This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Date of publication xxxx 00, 2019, date of current version July 19, 2019.
Digital Object Identifier 10.1109/ACCESS.2019.DOI

Channel Modeling for Satellite

Communication Channels at Q-Band in
High Latitude
LU BAI1 , CHENG-XIANG WANG2,3,4 (Fellow, IEEE), GEORGE GOUSSETIS4 (Senior
Member, IEEE), SHANGBIN WU5 , QIUMING ZHU6 (Member, IEEE), WENQI ZHOU7 , EL-HADI
M. AGGOUNE8 (Senior Member, IEEE)
1
Shandong Provincial Key Lab of Wireless Communication Technologies, School of Information Science and Engineering, Shandong Unviersity, Qingdao,
266237, Shandong, China (e-mail: [email protected])
2
National Mobile Communications Research Laboratory, School of Information Science and Engineering, Southeast University, Nanjing, 210096, China (e-mail:
[email protected])
3
Purple Mountain Laboratories, Nanjing, 211111, China
4
Institute of Sensors, Signals and Systems, School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, EH14 4AS, U.K. (e-mail:
[email protected])
5
Samsung R&D Institute UK, Staines-upon-Thames, TW18 4QE, U.K. (e-mail: [email protected])
6
College of Electronic and Information Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing, 211106, Jiangsu, China (e-mail:
[email protected])
7
Shandong Huahan Electronics Co., Ltd., No. 1768 Xinluo Street, Jinan High-Tech Zone, Jinan, Shandong, China (e-mail: [email protected])
8
Sensor Networks and Cellular Systems Research Center, University of Tabuk, Tabuk 47315/4031, Saudi Arabia (e-mail: [email protected])
Corresponding author: Cheng-Xiang Wang (e-mail: [email protected]).
The authors acknowledge the support of this work by the National Key R&D Program of China under Grant 2018YFB1801101, the
National Natural Science Foundation of China (NSFC) under Grant 61960206006, the Fundamental Research Funds for the Central
Universities under Grant 2242019R30001, the Taishan Scholar Program of Shandong Province, and the EU H2020 RISE TESTBED
project under Grant 734325.

ABSTRACT This paper proposes a three-dimensional (3D) channel model for satellite communications
at Q-band in a high latitude, including the path loss, shadowing, and small-scale fading. The shadowing
effect is modelled by a Markov chain. The three states in the Markov chain are separated by the threshold
of the received power level for the link budget and system optimization. The probability density function
(PDF) of shadowing amplitude is modelled by a mixture of two Gaussian distributions with parameters
obtained by the expectation-maximum (EM) algorithm. The small-scale fading is represented by a 3D
geometry-based stochastic model (GBSM) where scatterers are located on the spherical surface of a
hemisphere. The movement of the receiver and the Rician factor influenced by environment scattering are
considered. Statistical properties including the local temporal autocorrelation function (ACF) and Wigner-
Ville spectrum are derived. The satellite communication channel measurement at Q-band is conducted on
the campus of Heriot-Watt University (HWU) in Edinburgh, UK. The parameters of our proposed channel
model are estimated by the measurement data. Numerical and simulation results demonstrate that our
proposed channel model has the ability to reproduce main statistical properties which are also consistent
well with the corresponding theoretical and measurement results.

INDEX TERMS Satellite communications, Q-band, Markov chain, GBSM.

I. INTRODUCTION base stations and cannot achieve the worldwide coverage and
overcome the problems of special occasions, e.g., the mar-
The fifth generation (5G) wireless communication networks
itime communication scenario [5], [6] and natural disasters.
will come to the stage of commercial deployment in 2020.
With the acceleration of beyond 5G (B5G) wireless commu-
Its system performance, such as data rate, latency, energy
nication process, satellite communication technologies and
efficiency, and cost efficiency, has a great improvement com-
unmanned aerial vehicle (UAV) communication technologies
pared to the fourth generation (4G) networks [1]– [4]. How-
have attracted wide attention for their reduced vulnerability
ever, 5G wireless communication networks are still based on

VOLUME 4, 2019 1

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

of natural disasters and physical attacks [7]– [10]. As a tech- hidden Markov models and dynamic higher order Markov
nology that has been proved and deployed for a long time, state models provide better balance between complexity
satellite communications stand out for its capacious service and accuracy. The channel models in [24]– [27] proposed
coverage capabilities. In the past years, satellite communi- Markov-chain-based channel models for satellite systems at
cations have been widely used for a variety of applications L-band, Ku-band, Ku-band, and X-band, respectively. While
such as navigation, earth observations, and broadcasting. the traditional satellite communications service at L-band (1-
There are higher requirements for satellite communication 2 GHz), C-band (4-8 GHz), X-band (8-12 GHz), and Ku-
systems, such as the higher capacity, improved quality of band (12-18 GHz), the increasing need for higher bandwidth
service (QoS), and ubiquitous connectivity, when there is for reducing costs drives the exploitation of millimeter wave
considerable interest in the application of next generation ter- (mmW) bands. Q-band (33-50 GHz) is used for the feeder
restrial wireless communication systems. For the design and link of satellite communication systems in order to free
performance evaluation of satellite communication systems, lower band spectrum to revenue user links and reduce cost
it is necessary to get a general, accurate, and low-complexity of the terrestrial segment [28]–[30]. Additionally, compared
model to depict the underlying realistic propagation channel with 5G wireless communications at mmW bands, satellite
characteristics [11]. communications at Q-band have better performance in global
Most researchers mainly focus on using different dis- seamless coverage and reliability at disaster scenarios. Al-
tributions to represent the shadowing and small-scale fad- though none of [22]– [27] has considered the satellite prop-
ing and how the shadowing affects the line-of-sight (LoS) agation at Q-band in high latitude, variations at frequency
and scattered components. The authors in [12]– [16] used band have a major effect on space-to-earth channel. They
Rayleigh distribution and lognormal distribution to model used conventional distributions in traditional satellite channel
small-scale fading and shadowing, respectively. In [12] and models to generate parameters, such as Rice distribution, log-
[13], the authors associated the above two distributions as ad- normal distribution, Nakagami-Rice distribution, Loo distri-
ditive. The envelope of LoS component faded by shadowing bution, and lognormal-Rayleigh distribution. However, none
and multipath scattered component are assumed independent of them focued on the improvement of parameter estimation
with each other. The phases are also independent with each algorithm. The current small-scale channel models based on
other. The major difference between [12] and [13] is that the Rayleigh or Nakagami family are insufficient to take the
LoS component in [13] is Doppler shifted, so the model in propagation mechanism and the correlations among ampli-
[13] has a higher degree of freedom. The authors in [14]– tude, angle, and Doppler frequency into account. The ratio of
[16] took the association of the two distributions as mul- LoS component to non-line-of-sight (NLoS) component was
tiplicative. The envelope of multipath scattered component also not considered.
is effected by shadowing and LoS component is ignored in In this paper, we propose a three-dimensional (3D) channel
[14]– [16]. The channel models in [17]– [21] used Nakagami- model for satellite communications at Q-band in a high lati-
m/Nakagami-q distribution instead of Rayleigh distribution tude. The main contributions of this paper are summarized
to model small-scale fading because they represent more as follows:
diverse fading conditions compared to Rayleigh distribution.
1) The mixture of Gaussian (MoG) distribution is firstly
As Rayleigh distribution is a special case of Nakagami family
used to depict the satellite shadowing channel model.
distributions, the channel models in [17]– [20] are viewed
Expectation-maximum (EM) algorithm is used to es-
as generalizations of the multiplicative Rayleigh-lognormal
timate parameters of probability density functions
models in [14]– [16].
(PDFs) of amplitude of Markov chain process states.
Aforementioned channel models composed of two dis-
2) The received power level of the channel is modeled via
tributions are oversimplified and not flexible to depict the
a Markov process.
channel characteristics under kinds of weather conditions
3) 3D geometry-based channel model (GBSM) is firstly
and environments separately and clearly. For classifying
used to model satellite small-scale fading. The corre-
different channel states and describing the received power
lations among channel parameters (distances of mul-
level changes over time, the Markov chain process has been
tipaths, azimuth angle of arrival (AAoA), elevation
widely used to model satellite communication channels, such
angle of arrival (EAoA), azimuth angle of depar-
as [22]– [27]. It defines specified number of states with
ture (AAoD), elevation angle of departure (EAoD),
specified probability depending only on the previous state.
Doppler frequency, and phase) are considered. The
The authors in [22] modeled the satellite propagation fading
movements of receiver (Rx) and cluster are also taken
channels by hidden Markov models of 10 states. The number
into account.
of states was too large to explain the physical mechanism of
4) The measurement data of satellite-to-earth propagation
these states. Meanwhile, the International Telecommunica-
channel at Q-band in high latitude has been obtained
tion Union (ITU) recommended it is better to use a three-
and analyzed to model.
state model to depict the satellite signal propagation [32].
A dynamic higher order Markov state model for multiple The rest of the paper is organized as follows. The 3D chan-
satellite broadcasting systems was proposed in [23]. The nel model for satellite communications at Q-band in high
2 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

latitude is shown in Section II. In Section III, we describe fading which can be detected by the satellite communication
the measurement setup of satellite communications at Q- receiver system. The good state corresponds to a small signal
band. The parameter estimation via measurement data is also lever oscillation with low scattering contributions.
given in this section. The statistical properties of the proposed
reference model are derived in Section IV. Section V presents The first order Markov model can be represented by a
the corresponding simulation model. Simulation results and stationary state probability vector (SSPV) which contains
analysis are also given in Section V. At last, conclusions are the limited probabilities of three states and a state transi-
drawn in Section VI. tion probability matrix (SPTM) which contains all transition
probabilities between any two of three states. The SSPV is
II. 3D SATELLITE CHANNEL MODEL presented as
The proposed 3D channel model for satellite communica- P = [PG , PM , PB ]T (5)
tions at Q-band in high latitude can be decomposed into three
Ni
parts, and can be presented as , i ∈ G, M, B
Pi = lim (6)
N →∞ N
h(t) = h1 (t) × h2 (t) × h3 (t) (1) where PG , PM , and PB denote the limited probabilities of
where h1 (t) and h2 (t) denote the free space path loss (FSPL), good state, moderate state, and bad state, respectively, [·]T
and the shadow fading, respectively, h2 (t) is a Markov chain denotes the transpose operator, Ni is the number of the i-th
process based on measurement will be explained in Section state and N is the total number of three states.
III, h3 (t) denotes the small-scale fading, which is caused
by scattering multipaths and the movement of Rx. The def- The transitions between any two states are based on SPTM.
initions of key parameters for the proposed 3D theoretical The state transition diagram of the Markov chain is shown in
satellite channel model are given in Table 1. Fig. 2. The SPTM is presented as
 
PGG PGM PGB
A. FSPL Pt =  PMG PMM PMB  (7)
FSPL is mainly determined by the distance and frequency. It PBG PBM PBB
predicts how the area mean varies with the distance between
Nij
the satellite and Rx [33]. FSPL can be expressed as Pij = lim , i ∈ G, M, B, j ∈ G, M, B (8)
q N →∞ Ni
4πfc kD LoS (t)k
h1 (t) = 10 10c (2) where Pij denotes the transition probability from the i-th
state to the j-th state, Nij is the number of the transition from
where fc is carrier frequency, c is speed of light, the initial the i-th state to the j-th state.
distance DLoS between the satellite and Rx can be calculated
by
In the measurement data, the number of states is suffi-
cosϕLoS (t) cosθLoS (t)
 
ciently large but finite. Therefore,
DLoS (t) = kDLoS k  sinϕLoS (t) cosθLoS (t)  (3) Ni
sinθLoS (t) Pi = , i ∈ G, M, B (9)
N
with Nij
q Pij = , i ∈ G, M, B, j ∈ G, M, B. (10)
kDLoS k = (RE2 sin2 θ LoS + L2 + 2LR ) − R sin θ LoS Ni
E E
(4)
where kDLoS k is the initial distance between satellite and
Rx [31], RE is the radius of the earth (in the range of
6378.1–6356.8 km, depending on the latitude), L and θLoS C. SMALL-SCALE FADING
respectively present the altitude of the satellite and elevation
angle as shown in Fig. 1. In the theoretical small-scale model shown in Fig.1, the Rx
is located at the origin point of the 3D reference coordinate
B. SHADOW FADING system. The cluster is located at the spherical surface of
The long-term variations of the amplitudes are modeled as the hemisphere whose center is also located at the original
a chain of distinct states using a first order Markov-chain point of the 3D reference coordinate system. The vector of
process. The differentiated received signal amplitude level is movement of the cluster and Rx, is defined as vc and v.
related to the underlying signal propagation condition, which The maximum Doppler frequency and carrier wavelength are
is represented as a state of the Markov model. The bad state denoted as fmax and λ. Also, let S denotes the total number
corresponds to the situation that the received signal cannot be of rays within a cluster, which obeys a Poisson distribution
detected because of the link budget. The moderate state oc- in the millimeter wave frequency band [34]. The channel
curs in bad weather conditions, and corresponds to the deep impulse response of small-scale fading can be presented as
VOLUME 4, 2019 3

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

TABLE 1: Summary of key parameter definitions.

Symbol Definition
h(t) Channel impulse response for satellite communication at Q-band in high latitude
h1 (t) FSPL
L Altitude of the satellite
RE Radius of the earth
fc /c Carrier frequency/ speed of light
h2 (t) Shadowing and modification of path loss for satellite communication
P Stationary state probability vector (SSPV)
Pt State transition probability matrix (SPTM)
α(t) Amplitude of h2 (t)
h3 (t) Small-scale fading for satellite communication
LoS
fmax Maximum Doppler frequency of the LoS component
NLoS
fmax Maximum Doppler frequency of the non-line of sight (NLoS) component
hLoS
3 (t) LoS path of small-scale fading
R(t) Radius of the hemisphere
K/S Rician factor and number of rays within one cluster
f LoS (t) Doppler frequency of LoS path
φLoS (t) Phase of LoS path
DLoS (t) Vector from Rx to satellite
ϕLoS (t) AAoA of LoS path
θLoS (t) EAoA of LoS path
hNLoS
3 (t) NLoS path of small-scale fading
f NLoS (t) Doppler frequency of NLoS path
φNLoS
s (t) Phase of NLoS path via the s-th ray
DsR (t) Vector from cluster to Rx via the s-th ray
ϕR s (t) AAoA of NLoS path from cluster to Rx via the s-th ray
θsR (t) EAoA of NLoS path from cluster to Rx via the s-th ray
DsT (t) Vector from satellite to cluster via the s-th ray
ϕT s (t) AAoD of NLoS path from satellite to cluster via the s-th ray
θsT (t) EAoD of NLoS path from satellite to cluster via the s-th ray
vc Velocity vector of cluster
ϕvc AAoD of movement of cluster
θcv EAoD of movement of cluster
v Velocity vector of Rx
ϕv AAoD of movement of Rx
θv EAoD of movement of Rx

The received phase of LoS path can be expressed as

h3 (t) = hLoS (t) + hNLoS (t) (11) 2π
r3 3 kDLoS (t)k.
φLoS (t) = φ0 + (15)
λ
K j (2π t f LoS (τ )dτ +φLoS (t))
R
= e 0
The Doppler frequency of NLoS path can be calculated as
K +1
< DsR (t), (v − vc ) >
| {z }
LoS fsNLoS (t) = fmax
NLoS
r S
! kDsR (t)kk(v − vc )k
1 1 X j (R t fsNLoS (τ )dτ +φNLoS (t)) (16)
+ lim √ e 0 s . k(v − vc )k < DsR (t), (v − vc ) >
K + 1 S→∞ S s=1 =
| {z } λ kDsR (t)kk(v − vc )k
NLoS
(12) where
kDsR (t)k = R(t) (17)
The Doppler frequency of LoS path is expressed as
cosϕR R
 
s (t) cosθs (t)
DsR (t) = kDsR (t)k  sinϕR R
s (t) cosθs (t)
 (18)
< DLoS (t), v > R
sinθs (t)
f LoS (t) = fmax
LoS
(13)
kDLoS (t)kkvk
cosϕ cosθv
v
 
T 
cosϕLoS (t) cosθLoS (t) cosϕv cosθv
 
v v = kvk  sinϕv cosθv  (19)
=  sinϕLoS (t) cosθLoS (t)   sinϕv cosθv  . sinθv
λ
sinθLoS (t) sinθv
(14)
4 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

D LOS
DcT

Clusterc

q
vc

LO
DcR

q c LOS
R
j
RE jc R
v

q
v
j
v

FIGURE 1: The 3D channel model for satellite communication channel at Q-band.

cosϕvc cosθcv

PGG vc = kvc k  sinϕvc cosθcv  . (20)
sinθcv
The phase of received NLoS path via the s-th ray can be
*RRG6WDWH expressed as
2π
φNLoS
s (t) = φ0 + (kDsT (t)k + kDsR (t)k) (21)
λ
PMG where
PBG
DsT (t) = DLoS (t) + DsR (t) (22)
PGM PGB  T T

cosϕs (t) cosθs (t)
DsT (t) = kDsT (t)k  sinϕT T
s (t) cosθs (t)
 (23)
T
sinθs (t)
PMB
cosϕLoS (t) cosθLoS (t)
 
0RGHUDWH
%DG6WDWH
6WDWH DLoS (t) = kDLoS (t)k  sinϕLoS (t) cosθLoS (t)  . (24)
sinθLoS (t)
PBM
PMM PBB
III. MEASUREMENT SETUP AND PARAMETER
FIGURE 2: The state transition diagram of the Markov chain. ESTIMATION
VOLUME 4, 2019 5

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

loop tracking systems is used for the antenna to track the

inclined orbit of the Alphasat. The first downconversion to
a conventional intermediate frequency (IF) of 70 MHz hap-
pends within the temperature controlled the radio frequency
(RF) installed directly behind the antenna. Independent tem-
perature controls of the low noise amplifier (LNA) is used
to maintain a temperature stability of +/ − 0.01◦ C. The
LNA has a noise figure of 2.7 dB, which results in a reduced
dynamic range of 35 dB. From the RF box which is installed
behind the antenna, the signal is routed to a secondary
temperature controlled IF box where the final downconver-
sion stages take place to the 5 MHz IF. The temperature
(a) The Alphasat beacon Rx on the top of stability of the IF box is maintained to within +/ − 0.25◦ C.
a building at HWU. The estimated system temperature of the AlphaSat beacon
receiver based on the measured component performance is
calculated as 908 K. A common ultra–stable 10 MHz refer-
ence oscillator drives all local oscillators utilized in the three-
stage downconversion process. The system parameters of the
AlphaSat beacon receiver system are shown in Table 3.
The 5 MHz IF of the receiver requires modifications in
order to maintain similar performance from the frequency
estimation routine employed in all NASA GRC-based beacon
receivers [36], [37]. The 5 MHz IF signal is sampled by
a 12-bit National Instruments 5124 data acquisition (DAQ)
card at a sampling frequency of 11.111 MHz. For the 10
(b) The relative location of the receiver. Hz data measurement rate, 220 samples are collected for
FIGURE 3: The photographs of measurement campaign. a final fs/N resolution of 10.6 Hz. In order to perform the
frequency estimation routine and record the signal power
TABLE 2: Satellite location and specifications. every 0.1 s, it was required to digitally filter and decimate
the sampled data by a factor of 32 to reduce processing time.
Location Parameters Specification A 50 kHz 10-th order Type 2 Chebyshev digital bandpass
Latitude/ Longitude 55.91◦ N/ 3.32◦ W
Altitude 130 m filter is employed prior to decimation. The frequency track
Elevation Angle 21.3◦ (nom.) +/- 1.2◦ (track) of the signal is maintained to center the Chebyshev filter at
Azimuth Angle 147.2◦ (nom.)+/- 0.5◦ (track) the current tracked frequency. This also allows for tracking
of deep fades by reducing the tracking bandwidth window
around the nominal beacon IF frequency when the signal
A. MEASUREMENT SETUP strength is reduced to within 10 dB of the noise floor level.
Alphasat (also referred to as Inmarsat-4A F4) is a geosta-
tionary orbit satellite which is located at 25.0◦ E. Since May B. MEASUREMENT DATA
1st , 2016, the Alphasat beacon Rx at Q-band (39.402 GHz) We take the measurement data from June 2016 to May 2017.
shown in Fig. 3-(a), has been installed at the roof of the After deleting the invalid data, the time series of received sig-
Earl Mountbatten Building on the campus of Heriot-Watt nal power is shown in Fig. 4. The blue line shows the received
University (HWU) in Edinburgh, UK. The azimuth angle and signal in 1Hz sample rate. For decreasing the complexity of
elevation angle of observation is approximately 147.2◦ and data pre-processing, we get the received signal in 1/60Hz
21.3◦ , respectively. The measurement campaign is a joint sample rate marked by red dash line. To a certain degree, the
effort between HWU and National Aeronautics and Space fluctuation of the red dash line can show that of the blue line.
Administration (NASA) Glenn Research Center (GRC) to The dark line and green dash line show the signal mean power
characterize the satellite channel attenuation at the Q-band. and the signal power after FSPL and losses of the Alphasat
The location of the Alphasat beacon terminal is shown in beacon terminal system, respectively. The light blue dash line
Fig. 3-(b). The location information and specifications for the is the minimum detection of signal power. The PDF of the
installation site are summarized in Table 2. received signal power is shown in Fig. 5.
The basis design of the AlphaSat beacon receiver at Q-
band is similar to the one installed and in operation since C. STATE DURATION AND THRESHOLD
April 2014 at the Politecnico di Milano (POLIMI) described According to the performance of satellite system and link
in [35]. It is a 0.6 m Q-band Cassegrain reflector with equiv- budget [38], we set the threshold to distinguish the good
alent antenna beamwidths of 0.9◦ . The independent open- state and the moderate state (TGM) by 3 dB, 4 dB and 5 dB
6 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

TABLE 3: The parameters of AlphaSat beacon receiver.

Receiver Parameters Performance Specification
Antenna Gain 45.6 dBi
EIRP (Ptx +Gtx -Losses) 59.5 dBm
Lf r (Rx antenna feed losses) 0.2 dB
Lr (Rx antenna depointing losses) 1 dB
IF Amplifier Gain 20 dB (29 dB meas.)
Low-Noise Amplifier Gain 35 dB (38 dB meas.)
Rx C/N (dB,Carrier to Noise ratio) 45.76 dB
Detection Threshold (The minimum C/No for signal detection) 10 dB
Dynamic Range (Demonstrated) 35.76 dB
Q-band Mixer Loss 9 dB
1st Stage Bandpass Filter (20.2 GHz) Loss 1 dB
IF Bandpass Filter (70 MHz) Loss 2.7dB (1.5dB meas.)
IF Lowpass Filter Loss 1.8dB (1.0dB meas.)
Measurement Sampling Rate 1Hz

Signal power (2016.06-2017.05)

-60
-61.844
-65 15

-70
SD10_TGM3
SD15_TGM3
-75 SD20_TGM3
-61.5742
-80 10 SD25_TGM3
Power, P (dBm)

-65.0399 SD30_TGM3
-85

PDF
-90

-95 -97.6122 5

-100
Received power_1 Hz
-105 Received power_1/60 Hz
Mean power
-110 Detection threshold
0
0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78
-115 After FSPL and losses of system The signal amplitude of good state, , (V)
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500
Time, T (h) (a) Good state.
8
FIGURE 4: The time series of the received signal power. SD10_TGM3
7
SD15_TGM3
6 SD20_TGM3
SD25_TGM3
5
1 SD30_TGM3
PDF

4
0.9
3
0.8
2
0.7
1
0.6
PDF

0
0.5 0.3 0.4 0.5 0.6 0.7 0.8
The signal amplitude of moderate state, , (V)
0.4

0.3
(b) Moderate state.
22
0.2
20 SD10_TGM3
0.1 SD15_TGM3
18
SD20_TGM3
0 16
-70 -69 -68 -67 -66 -65 -64 -63 -62 SD25_TGM3
14
Signal power, P (dBm) SD30_TGM3
PDF

12

FIGURE 5: The PDF of the received signal power. 10

2
lower than signal maximum (-61.844 dB) and the threshold 0
0 0.05 0.1 0.15 0.2 0.25
between moderate state and bad state (TMB) as 10 dB lower The signal amplitude of bad state, , (V)
than TGM. For the resolution and accuracy, we set up the (c) Bad state.
state duration (SD) as 10 min, 15 min, 20 min, 25 min, and 30
min. The PDFs of signal amplitudes of good state, moderate FIGURE 6: The PDFs of signal amplitudes of good state,
state and bad state with different SDs and TMBs are shown moderate state and bad state (SD=10, 15, 20, 25, 30 min;
as follows. TGM=3 dB).
The comparison among Fig. 6, Fig. 7, and Fig. 8 indicates
VOLUME 4, 2019 7

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

14 14

SD10_TGM4 SD10_TGM5
12 12
SD15_TGM4 SD15_TGM5
SD20_TGM4 SD20_TGM5
10 10
SD25_TGM4 SD25_TGM5
8
SD30_TGM4 8
SD30_TGM5
PDF

PDF
6 6

4 4

2 2

0 0
0.55 0.6 0.65 0.7 0.75 0.5 0.55 0.6 0.65 0.7 0.75
The signal amplitude of good state, , (V) The signal amplitude of good state, , (V)

(a) Good state. (a) Good state.

6 6

SD10_TGM4 SD10_TGM5
5 SD15_TGM4 5 SD15_TGM5
SD20_TGM4 SD20_TGM5
4 SD25_TGM4 4 SD25_TGM5
SD30_TGM4 SD30_TGM5
PDF

PDF
3 3

2 2

1 1

0 0
0.2 0.3 0.4 0.5 0.6 0.7 0.2 0.3 0.4 0.5 0.6 0.7
The signal amplitude of moderate state, , (V) The signal amplitude of moderate state, , (V)

(b) Moderate state. (b) Moderate state.

22

20 SD10_TGM4 SD10_TGM5
20
18 SD15_TGM4 SD15_TGM5
16
SD20_TGM4 SD20_TGM5
SD25_TGM4 15 SD25_TGM5
14
SD30_TGM4 SD30_TGM5
PDF

PDF

12

10 10
8

6
5
4

0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
The signal amplitude of bad state, , (V) The signal amplitude of bad state, , (V)

(c) Bad state. (c) Bad state.

FIGURE 7: The PDFs of signal amplitudes of good state, FIGURE 8: The PDFs of signal amplitudes of good state,
moderate state and bad state (SD=10, 15, 20, 25, 30 min; moderate state and bad state (SD=10, 15, 20, 25, 30 min;
TGM=4 dB). TGM=5 dB).

the PDF of signal amplitudes for moderate state is smoother

The EM algorithm for parameter estimation of PDFs of
and more stable when the TGM is set up as 3 dB and the SD
signal amplitudes of good state and moderate state is detailed
is set up as 10 min or 15 min. The SSPVs and SPTMs with
as follows.
different TGM and SD which are shown in Table 4.
1) Initialize ω = [ω1 , ω2 ], µ = [µ1 , µ2 ], and δ = [δ1 , δ2 ].
D. PARAMETER ESTIMATION 2) E step: compute
By analyzing the measurement data, we model the PDF of ωk ℵ(αm ; µk , δk )
good state as an mixture of two Gaussian distribution, which γ(m, k) = P2 , for all k (26)
is presented as n=1 ωn ℵ(αm ; µn , δn )
2 1 1 1 T −1
X ℵ(α; µ, δ) = D 1 e− 2 (α−µ) δ (α−µ)
(27)
p(α) = ωk ℵ(α; µk , δk2 ) (25) (2π) 2 |δ| 2

k=1
3) M step: update
where ω = [ω1 , ω2 ], µ = [µ1 , µ2 ], and δ = [δ1 , δ2 ] denote N
weight vector, mean vector, and variance vector, respectively. 1 X
µk = PN γ(m, k)αm , for all k
The EM algorithm is used here to estimate parameters of m=1 γ(m, k) m=1
MoG distributions [39]. (28)
8 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

TABLE 4: The SSPVs and SPTMs.

State Duration SSPV (P) SPTM (P)

0.996602 0.003375 0.000233
10 min 0.882108 0.106909 0.010983 0.027765 0.970755 0.00148
0.002668 0.013607 0.983725
0.966465 0.032974 0.000562
15 min 0.876565 0.11192 0.011515 0.258561 0.72793 0.013509
0.039695 0.134351 0.825954

be presented as
25
Measurement_SD10_TGM3_Good State
Fitting_SD10_TGM3_Good State E{h∗ (t − τ2 )h3 (t + τ2 )}
Measurement_SD15_TGM3_Good State ρ(t, τ ) = p 3 (31)
20 Fitting_SD15_TGM3_Good State
Measurement_SD10_TGM3_Moderate State
E{|h3 (t)|}2 E{|h3 (t)|2 }
Fitting_SD10_TGM3_Moderate State
Measurement_SD15_TGM3_Moderate State
Fitting_SD15_TGM3_Moderate State
where (·)∗ denotes complex conjugate operation and E{·}is
15
Measurement_SD10_TGM3_Bad State expectation operator. By substituting (12) and (31), the local
PDF

Fitting_SD10_TGM3_Bad State
Measurement_SD15_TGM3_Bad State temporal ACF can be calculated as
10 Fitting_SD15_TGM3_Bad State

ρ(t, τ ) = ρLoS (t, τ ) + ρNLoS (t, τ ) (32)

5
where the LoS component of the local temporal ACF can
be calculated as (33) and the NLoS components of the local
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 temporal ACF can be calculated as (34).
The signal amplitude of states, , (V)

FIGURE 9: The fitting between MoG distributions and mea- Please note that pϕ (ϕR ) is the PDF of the AAoA ϕR and
surement data. pθ (θR ) is the PDF of the EAoA θR . A number of distributions
have been adopted to describe the angle parameters, such as
the Gaussian distribution [40], the uniform distribution [41],
N
1 X and the Laplacian distribution [42], etc. In this simulation
δk = PN γ(m, k)(αm −µk )(αm −µk )T
γ(m, k) model, the AAoA ϕR and EAoA θR are described using the
m=1 m=1
(29) von-Mises distribution [43]. The von-Mises distribution can
PN
m=1 γ(m, k)
approximately transform to many distributions [44], which
ωk = , for all k (30) has been successfully validated by measurement data [43].
N
where N is the number of measurement sample. The PDF of the von-Mises distribution is given by
4) Repeat E step and M step until the convergence condi- eκ cos(ξ−ς)
tion is met. p(ξ) = , ξ ∈ [−π, π) (35)
2πI0 (κ)
The fitted results by the EM algorithm with the measure- where I0 (·) is the zero order modified Bessel function of
ment data are shown in Fig. 9. the first kind, ς is the mean value of the angles, and κ
The parameters of the mixture of two Gaussian distribution corresponds to the angular spread. The uniform distribution
in three states, such as the weight vector, mean vector, and is a special case of the von-Mises distribution when κ = 0,
variance vector, are shown in Table 5. 1
i.e., p(ξ) = 2π . As κ become larger, the angles become more
concentrated around the mean value ς, representing a non-
IV. STATISTICAL PROPERTIES isotropic scattering environment.
In this section, the statistical properties of the proposed
theoretical channel model in Section II will be derived based
on the expressions (12)–(20) under the non-isotropic scat-
tering condition. The statistical properties can capture the B. WIGNER-VILLE SPECTRUM (DOPPLER POWER
effects of the movements of Rx and clusters that in a non- SPECTRUM DENSITY)
stationarity behavior. There is significant meaning for the
satellite communication technology applied in the beyond 5G The Wigner-Ville distribution is presented as [45]
wireless systems. Z +∞
τ τ
W (t, f ) = h∗3 (t − )h3 (t + )e−j2πf τ dτ. (36)
A. LOCAL TEMPORAL ACF −∞ 2 2
The normalized temporal autocorrelation function between The Wigner-Ville spectrum is the expectation value of the
two complex fading envelope h3 (t − τ2 ) and h3 (t + τ2 ) can Wigner-Ville distribution, which is the Fourier transform of
VOLUME 4, 2019 9

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

TABLE 5: The parameters of the mixture of two Gaussian distribution in three states.

Weight Vector Mean Vector Variance Vector

State Duration State
ω1 ω2 µ1 µ2 δ1 δ2
Good State 0.347970392 0.65203 0.68128 0.726409 0.001035 0.000393
10 min Moderate State 0.296766172 0.703234 0.471615 0.603069 0.010054 0.002856
Bad State 0.29866231 0.701338 0.290238 0.03809 0.03842 0.001081
Good State 0.63378441 0.366216 0.727214 0.68443 0.000381 0.000991
15 min Moderate State 0.743162845 0.256837 0.610572 0.470644 0.003085 0.009401
Bad State 0.324595203 0.675405 0.313376 0.040384 0.042724 0.001149

R t− τ2 R t+ τ
   
K −j 2π 0 f LoS (ε)dε+φLoS (t− τ2 ) +j 2π 0 2 f LoS (ε)dε+φLoS (t+ τ2 )
LoS
ρ (t, τ ) = e (33)
K +1

τ τ
   
Z π Z π −j 2π R t− 2 f NLoS (ε)dε+φNLoS (t− τ ) +j 2π R t+ 2 f NLoS (ε)dε+φNLoS (t+ τ )
NLoS 1 0 0
pϕ (ϕR )pθ (θR )dθR dϕR .
2 2
ρ (t, τ ) = e
K +1 −π 0
(34)

the local temporal ACF ρ(t, τ ) B. RECEIVED SIGNAL LOCAL MEAN STATISTICAL
Z +∞ PROPERTIES
τ τ
S(t, f ) = E{ h∗3 (t − )h3 (t + )e−j2πf τ dτ } In the proposed theoretical model for small-scale fading, the
−∞ 2 2
Z +∞ number of rays is assumed to be infinity (S → ∞). It is
(37) impossible to implement this model directly. With respect
= ρ(t, τ )e−j2πf τ dτ
−∞ to a channel simulator, the simulation model, which is the
= S LoS (t, f ) + S NLoS (t, f ) discrete realization of the theoretical model with a finite
number of rays, should be used. The simulation model of the
where the LoS component of the Wigner-Ville spectrum
proposed small fading channel model is expressed as
can be calculated as (38) and the NLoS components of the
Wigner-Ville spectrum can be calculated as (39).
h3 (t) = hLoS (t) + hNLoS (t) (40)
r3 3
V. SIMULATION RESULTS AND ANALYSIS K j (2π f LoS (τ )dτ +φLoS (t))
R t
In this section, the generation and comparison with measure- = e 0 (41)
K +1
ment data of the received signal area mean power (corre- | {z }
LoS
sponding FSPL h1 (t) and shadowing h2 (t)) are presented. !
r S
The numerical and simulation results of the received signal 1 1 X j (R t fsNLoS (τ )dτ +φNLoS (t))
+ √ e 0 s .
local mean statistical properties of (corresponding small- K +1 S s=1
scale fading h3 (t)) are also provided. The effects of the | {z }
NLoS
movements of Rx is evaluated.
(42)
A. RECEIVED SIGNAL AREA MEAN POWER In the simulation, the modified method of equal areas
The Markov process s[n] which represents the switch posi- (MMEA) [46] is adopted to obtain the discrete AAoAs ϕR s
tion in Fig. 10 is obtained by the state transition diagram as (EAoAs θsR ) of finite rays through the following equations
depicted in Fig. 2. The initial state is generated by SSPV, and Z ϕRs
1 1
the following states are only determined by the last state and p(ξ)dξ = (s− ), s = 1, 2, · · · , S, ϕRs ∈ [−π, π)
−π S 4
SPTM. We can generate the received amplitude level α(t) by (43)
the state chain according to the PDFs of amplitudes of good Z θsR
1 1
state, moderate state, and bad state. p(ξ)dξ = (s− ), s = 1, 2, · · · , S, θsR ∈ [0, π)
0 S 4
To evaluate the performance of simulation, the PDFs of (44)
signal amplitudes of good state, moderate state, and bad where p(ξ) is the PDF of von-Mises distribution in (35).
state in this simulation is calculated and compared with The theoretical and simulated absolute values of the local
measurement data and the MoG distribution fitting based on temporal ACFs with different initial time t are shown in
EM algorithm in Fig.11. Fig. 12-(a). The theoretical and simulated results align well,
which demonstrates the correctness of our derivation and
10 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

R t− τ2 R t+ τ
   
Z +∞ f LoS (ε)dε+φLoS (t− τ2 ) +j 2π 0 2 f LoS (ε)dε+φLoS (t+ τ2 )
K −j 2π
e−j2πf τ dτ
LoS 0
S (t, f ) = e (38)
−∞ K +1

τ τ
   
Z +∞ Z π Z pi −j 2π R t− 2 f NLoS (ε)dε+φNLoS (t− τ ) +j 2π R t+ 2 f NLoS (ε)dε+φNLoS (t+ τ )
1 0 0
S NLoS (t, f ) =
2 2
( e
−∞ K +1 −π 0 (39)
R R R R −j2πf τ
pϕ (ϕ )pθ (θ )dθ dϕ )e dτ.

a G [t ] 1

Absolute value of local temporal ACF, | (t, )|

Theoretical, t=0s
Good State 0.9 Simulation, t=0s
Theoretical, t=5s
0.8 Simulation, t=5s

a M [t ] s[t ] 0.7

0.6
Theoretical, t=10s
Simulation, t=10s

Moderate State
a [t ] 0.5

0.4

a B [t ] 0.3

0.2
Bad State
0.1

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Time difference, (s)
FIGURE 10: Generation of the observed sequence in the
(a) With different initial time. (fc =39.402
Markov-chain-based channel model.
GHz, kvk=3 m/s, kvc k=0.5 m/s)
1

Absolute value of local temporal ACF, | (t, )|

Theoretical, V=2m/s
0.9 Simulation, V=2m/s
16
Theoretical, V=2.5m/s
Measurement_SD15_TGM3_Good State 0.8 Simulation, V=2.5m/s
14 Fitting_SD15_TGM3_Good State Theoretical, V=3m/s
0.7 Simulation, V=3m/s
Simulation_SD15_TGM3_Good State
Measurement_SD15_TGM3_Moderate State 0.6
12 Fitting_SD15_TGM3_Moderate State
Simulation_SD15_TGM3_Moderate State 0.5
Measurement_SD15_TGM3_Bad State
10 0.4
Fitting_SD15_TGM3_Bad State
Simulation_SD15_TGM3_Bad State 0.3
PDF

8
0.2

6 0.1

0
4 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18
Time difference, (s)

2
(b) With different velocity of Rx.
0 (fc =39.402 GHz, t=10 s, kvc k=0.5 m/s)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
The signal amplitude of states, , (V)
FIGURE 12: The theoretical and simulation absolute value
FIGURE 11: The comparison among measurement data, of the local temporal ACFs.
MoG distribution fitting, and simulation.

The theoretical and simulation absolute values of the Wigner-

simulation. The diversity of absolute values of the local Ville spectrums with different velocity of Rx (v) are shown
temporal ACFs with different initial time shows the non- in Fig. 13-(b). As velocity of Rx (v) increases, the Wigner-
stationarity of this model. The theoretical and simulated Ville spectrums become more dispersive and the peak value
absolute values of the local temporal ACFs with different become lower.
velocity of Rx (v) are shown in Fig. 12-(b). With a higher
speed of Rx, the absolute values of the local temporal ACF VI. CONCLUSIONS
drops fast, which can result in a shorter coherence time of the In this paper, we have proposed a 3D channel model for
channel. satellite communications at Q-band in high latitude. It con-
The theoretical and simulated absolute values of the tains three parts, i.e., FSPL model, a modified shadowing
Wigner-Ville spectrums with different initial time t are shown model based on a first order Markov-chain process, and a
in Fig. 13-(a). The theoretical and simulation results align small-scale fading based on GBSM. In the modified shad-
well with all selected initial time, which clearly demonstrats owing model, the SSPV and SPTM have been calculated
that the derivations and simulations are correct. The non- by the measurement data. The PDFs of amplitudes of three
stationarity of this model is also observed through the di- states (good state, moderate state, and bad state) follow the
versity of Wigner-Ville spectrums with different initial time. MoG distribution. The parameters of the MoG distributions
VOLUME 4, 2019 11

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

0.015
[10] T. Qi, W. Feng, and Y. Wang, “Outage performance of non-orthogonal
Theoretical, t=0s multiple access based unmanned aerial vehicles satellite networks,” China
Simulation, t=0s
Theoretical, t=5s Commun., vol. 15, no. 5, pp. 1–8, May 2018.
Wigner-Ville spectrum, S(t,f) Simulation, t=5s
Theoretical, t=10s
[11] A. Molisch, Wireless communications. UK: John Wiley & Sons, 2011.
0.01 Simulation, t=10s
[12] M. Milojevic, M. Haardt, E. Eberlein, and A. Heuberger, “Channel mod-
eling for multiple satellite broadcasting systems," IEEE Trans. Broadcast.,
vol. 44, no. 4, pp. 705–718, Dec. 2009.
[13] M. Pätzold, Y. Li, and F. Laue, “A study of a land mobile satellite
0.005
channel model with asymmetrical Doppler power spectrum and lognormally
distributed line-of-sight component," IEEE Trans. Veh. Technol., vol. 47, no.
1, pp. 297–310, Feb. 1998.
0
[14] A. Moro and U. Spagnolini, “Error probability of direct sequence-code
-150 -100 -50 0 50 100 150 division multiple access systems with adaptive antenna minimum mean-
Frequency, f (Hz)
square error multiuser receivers in Rayleigh-lognormal fading," IET Com-
(a) With different initial time. (fc =39.402 mun., vol. 3, no. 10, pp. 1649–1658, Oct. 2009.
GHz, kvk=3 m/s, kvc k=0.5 m/s) [15] C. Yoon, H. Lee, and J. Kang, “Performance evaluation of space-time
block codes from coordinate interleaved orthogonal designs in shadowed
0.015
Theoretical, V=2m/s
fading channels," IEEE Trans. Veh. Technol., vol. 60, no. 3, pp. 1289–1295,
Simulation, V=2m/s Mar. 2011.
Theoretical, V=3m/s
Simulation, V=3m/s [16] S. Scalise, H. Ernst, and G. Harles, “Measurement and modeling of the
Wigner-Ville spectrum, S(t,f)

Theoretical, V=6m/s
0.01 Simulation, V=6m/s land mobile satellite channel at Ku-band," IEEE Trans. Veh. Technol., vol.
57, no. 2, pp. 693–703, Mar. 2008.
[17] C. Fischione, M. Dangelo, and M. Butussi, “Utility maximization via
power and rate allocation with outage constraints in Nakagami-lognormal
0.005 channels," IEEE Trans. Wireless Commun., vol. 10, no. 4, pp. 1108–1120,
Apr. 2011.
[18] L. Yang and W. Fang, “Performance of distributed-antenna DS-CDMA
systems over composite lognormal shadowing and Nakagami-m-fading
0 channels," IEEE Trans. Veh. Technol., vol. 58, no. 6, pp. 2872–2883, Jul.
-150 -100 -50 0 50 100 150
Frequency, f (Hz) 2009.
[19] S. Hazra and A. Mitra, Advances in computing and communications.
(b) With different velocity of Rx. Berlin: Springer, 2011.
(fc =39.402 GHz, t=10 s, kvc k=0.5 m/s) [20] N. Youssef, C.-X.Wang, and M. Pätzold, “A study on the second order
statistics of Nakagami-Hoyt mobile fading channels," IEEE Trans. Veh.
FIGURE 13: The theoretical and simulation absolute value Technol., vol. 54, no. 4, pp. 1259–1265, Jul. 2005.
of the Wigner-Ville spectrums. [21] Q. Zhu, X. Dang, D. Xu, and X. Chen, “High efficient rejection method
for generating Nakagami-m sequences," Electronic Letter, vol. 49, no. 19,
pp. 1100–1101, Sept. 2011.
[22] H. Lin and M. Tseng, “Modeling fading properties for mobile satellite link
are estimated by EM algorithm from the measurement data channels using non-stationary hidden Markov model," IET Microwaves,
which is measured on the campus of HWU in Edinburgh, Antennas & Propag., vol. 3, no. 1, pp. 171–180, Feb. 2009.
UK. The first order statistical properties of received signal [23] M. Milojevic, M. Haardt, E. Eberlein, and A. Heuberger, “Channel mod-
eling for multiple satellite broadcasting systems," IEEE Trans. Broadcast.,
area mean power and the second order received signal local vol. 55, no. 4, pp. 705–718, Dec. 2009.
mean statistical properties have been derived and simulated [24] L. E. Braten and T. Tjelta, “Semi-Markov multistate modeling of the
to verify and analyze this model. land mobile propagation channel for geostationary satellites," IEEE Trans.
Antennas Propag., vol. 50, no. 12, pp. 1795–1802, Dec. 2002.
[25] S. Scalise, H. Ernst, and G. Harles, “Measurement and modeling of the
REFERENCES land mobile satellite channel at Ku-band," IEEE Trans. Veh. Technol., vol.
[1] C.-X. Wang, F. Haider, X. Gao, X.-H. You, Y. Yang, D. Yuan, H. Aggoune, 57, no. 2, pp. 693–703, Mar. 2008.
H. Haas, S. Fletcher, and E. Hepsaydir, “Cellular architecture and key [26] E. Cid, A. Vzquez Alejos, M. Garcla Snchez, and S. Garcla Fernndez,
technologies for 5G wireless communication networks," IEEE Commun. “Statistical model for satellite to helicopter radio channel," in Proc. 2014
Mag., vol. 52, no. 2, pp. 122–130, Feb. 2014. 44th European Microwave Conference, Rome, 2014.
[2] X. Ge, S. Tu, G. Mao, C.-X. Wang, and T. Han, “5G ultra-dense cellular [27] E. Cid, M. G. Sanchez, and A. V. Alejos, “Wideband analysis of the
networks," IEEE Wireless Commun., vol. 23, no. 1, pp. 72–79, Feb. 2016. satellite communication channel at Ku- and X-bands," IEEE Trans. Veh.
[3] S. Wu, C. X. Wang, e. H. M. Aggoune, M. M. Alwakeel, and X. H. You, Technol., vol. 65, no. 4, pp. 2787–2790, Apr. 2016.
“A general 3D non-stationary 5G wireless channel model,” IEEE Trans. [28] W. Fan, P. KyÃűsti, L. HentilÃd’, and G. F. Pedersen, “A flexible mil-
Commun., vol. 66, no. 7, pp. 3065–3078, Jul. 2018. limeter wave radio channel emulator design with experimental validations",
[4] C.-X. Wang, J. Bian, J. Sun, W. Zhang, and M. Zhang, “A survey of 5G IEEE Trans. Antennas Propag., accepted.
channel measurements and models,” IEEE Commun. Surveys Tuts., vol. 20, [29] A. W. Mbugua, W. Fan, Y. Ji and G. F. Pedersen, “Millimeter wave
no. 4, pp. 3142–3168, 4th Quart., 2018. multi-user performance evaluation based on measured channels with virtual
[5] T. Wei, W. Feng, J. Wang, N. Ge, and J. Lu, “Exploiting the shipping lane antenna array channel sounder," IEEE Access, vol. 6, pp. 12318–12326,
information for energy-efficient maritime communications,” IEEE Trans. 2018.
Veh. Technol. vol. 68, no. 7, pp. 7204–7208, Jul. 2019. [30] H. Fenech, A. Tomatis, S. Amos, J. S. Merino, and V. Soumpholphakdy,
[6] C. Liu, W. Feng, T. Wei, and N. Ge, “Fairness-oriented hybrid precoding for “An operator’s perspective on propagation," in Proc. IEEE EUCAP’14, The
massive MIMO maritime downlink systems with large-scale CSIT,” China Hague, Netherlands, Apr. 2014, pp. 2164–3342.
Commun., vol. 15, no. 1, pp. 52–61, Jan. 2018. [31] 3GPP, R1-1804479, “Path loss model for NTN", Nokia, Nokia Shanghai
[7] 3GPP, TR 38.811, “Study on new radio (NR) to support non-terrestrial Bell, 3GPP TSG RAN WG1 Meeting #92bis, Sanya, China, Apr. 16–20,
networks," Dec. 2017. 2018.
[8] W. Feng, J. Wang, Y. Chen, X. Wang, N. Ge, and J. Lu, “UAV-aided [32] ITU-R P.681-7, “Propagation data required for the design of earth space
MIMO communications for 5G Internet of Things,” IEEE Internet of Things land mobile telecommunication systems," Oct. 2009.
Journal, vol. 6, no. 2, pp. 1731–1740, Apr. 2019. [33] ITU-R P.525-3, “Calculation of free-space attenuation", Sept. 2016.
[9] Y. Chen, W. Feng, and G. Zheng, “Optimum placement of UAV as relays,” [34] J. Huang, C.-X. Wang, R. Feng, J. Sun, W. Zhang, and Y. Yang, “Multi-
IEEE Commun. Lett., vol. 22, no. 2, pp. 248–251, Feb. 2018. frequency mmWave massive MIMO channel measurements and charac-

12 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

terization for 5G wireless communication systems," IEEE J. Sel. Areas CHENG-XIANG WANG [S’01-M’05-SM’08-
Commun., vol. 35, no. 7, pp. 1591–1605, Jul. 2017. F’17] received the BSc and MEng degrees in
[35] J. Nessel, J. Morse, M. Zemba, C. Riva, and L. Luini, “Preliminary Communication and Information Systems from
results of the NASA beacon receiver for AlphaSat aldo paraboni TDP5 Shandong University, China, in 1997 and 2000,
propagation experiment," in Proc. 20th Ka and Broadband Communications respectively, and the PhD degree in Wireless Com-
Conference, Salerno, Italy, Oct. 2014. munications from Aalborg University, Denmark,
[36] M. Zemba, J. Morse, J. Nessel, “Frequency estimator performance for a in 2004.
software-based beacon receiver," in Proc. IEEE Antennas and Propagation
He was a Research Assistant with the Hamburg
Conference’14, Jul. 6–12, 2014.
University of Technology, Hamburg, Germany,
[37] B. Quinn and J. Fernandes, “A fast technique for the estimation of
frequency," Biometrika, vol. 78, no. 3, pp. 489–497, Sept. 1991. from 2000 to 2001, a Visiting Researcher with
[38] European Standard Telecommunications Series, ETSI, EN 302 307–2 Siemens AG Mobile Phones, Munich, Germany, in 2004, and a Research
(v1.1.1), “Digital video broadcasting (DVB); Second generation framing Fellow with the University of Agder, Grimstad, Norway, from 2001 to
structure, channel coding and modulation systems for broadcasting," Nov. 2005. He has been with Heriot-Watt University, Edinburgh, U.K., since
2014. 2005, where he was promoted to a Professor in 2011. In 2018, he joined
[39] C. M. Bishop, Pattern recognition and machine learning, Springer, New Southeast University, China, as a Professor. He has authored three books,
York, 2007. one book chapter, and over 350 papers in refereed journals and conference
[40] P. Kyösti, et al., “WINNER II channel models," IST–4–027756, WINNER proceedings, including 23 Highly Cited Papers. He has also delivered 17
II D1.1.2, v1.2, Apr. 2008. Invited Keynote Speeches/Talks and 7 Tutorials in international conferences.
[41] D. Shiu, G. Foschini, M. Gans, and J. Kahn, “Fading correlation and His current research interests include wireless channel measurements and
its effect on the capacity of multielement antenna systems," IEEE Trans. modeling, (B)5G wireless communication networks, and applying artificial
Commun., vol. 48, no. 3, pp. 502–513, Mar. 2000. intelligence to wireless communication networks.
[42] 3GPP, TR 36.873, V12.0.0, “3rd generation partnership project, technical Dr. Wang is a fellow of the IET, an IEEE Communications Society
specification group radio access network, Study on 3D channel model for
Distinguished Lecturer in 2019 and 2020, and a Highly-Cited Researcher
LTE (Release 12)," Jun. 2015.
recognized by Clarivate Analytics in 2017 and 2018. He is currently an
[43] A. Abdi, J. Barger, and M. Kaveh, “A parametric model for the distribution
of the angle of arrival and the associated correlation function and power Executive Editorial Committee member for the IEEE TRANSACTIONS ON
spectrum at the mobile station," IEEE Trans. Veh. Technol., vol. 51, no. 3, WIRELESS COMMUNICATIONS. He has served as an Editor for nine
pp. 425–434, May 2002. international journals, including the IEEE TRANSACTIONS ON WIRE-
[44] K. V. Mardia, Statistics of directional data. London: Academic, 1972. LESS COMMUNICATIONS from 2007 to 2009, the IEEE TRANSAC-
[45] K. Gröchenig, Foundations of time-frequency analysis. New York: TIONS ON VEHICULAR TECHNOLOGY from 2011 to 2017, and the
Springer Science & Business Media, 2013. IEEE TRANSACTIONS ON COMMUNICATIONS from 2015 to 2017.
[46] C. A. Gutierrez–Diaz–de–Leon and M. Pätzold, “Sum-of-sinusoids-based He was a Guest Editor for the IEEE JOURNAL ON SELECTED AREAS
simulation of flat fading wireless propagation channels under non-isotropic IN COMMUNICATIONS, Special Issue on Vehicular Communications and
scattering conditions," in Proc. IEEE GLOBECOM’07, Washington, USA, Networks (Lead Guest Editor), Special Issue on Spectrum and Energy
Nov. 2007, pp. 3842–3846. Efficient Design of Wireless Communication Networks, and Special Issue
[47] H. Suzuki, “A statistical model for urban radio propogation," IEEE Trans. on Airborne Communication Networks. He was also a Guest Editor for
on Commun., vol. 25, no. 7, pp. 673–680, Jul. 1977. the IEEE TRANSACTIONS ON BIG DATA, Special Issue on Wireless
Big Data, and is a Guest Editor for the IEEE TRANSACTIONS ON
COGNITIVE COMMUNICATIONS AND NETWORKING, Special Issue
on Intelligent Resource Management for 5G and Beyond. He has served as
a TPC Member, TPC Chair, and General Chair for over 80 international
conferences. He received ten Best Paper Awards from IEEE GLOBECOM
2010, IEEE ICCT 2011, ITST 2012, IEEE VTC 2013-Spring, IWCMC
2015, IWCMC 2016, IEEE/CIC ICCC 2016, WPMC 2016, and WOCC
2019. E-mail: [email protected].

LU BAI received the B.Sc. degree in Electronic

Information Engineering from Qufu Normal Uni-
versity, China, in 2014. She is currently pursu-
ing the Ph.D. degree in Information and Com-
munication Engineering with Shandong Univer-
sity, China. Her research interests include mas-
sive MIMO channel measurements and modeling,
satellite communication channel modeling, and
wireless big data. From 2017 to 2019, she is also
a visiting Ph.D. student at Heriot-Watt University,
UK. E-mail: [email protected].

VOLUME 4, 2019 13

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2019.2941975, IEEE Access

Lu Bai et al.: Preparation of Papers for IEEE Access

GEORGE GOUSSETIS [S’99, M’02, SM’12] WENQI ZHOU received the B.Sc. degree from
received the diploma degree in Electrical and Shandong Architecture University in 2001 and the
Computer Engineering from the National Techni- M.Sc. degree in Control Science and Engineering
cal University of Athens, Greece, in 1998, and the from Shandong University in 2006. He has been
Ph.D. degree from the University of Westminster, working at Shandong Huahan Electronics Com-
London, UK, in 2002. In 2002 he also graduated pany Limited since 2013 and engaged in the re-
B.Sc. in physics (first class) from University Col- search of information security, wireless communi-
lege London (UCL), UK. In 1998, he joined the cations, and intelligent robots. His current research
Space Engineering, Rome, Italy, as RF Engineer interests include wireless channel characterization
and in 1999 the Wireless Communications Re- and modeling. E-mail: [email protected].
search Group, University of Westminster, UK, as a Research Assistant.
Between 2002 and 2006 he was a Senior Research Fellow at Loughborough
University, UK. He was a Lecturer (Assistant Professor) with Heriot-Watt
University, Edinburgh, UK between 2006 and 2009 and a Reader (Associate
Professor) with Queen’s University Belfast, UK, between 2009 and 2013.
In 2013 he joined Heriot-Watt as a Reader and was promoted to Professor
in 2014, where he currently directs the Institute of Sensors Signals and
Systems. He has authored or co-authored over 500 peer-reviewed papers five
book chapters one book and four patents. His research interests are in the
area of microwave and antenna components and subsystems. Dr. Goussetis
has held a research fellowship from the Onassis foundation in 2001, a
research fellowship from the UK Royal Academy of Engineering between
2006-2011 and European Marie-Curie experienced researcher fellowships in
2011-12 and again in 2014-17. He is the co-recipient of the 2011 European
Space Agency young engineer of the year prize, the 2011 EuCAP best
student paper prize, the 2012 EuCAP best antenna theory paper prize and
the 2016 Bell Labs prize. He has served as Associate Editor to the IEEE
Antennas and Wireless Propagation Letters. E-mail: [email protected].

EL-HADI M. AGGOUNE [M’83-SM’93] re-

ceived the M.S. and Ph.D. degrees in Electrical
Engineering from the University of Washington
(UW), Seattle, WA, USA. He taught graduate and
SHANGBIN WU received his B.Sc. degree in undergraduate courses in electrical engineering at
Communication Engineering from South China many universities in the USA and abroad. He
Normal University, Guangzhou, China, in 2009, served at many academic ranks, including an En-
M.Sc. degree in Wireless Communications with dowed Chair Professor. He is listed as Inventor
distinction from University of Southampton, in two patents assigned to the Boeing Company,
Southampton, UK, in 2010, and Ph.D. degree in USA, and the Sensor Networks and Cellular Sys-
Electrical Engineering from the Heriot-Watt Uni- tems Research Center, University of Tabuk, Saudi Arabia. He is a Profes-
versity, Edinburgh, UK in 2015. From 2010 to sional Engineer registered in the State of Washington. He is currently serving
2011, he worked as a LTE R&D engineer respon- as a Professor and the Director of the SNCS Research Center, University
sible for LTE standardization and system level of Tabuk. His research is referred to in many patents, including patents
simulation in New Postcom Equipment Ltd., Guangzhou, China. From assigned to ABB, Switzerland, and EPRI, USA. He authored many papers
October 2011 to August 2012, he was with Nokia Siemens Network, where in IEEE and other journals and conferences. His research interests include
he worked as a LTE algorithm specialist, mainly focusing on LTE radio wireless sensor networks, energy systems, and scientiïňAc˛ visualization. He
resource management algorithm design and system level simulations. He has is serving on many technical committees for conferences worldwide as well
been with Samsung R&D Institute UK as a 5G researcher since November as reviewer for many journals. One of his Laboratories won the Boeing
2015. E-mail: [email protected]. Supplier Excellence Award. He won the IEEE Professor of the Year Award,
UW Branch. E-mail: [email protected].

QIUMING ZHU received his B.Sc. degree in

Electronic Engineering from Nanjing University
of Aeronautics and Astronautics (NUAA) in Nan-
jing, China in 2002 and his M.S. and Ph.D. de-
grees in Communication and Information Systems
in 2005 and 2012, respectively. Since 2012, he
has been an associate professor in Wireless Com-
munications. From 2016 to 2017, he was also a
visiting academic at Heriot-Watt University. His
research interests include channel modeling for 5G
communication systems and wireless channel emulators. E-mail: zhuqium-
[email protected].

14 VOLUME 4, 2019

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/.