International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013
A Novel Geometrical Mixed-Bounce Two-Ring
Deterministic Model for MIMO Mobile-to-Mobile Non-
Isotropic Communication Channel
MALA Umar Mustapha Bala Alhaji Salihu
Bakura Beijing University of Posts and
University of Maiduguri, Nigeria Telecommunications China
ABSTRACT
An accurate modeling of multiple-input multiple-output
(MIMO) channels is a prerequisite for designing multi-
antenna system. In this paper, we propose a geometrical
mixed-bounce two-ring deterministic model for mobile-to-
mobile channel considering the non-isotropic scattering while
adopting the Von Mises probability density function for both
the angle of departure AoD and angle of arrival AoA
surrounding the transmitter and receiver respectively.
Beginning with isotropic scattering, the expression for two
dimensional (2-D) space time cross-correlation (STCC)
function between any two sub channels is derived.
Comparison between the statistical properties of the
deterministic model under the assumption of non-isotropic
scattering with that of the reference model indicates a good
agreement, thus, the importance of this model. The result
obtained is in conformity with that of the double bounce and
single bounce two ring models.
General Terms
Model.
Keywords
Mixed bounce, mobile-to-mobile channel, space time
correlation, Isotropic scattering.
1. INTRODUCTION
Designing a very high speed minimum error wireless link
having good quality of service (QoS) and with considerable
range capabilities in a highly fading channel attracts
significant research interest and poses a great challenge in
communication engineering. Assuming no fading for the
moment, we can, in principle, meet the 1-Gb/s data rate
requirement if the product of bandwidth (Hz) and spectral
efficiency (b/s/Hz) equals 109. As clearly contained in
literature, a variety of cost, technology, and regulatory
constraints make such a brute force solution unattractive, if
not impossible. In an effort to achieve high data rate wireless
communication services with less interference and fading,
communication engineers have recently concentrated much of
their resources and time on the new dimension- the space. It is
now a well-established fact that higher data rates are
achievable by employing MIMO which entails the application
of array of antennas both at the transmitter and receiver
utilizing the space-time technology. The analysis and
simulation of space-time wireless communication systems
need accurate, yet tractable spatio-temporal channel models.
Here we aim at providing a novel mixed-bounce mobile-to-
mobile deterministic channel model assuming non-isotropic
Suleiman Zubair
UTM MIMOS CoE Universiti
Teknologi Malaysia Johor,
Malaysia
scattering of electromagnet waves. Non isotropic scattering is
a good approximation as it is more near to real phenomena. It
is natural that the multipath signal components received at any
antenna may consist of the single-bounce, the double-bounce
and even both of them. Single-bounce literally means signals
from the transmitting antenna undergo a single hop through a
scatterer before arriving at the receiving antenna while double
bounce involves two hops before arriving at the receiver. We
in this paper are incorporating all the signal paths in
determining the channel transfer function and further looking
at some of its statistical properties. In literature many
geometrically based MIMO channel models were proposed
such as the one ring, two ring, multi-ring and spherical
models. In the two ring model both single bounce [1] and [2]
and double bounce [3] were discussed. To the best of our
knowledge no work was done in mixed-bounce where the link
between the antenna-scatterer-antenna and antenna-scatterer-
scatterer-antenna are considered together which constitutes
the central point of concentration of this paper.
The rest part of the paper is arranged as follows. Section two
discusses the two-ring model; section three gives the cross-
correlation; Section four for computation of AoD and AoA
while section five is for simulation and result and conclusion
is in section six..
2. THE GEOMETRY OF THE TWO-
RING MIXED-BOUNCE MODEL
The geometry of the two-ring mixed-bounce is shown in fig.1
for a MIMO mobile-to-mobile channel. For simplicity two
transmit antennas and two receive antennas are considered
where local scatterer of sTk and sRl are equally distributed on
the circumference of a separate rings around the transmitter
and receiver respectively. The paramount difference between
our model and other two ring models is that in our model both
the single and double bounce rays were considered
simultaneously, since each multipath contributes to the total
system’s data rate. As could be easily observed from Fig. 1,
the local scatterers around the transmitter represented as sTk
(k=1,2,3….,M) are located on a ring of radius RT and the
local scatterers around the receiver denoted as sRl
(l=1,2,3…..N) lie on a separate ring with radius RR . The
symbols tk and  Rl represent the main angle of departure
1
 International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013

vT vR
d2 SLR
SKT
d1 d4 d3
d5
A(q)R
A(P)T d6
Φ kT d7 ψ kT
δT α T βT
αR
Φ lR βR
X
ψl R

A(P’)T A(q’)R

δR
Rt Rr

Fig. 1 Mixed Bounce Two Ring Model

(AoD) and main angle of arrival (AoA) respectively and the
auxiliary AoD and AoA are represented by Rl and  Tk
respectively. It is assumed that the radii RT and RR are small in
comparison with D, which is the distance between the
transmitter and receiver.
The symbols T and  R are used to represent the antenna
spacing at the transmitter and receiver respectively. Based on
our assumption the inequalities max RT , RR  min T ,  R  and
min RT , RR  max T ,  R  holds. The tilt angle between the
x-axis and the orientation of the transmit antenna arrays is
denoted by T and the tilt angle of the receiver antenna array
is given by  R .
Both the transmitter and receiver are assumed to move with
the velocities of vT and vR respectively in the direction
denoted T and  R . Furthermore, 2ΔT is the maximum angle
spread at the transmitter determined by the scatterers around
the receiver. Similarly 2Δ R represents the maximum angle
spread at the receiver determined by the scattering around the
transmitter. The geometrical two ring model is an appropriate
approximation for scenarios in which neither the transmitter
nor receiver is elevated but both are surrounded by a large
number of scatterers. Such scattering condition may exist both
in indoor and outdoor environment. Using cosine rule and
other mathematical simplifications, the effective channel
transfer function is

h pq  t  
M
PTS lim
k 
dconst
M g k 
T am exp  j

 R
 
cos  R   Tk  jΨT
k
   
 j 2πt ( fTmax cos T  Tk  f Rmax cos( R  Tk )) 

K 1
N
  T  Rl   f R
(1)

d  T 
 PRS lim const gR l
bn exp  j cos T  R
l
 jΨlR  j 2πt ( fTmax cos cos( R  R
l
)) 
l  N   max

l 1
M ,N
 PSTR lim
k ,l  MN
1 kl
gTR

g mn exp j0  jΨTR
kl

 j 2πt( fT( m)  f R( n) ) 
k ,l 1

Where

gmn  ambncmn (2) 

fT(m)  fTmax cos T  Tk  (7)
 
  

am  exp( j  T  cos T  Tk  (3) f R(n)  f Rmax cos  R  R
l
 (8)

2
 4R    4R  
dconst  exp  j T   exp  j R  (5) 0  

 RT  RR  D  (9)
     

cmn =exp{ j
2
  
(RT cos Tk  Rl  RR cos R
l
)} (6)

2
 International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013

R pq, p'q' T ,  R ,   R SBT ' T ,  R ,   RSBR  ,  ,   Rpq

DB
 ,  ,  (1)
pq, p q' pq, p'q' T R , p 'q ' T R
R pq, p' q ' T ,  R ,  
M
lim
PTS
k  M E  g
 k 2  2 
T  exp  j
 
 k
 k 
 T cos T  T   R cos(  R  T   j 2πt ( fTmax cos T  T  f Rmax cos( R  T )) 
 
k

k

 
K 1
N


 l 2  2 
     
P l  l 
(11)
  R cos  R  R  T cos T  R   j 2πt ( fTmax cos T  R  f Rmax cos( R   R )) 
l l
 lim RS E  gR  exp  j
l  N      
l 1

 
M ,N
 kl 2  2 2  j 2 t fT( m )  f R( n )

PSTR
 lim E{ gTR  ambn e
k ,l  MN
k ,l 1
 

cos( R  TK )   cos  R  T sin  R sin ΘT ,cos( R  TK )   cos  R  T sin  R sin ΘT (12)

cos(T  R
l
)  cos T   R sin T sin ΘR andcos(T  Rl )  cos T  R sin T sin ΘR (13)


 exp  x sin   y cos  d  2 I 

 
0 x2  y 2  (14)



 2   2 
Let AT  exp   j  R cos  R  j 2πtf Rmax cos  R  , AR  exp  j T cos T  j 2πtfTmax cos T  (15)
     
2 2
BcosT  j T cos  T   j 2πtfTmax cos T ,BcosR  j  R cos   R   j 2πtf Rmax cos R (16)
 
2π 2
BsinR   j 2πtf Rmax sin  R  j  R sin  R  j T  R sin T  j 2πtfTmax  R sin T (17)
λ 
2 2
CDT  j T cos  T   j 2πtf Rmax cos T and CDR  j  R cos   R   j 2πtfTmax cos  R (18)
 
2 2
DDT  j T sin T  j 2πtf Rmax sin T and DDR  j  R sin  R  j 2πtfTmax sin  R (19)
 

As in [1] the diffuse component of h pq  t  follows that the continuous random variable AOD and continuous random
variable AOA where each of which are characterize by
mean value and mean power are equal to 0 and 1respectively certain distribution represented as and respectively. The
hence the central limit theorem state that h pq  t  is a zero-mean infinitesimal power contribution due from each scatterer
complex Gaussian process with unit variance. Consequently around the transmitter and receiver corresponding to the
differential angle and is given as:
the envelope h pq  t  is a Rayleigh fading process. One can
 2  l 2
show that the diffuse component of the Link E  gTk  E  gR 
' ' 
 
  p ( )d
T T
, 
 
  p R ( R )d R and
from ATp to ARq can be obtained from (1) by replacing by their M
T
N
 kl 2 
respective complex conjugates. The rest elements of the E  gTR 
matrix could be obtained as in [3].    p ( ) p ( )d d .
T T  R R T R
MN

2.1 The Space-Time Cross Correlation For non-isotropic scattering, the AoD and AoA have non-
Function CCF uniform distribution over  ,   . From [9] we get equation
The non-normalized space-time correlation function between (14) - (19).
two complex faded envelopes hpq and hp'q' is defined
The Von Mises also known as circular normal distribution
as R pq, p'q' T ,  R ,  
 E[hq t hp'q' t   ]
  
 . From [5] the describes a "normal" distribution on a circular path of 2.
Given by formula in equation (20).
effective space-time correlation function for the mixed bounce
exp  k cos    
can be calculated using the formula in equation (10) where the
superscripts SBT, SBR, and DB stands for Single Bounce  
f  , k 
2 I 0  k 
(20)
Transmitter, Single Bounce Receiver and Double Bounce
respectively. The effective STCC for the reference model For non- isotropic distribution the Von Mises distribution is
reduces to the following expression after some mathematical considered. For represents isotropic distribution and I0 is the
operations and reductions as in equation (11). Furthermore,
setting sin   0and cos   1 we obtain equations (12) and (13). Various other non-uniform distributions such as Gaussian,
quadratic, Laplacian and cosine were discussed in other
For values of M and N approaching infinity the discrete angle literatures but due to mathematical suitability the Von Mises
of departure and discrete angle of arrival becomes

3
 International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013

R pq, p' q ' T ,  R , 


I 0  k R 2  CD

2
R
2
 DD
R
 


 2k R CDR cos  R  DDR sin  R   I 0   kT 2  C D

2
T
2
 DD
T



 2kT CDT cos T  DDT sin T   PSTR 

I 0  kT   I 0  k R 

I 0  kT 2  Bcos

2
T
2
 BSin
T
 
 2kT BcosT cos T  BsinT sin T  PTS AT

 (21)

I 0  kT 

I 0  k R 2  Bcos

2
R
2
 Bsin
R
 
 2k R BcosR cos  R  BsinR sin  R  PRS AR



I0  kR 

is adopted in this paper. Substituting equations (11), (12) and

(15) to (19) into 11 and undergoing other manipulations we
obtain equation (21). Next, we introduce the information theoretic channel capacity
of the SSM. The capacity [4] in the absence of channel
3. THE SIMULATION MODEL knowledge at the transmitter is given
 
It is very difficult if not impossible to implement the reference by Cˆ  t   log2 det  I M  PT Hˆ  t  Hˆ H (t )   where Hˆ  t   hˆ pq  t 
 R  
  MT PN  
model either using hardware or software since it involves
infinity harmonics, therefore the need to come up with the assuming MT  N R , I M R is the M R  M R identity matrix, PN
simulation model. We begin by the stochastic simulation is the noise power, PT is the total transmitted power allocated
model considering some of its properties such as its space-
time CCF after which we also discuss the deterministic uniformly to all M T transmit antenna element.  . H is the
simulation model by applying the principles of deterministic conjugate complex transpose operator. The signal to noise
channel modeling along with some of its basic properties as in
SSM [11]. ratio (SNR) is given by PT gives the statistical average of the
PN

3.1 The Stochastic Simulation Model SSM stochastic capacity with respect to the random phases.

The SSM system can easily be derived from the reference
model by considering finite number of scatterers both around
the transmitter M and the receiver N. Of course the scatterers
are responsible for the multipath propagation ensuring
multiple copies of the transmitted signals at the receiver. Even
though we have finite harmonics at the receiver, the phase
angles n ,m and mn are still i.i.d. random variables each
3.2 The Deterministic Simulation Model
DSM
By making the phase angles and all other parameters of SSM
constant, a new model called Deterministic Simulation Model
is obtained. Observing rightly that the diffuse component of
the link between AT  and AR  is an ergodic process permits
p q

with uniform distribution on the interval (0,2] thus we have

us to consider the properties of the single sample of hˆ pq  t  to
stochastic simulation model SSM where its channel transfer
function is given by hˆ pq  t  . This is for the link between, represent the ensemble properties of the system denoted
by h pq  t  . If the phases are outcome of a random generator
The 3-D space –time CCF between two different with uniform distribution in the interval (0, 2],
links hˆ pq  t  and hˆ p'q'  t  is calculated using the n ,m and mn are no longer a random variables but constant


formula ˆ pq, p'q' T ,  R ,   E hˆ pq  t  hˆ*p'q'  t    . In the quantities. Appropriate choice of parameters of h pq  t  leads
to very good and close approximation of h pq  t  . h pq  t  is a
equation above . represent complex conjugate and E 
*
. is
deterministic function of time therefore statistical properties
the expectation operator, which applies to the random of DSM is analysed using time averages rather than statistical
phases n ,m and mn . The closed form expression is therefore averages. The 3-D space-time CCF for example is computed
thus  pq, p'q' T ,  R ,   hˆ pq  t  hˆ*p'q'  t    where . denotes
the time average operator. According to [11, Pg. 66], the
ˆ pq, p 'q '
relation  pq, p'q' T ,  R ,   ˆ pq, p'q' T ,  R ,  holds .This
 2 
 PTS exp   j  R cos  R  j 2 f Rmax  cos  R 
   is applicable to ergodic systems. The DSM’s channel
 1 capacity is written as
 2 2 
   2   Δ sin   Δ sin   2 sin       
     R T R R T R T T
  C  t   log 2 det  I M R 
PT
H  t  H H (t )  
  
   2  2
    MT PN  
 
 
2
  T  T  2 fTmax  cos T  T    2 fTmax   
I0  

     



where H  t   hpq  t  is deterministic and MT  N R . The
   2 2  
 
 
 
 R ΔT  sin  R fTmax sin T  ΔT f Rmax sin  R 


 time average mC of channel capacity C  t  is given by the
  2

 
   2  f Rmax  ΔT sin  R  f Rmax τΔT sin  R  T sin T  2 fTmax  sin T
2
  T
      (22)

 C t  dt .
1
  relation mC   C  t    lim The stochastic
T  2T
T

4
 International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013

capacity Ĉ  t  is mean ergodic if mCˆ (t ) is independent of wavelength   0.15m . Under an isotropic scattering we arrive
at equations (26) and (27).
time t and equal mC thus mCˆ  mC [11, Pg. 79].
Fig. 2 is the mixed bounce composite cross correlation of the
entire channel. When we critically observe the figure we find
4. COMPUTATION of AoD and AoA that the it is averaging around zero which clearly indicates
ASSUMING NON-
ISOTROPIC  2 2
R pq, p q T ,  R ,    4  I 02   fR     
 fT    
 R  
2 2 2
  T   2 T f Rmax  
SCATTERING 

' '
max
    max
   
 
On choosing appropriate values for M 
2 j 2 f Rmax   2 T  R  T  R  T  R  T  (27)
and N, in the simulation model we are   2  e I0  fTmax   fTmax   T     

   2     2  2 
only left with AoD Tk and AoA    
 1   R 2 2 
 Rl 2 j 2 fTmax
I0   fR    2  RT   R  2T  R 
to be determined. For non- 2
  2  e  fTmax   fTmax   R      
 2  2        
isotropic distribution the Modified 
max
   2

 
Method of Equal Areas (MMEA) is
applicable in order to calculate the
parameters. This method involves
finding the set of AoAs or AoDs   such that the ACF for
k
T The Correlation Function Diagram

the reference model is approximately equal to ACF of the

deterministic model r    r   ,   0, max 
5
for N  M  20 . Using numerical root finding techniques [12],
mixed bounce correlation function.  (T,R, )

4
the AoD and AoA are determined from the expressions (23) 3
and (24) 2

T( m ) 1

m 1

0
4 pT (T )dT  0, m  1, 2,............, M (23)
M -1

T(0)  -2

-3
T( m ) 0

n 1

2 8
4 pR (T )d R  0, n  1, 2,............, N (24) 7
6
4
N 4
5

T(0)  6
2
3
1
8 0

The closed form AoA and AoD are computed if the inverse of .fTR
/  max

the function exist by

 n 1  that the correlation is zero and thus it means that the

T   R  F 1 T   4 (25) interference between co-channel is less when using MIMO
 N 
  system.
The MMEA has the advantages that the statistical properties pT T  dT  p R  R  d R  1 (26)
matches those of the reference model for only small time  2 
delay while requiring higher values of scatterers (M,N = 40)
to match the properties of the reference model. Fig 2 Mixed bounce cross correlation function

5. SIMULATION AND RESULTS 6. CONCLUSION

After determining the composite space time cross correlation Here a mixed bounce MIMO mobile-to-mobile deterministic
functions for the mixed bounce two ring model, we hereby channel model is presented using the two rings as a basis
present the result of a special case of isotropic scattering using structure. Our simulation model was derived from non-
the following model parameters. The antenna tilt realizable reference model by applying the concept of
deterministic channel modeling. It is assumed that both the
angle T   R   2 . The transmit antenna is assumed to move
transmitter and the receiver antenna are surrounded by finite
at T   4 while the receiver antenna moves at  R  0 . number of scatterer with a given radius and are also assumed
to be moving at a given Doppler frequency. The homogenous
Identical maximum Doppler frequency is assumed at both the channel transfer function is presented while the space-time
transmitter and receiver given as 91 Hz and cross correlation was derived. Our model particularly
extended the work done by Matthias Pätzold et al by
considering mixed bounce.

5
 International Journal of Computer Applications (0975 – 8887)
Volume 70– No.28, May 2013

7. REFERENCES
[1] huangquan Wang, Raghukumar K., Abdi A.,
Wallace J., Jensen M., Indoor MIMO channels: a
parametric correlation model and experimental
results. Symposium on Advances in Wired and
Wireless Communication, 2004 IEEE/Sarnoff.
Publication Year: 2004 , Page(s): 1 - 5.
[2] Tang Z., Mohan A.S., A correlated indoor MIMO
channel model. Canadian Conference on Electrical
and Computer Engineering, 2003. IEEE CCECE
2003. Publication Year: 2003 , Page(s): 1889 - 1892
vol.3
[3] Patzöld M.,Hogstad B.O., Youssef N., Dongwoo
Kim. A MIMO Mobile-To-Mobile Channel Model:
Part I - The Reference Model. IEEE 16th
International Symposium on Personal, Indoor and
Mobile Radio Communications, 2005. Volume: 1
PIMRC 2005. Page(s): 573 - 578
[4] Hogstad B.O., Patzold M.,YoussefN., Dongwoo
Kim. A Mimo Mobile-To-Mobile Channel Model:
Part II - The Simulation Model. IEEE 16th
International Symposium on Personal, Indoor and
Mobile Radio Communications, 2005. PIMRC 2005
Volume: 1 Publication Year: 2005 , Page(s): 562 –
567
[5] Zajic A.G., Stubber G. Space-Time Correlated
Mobile-To-Mobile Channels: Modelling And
Simulation. IEEE Transactions on Vehicular
Technology, Volume: 57 , Issue: 2. Publication
Year: 2008 , Page(s): 715 – 726
[6] Xiang Cheng, Cheng-Xiang Wang; Laurenson,
D.I.; Hsiao-Hwa Chen; Vasilakos, A.V.; A Generic
Geometrical-Based MIMO Mobile-to-Mobile
Channel Model. International Wireless
Communications and Mobile Computing
Conference, 2008. IWCMC '08. Publication Year:
2008 , Page(s): 1000 – 1005
[7] Bakhshi, G.; Saadat, R.; Shahtalebi, K.; A modified
two-ring reference model for MIMO Mobile-to-
Mobile communication channels. International
Symposium on Telecommunications, 2008. IST
2008. Publication Year: 2008 , Page(s): 409 – 413
[8] Gregory D. Durgin “Space-Time Wireless Channel”
Prentice Hall PTR, Upper Saddle River, NJ07458.
[9] I.S Gradshteyn and I.M Ryzhik “Table of Integrals,
Series, and Product” (Seventh Edition).Academic
Press.
[10] David Tse, Pramod Viswanath “Fundamentals of
Wireless Communication” Cambridge
[11] M. Patzöld. Mobile Fading Channels, John Willey
and Sons. 2002

6
IJCATM : www.ijcaonline.org