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.
Time-Domain Nonstationary Intra-Car Channel
Measurement in 60 GHz Band
Ales Prokes, Josef Vychodil, Martin Pospisil, Jiri Blumenstein, Tomas Mikulasek, Aniruddha Chandra
Department of Radio Electronics
Brno University of Technology
Brno, Czech Republic
prokes@feec.vutbr.cz
Abstract— The paper deals with a time varying intra-vehicle
channel measurement in the 60 GHz millimeter wave (MMW)
band using a unique time-domain channel sounder built from off-
the-shelf components and standard measurement devices. The aim
of the presented work is to describe the sounder architecture, the
primary data processing technique, and preliminary results of
measurements aimed at the effect of vehicle vibrations and
twisting, and passengers moving in a car cabin. As the amplitude
of the car cabin vibration and twisting had been supposed to be
comparable with the MMW wavelength, some effect on the
channel impulse response (CIR) and consequently on the delay-
Doppler spread (DDS) was expected. Preliminary results confirm
the correctness of this assumption and allow us to assess the effect
of different above-mentioned phenomena. We tested the effect of
driving the car on different types of road (bumpy road, flat road,
highway, etc.). For comparison purposes we use DDS and statistics
of CIR amplitudes calculated using the correlative technique
applied to the pseudorandom binary sequence.
Keywords—millimeter wave; channel measurement; channel
sounder; channel impulse response; Doppler spread
I. I
NTRODUCTION
Recent societal trends show a continuously increasing
demand for transport of people and goods. To make transport
safer, more efficient, and cleaner, various traffic telematics
services are currently under development. Besides the above
requirements we have to consider the fact that traveling people
spend a lot of time in public and private transportation. This time
can be made more enjoyable by ensuring internet access or by
installing some entertainment systems in vehicles. Hence in
recent years there has been a growing demand for WPANs
offering high data rates for short range intra-vehicle
communication applications. Manufacturers of vehicles,
aircraft, combat vehicles, etc. have a great interest in replacing
wired communication links by wireless ones in order to save
installation costs. These systems allow, for example, connecting
the (rear) seat entertainment system without using cables,
integrating the passenger mobile devices into the vehicle
network or interconnecting a variety of car sensors with a central
unit and actuators. There are many locations in the vehicle where
the wired connection is impractical or even impossible (steering
wheel, tires, and windshields). Moreover, as the number of intra-
vehicle sensors and devices steadily increases, the wiring
harness is becoming one of the heaviest components in a modern
vehicle and as such is considered to have a significant impact on
fuel consumption and ecology [1], [2].
There are several technologies such as Bluetooth or ZigBee
available today that can be considered as candidates for intra-
vehicle communication [2]. The Bluetooth provides sufficient
bandwidth for multimedia but requires high transmitting power
and connecting a large number of nodes is not straightforward.
The ZigBee lacks throughput for multimedia applications and
exhibits significant latency, which is unacceptable for certain
intra vehicle scenarios. Due to the limitations of the above
systems the other frequency bands such as ultra-wide band
(UWB) defined by the IEEE 802.15.x recommendation (3.1 to
10.7 GHz) or millimeter wave (MMW) band (57 to 64 GHz) are
being investigated for short-range communication purposes.
The intra-vehicle environment exhibits very specific
propagation characteristics such as multipath propagation,
shadowing, and non-stationary behavior. However, its
investigation has been pursued predominantly for stationary
case. Many works are based on the frequency-domain
measurement using a vector network analyzer, which offers an
excellent dynamic range (100–120 dB) but very long
measurement times. With this approach, the complex transfer
function (scattering parameter s
21
referred to as the forward
voltage gain) being measured is converted into a channel
impulse response (CIR) using a windowed inverse fast Fourier
transform. An averaged set of the squared CIRs then forms a
power delay profile (PDP), which is the most frequently used
characteristic for both small- and large-scale channel feature
evaluation.
In [3] and [4] the UWB and the MMW intra-car channels are
compared for different scenarios (different antenna
configurations, empty and occupied car). The comparison is
performed through the delay spread and the path-loss calculated
from CIR or PDP. It is obvious that on the one hand the MMW
band suffers from higher propagation and penetration losses
(stronger shadowing effect) compared to the UWB but on the
second hand the lower delay spread implies a lower complexity
channel estimation and the usage of high-gain steerable antennas
in a small physical form allows an easier MMW system
implementation. The suitability of the MMW band for the intra-
car communication based on signal- to-noise ratio (SNR) and
interference evaluation is given in [5].
Fig. 1. Block diagram of time-domain channel sounder.
A time-domain intra-car channel measurement, which
allows very fast measurement periods, but at the cost of low
dynamic range (30 – 40 dB), has (according on our survey) been
performed only for the UWB. Some experiments are performed
via sounding the channel with narrow (picoseconds) Gaussian
or common impulses. The CIR is then calculated from the
received signal usually using the CLEAN deconvolution
algorithm. While in [6] and [7] the authors deal with stationary
channel characterization using the RMS delay spread, the cluster
arrival rate, or the mean excess delay, in [8] a stochastic tapped
delay line model for stationary but also for moving car scenario
is created. Another time-domain measurement technique results
from sounding the channel with a pseudo-random binary
sequence (PRBS). A correlation principle used for the CIR
calculation provides a relatively high correlation gain which is
the biggest advantage over the narrow impulse channel sounding
approach. The PRBS based channel sounder for multiple time-
variant radio propagation channel measurement in different
UWB bands and in the 59.5 – 66.5 GHz MMW band is
introduced in [9]. It performs up to 100 measurements per
second and offers 67 dB of instantaneous dynamic range and
7 GHz of bandwidth. Another UWB channel sounder using
correlation technique can be found in [10]. However, intra-
vehicle channel measurement using such a sounder has probably
not been published yet.
The main contributions of the paper are
• introduction of a novel concept of time-domain PRBS
based MMW channel sounder built from off-the-shelf
components and standard measurement devices,
• presentation of original preliminary results of
measurements aimed at the effect of vehicle vibrations
and twisting due to driving the car on different types of
roads, and passengers moving in the car cabin.
We analyzed the above effects through the delay-Doppler
spreading function (DDSF) and through the statistical
distribution of the CIR magnitudes.
The rest of the paper is organized as follows. Section 2
describes the measurement setup including the antenna radiation
characteristic and the list of the fundamental parameters. Section
3 briefly informs about the techniques of signal processing used
for both the CIR and the DDSF calculation. Then, in the most
important section, Section 4, the DDSF evaluated for different
scenarios and statistics of the CIR are given. A summary of the
paper is given in the Conclusion.
II. M
EASUREMENT
S
ETUP
The time-domain measurement testbed (see Fig. 1) is created
using a Tektronix MSO72004C (20 GHz, 50 GS/s) Mixed
Signal Oscilloscope, an Anritsu MP1800A Signal Quality
Analyzer (working as a generator of PRBS), a SiversIma
FC1003V/01 up/down converter changing the frequency from
baseband to the MMW band in transmitter (TX) and back in
receiver (RX), and an Agilent 83752A a frequency stable, low
phase noise generator (carrier generator for up/down
conversion). In order to compensate for large propagation loss
in the MMW band the Quinstar QLW-50754530-I2 low noise
amplifier and QPW-50662330-C1 power amplifier are used. To
achieve good time synchronization between the MP1800A and
MSO72004C both devices are supplied with a 10 MHz reference
frequency generated by the oscilloscope. For the case of moving
car channel measurement the sounder is placed in the car and
supplied with uninterruptible power supply (DELL 5600W 4U
230V). The filter 0.1 – 5 GHz is optional and limits together with
the internal up/down converter filter the frequency band of
PRBS.
ABCD
MIXER
Input I
Input Q
Output I
Output Q
Isolator
Quinstar QLW-
50754530-I2
Adapter
PA
LNA
Adapter
Phase stable coaxial cable
MMW
RX
MMW
TX
filter
Phase stable coaxial cables
Tektronix MSO72004C
Quinstar QPW-
50754530-I2
PC & LabView
Anritsu MP1800A
14.7500 GHz
Agilent 83752A
Ethernet cable
Siversima
FC1003V/01
15 dBm
Cooler
0.1 - 5.0 GHz
59.1-64 GHz
12.0 V8.0V
To PA
To LNA
DELL 5600W 4U 230V
Power 230V DIAMETRAL L240R51D
Fig. 2. Simulated radiation patterns of the open waveguide.
The operation of the testbed is straightforward. The seamless
repeating PRBS of the length N
bit
= 2
k
– 1 bits, where k can in
our system vary from 7 to 31, and data rate R
PS
= 12.5 Gbit/s is
frequency limited to the 0.1 – 5 GHz bandwidth, up-converted
into the 59.1 – 64 GHz MMW band, and then fed to the power
amplifier with a gain of 30 dB through a 2.5 m phase stable
coaxial cable (Megaphase TM67-V1V1-138) with an
attenuation of 26 dB. The MMW signal is irradiated and
received using the open-ended waveguide antennas QuinStar
QWS-V02000. The radiation pattern of the waveguide is shown
in Fig. 2 [11]. It is obvious that radiation in the H-plane is more
directional but due to the large propagation loss of the MMW
signal the effect of indirect components is less important.
However, for the first experiments such a solution is reasonable.
The received signal then passes through the amplifier with a gain
of 30 dB and a noise figure of 4.5 dB. The waveguide isolator
prevents the receiver from oscillating. The quadrature down-
conversion produces two baseband signals, I and Q, which are
digitized and stored in the internal oscilloscope memory.
Because the converter includes the frequency multiplier by a
factor of 4, the generator output frequency is set to 14.75 GHz
(see e.g. upper frequency calculation: 5 GHz + 4×14.75 GHz =
64 GHz).
The repeating PRBS is chosen as the excitation signal due to
the very good circular correlation properties of selected PRBSs
which invoke using a very convenient circular auto-correlation
technique for the CIR calculation. The PRBS length then defines
the maximum observable time span T
max
= N
bit
/R
PS
. For the
chosen k = 11 we get N
bit
= 2047 and T
max
= 163.8 ns. Assuming
the speed of light c = 3×10
8
m/s we can obtain the maximum
observable distance L
max
= 49.13 m. Selected system parameters
are summarized in Table 1.
In order to suppress the oscilloscope wideband noise, we use
the 8 GHz user-selectable bandwidth limit filter. This option can
only be used when the full maximum sampling rate R
S
= 50 GS/s
is chosen. Considering the memory depth of the oscilloscope M
D
= 31.25 MSa per channel and the above mentioned time span
T
max
= 163.76 ns we can calculate important system parameters
such as the maximum measurement rate (number of
measurements per second), number of samples per CIR, number
of saved CIRs, and total measurement time (see Table 1).
Because the maximum measurement rate (e.g. 6.1 × 10
6
Meas/s
for k = 11) is too high for our needs we insert between two
consecutive CIR measurements the time delay T
D
= 1 ms.
TABLE I. L
IST OF
S
ELECTED
S
YSTEM
P
ARAMETERS
Relation k = 10 k= 11 k= 12
Number of PRBS bits [-] N
bit
= 2
k
– 1 1023 2047 4095
Maximum observable
time [ns] T
max
= N
bit
/R
PS
81.84 163.76 327.60
Maximum observable
distance [m] L
max
= cT
max
24.53 49.13 98.21
Max. measurement rate
[Meas/s × 10
6
]R
M
= 1/T
max
12.22 6.10 3.053
Number of samples per
CIR [-] N
CS
= N
bit
R
S
/R
PS
4092 8188 16380
Number of saved CIRs [-] N
CIR
= M
D
/N
CS
7636 3816 1907
Total measurement time
[μs] T
M
= N
CIR
T
max
624.8
Correlation gain [dB] G
C
= 20log(N
bit
) 60.20 66.22 72.25
The time delay is controlled by the oscilloscope and precisely
triggered by the signal quality analyzer. Such a sampling period
allows us to analyze the Doppler spread up to approximately
500Hz from the 3816 CIR records corresponding to the 3.816 s
time interval. For additional information about the channel
sounder modified for the UWB see [12].
The installation of the test-bed in Skoda Octavia 1.9 TDI car
is shown in Fig. 3. For all the measurements we placed the
transmitting antenna behind the rear seats on the right side of the
car while the receiving antenna was situated on the right side of
the dash board (see Fig. 4). Both antennas were aligned to the
center of the car cabin. The line-of-sight communication is
blocked by the passenger siting on the front seat.
III. S
IGNAL PROCESSING TECHNIQUES
As mentioned above the proposed channel sounder benefits
from the correlation technique which increases significantly the
dynamic range because the SNR of the receiver is improved by
the correlation gain. Since the dynamic range of the
MSO72004C oscilloscope is about 35 dB [13], the total dynamic
range approaches 100 dB theoretical limit (for k = 11).
Unfortunately, the CIR calculated using correlation of PRBS
passing through a nonlinear system (represented in our case by
amplifiers and mixers) exhibits plenty of spurs [14], which
markedly decreases the dynamic range. This phenomenon that
depends on the signal level and PRBS length and limits the
dynamic range of the sounder to 30–50 dB is now intensively
studied and suitable technique for the spurs mitigating is sought.
Fig. 3. Installation of the tesbed into car Skoda Octavia 1.9 TDI.
Fig. 4. Installation of the receiver (left) and transmitter (right).
The channel is assumed to behave as a linear time invariant
(LTI) system having an impulse response h(t). To estimate h(t)
the cross correlation R
xy
(t) between the LTI system output and
input may be employed in accordance with [15]
()ℎ
()∗
()(), (1)
where R
xx
(
τ
) is the autocorrelation function of the input signal
and ∗ denotes convolution. In order to obtain high correlation
gain we employ m-sequences [16] which belong to the family
of PRBS. The autocorrelation of the m-sequence is actually a
sharp triangle function (note that the theoretical circular
autocorrelation of an m-sequence is a Kronecker delta
function), however, for our purpose we can write R
xx
(
τ
) ≈
p
s
δ
(
τ
), where p
s
is a constant related to a signal power and
δ
(
τ
)
is the Dirac pulse. Thus, the cross correlation function is equal
to the system impulse response as follows
()ℎ
()∗
()()
ℎ(). (2)
The cross correlation R
xy
(t) can be calculated using
convolution. For this purpose let s
m
(t) denotes the periodical
excitation PRBS signal (LTI system input) and r(t) denotes the
received signal (LTI system output). Then under simplifying
condition
1 and considering the relation between
correlation and convolution [15] the CIR can be expressed as
ℎ()
(),
()
()∗
∗
(−). (3)
To accelerate the CIR calculation, equation (3) is transformed to
the frequency domain and back using the complex fast (inverse)
Fourier transform ((I)FFT)
ℎ() ()
∗
(−). (4)
An appropriate metric for the assessment of the car
vibrations and passenger movements is the delay-Doppler
spreading function [17]. The CIR calculation (4) corresponds to
a single measurement. As mentioned above we measured set of
N
CIR
responses in regular 1 ms intervals. Let us designate this
time by the letter . Then we have to distinguish the sampling
time (oscilloscope recording time) t from the measuring time
(time of regular CIR measurements) , as shown in Fig. 5, and
write the CIR as a two dimensional function ℎ(,). The DDSF
can then be written in the form
Fig. 5. Example of CIR magnitude for driving the car on a highway.
(,
)ℎ(, )
, (5)
where f
D
is the Doppler frequency and ℎ(, ) can be expressed
as a sum of P multipath components (paths) at any time , i.e.
ℎ(, )∑
()
()
,()
−
(), (6)
where
α
p
is the complex attenuation factor, f
D,p
is the Doppler
shift, and t
p
is the time delay associated with the p-th path. In
reality we calculate the DDSFs applying FFT to the measured
series where t is constant and varies from 0 to N
CIR
T
D
. For
this purpose we select only the cases where the CIR exceeds the
noise (e.g. for ∈〈40, 50〉 as is obvious from Fig. 4) and
in the following step we average them. Considering the fact,
that both the times t and are discrete (t = n/R
S
and
,
where n and k are integers), the average DDSF can be expressed
in the form
(
)
∑(
⁄,
)
, (7)
where n
1
is the sample ordinal number calculated as a median of
the ordinal numbers of all the measured CIR maxima in a single
record, and n
2
corresponds to the coordinate of the maximum
excess delay value.
Other metrics used to evaluate the time variant behavior of
the channel are the probability distribution function (PDF) and
the cumulative distribution function (CDF) of the CIR for
constant sampling time t.
IV. R
ESULTS
First we carried out a few reference measurements inside an
empty car parked in an underground garage. The measurement
devices were placed outside the car. We investigated the effects
of running engine at different revolutions per minute (rpm) and
the effects of a very loudly playing built-in audio system on the
car cabin vibrations. Then we installed the test bed into the car
as shown in Figs 3 and 4 and conducted a set of 120
measurements for the car traveling over different types of road
(flat road, cobblestone road, bumpy road with potholes, old and
new highways, etc.) at different speeds. The car was occupied
by a driver, a front passenger, and a passenger sitting behind the
driver (controlling the test-bed).
Fig. 6. Comparison of the DDSF magnitudes for different chirps.
A. Delay Doppler spread
For the reference measurements we prepared a set of MP3
sound files (chirps and pure sine waves) at frequencies between
20 Hz and 400 Hz. We also tested the influence of different
antenna locations. The DDSF for three different chirps
(harmonic signal swept from lower to upper frequency) is shown
in Fig. 6. It is evident that the effect of the lower frequencies is
more marked than the effect of the upper ones. There are also
some local minima and maxima caused probably by mechanical
resonances of particular objects in the car cabin and possibly
also by mechanical resonances of the transmitting or receiving
antennas. Frequencies above 150 Hz have almost no influence
on the channel. Note that the voltage density drop in dependence
on frequency is influenced probably also by the acoustic
pressure, generated by speakers, which is higher at the lower
frequencies.
The effect of the running engine is very similar. It can be
observed dominantly at low rpm as is obvious from Fig. 7. Note
that the four-stroke engine running at 800 rpm produces 1600
ignitions per minute which corresponds to 13.3 revolutions and
26.6 ignitions per second. It is obvious that the effect of the car
body vibration is minimal at the engine rpm corresponding to
common ride, which testifies to the good damping of engine-
induced vibrations in recent cars.
Fig. 7. Effect of running engine at different rpm on the DDSF magnitude.
Fig. 8. Effect of passengers moving quickly their hands and bodies.
An interesting result was obtained when the passengers were
moving their hands and bodies inside the stationary (parked) car.
As is obvious from Fig. 8 up to 45 Hz there are significant
components in the spectrum. This phenomenon was taken into
account during the measurements in the moving car. The
passengers tried to remain as motionless as possible, which was
very difficult particularly for the driver and the person
controlling the test-bed.
In the next step we evaluated all the measurements carried
out in the moving car and compared the DDSFs. The results are
ambiguous. The same scenario (the same road and speed) can
produce very different results. It is probably caused by
unintentional passenger movements. Due to the large amount of
measurements we can exclude such anomalous results and state
typical representatives for particular situations. An example of
DDSF representatives for two different roads is shown in Fig. 9.
It is obvious that the bumpy road causes more DDSF spectral
components. It is also evident that the magnitude of DDSF in
Figs. 8 and 9 is much larger than the magnitude in Figs. 6 and 7.
Hence the effect of the running engine on the moving car DDSF
is negligible. For completeness let us mention that the influence
of the surrounding environment (other vehicles) was not proved.
Fig. 9. DDSF magnitude calculated for the two different road surfaces.
DDSF voltage density (V/Hz½)
0 102030405060708090100
Frequency (Hz)
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
800 rpm
2520 rpm
DDSF voltage density (V/Hz½)
DDSF voltage density (V/Hz½)
Fig. 10. The PDF and the CDF of the CIR magnitude (stair plot) and their
fitting using GEV distribution (dashed line).
B. Probability density and cumulative density functions
To create a stistical channel model in the time-domain it is
nesesary to know the amplitude distribution ofℎ(, ). We tried
to find the best fit between all the measured data and a variety
of probability distributions. The best fit was obtained using the
generalized extreme value (GEV) distribution. It is defined by
the location parameter
μ
, the scale parameter
σ
> 0, and the
shape parameter
ξ
[18]. Examples of the PDF and the CDF of
the CIR magnitude and their fitting using the GEV distribution
for the case of bumpy road are shown in Fig 10. The plots are
depicted for the CIR coordinates n = tR
S
= n
1
+ 5i, where n
1
is
defined in Section III and i = 0,1,2… For all the CIRs we
obtained the following limits ∈〈0.001, 0.03〉,∈ 〈0.0005,
0.004〉,and ξ ∈〈−0.4, 0.1〉.
V. C
ONCLUSION
We propose a channel sounder realized without the need for
designing and manufacturing any electronic circuits or
equipment. It consists of standard measurement devices and
common microwave components, and offers very good
properties such as 5 GHz bandwidth, up to several millions
measurements per second, up to a few seconds record length (we
plan to extend it), and a satisfactory dynamic range.
By analysing the measured data we discovered that the
effects of a running engine or a loudly playing built-in audio
system are negligible when compared with the influence of car
body vibration or twisting caused by the car motion. Of
considerable effect on the channel time variance (expressed e.g.
by the DDSF) are the movements of passengers in the car-cabin,
which, among other things, are also caused by driving the car on
a bumpy road or negotiating a curve, or by accelerating or
braking. Unintentional passenger movement probably causes
certain randomnes in the data measured, and some records have
to be discarded when other effects such as car speed or road
quality are evaluated. The CIR probability distribution in
measurement time can be obtained satisfactorily using the GEV
distribution. Note that here unpublished analyses of the CIR in
sampling time can be found in many other works [3] - [7].
A
CKNOWLEDGMENT
The research described in this paper was financed by the
Czech Science Foundation, Project No. 13-38735S, by the
SoMoPro II programme, Project No. 3SGA5720 Localization
via UWB, co-financed by the People Programme (Marie Curie
action) of the Seventh Framework Programme (FP7) of EU
according to the REA Grant Agreement No. 291782 and by the
South-Moravian Region, and by National Sustainability
Program under grant LO1401. For the research, the
infrastructure of the SIX Center was used.
R
EFERENCES
[1] G. Leen and D. Heffernan, "Expanding automotive electronic systems,"
Computer, vol. 35, no. 1, pp. 88–93, Jan. 2002.
[2] N. Lu, N. Cheng, N. Zhang, X. Shen, J.W. Mark: "Connected Vehicles:
Solutions and Challenges," IEEE Internet Things J, vol. 1, no 4, pp. 289–
299, Aug. 2014.
[3] M. Schack, M. Jacob, and T. Kurner, "Comparison of in-car UWB and 60
GHz channel measurements," in Proc. IEEE EuCAP, Barcelona, Spain,
Apr. 2010, pp. 1–5.
[4] J. Blumenstein, T. Mikulasek, A. Prokes, T. Zemen, and C.
Mecklenbrauker, "Intra-vehicular path loss comparison of UWB channel
for 3-11 GHz and 55-65 GHz," in Proc. IEEE ICUWB, Montreal, Canada,
Oct. 2015, pp. 1–4
[5] M. Peter, R. Felbecker, W. Keusgen and J. Hillebrand, "Measurement-
Based Investigation of 60 GHz Broadband Transmission for Wireless In-
Car Communication," in Proc. IEEE VTC Fall, Anchorage, AK, USA,
Sep. 2009, pp. 1–5.
[6] W. Niu, J. Li and T. Talty, "Intra-Vehicle UWB Channel Measurements
and Statistical Analysis," in Proc. IEEE GLOBECOM, New Orleans, LO,
USA, Dec. 2008, pp. 1–5.
[7] A. Chandra, J. Blumenstein, T. Mikulášek, J. Vychodil, R. Maršálek, A.
Prokeš, T. Zemen, C. A. Mecklenbräuker, "Serial subtractive
deconvolution algorithms for time-domain ultra wide band in-vehicle
channel sounding," IET Intel Transport Syst, vol. 9, no. 9, pp. 870–880,
Nov. 2015.
[8] W. Niu, J. Li and T. Talty, "Intra-vehicle UWB channels in moving and
stationary scenarios," in Proc. IEEE MILCOM, Boston, MA, USA, Oct.
2009, pp. 1–6.
[9] R. Zetik, M. Kmec, J. Sachs, and R. S. Thomä, "Real-Time MIMO
Channel Sounder for Emulation of Distributed Ultrawideband Systems,"
Int. J. Antennas Propag., vol. 2014, Article ID 317683, pp. 1–16, 2014.
[10] A. Dezfooliyan, A. M. Weiner, "Evaluation of Time Domain Propagation
Measurements of UWB Systems Using Spread Spectrum Channel
Sounding," IEEE Trans. Antennas Propag., vol. 60, no. 10, pp. 4855–
4865, Oct. 2012.
[11] T. Mikulasek, J. Blumenstein, and A. Prokes, “Antennas utilized for intra-
vehicle 3-11 GHz and 55-65 GHz channel measurement,” in Proc. PIERS,
Shanghai, China, Aug. 2016, pp. 1–5.
[12] J. Vychodil, A. Chandra, T. Mikulasek, A. Prokes and V. Derbek, "UWB
time domain channel sounder," in Proc. Radioelektronika, Pardubice,
Apr. 2015, pp. 268–271.
[13] MSO/DPO70000 Series Digital and Mixed Signal Oscilloscopes
Datasheet, Available at http://www.tek.com/datasheet/mso-dpo70000-
series-digital-and-mixed-signal-oscilloscopes-datasheet. 2016.
[14] S. A. Billings , S. Y. Fakhouri, "Identification of non-linear systems
using correlation analysis and pseudorandom inputs," Int. J. Syst. Sci., vol.
11, no. 3, pp. 261–279, 1980.
[15] R. J. Polge and E. M. Mitchell, "Impulse Response Determination by
Cross Correlation," IEEE Trans. Aerosp. Electron. Syst., vol. AES-6, no.
1, pp. 91–97, Jan. 1970.
[16] S. W. Golomb et al., Shift register sequences. Aegean Park Press, 1982.
[17] F. Hlawatsch and G. Matz, Wireless communications over rapidly
timevarying channels. Academic Press, 2011.
[18] S Kotz, S Nadarajah, Extreme Value Distributions, World Scientific,
Washington, D.C, 2000.
PDF(-)
CDF(-)
