See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/336728101

CDL-based Channel Model for 5G MIMO Systems in Remote Rural Areas

Conference Paper · August 2019

DOI: 10.1109/ISWCS.2019.8877334

CITATIONS READS

4 857

8 authors, including:

Alexandre Pessoa Bruno Sokal

Universidade Federal do Ceará Universidade Federal do Ceará
13 PUBLICATIONS   20 CITATIONS    11 PUBLICATIONS   18 CITATIONS   

SEE PROFILE SEE PROFILE

Tarcisio F. Maciel André L. F. de Almeida

Universidade Federal do Ceará Universidade Federal do Ceará
132 PUBLICATIONS   766 CITATIONS    263 PUBLICATIONS   2,578 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Tensor Factorizations for MIMO Cooperative Systems View project

TIDE5G - Dynamic Time Division Duplex for 5G Systems View project

All content following this page was uploaded by Alexandre Pessoa on 04 February 2020.

The user has requested enhancement of the downloaded file.

CDL-based Channel Model for 5G MIMO Systems
in Remote Rural Areas
Alexandre M. Pessoa, Bruno Sokal, Carlos F. M. e Silva, Tarcisio F. Maciel, André L. F. de Almeida,
Diego A. Sousa, Yuri C. B. Silva and Fco. Rodrigo P. Cavalcanti
Emails: {alexandrematos, brunosokal, cfms, maciel, andre, yuri, diego, rodrigo}@gtel.ufc.br

Abstract—In the fifth generation (5G) mobile networks, it is for the digital inclusion and limits the democracy in accessing
expected that users experience high throughputs with an ultra to mobile services and information.
low-latency network, while a massive number of devices are To overcome the possible high costs of these networks in
connected to the network. However, in remote rural areas, there
is still a large number of people that do not have access to rural areas, some alternatives may be employed. For example,
Internet broadband. To overcome this issue, a possible strategy of the network or some nodes, using the software-defined
is to exploit the excellent propagation conditions of very high network (SDN) and network function virtualization (NFV)
frequency (VHF) and ultra high frequency (UHF) bands by concepts, in order to have, e.g., a cloud-radio access network
allowing a secondary spectrum reuse in the TV white space (C-RAN) [5], [6]; use new sources of renewable energy
(TVWS) channels. But for that, a reliable channel model is
required to perform valid coverage and capacity prediction and efficient solutions that switch-off the base station (BS)
studies. In this paper, a channel model is proposed that takes when there is no traffic; and exploit the excellent propagation
into account large and small scale fading effects, as well as the conditions of very high frequency (VHF) and ultra high fre-
particularities of such remote rural areas. The proposed model is quency (UHF) bands by enabling the spectrum sharing in TV
simple to implement and can be used to provide fast link-level and white space (TVWS) channels, so that secondary users reuse
system-level simulations. Numerical simulations are provided to
validate the model and a capacity evaluation is done considering the licensed spectrum from primary users, without causing
single input single output (SISO) and different multiple input harmful interference to the latter ones (namely, television (TV)
multiple output (MIMO) configurations. broadcasters) [7].
Index Terms—Channel modeling, rural remote areas, MIMO, In November 2017, a partnership project between Europe
VHF, UHF, TVWS. and Brazil, called Remote area Access Network for 5th
GEneration (5G-RANGE), was started, aiming to meet the
I. I NTRODUCTION Internet broadband demands in remote rural areas. The 5G-
It is well known that mobile users demand higher through- RANGE project adopts the TVWS strategy, which not only
puts and better quality of service (QoS) every year. A study significantly reduces the deployment and operational costs,
conducted by Qualcomm and Nokia [1] shows that since but also makes it possible for the network to cover a wide
2014 the mobile traffic is almost doubling every year. The area, due to the lower signal attenuation in the VHF and UHF
next generation of mobile networks, fifth generation (5G), bands. For this network design, the channel characterization
shall not only attend this demand for smartphones, but also is of utmost importance, since it allows operators to perform
bring to the society a new experience, where “everything” is valid coverage and capacity prediction studies.
connected. The 5G mobile networks can be classified into three The project specification considers a single cell with radius
categories, according to the International Telecommunication (coverage) up to 50 km and frequency ranges from 170 to
Union (ITU) [2]: the enhanced mobile broadband (eMBB) 400 MHz and from 450 to 700 MHz (VHF and UHF). The
which consists in services that demand a high bandwidth, bandwidth of each TVWS channel is 6 MHz in Brazil, while
such as videos in ultra-high definition (UHD); the ultra-reliable in Europe it is 8 MHz. More details can be found in the first
low latency (uRLL) which includes services that are sensitive 5G-RANGE report on application and requirements [8].
to latency and delay, such as remote surgery, self-driving In this paper, a clustered delay line (CDL)-based channel
cars; and the massive machine-type communications (mMTC), model for 5G multiple input multiple output (MIMO) systems
where a massive number of devices are connected, bringing the in remote rural areas is proposed, wherein the large and
concept of smart cities and Internet of things (IoT). However, small scale fading parameters are based on measurements
these new experiences are usually projected for urban areas. performed by Telstra and Ericsson [9] and in the rescaled
In this manner, one may think that the deployment and values of CDL model in 3rd Generation Partnership Project
operational costs [3] of 5G mobile networks in rural areas are (3GPP) standards [10], respectively. The model is simple to
not profitable (costs vs. incomes) and, therefore, do not attract be implemented and may be used in both link-level and simple
investments from operators [4]. It is fact that the user density system-level simulations. In order to validate the proposed
in urban areas can be very high when compared to rural areas. channel model, the coherence bandwidth and coherence time
But with the lack of investment the number of unconnected are computed numerically and compared with theoretical re-
people in rural areas may be large, which is a serious drawback sults. The numerical results also include a capacity evaluation
considering single input single output (SISO) and MIMO (2×2 argue that the path loss in (1) is simpler than the other models
and 4 × 4) configurations. and shows a good agreement with the measured data.
The rest of the paper is organized as follows: section II Using the measurements from [9], the authors of [17] have
briefly presents channel models in the literature for 5G that used the CDL-A from [10], which is for non-line-of-sight
take into account rural areas, as well as the measurement cam- (NLOS) case, to model the fast fading in link-level and system-
paign in [9]; in section III the proposed large and small scale level simulations. Therein, it was concluded that the RMS
fading models are presented, which characterize the channel delay spread does not change as function of distance and, due
for the 5G-RANGE project; channel analysis simulations are to the long distance and high effective antenna height, a small
done and discussed in section IV; and in section V the final angular spread is expected.
conclusions are drawn.
III. P ROPOSED C HANNEL M ODEL
II. RURAL C HANNEL M ODELS A. Large Scale Fading
As in [13], the path loss model in (1) is here adopted due
Rural areas are usually characterized by high effective BS
to its simplicity. However, the value of K, in this case, is
heights, small cities with few buildings, or farms, natural
computed for the measured data of the four scenarios in [9],
scatterers and low density of users [4]. In the literature
illustrated in fig. 1, covering only the distance of interest, i.e., 1
related to 5G mobile networks, Wireless World Initiative New
to 50 km. For this purpose, the unconstrained minimum mean
Radio (WINNER) II [11], Quasi Deterministic Radio channel
squared error (MMSE) method is used in this work, which is
Generator (QuaDRiGa) [12], and 3GPP TR 38.901 [10] can
given by
be cited as geometric-stochastic channel models (GSCMs) ( I
)
that account for rural macro (RMa) areas. In table I some 1X
2
min f (K) = e(di ) , (2)
specifications of these channels are shown. However, none of K I i=1
these models can be directly applied in 5G-RANGE project, where I = 191 is the number of measured points in fig. 1 and
due to coverage and/or carrier frequency fc limitations. e(di ) = PL(di , fc ) − y(di ) is the error between the measured
point y(di ), which is an average of thousands of samples to
TABLE I remove the fast fading fluctuations, and the proposed path loss
S PECIFICATIONS OF CHANNEL MODELS FOR RURAL AREAS .
model in (1). Thus, differentiating (2) with respect to K and
Channel model Max. distance (km) fc (GHz) equaling it to zero yields
WINNER II 10 2–6 I
1X

QuaDRiGa 10 2–6 K= y(di ) − FSPL(di , fc ) = 29.38 dB, (3)
3GPP TR 38.901 10 0.5–100 I i=1
which is the unique optimal solution of (2) since it is a convex
As previously mentioned, in [9] a measurement campaign function in K.
was performed by Telstra and Ericsson for remote rural areas
on third generation (3G) live cells (band 5: 850 MHz), with Measurements Outliers FSPL
distances up to 200 km. The path loss and the root-mean- 3GPP-RMa Hata-Davidson Proposed
square (RMS) delay spread (DS) were measured using con- 170
nected user equipments (UEs) over 50 cells in four different
scenarios, with 500 to 600 m as the relative height difference 150
Path loss (dB)

between the BS and the UE. Based on these measurements, a

path loss model is proposed in [13] for the remote rural areas
130
as
PL(d, fc ) = FSPL(d, fc ) + K, (1)
110
where FSPL(d, fc ) = 32.45+20 log10 (d)+20 log10 (fc ) is the
free space path loss (FSPL) at distance d in km for a carrier 90
1 8 15 22 29 36 43 50
frequency fc in MHz, and K is an offset in dB. The values of
Distance (km)
offset K and shadowing that the authors have recommended
are in table II. Fig. 1. Proposed path loss model and measurements from [9]. Due to the
It is important to mention that in [9], the authors considered low statistical relevance, the measured points plotted as diamonds (outliers)
were removed to derive the proposed path loss and shadowing. Moreover, the
as distance of interest the range between 45 and 115 km in 3GPP-RMa for NLOS [10] and the Hata-Davidson [16] path loss models are
order to predict the value of offset K in (1). Moreover, it is also presented, showing that both models do not fit well the measurements.
well known that in the literature there are many models that
could be applied to a large range of frequencies and distances, The extraction of the shadowing is done by subtracting the
such as Longley & Rice model [14], ITU-P-1546 [15], and measured data from the proposed path loss, i.e., SF(di ) =
Hata-Davidson model [16]. Nevertheless, the authors in [13] y(di ) − PL(di , fc ).
 But, since the distribution of shadowing is unknown, its denote the antenna gains of the antenna elements u and s in
histogram is plotted, as shown in Fig. 2, where the number
√  the UE and BS, respectively, considering the ray m within the
of bins is calculated following a simple rule kbin = 171 cluster n. Also, fBS,s,n,m is given by
(I = 171 is the number of measured points after remov-  
ing the outliers). As it can be seen, the tendency is to cos(ξs )
q
fBS,s,n,m = ZOD , φAOD )
Rs (θn,m
follow a normal distribution. In fact, it is what shows the n,m sin(ξs )
empirical probability density function (PDF) of the SF(di ), 
2πj T ZOD AOD

obtained from the kernel smoothing method [18], and the exp r (θn,m , φn,m )ds , (5)
2
λ
PDF of a normal distributed random variable N (µSF , σSF )
with µSF = avg{SF(d1 ), . . . , SF(dI )} = 0 dB and σSF = which considers only linear polarization (for details see [10],
std{SF(d1 ), . . . , SF(dI )} = 4.47 dB, where avg{·} and section 7.3.2, model 2). In (5), Rs is the radiation power
ZOD
std{·} denotes the average and standard deviation operators, pattern of the antenna element s, θn,m and φAOD
n,m are the zenith
respectively. angle of departure (ZOD) and azimuth angle of departure
(AOD) of ray m within the cluster n, respectively. ξs ∈
Histogram Emp. PDF N (0, 4.472 )
[0°, 90°] is the polarization angle of the antenna element s,
0.10 rT (θn,m
ZOD
, φAOD ZOD AOD
n,m ) represents the pair of angles (θn,m , φn,m ) in
the Cartesian coordinate system, and ds is a vector that points
from the reference point of the antenna array to the antenna
0.07
element s. fUE,u,n,m is also calculated using (5), but the
antenna element index s has to be replaced by u and the pair
PDF

ZOD
0.05 of angles (θn,m , φAOD ZOA AOA
n,m ) has to be replaced by (θn,m , φn,m ),
which are the zenith angle of arrival (ZOA) and azimuth angle
0.03 of arrival (AOA), respectively. Xn,m from (4), is the cross
polarization power ratio (XPR) matrix of dimension 2 × 2
of the ray m within the cluster n. The XPR is independently
0.00
−14 −10 −6 −2 2 6 10 14 generated for each ray m within each cluster as in [10] (section
Shadowing (dB) 7.7, step 3). dUE,n,m (t) is the Doppler coefficient at time t
between the UE and the ray m within the cluster n and it is
Fig. 2. Estimated shadowing from the measured data in fig. 1.
given by
As shown in Fig. 2, there is a good agreement between the 
2πjt T ZOA AOA

empirical and theoretical models. Finally, the values of the dUE,n,m (t) = exp r (θn,m , φn,m )vUE , (6)
λ
proposed offset K and shadowing standard deviation σSF are
summarized in table II. where vUE is the UE speed in three dimensional (3D) space
and λ is the wavelength. Pn /M and τn are the power and
TABLE II delay of each ray within the cluster n, respectively.
O FFSET AND SHADOWING VALUES FOR THE FSPL MODEL .

Offset (dB) Shadowing std. (dB) Range (km) C. Channel Generation

Ref. [13] 20 8 45–115 In order to generate the powers, angles and delays for the
Proposed 29.38 4.47 1–50 clusters and rays of the channel model in (4), a rescaled
version of the CDL-A model of [10] (section 7.7, table 7.7.1-
1) was used. The rescaling procedure is necessary because the
B. Small Scale Fading original delay and angular spreads are widely different from
The general channel impulse response that describes the fast the ones expected in a remote rural area.
fading for the proposed channel model for remote rural areas In this context, the values of table III, which are a rec-
is given by [10], [19] as ommendation from [17], and the procedure described in [12]
N X
M 
(sections 3.2 and 3.3) were considered to rescale the CDL-
X A model. Next, the rescaled model was used to generate
Hu,s (t, τ ) = fUE,u,n,m T Xn,m fBS,s,n,m
the channel impulse response in (4) following the procedure
n=1 m=1
r  described in [10] (section 7.7, steps 1–4).
Pn
dUE,n,m (t)δ(τ − τn ) , (4)
M IV. S IMULATION R ESULTS
where u is the antenna element index of the receiver, s is the
A. Metrics and Simulation Parameters
antenna element index of the transmitter, N is the number of
clusters between the BS and the UE and M is the number of In order to evaluate the channel model proposed in sec-
rays within each cluster. The variables fUE,u,n,m and fBS,s,n,m tion III, the metrics in the following paragraphs were analyzed.
 TABLE III transmit power and γ 2 is the noise power in a PRB. If the
D ELAY AND ANGULAR SPREADS FOR REMOTE AREAS . channels in (4) are spatially uncorrelated, then (10) simplifies
to
DS (ns) ASD (°) ASA (°) ZSD (°) ZSA (°)
rMIMO = min{NBS , NUE }rSISO . (11)
100 1 30 0.10 1
ASD: azimuth spread of departure, ASA: azimuth The previous metrics were analyzed using the simulation
spread of arrival, ZSD: zenith spread of departure, parameters presented in table IV.
ZSA: zenith spread of arrival.
TABLE IV
M AIN SIMULATION PARAMETERS .
1) The autocorrelation function (ACF) of the proposed
channel model, which is assumed to be a wide-sense stationary Parameter Value Unit Ref.
(WSS) random process [20], is given by Cell radius 50 km [8]
BS transmit power 49 dBm [8]
Ru,s (∆t, ∆f ) = Eff. channel bandwidth 23.40 MHz –
n o PRB bandwidth 180 kHz [19]
∗ Total number of PRBs 130 –
E H̃u,s (t, f )H̃u,s (t + ∆t, f + ∆f )
n o , (7) UE speed 3/120 km/h –
∗ (t, f ) Carrier frequency 700 MHz –
E H̃u,s (t, f )H̃u,s Path loss model FSPL + 29.38 dB –
Shadowing std. 4.47 dB –
∗
where E{·} and (·) are the expectation and complex con- Fast fading model CDL-A – –
jugate operators, respectively, and H̃u,s (t, f ) is the channel Antenna config. 1 × 1, 2 × 2 and 4 × 4 – [8]
Rad. pattern BS/UE isotropic – –
frequency response obtained using the non-discrete Fourier Antenna gain BS/UE 9/0 dBi –
transform (NDFT) [21] over (4). In this manner, (7) is used to Polarization BS/UE vertical – –
determine the coherence bandwidth and the coherence time of Noise figure 5 dB [8]
Noise power per PRB −116.45 dBm –
the proposed channel model. Such metrics have high impact Number of UEs 1 – –
on the system design and impose practical limitations to the Monte Carlo runs 5000 – –
time of symbol and the bandwidth of each resource block. It
is also desired to compare the coherence time and coherence
bandwidth with theoretical results, ensuring a confidence level B. Results
to the results provided by the proposed channel model. As- Fig. 3 shows the numerical evaluation of the ACF in (7) for
suming 50 % of correlation, the coherence time and coherence ∆f = 0 Hz and a UE speed vector of
bandwidth [20] are respectively given by  T
vUE = kvUE k cos(β) sin(β) 0 , (12)
arccos(0.5) arccos(0.5)
Tc = and Bc = , (8) where kvUE k = 120 km/h and β was uniformly generated in
2πSν 2πSτ
the range [0°, 360°] for each Monte Carlo run. This figure
where Sν and Sτ are the RMS Doppler and delay spreads. highlights the theoretical coherence time from (8) and the
2) To analyze the maximum achievable data rate and also empirical coherence time when the ACF is equal to 0.5. As it
the spatial correlation of the MIMO channels, the cumulative can be seen, such coherence times are very close presenting a
distribution function (CDF) of the throughput was calculated small relative error of 1.60 %.
considering SISO and MIMO (2 × 2 and 4 × 4) configurations, Fig. 4 shows the numerical evaluation of the ACF in (7)
with the UE placed at cell-edge. That metric was computed for ∆t = 0 s and the UE speed vector was given by (12), but
using the Shannon capacity formula for SISO and single user using kvUE k = 3 km/h. This figure highlights the theoretical
MIMO (SU-MIMO) channels1 , i.e., coherence bandwidth from (8) and the empirical coherence
N
X PRB
bandwidth when the ACF is equal to 0.5. The relative error
rSISO = B log2 (1 + SNRk ), and (9) between such results is about 2.50 %, which is slightly larger
k=1 (but still small) than the relative error of the coherence time.
N To finalize this section, fig. 5 shows the CDF of the
PRB
PBS Hk HH
X   
k throughput for three different antenna configurations with
rMIMO = B log2 det I + , (10)
NPRB NBS γ 2 the UE placed at cell-edge. In this figure, the ratios of the
k=1

where NPRB is the number of physical resource blocks (PRBs), throughput for MIMO 2 × 2 and 4 × 4 configurations over
B is the bandwidth of each PRB, INUE ×NUE is the identity SISO configuration are 1.52 and 2.22 in the 50th percentile,
UE ×NBS ∆ respectively. If uncorrelated MIMO channels are assumed,
matrix, HNk = {H̃u,s (0, fk ), ∀(u, s)} is the channel
such ratios can be computed using (11), which results in 2
matrix in PRB k, NBS and NUE are the number of antenna
and 4, respectively. Comparing these ratios, it is noticeable
elements (AEs) at the BS and UE, respectively, PBS is the BS
that the proposed channel model has a considerable spatial
1 Herein, the maximum achievable data rate is a synonymous of the digital correlation, mainly in the MIMO 4 × 4 case, which presented
bandwidth capacity, obtained by the Shannon capacity formula. a gain of only 2.22 over the SISO case. This is explained due
 1.00 proposed model is simple to implement and useful to perform
Coherence time Tc fast link-level and system-level simulations.
0.63 Theoretical 3.12 ms To validate the channel model, the coherence time and
(3.07, 0.50) Empirical 3.07 ms coherence bandwidth were computed numerically and com-
pared with theoretical results, showing only a small difference
ACF

0.25
between them. Furthermore, the throughput at cell-edge for
SISO and MIMO configurations (2×2 and 4×4) was computed
−0.12
numerically using the proposed model. In this case, the gains
of MIMO over SISO are small, showing a considerable spatial
−0.50
0 3 6 9 12 15 18 21
correlation of the MIMO channels.
Time shift (ms) As a future work, it is foreseen the need to analyze the
eigenvalues of the channel matrix for single and dual polarized
Fig. 3. ACF in (7) as a function of the time shift with ∆f = 0 Hz. antennas, aiming to exploit the capability of the channel to
transmit multiple data streams. The use of dual polarized
1.00 antennas is relevant in VHF and UHF bands because they
Coherence bandwidth Bc would allow the physical dimension reduction of the antennas.
0.63 Theoretical 1.59 MHz
(1.63, 0.50) Empirical 1.63 MHz ACKNOWLEDGMENT
This work has been carried out within the 5G-RANGE project and received
ACF

0.25 funding from the European Union Horizon 2020 Programme (H2020/2017-
2019) under the grant agreement no. 777137 and from the Ministry of Science,
Technology and Innovation (MCTIC) of Brazil through Rede Nacional de
−0.12 Ensino e Pesquisa (RNP) under the 4th EU-BR Coordinated Call Information
and Communication Technologies.
This study was also financed in part by the Coordenação de Aperfeiçoa-
−0.50 mento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001.
0 5 10 15 20 25 30 35
Frequency shift (MHz) R EFERENCES
[1] Qualcomm and Nokia, “Making 5G a reality: Addressing the strong
Fig. 4. ACF in (7) as function of the frequency shift with ∆t = 0 s. mobile broadband demand in 2019 & beyond,” White paper, Sep. 2017.
[2] ITU-R, “Minimum requirements related to technical performance for
IMT-2020 radio interface(s),” Report ITU-R M.2410-0, Nov. 2017.
to small angular spreads in the zenith (ZSD and ZSA), which [3] C. F. M. e Silva and F. R. P. Cavalcanti, “Techno-economic evaluation
for TV white spaces,” in TV White Space Communications and Net-
make most of the multipath components widely correlated in works, R. Stewart, D. Crawford, and A. Stirling, Eds., ser. Electronic
the angular space. and Optical Materials. Woodhead Publishing, 2018, ISBN: 978-0-08-
100611-5.
[4] L. Chiaraviglio et al., “5G in rural and low-income areas: Are we
SISO MIMO 2 × 2 MIMO 4 × 4 ready?” In ITU Kaleidoscope: ICTs for a Sustainable World (ITU WT),
1.00 IEEE, Nov. 2016. DOI: 10.1109/ITU-WT.2016.7805720.
[5] P. Rost et al., “Cloud technologies for flexible 5G radio access net-
works,” IEEE Communications Magazine, vol. 52, no. 5, pp. 68–76,
0.75 May 2014. DOI: 10.1109/MCOM.2014.6898939.
[6] P. Demestichas et al., “5G on the horizon: Key challenges for the radio-
access network,” IEEE Vehicular Technology Magazine, vol. 8, no. 3,
CDF

0.50 Percentiles/Throughput (Mbps) pp. 47–53, Sep. 2013. DOI: 10.1109/MVT.2013.2269187.

[7] H. Bogucka, M. Parzy, P. Marques, J. W. Mwangoka, and T. Forde,
SISO MIMO 2 × 2 MIMO 4 × 4
“Secondary spectrum trading in TV white spaces,” IEEE Communica-
0.25
5th 6.14 10.79 19.06 tions Magazine, vol. 50, no. 11, pp. 121–129, Nov. 2012. DOI: 10.1109/
50th 23.31 35.44 51.80 MCOM.2012.6353691.
95th 55.16 74.39 96.28 [8] 5G-RANGE, “Application and requirements report,” Deliverable 2.1,
Apr. 2018. [Online]. Available: http://5g-range.eu/wp-content/uploads/
0.00
0 22 44 66 88 110 132 154 2018/04/5G- Range_D2.1_Application_Requirement_Report_v1.pdf
Throughput (Mbps) (visited on 05/2018).
[9] Telstra and Ericsson, “Measurements of extreme rural scenarios,” R1-
Fig. 5. CDF of the throughput for SISO, MIMO 2 × 2 and MIMO 4 × 4 166599, Aug. 2016, Draft.
with the UE placed at cell-edge. The outliers were removed following the [10] 3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,”
procedure described in [22]. TR 38.901, Mar. 2017, Release 14.
[11] P. Kyösti et al., “WINNER II channel models,” Deliverable 1.1.2, Sep.
2007, V1.2.
[12] S. Jaeckel, L. Raschkowski, L. Thiele, F. Burkhardt, and E. Eberlein,
V. C ONCLUSION “Quadriga - quasi determinisc radio channel generator, user manual and
In this paper, a CDL-based channel model was proposed documentation,” Technical Report, 2017, v2.0.0.
[13] Telstra and Ericsson, “Path loss model for extreme rural scenario,” R1-
for MIMO systems (up to 4 × 4) in remote rural areas with 167452, Aug. 2016.
a maximum distance of 50 km. The model was derived by [14] P. L. Rice, A. G. Longley, K. A. Norton, and A. P. Barsis, “Transmission
combining the measured data from [9], which led to the large loss predictions for tropospheric communication circuits,” Technical
Note 101, May 1965.
scale fading characterization, with the CDL-A from [10]. The
[15] ITU-R, “Method for point-to-area predictions for terrestrial services in
the frequency range 30 MHz to 3000 MHz,” Recomendation ITU-R
P.1546-5, Sep. 2013.
[16] L. Gavrilovska, P. Latkoski, V. Atanasovski, R. Prasad, A. Mihovska,
O. Fratu, and P. Lazaridis, “Radio spectrum: Evaluation approaches,
coexistence numbers and monitoring,” Computer Networks, vol. 121,
pp. 1–12, Jul. 2017. DOI: 10.1016/j.comnet.2017.04.017.
[17] Telstra and Ericsson, “Channel model for extreme rural scenario,” R1-
167451, Aug. 2016.
[18] A. W. Bowman, “An alternative method of cross-validation for the
smoothing of density estimates,” Biometrika, vol. 71, no. 2, pp. 353–360,
Aug. 1984. DOI: 10.2307/2336252.
[19] 3GPP, “Public safety broadband high power user equipment (UE) for
band 14,” TR 36.837, Dec. 2012, Release 11.
[20] A. F. Molisch and M. Steinbauer, “Condensed parameters for char-
acterizing wideband mobile radio channels,” International Journal of
Wireless Information Networks, vol. 6, no. 3, pp. 133–154, Jul. 1999.
DOI : 10.1023/A:1018895720076.
[21] D. Potts, G. Steidl, and M. Tasche, “Fast Fourier transforms for nonequi-
spaced data: A tutorial,” in Modern Sampling Theory, Mathematics and
Applications, J. J. Benedetto and P. J. Ferreira, Eds. Springer, 2001,
pp. 247–270. DOI: 10.1007/978-1-4612-0143-4.
[22] R. Mcgill, J. W. Tukey, and W. A. Larsen, “Variations of box plots,”
The American Statistician, vol. 32, no. 1, pp. 12–16, 1978. DOI: 10 .
2307/2683468.

View publication stats