![Antennas layout, receiving antenna [red], transmitting antenna [blue]. We utilize one transmitting antenna, which has been manually moved over the grid depicted in Figure 2, ensuring defined antenna positions. We employed two specific measurement grid positions, the seat behind the driver (κ = 1) and the middle seat (κ = 2). The grid measurements also allow averaging out the effect of small-scale signal fading. The grid distance in both directions is 3 cm for 3-11 GHz and 4 mm for 55-65 GHz. The grid allows measuring 10×10 positions.](https://www.researchgate.net/profile/Aniruddha-Chandra/publication/309155142/figure/fig1/AS:613914123452431@1523379941474/Antennas-layout-receiving-antenna-red-transmitting-antenna-blue-We-utilize-one_Q320.jpg)
Content uploaded by Aniruddha Chandra
Author content
All content in this area was uploaded by Aniruddha Chandra on Feb 15, 2018
Content may be subject to copyright.
1
In-Vehicle Channel Measurement, Characterization
and Spatial Consistency Comparison of 3-11 GHz
and 55-65 GHz Frequency Bands
Jiri Blumenstein, Ales Prokes, Aniruddha Chandra, Tomas Mikulasek, Roman Marsalek, Thomas Zemen,
and Christoph Mecklenbr¨
auker
Abstract—The paper provides real-word wireless measurement
data of the intra-vehicular channel for both the 3-11 GHz and
the 55-65 GHz frequency band under similar conditions.
By spatially averaging channel impulse response (CIR) real-
izations within a 10×10 grid, we obtain the power-delay profile
(PDP). The data measured at 3-11 GHz and 55-65 GHz exhibit
significant differences in terms of root mean square (RMS)
delay spread, number of resolvable clusters and variance of
the maximal excess delay. Moreover, we evaluate the spatial
stationarity via the Pearson correlation coefficient and via the
PDP collinearity depending on the distance in the grid. The mea-
sured and calculated results indicate that a strong reverberation
inside the vehicle produces similar PDPs within the range of
approximately 10 wavelengths.
We also provide a linear piecewise model of the PDP in
logarithmic scale and a generalized extreme value (GEV) model
of small-scale signal fading. Our channel model is validated
utilizing the Kolmogorov-Smirnov (K-S) test.
Index Terms—In-vehicle channel model, UWB, mm-wave,
PDP, frequency domain channel sounding, channel stationarity.
I. INTRODUCTION
INTRA-and inter-vehicular wireless connectivity is the key
enabler for enhancing traffic safety, reliability, fluidity and
efficiency of future transportation systems [1], [2].
A modern vehicle contains a cable harness weighting several
tens of kilograms [3]. Since weight savings are crucial in
the car industry in terms of fuel and power consumption,
future vehicles will substitute at least part of the metallic
cables with wireless connections. Thus, an in-vehicle wireless
communication system is beneficial not only when connecting
moving parts such as wheels or seats [4]. Moreover, the
design of the cable harness is often platform specific and
its manufacture and installation are a time consuming and
technically demanding process. Also, assuring the reliability
of the cable harnesses is not easy in the presence of moisture
Copyright (c) 2015 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be
obtained from the IEEE by sending a request to pubs-permissions@ieee.org.
J. Blumenstein, A. Prokes, A. Chandra, T. Mikulasek and R. Marsalek are with
the Department of Radio Electronics, Brno University of Technology, Czech
Republic. T. Zemen is with AIT, Austrian Institute of Technology, Vienna,
Austria, and Ch. Mecklenbr¨
auker is with Institute of Telecommunications,
Christian Doppler Lab, Vienna University of Technology, Vienna, Austria.
e-mail: blumenstein@feec.vutbr.cz.
Manuscript received December 23, 2015; revised March 10, 2016, accepted
Jul 31, 2016.
or bending. Another application of the intra-vehicular wireless
backbone could be seen in the possibility of retrofitting state-
of-the-art driver assistance features also into older vehicles.
Thus, the in-vehicle wireless connection could bring together
functionalities such as being the wireless backbone for the
vehicle’s operational data, providing local Internet connection
for the passengers, enabling the connection of vehicle parts
that are otherwise difficult to connect or supporting an on-
board infotainment system.
To contribute to the research into the in-vehicle wireless
connectivity, we compare in detail wireless channels in the
range of the 3-11 GHz and the 55-65 GHz frequency bands in
the in-vehicle environment under very close conditions (differ-
ences will be clarified later in the paper and are given only by
inevitably different channel sounders). We also parametrize
our measured data in order to provide a specialized and
detailed channel model for the typical in-vehicle environment
of a mid-sized passenger car. Our findings could help to
design an optimized physical layer for the broadband in-
vehicle wireless communication system.
While both bands provide around 10 GHz of unlicensed
bandwidth (depending on the local spectrum management
authorities), the differences between the 3-11 GHz and the
55-65 GHz frequency band are significant and presented by
this paper.
In order to enhance the credibility and the reproducibility
of our research, the measured data are freely available on the
link: http://www.radio.feec.vutbr.cz/GACR-13-38735S/
A. 3-11 GHz frequency band
In the 3-11 GHz frequency range, authors in [5], [6]
deduce that the beneficial radio environment characteristics
for indoor areas (such as robustness against multipath fading
and low transmit power) could be extrapolated to the vehicle
passenger compartment. Intra-vehicle channel measurements
are performed in [7]–[13] and channel modeling in [6], [14].
The spatial stationarity and collinearity have, however, not
been studied and compared with the millimeter wave band.
In [15] the channel stationarity has been shown for 3-11 GHz,
but the 55-65 GHz band is not mentioned.
Obstructed by seats and passengers, the in-vehicle environ-
ment will hardly offer a line of sight (LOS) component. This
might be of significant importance when compared with indoor
environment in [5], [6]. On the other hand, please note that
2
the 1.99 - 10.6 GHz band is deregulated for communication
and wall penetrating radars in the U.S [16]–[18]. Thus, one
of the research objectives for the lower band is to find out
whether the multipath components could sufficiently support
the intended wireless application.
B. 55-65 GHz frequency band
In [19]–[21] the ultra-wide bandwidth (UWB) and the
mm-wave in-vehicle channels are compared. The comparison
includes the Rician K-factor, the root mean square (RMS)
delay spread and the path-loss, but no spatial stationarity
evaluation is given. No specific channel model parametrization
for the in-vehicle environment is offered in [20].
The 55-65 GHz band enables the usage of high-gain
steerable antennas in a small physical form allowing for
a beamforming, a beamsteering and a spatial multiplexing
[19], [22]. By contrast, the 60 GHz band suffers from
high penetration loss and insignificant diffraction contribution.
Thus, shadowing effects are expected to be more significant
compared to the 3-11 GHz band [23]. Also, considering the
millimeter wavelength, the vibrations usually occurring in the
transportation may cause difficulties in terms of increased
Doppler spread [24].
However, expecting a lower delay spread of the power-delay
profile (PDP), the 55-65 GHz frequency range implies a lower
complexity channel estimation [25] compared to the 3-11 GHz
band. Our results also show higher spatial correlation of the
55-65 GHz band for a distance of around 5 wavelengths,
indicating a lower achievable transmit diversity gain for ap-
propriate Multiple-Input Multiple-Output (MIMO) schemes. A
channel model for 55-65 GHz of the in-vehicle environment
is presented in [26]; however, a comparison with other bands
is lacking.
C. Contribution of the paper
•Under similar conditions we compare the 3-11 GHz
with the 55-65 GHz frequency band in terms of RMS
delay spread, maximal excess delay, number of resolvable
clusters in the PDP, and the cluster decay rate [27].
•We show a linear piecewise model of the PDP in logarith-
mic scale and a generalized extreme value (GEV) model
of small-scale signal fading.
•The spatial stationarity of the measured channel parame-
ters is analyzed. We evaluate:
1) the Pearson correlation coefficient of the channel
impulse responses (CIRs) in order to see the poten-
tial diversity gains.
2) the collinearity of the PDP in order to evaluate
the number of realizations of the CIR needed for
calculating the PDP.
•We validate our statistical channel model utilizing the
Kolmogorov-Smirnov (K-S) test.
D. Organization of the paper
The organization of the paper is as follows. First we
state Introduction section, and then we present the section
Measurement Setup. Next sections are Channel Description,
Spatial Stationarity of Measured Channel Impulse Responses
and Channel Model Validation. Conclusion section sums up
the paper.
Fig. 1: Antennas layout, receiving antenna [red], transmitting
antenna [blue]. We utilize one transmitting antenna, which
has been manually moved over the grid depicted in Figure 2,
ensuring defined antenna positions. We employed two specific
measurement grid positions, the seat behind the driver (κ=1)
and the middle seat (κ=2). The grid measurements also allow
averaging out the effect of small-scale signal fading. The grid
distance in both directions is 3 cm for 3-11 GHz and 4 mm
for 55-65 GHz. The grid allows measuring 10×10 positions.
II. MEASUREMENT SETUP
Measurements are performed in a mid-sized Skoda Oc-
tavia III car using transmit and receive antennas marked with
blue and red colors in Figure 1, respectively. The receiving
(RX) antenna is placed next to the rear-view mirror on the front
windshield according to Figures 1, 5a and 5b. The transmitting
antenna is placed at 10×10 spatial points, using a polystyrene
rack in the case of the 3-11 GHz band and a metallic X-Y
table covered with absorbers in the case of the 55-65 GHz
band.
Given the absorbers and the fact that the relative polystyrene
permittivity is close to the relative air permittivity, we en-
sure that the locations of the receive antennas are well-
defined, while introducing as few artificial reflections from
the measurement device as possible. Between neighboring
measurement locations, the polystyrene rack utilizesa3cm
grid distance while the X-Y table step was 4 mm. Moreover,
the measurement grids (for both examined bands) ware placed
at two different positions on the rear bench. As seen in Figure
1, the positions of the measurement grid are denoted with
index κ.
As will be discussed later, apart from the different antennas
utilized for the two frequency bands examined, the different
antenna holding racks are one of the inevitable dissimilarities
introduced into our measurement. Without considering the an-
tenna manipulation overhead, the measurement of one channel
transfer function (CTF) took approximately 15 seconds. In
total, 200 CTFs were captured for both examined frequency
bands and one grid position κ.
3
Fig. 2: 10 ×10 measurement grid. The distance between grid
positions is 3 cm for 3-11 GHz and 4 mm for 55-65 GHz.
A. 3-11 GHz measurement apparatus
The Agilent Technologies E5071C four-port vector network
analyzer (VNA) is used for measuring the transmission co-
efficient between two antennas in the 3-11 GHz frequency
band. The omni-directional conical monopole antennas having
a radiation pattern as depicted in Figure 3a are used as
transmitting and receiving antennas. To achieve the highest
possible VNA dynamic range, we chose a maximum available
output power of 5 dBm at the test port of the VNA and the
intermediate frequency (IF) bandwidth of 100 Hz. The choice
of the IF bandwidth value is motivated by keeping the high
system dynamic range (>90 dB) while maintaining reasonable
time (≈15 s for one CTF) of one measurement runtime.
The measurement is performed in the frequency domain with
10 MHz frequency steps (as in [28]).
After temperature stabilization of VNA, the coaxial cables
(antennas disassembled) were connected to the electronic cal-
ibration module and full four port calibration was performed.
B. 55-65 GHz measurement apparatus
The R&S ZVA67 four-port VNA is utilized for measuring
the transmission coefficient between two antennas in the
55-65 GHz frequency band. The system dynamic range of
the measurement setup was extended utilizing a broadband
power amplifier (QuinStar QPW-50662330, measured gain of
35 dB in the band of interest ) on the transmitting side. The
WR15 open waveguide having a radiation pattern as depicted
in Figure 3b was used as a transmitting and receiving antenna
(similarly as in [29]). The system’s dynamic range of this mea-
surement setup is ≈50 dB. The VNA output power of 5 dBm
and IF bandwidth of 100 Hz was chosen in the same way as
in the case of the measurement apparatus for 3-11 GHz.
Note that the measurement is performed in the frequency
domain with 10 MHz frequency step. We transform the
channel frequency response into the time domain, utilizing
the inverse Fourier transform with Blackman windowing (as
in [30]). More about the influence of the windowing is
elaborated in [31].
In order to avoid a degradation of the measured phase
accuracy due to movements of the TX antenna, phase-stable
coaxial cables were used. The full four port calibration process
was performed including the power amplifier. The flanges of
-20dBi
-10dBi
0dBi
10dBi
90
o
60
o
30
o
0
o
-30
o
-60
o
-
90
o
-120
o
-150
o
1
80
o
150
o
120
o
E
-p
l
ane f = 3.0 GHz
f = 6.5 GHz
f = 10.0 GHz
-20dBi
-10dBi
0dBi
10dBi
90
o
60
o
30
o
0
o
-30
o
-60
o
-90
o
-120
o
-150
o
1
80
o
150
o
120
o
H
-p
l
ane
(a) Measured gain pattern of the conical monopole antennas
for a frequency range of 3-11 GHz.
-20dBi
-10dBi
0dBi
10dBi
90
o
60
o
30
o
0
o
-30
o
-60
o
-90
o
-120
o
-150
o
180
o
150
o
120
o
H
-p
l
ane
-20dBi
-10dBi
0dBi
10dBi
90
o
60
o
30
o
0
o
-30
o
-60
o
-
90
o
-120
o
-150
o
180
o
150
o
120
o
E
-p
l
ane f = 55 GHz
f = 60 GHz
f = 65 GHz
(b) Measured gain pattern of the open-ended waveguide in
the E–plane and the H–plane for open-ended waveguide
antenna at 55-65 GHz.
Fig. 3: Measured radiation patterns of the antennas used.
the TX and RX antennas were connected together and the
VNA transmission was normalized in the forward direction.
Complete list of used components is available on-
line: http://www.radio.feec.vutbr.cz/GACR-13-38735S/mmw-
measurement-in-frequency-domain-with-rotagrip/.
C. Antenna pattern influence
The conical monopole antenna used for the channel sound-
ing at 3-11 GHz is omnidirectional in the H-plane and 120◦
wide in the E-plane, while the open-ended waveguide used
for the measurement at 55-65 GHz has a 120◦wide radiation
pattern in the E-plane and 60◦wide radiation pattern in the
H-plane (for 3 dB decay). This makes the comparison of
the two bands seemingly problematic. On the other hand,
however, there is no of-the-shelf omnidirectional antenna for
the millimeter waves with 10 GHz bandwidth.
D. Comparability of the measurements with omnidirectional
and directional antennas
To provide a further insight on the antenna pattern influence,
the open-ended waveguide transmitting antenna has been ro-
tated clockwise with a 60◦step from 0◦to 180◦.
The results of this measurement are visible in Figure 4. The
60◦curve, apart from being 8 dB weaker compared to the 0◦
heading direction, shows no significant difference. The 120◦
and 180◦data exhibit a different behavior. The first arrival
(LOS) components are 20-30 dB weaker or simply not visible
(again, compared with 0◦). This is predominant namely in the
range of τ∈{3−8}ns.
4
To summarize, no additional resolvable clusters of the multi-
path components (MPCs) are visible in the data measured with
the angular misalignment to LOS. Thus, the measurements
with the open-ended waveguide antennas in the frequency
range of 55-65 GHz are comparable with the measurements
performed with the omnidirectional monoconical antenna
for 3-11 GHz.
0 5 10 15 20 25 30 35
−130
−120
−110
−100
−90
−80
τ [ns]
Relative power [dB]
0°
120°
60°
180°
2 3 4 5 6 7 8
−140
−120
−100
−80
τ [ns]
Rel. power [dB]
Fig. 4: The measured PDPs for following heading directions
(relative to the LOS): 0◦,60◦,120◦and 180◦.
III. CHANNEL DESCRIPTION
The measured in-vehicle radio channel is considered as time
invariant. Thus, to fully describe the radio channel, we write
the CTF as:
κHα(k)=κsα
21(k),(1)
where kstands for the measurement index identifiable with
certain frequency and where αdenotes the position of the
receiving antenna within the measurement grid, as visible in
Figure 2 and κstands for the location index of the whole
measurement grid within the vehicle. Utilizing an inverse
Fourier transform we convert the CTF into a CIR according
to:
κhα(n)=
N−1
k=0
κHα(k)ejkn2π/N,(2)
where nis discrete time in the delay domain.
The PDP is then calculated as spatial average of |κhα(n)|2
according to:
κP(n)=E
α{|κhα(n)|2}.(3)
The spatial measurement grid is of square layout with dimen-
sions of approximately 10 ×10 wavelengths.
Please note that it holds that:
τ=n1
B,(4)
where τis the time in the delay domain, Bis the bandwidth
and 1/B is the time resolution. Please note that the symbols
and notations used are listed in Table I.
TABLE I: Table of selected symbols and notations.
κHα(k)channel transfer function for α-th in the measurement grid
and κ-th position of the grid within the vehicle
κhα(n)channel impulse response for α-th in the measurement grid
κ-th position of the grid within the vehicle
ρPearson correlation coefficient
sα
21(k)scattering parameter, the forward transmission coefficient
αindex denoting the spatial position within the measurement grid
ntime in delay domain
kfrequency index
κindex denoting the position of the measurement
grid within the vehicle
P∗
dB(n)PDP model in logarithmic scale
(.)∗model variable
E{.}expected value
||.||FFrobenius norm
(.)Hcomplex conjugate transpose
tr(.)trace (the sum of the elements on the main diagonal)
TABLE II: Tabulated average delay and RMS delay spread.
The mean and variance values for both 3-11 GHz and
55-65 GHz. (The calculation is done for 100 measurements
in each examined band.)
position average delay [ns] RMS delay spread [ns]
mean variance mean variance
3-11 GHz κ=1 48.07 4.50 23.34 <0.01
κ=2 47.41 4.2 23.27 <0.01
55-65 GHz κ=1 13.75 0.01 5.02 <0.01
κ=2 12.47 0.01 5.03 <0.01
A. Moments of PDP
For practical reasons, the preferred parameters evaluating
the wireless channel are the first and the second moment of
the PDP, i.e., the average delay is given as:
D=nmax
nP(n)n
nmax
nP(n),(5)
where the maximal delay of P(n),nmax is detected as a time
in the delay domain where the PDP crosses the noise floor.
Please note that the noise floor is estimated as the RMS value
of the PDP before the rising edge of the first arriving MPC.
For the case of 3-11 GHz band, the noise floor is -109 dBm
and for the 55-65 GHz band the noise floor is -134 dBm. The
RMS delay spread is defined according to:
Sn=nmax
nP(n)n2
nmax
nP(n)−D2.(6)
Our measured data are stated in Table II. The 3-11 GHz
band is affected by an approximately 5 times higher delay
spread than the 55-65 GHz band in a system with a peak
signal-to-noise ratio of around 50 dB for both bands. We uti-
lized 100 CIRs for the calculation of one examined frequency
band and one position κ.
One reason why the 55-65 GHz channels have lower delay
spread is because of the generally higher path-loss that atten-
uates MPCs with a large delay. Also, the non-omnidirectional
antenna (open-ended waveguide) does not radiate in angles
from -80◦to 80◦, which means that it produces a lower number
of highly delayed MPCs, thus the resulting PDP might be
shortened.
5
(a) 3-11 GHz monoconical antenna mounted on the front
windshield next to the rear-view mirror.
(b) 55-65 GHz open-ended waveguide antenna fastened using
a suction cap.
(c) X-Y table for precise movement of the open-ended waveguide antenna
within the measurement grid. The metallic X-Y table, the antenna with power
amplifier and heat sink are covered with absorbers in order to reduce the
unwanted reflections and thus the influence of the measurement. The grid
position corresponds to κ=1.
(d) Polystyrene 10×10 antenna holder and
3-11 GHz monoconical antenna. The grid
position corresponds to κ=1.
Fig. 5: Photographs taken during the measurement campaign.
B. Comparison with other in-vehicle measurement campaigns
To give a perspective on the data presented in Table II, we
provide a brief overview on findings of different authors. The
other measurement campaigns have usually been performed
under notably different conditions. We present the most simi-
lar ones.
For the much more confined area of the engine bay, the
authors in [7] report RMS delay spread values of around 5 ns
for the 3.1-10.6 GHz band. For the same band, the author
in [9] reports around 10 ns of the RMS delay spread for
the passenger compartment. In [14], the reported values of
RMS delay spread are around 15 ns while the frequency range
is not exactly given, but 802.15.3a is claimed as the target
application. The authors in [20] compare the 5-8.5 GHz bands
and the 67-70.5 GHz bands, but they report significantly lower
RMS delay spread values, which range from 1 ns to 5 ns while
the lower band exhibits higher RMS delay spread values. Such
a notable difference might probably arise due to the different
dynamic range of the channel sounder used.
C. Channel approximation
The small dimensions and metallic surroundings of the in-
vehicle environment create conditions which are best described
by a dense scattering model. Thus, with the model validated
via the K-S test, we parametrize the well-known linear piece-
wise PDP model [25]. The specular components which form
the leading edge of the clusters and, hence, the breakpoints of
the linear piecewise characteristics, are estimated by a peak
finder. The peak finder is adjusted based on visual inspection.
According to the observations and with the aim to provide
a reproducible channel model, we operate with the hypothesis
that the CIR is composed of PDP and small scale fading (SSF)
according to:
|hdB(n)|2=PdB (n)+ξ(n),(7)
6
where PdB(n)and ξ(n)stand for the PDP and SSF, respec-
tively. As proposed below, SSF is modeled by an appropriate
random process while the PDP is described by a piecewise
linear approximation. Note that the PDP is interchangeable
with a large scale fading (LSF). Since recent measurements
presented in [13], [20], [21] report a very low path-loss
coefficient, thus a low dependency of the path-loss on the
actual antenna separation within the vehicle, we decided to
introduce LSF as a function of time in the delay domain
instead of the more traditional function of antenna separation.
Figure 6 presents the measured |1,2hα
dB(n)|2∀α, each
aligned according to the first arrival MPC. It also presents
the averaged PDP 1,2PdB(n)and the piecewise linear approx-
imation 1,2P∗
dB(n).
1) Power-delay profile: Our measured data are presented
in logarithmic scale. Thus, the generally accepted one-sided
exponential PDP model [25] is reduced to a linear function
according to:
P∗
dB,c(n)=−nPDR,c /Tm,c Tm,c−1<n<T
m,c
0 otherwise,(8)
where cindexes the clusters, and PDR,c and Tm,c stand for the
the dynamic range and maximal excess delay of the relevant
cth cluster. The multiple cluster PDP model is written as:
P∗
dB(n)=
cPdB,c(n)+Cc,(9)
where Ccrepresents the cluster decay ratio in decibels. The
quantities from (8) and (9) are visualized in Figure 7. Note
that the upper index ∗denotes the artificial model variable
derived from a measured variable. The corresponding param-
eters of the channel, as derived from our data, are presented
in Table III.
In the case of the position κ=1, the origin of the three
MPC clusters in the 55-65 GHz band is as follows: the first
cluster represents the direct/dominant component, the second
cluster stands for a strong reflection and the third cluster is
a reverberation diffused tail. As for the position κ=2, three
recognizable reflections are present. Please note that the term
reverberation diffused tail is discussed and defined in [32].
Now, a notable difference from the 3-11 GHz band is the lack
of higher number of MPC clusters on one hand, but on the
other hand it features a significantly longer PDP in the delay
domain.
2) Small scale fading: Exploiting the maximum likelihood
estimate (MLE), we parametrize the superimposed SSF signal
ξ(n), using the generalized extreme value (GEV) distribution
[33]. The probability density function (PDF) of the GEV is
given by:
f(x|K, Υ,Γ) = 1
Γexp−β−1
Kβ−1−1
K,∀β=1+Kx−Υ
Γ.
(10)
The estimated parameters of the spatially universal small-
scale fading model are presented in Table IV.
Due to the high flexibility of the SSF model ξ∗(n), which is
given by three input parameters as opposed to the usual two
parameters, the MLE metric recommends GEV distribution.
On the other hand, the authors in [34] claim that there is
no theoretical explanation for encountering this distribution
type. It is, however, worth noting, that the GEV contains the
accepted Weibull distribution as a special case for K<0.
Please note that in [12] the authors report the Weibull distri-
bution for their measurement results for the 2.4 GHz intra-
vehicle channel. Similar results are presented in [13]. Note
that there is practically a very small difference between the
data for position κ=1and κ=2.
D. Physical meaning of SSF parameters
The parameters K, Υ,Γof the GEV process are commonly
known as location,scale and shape respectively. It holds that
K∈R,Υ>0and Γ∈R.
The physical meaning of these parameters might not be
exactly clear from this notation, therefore we provide the
following insight into the effect of these parameters. The pa-
rameter Γdrives the tail behavior while the scale parameter Υ
determines the spread of the distribution. Next, the parameter
Kprescribes the impulsiveness of the signal. The higher |K|,
the higher rate of occurrence and power of an impulsive noise
superimposed on the random GEV signal with parameters Υ
and Γ. The polarity of these impulses is the same as the
polarity of the parameter K.
It is worth noting that for the 3-11 GHz band the parameter
Kis almost zero (agrees with [13], [34], see Table IV),
but for the 55-65 GHz frequency band the parameter K
is significantly lower, implying that clusters of MPCs are
more resolvable in the delay domain. Moreover, this fact also
supports the finding that for the higher frequency band the
popular assumption of uncorrelated scattering is not exactly
valid (as presented for a vehicle-to-vehicle example in [35]).
IV. SPATIAL STATIONARITY OF MEASURED CHANNEL
IMPULSE RESPONSE
The spatial stationarity of the measured wireless channel
is studied via an evaluation of the Pearson correlation coef-
ficient |ρα|of the measured CIR hα
dB(n)similarly as in [36]
and via an evaluation of the collinearity of the PDP PdB(n).
A. Pearson correlation coefficient evaluation of the 3-11 GHz
and 55-65 GHz bands
In this subsection we evaluate the spatial stationarity of the
whole measured bandwidths. This means that we compare the
8 GHz bandwidth of the 3-11 GHz band (later designated
with uwb) with the 10 GHz bandwidth of the 55-65 GHz band
(later designated with mmw). As will be shown in a following
section, we also provide a comparison of spatial stationarity
of sub-bands with equal relative bandwidths.
The Pearson correlation coefficient is given as:
ρα=E[(hα(n)−μα)(h45(n)−μ45 )]
σασ45 ,(11)
where σαdenotes the standard deviation and μαthe mean
of hα(n). The correlations are calculated between the spatial
7
0 20 40 60 80
−110
−100
−90
−80
−70
−60
τ [ns]
Relative power [dB]
|1hα(n)|2∀α
PDP model 1P*
dB(n)
Noise floor
PDP meas. 1PdB(n)
(a) 3-11 GHz, κ=1.
0 5 10 15 20 25 30 35
−140
−130
−120
−110
−100
−90
−80
τ [ns]
Relative power [dB]
Noise floor
|1hα(n)|2∀α
PDP model 1P*
dB(n)
PDP meas. 1PdB(n)
(b) 55-65 GHz, κ=1.
0 5 10 15 20 25 30 35
−140
−130
−120
−110
−100
−90
−80
Relative power [dB]
τ [ns]
|2hα(n)|2∀α
Noise floor
PDP model 2P*
dB(n)
PDP meas. 2PdB(n)
(c) 55-65 GHz, κ=2.
Fig. 6: Measured and calculated channel parameters in the delay domain for both investigated frequency bands, i.e., measured
absolute value of squared CIR |1hα
dB(n)|2, 100 realizations; Spatially averaged power-delay profile 1PdB(n); Piecewise linear
approximation 1P∗
dB(n)and estimated noise floors. The data for the 3-11 GHz band and the grid position κ=2exhibit no
visual difference compared to κ=1plotted in Figure 6a.
TABLE III: Table of the PDPs measured. The values of dynamic range PDR,c, maximal excess delay Tm,c and cluster
decay rate Ccare given. (The PDP model for the 3-11 GHz band and for position κ=2match the model for κ=1.)
3-11 GHz 55-65 GHz
position cluster
number 12123
κ=1 PDR,c [dB] 11.56±0.30 33.2 ±0.31 18.31 ±0.32 19.92 ±0.22 30.83 ±0.25
Tm,c [ns] 1.00 80.23 1.03 1.33 27.76
Cc,[dB] 0 9.63 ±0.34 0 3.63 ±0.30 15.68 ±0.26
55-65 GHz
position cluster
number 12345
κ=2 PDR,c [dB] 16.21±0.20 9.1 ±0.24 11.11 ±0.21 6.54 ±0.32 29.44 ±0.35
Tm,c [ns] 1.40 1.20 1.20 1.20 31.20
Cc,[dB] 0 9.36 ±0.33 3.02 ±0.31 8.06 ±0.20 6.00 ±0.36
positions α∈{1...100}and α=45, which is located
approximately in the middle of the measurement grid.
Reflecting the square geometry of our antenna holding
rack, the maps of the correlation coefficient ραare plotted
in Figure 8.
1) 3-11 GHz: Figure 8a shows that 1hα(n)is spatially
stationary with a mean value of |1ρα|≈0.17. For the
position κ=2, the data are |2ρα|≈0.15. The coefficients
|1,2ρα|slightly decrease with the distance form the reference
point α=45.
8
TABLE IV: Estimated parameters with confidence intervals of the superimposed SSF random process. We also present a best
fit for each parameter ensuring the highest score in the K-S test.
position K best fit
of K Υbest fit
of ΥΓbest fit
of Γ
3-11 GHz κ=1 -0.0239 ±0.0062 -0.0282 4.541 ±0.0451 4.5682 -2.236 ±0.1070 -2.7938
κ=2 -0.0178 ±0.0055 -0.0279 4.476 ±0.0554 4.0169 -2.465 ±0.1102 -2.6296
55-65 GHz κ=1 -0.4657 ±0.0081 -0.5279 6.016 ±0.0488 6.2783 -2.532 ±0.0809 -2.6514
κ=2 -0.4878 ±0.0052 -0.4672 6.840 ±0.0514 6.4625 -2.734 ±0.0721 -2.4264
2) 55-65 GHz: Figure 8b depicts the significantly higher
spatial stationarity of 1hα(n)with a mean value of |1ρα|≈
0.30. Also, the variance of |1ρα|is notably higher. As for the
position κ=2, the data are |2ρα|≈0.38. The coefficients
|1,2ρα|do not decrease significantly with distance from the
reference point α=45.
B. Pearson correlation coefficient comparison of the 3-11 GHz
sub-bands with equal relative bandwidth
Here we evaluate and depict the absolute value of the
Pearson correlation coefficient |1ρα|for selected sub-bands
Fig. 7: An example of the piecewise linear approximation
having two dominant clusters of PDP in logarithmic scale.
2 4 6 8 10
2
4
6
8
10
0
0.2
0.4
0.6
0.8
|ρ|
2 4 6 8 10
2
4
6
8
10
0
0.2
0.4
0.6
0.8
|ρ|
0 0.2 0.4 0.6
0
5
10
15
20
|ρ|, mean=0.17
The frequency of occurrence
(a) 3-11 GHz.
0 0.2 0.4 0.6
0
5
10
15
20
|ρ|, mean = 0.36
The frequency of occurrence
(b) 55-65 GHz.
Fig. 8: Spatial maps and histograms of the calculated Pearson
correlation coefficient |1ρα|∀α.
of the lower band from 3 GHz to 11 GHz. The sub-bands are
identified such that their relative bandwidth to theirs center
frequency is kept equal to the case of the relative bandwidth
of the upper band from 55 GHz to 65 GHz. The relative
bandwidth is given as:
Br=fmmw
2−fmmw
1
fmmw
c
,(12)
where fmmw
1=55GHz, fmmw
2=65GHz and
fmmw
c=60 GHz is the center frequency. Now, the calculation
of the sub-bands of the 3-11 GHz band is as follows:
fuwb
i+1 =fuwb
i(−Br−2)
Br−2,(13)
where fuwb
1=3GHzand fuwb
max(i)=11GHz. The calculated
values fuwb
i∀iare stated in Figure 9 together with maps of
the Pearson correlation coefficients and theirs histograms. It is
notable that the correlation coefficients are higher than in the
previous case, where we compare the entire 3-11 GHz band
and also, according to the histograms, theirs distributions are
similar to the case of the correlation of the 55-65 GHz band (by
visual inspection, the span of |ρα|is approximately from 0.1
to 0.6). It is also visible that the correlation slightly decreases
with increasing center frequency fc.
Please note that the last uwb sub-band from
8.19 GHz to 11 GHz exhibits a slightly narrower relative
bandwidth Brcompared to other examined sub-bands. For
the equal relative bandwidth among all sub-bands, an upper
frequency of 11.41 GHz is required, but the measurement is
limited to 11 GHz. Data for 2hα(n)do not deviate from the
presented 1hα(n).
C. Collinearity of PDP
As opposed to the previously exploited definitions of PDP
in (3), the PDPs are now calculated from a specified subset
of measurement grid points. The subset is defined by an
averaging window wg, where gdenotes the spatial location
of the window wgrespective to the measurement grid. The
window wgslides over the measurement grid with a step
of one measurement grid point. The graphical representation
of the sliding window wgand its movement within the
measurement grid is in Figure 10. Please note that the window
wgis of square shape and of variable size ranging from 2 to 9.
Now, taking into account only the measured data covered
by the averaging window wg, the PDP is calculated as defined
in (3). Then, in order to determine a sufficient size of the
averaging window wg, we evaluate the collinearity of the PDPs
9
0 0.2 0.4 0.6
0
5
10
15
20
ρ for 9.68 GHz to 11 GHz
|ρ|, mean=0.26
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 9.68 GHz to 11 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6
0
5
10
15
20
ρ for 8.19 GHz to 9.68 GHz
|ρ|, mean=0.18
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 8.19 GHz to 9.68 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6
0
5
10
15
ρ for 6.93 GHz to 8.19 GHz
|ρ|, mean=0.21
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 6.93 GHz to 8.19 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6
0
5
10
15
ρ for 5.86 GHz to 6.93 GHz
|ρ|, mean=0.27
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 5.86 GHz to 6.93 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6
0
5
10
15
20
ρ for 4.96 GHz to 5.86 GHz
|ρ|, mean=0.24
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 4.96 GHz to 5.86 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6
0
5
10
15
ρ for 4.2 GHz to 4.96 GHz
|
ρ
|, mean=0.26
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 4.2 GHz to 4.96 GHz
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8
0
5
10
15
ρ for 3.55 GHz to 4.2 GHz
|ρ|, mean=0.39
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 3.55 GHz to 4.2 GHz
0
0.2
0.4
0.6
0.8
0 0.2 0.4 0.6 0.8
0
5
10
15
ρ for 3 GHz to 3.55 GHz
|ρ|, mean=0.38
The frequency of occurrence
2 4 6 8 10
2
4
6
8
10
|ρ| for 3 GHz to 3.55 GHz
0
0.2
0.4
0.6
0.8
1
Fig. 9: Spatial maps and histograms for position κ=1of the sub-bands (selected from the 3-11 GHz band) having equal
relative bandwidth to the relative bandwidth of the 55-65 GHz band.
for all available window sizes. The collinearity of PDP [35]
is defined as:
cg(PdB,ref ,P
dB,g)= |tr(PdB,ref PH
dB,g)|
||PdB,ref ||F||PdB,g||F
,∀g, (14)
where PdB,ref is the selected reference PDP for g=1and
PdB,g are the PDPs calculated from the respective spatial
windows wg. Next, ||.||Fdenotes the Frobenius norm, (.)H
represents the complex conjugate transpose and tr(.)is the
matrix trace.
For both the frequency bands examined, the PDP collinear-
ity results are plotted in Figure 11. In addition, the PDP
collinearity has also been evaluated for different averaging
window sizes with practically the same results. The mean and
standard deviations of the PDP collinearity are shown in Figure
Fig. 10: 10 ×10 measurement grid with marked averaging
windows wgfor PDP calculation. The averaging window size
ranges from 2 to 9.
10
12. As can be seen, the collinearity difference for various
window sizes is negligible, thus we claim that for the examined
environment a 2×2 window is sufficient for averaging out the
effect of small scale variations.
For all cases, the collinearity slightly decreases with the
distance between examined PDPs in both directions of the
measurement grid. This phenomenon is caused by a strong
reverberation producing very similar PDPs within up to 10
wavelengths. Here, please keep in mind that the term wave-
length is not rigorous as the wavelength is, of course, variable
within the frequency bands examined.
2 4 6 8 10
2
4
6
8
10
0.9985
0.999
0.9995
1
(a) 3-11 GHz.
2 4 6 8 10
2
4
6
8
10
0.9985
0.999
0.9995
1
(b) 55-65 GHz.
Fig. 11: The PDP collinearity evaluation for the 3-11 GHz
and 55-65 GHz bands. In this case, the size of the averaging
window wgis 2. Figures are plotted for κ=1.
V. C HANNEL MODEL VALIDATION
In order to validate the presented in-vehicle channel model,
we have visualized the two-sample K-S test in Figure 13.
The comparison is performed for all the measured squared
channel impulse responses |1,2hα
dB(n)|2with its corresponding
piecewise linear model of PDP 1,2P∗
dB(n)+ξ.Wehave
evaluated the two-sample K-S test [37] according to:
κDmax =sup
α
{sup
nF(|κhα
dB(n)|2)−F(|κh∗
dB(n)|2),∀α},
(15)
23456789
Grid size
0.9986
0.9988
0.999
0.9992
0.9994
0.9996
0.9998
1
Mean collinearity
UWB, =1
UWB, =2
MMW, =2
MMW, =1
Fig. 12: Comparison of the mean PDP collinearity for both the
frequency bands examined and for the averaging window size
wgshowing a negligible difference between the listed window
sizes. Note: the error bars are negligible for larger widow sizes
due to growing averaging effect.
where Fis the cumulative distribution function and the sup
operator stands for supremum . Another useful metric is a
mean square error (MSE) between the cumulative distributions
of measured |κhα
dB(n)|2and modeled |κh∗
dB(n)|2:
κMSE =E{EF(|κhα
dB(n)|2)−F(|κh∗
dB(n)|2)2,∀α}.
(16)
Finally, Table V evaluates the goodness-of-fit metrics calcu-
lated according to (15) and (16). The maximal deviation of the
data measured compared to the model Dmax is highest for the
55-65 GHz band κ=2probably due to the highest number of
resolvable clusters. Visual inspection of the presented metrics
could be performed via Figure 13. Considering the position
κ=1, the value of Dmax is higher for the 3-11 GHz band
since the maximal deviation of the model is higher, compared
to the 55-65 GHz band. On the other hand, the MSE is higher
for the 55-65 GHz as the model shows slightly wider range
mismatch (see Fig. 13).
TABLE V: Table of goodness-of-fit metrics evaluation
position Dmax MSE
3-11 GHz κ=1 39.41 4.83
κ=2 31.46 5.31
55-65 GHz κ=1 30.37 11.93
κ=2 41.65 15.56
-130 -120 -110 -100 -90 -80 -70 -60
Received signal amplitudes [dB]
0
0.2
0.4
0.6
0.8
1
Cumulative probability
55-65 GHz
|1hdB
(n)|2
model 1P*
dB(n) +
3-11 GHz
Fig. 13: [red] Amplitudes of the piecewise model of PDP
|1h∗
dB(n)|2=1P∗
dB(n)+ξand [blue] the measured absolute
values of the CIRs |1hα
dB(n)|for both examined frequency
bands.
VI. CONCLUSION
In the environment under investigation, the frequency band
of 3-11 GHz is without the appearance of a high number of
resolvable clusters (agrees with [38]). The two visible clusters
are considered as direct/dominant component and reverberant
diffuse tail. Hence, the number of clusters is 2, with a cluster
decay rate of around 10 dB. On the other hand, the 55-65 GHz
11
frequency band exhibits a more evident clustering behavior.
The number of resolvable clusters ranges from 3 to 5, with a
cluster decay rates of 3.02 dB and 15.6 dB.
The RMS delay spread parameter is, due to the strong re-
vibration of the environment, rather high. For the 3-11 GHz
band the RMS delay spread is 23 ns, whereas for the
55-65 GHz band the RMS delay spread is almost 5-times
smaller, around 5 ns. The average delay is for both cases more
than twice the RMS delay spread.
Regarding the spatial stationarity of the measured CIRs for
the two frequency bands examined, we have observed a sig-
nificantly lower achievable diversity in the 55 - 65 GHz band.
If we compare the overall bandwidths, the lower achievable
diversity with a mean value of the Pearson correlation of
around 0.36, in the 55-65 GHz band might be problematic
in terms of the exploitation of transmit diversity MIMO
schemes. On the other hand, the 3-11 GHz band exhibits lower
correlations in the range of 0.18. Now, if the comparison
is done such that the 3-11 GHz band is divided into sub-
bands exhibiting equal relative bandwidth as the 55-65 GHz
band, the achievable diversity is approximately the same.
The collinearity slightly decreases with the distance between
examined PDPs in both directions of the measurement grid.
This phenomenon is common to both the 3-11 GHz and 55-65
GHz bands and is probably caused by a strong reverberation
producing very similar PDPs within up to approximately 10
wavelengths. We also show that increasing the size of the
averaging window wgproduces higher collinearity values;
however, the increase is very small, it is in the order of 10−4.
The 55-65 GHz band exhibits higher PDP collinearity than the
3-11 GHz band.
In this paper we propose a linear piecewise model of PDPs
for both frequency bands. To authentically simulate the in-
vehicle radio environment, we also provide parameters of the
GEV random process in order to model small-scale signal
fading in the delay domain. The resulting composition of
the linear piecewise model of PDP and the superimposed
model of SSF is validated by a comparison of cumulative
probabilities as in the case of the Kolmogorov-Smirnov test.
Our models show a very good match of the model with real-
world measured data.
Although the 55-65 GHz band shows a higher atmospheric
attenuation and is generally presumed to suffer by high
shadowing losses, this paper shows the possibility to exploit
this band for short-range communication even for non-line of
sight (NLOS) cases in a tight and reflections-rich environment.
The findings of this measurement campaign are exactly valid
only for one particular vehicle type. On the other hand,
authors expect similar results in all examined parameters also
for vehicles sharing similar characteristics as the geometry
or interior paneling. To enhance the reproducibility of our
research, measured data are publicly available at: http://www.
radio.feec.vutbr.cz/GACR-13-38735S/
ACKNOWLEDGMENT
This work was supported by the Czech Science Foundation
project No. 13- 38735S Research into wireless channels for
intra-vehicle communication and positioning, and was per-
formed in laboratories supported by the SIX project, No.
CZ.1.05/2.1.00/03.0072, the operational program Research
and Development for Innovation. The cooperation with Chris-
tian Doppler Laboratory for Wireless Technologies for Sus-
tainable Mobility is also gratefully acknowledged. This work
was supported by the SoMoPro II programme, Project No.
3SGA5720 Localisation via UWB, co-financed by the People
Programme (Marie Curie action) of the Seventh Framework
Programme of EU according to the REA Grant Agreement No.
291782 and by the South-Moravian Region. This publication
reflects only the author’s views and the Union is not liable
for any use that may be made of the information contained
therein.
REFERENCES
[1] C. Sommer and F. Dressler, Vehicular Networking. Cambridge Univer-
sity Press, 2014.
[2] N. Lu, N. Cheng, N. Zhang, X. Shen, and J. Mark, “Connected vehicles:
Solutions and challenges,” IEEE Journal Internet of Things, vol. 1, no. 4,
pp. 289–299, Aug 2014.
[3] G. Leen and D. Heffernan, “Expanding automotive electronic systems,”
Computer, vol. 35, no. 1, pp. 88–93, 2002.
[4] G. Lasser and C. Mecklenbrauker, “Channel model for tyre pressure
monitoring systems (TPMS),” in Proceedings of the Fourth European
Conference on Antennas and Propagation (EuCAP). IEEE, 2010.
[5] M. Win and R. Scholtz, “Characterization of ultra-wide bandwidth wire-
less indoor channels: a communication-theoretic view,” IEEE Journal on
Selected Areas in Communications, vol. 20, no. 9, pp. 1613–1627, 2002.
[6] R.-M. Cramer, R. Scholtz, and M. Win, “Evaluation of an ultra-
wide-band propagation channel,” IEEE Transactions on Antennas and
Propagation, vol. 50, no. 5, pp. 561–570, 2002.
[7] U. Demir, C. Bas, and S. Ergen, “Engine compartment UWB channel
model for intravehicular wireless sensor networks,” IEEE Transactions
on Vehicular Technology, vol. 63, no. 6, pp. 2497–2505, July 2014.
[8] M. Schack, J. Jemai, R. Piesiewicz, R. Geise, I. Schmidt, and T. Kurner,
“Measurements and analysis of an in-car UWB channel,” in Vehicular
Technology Conference, 2008. VTC Spring 2008. IEEE, 2008, pp. 459–
463.
[9] T. Kobayashi, “Measurements and characterization of ultra wideband
propagation channels in a passenger-car compartment,” in 2006 IEEE
Ninth International Symposium on Spread Spectrum Techniques and
Applications, 2006, pp. 228–232.
[10] T. Tsuboi, J. Yamada, and N. Yamauchi, “UWB radio propagation
inside vehicle environments,” in 7th International Conference on ITS
Telecommunications, 2007. ITST ’07., 2007, pp. 1–5.
[11] M. Schack, R. Geise, I. Schmidt, R. Piesiewiczk, and T. Kurner, “UWB
channel measurements inside different car types,” in 3rd European
Conference on Antennas and Propagation, 2009. EuCAP 2009., 2009,
pp. 640–644.
[12] A. Moghimi, H.-M. Tsai, C. Saraydar, and O. Tonguz, “Characterizing
intra-car wireless channels,” IEEE Transactions on Vehicular Technol-
ogy, vol. 58, no. 9, pp. 5299–5305, 2009.
[13] J. Blumenstein, T. Mikulasek, R. Marsalek, A. Prokes, T. Zemen, and
C. Mecklenbruker, “Measurements of ultra wide band in-vehicle channel
- statistical description and TOA positioning feasibility study,” EURASIP
Journal on Wireless Communications and Networking, p. 15, 2015.
[14] P. C. Richardson, W. Xiang, and W. Stark, “Modeling of ultra-wideband
channels within vehicles,” IEEE Journal on Selected Areas in Commu-
nications, vol. 24, no. 4, pp. 906–912, 2006.
[15] J. Blumenstein, T. Mikulasek, R. Marsalek, A. Chandra, A. Prokes,
T. Zemen, and C. Mecklenbrauker, “In-vehicle UWB channel mea-
surement, model and spatial stationarity,” in Vehicular Networking
Conference (VNC), 2014 IEEE. IEEE, 2014, pp. 77–80.
[16] Federal Communications Commission and others, “Revision of part
15 of the commission’s rules regarding ultra-wideband transmission
systems,first report and order, FCC 02,” V48, April, 2002.
[17] R. Thoma, O. Hirsch, J. Sachs, and R. Zetik, “UWB sensor networks
for position location and imaging of objects and environments,” in
The Second European Conference on Antennas and Propagation, 2007.
EuCAP 2007., 2007, pp. 1–9.
12
[18] R.-R. Lao, J.-H. Tarng, and C. Hsiao, “Transmission coefficients mea-
surement of building materials for UWB systems in 3-10 GHz,” in
Vehicular Technology Conference, 2003. VTC 2003-Spring. The 57th
IEEE Semiannual, vol. 1. IEEE, 2003, pp. 11–14.
[19] E. Ben-Dor, T. Rappaport, Y. Qiao, and S. Lauffenburger, “Millimeter-
wave 60 GHz outdoor and vehicle AOA propagation measurements
using a broadband channel sounder,” in Global Telecommunications
Conference (GLOBECOM 2011), 2011 IEEE, Dec 2011, pp. 1–6.
[20] M. Schack, M. Jacob, and T. Kurner, “Comparison of in-car UWB and
60 GHz channel measurements,” in 2010 Proceedings of the Fourth
European Conference on Antennas and Propagation (EuCAP). IEEE,
2010, pp. 1–5.
[21] J. Blumenstein, T. Mikulasek, A. Prokes, T. Zemen, and C. Mecklen-
brauker, “Intra-vehicular path loss comparison of UWB channel for
3-11 GHz and 55-65 GHz,” in 15th IEEE International Conference on
Ubiquitous Wireless Broadband (ICUWB). IEEE, Sep 2015.
[22] T. Rappaport, J. Murdock, and F. Gutierrez, “State of the art in 60-
GHz integrated circuits and systems for wireless communications,”
Proceedings of the IEEE, vol. 99, no. 8, pp. 1390–1436, Aug 2011.
[23] C. Gustafson, K. Haneda, S. Wyne, and F. Tufvesson, “On mm-wave
multipath clustering and channel modeling,” IEEE Transactions on
Antennas and Propagation, vol. 62, no. 3, pp. 1445–1455, March 2014.
[24] J. Blumenstein, J. Vychodil, M. Pospisil, T. Mikulasek, and A. Prokes,
“Effects of vehicle vibrations on mm-wave channel: Doppler spread
and correlative channel sounding,” in IEEE PIMRC 2016 - accepted
for publication.
[25] A. F. Molisch, Wireless communications. John Wiley & Sons, 2010,
vol. 15.
[26] J. Blumenstein, T. Mikulasek, R. Marsalek, A. Prokes, T. Zemen, and
C. Mecklenbrauker, “In-vehicle mm-wave channel model and measure-
ment,” in Vehicular Technology Conference (VTC Fall), 2014 IEEE 80th.
IEEE, 2014, pp. 1–5.
[27] A. A. M. Saleh and R. Valenzuela, “A statistical model for indoor mul-
tipath propagation,” Selected Areas in Communications, IEEE Journal
on, vol. 5, no. 2, pp. 128–137, 1987.
[28] A. Chandra, A. Prokes, T. Mikulasek, J. Blumenstein, P. Kukolev,
T. Zemen, and C. F. Mecklenbrauker, “Frequency-domain in-vehicle
UWB channel modeling,” IEEE Transactions on Vehicular Technology,
vol. 65, no. 6, pp. 3929–3940, June 2016.
[29] H. Xu, V. Kukshya, and T. Rappaport, “Spatial and temporal character-
istics of 60-GHz indoor channels,” IEEE Journal on Selected Areas in
Communications,, vol. 20, no. 3, pp. 620–630, Apr 2002.
[30] H. Hashemi, “Impulse response modeling of indoor radio propagation
channels,” IEEE Journal on Selected Areas in Communications, vol. 11,
no. 7, pp. 967–978, 1993.
[31] F. J. Harris, “On the use of windows for harmonic analysis with the
discrete Fourier transform,” Proceedings of the IEEE, vol. 66, no. 1, pp.
51–83, 1978.
[32] T. Pedersen, G. Steinbock, and B. Fleury, “Modeling of reverberant radio
channels using propagation graphs,” IEEE Transactions on Antennas and
Propagation, vol. 60, no. 12, pp. 5978–5988, Dec 2012.
[33] S. Kotz and S. Nadarajah, Extreme value distributions. World Scientific,
2000.
[34] H. Hashemi, “The indoor radio propagation channel,” Proceedings of
the IEEE, vol. 81, no. 7, pp. 943–968, 1993.
[35] A. Paier, T. Zemen, L. Bernado, G. Matz, J. Karedal, N. Czink, C. Du-
mard, F. Tufvesson, A. Molisch, and C. Mecklenbrauker, “Non-WSSUS
vehicular channel characterization in highway and urban scenarios at 5.2
GHz using the local scattering function,” in International ITG Workshop
on Smart Antennas, 2008. WSA 2008, Feb 2008, pp. 9–15.
[36] R. He, O. Renaudin, V. M. Kolmonen, K. Haneda, Z. Zhong, B. Ai, and
C. Oestges, “Characterization of quasi-stationarity regions for vehicle-
to-vehicle radio channels,” IEEE Transactions on Antennas and Propa-
gation, vol. 63, no. 5, pp. 2237–2251, May 2015.
[37] F. J. Massey Jr, “The Kolmogorov-Smirnov test for goodness of fit,”
Journal of the American statistical Association, vol. 46, no. 253, pp.
68–78, 1951.
[38] A. Ott, U. Siart, T. Eibert, O. Klemp, and R. Steffen, “Enhanced
investigations of 433 MHz - 10 GHz time invariant wireless in-car
communication channels,” in Microwave Conference (GeMIC), 2011
German, March 2011, pp. 1–4.
Jiri Blumenstein was born in Prostejov, Czech
Republic, in 1984. He received his Ph.D. degree in
electrical engineering at Brno University of Technol-
ogy in 2013.
In 2011 he worked as a researcher at the Institute
of Telecommunications, Vienna University of Tech-
nology, Austria. At present, he is a researcher at the
Department of Radio Electronics, Brno University
of Technology, Czech Republic. He cooperates with
several companies in the area of applied research
into wireless systems as well as in the area of the
fundamental research financed by the Czech Science Foundation. His research
interests are signal processing, physical layer of communication systems,
channel measurement and modeling, and wireless system design.
Ales Prokes , Prof. Ing., Ph.D., *1963, graduated
from BUT in 1989, obtained the Ph.D. degree in
1999. In 2006 he habilitated and in 2011 he became
Professor at BUT. He has been joint researcher in
several grant projects concerned with problems of
optical an RF communications. He has dealt with
the analysis and optimization of optical receivers,
atmospheric channel effects on optical signal, optical
channel modelling, evaluation of free space optics
availability and reliability, further with millimeter
wave (MMW) and ultra-wideband (UWB) channel
measurement, analysis and modeling and also with the ranging accuracy
evaluation and localization in both the UWB and MMW bands.
Aniruddha Chandra (M’08-SM’16) received B.E.,
M.E., and Ph.D. degrees from Jadavpur University,
Kolkata, India in 2003, 2005, and 2011 respectively.
He joined Electronics and Communication Engineer-
ing department, National Institute of Technology,
Durgapur, India in 2005 as a Lecturer. He is cur-
rently serving as an Assistant Professor there. From
August 2011 to January 2012 he was a visiting
Assistant Professor at Asian Institute of Technol-
ogy, Bangkok. In 2014, he received Marie Curie
fellowship to pursue postdoctoral studies in Brno
University of Technology, Czech Republic. Dr. Chandra published about 80
research papers in referred journals and peer-reviewed conferences. He has
also delivered several invited lectures. His primary area of research is physical
layer issues in wireless communication.
Tomas Mikulasek received masters degree and doc-
toral degree from the Brno University of Technology
in 2009 and 2013, respectively. At the present, he
is with the Department of Radio Electronics, Brno
University of Technology in the Czech Republic, as a
researcher. His research interest is focused on design
of centimeter and millimeter wave antennas.
13
Roman Marsalek graduated at the Brno Univer-
sity of Technology in 1999 and received the doc-
toral degree from Universit´
e de Marne-La- Vall´
e,
´
Ecole Suprieure d’Ing´
enieurs en ´
Electronique et
´
Electrotechnique de Paris, France in 2003. He is
currently assistant professor at the Department of
Radio Electronics, Brno University of Technology
in the Czech Republic. His research interests are in
wireless communications theory and applied digital
signal processing
Thomas Zemen (S’03-M’05-SM’10) received the
Dipl.-Ing. degree (with distinction) in electrical en-
gineering in 1998, the doctoral degree (with distinc-
tion) in 2004 and the Venia Docendi (Habilitation)
for ”Mobile Communications” in 2013, all from
Vienna University of Technology. From 1998 to
2003 he worked as Hardware Engineer and Project
Manager for the Radio Communication Devices
Department, Siemens Austria. From 2003 to 2015
Thomas Zemen was with FTW Forschungszentrum
Telekommunikation Wien and Head of the ”Signal
and Information Processing” department since 2008. Since 2014 Thomas
Zemen has been Senior Scientist at AIT Austrian Institute of Technology.
He is the author or coauthor of four books chapters, 20 journal papers and
more than 70 conference communications. His research interests focuses
on reliable, low-latency wireless machine-to-machine communications for
highly automated vehicles, sensor and actuator networks, vehicular channel
measurements and modeling, time-variant channel estimation, cooperative
communication systems and interference management. Dr. Zemen is docent
at the Vienna University of Technology and serves as Associate Editor for the
IEEE Transactions on Wireless Communications.
Christoph Meckenbr¨
auker (S’88-M’97-SM’08)
was born in Darmstadt, Germany, in 1967. He re-
ceived the Dipl-Ing. degree in Electrical Engineer-
ing from Vienna University of Technology in 1992
and the Dr.-Ing. degree from Ruhr-University of
Bochum in 1998, respectively. His doctoral thesis
was awarded with the Gert Massenberg Prize.
From 1997-2000, he worked for the Mobile Net-
works Radio department of Siemens AG Austria
where he participated in the European framework of
ACTS 90 FRAMES. He was a delegate to the Third
Generation Partnership Project (3GPP) and engaged in the standardization of
the radio access network for UMTS.
Since June 2000, he was a senior researcher at the Telecommunications
Research Center Vienna (ftw.) in the field of mobile communications, key
researcher since November 2002, and proxy since July 2003. Between 2006
and 2009, he coordinated the Sixth Framework project ”Multiple-Access
Space-Time Coding Testbed” (MASCOT) on behalf of ftw. He leads the
Special Interest Group on mobile-to-mobile communications within COST
Action 2100 Pervasive Mobile and Ambient Wireless Communications.
In 2006, he joined the Institute of Communications and Radio Frequency
Engineering at Vienna University of Technology as a full professor. Since July
2009, he leads the newly founded Christian Doppler Laboratory for Wireless
Technologies for Sustainable Mobility. His current research interests include
radio interfaces for future peer-to-peer networks (car-to-car communications,
personal area networks, and wireless sensor networks), ultra-wideband radio
(UWB) and MIMO-OFDM based transceivers (UMTS long term evolution,
WiMax, and 4G).
Christoph Mecklenbr¨
auker is a member of the IEEE, the Antennas and
Propagation Society, the Vehicular Technology society, the Signal Processing
society, and EURASIP. He is the councilor of the IEEE Student Branch Wien.
He is associate editor of the EURASIP Journal of Applied Signal Processing.
