Vous êtes sur la page 1sur 5

Performance Analysis of Indoor MIMO Visible

Light Communication Systems

Ngoc-Anh Tran, Duy A. Luong, Truong C. Thang, and Anh T. Pham
Computer Communications Lab., University of Aizu
Aizuwakamatsu, Fukushima, Japan 965-8580
Emaill:{sI180247.m5172106.thang. pham}@u-aizu.ac.jp

Abstract-In this paper, we study the performance of Repeti

stemmed from multi path propagation, which both significantly

degrade the system performance. It is important to note that
in current and foreseen lighting technology, diffused system
have been being essential for indoor scenarios [6].
In order to provide sufficient illumination, multiple LEDs
arrays are usually deployed, thus the utilization of multiple
input-multiple-output (MIMO) systems in VLC is ready to
be realized. In optical wireless communications (OWC) with
outdoor scenarios, the use of MIMO systems can substantially
improve the system performance by providing the spatial
diversity gain [7]. Nevertheless, in indoor scenarios the fading
effects caused atmospheric turbulence does not present, thus
the use of classical MIMO transmission scheme of repetition
code (RC) can only provide a minor diversity gain. On the
other aspect, by means of spatial multiplexing gain, two other
well-known MIMO schemes Spatial Modulation (SM) and
Spatial Multiplexing (SMP) can offer substantial performance
improvement [8]-[10]. In VLC, SM and SMP are expected to
be an effective and low-complexity solution for realizing the
benefit of MIMO technique [II].
In this regard, in [10] the BER performance of three above
MIMO techniques are compared in various scenarios including
different transmitter spacing, link blockage and imbalance
power. In each case, the BER is derived in terms of electrical
signal-to-noise ratio and different level of spectral efficiencies.
However, the NLOS was not considered in this study. As a
result, the effects of reflection and lSI phenomenon is ignored
in the calculations. Moreover, the consideration of a quite
small transmitter spacing (ranging from 0.2 m to 0.7 m)
together with the assumption of signal independent receiver
noise make the system far from the practical lighting systems.
These factors raise the channel gain matrix to the region of
10-4 and, thus, do not reflect the high path loss property of
an actual indoor VLC link.
Therefore, in this paper, we apply two MIMO techniques
SMP and RC to NLOS VLC for a 5 m x 5 m x 3 m room. The
receiver noise is modeled as a AW GN noise taking to account
thermal noise, signal-dependent shot noise and lSI power. It is
worth noting that the BER strongly depends on the location of
the receiver, hence it is important to examine the distribution
of the achievable BER within the room. Additionally, the
normalized area in which the system performance falls below a
required BER level is considered in order to find the optimum
transmitter separation. We also compare SMP and RC in terms

tion Code (RC) and Spatial Multiplexing (SMP) applied to Visible

Light Communication (VLC) in indoor scenarios with non-direct
line-of-sight (LOS) characteristics. Considering a VLC MIMO

system with 4 transmitters within a typical 5 m x 5 m x 3 m

room, by means of simulation, we investigate the distribution of
Bit-error-rate (BER) when the receiver moves around the room.
For various transmitter separations, the normalized area at which
the system performance does not satisfy a required BER level is
illustrated. The optimal values of transmitter separation for RC

and SMP are 1.5 m and 3.5 m, respectively. Comparing between

RC and SMP, the results show that RC outperforms SMP
at low spectral efficiency. However, at relatively high spectral
efficiency RC does not perform well since it requires a very high
constellation size, while SMP with its spatial multiplexing gain
can offer a much better performance.



Over the past decade or so, solid state lighting technology

using light emitting diodes (LEDs) has emerged as a promising
illumination solution for both commercial and home uses.
It is expected that LED lighting will progressively replace
conventional phosphorescent light bulbs and incandescent
lamps due to its advantages of lower power consumption,
longer lifetime, non-toxic and safety from eavesdropper [I],
[2]. In addition, thanks to the fast-switching property, LED
can also be used as a transmitter for data transmission, and
this technology is referred to as Visible light communications
(VLC). Especially, in typical indoor scenarios, e.g. office or
home, VLC is expected as a strong contender to WiFi for
home networking and a reliable successor in case of the use
of WiFi is concerned.
The indoor VLC however has to overcome several major
challenges, including limited bandwidth of white LEDs, severe
path loss, multi-path dispersion and unavoidable background
light noise [3], [4]. Moreover, performance of VLC systems
strongly depends on the link configurations. In general, there
are two main types of link configurations which are non-direct
line-of-sight (NLOS), namely diffused link, and direct line-of
sight (LOS). LOS systems with their directionality are able to
offer a high bit-rate up to several Gbps [5], however the system
is sensitive to the link blockage. On the other hand, in diffused
VLC systems the signal propagates from the transmitter to
the receiver by several paths making the system more robust
to link blockage and moving objects. This comes with the
expenses of higher path loss and inter symbol interference

978-1-4799-5051-5/14/$31.00 2014 IEEE


of the BER performance for different efficiency levels. The

remainder of the paper is organized as follows: Section II
introduce the system model with VLC link profile and the
receiver noise. Section III presents SMP and RC principle
and then BER performance is derived. Section IV results are
shown. Finally, section V concludes the paper.


Continuously, we assume the surface reflector coefficient

p = 0.8 corresponding to the plaster surface. Channel gain
corresponding to each small reflective area dArej is calculated



A. VLC Channel Model

L2 pdArej cos(a) cos(jJ) cosm(r)Ts(1)r)


cos(1)r ) , 0 S 1j;r S we
0, 1Pr > We

B. Receiver Noises with lSI

In our system, we consider a room with the dimension of

5m x 5m x 3m, thus the longest path length difference between
the longest reflected and directed path can exceed 3 meters.
In this case, the delay becomes comparable to the symbol
duration at the bit-rate of interest. Therefore, lSI phenomenon
should be taken into account. Consequently, the receiver noise
variance is treated as a combination of thermal noise, shot
noise and lSI power. The received lSI power can be given as

receivers (Nr = 4) is described in Fig.l. Four LED arrays are

hanged on the ceiling at the height of 2.50 m from the ground
and their separation will be investigated for the optimal value
in Section IV. Four receivers are located at the table height
z = 0.75 m and aligned 2 by 2 on 10 cm x 10 cm plane
which is small enough to enable mobility and possibility of
integration into usual end-to-end devices, such as a router or
a laptop computer.
The received signal vector can be expressed as
x +

where D1, D2 , a and jJ are described in the right side of Fig.


We consider a MIMO indoor VLC system with intensity

modulation and direct detection (IMIDD) where Nt LEDs
arrays, as transmitters, and N r photo-detectors, as receivers,
are deployed. As each LED chip is small and close enough to
others in same array, we can assume that all optical signals
coming from a same source can be synchronized and well
represented by one point at the center of that LEDs array.
Since the non-direct LOS link is assumed, large values of both
transmitter semiangle <1:>1/2 (at half power) and field-of-view
(FOV) semiangle W 1/2 of the receiver are chosen, which are
60 and 75 degrees, respectively. For the sake of convenient
calculation, we only consider up to the first order reflection.
The VLC channel with 4 transmitters (Nt = 4) and 4

y =

00 h(t)



where h(t) is the inverse Fourier transform of Hand

denotes the convolution operator. The shot noise variance CTshot
and thermal noise variance CTthermal are given in [1]. Finally,
the noise power spectral density is summarized by
No =

( CT ;hermal

+ CT hot +

r2 Pii)/ E,


where r is electrical optical conversion efficiency of photo

detector and E is the bandwidth.


where x = [ Xl ... XN, ] T is the transmitted signal vector where

[.] T denotes transpose operator. n is total noise vector. H
In this section, we analyze the BER of indoor VLC chan
indicates the channel transmission matrix and can be given
considering two MIMO techniques, RC and SMP. For
the selection of modulation scheme,unipolar /vI level pulse
amplitude modulation (M-PAM) is chosen thanks to its suit
(2) ability with IMIDD channel and higher bandwidth efficiency
compared to other modulation scheme such as OOK, PPM or
PWM. The intensity levels in M-PAM is given as [10]
where Hij represents the channel gain on VLC link from the
2 Is
ith transmitter to the jth receiver.
1m = /vI 1 m wIth m = 0, 1, ... , ( /vI - 1),
Each component in the channel matrix Hcan be given as



Hd + dHrej,

where Is is mean emitted optical power.

RC is the simplest MIMO techniques in which all trans
mitters simultaneously transmit the same signal. RC together
with M-PAM can achieve a bandwidth efficiency of log2 (M)
bits/s/Hz. The BER can be calculated as [3]


where the DC gain on the directed path Hd is described in [I]



{(;::-lA cosm()Ts(1j;) cos(1))

0, 1j; > We

, 1) We.




A and 1 (1)) are the physical area of the photodetector

and gain of optical filter, respectively. m is the order of
Lambertian emission and related to <1:>1/2 as m = I n (
co s

2(M - 1 )
-. Mlog2 M. Q M




Here, SINR is electrical signal to interference and noise ratio

after being combined from Nr received signals by applying






" :









5m --------''-''

Fig. l: Geometrically model of Indoor VLC channel with reflection

maximum ratio combining (MRC). With a Nr
we obtain SINR as given in [10]


= N


:l 8

( )

f; hij



I V.



Some major parameters used in our calculation are listed in

Table I.


TABLE I: System parameters


where PTX = ( r1hx)2 is the average emitted electrical

power where 1 is symbol duration.

LEDs chip power

Total photo-detector area
Refractive index
OlE conversion efficiency
Background current

In the case of SMP, each transmitter transmits independent

data stream simultaneously. In contrast to RC, SMP offers
the spatial multiplexing gain. Therefore, to provide the same
spectral efficiency SMP needs much lower constellation size
than RC. Precisely, SMP achieves the bandwidth efficiency of
R = Nt *log2 (M) bits/s/Hz with M-PAM modulation scheme.
However, to gain those high values, SMP requires sufficient
low correlation among active data streams. For SMP, pairwise
error probability (PEP) is the probability that the receiver
erroneously detects the signal from other transmitters. The
PEP in which Xm(2) is erroneously detected while Xm(l) is
actually transmitted can be derived as [10]



z= Adct



10 mW
1 cm 2
5.1 X 10-3 A

A. BER Distribution

Figure 2 illustrates the distribution of BER when the re

ceiver moves within the room where the transmitter separation
dTX = 2.5 m, the bandwidth B = 100 MHz and the efficiency
R = 4 bits/s/Hz are selected. In case of RC, when the receiver
locates right under the source BER becomes very low and the
worst BER appears at the corner of room. Regarding SMP,
it offers wide range of BER value due to the sensitivity to
different transmitter receiver alignments.
B. Transceiver Optimal Positions

In the previous discussion, we chose the position of LED

arrays on purpose of having an ideal alignment of LEDs-to
LEDs and LEDs-to-walls. As expected, the BER increases
severely when the receiver locates near walls and comers of
the room. This configuration needs relatively high power to ob
tain uniform VLC link quality. Thus, the relationship between
the normalized area with BER2>: S'J and dTX is investigated
in Fig 3. Here S'J denotes the desired BER level and it can
be 10-3 with the use of FEC or 10-4 as the requirement of
a typical wireless link. The transmitted electrical power from
each LEDs array is assumed to be 13 W and the bandwidth
efficiency is R = 4 bits/s/Hz. The normalized area with
BER 2>: 10-3, where the BER is considered to be unsatisfied,
occupies at least 5% entire room's area with SMP. We can see
that RC can completely satisfy the requirement of S'J = 10-3

where II.IIF denotes Frobenius norm. Then, the BER is well
approximated by

where dH(.,.) indicates the Hamming distance of two bit

assignments corresponding to two concerned signal vectors.
Thus, dH(bm(1),bm(2) is the number of bit errors when pair
wise error happens between Xm(1) and Xm(2).




a: 1E-10
w 1E-15





o. "






Pave for each array (W)








Fig. 4: BER versus Pave at the comer of room with dTXnc

1.5 m and dTXSMP = 1.0 m with the same actual bit-rate.

needs much higher constellation size, in this case lvI = 256,

while SMP with its spatial multiplexing gain only have to work
on 4-PAM to achieve the same bit-rate.






.... 'V..



In this paper, we have investigated the performance of VLC

systems with two MIMO techniques RC and SMP in non
direct LOS environment. The results showed how the BER
variates within the room and therefore the optimal transmitter
separation was found. Regarding MIMO techniques, it is seen
that RC can only perform well is the case of low efficiency. On
the other hand, SMP with its spatial multiplexing gain is shown
to be a promising solution to provide high efficiency with low
complexity in the presence of high data-rate demandings.

4 at B

Fig. 2: BER distribution with RC & SMP, R

100MHz dTX = 2.5 m Pave = 9 W.






[I] Komine, T; Nakagawa, M., 'Fundamental analysis for visible-light
communication system using LED lights," Consumer Electronics, IEEE
Transactions on , vo1.50, no.I, pp.100,I07, Feb 2004.
[2] Tanaka, Y; Komine, T; Haruyama, S.; Nakagawa, M., "Indoor Visible
Light Data Transmission System Utilizing White LED Lights," TEICE
Trans. Commun., vol. E86-B, no. 8, Aug. 2003, pp. 244054.
[3] Grubor, J.; Randel, S.; Langer, K.-D.; Walewski, J.w., "Broadband In
formation Broadcasting Using LED-Based Interior Lighting," Lightwave
Technology, Joumal of , vo1.26, no.24, pp.3883,3892, Dec.15, 2008
[4] Oswaldo Gonzalez (2012). Multiple-Input Multiple-Output (MIMO)
Optical Wireless Communications, Optical Communication, Dr. Narot
tam Das (Ed.), ISBN: 978-953-51-0784-2, InTech, DOl: 10.5772/46079.
[5] Akbulut, M.; Chen, C. H.; Hargis, M. c.; Weiner, A.M.; Melloch, M.R.;
Woodall, J.M., "Digital communications above I Gb/s using 890-nm
surface-emitting light-emitting diodes," Photonics Technology Letters,
TEEE , vo1.13, no.l, pp.85,87, Jan. 2001
[6] C. G. Lee (2011). Visible Light Communication, Advanced
Trends in Wireless Communications.
Dr. Mutamed Khatib
[7] Navidpour, S.M.; Uysal, M.; Kavehrad, Mohsen, "BER Performance
of Free-Space Optical Transmission with Spatial Diversity," Wireless
Communications, IEEE Transactions on , vo1.6, no.8, pp.2813,2819,
August 2007 doi: 1O.1109/TWC.2007.06109

dTX Transmitter separation

Fig. 3: The normalized area with BER:;:, S'J versus the separa
tion of transmitter

when dTX = 3.5 m. The optimal value of dTX in RC and

SMP is recognized as 1.5 m and 3.5 m, respectively.
Figure 3 also shows that RC outperforms to SMP at low
efficiency R = 4 bits/s/Hz. That result is verified when we
examine a certain position. In figure 4, we plot the BER when
the receiver is located at the comer of room with two different
efficiencies but the same actual bit-rate. The result shows that
to achieve BER below 10-3, SMP needs more transmitted
power than RC when R = 4 bits/s/Hz. However, at a higher
efficiency, i.e., R = 8 bits/s/Hz, SMP shows much better
performance than RC. This fact can be explained that RC


[8] Mesleh, R; Mehmood, R; Elgala, H.; Haas, H., "Indoor MIMO Optical
Wireless Communication Using Spatial Modulation," Communications
(TCC), 2010 IEEE Intemational Conference on , vol., no., pp.l,5, 23-27
May 2010 doi: 1O.1109/ICC.201O.5502062
[9] Fath, T.; Haas, H.; Di Renzo, M.; Mesleh, R., "Spatial Modulatiou
Applied to Optical Wireless Commuuicatious in Indoor LOS Environ
ments," Global Telecommunications Conference (GLOBECOM 2011),
2011 IEEE , vol., no., pp.l,5, 5-9 Dec. 2011


[10] Fath, T.; Haas, H., "Performance Comparison of MIMO Techniques for
Optical Wireless Communications in Indoor Environments," Communi
cations, IEEE Transactions on , vo1.61, no.2, pp.733,742, February 2013
[11] Di Renzo, M.; Haas, H.; Ghrayeb, A.; Sugiura, S.; Hanzo, L., "Spatial
Modulation for Generalized MIMO: Challenges, Opportunities, and
Implementation," Proceedings of the IEEE , vol.102, no.1, pp.56,103,
Jan. 2014 doi: 10.1109/JPROC.2013.2287851