Page 57 - ITU Journal: Volume 2, No. 1 - Special issue - Propagation modelling for advanced future radio systems - Challenges for a congested radio spectrum

ITU Journal: ICT Discoveries, Vol. 2(1), December 2019

MS and BS does not have a signi icant impact on our re- Parameters obtained by curve itting are presented in Ta-
sults. The main reason is that we set the simulation area ble 3. Noting that due to the curve itting, we end up with
as a square with a side length of 1.6 km, and farthest dis- the result for the model such that ∫ 180 ( ) ≈ 1.
m
m

tance from MS to BS is 600 m. The distance variation in −180
our simulation is not very big, but it is enough to cover a 5.4 Cluster power ratio
reasonable range. In other words, our model is distance-
independent, valid for a range of a few hundred meters, The power per cluster can be solved by two sub-problems,
such as an outdoor urban microcell scenario. In the fol- the total received power, P at the MS, and the power dis-
r
lowing parts of this section, the directional channel model tribution among ℳ clusters. In this paper, we are more
is represented with two carrier frequencies, 2 GHz and 28 interested in the latter problem. For the irst one, based
GHz, to cover both examples of microwave and mmWave on the speci ic scenarios, the P calculation can be solved
r
spectrum. by many widely accepted omnidirectional path loss mod-
els in the existing literature, such as the log-normal shad-
5.2 Number of clusters owing model, two-ray model, Okumura–Hata model etc.
In this paper, we do not recommend any path loss expo-
Once the cluster gap is set, all the rays at one receiver nent or shadowing parameters, because the main contri-
can be grouped into ℳ clusters in the azimuth domain. bution from us is the directional features of the power at
ℳ is characterized by its PMF, (ℳ = ), =1,...,5, pre- the receiver end. To model the power per cluster, we de-
sented in Table 2. ine a cluster power ratio as
P m
Table 2 – Probability mass function of ℳ = . (6)
m
P r
ℳ 1 2 3 4 5 Note that 0 < m ≤ 1, and if ℳ = 1, m = 1. For
2 GHz 0.2496 0.4389 0.2551 0.05419 0.002193 other cases given ℳ > 1, we ind that no matter what
28 GHz 0.3821 0.3836 0.2005 0.03124 0.002588 the value of ℳ is, the PDF of follows a U-shape dis-
m
tribution, which can be very well itted by Kumaraswamy
distribution [26].
5.3 Cluster center

To further outline the azimuth feature of the clusters, we
introduce cluster center, , as the average azimuth of all ( |ℳ) = { 1, ℳ = 1 (7)
m
m
−1
the rays within the same cluster −1 (1 − ) , ℳ > 1.
m
m
A summary of and parameters conditioned by ℳ is
∑
ℛ m
= =1 m (3) listed in Table 4. Another important condition of the
m
ℛ
power ratio is that the summation of all the at one MS
where ℛ is the total number of rays within the same is 1, ∑ ℳ = 1. m
m
m=1
m
cluster. is the DOA per ray within this cluster in de-

gree. Then based on our RTS, we extract the PDF of as
m
a piecewise exponential function, Table 4 – Parameters of m
ℳ = 2 ℳ = 3 ℳ = 4 ℳ = 5

⎧ exp(− ( + 90)), −180 ≤ < −90 2 GHz 0.17 0.37 0.15 0.45 0.20 0.55 0.03 0.50
2
2
m
m
{ 28 GHz 0.17 0.37 0.12 0.55 0.05 0.50 0.01 0.55
{ exp( ), −90 ≤ < 0
( ) = ⎨ 1 1 m m

m
exp(− ),
{ 1 1 m 0 ≤ < 90
m
{ 6. SIMULATION RESULTS
⎩ exp( ( − 90)), 90 ≤ < 180.
2
2
m
m
(4) In the RTS, we consider the system models presented in
To make (4) right continuous and symmetric on = 0, section 3 using the parameters in Table 5.
m
= ( = −90) = exp(−90 ) (5)
1
1
m
2

We generate ive realizations of the virtual city models
with the same set of , and , and launch RTS in

ive such environments to average the randomness. At
Table 3 – Parameters of ( m )
each MS, we consider the maximum of 100 incident rays.
1 1 2 However, the power of some rays can be low, so that they
2 GHz 0.005658 0.01340 0.007666 hardly contribute to the total received signal. Therefore,
28 GHz 0.007494 0.02485 0.02121 we remove such rays with power below a threshold, P off =
-150 dBm. The RTS returns in the output iles and the
i
© International Telecommunication Union, 2019 41