www.ietdl.

org

Published in IET Communications

Received on 22nd July 2009
Revised on 21st February 2010
doi: 10.1049/iet-com.2009.0465

ISSN 1751-8628

Analysis of power control for indoor optical

wireless code-division multiple access
networks using on – off keying and binary
pulse position modulation
S. Khazraei M.R. Pakravan A. Aminzadeh-Gohari
Optical Communications and Data Networks Research Lab, Advanced Communications Research Institute, Electrical
Engineering Department, Sharif University of Technology, Tehran 11365-9363, Iran
E-mail: [email protected]

Abstract: Wireless infrared optical code-division multiple access (W-OCDMA) is a new developing technique with
many useful applications. Noting the limitation on power consumption and eye-safety requirements, wireless
optical systems are power limited. Therefore control and efficient use of optical power is a key issue in
analysis and design of these systems. Also, multi-user interference is the major source of impairment in these
systems and power control is required to control and reduce this interference. Power control and the
inevitable errors in its algorithms play an important role in design and implementation of these systems. In
this article the authors study the uplink performance of W-OCDMA networks employing on – off keying (OOK)
and binary pulse position modulation (BPPM) schemes without any power control algorithm. The performance
improvement as a result of using perfect power control is calculated. Then, the impact of imperfect power
control that is the result of channel estimation error is analysed. The results clearly illustrate that deploying a
proper and accurate power control algorithm can increase network capacity and reduce network average
power consumption. The authors show that systems with non-ideal power control still perform considerably
better than no-power controlled systems.

1 Introduction multiple access (CDMA)-based infrared network prototype

considering the practical aspects of system implementation.
Indoor wireless infrared communication systems have been The results of various stages of the proposed wireless optical
the subject of considerable research and development CDMA (OCDMA) LAN strongly indicate the viability and
activities during recent years [1 – 21]. It is believed that the importance of such networks in certain applications [13].
wireless infrared networks receive more attention where
security is an issue or using radio band would not be In this paper, we consider the performance analysis for the
economical or practical because of bandwidth regulation or uplink of indoor wireless infrared networks employing optical
electromagnetic interferences [1, 2]. orthogonal codes (OOC) [22]. It has been shown that using
OOCs can provide a suitable modulation and multiple access
Analysis of various aspects of wireless indoor infrared algorithm in fibre-optic communication systems [22, 23].
communication systems and local area networks (LAN) has However, applying this technique to wireless optical
been presented in the literature [3–8]. Applying spread communication raises many important and interesting
spectrum techniques to infrared networks has received more questions [10 – 17], some of which are analysed in this
attention recently [9–17]; Ghaffari et al. studied the digital paper. The scenario of interest for propagation channel in
design concepts and structure in a typical code-division this paper is a diffused system [1] that connects several

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1919
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

portable nodes to a central fixed node or base station (BS) [8] 2 Infrared CDMA system model
which provides backbone connectivity through the wired
infrastructure of the building. The indoor infrared system is assumed to be deployed using a
cellular architecture that consists of a few base stations placed
at the centre of each cell. Base stations act as a bridge between
The importance of power control in wireless infrared the users in wireless environment and the wired backbone
optical CDMA (W-OCDMA) networks is because of
infrastructure. Transceivers use OCC [22] as the multiple
three important facts. First and foremost, is the multi-user access method. The system is assumed to use a set of OOC
interference at the receiver, which can dramatically affect sequences with length Lc and weight K having minimum
the system performance. High dynamic range in received
auto- and cross-correlation (la ¼ lc ¼1) to minimise
power just like traditional wireless CDMA networks interference [23]. A unique code should be assigned to
necessitates the implementation of appropriate power each user for transmission. Two modulation schemes are
control mechanisms in the transmitters. Second, the
considered: OOK and BPPM. In OOK the user sends its
amount and pattern of optical power radiated from a unique OOC signature for data bit ‘1’ and sends nothing
transmitter should meet eye-safety standards. That requires for data bit ‘0’. In BPPM the duration of each chip, is
careful control of emitted optical power from each
divided into two equal sub-chip parts. Transmission of K
transmitter [1]. Last but not least, power limitations of marked sub-chips, all in the first part of the divided chip
mobile wireless systems puts severe restrictions on their times, represents the data bit ‘0’ whereas the data bit ‘1’ is
possible radiated optical power levels. The objective of this
represented by choosing the second K sub-chip times [25].
paper is to analyse the power control issues in the uplink of This pattern is therefore called chip-level BPPM that has
an OOC-based indoor diffused infrared network. The more uniform power distribution in comparison with bit-level
analysis has been done in two different scenarios to
BPPM. For both cases the number of associated users in each
demonstrate the important effects of power control cell (M + 1) is limited to Nmax = (Lc − 1)/K (K − 1) [22].
implementation. In the first scenario, the system is analysed Furthermore, if the irradiance of the received signal is Ir (W/
assuming no power control mechanism in the uplink
cm2), the average received power can be expressed by
direction. Results demonstrate the system deficiencies and Pr = (K /2Lc )ir Ad , where Ad is the area of detector [10].
the need for a power control algorithm for both OOK and
binary pulse position modulation (BPPM) schemes. In the
second scenario, a power controlled system is analysed in
which all users estimate their uplink channel path loss and 2.1 Channel model
correspondingly send sufficient uplink power to ensure
equal received power level from all transmitters at the base Diffuse infrared links are known to be the most robust link
station. However, the imperfect channel estimation by users configuration in wireless infrared communication systems
during the power control process causes an undesired error [1]. However, they provide a challenging environment for
in the power control algorithm. This makes it impossible to system designers because of their high path loss and long
ensure exact parity of received powers at the base station; delay spreads [3 – 6]. One of the most important parameters
the limitations caused by this error on the performance is that affects the performance of infrared systems is the
calculated and demonstrated as well. Furthermore, we have channel path loss that is the gain of the channel transfer
compared the performance of the network using on – off function H( f ) at f ¼ 0 which is used to evaluate the
keying (OOK) and BPPM modulation schemes each of relation between transmitted and received average power to
which is shown to have some advantages and drawbacks [24]; be Pr = H0 Pt . To compute the path loss of a diffuse
for instance, employing OOK improves the bandwidth channel, as a first-order approximation, only the first
efficiency of the system (i.e. it can achieve higher bit rates) bounce off the ceiling can be considered [1].
while BPPM improves the power efficiency, in other words
the network is more robust regarding the amount of To simplify and evaluate the channel path loss, it is
received noise and it can perform better when the assumed that both transmitter and receiver are pointed
transmission power is low. However, it is concluded that straight upward and transmitter emits a Lambertian
both OOK and BPPM schemes dramatically suffer from pattern. It has been shown in this case that the numerical
near – far problem as stated above. and experimental evaluation of path loss can be found as a
function of r (horizontal separation between transmitter
and receiver) in [1, 3]. To simplify the calculations, H0(r)
The rest of this paper is organised as follows. Section 2 is approximated from the one-bounce curve by a function
describes the indoor infrared CDMA system model −10 log H0 (r) = a0 + a1 r + a2 r2 + a3 r3 + a4 r4 in which
including signalling, channel model and noise sources. a0 – a4 are chosen so that the fitted curve minimises the
Section 3 provides analytically, in both scenarios, the BER mean square difference between the logarithm of main data
performance of the system as well as average power and fitted data. In the fitted curve, which is derived from
consumption. Section 4 derives some numerical the above equation and used for path loss calculations, the
comparisons for two strategies while the conclusions are values for the a0 to a4 coefficients are: 53.555, 1.559,
presented in Section 5. 0.412, 20.04 and 0.0011, respectively [14].

1920 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

2.2 Photon counting characteristics 2.2.3 Photodetector dark current noise:

and noise sources Photodetector’s dark current (id) has been shown to be a
Gaussian process and the mean number of photon count
The infrared medium has its own drawbacks. Random nature produced by this source in t seconds can be expressed by
of number of photons and other noise sources are inevitable md = id t/e where e is the electron charge [1].
in wireless optical systems.
2.2.4 Gaussian thermal noise: When little or no
2.2.1 Photon counting statistical characteristics: ambient light is present, the dominant noise source will be
In infrared communication systems the signal detection is the receiver’s preamplifier noise. In a well-designed
done by a photodetector, having quantum efficiency of h, practical system, this rarely happens and the system is
which is the probability that an incoming photon releases an mostly dominated by ambient light noise.
electron in the photodetector [26]. Therefore the number of
photoelectrons at the output of the photodetector is a
Poisson process which can be approximated by a Gaussian
2.3 Transmitter and receiver
process in many cases [26]. If the intensity of received power Fig. 1 shows the block diagram of transmission and detection
to a photodetector with the quantum efficiency h and the architecture. The system uses OOK or BPPM intensity
area of Ad is given by Ir , the mean number of photoelectron modulation at the transmitter and direct detection at the
count at the output of the photodetector in t seconds will be receiver. Duration of each chip is Tc = Tb /Lc , where Lc is
mr = hAd Ir t/hv, where n is frequency of light and h is the code length and Tb is the bit duration. At the receiver,
Plank’s constant [25]. the desired signal along with the noise and interference
from all other M active users (interferers) is detected. The
2.2.2 Ambient light noise: In many applications, receiver’s structure is assumed to be based on correlation
infrared links operate in the presence of intense infrared detection. It should be noted that one of the system design
and visible background light. Although received objectives here is to design a high-performance uplink
background light can be minimised by optical filtering, it minimising the interference as well as the average power
still adds shot noise, which is usually the limiting noise consumption of the network.
source in a well-designed receiver. Owing to its high
intensity, this shot noise can be modelled as white, 3 Performance analysis of
independent of Gaussian noise [1]. Therefore the mean
number of photon counts produced by ambient background
wireless optical CDMA uplink
noise at the output of photodetector in t seconds will be In this part we derive analytical formulation for the uplink
mb = hAd Ib t/hn, where Ib is the intensity of radiated performance of the introduced wireless optical CDMA
ambient light. Infrared data association (IrDA) links should network for both OOK and BPPM modulations in the
operate at BER of 1026 when the maximum intensity of two presented scenarios. First, when there is no control on
radiated ambient light is limited to Ib ¼ 490 mW/cm2 [27]. the users’ transmitted power and then assuming an

Figure 1 Infrared CDMA system architecture block diagram

a Employing OOK modulation and detection
b Employing chip-level BPPM detection diagram

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1921
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

imperfect power control algorithm defined for transmission. Rj=0 or 1 and s2Rj (the mean and variance of received photon
A simplifying assumption is to consider signals of different count) can be expressed as ( j is the information bit
users to be chip synchronous. This represents a pessimistic transmitted by the desired user)
case and gives an upper bound to the bit error rate (BER)
of the real chip-asynchronous system [22]. Additionally, 
l
the average power consumption of the active users in each Rj = Kmr j + Kmb Kmd + mi
cell is derived in the above cases. i=1 (4)
s2Rj = Rj + K s2T
3.1 Upper bound BER analysis – no
control on transmission power The number of photoelectrons at the output of the
photodetector is the sum of Gaussian variables
3.1.1 OOK modulation scheme: Since we use
(photoelectrons from different sources). Therefore the final
correlation receiver, interference pattern is not important
decision variable is also a Gaussian variable with mean and
and just number of interferers should be taken into
variance equal to the sum of means and variances of each
account. In OOC with minimum auto- and cross-
variable. Decision variable is composed of K pulse positions
correlation, two code words cannot overlap at more than
consisting desired signal, noise and sum of interferences
one pulse position. Therefore the probability that two code
from other users. mr is the mean received photon count of
words overlap at one pulse position is q ¼ K 2/2Lc where
the desired user, mb is the mean photon count of ambient
factor 1/2 accounts for the probability that interfering user
light noise, s2T is the variance of the electron count
has sent ‘one’. Assuming M interfering users, the BER of
produced by thermal circuit noise and md represents
the desired user’s detected information can be expressed
photodetector dark current noise. Introduced equations in
as [28]
Section 2 completely determine mr , mb and md . Furthermore,
M   mi is the mean photon count produced by ith interferer
 M l
PE = q (1 − q)M−l PE (l ) (1) which is a function of ri , path loss and transmission power.
l =0
l
By replacing Pr = (K /2Lc )Ir Ad in mr = h Ad Ir Tb /hnLc ,
where PE(l ) is the probability of error when there exists l we will have mr = 2hPr /KhnRb , mb = h Ad Ib Tc /hn
interferes in the system.
i T 2hPr,i
md = d e c , mi = (5)
If the position vector of l interferers is considered to be r, KhnRb
that is, ri is the distance of ith user from the base station
(placed at the center of the cell), the value of PE(l ) can be Pr,i is the power that base station receives from ith user and
obtained by averaging the conditional probability PE (l |r) can be expressed as Pr,i = H0 (ri )Pt . Rb is the transmitter
with respect to r bit rate and Tc is chip duration in seconds. Notice that Pr,i
 is the average received power. By dividing Pr,i to code
PE (l ) = PE (l |r)fr (r) dr (2) weight (and applying other coefficients), average number of
r photons in every chip is obtained. Finally, the error
probability of detected information can be derived using
Since the position of each user is independent of the others; equations (1) – (5).
dr = li=1 dri and correspondingly fr (r) = li=1 f (ri ),
where f (r) is the probability distribution function (PDF) of
3.1.2 BPPM scheme: For the case of BPPM even the
horizontal separation distance between each user and base
users transmitting information bit ‘0’ should be considered
station. f (r) is the probability of user existence in a thin
in the interference pattern. Consider the number of
ring with thickness of dr. Assuming the distribution of all
interfering pulses in the first and second sub-chips as l0
users over the area of the cell is uniform, f (r) can be easily
2 and l1 . Following the same methodology of (1) and (2) the
expressed by: f (r) = 2r/rcell . Mention that rcell is the
bit error probability can be obtained using
radius of the cell. So that integration of f (r) over r
becomes 1.
 0
M M−l
M!
PE =
As discussed in Section 2.2, using Gaussian approximation l0 =0 l1 =0
l0 !l1 !(M − l0 − l1 )!
for photon detection in [25, 26], PE (l |r) can be written as
× ql0 +l1 (1 − 2q)M−l0 −l1 PE (l0 , l1 ) (6)
1 1
PE (l |r) = PE (l |r, 0) + PE (l |r, 1)
2 2 l0 and l1 users cause interference sending information bit ‘0’
   
Th − R0 R1 − Th and ‘1’ respectively among M users. This follows a
= 0.5 ∗ Q + 0.5 ∗ Q (3) trinomial distribution with parameters M (total number of
sR0 sR1
possible interferers) and q = q1 = q2 = K 2 /2Lc .

1922 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

As illustrated in Fig. 1 the decision about the received data of the optimum threshold for OOK would be a difficult
for BPPM is completed by comparing two decision variables task because of the inaccuracy in estimating the proper
U0 and U1 . Since the probability of transmitting ‘0’ and ‘1’ is signal value in noisy environments. For the case of BPPM
equal, considering symmetry we have no threshold calculation is required at the receiver;
therefore the results do not suffer from possible errors of
PE (l0 , l1 ) = PE (l0 , l1 |0) = PE (l0 , l1 |1) optimum threshold computation. Additionally, (10)
= P(U0 , U1 |l0 , l1,0 ) (7) predicts that the mean of the noise is cancelled when
comparing decision variables in detection process and gives
To finalise the computation, we have to consider the location better performance and robustness in noisy environments.
vectors of interfering users r0 and r1 , where r0,i and r1,j are These advantages of the BPPM system come at the cost of
the position of ith first sub-chip interfering user (of l0) and additional bandwidth requirement of this modulation
jth second sub-chip interfering user (of l1), respectively. scheme.
Therefore the error probability of P(U0 , U1 |l0 , l1,0 ) can
be obtained by averaging P(U0 , U1 |l0 , l1,0 , r0 , r1 ) with 3.2 Upper bound BER analysis – power
respect to r0 and r1 controlled network
Power control is required for all realistic CDMA systems,
PE (l0 , l1 ) = P(U0 , U1 |l0 , l1,0 )
mainly because of what is known as the near – far problem,

that is, users far from the base station experience greater
= P(U0 , U1 |l0 , l1,0 , r0 , r1 ) path loss than users that are near the base station.
r0 , r1
Optimum power control is achieved when all users are
× fr0 (r0 )fr1 (r1 ) dr0 dr1 (8) decoded with the same signal-to-interference ratio (SIR),
that is, the users’ signals all arrive at the base station with
where f (r) is the PDF of users’ distribution in the area which approximately the same power. Otherwise, a user with a
is described in the previous section. Finally we have high SIR dominates the BER performance of the system [29].

P(U0 , U1 |l0 , l1,0 , r0 , r1 ) = P(x . 0|r0 , r1 ) (9) The power control algorithms are generally based on the
channel estimation in portable users or base station, thus if
where x is a Gaussian random variable with mean and there is a perfect power control algorithm with no errors,
variances equal to finding the BER performance is straightforward as
described in [22, 28, 30]. In the ideal uplink power control

l0 0 +l1
l scenario, the transmission power of all users is set to a
mx = −Kmr − mi + mi suitable value to assure the received power of all users at
i=1 i=l0+1 the base station is fixed to a desired predefined value (Pr,d).

l0 0 +l1
l Therefore the user located at the edge of the cell emits the
s2x = Kmr + mi + mi + 2K (mb + md + s2T ) maximum transmission power of all users (Pt,max) which is
i=1 i=l0+1 related to Pr,d by Pr,d = Pt, max H0 (Rcell ).
(10)
Unfortunately, because of large ambient light noise, and
mr , mi , mb and md are determined using (5). However, for other sources of noise or interference, there will be an error
BPPM, chip time is different from what is stated for in estimation that causes power control error (PCE). PCE
OOK; in other words, as Fig. 1 predicts the chip rate of a can be considered to follow a lognormal distribution [29,
BPPM system operating at the same bit rate with OOK, is 31]. The impact of PCE on the system performance should
twice the OOK system. If two systems are considered to therefore be investigated.
have the same chip rate, OOK can achieve twice the bit
rate of BPPM enabled network. In this article, we assume Performance evaluation of a system with PCE is similar to
the same chip rate for both systems, so that both systems previous calculation; in which because of the considered
use equal amount of system resources, for example lognormal distribution of PCE, the received power from
bandwidth. For the case of same bit rates, BPPM needs ith user (Pr,i) is given by
twice bandwidth. If we assume that the uplink direct
channel does not degrade at higher frequencies, both Pr,i = Pr,d exi , where xi  N (0, s2PCE ) (11)
systems can be considered to be equal. However, in reality,
the channel frequency response degrades at higher Therefore the desired received mean photon count of ith user
frequencies and that has an impact on system performance. will be

Furthermore, in BPPM, two decision variables are mr,i = mr,d exi , where xi  N (0, s2PCE ) (12)
compared together, as provided in (8), which leads to an
interesting phenomenon in the system design. Calculation where Pr,d is the desired power received from all users,

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1923
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

including ith user (if there is no estimation error) and mr,d is Similar to (9) the above equation can be reduced using the
the desired mean photon count (for which a proper BER is following form
obtained). Mention that xi is a random variable whose
randomness is because of error in estimation of power in P(U0 , U1 |l0 , l1,0 , X ) = P(y . 0|X , 0) (17)
each receiver and its variation would depend on the update
frequency of the much slower power measurement process. where y is a Gaussian random variable with mean and
Therefore xi has a much slower rate of variation compared variances equal to
to system bit rate.
 

l0 0 +l1
l

3.2.1 OOK modulation scheme: Since interfering my = −K ex0 − e xi + exi mr,d

users are not distinguishable in correlation receiver, we i=0 i=l0 +1

define PE (l |X ) the conditional error probability of PE (l ) if  


l0 0 +l1
l (18)
the PCE for every interfering user (xi) is known to be the s2y = Ke + x0
e +
xi
e xi
mr,d
vector X. Hence PE (l ) can be easily derived as follows i=0 i=l0 +1

 + 2K (mb + md + s2T )
PE (l ) = PE (l |X )fX (X ) dX (13)
X
l Finally, mb , md and s2T can be found directly, using (5) and
l l l =0
xi
mr,d is determined by replacing Pr by Pr,d in(5) as well.
where dX = i=0 dxi and fX (X ) = i=0 f (xi ) = e .

Therefore using the same methodology in (3) and 3.3 Average power consumption
considering Gaussian approximation for photon detection,
Power efficiency is considered to be an important factor in
PE (l |X ) is clearly extracted as follows
designing any energy limited network, because source
1 1 power limitations of mobile wireless systems puts major
PE (l |X ) = PE (l |X , 0) + PE (l |X , 1) restrictions on their possible radiated power levels for long
2 2
    time. Hence, deriving the average power consumption
Th − R0 R1 − Th should be considered for such networks beside the analysis
= 0.5 ∗ Q + 0.5 ∗ Q
s R0 sR1 of their performance.
(14)
3.3.1 Basic structure employing no power
 j and s2R can be expressed as
Finally, the values of R control: In the first scenario, there is no control on
j
transmission power of the users and it is logical to assume
 

l all users consume the same energy to transmit at the same
Rj = Kjex0 + exi mr,d + Kmb + Kmd power (Pt) independently. Using (3) – (5) for OOK
i=1 (15) modulation and equivalently (6) – (10) for BPPM, the only
s2Rj = Rj + K s2T way to acquire desired error probability is to change Pt for
all users; hence, the average power consumption of each
user is equal to the required transmission power to achieve
In this case mb , md and s2T can be found directly, using (5),
the desired BER. Although the error probability for all
whereas mr,d is determined by replacing Pr by Pr,d in (5). It is
users located within the cell area is less (better) than the
required to note that the value of Pr,d should be adjusted by
user(s) located on the edge of the cell (i.e. worst-case user),
the power control algorithm to obtain the desired BER, for
it is mandatory to adjust the transmission power of all users
example, 1026. Furthermore, Pr,d is related and limited by
to attain desired BER for the worst-case user(s). It has to
the maximum possible transmission power of an infrared
be noted that in this case the average power consumption is
source (Pt,max); in other words the user located at the edge
not related to the distribution of users in the cell.
of the cell is permitted to emit Pt,max while the power
control algorithm forces all closer users to decrease their
transmission power with respect to their path loss. 3.3.2 Power controlled network: The average power
consumption of users in the power controlled network can
3.2.2 BPPM scheme: To compute the error probability be calculated using the location vector (r) and the PCE
for BPPM power controlled network, (6) and (7) are valid vector (X ) both of which are (M + 1)-dimensional because
and using the same methodology of obtaining (13) we can of a desired user plus M interferers. Considering
easily compute PE (l0 , l1 ) as follows Pr,d = H0 (ri )Pt and using (11), the transmission power for
the ith user is as follows
PE (l0 , l1 ) = P(U0 , U1 |l0 , l1,0 )
 Pr,d = Pt,i H0 (ri ), where xi  N (0, s2PCE ) (19)
= P(U0 , U1 |l0 , l1,0 , X )fX (X ) dX (16)
X This leads to the following equation for the transmission

1924 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

power of the ith user 4 Numerical results

In this part we present some numerical result on performance
1 of the introduced cases and scenarios using the previously
Pt,i = Pr,d e L0 (ri ),
xi
where L0 (ri ) = (20)
H0 (ri ) presented analytical evaluations. We also present Monte
Carlo simulations to verify the numerical results. We
assume an infrared CDMA system with Rb ¼ 2 Mbps and
Regardless of how power control algorithm works, from (12)
Rb ¼ 1 Mbps per user for OOK and BPPM cases,
to (15) we know how much power is required at base station
respectively. These bit rates are chosen to allocate equal
(Pr,d) to achieve acceptable BER. Using (20), the required
bandwidth to the two analysed schemes. Furthermore, the
transmit power of each user can be obtained. It should be
OOC code words with length Lc ¼ 256 and weight K ¼ 5
noted that in the uplink, received signals from all users
(which result in Nmax ¼ 12) is considered for computations.
experience the same SINR because they are all detected by
The system is assumed to operate at l ¼ 850 nm and the
one photodetector at the base.
cell radius is fixed to 4 m. Photodetector’s efficiency, area
and dark current are assumed to be h ¼ 0.77, Ad ¼ 1 cm2
As we know the users are distributed independently on the
and id ¼ 160 nA, respectively, as in [1, 11, 21]. Although,
cell area and their PCE is independent of each other, too.
background light irradiance can be as strong as
Therefore the average power consumption is expressed by
Ib ¼ 10 000 mW/cm2 because of direct sunlight without
averaging of (20) over position and PCE variance, which
optical filtering in the field of view of the receiver, the
leads to
IrDA standard places the limit of Ib ¼ 490 mW/cm2 on the
background light for the operation of an infrared receiver
 
1 2 2 2ri [27]. This amount of background light is the equivalent of
Pav = Pr,d exi L(ri ) √

e−xi /sPCE dxi dri direct sun in the field of view of the receiver with a
ri xi 2p s PCE
R cell2
  developed unexposed film such as a long-pass optical filter
Rcell
1 −x2i /s2PCE dxi [11]. In all cases, optimum threshold is assumed for OOK
= Pr,d L(ri ) √

e detection system [10, 30].

0 xi 2psPCE
× f (ri ) dri Fig. 2 illustrates the near –far problem in a basic OOC
Rcell wireless infrared network without power control employing
2
= Pr,d esPCE /2 L(rt )f (rt ) drt (21) OOK or BPPM schemes using (2) and (8). In this figure
0
the effect of user’s position (r) on BER is plotted for five
active interfering users (M ¼ 5). This figure clearly
2
where f (r) = 2r/rcell . Additionally, for the ideal power demonstrates the need for control on transmission power.
controlled network, finding the average power consumption As an example when user is located at the radius of 1.5 m
is straightforward and is written as follows or more, BER is worse than 1024 for Pt ¼ 5 mW and
therefore users at distance over 1.5 m cannot properly
Rcell connect to base. It can be seen that the performance of the
network for both OOK and BPPM degrades rapidly as the
Pav = Pr,d L(ri )f (ri ) dri (22)
0 user moves towards the boundary of the cell, in fact even
if a user changes its position a few centimetres, its
performance might degrade or improve about ten times.
It is important to note that there are two factors that affect This demonstrates that the users who are located near the
average power consumption in (21), first and obvious one is base station obtain much better BER than required,
2
esPCE /2 that directly increases the average power whereas the data from farther users cannot be detected with
consumption and the power control error variance. a desired BER. The designer should consider the far users
However, one should note that the desired received power to have satisfactory performance but much power will be
at the receiver (Pr,d) in (21) is highly affected by PCE to wasted because of over design for near users. Also, in this
achieve a desired error probability as well. To elaborate this case the transmission power is the only parameter to
effect consider, for instance, an ideal power controlled improve the network’s performance, however, as Fig. 2
network that works having BER ¼ 1029 and consequently illustrates clearly, the effect of change in transmission
its average power consumption is given by (22). Now if in power has its minimum effect just when it is needed the
this case PCE is added to the network, BER increases, for most, that is when the user is at the edge of the cell. For
example, to 10226 while average power consumption is example, for a user of OOK system at a distance of 1 m
increased by esPCE /2 . However, it is clear that in order to from the source, increasing the transmit power from 5 mW
compare the two cases the power control algorithm should to 50 mW improves the performance from around 1024 to
increase the desired received power (Pr,d) for the latter case less than 1029 whereas the same amount of increased
to attain BER ¼ 1029 which consequently increase the power for a user at r ¼ 3 m changes the BER from 1021
average power consumption to a higher value. to 1022 approximately. Furthermore, although BPPM

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1925
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 2 System performance against the user’s distance from base station (r) for several transmission power ranges
Results are presented for a case with M ¼ 5 interfering users

performs better at lower transmission powers than OOK wireless system is a band limited system. Channel
because of mean noise cancellation in detection, both frequency response will degrade when using large
modulation schemes show the same weakness when the bandwidth. So the designer selects code length as long as
network becomes interference limited at large r values. possible (limited by channel) to have better performance
and looks for other solutions to improve system output.
Now let us find out whether code parameters can improve
system performance or not. Fig. 3 depicts BER against code Some other performance degradation issues of the network
length for different transmission powers. There are five without control on the transmission power are shown in
interfering users, the desired user is located at the middle Fig. 4 which is also obtained from (2) and (8). The upper
of the cell (r = Rcell /2), and code weight is kept fixed. It is bound of BER against number of uniformly distributed
seen that larger codes can have a slight better performance. interfering users for several values of transmission power is
But can we increase code length to improve performance to plotted in Fig. 4a for a user located at the middle of the
the desired value? The answer is NO! Because optical cell. Transmission power is controlled through (6)
considering appropriate channel loss.

It is apparent form Fig. 4a that increasing the number of

interfering users degrades the link performance because of
the increase of multiple access interference (MAI). It is also
clear that dependence of performance on the number of
users is less for lower transmission powers because of noise
domination and less interference. It can also be seen from
Fig. 4b that when the system is MAI limited, increasing
the transmission power of all users does not solve the
problem. For example, there is little BER improvement in
going from 40 mW of transmit power per user to 100 mW
of transmit power. Moreover, as stated above and obviously
demonstrated in Fig. 4b, BPPM gives better performance
in lower transmission powers where the system is noise
limited, nevertheless, increasing the transmission power
fades BPPM strength over OOK because MAI dominates.

Figure 3 Effect of code length on BER for different Finally, Fig. 5 shows the effect of cell radius on BER for
transmission powers several values of uplink transmission power when there is

1926 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

Figure 4 Performance degradation issue of a non-power controlled network

a BER against the number of interferers (M), user is placed at the half cell radius, Rcell ¼ 4 m
b BER against transmission power in the same situation

no power control. Users are assumed to be uniformly decreasing cell radius, or increasing transmission power
distributed over cell area. Cell radius impacts f (r) in (2) confirming the previous stated system behaviour. This
and (8). This curve can be used to obtain an appropriate clarifies the importance of power control for proper
range of coverage for a given target BER value and uplink operation of the system.
transmission power level. For example, for a target BER of
1023, a cell radius of around 2.2 m can be achieved if The performance results of the power controlled network
transmit power is set to 50 mW. It is clear that low target using (13) and (16) are depicted in Figs. 6 and 7. In this
values of BER (less than 1024) cannot be achieved even by case the limiting factor is the maximum allowed

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1927
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 5 BER against cell’s radius (Rcell) for M ¼ 5 and different values of transmission power

transmission power that should be set to a proper value where enhance the system performance in comparison to what
all users in the cell (including users at the edge) attain Fig. 2 (curves with Pt ¼ 50 mW) stated in worst case
appropriate BER. Users located closer to the edge (worst-case happens when user is placed in the farthest
experience higher path loss and therefore should transmit position from centre), considering points with five
higher power in comparison to other users in the cell. interferers in Fig. 6b. Moreover, the capturing effect and
Regardless of how power control algorithm operates, the near – far problem (Fig. 2) is solved in this scenario and all
maximum allowed transmission power is directly related to the users disregarding to their location achieve the same
desired received power at the base station with BER as predicted in Fig. 6b. Estimation errors (higher
Pr,d ¼ H0(Rcell)Pt,max , which means all in-cell users have values of PCE, i.e. sPCE or call it s) degrade this
been forced to reduce their transmission power so that the performance as it can be seen in this figure, but even with
received power of all users at BS is equal to the ones at the s ¼ 0.5 results are better than the case of no power
edge (sending Pt,max). control. Fig. 6 also demonstrates the improvement on
system capacity when using power control. For example to
Fig. 6a illustrates the effect of maximum transmission achieve BER ¼ 2 × 1024, system can tolerate only one
power (Pt,max) on the error performance of network when interferer when PCE is high (s ¼ 0.5), but eight users
the power control procedure is assumed to be ideal using (seven active interferers) can coexist in the system when
perfect channel estimation (s2PCE = 0). This figure shows PCE is low (s ¼ 0.05). One of the interesting observations
that employing power control greatly enhances the system from the results of this figures is that, even though BPPM
performance. It should be noted that as this figure shows performs much better than OOK in lower PCE, for
for both OOK and BPPM, when M is less than code example s ¼ 0.25, it starts to lose its attitude for higher
weight, that is, M , 5, MAI is negligible and BER can be errors in power control algorithm and gives unacceptable
reduced as much as desired by increasing the transmission performance as PCE increases. This is verified by Fig. 7 in
power to overcome noise sources. Perfect power control is which the performance degradation of both OOK and
an idealistic assumption because channel estimation is BPPM modulation schemes is plotted against s for
subject to estimation error and that impacts the different number of interfering users and two different
performance of the power control algorithm. Analysis of Pt,max values. As predicted by Fig. 6, Fig. 7a states that
the impact of power control error on system performance BPPM performs better than OOK when the system is
has also been performed using Monte Carlo simulations. working near ideal power control scheme; however,
The results of simulation are very close to the analytical increasing the amount of PCE leads the performance of
results and are presented in Fig. 6a. BPPM to degrade faster than OOK until it gets worse for
larger amount of error in estimation. The cross point varies
The effect of PCE on BER is shown in Fig. 6b for a case for different number of interfering users. In BPPM
where the maximum transmission power of a node is set to decision is based on comparing two decision variables, both
50 mW. It can be seen that power control can considerably of which suffer from power control error, so the total

1928 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

Figure 6 Performance evaluation of a power controlled network

a BER against maximum transmission power in ideal power control case for various values of M
b BER against number of interferers (M) for several values of PCE variance
Maximum transmission power is set to 50 mW

variance of error is actually doubled in BPPM. This leads to transmission powers (in low transmission power, BPPM
sharper performance degradation in BPPM over OOK. The performs better than OOK).
performance degradation of BPPM over OOK starts sooner
if we increase the maximum allowed transmission power Mention that obtaining the right threshold level for OOK
which is illustrated by Fig. 7b. In high transmission power, requires relatively accurate levels of received power estimation
where the effect of noise is cancelled and interference is the which is exactly what is required for a good power control
limiting factor, performances of BPPM and OOK are close algorithm. Fig. 8 shows the effect of threshold on OOK
to each other. So power control error makes BPPM to have BER for different PCE. Each curve has an optimum point
worse performance than OOK in comparison to low in which BER is at its minimum value. It can be seen that

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1929
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

Figure 7 BER of a non-ideal power control algorithm against PCE for different number of interfering users (M )
a Maximum transmission power is set to Pt,max ¼ 50 mW
b Pt,max ¼ 100 mW

when PCE is high, BER is less susceptible to threshold

selection. It means that in a situation where we have error
in estimating received power, although threshold cannot be
accurately obtained, system performance is less susceptible
to threshold value selection. So we expect OOK to perform
better than BPPM when PCE is high.

Finally, it is essential to find out the average power

consumption of the systems for possible errors in power
control procedure employing OOK and BPPM. In Fig. 9
Pav is plotted against s for a system achieving BER ¼ 1024.
Each point in this figure is obtained through (21) using the
following procedure: A plot of BER against Pr,d for different
values of s should be provided using (14) and (16) for
Figure 8 BER against threshold for different values of PCE OOK and BPPM, respectively. The required Pr,d for
Threshold is expressed as the percentage of maximum received level BER ¼ 1024 and for a specific s is extracted from that plot.

1930 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

Figure 9 Average power consumption of the network (Pav) against s to achieve BER ¼ 1024, for different number of
interfering users (M)

This Pr,d is used in (21) to calculate average power that these systems have to employ some form of power
consumption for the specified s. control with proper accuracy level in their uplink to
enhance their performance for both BPPM and OOK
Using Fig. 9, one can determine the maximum acceptable modulation schemes. Analytical equations and results were
error in power control algorithm to design a power efficient included to evaluate and emphasise the advantages of
network. In other words, the trade-off between design power control on the system performance which are
complexity of the power control loop and energy generally overcoming near – far effect and decreasing average
consumption of nodes can be addressed and numerically transmission power. Since error in power control
calculated by such plots. For example to design a BPPM algorithms is inevitable, we focused on PCE to see how
system with average power consumption less than 30 mW, PCE impacts system performance. Analytical evaluation of
PCE should be kept less than 0.15. Designing an imperfect power control and numerical results for some
inaccurate power control loop, results in higher average special cases were presented. Results show that the use of
power consumption to achieve BER of 1024. Furthermore, power control, even with channel estimation errors can still
Fig. 9 confirms that PPM appropriately enhances the significantly enhance the performance of the uplink
average power consumption of the network in lower channel in these systems while the effect of power control
deficiencies in power control algorithm while this behaviour error is quite different for PPM and OOK schemes. The
rapidly diminishes when PCE is increased and OOK results demonstrated that PPM performs better than OOK
becomes the first choice to decrease average power in a typical non-power control enabled network as well as a
consumption of users. power controlled network with lower errors in channel
estimation. But power control error has a larger effect on
PPM than on OOK and as a result OOK can be
5 Conclusions considered optimal when the system suffers from large
In this paper we analysed the performance of an optical amounts of error in its power control process. As stated in
CDMA wireless LAN with and without control on the paper, it should be noted that although OOK shows
transmission power. The system uses OOC with minimum better robustness in some cases, the main practical
auto- and cross-correlation as an implementation of disadvantage of this modulation scheme is its requirement
CDMA concept. Both OOK and BPPM modulation for an adaptive optimal threshold level to achieve the best
schemes are employed. Intense background light and other performance.
noise sources are considered. In a wireless infrared medium
with line of sight, path loss is the main issue; therefore The success of OCDMA will definitely rest upon
power control becomes the main concern to reduce maturing technologies but more importantly it depends on
interference between users, each using a specific signature finding the right application where OCDMA features will
sequence. It is proposed and proved by numerical results stand out when compared to other multiple access

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1931
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010
www.ietdl.org

techniques. We have highlighted and discussed such an [11] ZAHEDI S., SALEHI J.A., NASIRI-KENARI M.: ‘M-ary infrared CDMA
application in this paper and reported some analytical and for indoors wireless communications’. Proc. IEEE PIMRC’01,
numerical computations to overcome some basic San Diego, CA, USA, September 2001, vol. 2, pp. 6–10
inefficiencies.
[12] GHAFFARI B.M., MATINFAR M.D., SALEHI J.A.: ‘Wireless optical
CDMA LAN Part I: digital implementation analysis’, IEEE
6 Acknowledgment J. Sel. Areas Commun., 2009, 27, (9), pp. 1676 – 1686
This paper was presented in part at the IEEE 25th
International Performance Computing and Communication [13] GHAFFARI B.M., MATINFAR M.D., SALEHI J.A.: ‘Wireless optical
Conference, Phoenix, AR., USA, April 2006 and in part at CDMA LAN Part I: digital design concepts’, IEEE Trans.
the IEEE International Conference on Telecommunications, Commun., 2008, 56, (12), pp. 2145 – 2155
Penang, Malaysia, May 2007.
[14] AMINZADEH-GOHARI A., PAKRAVAN M.R.: ‘Analysis of power
control for indoor wireless infrared CDMA communication’.
7 References Proc. IEEE 25th Int. Performance Computing and Commun.
Conf., IPCCC’06, Phoenix, AR, April 2006, pp. 246–252
[1] KAHN J.M., BARRY J.R.: ‘Wireless infrared communications’,
Proc. IEEE, 1997, 85, (2), pp. 265– 298 [15] KHAZRAEI S., PAKRAVAN M.R.: ‘Power control analysis for
indoor wireless infrared CDMA networks using BPPM’.
[2] GREEN R.J., JOSHI H. , HIGGINS M.D. , LEESON M.S. : ‘Recent Proc. IEEE 14th Int. Conf. on Telecomm. ICT’07, Penang,
developments in indoor optical wireless’, IET commun., Malaysia, May 2007, pp. 402 – 406
2008, 2, pp. 3 – 10
[16] KHAZRAEI S., PAKRAVAN M.R.: ‘Analysis of generalized optical
[3] PAKRAVAN M.R., KAVEHRAD M.: ‘Design considerations for orthogonal codes in wireless optical networks using BPPM’.
broadband indoor infrared wireless communication Proc. IEEE 15th Int. Conf. Networks, ICON’07, Adelaide,
systems’, Int. J. Wirel. Inf. Netw., 1995, 2, (4), pp. 223– 238 Australia, November 2007, pp. 170– 175

[4] PAKRAVAN M.R., KAVEHRAD M., HASHEMI H.: ‘Indoor wireless [17] KHAZRAEI S., PAKRAVAN M.R.: ‘Comparison of AIr and IEEE
infrared channel characterization by measurements’, IEEE 802.11 with optical CDMA in wireless infrared
Trans. Veh. Technol., 2001, 50, pp. 1053 – 1073 communication’. Proc. IEEE 15th Int. Conf. Telecomm.
ICT’08, St. Petersburg, Russia, June 2008
[5] PAKRAVAN M.R., KAVEHRAD M., HASHEMI H.: ‘Effects of rotation
on the path loss and the delay spread in indoor infrared [18] KHAZRAEI S., PAKRAVAN M.R.: ‘Analysis of generalized
channel’. Proc. IEEE ICC’98, June 1998, vol. 2, pp. 817– 820 optical orthogonal codes in optical wireless local area
networks’, IEEE J. Sel. Areas Commun., 2009, 27, (9),
[6] HASHEMI H., GANG Y., KAVEHRAD M., BEHBAHANI F., GALKO P.A.: pp. 1572 – 1581
‘Indoor propagation measurements at infrared
frequencies for wireless local area networks applications’, [19] AKHAVAN K., KAVEHRAD M., JIVKOVA S.: ‘High-speed power-
IEEE Trans. Veh. Technol., 1994, 43, (Part 1 – 2), pp. 562– 576 efficient indoor wireless infrared communication using
code combining’, IEEE Trans. Commun., 2002, 50,
[7] RAJBHANDARI S., GHASSEMLOOY Z., ANGELOVA M.: ‘Bit error pp. 1098 – 1109
performance of diffuse indoor optical wireless channel
pulse position modulation system employing artificial [20] KAVEHRAD M., JIVKOVA S.: ‘Indoor broadband optical
neural networks for channel equalisation’, IET wireless communications: optical subsystems designs and
Optoelectron., 2009, 3, (4), pp. 169– 179 their impact on channel characteristics’, IEEE Wirel.
Commun., 2003, 10, pp. 30– 35
[8] MARSH G.W., KAHN J.M. : ‘Channel reuse strategies for
indoor infrared wireless communications’, IEEE Trans. [21] AMINZADEH-GOHARI A., PAKRAVAN M.R.: ‘Power control to
Commun., 1997, 45, (10), pp. 1280 – 1290 enable QoS for indoor wireless infrared CDMA networks’.
Proc. IEEE Int. Conf. Communications Electronics, ICCE’06,
[9] MARTIN-GONZALEZ J.A., POVEZ E., LOPEZ-HERNANDEZ F.J.: ‘Random Hanoi, Vietnam, October 2006, vol. 1, pp. 246 – 252
optical codes used in optical networks’, IET Commun., 2009,
3, pp. 1392 – 1401 [22] SALEHI J.A.: ‘Code division multiple access techniques in
optical fiber networks - Part I: fundamental principles’, IEEE
[10] ZAHEDI S., SALEHI J.A., NASIRI-KENARI M.: ‘A Photon counting Trans. Commun., 1989, 37, pp. 824– 833
approach to the performance analysis of indoors wireless
infrared CDMA networks’. Proc. IEEE PIMRC’00, London, [23] SALEHI J.A., BRACKETT C.A.: ‘Code division multiple access
UK, September 2000, vol. 2, pp. 928– 932 techniques in optical fiber networks - Part II: Systems

1932 IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933
& The Institution of Engineering and Technology 2010 doi: 10.1049/iet-com.2009.0465
 www.ietdl.org

performance analysis’, IEEE Trans. Commun., 1989, 37, [28] AZIZOGLU M., SALEHI J.A. , LI Y. : ‘Optical CDMA via
pp. 834– 842 temporal codes’, IEEE Trans. Commun., 1992, 40,
pp. 1162 – 1170
[24] SHALABY H.M.H.: ‘Chip-level detection in optical code-
division multiple-access’, J. Lightwave Technol., 1998, 16, [29] VITERBI A.J.: ‘CDMA - principles of spread spectrum
pp. 1077 – 1087 communications’ (Addison-Wesley, Reading, MA, 1995)

[25] JAZAYERIFAR M., SALEHI J.A.: ‘Atmospheric optical CDMA [30] ZAHEDI S., SALEHI J.A.: ‘Analytical comparison of various
communication systems via optical orthogonal codes’, fiber– optic CDMA receiver structures’, IEEE J. Lightwave
IEEE Trans. Commun., 2006, 54, pp. 1614 – 1623 Technol., 2000, 18, pp. 1718 – 1727

[26] GAGLIARDI R.M., KARP S.: ‘Optical communications’ (John [31] ANDREWS J.G., MENG T.H.: ‘Optimum power control for
Wiley and Sons, 1995, 2nd edn.) successive interference cancellation with imperfect
channel estimation’, IEEE Trans. Wirel. Commun., 2003, 2,
[27] IrDA standards can be obtained at http://www.irda.org/ (2), pp. 375– 383

IET Commun., 2010, Vol. 4, Iss. 16, pp. 1919 – 1933 1933
doi: 10.1049/iet-com.2009.0465 & The Institution of Engineering and Technology 2010