This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

Analysis and Design of Adaptive OCDMA

Passive Optical Networks
Mohammad Hadi, Member, IEEE, and Mohammad Reza Pakravan, Member, IEEE

Abstract—OCDMA systems can support multiple classes Division Multiple-Access (OFDMA) and OCDMA (known
of service by differentiating code parameters, power as Multi-Carrier OCDMA) [3], [6] or Wavelength Division
level and diversity order. In this paper, we analyze BER Multiplexing and OCDMA (known as WDM/OCDMA) to
performance of a multi-class 1D/2D OCDMA system and
propose a new approximation method that can be used increase the system performance and capacity [7], [8].
to generate accurate estimation of system BER using a Multiple-Code (MC) and Multicode-Keying (MK) schemes
simple mathematical form. The proposed approximation are also proposed to support multirate transmission in
provides insight into proper system level analysis, system OCDMA systems [2], [9]. Furthermore, some researchers
level design and sensitivity of system performance to have concentrated on designing new code families and
the factors such as code parameters, power level and
diversity order. Considering code design, code cardinality efficient coding/decoding structures for OCDMA systems to
and system performance constraints, two design problems reduce design complexity and increase system capabilities
are defined and their optimal solutions are provided. We [10]–[14]. Performance analysis of different OCDMA-based
then propose an adaptive OCDMA-PON that adaptively systems is another attractive research topic [1]–[4], [6],
shares unused resources of inactive users among active [15].
ones to improve upstream system performance. Using
the approximated BER expression and defined design As examples of OCDMA-based systems, we can refer to
problems, two adaptive code allocation algorithms for the the demonstrations of OCDMA Passive Optical Networks
adaptive OCDMA-PON are presented and their perfor- (OCDMA-PONs) for next generation high speed access
mances are evaluated by simulation. Simulation results networks [10], [16]–[18]. OCDMA-PON as a candidate for
show that the adaptive code allocation algorithms can Next-Generation (NG)-PON can provide full asynchronous
increase average transmission rate or decrease average
transmitted optical power of ONUs for dynamic traffic communication, data confidentiality, and symmetric band-
patterns. According to the simulation results, for an width for downlink and uplink [10]. Noting the exponen-
adaptive OCDMA-PON with BER value of 10−7 and user tial growth of highly-dynamic traffic patterns due to emer-
activity probability of 0.5, transmission rate (transmitted gence of social networking, real-time gaming and high
optical power) can be increased (decreased) by a factor definition audio-video streaming in modern networks, on-
of 2.53 (0.25) compared to fixed code assignment.
demand resource allocation is essential for cost reduction
Index Terms—Green Communication, Passive Optical and revenue generation. Consequently, on-demand and
Networks (PON), Optical Code Division Multiple Access agile resource provisioning in PONs is rapidly becoming
(OCDMA), Optical Codes, System Design, Adaptive Code
Allocation. a significant technical and economic priority. Software
Defined Networking (SDN) as an emerging technology is
a promising candidate for PON control plane to provide
I. I NTRODUCTION the favorable online resource assignment. Through the
PTICAL Code Division Multiple-Access (OCDMA) is separation of control and data plane in network ele-
O a well-known multiple access technique considered
in optical networks. In OCDMA networks, users share a
ments, SDN provides dynamic and fine-grained traffic
control that enhances total controllability, manageability,
common physical channel and each user sends its data efficiency and resource utilization of PONs [19], [20]. Re-
using a code string. OCDMA has some advantages over ducing the complexity of resource management algorithms
other multi-access methods such as simple and asyn- is the main challenge that should be tackled to let an
chronous operation, high resource utilization and capabil- OCDMA-PON provide on-demand and adaptive resource
ity of providing differentiated classes of Quality of Service allocation [19]–[21]. There is a strong desire to increase
(QoS) for different users. The performance of OCDMA the bandwidth while decreasing the power consumption
networks is mainly limited by Multi-Access Interference of NG-PON systems. Putting network element into sleep
(MAI) from other users [1]–[5]. There are research results mode is a cost-effective and attractive method developed to
that have combined the concepts of Orthogonal Frequency decrease the power consumption in NG-PON. Performance
of different green architectures for NG-PONs has been
Mohammad Hadi is a PhD student at department of Electrical Engi- compared in [22]. Many researchers have also proposed
neering, Sharif University of Technology, e-mail: mhadi@ee.sharif.edu. various green bandwidth allocation frameworks and al-
Mohammad Reza Pakravan is faculty member of Electrical Engi-
neering department, Sharif University of Technology, e-mail: pakra- gorithms for WDM- or TDM-PONs [22]–[24] but green
van@sharif.edu. resource allocation for OCDMA-PON needs more research.
Copyright ©2017 IEEE. Personal use of this material is permitted. However, permission to use this material for any other purposes must be obtained from the IEEE by sending a request to pubs-permissions@ieee.org.

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

Initially, OCDMA system was introduced by of inactive users to increase the transmission rate of
1D Strict/Generalized Optical Orthogonal Codes active ones while the power-efficient algorithm provides
(SOOC/GOOCs) as the user code strings. 2D wavelength- a green code allocation mechanism in which code weights
time optical coding schemes have also been studied are dynamically reduced to decrease transmitted optical
for OCDMA system which improves overall network power. Simulation results show that the code allocation
throughput and code cardinality [11], [12], [25], [26]. algorithms can use the randomness of the user activities to
Multi-class OCDMA systems have been proposed increase transmission rate or decrease transmitted optical
to provide different QoS levels that distinguish power and the amount of improvement is a function of
between users in terms of performance metrics such system BER.
as transmission rate and Bit Error Rate (BER). There are The rest of the paper is organized as follows. The system
many techniques that can be used to design multi-class model and the methods of QoS provisioning are described
OCDMA systems [27]–[30]. System level concepts such in Section II. The approximated BER expression is derived
as Multi-Length OCDMA (ML-OCDMA), Variable-Weight and validated in Section III. In Sections IV and V, we
OCDMA (VW-OCDMA) and Multi-Length Variable-Weight show the applications of the proposed BER expression in
OCDMA (MLVW-OCDMA) have been introduced for this developing systematic design and adaptive code allocation
purpose [30]. Differentiating in code parameters plays the procedures. Simulation results are included in Section VI.
key role in QoS provisioning of a MLVW-OCDMA system. Finally, we conclude the paper in Section VII.
Multi-level signaling technique has also been proposed
in which multiple classes of QoS can be created by II. S YSTEM M ODEL
differentiating in transmitted power levels [28]. Diversity The system model considered in this paper is the com-
in OCDMA system can also be used to provide various mon On-Off Keying (OOK) OCDMA with chip rate R c and
levels of QoS differentiated by BER performance without N active users [25], [31]. Each user has a transmitter
changing the code parameters [29]. that consists of an OOK modulator and a 1D/2D optical
Simple and efficient dynamic resource allocation and delay line encoder. A unique 2D optical code (ML,W, λ) is
system redesign is a key requirement for adaptive on- assigned to each user where M, L, W and λ are number
demand control of OCDMA systems. Since BER equations of wavelengths, code length, code weight and code cross-
of OCDMA systems are not simple to calculate and are correlation, respectively. To guarantee the unique code
not easy to reverse, the redesign and dynamic resource assignment, the number of users N should be less than
assignment problems are consequently not easy to solve code cardinality Φ(M, L,W, λ) which is in turn upper-
using few simple design equations and algorithms. In this limited by Johnson bound [26], [32]:
paper, an approximated simple expression for the BER
of a general multi-class OCDMA system is derived and M(ML − 1) · · · (ML − λ)
Φ(M, L,W, λ) É (1)
its accuracy is verified by comparing its results against W(W − 1) · · · (W − λ)
the previously published BER equations which are spe- Transmitters send the assigned optical codes of their
cial forms of the considered general case. The proposed intended users for each "1" data bit and no thing is sent
approximation is valid for OCDMA systems with 1D/2D during the bit interval of each "0" data bit. Users are
optical codes. It also supports multi-class OCDMA systems transmitting and receiving their chip sequences using opti-
having QoS classes differentiated by the mentioned system cal fiber cables that are connected together using a passive
level techniques. The proposed BER equation provides star coupler. Transmitted signals are positively added in
insight into proper system level analysis, design and their the star coupler and then a mixed signal is delivered to
corresponding sensitivity analysis. Considering code de- receivers. Receiver structure is based on the well-known
sign, code cardinality and system performance constraints, optical AND logic gate receiver [25]. We assume that the
we formulate two design problems that aim at max- main source of system performance degradation is MAI
imizing transmission rate and minimizing transmitted from other users and other system impairments such as
optical power. We investigate heuristic and brute-force background noise, shot noise and thermal noise are negli-
search approaches for the optimal solutions of the design gible. In fact, a practical OCDMA system can be considered
problems and use them to provide design guidelines and in the MAI-limited regime if optical power is sufficiently
procedures for a typical OCDMA system. The procedures higher than noise floor [33], which is guaranteed in our
are simple, easy to apply and can easily be used to adap- work. The effects of noise in OCDMA systems are studied
tively derive the required system parameters during on- in many papers such as [34] and can be combined with
demand resource provisioning for dynamic traffic patterns. our MAI-limited analysis.
We also use the BER equation and design procedures Since all users of a conventional OCDMA system are
to propose two simple code allocation algorithms, named given the same code parameters, power level and di-
"rate-efficient" and "power-efficient" code allocation algo- versity order, they experience equal transmission rate
rithms, for an adaptive OCDMA-PON system that dynam- and BER or in other words, the same QoS. In order to
ically reallocates user codes according to the activity of provide differentiated classes of QoS, ML-OCDMA has
users and dynamic traffic fluctuations. The rate-efficient been introduced in which code words have the same code
code allocation algorithm exploits the unused resources weight but different code length. VW-OCDMA is another

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

multi-class OCDMA system in which code words have OLT

ONU Transmitter 1 OLT Receiver 1
the same code length but different code weight. Multiple
classes of QoS can also be provided using simultaneous ONU Transmitter 2 OLT Receiver 2
differentiation in code weight and code length in MLVW- Combiner Splitter

OCDMA system. In a MLVW-OCDMA system, the length

and the weight of code words are chosen based on the Control
ONU Transmitter N Unit OLT Receiver N
desired QoS level such that high-weight and short-length
code words are assigned to users having lower BER and
higher transmission rate [30]. Multi-level signaling has (a) System block diagram
been proposed to improve the performance of OCDMA
systems and to provide multiple classes of QoS. In a multi- Tunable Intensity
Narrow 2D
level OCDMA system, users are divided into different Multiple-Wavelength
Delay Lines Modulators
Wavelength-Time
groups each transmitting at different power levels. In a Pulse 1 Optical Code
multi-level OCDMA system, high-power classes have bet- 2 2 M 1 M 1
ter BER performance than low-power ones [28]. Diversity t t
WDM WDM
in OCDMA deals with transmission of multiple copies of DEMUX MUX
a bit through spatially independent optical paths and can M 1
improve BER performance without affecting code parame-
M
ters. Diversity scheme proposed in [29] can be used to dif-
ferentiate users in terms of their BER. We use the notation
OCDMA(K, M, N, L, W, Γ, C, B) to characterize a general K-
class M-wavelength 1D/2D OCDMA system where vectors (b) ONU Transmitter
N[N i ]1×K , L[L i ]1×K , W[Wi ]1×K and B[b i ]1×K are number
of users, code length, code weight and diversity order vec- Tunable Hard
2D Delay Lines Limiters
tors, respectively. Matrices Γ[λ i j ]K ×K and C[c i j ]K ×K are Wavelengtn-Time
1
Optical Code
code cross-correlation and power ratio matrices. Clearly,
2 M 1 M 1 2
element i of each vector corresponds to i-th class and λ i j

Photo Detector
Received
t
denotes for cross-correlation between classes i and j code WDM AND
Bits

words. Furthermore, c i j shows the power ratio of class i DEMUX Gate

M 1
to class j defined as:
M
( P
b Pi c if P i Ê P j
j
ci j = (2)
1 if Pi < P j
(c) OLT Receiver
Figure 1: Adaptive OCDMA-PON system block diagram in upstream which includes
where P i and P j are the power levels of classes i and j, ONU transmitters, OLT receivers, optical combiner and optical splitter. Block
respectively. Clearly, the transmission rate of k-th class in diagram of the ONU transmitter and OLT receiver are also shown in which tunable
optical delay lines can be adjusted to fit a desired optical code.
a multi-class OCDMA system is given by:

Rc
Rk = (3) III. A PPROXIMATED BER E XPRESSION
Lk
The BER of k-th class in an OCDMA system character-
where R c is the mentioned common chip rate of the
ized by OCDMA(K, 1, N, L, W, Γ, 1, 1) is given by [30]:
system.
à !
Without loss of generality, we apply the concept of 1W Xk n i Wk
analysis and design of an OCDMA system for the case of P e(k) = (−1) (4)
2 i=0 i
an adaptive OCDMA-PON system as shown in Fig. 1. In λkq λkq
à !à !
K h i ko
this system, Optical Network Units (ONUs) are generating j i Wk − m kq N q
Y X X
1+ (−1) Pm
their upstream data using an 1D/2D tunable transmitter q=1 j =1 m= j j m− j
in which the code parameters are programmable. As in
other PON systems, all upstream data from ONUs are where N qk is the number of interfering users with values
kq
mixed together in the passive combiner and the combined N qk = N q , q 6= k and N kk = Nk − 1. P m is the probability that
signal reaches Optical Line Terminal (OLT). In OLT, a a class q user makes m interferences on a class k user.
splitter generates N versions of the received signal which Obviously, (4) is a complicated function of code parameters
are fed to a bank of N tunable 1D/2D receivers that decode and number of users. Furthermore, 2D, multi-level and di-
the received upstream signal [35], [36]. We only consider versity signaling severely increase the complexity of BER
MAI in our analysis and assume that ONU transmit expression [28], [29]. To relax the inherent complexity
powers are adjusted such that the received optical power of the BER expression, we have derived a simple form
from all ONUs are equal [37]. approximation that can provide accurate estimation of the

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

-1
10
maximized (or equivalently code length L is minimized):
-2
10

10
-3 max R (7)
W,L,λ

10
-4


 C1 : λ É W É ML
− 1) · · · (W − λ) É M(ML − 1) · · · (ML − λ)

C2 : NW(W


10-5
p

s.t. C3 : W É 2M λL
Pe

10-6 
C4 : P e (1, M, N, L,W, λ, 1, 1) É P e th





10
-7 C5 : W, L, λ ∈ N


10-8 Approximated BER for OCDMA(1; 1; N; 900; 5; 3; 1; 1)

Accurate BER for OCDMA(1; 1; N; 900; 5; 3; 1; 1)
Approximated BER for OCDMA(1; 1; N; 900; 5; 2; 1; 1)
Accurate BER for OCDMA(1; 1; N; 900; 5; 2; 1; 1)
C1 is an obvious constraint on optical code deign and C2
10-9
Approximated BER for OCDMA(1; 1; N; 900; 5; 1; 1; 1)
Accurate BER for OCDMA(1; 1; N; 900; 5; 1; 1; 1)
comes from cardinality constraint (1). C3 is an extension
10-10 of the proposed bound in [25] to 2D optical codes. Finally,
10 20 30 40 50 60 70 80 90
N C4 specifies BER performance requirement. Brute force
search over 3D region limited to Wmax , L max and λmax
Figure 2: Comparison of accurate BER and its approximation in terms of the
number of users for three single-class 1D OCDMA systems distinguished by needs Wmax L max λmax evaluations of constraints which
different values of cross-correlation λ = 1, 2, 3. In this comparison, we assume that may be complex and unacceptable for online redesign
P m = 0, m = 1, · · · , λ − 1 to compare the worst case BER performance.
or reoptimization of the system. We provide a heuristic
design routine shown in Alg. 1 which finds the solution
in a much simpler and faster method. For fixed value of
BER value for a general multi-class OCDMA system:
λ over continuous 2D region of L and W, C1, C2 and
K ³ N W
1h X ´ ki iWk b k c C3 yield ascending boundary functions for L in terms of
i i λ ki
P e(k) ≈ (5) W while C4 provides a descending one (remember that
2 i=1 2M λki L i
for a fixed value of P e th , L is a descending function of
The method of deriving the approximated BER expression W). Therefore, the optimum point is certainly placed on
is included in the Appendix. For a single class OCDMA, the intersections of C4 boundary with boundaries C1,
the approximated BER expression is simplified to: C2 and C3. This implies that for the discrete case, the
optimum solution is around these intersections. As can be
1 NW W
Pe ≈ ( )λ (6) seen from (6), BER is exponentially increased for higher
2 2MLλ values of λ and as code cardinality constraint C2 allows,
As a special case, for a single class OCDMA system with we should choose the smallest λ. In each iteration of Alg.
NW W
SOOC, the approximation is reduced to P e ≈ 12 ( 2ML ) 1, a limited region around the mentioned intersections is
which is equal to the approximation proposed in [29]. searched for the smallest L and the algorithm stops when
The accuracy of the approximation can be verified by the smallest λ and L are obtained. p Note that for λ Ê 2,
λ
comparing its values against the values offered by pre- we approximate C2 by LM Ê NW(W − 1) · · · (W − λ)/M
viously published BER equations [28]–[30]. As example, to obtain an explicit boundary expression for L. This
Fig. 2 compares the accurate BER and its approximation approximation is sufficiently accurate if L À λ which is a
in terms of the number of users for three single-class valid statement for practical values of L and λ. For λ = 1,
1D OCDMA systems distinguished by different values of C2 provides an explicit expression for L and there is no
cross-correlation λ = 1, 2, 3. need to apply the approximation. In addition, one can use
the approximated BER (6) to provide an initial value to
speed up local searches around the intersections.
IV. S YSTEMATIC D ESIGN P ROCEDURES
In this part, we formulate two design problems to
find the optimum values of code parameters that result
in maximum transmission rate or minimum transmitted B. Power-Optimized Design Problem
optical power. We refer to these formulations as "Rate-
optimized" and "Power-optimized" design problems. Again consider the 1D/2D OCDMA-PON system with
chip rate R c , N users and M wavelengths with GOOCs
and now assume that code length L (equivalently trans-
A. Rate-Optimized Design Problem
mission rate R) is given. Clearly for fixed values of L
Assume a 1D/2D OCDMA-PON system having chip rate and M, average transmitted optical power is proportional
R c , N users, M wavelengths using GOOC codes. We want to code weight W. We want to find a pair of W and λ
to determine the values of L, W and λ such that code that minimizes transmitted optical power (or equivalently
design and cardinality constraints are met, BER is less minimizes code weight) and holds BER, code design and
than a predefined value P e th and transmission rate is code cardinality constraints. The problem is formulated as

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

Algorithm 1 Heuristic procedure for designing a rate- user codes to exploit the unused resources of inactive
optimized OCDMA system. users. The algorithms have two main parts which are
Input: Number of users N, Number of wavelength M, Error named offline and online stages. The first algorithm shares
threshold P e th the unused resources of the inactive users among active
Output: Code Length L, Code weight W, Code cross-correlation
λ ones to increase the transmission rate while the other
algorithm adaptively reduces the code weight to decrease
1: λ ←− 1, λ t ←− 1; the amount of transmitted optical power when the number
2: search for the smallest W such that P e (1, M, N, L,W, λ, 1, 1) É of inactive users falls in predefined integer intervals.
†
P e th where L = (NW(W − 1) + M)/M 2 ;
3: while 1 do
4: λ t ←− λ t + 1; A. Rate-Efficient Adaptive Code Allocation Algorithm
5: search for the smallest Wt1 such that Consider a single class 1D/2D OCDMA-PON system
P e (1, M, N, L t1 ,Wt1 , λ t , 1, 1) É P e th where that works with M wavelengths and GOOCs, has at most
λ †
L t1 = ( t NWt1 · · · (Wt1 − λ t )/M)/M ± j, j = 1, · · · , λ t ;
p
N active users and its BER is upper-limited by P e th .
2 /(2M λ ) > L
if Wt1 Assume that the system has a bank of optical codes that
6: t t1 then
7: L t1 ←− ∞; includes different code books corresponding to various code
8: end if parameters L, W and λ. Each code book has a unique
9: search for the smallest Wt2 such that ID number. There are many methods that can be used
P e (1, M, N, L t2 ,Wt2 , λ t , 1, 1) É P e th where to construct the code bank [32], [38]. Based on (6), BER
L t2 = Wt22 /(2M λ† ); approximately remains unchanged while the ratio N L is
t
10: if N > [M(ML t2 − 1) · · · (ML t2 − λ t )]/[Wt2 · · · (Wt2 − λ t )] then fixed which implies that code length and consequently
11: L t2 ←− ∞; transmission rate can adaptively be improved according to
12: end if the number of active users. In our rate-efficient OCDMA-
13: if min{L t1 , L t2 } < L then PON, the control unit which resides in OLT, dynamically
14: λ ←− λ t ; monitors the system nodes, receives control messages,
15: if L t1 < L t2 then runs the proposed routine in Alg. 2 and allocates available
16: W ←− Wt1 , L ←− L t1 ; resources among the active users as illustrated in Fig. 3.
17: else The algorithm has an offline stage in which the optimum
18: W ←− Wt2 , L ←− L t2 ; values of code parameters for maximizing transmission
19: end if rate in an OCDMA-PON with i, i = 1, · · · , N active users
20: else are computed using formulation (7) and Alg. 1. The offline
21: break; stage runs when an important change in the status of
22: end if the OLT and/or ONUs occurs such as rebooting the OLT
23: end while or adding the ONU of a new user. In online stage, the
24: store L, W, λ; control unit polls all registered ONUs and updates the
† use approximation (6) to choose an initial value and speed up local
number of active users by means of the received activation
search.
and deactivation messages. At predefined time intervals
separated by code reallocation time T, a new code book
below: corresponding to the number of active users is selected and
announced. Next, optical decoders and encoders are tuned
min W (8) according to the announced code assignment messages.
W,λ
 The complexity of Alg. 2 mainly depends on the offline


 C1 : λ É W É ML stage which in turn is related to the maximum number
− 1) · · · (W − λ) É M(ML − 1) · · · (ML − λ)

C2 : NW(W of active users and the search space of the rate-optimized


p

s.t. C3 : W É 2M λL design problem.

C4 : P e (1, M, N, L,W, λ, 1, 1) É P e th




B. Power-Efficient Adaptive Code Allocation Algorithm

C5 : W, λ ∈ N

Again, assume that we have a single class 1D/2D
A brute force search over 2D region limited to Wmax and
OCDMA-PON with M wavelengths, at most N users and
λmax can yield the optimal solution of (8) by Wmax λmax
BER threshold P e th . If an OCDMA transmitter uses a code
evaluation of constraints. Since the number of search
with smaller code weight W, it needs lower transmitted
points is small, it is easy to find the optimum point using
optical power. As can be seen from (6), if the number of
a simple search process.
active users is sufficiently low, transmitted optical power
can be reduced by reallocating a new code book with
V. A DAPTIVE C ODE A LLOCATION A LGORITHMS lower code weight such that BER remains less than the
Here, we design an OCDMA-PON using the introduced target value. In offline phase of Alg. 3, an OCDMA system
design procedures and then propose two simple and adap- with given parameters N, M and P e th are design using
tive code allocation algorithms that dynamically reallocate Alg. 1. Then, formulation (8) is used to find optimum

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

Activation Algorithm 3 Power-efficient adaptive code allocation al-

Message:
My transmit gorithm for OCDMA-PON system.
buffer is full

Deactivation
Input: User activity information, Number of wavelength M,
Message: OLT Maximum number of active users N, error threshold P e th ,
ONU Transmitter 1 My transmit OLT Receiver 1
buffer is empty code reallocation interval T
Output: Code assignment messages
ONU Transmitter 2 OLT Receiver 2
Combiner Splitter
Offline Stage:
Code 1: design an OCDMA system with N users, M wavelengths and
Assignment
Message: Control error threshold P e th using formulation (7) and name obtained
ONU Transmitter N Unit OLT Receiver N
Use code book
with IDXXX
code length, code weight and cross-correlation L, W and λ;
2: for i = 1, · · · , N do
3: design an OCDMA system with i users, M wavelengths,
Figure 3: Control unit and messages in the proposed adaptive OCDMA-PON
system running algorithm Alg. 2 or Alg. 3. code length L and error threshold P e th using formulation
(8) and name obtained code weight and cross-correlation
Wi and λ i ;
4: end for
Algorithm 2 Rate-efficient adaptive code allocation algo- 5: n ←− N;
rithm for OCDMA-PON system. 6: choose code book (L,Wn , λ n );
7: send code assignment messages;
Input: User activity information, Number of wavelength M,
8: T cr ←− time() + T;
Maximum number of active users N, error threshold P e th ,
code reallocation interval T Online Stage:
9: while 1 do
Output: Code assignment messages
10: if a user activation message is received then
Offline Stage: 11: n ←− n + 1;
1: for i = 1, · · · , N do 12: else if a user deactivation message is received then
2: design an OCDMA system with i users, M wavelengths 13: n ←− n − 1;
and error threshold P e th using formulation (7) and name 14: end if
obtained code length, code weight and cross-correlation L i 15: if time() Ê T cr then
, Wi and λ i ; 16: choose code book (L,Wn , λn );
3: end for 17: send code assignment messages;
4: n ←− N;
18: T cr ←− time() + T;
5: choose code book (L n ,Wn , λ n );
6: send code assignment messages;
19: end if
20: end while
7: T cr ←− time() + T;
Online Stage:
8: while 1 do
9: if a user activation message is received then
VI. S IMULATION R ESULTS
10: n ←− n + 1;
11: else if a user deactivation message is received then Without loss of generality and for the sake of simplicity,
12: n ←− n − 1; we restrict our simulations to a 1D OCDMA system. We
13: end if first focus on rate-optimized and power-optimized design
14: if time() Ê T cr then problems and then investigate the performance of the
15: choose code book (L n ,Wn , λn ); proposed rate-efficient and power-efficient adaptive code
16: send code assignment messages; allocation algorithms.
17: T cr ←− time() + T; To evaluate the performance of Alg. 1, its results are
18: end if compared with brute force search. Fig. 4, Fig. 5 and Fig. 6
19: end while show the optimum values of W, λ and L in a 1D OCDMA
system design for various P e th = 10−7 , 10−9 , 10−11 versus
number of users N obtained by Alg. 1. Simulation results
show that there is no difference between the results of
Alg. 1 and brute force search approach for W and λ. For
values of W and λ for a fixed value of L and various
large values of N, the offered L by Alg. 1 is slightly greater
values of number of users i, i = 1, · · · , N. Similarly in
than the optimum value obtained by the brute force search
the online stage, number of active users is determined
but its relative error is less than 3.29%. Our simulation
using activation and deactivation messages and at certain
results show that λ mainly takes 2 or 3 as its optimum
time interrupts, a new code book that corresponds to the
value which is the same as the results published in [25].
number of active users is selected and announced. Fig.
However, for lower values of N, the optimum value of λ
3 illustrates the adaptive OCDMA-PON, its control unit
may be greater than 3. To investigate the complexity of
and control messages. The complexity of Alg. 3 mainly
Alg. 1, we define complexity gain G com as:
depends on the offline stage which in turn is related to the
maximum number of active users and the search space of M AC B
the power-optimized design problem. G com = (9)
M AC H

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

120
where M AC B and M AC H are number of required Multi-
ply–ACcumulate (MAC) operations in the rate-optimized
brute force search and Alg. 1, respectively. G com shows 100

how Alg. 1 is faster than the brute force search which is

an important factor for on-demand and adaptive system 80
design and resource management. Fig. 7 shows G com for
Alg. 1 compared to the brute force search over the region

Wopt
limited to Wmax , λmax and L max . Clearly, Alg. 1 reduces the 60

computational complexity. For example when L max = 4000,

Wmax = 100 and λmax = 5, the complexity gain G com is 40
more than 104 which means that Alg. 1 is 104 times faster
than its brute force search counterpart.
The optimum values of W and λ versus number of users 20
Peth = 10!11
N for power-optimized design problem are plotted in Fig. Peth = 10!9
Peth = 10!7
8 and Fig. 9, respectively. To obtain these figures, we first 0
10 20 30 40 50 60 70 80 90 100
design three OCDMA systems with N = 70 and P e th = N
10−7 , 10−9 , 10−11 using the rate-optimized formulation (7)
Figure 4: Optimum value of W in a rate-optimized 1D OCDMA system design for
to obtain code lengths which are L = 1334, 1718, 2096. various P e th = 10−7 , 10−9 , 10−11 versus number of users N .
Then, power-optimized brute force search for N = 1, · · · , 70
and the obtained values of L is executed. Note that the
jumps for W in Fig. 4 and Fig. 8 correspond to points rate) and the amount of energy injected to each ON chip.
in which λ changes. Based on (6), the exponent W λ
has a Fig. 10 shows G rat versus P active for different values of
significant effect on BER values and for any change in λ, error threshold P e th . Running Alg. 2, the transmission rate
W can change such that W λ
and consequently BER remain increases for lower values of P active since the control unit
constant. shares the unused resources of the inactive users among
To evaluate the performance of the rate-efficient and active users to increase transmission rate. For instance,
power-efficient code allocation algorithms, consider a 1D when P e th = 10−7 and P active = 0.5, G rat = 2.53 which
OCDMA system with at most N = 70 active users in means that the transmission rate is 2.53 times higher
which each user is active in code reallocation interval T than the scenario in which no adaptive code allocation is
with the probability of P active . Also assume that P active ’s available. Furthermore, the amount of transmission rate
are i.i.d over different code reallocation intervals and improvement is enhanced for higher values of P e th . Fig.
different users. We define two performance metrics named 11 shows G pow versus P active for various values of error
transmission rate gain G rat and optical power gain G pow threshold P e th . Clearly using Alg. 3, the amount of optical
to quantize the amount of improvements due to the rate- power reduces as P active decreases and the amount of
efficient and power-efficient code allocation algorithms, power reduction improves if the system can tolerate higher
respectively. G rat equals to: BER values. For instance, when P e th = 10−7 and P active =
0.5, the optical power is reduced by a factor of G pow = 0.25
RA
G rat = (10) for adaptive code reallocation scenario compared to fixed
RF
code assignment one. As demonstrated by the simulation
where R A is the transmission rate in the proposed adap- results, the proposed algorithms reduce the computational
tive OCDMA-PON system running Alg. 2 and R F is the complexity required for dynamic resource allocation on a
transmission rate in the simple OCDMA with fixed code PON system. Therefore, they are more useful for prac-
allocation. G pow is: tical implementation in PON systems. Furthermore, the
proposed algorithms can be combined with other green
PA
G pow = (11) communication PON ideas such as using sleep modes in
PF Optical Network Terminal (ONT) operation.
where P A is the average transmited optical power in the
proposed adative OCDMA-PON system running Alg. 3 VII. C ONCLUSION
and P F is the average transmited optical power in the In this paper, we provide an approximate simple expres-
simple OCDMA-PON with fixed code allocation. Clearly, sion for BER of a multi-class OCDMA system and verify
higher values of G rat are desired while lower values of its accuracy against the previously published complex
G pow are more favorable. In our simulations, we have BER expressions. The proposed approximation is valid for
increased transmitted optical power in each chip such that 1D/2D OCDMA systems with generalized or strict optical
MAI-limited condition becomes valid [33]. P F and P A are orthogonal codes. It also supports multi-class OCDMA
the amount of optical power used for filling ON chips systems having QoS classes differentiated by system level
in the adaptive and fixed OCDMA-PON and their values techniques such as multi-length variable-weight, multi-
are related to the number of ON chips per time (which level and diversity signaling. We then provide some ex-
in turn is related to code weight, code length and chip amples to show the applications of the derived BER

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

5 40

4.5 35

4 30

3.5 25

Wopt
6opt

3 20

2.5 15

2 10

1.5 Peth = 10!11 5 Peth = 10!11 ; L = 2096

Peth = 10!9 Peth = 10!9 ; L = 1718
Peth = 10!7 Peth = 10!7 ; L = 1334
1 0
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70
N N

Figure 5: Optimum value of λ in a rate-optimized 1D OCDMA system design for Figure 8: Optimum value of W in power-optimized 1D OCDMA system design
various P e th = 10−7 , 10−9 , 10−11 versus number of users N . versus number of users N for fixed values of L and various error thresholds P e th .

3
3500

3000

2500
2

2000
6opt
Lopt

1500

1000

500 Peth = 10!11 ; L = 2096

Peth = 10!11
Peth = 10!9 ; L = 1718
Peth = 10!9
Peth = 10!7 ; L = 1334
Peth = 10!7
0
0 10 20 30 40 50 60 70
10 20 30 40 50 60 70 80 90 100
N
N

Figure 9: Optimum value of λ in power-optimized 1D OCDMA system design

Figure 6: Optimum value of L in a rate-optimized 1D OCDMA system design for
versus number of users N for fixed values of L and various error thresholds P e th .
various P e th = 10−7 , 10−9 , 10−11 versus number of users N .

10
5
10 Peth = 10!11
Peth = 10!9
9 Peth = 10!7

7
104

6
Grat
Gcom

4
3
10

2
Wmax = 100; 6max = 15
Wmax = 100; 6max = 10
Wmax = 100; 6max = 5 1
102 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Pactive
Lmax

Figure 10: Transmission rate gain G rat versus active probability P active for various
Figure 7: Complexity gain G com of Alg. 1 compared to the rate-optimized brute values of BER threshold P e th and maximum number of users N = 70 in an adaptive
force search over 3D region limited to Wmax , L max and λmax . OCDMA-PON system running rate-efficient adaptive code allocation Alg. 2.

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

1
where: ¡W − i¢
0.9 W2 h λ
i
f (i) = 1 − 1 − ¡W (14)
2M λL
¢
0.8 λ

Obviously, f (i) can be considered as a λ-th order polyno-

0.7
mial with coefficients f t , t = 0, · · · , λ and λ is usually chosen
0.6 between 1 to 3 [25]. The constant coefficient of f (i) is
Gpow

f 0 = 1. In practical implementations of an OCDMA system,

0.5
the code length L is usually a large value and therefore,
0.4
| f t | ¿ 1, t = 1, 2, · · · , λ. Furthermore, one can check that f t
is positive and negative for even and odd values of t, i.e.
0.3 f t = (−1) t | f t |. Now, we use Newton multinomial expansion
Peth = 10!11
to reform (13), as can be seen in (15) at the top of the
0.2
Peth = 10!9
Peth = 10!7
next page. Noting the identity in (12), the first non-zero
Pλ
0.1
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 term of (15) is obtained when tk t = W. We use this first
Pactive t=1
non-zero term to approximate BER:
Figure 11: Optical power gain G pow versus active probability P active for various
λ
à !
values of BER threshold P e th and maximum number of users N = 70 in an adaptive W! N −1
| f t |k t
X Y
OCDMA-PON system running power-efficient adaptive code allocation Alg. 3. Pe ≈ (16)
2 k0 +k1 +...+kλ = N −1 k 0 , k 1 , ..., k λ t=1
k 1 +2k 2 +...+λ k λ =W

equation in providing guidelines for analysis and design For sufficiently large values of N, we approximately have
¡ N −1 ¢ N k1 +···+kλ
of OCDMA-based systems. We especially formulate two k 0 ,k 1 ,...,k ≈ k !...k ! . Therefore:
λ 1 λ
design problems for maximizing transmission rate or mini-
1 W! λ
mizing transmitted optical power under constraints of code | N f t |k t
X Y
Pe ≈ (17)
design, code cardinality and targeted system performance. 2 k1 +2k2 +...+λkλ =W k 1 !...k λ ! t=1
We also use the BER equation to interpret the sensitivity
Defining u i = ik i , i = 1, 2, · · · , λ and noting that | f t | ¿ 1, we
of BER performance to various system parameters and use
have:
the results to propose two adaptive code allocation algo-
1 W! λ ut
rithms for OCDMA-PONs. Simulation results demonstrate Pe ≈
X Y
|N f t | t
uλ
that we can benefit from user and traffic variation over 2 u1 +...+uλ =W (u 1 )!...( λ !) t=1
time to adaptively adjust system resources and improve 1³ Xλ 1 W
´ 1 W
the transmission rate (improvement up to a factor of 10) ≈ |N f t | t ≈ |N fλ| λ (18)
2 t=1 2
or reduce transmitted optical power (reduction by a factor
of up to 0.1). The proposed BER expression can be used in Substituting the value of f λ in (18) results in:
analysis and design of various types of OCDMA systems
1³ NW 2 ´W
λ 1 ³ NW ´ Wλ
and there is a good potential for its application in many Pe ≈ . (19)
2 2M λLW · · · (W − λ + 1) 2 2M λL
other scenarios.

A PPENDIX A B. 2-class OCDMA system

BER A PPROXIMATION Now, we derive the approximation for BER of the first
Through this appendix, we derive an approximated class in an OCDMA(2, M, N, L, W, Γ, 1, 1) system. The same
expression for BER of a K-class OCDMA system charac- method can be applied to the second class to yield its
terized by OCDMA(K, M, N, L, W, Γ, C, B). We first derive corresponding BER approximation. BER of the first class
the approximation for a simple 1-class OCDMA system of the desired 2-class OCDMA system is upper limited by
OCDMA(1, N, L,W, λ, 1, 1) and then extend it to work for a [30]: Ã !
W1
1X i W1
general 2-class and K-class OCDMA system. The following P e(1) É (−1) [ f (i)] N1 −1 [g(i)] N2 (20)
identity plays a key role in deriving the approximations 2 i=0 i
[29]: where:
à ! (
W λ
W12
à !à !
X i W t 0 if t = 0, · · · ,W − 1 i W1 − λ11
11
j
(−1) (− i) = (12)
X
f (i) = 1 + (−1) (21)
i =0 i W! if t = W 2M λ11 L 1 λW111 j=1
¡ ¢
j λ11 − j

λ
à !à !
A. 1-class OCDMA system W1 W2 j i W1 − λ12
X12
g(i) = 1 + (−1) (22)
The BER of a 1-class 2D OCDMA system is upper 2M λ12 L 2 W1 j=1 j λ12 − j
¡ ¢
λ12
limited by [25] (for 1D OCDMA system, simply set M = 1):
à ! Again, f (i) and g(i) can be considered as two polynomials
1X W
i W having orders of λ11 and λ12 with coefficients f s , s =
Pe É (−1) [ f (i)] N −1 (13)
2 i=0 i 0, · · · , λ11 and g t , t = 0, · · · , λ12 , respectively. Similar to the

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

10

λ
λ λ
à ! à ! à ! à !
1X W W
i W N −1 1 N −1 i W
P
tk t
t kt kt
X Y X Y X
Pe É (−1) (ft i ) = | f t| (−1) (− i)t=1
2 i=0 i k0 +...+kλ = N −1 k 0 , k 1 , ..., k λ t=0 2 k0 +k1 +...+kλ = N −1 k 0 , k 1 , ..., k λ t=1 i =0 i
(15)

λ λ
à ! à ! à !
W1
1X i W1 N1 − 1 11 N2 12
( f s i s )m s (g t i t )n t
X Y X Y
P e(1) É (−1) (23)
2 i=0 i m0 +...+mλ = N1 −1 m 0 , ..., m λ11 s=0 n 0 +...+ n λ = N2 n 0 , ..., n λ12 t=0
11 12
λP11 λP12
λ λ
à ! à ! à !
W1
1 N1 − 1 11 N2 12
i W1
sm s + tn t
| f s |m s nt
X Y X Y X
= | gt| (−1) (− i) s=1 t=1
2 m0 +...+mλ = N1 −1 m 0 , ..., m λ11 s=0 n 0 +...+ n λ = N2 n 0 , ..., n λ12 t=0 i =0 i
11 12

λ λ
W1 ! X 1 11 1 12
| N1 f s | m s | N2 g t | n t
X Y X Y
P e(1) ≈
2 u+v=W m1 +2m2 +···+λ11 mλ = u m 1 ! · · · m λ11 ! s=0 n 1 +2n 2 +···+λ12 n λ =v n 1 ! · · · n λ12 ! t=0
11 12
1 X W1 ! u v 1 1 1
≈ | N1 f λ11 | λ11 | N2 g λ12 | λ12 = (| N1 f λ11 | λ11 + | N2 g λ12 | λ12 )W1 (24)
2 u+v=W u!v! 2

1 h³ N1 W12 ´ 1
λ11
³ N2 W1 W2 ´ 1 iW1 1 h³ N W ´ 1
λ12 1 1 λ11
³ N W ´ 1 iW1
2 2 λ12
P e(1) ≈ + . +
2 2M λ11 L 1 W1 · · · (W1 − λ11 + 1) 2M λ12 L 2 W1 · · · (W1 − λ12 + 1) 2 2M λ11 L 1 2M λ12 L 2
(25)

previous derivation, we have f 0 = 1, g 0 = 1, f s = (−1)s | f s | [5] S. Khazraei, M. R. Pakravan, Analysis of Generalized Optical Orthog-
and g t = (−1) t | g t |. As shown in (23) at the top of the page, onal Codes in Optical Wireless Local Area Networks, IEEE Journal
on Selected Area in Communications, vol. 27, no. 9, pp. 1572-1581,
we use Newton multinomial expansion to reform (20). Dec. 2009.
λP 11 λP
12
Assume u = sm s and v = tn t . Noting the identity [6] M. H. Shoreh, A. Fallahpour, J. A. Salehi, Design Concepts and
s=1 t=1 Performance Analysis of Multicarrier CDMA for Indoor Visible Light
of (12), the first non-zero term of (23) is obtained when Communications, Journal of Optical Communications and Network-
ing, vol. 7, no. 6, pp. 554-562, Jun. 2015.
u + v = W. We proceed with the same way as the previous
[7] K. Kitayama, X. Wang, N. Wada, OCDMA Over WDM PON–Solution
sub-section to approximate BER with its first non-zero Path to Gigabit-Symmetric FTTH, Journal of Lightwave Technology,
term, as can be seen in (24) at the top of the page. vol. 24, no. 4, pp. 1654 -1662, Apr. 2006.
Substituting the values of f λ11 and g λ12 in (24) results [8] E. Wong, Next-Generation Broadband Access Networks and Technolo-
gies, IEEE Journal of Lightwave Technology, vol. 30, no. 4, pp. 597-
in the desired approximation expression, shown in (25) at 608, Feb. 2012.
the top of the page. [9] W. Chang, G. Yang, C. Chang, W. C. Kwong, Enhancing Optical-
CDMA Confidentiality With Multicode-Keying Encryption, Journal of
C. K-class OCDMA System Lightwave Technology, vol. 33, no. 9, pp. 1708-1718, Jan. 2015.
[10] R. Matsumoto, T. Kodama, S. Shimizu, R. Nomura, K. Omichi,
The same procedure as previous sub-sections can be N. Wada, K. Kitayama, 40G-OCDMA-PON System with an Asym-
used to provide the following approximated expression for metric Structure Using a Single Multi-port and Sampled SSFBG
Encoder/Decoders, Journal of Lightwave Technology, vol. 32, no. 6,
BER of an OCDMA(K, M, N, L, W, Γ, C, B) system: pp. 1132-1143, Mar. 2014.
K ³ N W
1h X ´ c ki iWk b k [11] J. Liu, G. Yang, H. Chen, Y. Chen, C. Chang, W. C. Kwong, New Tree-
i i λ ki
P e(k) . (26) Structured Optical Codes for Incoherent Optical-CDMA Applications,
2 i=1 2M λki L i Journal of Lightwave Technology, vol. 33, no. 19, pp. 3968-3979, Oct.
2015.
R EFERENCES [12] C. Chang, W. C. Kwong, Two New Families of Synchronous Optical
Codes for CDMA-based Passive Optical Networks, IEEE Transactions
[1] Y. Lin, G. Yang, C. Chang, W. C. Kwong, Design and Analysis of
on Communications, vol. 62, no. 5, pp. 1646-1657, May. 2014.
Asynchronous Incoherent Optical-CDMA Systems Using a New Code-
[13] H. Chen, S. Xiao, L. Yi, J. Shi, P. Yang, Y. Dong, W. Hu, Complemen-
Shifting Technique, IEEE Transactions on Communications, vol. 63,
tary Decoder Based on Polarization Modulation for the SAC-OCDMA
no. 7, pp. 2619-2631, Jul. 2015.
PON, IEEE Photonics Technology Letters, vol. 24, no. 5, pp. 335-337,
[2] H. Chen, M. Chang, G. Yang, .W. C. Kwong, Design of a Hybrid
Mar. 2012.
Multirate Asynchronous O-CDMA System : Performance , Spectral
Efficiency , and Throughput Analyses, Journal of Optical Communi- [14] F. Chen, F. Luo, B. Li, C. He, Modified Decoding and Detection
cations and Networkings, vol. 7, no. 11, pp. 1034-10481, Nov. 2015. Scheme for Spectral Amplitude Coding Optical CDMA Systems,
[3] M. H. Shoreh, H. Beyranvand, J. A. Salehi, Performance Evalua- Electronics Letters, vol. 51, no. 10, pp. 773–775, May. 2015.
tion of Asynchronous Multi-Carrier Code Division Multiple Access [15] A. G. Sabbagh, M. M. Kakhki, Performance Analysis of Two-Level
for Next-Generation Long-Reach Fibre Optic Access Networks, IET Asynchronous Optical CDMA Systems Utilizing Wrapped OPPM,
Optoelectronics, vol. 9, no. 6, pp. 1-8, Mar. 2015. Journal of Lightwave Technology, vol. 32, no. 1, pp. 122-129, Jan.
[4] J. S. Vardakas, I. D. Moscholios et al., Performance Analysis of 2014.
OCDMA PONs Supporting Multi-Rate Bursty Traffic, IEEE Trans- [16] T. Kodama, N. Wada, G. Cincotti, K. Kitayama, Asynchronous
actions on Communications, vol. 61, no. 8, Aug. 2013. OCDM-Based 10 G-PON Using Cascaded Multiport E/Ds to Sup-

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JLT.2017.2705035, Journal of
Lightwave Technology

11

press MAI Noise, Journal of Lightwave Technology, vol. 31, no. 20, [36] A. E. Willner, S. Khaleghi, M. R. Chitgarha, O. F. Yilmaz, All-
pp. 3258 - 3266 , Sep. 2013. Optical Signal Processing, Journal of Lightwave Technology, vol. 32,
[17] V. J. Hernandez, W. Cong, J. Hu, C. Yang,N. K. Fontaine, R. P. no. 4, pp. 660-680, Feb. 2014.
Scott, Z. Ding, B. H. Kolner, J. P. Heritage, S. J. B. Yoo, A 320- [37] D. Hood, E. Trojer, Gigabit-capable passive optical networks, John
Gb/s Capacity (32-user × 10 Gb/s) SPECTS O-CDMA Network Wiley & Sons, 2012.
Testbed with Enhanced Spectral Efficiency Through Forward Error [38] H. Cai, H. Liang, X. Tang, Constructions of Optimal 2-D Optical
Correction, Journal of Lightwave Technology, vol. 25, no. 1, pp. 79–86, Orthogonal Codes via Generalized Cyclotomic Classes, IEEE Trans-
Jan. 2007. actions on Informaton Theory, vol. 61, no. 1, Jan. 2015.
[18] G. Cincotti, N. Kataoka, N. Wada, X. Wang, T. Miyazaki, K. I.
Kitayama, Demonstration of Asynchronous, 10 Gbps OCDMA PON
System with Colorless and Sourceless ONUs, in ECOC, Vienna,
Austria, Sep. 2009.
[19] N. Cvijetic, A. Tanaka, P. N. Ji, K. Sethuraman, S. Murakami, and
T. Wang, SDN and OpenFlow for Dynamic Flex-grid Optical Access
and Aggregation Networks, Journal of Lightwave Technology, vol. 32,
no. 4, pp. 864-870, Feb. 2014.
[20] S. W. Lee, K. Li, M. Wuu, Design and Implementation of a GPON-
Based Virtual OpenFlow-Enabled SDN Switch, Journal of Lightwave
Technology, vol. 34, no. 10, pp. 2552-2561, Apr. 2016.
[21] S. McGettrick, F. Slyne, N. Kitsuwan, D. B. Payne, M. Ruffini,
Experimental End-to-End Demonstration of Shared N:1 Dual Homed
Protection in Long Reach PON and SDN-Controlled Core, Optical
Fiber Communication Conference, pp. Tu2E-5, Mar. 2015.
[22] M. Taheri, J. Zhang, N. Ansari, Design and Analysis of Green
Optical Line Terminals for TDM Passive Optical Networks, Journal Mohammad Hadi received his B.Sc. degree as
of Optical Communication and Networking, vol. 8, no. 4, pp. 221-228, a top-ranking graduate from the Isfahan Uni-
Apr. 2016. versity of Technology, Isfahan, Iran in 2011. He
[23] A. R. Dhaini, P. Ho, G. Shen, Toward Green Next-Generation Passive continued his studies in Sharif University of
Optical Networks, IEEE Communications Magazine, vol. 49, no. 11, Techonology, Tehran, Iran where he received his
pp. 4-101, Nov. 2011. M.Sc. in 2013 and now, he is a Ph.D. student.
[24] D. Suvakovic, H. Chow, N. P. Anthapadmanabhan, D. T. Veen, A. He is also a member of Data Networks Research
J. Wijngaarden, T. Ayhan, C. Praet, G. Torfs, X. Yin, P. Vetter, A Lab (DNRL). His main research interests are
Low-Energy Rate-Adaptive Bit-Interleaved Passive Optical Network, optical CDMA and optical OFDM communication
Journal on Selected Areas in Communications, vol. 32, no. 8, pp. systems.
1552-1565, Aug. 2014.
[25] S. Mashhadi, J. A. Salehi, Code-Division Multiple-Access Techniques
in Optical Fiber Network–Part III: Optical AND Logic Gate Receiver
Structure With Generalized Optical Orthogonal Codes, IEEE Trans-
actions on Communications, vol. 54, no. 8, pp. 1457-1468, Aug. 2006.
[26] W. C. Kwong, G. Yang, V. Baby, C. Bres, P. R. Prucnal, Multiple-
Wavelength Optical Orthogonal Codes under Prime-Sequence Permu-
tations for Optical CDMA, IEEE Transactions on Communications,
vol. 53, no. 1, pp. 117 - 123, Feb. 2005.
[27] M. Hadi, M. R. Pakravan, Adaptive Level Control to Improve QoS
in OCDMA Local Area Networks, in Proceedings of the 21st Iranian
Conference on Electrical Engineering, ICEE2013, May 2013.
[28] B. Ghaffari, J. A. Salehi, Multiclass, Multistage, and Multilevel
Fiber-Optic CDMA Signaling Techniques Based on Advanced Binary
Optical Logic Gate Elements, IEEE Transactions on Communica-
tions, vol. 57, no. 5, pp.1424-1432, May 2009.
[29] C. Chang, G.g Yang, C. Chang, W. C. Kwong, Study of a Diversity O- Mohammad Reza Pakravan was born in
CDMA Scheme for Optical Wireless, Journal of Lightwave Technology, Teheran, Iran. He received his B.Sc. degree as
vol. 30, no. 10, pp. 1549-1558, May 2012. a top-ranking graduate from the University of
[30] H. Beyranvand, B. M. Ghaffari, J A. Salehi, Multirate, Tehran, in 1991. He continued his studies in
Differentiated-QoS, and Multilevel Fiber-Optic CDMA System via Canada where he received his M.Sc. and Ph.D.
Optical Logic Gate Elements, Journal of Lightwave Technology, vol. degrees in Electrical Engineering from the Uni-
27, no. 19, pp. 4348-4359, May 2009. versity of Ottawa, in 1994 and 2000, respectively.
[31] A. E. Farghal, H. M. H. Shalaby, Z. Kawasaki, Performance Analysis He received the University of Ottawa’s excellence
of Multirate Multiservice Optical CDMA Networks Adopting Overlap- scholarship as well as three consecutive Ontario
ping PPM Signaling , Journal of Lightwave Technology, vol. 32, no.15, Graduate Scholarships for his academic achieve-
pp. 2649-2658, Aug. 2014. ments during his PhD research. He is currently
[32] J. A. Salehi, Emerging OCDMA Communication Systems and Data a faculty member of Electrical Engineering Department and director of
Networks, Journal of Optical Networking, vol. 6, no. 9, pp. 1138-1178, Data Networks Research Lab (DNRL), Sharif University of Technology,
Sep. 2007. Tehran, Iran. His research interests include optical CDMA and optical
[33] S. Galli, R. Menendez, E. Narimanov, P. R. Prucnal, A Novel OFDM communication systems.
Method for Increasing the Spectral Efficiency of Optical CDMA, IEEE
Transactions on Communications, vol. 56, no. 12, pp. 2133-2144, Dec.
2008.
[34] C. S. Bres, Y. K. Huang, D. Rand, I. Glesk, P. R. Pruc-
nal, T. M. Bazan, C. Michie, D. Harle, I. Andonovic, On the
Experimental Characterization of Beat Noise in 2-D Time-Spreading
Wavelength-Hopping OCDMA Systems, IEEE Photonics Technology
Letter, vol. 18, no. 21, pp. 2314-2316, Nov. 2006.
[35] P. Ebrahimi, K. Yu, M. C. Hauer, A. E. Willner, O. Solgaard, Tunable
Wavelength Demultiplexer and OCDMA Code Hopping Using a 10-
µ s-Tuning MEMS-Actuated Gires–Tournois Filter, IEEE Photonics
Technology Letters, vol. 18, no. 12, pp. 1398-1400, Jun. 2006.

0733-8724 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.