 14, no. 3, pp. 27–44, 2014.
[3] K. F. Warnick, B. Woestenburg, L. Belostotski, and P. Russer, “Mini-
cross-correlation spectra of the sampled electric field in the fre- mizing the noise penalty due to mutual coupling for a receiving array,”
quency interval from 0.9 to 1.1 GHz. The magnitude of the mea- IEEE Trans. Antennas Propag., vol. 57, no. 6, pp. 1634–1644, Jun.
sured near-field autocorrelation spectrum is plotted in Fig. 17. 2009.
Fig. 18 shows the magnitude of the field distribution in the [4] H. Rowe, “Amplitude modulation with a noise carrier,” Proc. IEEE,
vol. 52, no. 4, pp. 389–395, Apr. 1964.
object pane reconstructed from the field data sampled in the ob-
[5] H. Rowe, “Frequency or phase modulation with a noise carrier,” Proc.
servation line. The object plane field reconstruction has been IEEE, vol. 52, no. 4, pp. 396–408, Apr. 1964.
performed using parametric identification methods as described [6] P. Russer, J. A. Russer, F. Mukhtar, P. Lugli, S. Wane, D. Bajon, and W.
in [53] and [54]. Porod, “Integrated antennas for RF sensing, wireless communications
and energy harvesting applications,” presented at the Int. Workshop
Antenna Technol. (iWAT), Karlsruhe, Germany, Mar. 4–6, 2013.
XI. CONCLUSION AND OUTLOOK
[7] G. P. Szakmany, A. O. Orlov, G. H. Bernstein, W. Porod, M. Bareiss,
Field correlation spectra and the spectral energy density P. Lugli, J. A. Russer, C. Jirauschek, P. Russer, M. T. Ivrlac, and J. A.
distribution of stochastic electromagnetic fields with Gaussian Nossek, “Nano-antenna arrays for the infrared regime,” presented at the
18th Int. ITG Workshop Smart Antennas (WSA), Erlangen, Germany,
probability distribution can be described by field correlation Mar. 2014.
dyadics. The response of electromagnetic structures to a sto- [8] R. F. Harrington, Time Harmonic Electromagnetic Fields. New York,
chastic field excitation can be computed on the basis of the NY, USA: McGraw-Hill, 1961.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
88 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 63, NO. 1, JANUARY 2015

[9] P. Russer, Electromagnetics, Microwave Circuit and Antenna Design [33] J. V. Bladel, Electromagnetic Fields, 2nd ed. New York, NY, USA:
for Communications Engineering, 2nd ed. Boston, MA, USA: Artech Wiley, 2007.
House, 2006. [34] C. A. Balanis, Advanced Engineering Electromagnetics. New York,
[10] J. Shi, M. Cracraft, K. Slattery, M. Yamaguchi, and R. DuBroff, “Cal- NY, USA: Wiley, 1989.
ibration and compensation of near-field scan measurements,” IEEE [35] X. Tong, D. Thomas, A. Nothofer, P. Sewell, and C. Christopoulos,
Trans. Electromagn. Compat., vol. 47, no. 3, pp. 642–650, 2005. “Modeling electromagnetic emissions from printed circuit boards in
[11] A. Tankielun, U. Keller, E. Sicard, P. Kralicek, and B. Vrignon, “Elec- closed environments using equivalent dipoles,” IEEE Trans. Electro-
tromagnetic near-field scanning for microelectronic test chip investi- magn. Compat., vol. 52, no. 2, pp. 462–470, May 2010.
gation,” IEEE EMC Soc. Newslett., pp. 68–72, Oct. 2006. [36] R. F. Harrington, Field Computation by Moment Methods. San Fran-
[12] W. B. Davenport and W. L. Root, An Introduction to the Theory of cisco, CA, USA: IEEE Press, 1968.
Random Signals and Noise. New York: McGraw-Hill, 1958. [37] J. A. Russer and P. Russer, “Imaging of sources of radiated electro-
[13] J. A. Russer and P. Russer, “Stochastic electromagnetic fields,” in Proc. magnetic interference,” Frequenz, vol. 65, pp. 261–265, Sep. 2011.
German Microw. Conf. (GeMIC), Mar. 2011, pp. 1–4. [38] J. A. Russer and P. Russer, “Imaging of incoherent sources of ra-
[14] J. A. Russer and P. Russer, “An efficient method for computer aided diation,” presented at the Asia-Pac. Symp. Electromagn. Compat.
analysis of noisy electromagnetic fields,” in IEEE MTT-S Int. Microw. (APEMC), Singapore, May 21–24, 2012.
Symp. Dig., Jun. 2011, pp. 1–4. [39] J. W. Goodman, Introduction to Fourier Optics. New York, NY,
[15] J. A. Russer and P. Russer, “Network methods applied to the compu- USA: McGraw-Hill, 1968.
tation of stochastic electromagnetic fields,” in Proc. Int. Conf. Electro- [40] M. Kanda, “Standard probes for electromagnetic field measurements,”
magn. Adv. Appl. (ICEAA), Sep. 2011, pp. 1152–1155. IEEE Trans. Antennas Propag., vol. 41, no. 10, pp. 1349–1364, Oct.
[16] J. A. Russer, G. Scarpa, P. Lugli, and P. Russer, “On the modeling of 1993.
radiated EMI on the basis of near-field correlation measurements,” in [41] N. Masuda, N. Tamaki, T. Kuriyama, J.-C. Bu, M. Yamaguchi, and K.
Proc. Eur. Microw. Conf. (EuMC), Manchester, U.K., 2011, pp. 9–12. Arai, “High frequency magnetic near field measurement on LSI chip
[17] J. A. Russer, F. Mukhtar, O. Filonik, G. Scarpa, and P. Russer, “Mod- using planar multi-layer shielded loop coil,” in Proc. IEEE Int. Symp.
elling of noisy EM field propagation using correlation information of Electromagn. Compat., Aug. 2003, vol. 1, pp. 80–85.
sampled data,” presented at the IEEE Int. Conf. Numer. Electromagn. [42] N. Ando, N. Masuda, N. Tarnaki, T. Kuriyama, S. Saito, K. Kato, K.
Model. Optim. (NEMO), Pavia, Italy, May 14–16, 2014. Ohashi, M. Saito, and M. Yarnaguchi, “Miniaturized thin-film mag-
[18] A. Baev, A. Gorbunova, M. Konovalyuk, Y. Kuznetsov, and J. A. netic field probe with high spatial resolution for LSI chip measure-
Russer, “Planar stochastic sources localization algorithm in EMC ment,” in Proc. Int. Symp. Electromagn. Compat. (EMC), Aug. 2004,
problems,” in Proc. Int. Conf. Electromagn. Adv. Appl. (ICEAA), vol. 2, pp. 357–362.
2013, pp. 440–443. [43] N. Uddin, M. Spang, and A. Thiede, “Integrated magnetic loop probe
[19] A. Gorbunova, A. Baev, M. Konovalyuk, Y. Kuznetsov, and J. A. in GaAs technology for active near-field sensor,” in Proc. 38th Eur.
Russer, “Stochastic EMI sources localization algorithm based on time Microw. Conf. (EuMC), Oct. 2008, pp. 1070–1073.
domain planar near-field scanning,” in Proc. Int. Symp. Electromagn. [44] N. Uddin and A. Thiede, “Switchable double-sensor integrated active
Compat. (EMC EUROPE), Brugge, Belgium, Sep. 2–6, 2013, pp. probe for near-field scanner,” in Proc. Eur. Microw. Integr. Circuits
972–976. Conf. (EuMIC), Oct. 2011, pp. 37–40.
[20] L. B. Felsen, M. Mongiardo, and P. Russer, Electromagnetic Field [45] A. Thiede, N. Uddin, and A. Awny, “Integrated active miniature sen-
Computation by Network Methods. New York, NY, USA: Springer- sors for electro-magnetic near field measurement,” in Proc. Asia-Pac.
Verlag, Mar. 2009. Symp. Electromagn. Compat. (APEMC), May 2012, pp. 197–200.
[21] D. Thomas, C. Obiekezie, S. Greedy, A. Nothofer, and P. Sewell, [46] “Multichannel absorberless near field measurement system” U.S.
“Characterisation of noisy electromagnetic fields from circuits using Patent 8 502 546 B2 [Online]. Available: http://www.google.com/
the correlation of equivalent sources,” in Proc. Int. Symp. Electro- patents/US8502546, tUSClassification USPC 324/637.
magn. Compat. (EMC EUROPE), Sep. 2012, pp. 1–5. [47] F. Krug and P. Russer, “Ultra-fast broadband EMI measurement in time
[22] J. A. Russer, T. Asenov, and P. Russer, “Sampling of stochastic electro- domain using classical spectral estimation,” in Proc. IEEE MTT-S Int.
magnetic fields,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig., Jun. Microw. Symp. Dig., 2002, vol. 3, pp. 2237–2240.
2012, pp. 1–3. [48] S. Braun, T. Donauer, and P. Russer, “A real-time time-domain EMI
[23] M. Krumpholz and P. Russer, “A field theoretical derivation of TLM,” measurement system for full-compliance measurements according to
IEEE Trans. Microw. Theory Tech., vol. 42, no. 9, pp. 1660–1668, CISPR 16-1-1,” IEEE Trans. Electromagn. Compat., vol. 50, no. 2, pp.
1994. 259–267, 2008.
[24] M. Krumpholz, C. Huber, and P. Russer, “A field theoretical compar- [49] J. A. Russer, S. Braun, A. Frech, and P. Russer, “Time-domain mea-
ison of FDTD and TLM,” IEEE Trans. Microw. Theory Tech., vol. 43, surement of spectra of stochastic electromagnetic fields,” in Proc. Int.
no. 8, pp. 1935–1950, 1995. Conf. Electromagn. Adv. Appl. (ICEAA), 2013, pp. 1123–1126.
[25] H. W. König and H. Pötzl, “Schwankungsvorgänge in Elektronen- [50] A. Frech, A. Zakaria, S. Braun, and P. Russer, “Ambient noise
strahlen,” in Fortschritte der Hochfrequenztechnik. Frankfurt, Ger- cancelation with a time-domain EMI measurement system using
many: Akademische Verlagsgesellschaft, 1959, vol. 4, pp. 195–239. adaptive filtering,” in Proc. Asia-Pac. Symp. Electromagn. Compat.
[26] H. A. Haus and R. W. Adler, Circuit Theory of Linear Noisy Net- 19th Int. Zurich Symp. Electromagn. Compatibility (APEMC), 2008,
works. New York, NY, USA: Wiley, 1959. pp. 534–537.
[27] H. Bosma, “On the theory of linear noisy systems,” Philips Res. Rep., [51] A. Frech, S. Braun, and P. Russer, “Time-domain EMI measurements
Suppl., no. 10, pp. 1–189, 1967. in the presence of ambient noise,” in Proc. IEEE Int. Symp. Elec-
[28] H. Hillbrand and P. Russer, “An efficient method for computer aided tromagn. Compat. (EMC), Austin, TX, USA, Aug. 17–21, 2009, pp.
noise analysis of linear amplifier networks,” IEEE Trans. Circuits Syst., 139–142.
vol. 23, no. 4, pp. 235–238, Apr. 1976. [52] A. Frech, S. Braun, A. Zakaria, and P. Russer, “Time-domain EMI
[29] H. Hillbrand and P. Russer, “Correction to an efficient method for com- measurements at open test sites using ambient noise cancelation
puter aided noise analysis of linear amplifier networks,” IEEE Trans. methods,” presented at the URSI Gen. Assembly, Chicago, IL, USA,
Circuits Syst., vol. 23, no. 11, p. 691, Nov. 1976. August 7–16, 2008.
[30] P. Russer and S. Müller, “Noise analysis of linear microwave circuits,” [53] A. Baev, A. Gorbunova, M. Konovalyuk, Y. Kuznetsov, and J. A.
Int. J. Numer. Model., Electron. Netw., Devices, Fields, vol. 3, pp. Russer, “Stochastic EMI sources localization based on ultra wide
287–316, 1990. band near-field measurements,” in Proc. Eur. Microw. Conf. (EuMC),
[31] P. Russer, “Noise analysis of linear microwave circuits with general Nuremberg, Germany, Oct. 6–10, 2013, pp. 1131–1134.
topology,” in The Review of Radio Science 1993–1996. Oxford, [54] M. Konovalyuk, Y. Kuznetsov, and A. Baev, “The joint inverse fil-
U.K.: Oxford Univ. Press, 1996, pp. 887–890. tering and parametric identification for complex radar image,” in Proc.
[32] D. Middleton, An Introduction to Statistical Communication Theory. XXXth URSI Gen. Assembly Sci. Symp., 2011, pp. 1–4.
Los Altos, CA, USA: Peninsula, 1987, vol. 960.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.
RUSSER AND RUSSER: MODELING OF NOISY EM FIELD PROPAGATION USING CORRELATION INFORMATION 89

Johannes A. Russer (M'09) received the Dipl.-Ing. Peter Russer (SM'81–F'94–LF'13) received the
(M.S.E.E.) degree in electrical engineering and infor- Dipl.-Ing. (M.S.E.E.) degree in 1967 and the Dr.
mation technology from the Universität Karlsruhe, techn. (Ph.D.E.E.) degrees in 1967 and 1971,
Karlsruhe, Germany, in 2003 and the Ph.D.E.E. de- respectively, both from the Vienna University of
gree from the University of Illinois at Urbana-Cham- Technology, Vienna, Austria.
paign, Urbana-Champaign, IL, USA, in 2010. In 1971, he joined the Research Institute of
In 2004, he joined the University of Illinois at Ur- AEG-Telefunken, Ulm, Germany, where he worked
bana-Champaign as a Research Assistant. From 2007 on electronics and electronics and lightwave tech-
to 2010, he worked for Qualcomm, Inc,. as an in- nology. From 1981 to 2008, he was Full Professor
tern. Since 2010, he has been a Postdoctoral Research at the Technische Universität München (TUM),
Fellow at the Institute of Nanoelectronics of the Tech- München, Germany. From October 1992 to March
nische Universität München (TUM), München, Germany. 1995, he also was Director of the Ferdinand-Braun-Institut für Höchstfrequen-
Dr. Russer has received a Best Student Paper Award at the IEEE International ztechnik, Berlin, Germany. After his retirement in 2008, he was appointed
Microwave Symposium in 2008. He is a member of VDE and the Eta Kappa Nu as Emeritus of Excellence at the TUM, and since 2010, he has been with the
honor society. Institute of Nanoelectronics of the TUM. He has published five books and more
than 800 scientific papers in refereed journals and conference proceedings.
Dr. Russer received the Best Paper Award from the NTG (German Society
for Information Technology) in 1979. In 2006, he received the Distinguished
Educator Award of the IEEE MTT Society and in 2009 the Distinguished Ser-
vice Award from the European Microwave Association (EuMA). In 2007, he
received an honorary Doctor degree from the Moscow University of Aerospace
Technologies (MAI), Moscow, Russia. In 2010, he was awarded the Golden
Ring of Distinction of the VDE—the German Association for Electrical, Elec-
tronic and Information Technologies. He was elected member of acatech, the
German Academy of Science and Engineering, in 2006.

Authorized licensed use limited to: Southeast University. Downloaded on February 05,2021 at 02:42:32 UTC from IEEE Xplore. Restrictions apply.

You might also like