IEEE COMMUNICATIONS LETTERS, VOL. XX, NO.

X, JUNE 2015

Time Successive SSK-MPSK: A System Model to

Achieve Transmit Diversity
P. Maheswaran, Student Member, IEEE, M. D. Selvaraj, Member, IEEE

AbstractIn this letter, we consider an Nr Nt multipleinput multiple-output (MIMO) system in which M -ary phase
shift keying (MPSK) and space shift keying (SSK) are used
in successive time slots. We call this system Time Successive
SSK-MPSK (TSSM) and we show that this system achieves a
transmit diversity order of two. In TSSM, during the first time
slot, information is transmitted with MPSK symbols using all Nt
transmitter antennas with each antenna using Nt th of the total
transmitter power. During the second time slot, SSK modulation
is used to transmit the same information. We further derive the
optimal detection criterion, pairwise error probability (PEP) and
the average union bound of the bit error rate (BER) of TSSM.
We show the diversity order using asymptotic approximation of
the average union bound of the BER of TSSM. We compare
the BER of TSSM with that of Time Orthogonal Signal Design
assisted SSK (TOSD-SSK), Coherent Space Time Shift Keying
(CSTSK) and other schemes. The simulation results show that
TSSM performs considerably better.
Index TermsBit error rate (BER), M -ary phase shift keying (MPSK), Rayleigh fading, space shift keying (SSK), Time
Successive SSK/MPSK (TSSM), transmit diversity.

I. I NTRODUCTION
PACE Shift Keying (SSK) [1] is a multiple-input multipleoutput (MIMO) technique that conveys information by
activating an antenna out of Nt transmitter antennas thereby
achieving a spectral efficiency of log2 (Nt ). As SSK activates
only one transmitter antenna during any symbol duration,
inter channel interference is avoided at the receiver. Further,
only one radio frequency (RF) chain is required for the
implementation of SSK. The results in [1] show that only
receive diversity is achievable in SSK. By using wave shapes
with good auto correlation and cross correlation properties, a
transmit diversity order of 2 is reported for Time Orthogonal
Signal Design assisted SSK (TOSD-SSK) in [2]. Shaping
filters are used at the transmitter to realize the wave shapes
in TOSD-SSK. Coherent Space Time Shift Keying (CSTSK)
proposed in [3] achieves a maximum transmit diversity order
of min(Nt , T ) where T is the total number of time slots used
for symbol transmission. In Space-Time Block Coded Spatial
Modulation (STBC-SM) [4], a transmit diversity order of 2
is achieved by applying Alamoutis STBC in SM. Against
this background, we propose a two time slot scheme in this
letter that achieves a transmit diversity order of 2. This two
time slot scheme achieves the transmit diversity based on
the independence of two chi-square random variables (RVs)

P. Maheswaran and M. D. Selvaraj are with Indian Institute of Information

Technology, Design & Manufacturing Kancheepuram, Chennai-600127, India
(e-mail: [email protected], [email protected]).

Fig. 1. Time Successive SSK-MPSK system model

discussed in Section III-A without the use of any shaping

filters or STBC at the transmitter.
We present this letter with the following organization. In
Section II, we present the system model of TSSM and the
optimal detection criterion for the same. We derive the average
union bound of the BER of TSSM based on pairwise error
probability (PEP) in Section III. We also derive the asymptotic
expression for the BER of TSSM in Section III. We present the
simulation results and discussion in Section IV and conclude
the paper in Section V.
II. S YSTEM M ODEL
Consider an Nr Nt MIMO system as shown in Fig. 1.
In this system, we assume that the RF chain available at
the transmitter is power constrained to Es /Ts , where Es
and Ts are symbol energy and duration respectively. We
further assume that the transmitter is capable of switching
the modulation between SSK and MPSK with the number of
transmitter antennas Nt = M , where M is the total number
of symbols in MPSK constellation. A random sequence of
independent bits is given as input to a mapper. The mapper
groups log2 (M ) bits from this sequence and maps it to
the information symbol a {1, . . . , M }. In time slot one,
an MPSK symbol sa is transmitted simultaneously
from
 all

and
Nt transmitter antennas, where sa = exp j2(a1)
M

j , 1. As the transmitter RF chain is power constrained to

Es /Ts , each antenna transmits sa with energy Es /Nt . In time
slot two, transmitter antenna
i = a is activated with energy
Es . This is represented as Es ia in Fig. 1, where ia is the
Kronecker delta function. So, in both the time slots, the same

IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. X, JUNE 2015

information symbol a is transmitted and hence the overall

spectral efficiency achieved is log2 (M )/2 bits per channel use
(bpcu). Further, in Fig. 1, the Nr Nt matrix H characterizes
the slow flat Rayleigh fading channel. The individual entries
of H are assumed to be spatially independent and identically
distributed (i.i.d.) complex Gaussian RVs with zero mean and
variance 1 (i.e. CN (0, 1)). At the receiver we consider that the
signal is affected by Nr 1 complex white Gaussian noise nt
during time slots t = 1 and t = 2. Each i.i.d. entries of n1 and
n2 have CN (0, No ). Further, all entries of n1 and n2 are also
mutually independent RVs as they are two different samples
from additive white Gaussian noise process. Assuming that the
channel fading remains the same for two symbol duration, the
Nr 1 received signal yt at the receiver in time slots t = 1
and t = 2 are respectively given as
r
Nt
Es X
sa
hi + n1
(1)
y1 =
Nt i=1
p
y2 = Es ha + n2 ,
(2)
where hi is the ith column of the matrix H. We call this
scheme in which MPSK and SSK are used in time successive
fashion as TSSM. As the noise vectors n1 and n2 are independent of each other and by considering that all MPSK/SSK
symbols are equally likely, the maximum likelihood (ML)
detection criterion for TSSM is given as
a
= arg max f (y1 |a, H)f (y2 |a, H)

(3a)

a{1,...,Nt }

= arg max (No )2Nr exp (Da /No )

(3b)

a{1,...,Nt }

= arg min {Da },

(3c)

a{1,...,Nt }

A. Statistical properties of v1 and v2

In this subsection, by using the assumptions of H, we
investigate the statistical properties of v1 and v2 . We define

T
a 2Nr 1 complex random vector z , (hd )T , (hs )T ,
whereP
()T gives the transpose of a vector, hd , ha ha and
Nt
s
hi . The joint PDF of hd and hs can be expressed
h , i=1
from the PDF of z as [6, Eq. (A.23)],
f (hd , hs ) =

2Nr

1
exp z H Z 1 z ,
detZ

(7)

where ()H is the conjugate transpose of a vector, Z is the

2Nr 2Nr covariance matrix of the zero mean random vector
z which is represented as [6, Eq. (A.15)] Z = E[zz H ], with
E[] denoting the expectation operator. The determinant of Z
is given as detZ. Now we define the matrix Z in the following
form
 1

3
Z,
,
(8)
2 4
where k is an Nr Nr sub-matrix of Z. By using the
definitions of hd , hs and taking hr,c as the entry of H at
rth row and cth column, we have the entries kr,c of k as

where f (yt |a, H) is the probability density function (PDF) of

yt , conditioned on a and H. We define Da as

2
r
Nt

2


p
Es X




Da , y1
sa
hi + y2 Es ha , (4)

Nt i=1
where || || represents the Frobenius norm of a vector and Da
is used as the detection metric for TSSM. In the following
section, we derive the average union bound of the BER of
TSSM, based on which asymptotic analysis is performed.
III. P ERFORMANCE A NALYSIS
The average union bound of the BER of TSSM is expressed
based on PEP [5] as
PeT SSM

where Q() is the standard Gaussian Q-function [5, Eq. (4.1)]

Es
. The
and we define signal-to-noise ratio (SNR) , N
o
Euclidean distance between two MPSK symbols sa and sa
is given by daa , |sa sa |, where | | represents the absolute
value of a complex number.
define the RVs v1 ,
P We further
2
Nt
hi . The statistical properties
||ha ha ||2 and v2 , i=1
of v1 and v2 are discussed in the following subsection.

Nt X
Nt
X
1

P (a a
)n(a, a
),
Nt log2 (Nt ) a=1

(5)

a6=a
=1

where P (a a
) is the PEP of symbol a being detected as a

and n(a, a
) is the number of bits in error between the symbols
a and a
. By using (3c), the PEP of TSSM conditioned on H
is derived as
P (a a
|H) = P (Da > Da |H)
s 
!
d2aa

v1 +
v2
,
=Q
2
Nt

(6)

1r,c = E[(hr,a hr,a ) (hc,a hc,a ) ] = 2rc

(9)
" N
!
#
t
X

2r,c = E
hr,i (hc,a hc,a ) = rc rc = 0 (10)
i=1

3r,c = E (hr,a hr,a )

Nt
X

!
hc,i

= rc rc = 0 (11)

i=1

4r,c =

Nt X
Nt
X

i1 =1i2 =1

Nt X
Nt

 X
rc i1 i2 =Nt rc , (12)
E hr,i1 hc,i2 =
i1 =1i2 =1

where () is the complex conjugate of a complex number and

1 (r, c) Nr . In (9)-(12), we do not consider the equality
of a and a
as the condition a 6= a
is enforced in (5). Based
on (9)-(12), we rewrite (8) as


2INr
0
Z,
,
(13)
0
Nt INr
where INr and 0 are Nr Nr identity and zero matrices
respectively. Using (13) in (7) and after few algebraic manipulations we get



1
1
d
s
d 2

f (h , h ) =
exp h
2
( 2)2Nr


1
1
s 2

exp
kh k
Nt
( Nt )2Nr
= f (hd )f (hs ),

(14)

MAHESWARAN AND SELVARAJ : TIME SUCCESSIVE SSK-MPSK: A SYSTEM MODEL TO ACHIEVE TRANSMIT DIVERSITY

where f (hd ) and f (hs ) represent the PDFs of the complex

Gaussian random vectors hd and hs . From (14), we can infer
that the random vectors hd and hs are independent of each
other. As v1 and v2 are functions of hd and hs respectively, we
conclude that they are also independent. Now, the characteristic functions of v1 and v2 can be derived as [7, Eq. (2-1-109)]
v1 (j1 ) = (1j21 )Nr and v2 (j2 ) = (1jNt 2 )Nr .
We can write the joint PDF of v1 and v2 by taking the inverse
transform of their characteristic functions as




Nr 1
v1 v Nr 1 exp v2
v
exp 2 2
Nt
, (15)
p(v1 )p(v2 ) = 1 Nr
Nr
2 (Nr )
Nt (Nr )
where () is the gamma function defined in [7, Eq. (2-1-111)].
As we can see from (15), v1 and v2 are independent chisquare RVs with 2Nr degrees of freedom [7, Eq. (2-1-110)].
We observe that in (6), the Q-function has two independent
d2
terms v1 and Nata v2 as its argument. From this, we infer that
a transmit diversity order of two is achieved.
B. PEP of TSSM
We derive the PEP of
TSSM in this subsection. Based

on
d2
[8,Remark 2], v1 and Nata v2 can be modeled as Nr , 12 and
Nr , d21
respectively. Here, (, ) denotes the Gamma
a
a
distribution with shape parameter and rate parameter . If
d2
we define vaa , v1 + Nata v2 , by using (6), we can obtain the
p
PEP as P (a a
) = E[Q( vaa /2)], where the expectation
is taken with respect to vaa . Now, using [5, Eq. (4.2)] and [8,
Lemma 1], we get P (a a
) as

C. Asymptotic Analysis of BER

In order to know the diversity order of TSSM scheme, in
this subsection we derive the asymptotic BER expression of
TSSM in the higher SNR region. In (18), >> 1 in the higher
SNR region and sin2 1 in the interval [0, /2]. When
d2aa >> 1, we can approximate (18) as
2Nr
P (a a
)

d2aa
8

Nr Z/2
sin4Nr d

(19)

1
2Nr

1
2

d2aa
8

Nr

(4Nr 1)!!
(4Nr )!!

!
,

(20)

where we have obtained (20) from (19) by using [10,

Eq. (3.621.3)]. In (20), (4Nr 1)!! gives the product of all
odd integers between (4Nr 1) and 1, whereas (4Nr )!! gives
the product of all even integers between (4Nr ) and 1. We get
the asymptotic BER expression of TSSM by using (20) in (5)
as
Nt X
Nt  2 Nr
daa
2Nr (4Nr 1)!! X
T SSM
Pasy
=
n(a, a
).
2Nt log2 (Nt )(4Nr )!! a=1
8
a6=a
=1
(21)
As we can observe from (21), we get a diversity order of
2Nr in TSSM scheme, where 2 is due to transmit diversity
and Nr is due to receive diversity.

IV. S IMULATION R ESULTS AND D ISCUSSIONS

In this section we compare the BER of TSSM with TOSDSSK, CSTSK, MPSK and SSK using the assumptions of
channel matrix H discussed in Section II. We perform Monte
/2Z


Carlo simulation and compare the BER obtained in simulation


Z
N
r
2d2aa
vaa
vaa
2Nr 1
with analytical results. The BER plots of TSSM for 1 bpcu
P (a a
) =
va
exp

a
(2Nr )
4 sin2 d2aa
and 2 bpcu are shown in Fig. 2 and Fig. 3 respectively. We
0 0
T SSM




for TSSM. In the
also plot PeT SSM bound of (5) and Pasy
1
1
comparison
of
TOSD-SSK
system,
we
use
N
= 2 and Nt = 4
1 F1 Nr ; 2Nr ;

v
dv
d,
t
a
a
a
a
d2aa
2
transmitter antenna for 1 bpcu and 2 bpcu plots respectively.
(16)
In case of CSTSK, CSTSK(2,2,2,2) [3] system with BPSK and
where 1 F1 (; ; ) is the confluent hypergeometric function [9, CSTSK(2,2,2,4) system with QPSK are respectively used for 1
Sec. 9.2]. The inner integral in (16) can be simplified using bpcu and 2 bpcu plots. In single transmitter antenna (Nt = 1)
MPSK system, we use BPSK (1 bpcu) and QPSK (2 bpcu)
[10, Eq. (7.621.4)] as


transmission for comparison. In SSK system, we use Nt = 2
k
2Nr
,
(17) and Nt = 4 transmitter antenna system respectively for 1 and
(2Nr ) (s)
F Nr , 2Nr ; 2Nr ;
s
2 bpcu plots. Further, the BER of two time slot MPSK and



1
where in (17), we have defined s , 4 sin2 + d2 , k , SSK schemes is also compared. That is, in Fig. 1, instead of
a
a


using MPSK in time slot one and SSK in time slot two, we use
1
21 and F (, ; ; ) is the hypergeometric function [9, MPSK or SSK in both the time slots for the same information
d2a
a
Sec. 8.3]. Now by using [9, Eq. (8.21)] in (17) and substituting symbol transmission. Moreover, the total power constraint of
the simplified expression in (16), we get the PEP of TSSM as the transmitter discussed in Section II is still considered in
two time slot MPSK. Based on the assumption that the channel
Nr
Z/2
1
d2aa
fading remains the same for two symbol duration, the detection
2
P (a a
) =
1+
sin

4
metric for two time slot MPSK and SSK used in the simulation
0
can be derived respectively as

Nr

2
2
r
r
2



1 + sin
d.
(18)
Es
Es


M P SK
s
s
2
sa h + y2
sa h (22)
Da
, y1



Nt
Nt
Closed form expression for (18) is given in [5, Eq. (5A.58)]

2
2
p
p
and its supporting equations. Using (18) in (5), we get the




(23)
DaSSK , y1 Es ha + y2 Es ha .
bound of PeT SSM .

Fig. 2. BER performance of various 1 bpcu schemes versus SNR (dB) with
Nr = 2.

IEEE COMMUNICATIONS LETTERS, VOL. XX, NO. X, JUNE 2015

1 bpcu and 2 bpcu schemes. Moreover the asymptotic BER

T SSM
Pasy
of TSSM derived in (21) becomes tighter in the
T SSM
higher SNR region. We can observe from (21) that Pasy
is
also a function of the inverse of daa which is the Euclidean
distance between two MPSK constellation points sa and sa .
The minimum distance among any two

unit energy MPSK

and as the number of

constellation points is dmin = 2 sin M
constellation points M increases, dmin decreases. The double
T SSM
summation in Pasy
is dominated by (d2min )Nr terms. For
the approximation of P (a a
) given in (20) to be closer
to (18), we need >> d2
min . From this we infer that for
increasing spectral efficiency, SNR needed for (21) to be closer
to the bound of PeT SSM given by (5) and to the BER obtained
in Monte Carlo simulation also increases. This can be observed
from Fig. 2 and Fig. 3.
The transmit diversity order of 2 is achieved for any non
zero daa as we can observe from (18). From this we infer
that instead of MPSK, if we use M -ary quadrature amplitude
modulation (MQAM), we can still achieve the same transmit
diversity as the independence of v1 and v2 does not depend
on the constellation points used in first time slot.
V. C ONCLUSION
We propose a two time slot scheme named TSSM in this
letter which achieves a transmit diversity order of 2. For this
scheme we derive the optimal detection criterion, average
union bound of the BER from PEP. Further, we derive the
asymptotic expression of the BER to show the diversity order
of TSSM. Simulation results show that the derived union
bound of the BER is tight in the higher SNR region. Moreover
the TSSM scheme performs better compared to TOSD-SSK
and CSTSK. Due to its transmit diversity advantage, TSSM
performs considerably better than MPSK, SSK, two time slot
MPSK and two time slot SSK schemes. We infer that the
transmit diversity can also be achieved if we use MQAM
instead of MPSK in time slot one of TSSM.
R EFERENCES

Fig. 3. BER performance of various 2 bpcu schemes versus SNR (dB) with
Nr = 2.

The information symbol a

is detected by using (22) or (23)
in (3c). For 1 bpcu plots in Fig. 2, we use two time slot MPSK
modulation with M = 4 and two time slot SSK modulation
with Nt = M = 4. For 2 bpcu plots in Fig. 3, we use Nt =
M = 16. In all 1 bpcu and 2 bpcu plots, Nr = 2 antenna
receiver is considered.
From Fig. 2 and Fig. 3, we observe that TSSM performs 6
dB and 3 dB better than TOSD-SSK at BER = 103 for 1
bpcu and 2 bpcu schemes respectively. We also observe minor
performance gain in TSSM over CSTSK for both 1 bpcu and 2
bpcu schemes. Further we note that in both 1 bpcu and 2 bpcu
schemes, TSSM performs better than two time slot MPSK, two
time slot SSK, conventional MPSK and SSK schemes due to
its transmit diversity advantage.
We observe that the average union bound of PeT SSM given
by (5) is also tight for SNR greater than 7.5 dB for both

[1] J. Jeganathan, A. Ghrayeb, L. Szczecinski, and A. Ceron, Space

shift keying modulation for MIMO channels, IEEE Trans. Wireless
Commun., vol. 8, no. 7, pp. 36923703, Jul. 2009.
[2] M. D. Renzo and H. Haas, Space shift keying (SSK-) MIMO over
correlated rician fading channels: Performance analysis and a new
method for transmit-diversity, IEEE Trans. Commun., vol. 59, no. 1,
pp. 116129, Jan. 2011.
[3] S. Sugiura, S. Chen, and L. Hanzo, Coherent and differential spacetime shift keying: A dispersion matrix approach, IEEE Trans. Commun.,
vol. 58, no. 11, pp. 32193230, Nov. 2010.
Aygolu, E. Panayirci, and H. V. Poor, Space-time block
[4] E. Basar, U.
coded spatial modulation, IEEE Trans. Commun., vol. 59, no. 3, pp.
823832, Mar. 2011.
[5] M. Simon and M. Alouini, Digital Communication over Fading Channels, 2nd ed. Wiley, 2005.
[6] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.
Cambridge University Press, 2005.
[7] J. Proakis, Digital Communications, 4th ed. McGraw-Hill, 2001.
[8] J. Bausch, On the efficient calculation of a linear combination of chisquare random variables with an application in counting string vacua,
Journal of Physics A: Mathematical and Theoretical, vol. 46, no. 50, p.
505202, Nov. 2013.
[9] L. Andrews, Special functions for engineers and applied mathematicians. Macmillan, 1985.
[10] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
Products, 7th ed. Academic Press, 2007.