Spreading Codes for All-Optical Code

Division Multiple Access Communication

Systems

Seminar Report

by

Ravi
2014UEC01

Under the supervision of

Professor ....

Electronics & Communication Engineering Department

Malaviya National Institute of Technology, Jaipur
April 2017
1 Introduction
The development of the economy and society needs improved performances of communi-
cation networks, in particular optical communication networks. Optical backbone net-
works have been able to afford the requirement of bandwidth, whereas access networks
are still the bottleneck and will be of most importance in network design and construc-
tion. In code division multiple access (CDMA) network design, the error due to multiple
access interference (MAI) is an important factor which is mitigated to some extent by
generating almost orthogonal codes having very small autocorrelation and crosscorrela-
tion values. Unlike CDMA, the optical CDMA (OCDMA) design is based on binary,
unipolar spreading codes. The optical codes require considerably longer spreading length
in order to satisfy these correlation constraints. Spreading the optical codes in wave-
length, time and multiple fibers is observed to satisfy the constraints for accommodating
sufficiently large number of users with a comparatively smaller spreading code length.
In all-optical CDMA design, the information in electrical domain is spread by directly
assigning a pulse pattern in optical domain in order to accommodate Terabits per second
data rate. Generation of such patterns requires multiple number of delay lines stacked
in parallel [?]. Optical fiber delay lines of variable length are difficult to integrate along
with laser technology in miniaturized form which is expected to be a demand of network
technology in future. Based on different network architectures, types of code families
can be one-dimensional (1D) temporal spread codes, two-dimensional (2D) wavelength
- time spread or space - time spread codes or three-dimensional (3D) space - wavelength
- time spread codes. The 2D code families can be classified into single pulse per row
(SPR) and multiple pulse per row (MPR) codes. Similarly, the 3D code families can be
classified into single pulse per plane (SPP) and multiple pulse per plane (MPP) codes.

2 Literature Review
In optical CDMA, a mode-locked laser produces a low duty cycle, high intensity pulse
stream at the data rate. This sequence of pulses is modulated by an optical gate, which
is driven by the information waveform. At the receiver, correlation is performed by
optical fiber delay lines [?]. Code families for OCDMA having low autocorrelation and
crosscorrelation values are termed as optical orthogonal codes (OOCs) [?]. The length,

1
weight and cardinality parameters of optical code families with low correlation values
are inter-related. For a given weight, increasing the cardinality increases the length with
a multiplication factor dependent on the weight.
Multiwavelength OOCs (MWOOCs) [?] consist of 2D codewords with every pulse
of a codeword encoded in a distinct wavelength. Three classes of MWOOCs based on
OOCs, prime codes and Reed-Solomon codes have been constructed. Using multiple
wavelengths, the requirements of fiber ribbons and multiple stars in temporal/spatial
OCDMA networks are eliminated. Multiple pulse per row (MPR) codes with optimum
threshold detection maximizes the cardinality (maximum number of users in a code
family) and spectral efficiency (ratio of cardinality to code dimension). A family of 2D
wavelength-hopping time-spreading single pulse per row (SPR) codes, which employs
wavelength hopping algebraically under prime-sequence permutations on top of time-
spreading OOCs, has been studied and analyzed in [?]. These codes allow the number
of wavelengths and code length to be chosen independently and the code cardinality
is a quadratic function of the number of wavelengths without increasing the maximum
crosscorrelation value.
A family of space/ wavelength/ time spread 3D optical codes for OCDMA networks
is reported in [?]. Codes with single pulse per plane (SPP) and multiple pulses per plane
(MPP), based on a prime sequence algorithm are considered. A 3D OCDMA transmis-
sion system that encodes data on time, wavelength, and polarization is experimentally
demonstrated in [?]. Three dimensional perfect difference codes are constructed, and a
corresponding system structure for spectral/ time/ spatial OCDMA is described in [?].
The codes, generated from the perfect difference set, can suppress the phase-induced
intensity noise and possess the MAI cancellation property.

3 Objectives of the Thesis

Though various types of 1D, 2D and 3D OCDMA code families and network architectures
are reported, a practically realisable system for fiber to the home networks has not been
reported so far. The thesis aims to bridge the gap between the present and practically
realisable OCDMA systems. The objective of the thesis can be stated as:

Design, development and testing of spreading codes for all-optical code division
multiple access communication systems that possess the property of low correla-

2
 tion, high cardinality and miniaturized generation.

To meet the objective, the problems undertaken in the thesis can be described as
follows.

Construction, analysis and performance verification of new 2D OCDMA code fam-

ilies.

Implementation aspect of new 3D code families incorporating the above 2D code

families.

Optimization of 2D and 3D code construction from cardinality point of view.

To explore miniaturization of OCDMA code generation using integrated optics.

4 Work Done
Details of the work carried out in the thesis are summarized below.

4.1 OCDMA Code Families based on a Novel RWOP algorithm

In this part, the constructions and properties of existing 2D generalized multi-wavelength

prime codes (GMWPCs), generalized multi-wavelength Reed-Solomon codes (GMWRSCs)
[?], multi-wavelength optical orthogonal codes (MWOOCs) [?] and folded OOCs are dis-
cussed. The construction and properties of existing 3D SPP code families [?] are also
discussed. A row-wise orthogonal pairs (RWOP) algorithm for wavelength and/or fiber
allocation for each code is proposed to construct 2D multipulse per row (MPR) OCDMA
code families. The constructed 2D code families are extended to 3D multipulse per plane
(MPP) OCDMA code families.
The performance of the constructed code families is analyzed and compared with
the 2D GMWRSCs, MWOOCs and 3D SPP code families. A comparison between the
constructed 2D and 3D code families is also shown. The major performance metric
considered is the probability of error due to MAI. Probability of error due to MAI of the
code families is compared based on equivalent code dimension and equivalent probability
of error due to MAI at full cardinality.
The RWOP algorithm is elaborated with the help of a flowchart and demonstrated
with an example. The proposed 2D wavelength - time code families using the RWOP

3
algorithm are represented as (W T, K 0 , 2, 1), where W is the number of wavelengths,
T is the temporal length, K 0 is the weight of the 2D codes, 2 represents the maximum
out-of-phase autocorrelation value and 1 represents the maximum crosscorrelation value.
The probability of error due to MAI of the 2D code families is dependent on the
probability of error due to MAI of the 1D OOC families being used. The upper bound
of the probability of error due to MAI of the 1D OOCs having a crosscorrelation value of
1, for a hard-limiting receiver with a threshold value of detection equal to the weight,
assuming chip synchronous case is analyzed. The upper bound of the probability of
error due to MAI for the proposed 2D code families is derived using the analyzed 1D
probability of error. The 2D code families with weights K 0 = 6 and K 0 = 4 have
probability of error below 2 107 and 3 104 respectively.
The proposed 2D wavelength - time code families using the RWOP algorithm are
extended to 3D space - wavelength - time code families, which are represented as
(S W T, K 00 , 4, 1), where S is the number of fibers, K 00 is the weight of the 3D
codes, 4 represents the maximum out-of-phase autocorrelation value and 1 represents
the maximum crosscorrelation value.
The derivation of the probability of error due to MAI for the constructed 3D code
families is also done using the 1D probability of error. The 3D constructions give prob-
ability of error due to MAI below 1014 and 1.5 108 for code families with weights
K 00 = 12 and K 00 = 8 respectively. The probability of error due to MAI of these code
families is lower than that of previously reported code families. The spectral efficiency of
the lower weight (2D: K 0 = 4 and 3D: K 00 = 8) RWOP-based code families is comparable
to previously reported 2D and 3D code families. For higher weight codes (2D: K 0 = 6
and 3D: K 00 = 12), the spectral efficiency is observed to be lower for RWOP-based code
families.

4.2 OCDMA Code Families based on a Novel CRWOP algo-

rithm

Some limitations of the RWOP algorithm are overcome by a new algorithm named
as complete row-wise orthogonal pairs (CRWOP) algorithm. The limitations of the
RWOP algorithm follow. (1) It can only be used to construct code families based on
even numbers, which means the number of wavelengths and/or fibers have to be even.

4
This functionality of the RWOP algorithm can be improved by an algorithm which can
construct code families from any given number of wavelengths and/or fibers. (2) All the
rows of the array generated by the RWOP algorithm are not complete. Some pairs of
some rows in the array generated by the RWOP algorithm are discarded to preserve the
crosscorrelation property of c 1.
The 2D and 3D OCDMA code families constructed using the CRWOP algorithm
have the same representation and correlation constraints as those of the 2D and 3D
code families constructed using the RWOP algorithm. The mathematical models of the
probability of error due to MAI for these 2D and 3D code families are also the same as
that for the code families based on the RWOP algorithm.

Figure 1: Comparison of (a) MWOOC(5 41, 5, 1, 1); Nmax = 50; Cd = 205, (b)
RWOP(10 19, 4, 2, 1); Nmax = 39; Cd = 190 and (c) CRWOP(13 15, 4, 2, 1);
Nmax = 42; Cd = 195 2D code families

Figure 1 shows the comparison of probability of error among reported (a) MWOOC
[?], (b) RWOP-based and (c) CRWOP-based 2D code families. Comparing between (a)
and (c), the probability of error of (c) is lower by a factor of 102 at a spectral efficiency
of 0.2333 for CRWOP compared to 0.2439 for MWOOC. Comparison between (b) and
(c) shows that the probability of error is almost same at a spectral efficiency of 0.2333
for CRWOP compared to 0.2053 for RWOP.
Comparison of the probability of error due to MAI of (a) SPP [?], (b) RWOP-

5
based and (c) CRWOP-based 3D code families is shown in Fig. ??. Analyzing the
characteristics between (a) and (c), the probability of error of (c) is lower by a factor of
106 at a spectral efficiency of 0.1154 for CRWOP compared to 0.1423 for SPP. From
the characteristics of (b) and (c), the probability of error is almost same at a spectral
efficiency of 0.1154 for CRWOP compared to 0.09 for RWOP.
The proposed CRWOP-based code families give better performance than the RWOP-
based code families and also better than those published earlier. The code families are
suitable for networks which are to be deployed with low error probabilities. Due to the
completeness of the wavelength/space allocation algorithm, the cardinality and spectral
efficiency of the CRWOP-based code families are marginally higher than that of the
RWOP-based code families. The probability of error of the CRWOP-based and RWOP-
based code families is equivalent to that of MWOOCs and SPP codes at full cardinality.
However, the probability of error is higher than SPP codes for small number of active
users while yielding better spectral efficiency. A comparison based on code dimension
shows that the CRWOP-based 2D and 3D code families have lower probability of error at
higher spectral efficiency as compared to the RWOP-based code families. The CRWOP-
based 2D and 3D code families have lower probability of error at lower spectral efficiency
when compared with MWOOCs and SPP codes for equivalent code dimension.

4.3 Miniaturized All-Optical CDMA Code Generation

Temporal spread 1D OCDMA codes can be generated by using a TE-TM splitter or a

3dB power splitting Y-junction. The proposed miniature 1D OCDMA code generator
using the TE-TM splitter has a length of 50mm and a width of 10mm. When the TE-TM
splitter is replaced by a 3dB power splitting Y-junction, the maximum temporal delay
between the pulses reduces along with the length of the device. The maximum temporal
delay and the pulse width determine the cardinality supported by the code generators.
A miniaturized OCDMA code generator using a zero-gap directional coupler (ZDC)
based TE-TM splitter on a standard LiNbO3 crystal is shown in Fig. ??. The code gen-
erator is designed by extending and combining the output arms of the TE-TM splitter.
The code generator splits a single input pulse into two output pulses having orthogonal
polarizations (TE and TM) spread in the time domain. The maximum achievable delay
between the pulses for this device is 40 picoseconds.
The use of a 3dB power splitting Y-junction in place of the TE-TM mode splitter to

6
generate 1D OCDMA codes is shown in Fig. ??. The intensity of an input optical pulse
to a miniaturized OCDMA code generator using Y-junction would be split equally to
the two output arms. In this case also, the output consists of two pulses spread in the
time domain. The maximum achievable delay between the pulses for this device is 30
picoseconds.
The 1D code generators are extended to 2D wavelength - time and 3D space - wave-
length - time code generators. For T = 40, 19 1D OOCs of weight 2 can be generated
leading to a 20 wavelength weight 4 RWOP-based 2D MPR OCDMA code family or a
38 wavelength weight 4 CRWOP-based 2D MPR OCDMA code family. The cardinality
of a 20 wavelength RWOP-based 2D code family is 168 and that of a 38 wavelength
CRWOP-based 2D code family is 361. For T = 30, as would be in the case of the 3dB
power splitting Y-junction based 2D code generator, 14 1D OOCs of weight 2 can be
generated leading to a 14 wavelength weight 4 RWOP-based 2D MPR OCDMA code
family or a 28 wavelength weight 4 CRWOP-based 2D MPR OCDMA code family. The
cardinality of a 14 wavelength RWOP-based 2D code family is 79 and that of a 28
wavelength CRWOP-based 2D code family is 196.
For T = 40, 19 1D OOCs of weight 2 can be generated leading to 20 wavelength, 4
S 20 fiber, weight 8 RWOP-based 3D MPP OCDMA code families or 38 wavelength,
4 S 38 fiber, weight 8 CRWOP-based 3D MPP OCDMA code families. The
cardinality of 20 wavelength RWOP-based 3D code families would vary from 336 (S = 4)
to 1506 (S = 20) and that of 38 wavelength CRWOP-based 3D code families would vary
from 722 (S = 4) to 6498 (S = 38). For T = 30, as would be in the case of the 3dB
power splitting Y-junction based 3D code generator, 14 1D OOCs of weight 2 can be
generated leading to 14 wavelength, 4 S 14 fiber, weight 8 RWOP-based 3D MPP
OCDMA code families or 28 wavelength, 4 S 14 fiber, weight 8 CRWOP-based
3D MPP OCDMA code families. The cardinality of 14 wavelength RWOP-based 3D
code families would vary from 158 (S = 4) to 485 (S = 14) and that of 28 wavelength
CRWOP-based 3D code families would vary from 392 (S = 4) to 2744 (S = 28).
The quantitative results show that the number of users is lower for the 3dB power
splitting Y-junction based devices than the TE-TM splitter based devices. The opti-
cal power calculations show lower insertion loss for the 3dB power splitting Y-junction
based devices than the TE-TM splitter based devices. Hence, the 3dB power splitting
Y-junction based devices would be suitable for OCDMA networks where the priority is

7
given to low loss and the TE-TM splitter based devices would be suitable for OCDMA
networks requiring larger cardinality. A practical advantage of using the 3dB power split-
ting Y-junction based devices is the abscence of polarization mode dispersion/distortion.

5 Conclusions
The probability of error due to MAI of the proposed code families is lower than that of
previously reported code families. The probability of error of the CRWOP-based and
RWOP-based code families is equivalent to that of MWOOCs and SPP code families
with varying code dimensions when all users are interfering. The probability of error
however is higher than SPP codes for small number of active users while yielding better
spectral efficiency. Quantitative results for the miniaturized 2D and 3D code generators
show that CRWOP-based code families support larger cardinality than the RWOP-based
code families.