Interference-Limited Analysis of

the Convolutionally Coded Optical OOK/BPPM CDMA System

G. Farhadi S. H. Jamali
Dept. of Electrical & Computer Engineering Dept. of Electrical & Computer Engineering
University of Tehran University of Tehran
g.farhadi@ece.ut.ac.ir hjamali@ut.ac.ir

Abstract: The convolutional-coded Optical proposed by some research works [3], [4]. In [3],
CDMA systems with On-Off Keying (OOK) and the employment of convolutional code in an
Binary Pulse Position Modulation (BPPM) optical OOK-CDMA system without hard limiter
schemes are studied in this paper. In this work, the is suggested.
employment of a single hard limiter is also In this work we restrict our attention to the effect
considered. We evaluate the upper bound on the of MAI on the bit error performance of the
bit error probabilities of our proposed schemes convolutional coded optical OOK/BPPM-CDMA
under the interference-limited condition. For a systems with and without hard limiter and evaluate
convolutional coded optical OOK-CDMA system, the bit error probabilities. We come up with an
we drive a more accurate upper bound than that upper bound on bit error rates that is more accurate
obtained by Dale et.al in 1995. It is shown that the than the one obtained in [3]. We show that the
implementation of convolutional codes in an employment of the convolutional codes in optical
optical CDMA system provides significant CDMA systems can remarkably reduce the MAI
improvement in system performance. effect. Particularly, our proposed schemes enable
Furthermore, for a certain bit error rate, our the use of OOCs with less weight in the CDMA
proposed schemes enable the use of Optical system than that required in the comparable
Orthogonal Codes (OOCs) with less weight and uncoded systems.
hence increase the channel bit rate compared to Section 2, reviews a convolutional coded OCDMA
the one required in the uncoded system. system. In Section 3, the channel models are
presented for both OOK and BPPM schemes and
Keywords: Optical CDMA system, OOK and then probabilities of bit error are evaluated.
BPPM modulation, Convolutional Codes Furthermore, in this section we evaluate the
tightness of the derived upper bounds by means of
1 Introduction simulations. In Section 4, we compare the
performance of the uncoded and the proposed
Optical Code Division Multiple Access (OCDMA) coded optical systems. Finally, in Section 5,
has recently attracted great attention for optical concluding remarks are presented.
communication systems. Optical CDMA is an
asynchronous scheme, which does not require time 2 Coded Optical CDMA System
synchronization and frequency management [1,2].
Multiple access interference (MAI) is one of the Fig.1 shows a typical block diagram of a fiber
major factors for performance degradation in an optic CDMA system, which uses a convolutional
optical CDMA system [1-4]. Recently, the use of encoder right after the information source and a
channel coding with optical CDMA systems, for hard Viterbi decoder at the receiving part.
the performance improvement purpose, has been
Information Convolutional Encoder
r(t)
OOK or BBPM Optical Optical OOK or BBPM
Source j
th Encoder Modulator jth hard Correlator De-Modulator
Limiter

Fig. 2 Detector Structure.

Information Viterbi OOK or BBPM Decoder
th
Sink j Decoder De-Modulator jth
interference among users, which is the chip
Fig. 1. Convolutional coded optical CDMA system. synchronous case.
The optical CDMA channel can be regarded as a
Each convolutionally coded information bit binary channel as shown in Fig.3.
modulates a laser source using either OOK or
binary PPM (BPPM) modulation scheme. Then, p(0|0)
each modulated signal is spread by a signature bj=0 rj=0
sequence assigned to each user’s encoder. In this
paper, OOC codes with minimum auto- and cross p(0|1) p(1|0)
correlation constraints are used. Therefore, if N is
the number of active users, it can be shown that [5] bj=1 rj=1
p(1|1)
 F −1 
N≤  (1)
 w( w − 1)  Fig. 3. Binary channel
where F and w are the length and weight of OOC
codes, respectively and x denotes the integer In an OOK-CDMA system, p (0 1) =0 [1]. Hence,
part of the real number x. The perfect optimal the OOK system can be characterized as a binary Z
OOCs are also assumed in this paper. channel. Assuming equally likely input
At the receiving end, a copy of the desired signal
information bits, p (1 0) will be
along with N-1 interfering signals of all other users
is received. We use the detector structure as Fig. 2. p (1 0) = p = 2Pe (3)
The output function of the ideal optical hard limiter
(HL) is defined as follows where Pe denotes the probability of bit error for
OOK system.
1 ;t ≥ 1
g (t) =  (2) An alternative to OOK is PPM. A PPM system has
0 ;t < 1 the main advantage that there is no need to define a
where t is the normalized input optical power. It is threshold level as it is required in the case of OOK.
shown that the employment of hard limiter However, a PPM system suffers from the need for
enhances the system performance significantly [2]. the extra bandwidth which can be considered as a
In an optical CDMA system, detection is drawback. Assuming equally likely input
performed by counting the number of photons. In information bits, a BPPM system can be modelled
an OOK system with correlator type receiver, if the as a binary symmetric channel (BSC) with the
collected photon counts over the entire chip time following transition probabilities
duration is less than a predefined threshold, 0 is p (0 1) = p (1 0) = Pe (4)
detected, and otherwise 1 is declared. In a BPPM
system with correlator type receiver, if the sum of p (0 0) = p (11) = 1 − p (5)
e
the photon counts over the entire chip time where Pe denotes the probability of bit error for
duration of the first slot is greater than that of the
second slot, 0 is declared and otherwise, 1 is the BPPM system. Upper bounds on bit error
detected. Afterward, a hard input Viterbi decoder probabilities for both OOK/BPPM CDMA
decodes the recovered encoded information bits. systems, with and without hard limiter, are
presented in table 1 [1], [4]. The threshold level,
th, should satisfy the following inequality [1]
3 Performance Analysis
0 < th ≤ w (6)
In order to enhance the performance of an optical
Now, we evaluate the probability of bit error under
CDMA system, the employment of a convolutional
the quantum interference limited condition. We
code right after the binary source is proposed.
also consider a system with the worst-case
 Table 1. The Upper Bounds on the Bit Error Probabilities, Pe
Optical OOK-CDMA system Optical BPPM-CDMA system
System N −1− i 2m+ i 2 ( N −1 −m )− i
  w2     w2 
i
1 N −1  w 2  N −1 N −1 −i
 N − 1  N − 1 w 2 
without hard
∑ 
2 i =th  2 F 
1 −  
  2F   ∑∑      1 − 
  2F  
limiter (NHL)    i = w m= 0  m   m + i  2 F    
System with N −1 −i
  N −i

1  w  th−1   w   w
w 
hard limiter  ∏ 1 − 1 −   ∏ 1 − 1 −  
(HL) 2  th  i =0   2 F  


i =1   2 F  


The upper bound on the bit error probability of a equally likely, K is also a binomial random
convolutional code (k,n,m) is [6] variable with the parameter (d ,0.5) .Therefore the
1 probability of the occurrence of k 1’s in the d bits
Pb ≤ ∑ Bd p d
k d = d free
(7) of the correct sequence P( K = k ) ,can be written as
d  1
where Bd is the total numbers of nonzero P( K = k ) =   ( ) d (9)
information bits on all distance d paths and p d is k  2
the event error probability. Since N 10 and N 01 are statistically independent
random variables, then
3.1 Coded Optical OOK-CDMA System P( N10 = n10 , N 01 = n01 | K = k ) =
 d − k  n01 k 
A relation that incorporates Bd s to the bit error   p (1 − p) d −k − n01 .  q n10 (1 − q ) k −n10
probability calculation of a convolutional coded  n01   n10 
optical OOK-CDMA system has been suggested (10)
in [3]. Here we present a closed form expression where p and q denote p (1 | 0) and p (0 | 1) ,
for the upper bound on the bit error probability in respectively.
terms of the code generating function. To evaluate Therefore,
the event error probability we consider an 1 d  d  d − k  d − k  n01
incorrect path through the trellis, which differs Pd ≤ ( ) d ∑   ∑   p (1 − p) d − k − n01
from the correct path in d positions. We define the 2 k =0  k  n01= 0 n 01 
14444 4244444 3
following parameters, which are used in our d −k d− k
≤ 2d −k p 2 (1− p ) 2
theoretical evaluations.
k: the number of 1’s in the d bits of the correct k
k 
sequence ( 0 ≤ k ≤ d ) . ∑  q n10 (1 − q) k − n10
n10 = 0 n10 
n 01 : The number of 0 → 1 transitions from the 1444 424444 3
k k

correct to the received path ( 0 ≤ n01 ≤ d − k ) ≤ 2 k q 2 (1− q ) 2

k
n10 : The number of 1 → 0 transitions from the d
 d  q (1 − q) 
d d

≤ p (1 − p) ∑  
2 2 
correct to the received path ( 0 ≤ n10 ≤ k ) k =0  k  p (1 − p) 
Then, the event error probability can be written as d
d −k
d
 d
q (1 − q ) 
= p (1 − p ) 1 +
d k
p d = ∑ ∑ ∑ ( P( E K = k , N 01 = n01, N10 = n10 ).
2 2

k =0 n 01= 0 n10 = 0 144444

42444444 3  p (1 − p ) 
≤1
(11)
P ( N 01 = n01 , N10 =n10 K = k ) As an optical OOK-CDMA system can be
P ( K = k )) regarded as a Z channel, i.e. q = 0 , then
d d
(8)
Where E denotes the event error. N 01 and N10 are Pd ≤ p 2 (1 − p ) 2 (12)
the binomial random variables with the Denoting the generating function of the
parameters (d − k , p(1 | 0) ) and (k , p (0 | 1)) ,respecti- convolutional code by T ( D, I ) , it can be shown
vely. Assuming that the input information bits are that [6]
∂T ( D, I )
= ∑ Bd D d (13)
∂I I = 1 d = d free
Consequently,

1 ∂T ( D , I )
Pb−OOK ≤ I =1 (14)
k ∂I
D = p(1 − p )
where p is substituted with (3).
In order to verify the validity of the obtained
upper bound, we simulate a convolutional coded
system over a binary Z channel. We use optimal Fig.4 Performance of the convolutional code over the
convolutional code (2,1,3) with generator matrix binary Z channel.
G( D) = (1 + D 2 + D 3 ,1 + D + D 2 + D 3 ) [6].
T ( D, I ) of this convolutional code is [6] 4 Numerical Results
D 6 I 2 + D 7 I − D8 I 2 Now, we calculate the upper bounds on the bit
T ( D, I ) = (15)
1 − 2DI − D3 I error probabilities versus the number of active
Fig.4 shows the performance of the convolutional users.
code over a binary Z channel. It is observed that In an uncoded fiber optic CDMA system with
the simulation results tightly coincide with the OOCs (F,w,1,1) , the channel bit rate is
obtained theoretical upper bounds. To compare  1
our analytical approach with that of the previously  Nw( w − 1)T ; OOK

≅
reported one [3] we include the upper bound on c
Runcoded
the bit error probability using Dale ’s equation [3].  1
; BPPM
The required coefficients for evaluating the Dale’s  2 Nw( w − 1)Tc
equation, i.e. Bd s, can be easily obtained by (17)
expanding of the code generating function. It is It is observed that, for a certain OOC codes and
observed that our analytical approach predicts the chip time (Tc ), the required bandwidth of BPPM
performance of the codes more accurate than the
Dale’s approach. is doubled. In order to have the same bit rate for
the coded and uncoded systems the chip time and
3.2 Coded Optical BPPM -CDMA system the number of users are held fixed. In this case the
length and the weight of OOC codes, used in the
The optical channel of a BPPM-CDMA system is coded systems, should satisfy the following
modelled as a BSC. The upper bound on the equations [3]
probability of bit error of an arbitrary  Fcoded = Rc F 
convolutional code over BSC is recognized and 
 1 + 1 + 4 Rc w( w − 1)  (18)
expressed by [6]
 coded 
w = 
 
 2 
where F and w are the length and the weight of
1 ∂T ( D, I ) the OOC sequences used in the uncoded system,
Pb− BPPM ≤ I =1
k ∂I respectively, and Rc is the rate of the
D = 2 p (1 − p) convolutional code.
(16) We use the optimal convolutional code (2, 1, 3)
where p is substituted with the appropriate with the given generator matrix in Section 3.1, in
relation given in Table 1. all numerical evaluations.
Fig. 5. Performance of the optical OOK-CDMA system Fig. 6. Performance of various optical OOK-CDMA
for a fixed channel bit rate. systems.

Table 2. Weights of OOCs which is required in the

4.1 Optical OOK-CDMA System uncoded OOK system for a certain bit error
rate, number of users and chip time.
Figs.5 shows the BER of various systems with Coded w used in Rb
OOK scheme. In order to have the same channel system Pe the uncoded
bit rate we consider different values for the OOC (w=5) system Runcoded
code weights. The threshold values for these wNHL = 10 2.25
systems are set to their corresponding OOCs code Without 5 × 10 −11
weights. It is observed that the employment of the HL(NHL) wHL = 8 1.40
convolutinal code in an optical CDMA system wNHL = 12 3.30
−15
provides a remarkable performance improvements HL 5 ×10
compared to the uncoded system. Furthermore, wHL = 11 2.75
the performance of coded system with hard limiter
is superior with respect to the other systems. (According to (1)) and increases the available
In addition, for a certain bit error rate, our band width. A convolutional coded optical
numerical results (Fig.6 and Table 2) show that by CDMA system permits higher number of active
the use of a convolutional code, it is possible to users with el ss hardware requirements than the
implement an optical CDMA system with less ones required in the uncoded system.
OOC code weight. Consequently, the coded
system can operate at higher channel bit rates. 5 Conclusions
Furthermore, having a smaller w value is
beneficial from the hardware system In this paper, we evaluated the upper bounds on
implementation point of view [4]. the bit error probability of the convolutional
coded fiber-optic CDMA system assuming both
4.2 Optical BPPM -CDMA System OOK and binary PPM schemes. Systems with and
without hard limiter were considered. It was
Fig. 7 shows the BER of various BPPM systems shown that the presented upper bound on the bit
for a constant channel bit rate. Again, it is error probability of an OOK-CDMA system is
observed that the coded systems outperform the more accurate than that obtained by Dale et.al.
uncoded ones. As expected, the system Our numerical results showed that the
performance improves significantly in a convolutional codes significantly improves the
convolutional coded system with hard limiter. performance of the optical systems; particularly in
Furthermore, for a certain bit error rate, as shown those ones that employ hard limiter. In addition,
in Fig. 8 and table 3, smaller w values can be used for a certain bit error rate, the employment of
to implement the coded optical CDMA system. convolutional codes, whereas accommodating
Hence, for a given bit error rate and the number of higher number of users, enables the use of OOCs
active users, the employment of a convolutional with less weight and hence providing higher
code reduces the required OOC code length channel bit rates.
 Fig. 7. Performance of the optical BPPM-CDMA Fig. 8. Performance of various optical BPPM -CDMA
system for a fixed channel bit rate. systems.

References Table 3. Weights of OOCs required in the uncoded

BPPM system for a certain bit error rate,
[1] J.A. Salehi,´ Code Division Multiple Access number of users and chip time.
Techniques in Optical Fiber Networks-part I: Coded w used in Rb
Fundamental Principles”, IEEE Trans. on system Pe the uncoded
(w=5) system Runcoded
Communications, vol. 37, No.8, pp 824- 833,
Aug. 1989 wNHL = 9 1.80
−10
[2] J.A. Salehi, and C. Brackett,´Code Division Without 3 × 10 wHL = 8 1.40
Multiple Access Techniques in Optical Fiber HL(NHL)
Networks-part II: System Performance wNHL = 12 3.30
HL −13
Analysis”, IEEE Trans. on Communications, 10 wHL = 10 2.25
vol. 37, No.8, pp 834-842, Aug. 1989.
[3] M.R. Dale, and R.M. Gagliardi, “Channel
Coding for Asynchronous Fiberoptic CDMA
Communications”, IEEE Trans. on [5] Fan R.A Chung, Jawad. A. Salehi and Victor
Communications, Vol. 43, No.9, September K. Wei, “Optical Orthogonal Codes: Design,
1995. Analysis and Applications.” IEEE Trans. On
[4] C. Argon, and S.W. McLaughlin, “ Optical Information Theory, vol. 35, No. 3, PP 595-
OOK-CDMA and PPM-CDMA systems with 604, May 1989.
turbo product codes”, IEEE journal of [6] Shu Line, D.J. Costello, Jr., Error Control
lightwave technology, vol. 20, No. 9,PP Coding: Fundamentals and Applications,
1653-1663, Sep. 2002. Printice-Hall, 1983.