All-Optical Logic Circuits Based on the

All-Optical Logic Circuits Based on the Polarization Properties of Non-Degenerate Four-Wave Mixing - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
All-Optical Logic Circuits Based on the Polarization Properties of Non-Degenerate Four-Wave Mixing
Citation
Bhardwaj, Ashish Ishwar Singh
(2001)
All-Optical Logic Circuits Based on the Polarization Properties of Non-Degenerate Four-Wave Mixing.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/CMAB-WF06.
Abstract
This thesis investigates a new class of all-optical logic circuits that are based on the polarization properties of non-degenerate Four-Wave Mixing. Such circuits would be used in conjunction with a data modulation format where the information is coded on the states of polarization of the electric field. Schemes to perform multiple triple-product logic functions are discussed and it is shown that higher-level Boolean operations involving several bits can be implemented without resorting to the standard 2-input gates that are based on some form of switching. Instead, an entire hierarchy of more complex Boolean functions can be derived based on the selection rules of multi-photon scattering processes that can form a new class of primitive building blocks for digital circuits.
Possible applications of these circuits could involve some front-end signal processing to be performed all-optically in shared computer back-planes. As a simple illustration of this idea, a circuit performing error correction on a (3,1) Hamming Code is demonstrated. Error-free performance (Bit Error Rate of < 10⁻⁹) at 2.5 Gbit/s is achieved after single-error correction on the Hamming word with 50 percent errors. The bit-rate is only limited by the bandwidth of available resources. Since Four-Wave Mixing is an ultrafast nonlinearity, these circuits offer the potential of computing at several terabits per second. Furthermore, it is shown that several Boolean functions can be performed in parallel in the same set of devices using different multi-photon scattering processes. The main objective of this thesis is to motivate a new paradigm of thought in digital circuit design. Challenges pertaining to the feasibility of these ideas are discussed.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
all-optical signal processing; fiber-optic communications; four-wave mixing; semiconductor optical amplifiers
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Applied Physics
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Vahala, Kerry J.
Thesis Committee:
Vahala, Kerry J. (chair)
Fultz, Brent T.
Rutledge, David B.
Atwater, Harry Albert
Bridges, William B.
Defense Date:
18 May 2001
Non-Caltech Author Email:
ashishb (AT) bell-labs.com
Record Number:
CaltechETD:etd-07132001-112041
Persistent URL:
DOI:
10.7907/CMAB-WF06
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
2866
Collection:
CaltechTHESIS
Deposited By:
Imported from ETD-db
Deposited On:
13 Jul 2001
Last Modified:
01 Dec 2022 22:59
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
1MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
All-Optical Logic Circuits based on the Polarization Properties
of Non-Degenerate Four-Wave Mixing

Thesis by
Ashish Ishwar Singh Bhardwaj

In Partial Fulfillment of the Requirements
For the Degree of
Doctor of Philosophy

California Institute of Technology
Pasadena, California
2001
(Defended May 18, 2001)

Ashish Ishwar Singh Bhardwaj

iii

To my parents

Acknowledgements

I would like to express my gratitude to my advisor, Professor Kerry Vahala for giving me
the opportunity to be a member of his research group. I am indebted to him for the
knowledge I have gained in the areas of Photonics and Quantum Electronics during my
five years at Caltech.
I am extremely grateful to Prof. Vincent McKoy for his constant encouragement
and support. He has rekindled my passion for science and I will forever remain indebted
to him for being my mentor and for taking an active interest in the progress of my thesis.
I am also very grateful to Dr. Per Olof Hedekvist for teaching me the skills required to be
a good experimental scientist. Learning and working with him has been one of the most
enriching and fruitful experiences of my life.
I am grateful to my friends Nathan Good and Dane Boysen for their friendship
and moral support. I have shared many enlightening and fruitful discussions with them. I
would like to acknowledge the following members of the Vahala group for making my
stay at Caltech more enjoyable, Dr. David Geraghty, Dr. Dave Dougherty, Dr. Roberto
Paiella, Dr. Mark Brongermsa, Sean Spillane, and Tobias Kippenberg. I would also like
to thank Henrik Andersson, who is presently at Chalmers University of Technology in
Sweden, for his help with my experiments. I would like to thank the following members
of the Caltech staff, Jana Mercado, Linda Dosza, Greg Dunn, Lena Lenore, and Rosalie
Rowe for their help with administrative matters.

I would like to thank the Division of Applied Physics and the Division of
Chemistry for supporting my graduate studies in the form of teaching assistantships for
three and a half years. I would like to acknowledge the National Science Foundation for
its support in the form of a research assistantship for the last one and a half year.
My deepest thanks, of course, go to my parents for their unconditional support,
patience and encouragement. Their love and understanding have helped me overcome the
frustrations of graduate school.

Abstract

This thesis investigates a new class of all-optical logic circuits that are based on
the polarization properties of non-degenerate Four-Wave Mixing. Such circuits would be
used in conjunction with a data modulation format where the information is coded on the
states of polarization of the electric field. Schemes to perform multiple triple-product
logic functions are discussed and it is shown that higher-level Boolean operations
involving several bits can be implemented without resorting to the standard 2-input gates
that are based on some form of switching. Instead, an entire hierarchy of more complex
Boolean functions can be derived based on the selection rules of multi-photon scattering
processes that can form a new classes of primitive building blocks for digital circuits.
Possible applications of these circuits could involve some front-end signal
processing to be performed all-optically in shared computer back-planes. As a simple
illustration of this idea, a circuit performing error correction on a (3,1) Hamming Code is
demonstrated. Error-free performance (Bit Error Rate of < 10-9) at 2.5 Gbit/s is achieved
after single-error correction on the Hamming word with 50 percent errors. The bit-rate is
only limited by the bandwidth of available resources. Since Four-Wave Mixing is an
ultrafast nonlinearity, these circuits offer the potential of computing at several terabits per
second. Furthermore, it is shown that several Boolean functions can be performed in
parallel in the same set of devices using different multi-photon scattering processes. The

vii
main objective of this thesis is to motivate a new paradigm of thought in digital circuit
design. Challenges pertaining to the feasibility of these ideas are discussed.

viii

Contents

Acknowledgements

Abstract

Introduction

1.1

Background

1.2

Thesis Outline

Bibliography

Applications of Nonlinear Optics in High-Speed Digital Processing

2.1

Introduction

2.2

Switching Based on Cross-Phase Modulation (XPM)

A. The Nonlinear Optical Loop Mirror (NOLM)

B. Terahertz Optical Asymmetric Demultiplexer (TOAD)

C. Ultrafast Nonlinear Interferometer (UNI)

2.3

Switching Based on Cross-Gain Modulation (XGM)

2.4

Ultrafast Switching Using Four-Wave Mixing

Bibliography

Four-Wave Mixing in Semiconductor Optical Amplifiers

3.1

Introduction

3.2

Contributions to the FWM Susceptibility

3.3

Conversion Efficiency of FWM in a SOA

3.4

Optical-Signal-to-Noise Ratio of the FWM Converted Signal

3.5

Broad-Band Wavelength Conversion Using Long SOAs

3.6

Polarization Dependence of FWM

Bibliography

Advanced All-Optical Logic on a Spectral Bus

4.1

The Spectral Data Bus

4.2

All-Optical Front-End Processing on the Spectral Bus

4.3

Polarization Shift Keying (PolSK)

4.4

All-Optical Processing on PolSK Coded Spectral Data

4.5

Coding and Decoding using the (3,1) Hamming Code

4.6

Polarization Properties of ND-FWM Between 2-PolSK Coded

4.7

WDM Channels

Error Correcting Circuit for the (3,1) Hamming Code

Bibliography

The Error-Correcting Circuit for the (3,1) Hamming Code

5.1

Introduction

5.2

Characterization of the Extinction Ratio

5.3

The PolSK Transmitter

5.4

Building the Error-Correcting Circuit

A. The Pre-processing Elements

B. Polarization Maintaining EDFA (PM-EDFA)

5.5

Experiment

5.6

Results and Discussions

Bibliography

All-Optical Logic Circuits Based on Polarization Properties
of ND-FWM

6.1

Introduction

6.2

Generalization to a 3-bit Adder

6.3

Encoding the (7,4) Hamming Code

6.4

Decoder Circuit for the (7,4) Hamming Code

101

6.5

Comments on Generalization

108

6.6

Conclusion

110

Bibliography

112

List of Figures

Figure 2.1: All-optical switching using XPM in a nonlinear interferometer

Figure 2.2: The Nonlinear Optical Loop Mirror (NOLM)

Figure 2.3: The Terahertz Optical Asymmetric Demultiplexer (TOAD)

Figure 2.4: The Ultrafast Nonlinear Interferometer (UNI)

Figure 2.5: Schematic of the Mach-Zehnder Interferometer (MZI) gate that
uses a time delay between the control pulses entering each arm to
open a small switching window
Figure 3.1: Generation of the FWM sidebands in a SOA

16
25

Figure 3.2: Conversion efficiency of the FWM signal measured for wavelength
downshift in a 1.5 mm bulk SOA

Figure 3.3: Many cascaded FWM processes can occur due to high nonlinearity
in a SOA when two waves P1 and P2 are launched

Figure 3.4: Conversion efficiency versus detuning for a 2.2 mm and 4 mm long
SOA. The SOAs are biased at 1 A

Figure 3.5: 40 nm downshift in a 2.2 mm long SOA with 23.12 dB OSNR for
low error-rate detection at 10 Gb/s (from Ref. [14])

Figure 4.1: Schematic of the Spectral Data Bus for byte-wide transmission
(C1, C2, C3 etc. are different WDM channels)

Figure 4.2: All-optical logic circuits can be designed for some form of limited
on-the-fly front-end signal processing on a spectral bus (C1, C2, C3
etc. are different WDM channels)

xii
Figure 4.3: 2-PolSK coding using orthogonal linear states of polarization

Figure 4.4: Truth table and pictorial representation of the (3,1) Hamming Code

Figure 4.5: Implementation of the Boolean function for the (3,1) Hamming
Code using 2-input "AND" and "OR" gates

Figure 4.6: Diagrammatic representations of the non-degenerate FWM process
where in (a) C3 forms a grating with C1 that scatters off C2 and in (b) C3
forms a grating with C2 that scatters off C1, to form the FWM sideband
Figure 4.7: Polarization selection rules for ND-FWM between C1, C2, and C3

Figure 4.8: Layout of the error-correcting circuit showing the "non-correcting"
and the "correcting arm"

Figure 4.9: Working of the error-correcting circuit for different input Hamming
words

Figure 5.1: Schematic of the error-correcting circuit

Figure 5.2: The PolSK transmitter

Figure 5.3: BER versus received power (in 0.5 nm Resolution Bandwidth) for
PolSK modulated data at 2.5 Gbit/s measured immediately after
transmission on channel C1

Figure 5.4: Schematic of the birefringent element

Figure 5.5: Setup to measure the response of the birefringent element

Figure 5.6: Response of the birefringent element measured by the OSA

Figure 5.7(a): Schematic of the backward pumped PM-EDFA

Figure 5.7(b): The PM-EDFA

Figure 5.8(a): Setup to measure the ER of the PM-EDFA

xiii
Figure 5.8(b): Circle traced on the PoincarŽ sphere shows 19 dB ER for
PM-EDFA
Figure 5.9: Gain and Noise Figure of the PM-EDFA

79
80

Figure 5.10: Optical spectrum at the output of the SOA (in 0.1 nm Resolution
Bandwidth) showing the wavelengths channels [C1-C3] and EC
(from Ref. [7])
Figure 5.11: The Òcorrecting armÓ of the error-correcting circuit

82
84

Figure 5.12(a): Optical Spectrum after the SOA (in 0.1 nm Resolution Bandwidth)
in the "non-correcting arm" for the received word [1,1,1] (after Ref. [6])

Figure 5.12(b): Optical Spectrum after the SOA (in 0.1 nm Resolution Bandwidth)
in the "correcting arm" for the received word [1,1,1] (after Ref. [6])

Figure 5.13: Oscilloscope traces of the 8-bit pattern for (a) erroneous data on
channel C3, (b) data on channels C1 and C2 and (c) error-corrected FWM
signal (from Ref. [6])

Figure 5.14: Oscilloscope traces of (a) 16-bit patterns on channels C1, C2, and
C3 at 2.5 Gbit/s, (b) EC output from the non-correcting arm, (c) EC
output from the correcting arm and (d) EC output from both arms
combined (from Ref. [7])

Figure 5.15(a): BER versus received power (in 0.5 nm Resolution Bandwidth)
at 2.5 Gbit/s for random errors on C3 (from Ref. [7])

Figure 5.15(b): BER versus received power (in 0.5 nm Resolution Bandwidth)
at 2.5 Gbit/s for random errors on C1 (from Ref. [7])

Figure 5.16: Eye diagram after error-correction on the EC channel

xiv
Figure 5.17: BER versus received power (in 0.5 nm Resolution Bandwidth) at
2.5 Gbit/s for error correction on ill-defined states with 30% errors on
C3 (from Ref. [7])
Figure 6.1: Schematic of the full 3-bit adder (from Ref. [2])

94
98

Figure 6.2: Optical Spectrum at the output of the SOA (in 0.1 nm Resolution
Bandwidth) in the presence of four input waves (from Ref. [2])

100

Figure 6.3: Generating the parity bits for the (7,4) Hamming Code

101

Figure 6.4: Generator circuit for D4

105

Figure 6.5: Generator circuits for bits [D1-D3]

107

Figure 6.6: Two cascaded FWM processes can be used to implement a 5-bit
parity generator

109

Chapter 1

Introduction
________________________________________________________________________

1.1 Background

Data transmission speeds over fiber optic networks have been steadily increasing over the
past decade. Commercial systems capable of transmitting hundreds of gigabits per second
are available and systems demonstrations of over a terabit per second (one terabit is 1012
bits) transmission have been demonstrated experimentally [1-4]. Record transmission
rates of over 10 terabits per second have been recently demonstrated [5, 6]. Long-haul
transmission over thousands of kilometers is made possible with the availability of
extremely low loss optical fibers (typically 0.2 dB/km) and the invention of the ErbiumDoped Fiber Amplifiers (EDFAs) [7]. Thus it is now possible to develop high capacity
long-haul photonic networks and this is an active field of research in many leading
organizations across the world [8-12]. As the aggregate data rates increase, the efficiency
of the photonic network is limited by the speed of the electronics at the transmitters and
receivers in the network. Similar limitations are imposed by the presence of electronic
switches and routers. This is frequently referred to as the "electronic bottleneck" that has
led to an increased interest in trying to develop techniques to add optical functionality to

enable signal processing to be accomplished in the all-optical domain. Nonlinear optics
has been studied extensively to develop techniques to enable different forms of signal
processing in the optical domain and some of the techniques are reviewed in this thesis. A
third-order nonlinear process, namely Four-Wave Mixing (FWM) is studied in greater
detail. Applications of FWM for ultrafast optical logic are reviewed and a new class of
all-optical logic gates using the polarization properties of FWM is proposed and
demonstrated [13-16].

1.2 Thesis Outline

A review of optical switching technologies based on nonlinear effects is presented in
Chapter 2. FWM in Semiconductor Optical Amplifiers (SOA) is discussed in Chapter 3
and pertinent issues such as the device length and dependence of FWM on the
polarization of the electric fields involved are reviewed. The concepts of a Spectral Data
Bus and Polarization Shift Keying (PolSK) as new modes of data transmission on an
optical fiber are discussed in Chapter 4. This sets the premise to consider the feasibility of
using the polarization dependence of the FWM process to develop a new class of optical
logic gates. Keeping simple front-end applications on a spectral data bus in mind, on-thefly error detection and correction on a spectral bus is considered in detail and an Error
Detecting and Correcting circuit for a (3,1) Hamming Code based on the polarization
properties of FWM is proposed. Details of the construction of this circuit along with the
results obtained on its successful demonstration are presented in Chapter 5. The ideas

proposed for the (3,1) Hamming Code are further generalized in Chapter 6 where the
scheme is extended to implement a full 3-bit adder. New circuit designs based on the 3bit adder as fundamental building blocks for other higher level Hamming Codes, such as
the (7,4) Code are proposed.

Bibliography
1. T.N. Nielsen, A.J. Stentz, K. Rottwitt, D.S. Vengsarkar, Z.J. Chen, P.B. Hansen, J.H.
Park, K.S. Feder, S. Cabot, S. Stulz, D.W. Peckham, L. Hsu, C.K. Kan, A.F. Judy,
S.Y. Park, L.E. Nelson, and L. Gruner-Nielsen, "3.28 Tb/s Transmission over 3 x 100
km of Nonzero-Dispersion Fiber using Dual C- and L-band Distributed Raman
Amplification," IEEE Photonic. Tech. Lett., 12, 1079-1081 (2000).
2. A.R. Chraplyvy, "High-Capacity Lightwave Transmission Experiments," Bell Labs
Tech. J., 4, 230-245 (1999).
3. A.R. Chraplyvy, A.H. Gnauck, R.W. Tkach, J.L. Zyskind, J.W. Sulhoff, A.J. Lucero,
Y. Sun, R.M. Jopson, F. Forghieri, R.M. Derosier, C. Wolf, and A.R. McCormick,
"1-Tb/s Transmission Experiment," IEEE Photonic. Tech. Lett., 8, 1264-1266 (1996).
4. Y. Yambayashi and M. Nakazawa, "Terabit Transmission Technology," NTT Review,
11, 23-32 (1999).
5. K. Fukuchi, T. Kasamatsu, M. Morie, R. Ohhira, T. Ito, K. Sekiya, D. Ogasahara, and
T. Ono, "10.92 Tb/s (273 x 40-Gb/s) Triple-Band/Ultra-Dense WDM OpticalRepeatered Transmission Experiment," Post-deadline paper, PD-24, OFC' 2001,
Anaheim, CA, March 17-22 (2001).
6. S. Bigo,Y. Frignac, G. Charlet, S. Borne, P. Tran. C. Simonneau, D. Bayart, A.
Jourdan, J.P. Hamaide, W. Idler, R. Dischler, G. Veith, H. Gross, and W. Poehlmann,
"10.2 Tb/s (256 x 42.7 Gbit/s PDM/WDM) Transmission Over 100 km TeraLightTM
Fiber with 1.28 Bit/s/Hz Spectral Efficiency," Post-deadline paper , PD-25, OFC'
2001, Anaheim, CA, March 17-22 (2001).
7. R.J. Mears, L. Reekie, I.M. Jauncey, and D.N. Payne, "Low Noise Erbium-Doped
Fiber Amplifier Operating at 1.54 µm," Electron. Lett., 23, 1026-1028 (1987).

8. D.K. Hunter, and I. Andonovic, "Approaches to Optical Internet Packet-Switching,"
IEEE Commun. Mag., 38, 116-122 (2000).
9. S. Yao, B. Mukherjee, and S. Dixit, "Advances in Photonic Packet Switching: An
Overview," IEEE Commun. Mag., 38, 84-94 (2000).
10. R. S. Tucker, and W. De Zhong, "Photonic Packet Switching: An Overview," IEICE
T. Electron., E82C, 202-212 (1999).
11. C. Guillemot, M. Renaud, P. Gambini, C. Janz, I. Andonovic, R. Bauknecht, B.
Bostica, M. Burzio, F. Callegati, M. Casoni, D. Chiaroni, F. Clerot, S.L. Danielsen, F.
Dorgeuille, A. Dupas, A. Franzen, P.B. Hansen, D.K. Hunter, A. Kloch, R.
Krahenbuhl, B. Lavigne, A. Le Corre, C. Raffaelli, M. Schilling, J.C. Simon, and L.
Zucchelli, "Transparent Optical Packet Switching: The European ACTS KEOPS
Project Approach," J. Lightwave Technol., 16, 2065-2067 (1998).
12. V.W.S. Chan, K.L. Hall, E. Modiano, and K.A. Rauschenbach, "Architectures and
Technologies for High-Speed Optical Data Networks," J. Lightwave Technol., 16,
2146-2168 (1998).
13. A. Bhardwaj, P.O. Hedekvist, and K. Vahala, "All-Optical Logic Circuits Based on
Polarization Properties of Nondegenerate Four-Wave Mixing," J. Opt. Soc. Am. B,
18, 657-665 (2001).
14. P.O. Hedekvist, A. Bhardwaj, K. Vahala, and H. Andersson, "Advanced All-Optical
Logic Gates on a Spectral Bus," Appl. Opt., 40, 1761-1766 (2001).
15. P.O. Hedekvist, A. Bhardwaj, K. Vahala, and H. Andersson, "Multiple-Input AllOptical Logic Gates Utilizing Polarization Properties of Non-Degenerate Four-Wave

Mixing," Paper UFB2, presented at the Topical Meeting on Ultrafast Electronics and
Optoelectronics, Lake Tahoe, Nevada, Jan. 10-12 (2001).
16. A. Bhardwaj, P.O. Hedekvist, H. Andersson, and K.J. Vahala, "All Optical Front End
Error Correction on a Spectral Data Bus," Paper CWI5, presented at the Conference
on Lasers and Electro-Optics, San Francisco, CA, May 7-12 (2000).

Chapter 2

Applications of Nonlinear Optics in High-Speed
Digital Processing
________________________________________________________________________

2.1 Introduction

Nonlinear optics has been of increased interest for all-optical signal processing in highspeed photonic networks [1, 2]. Sample applications include all-optical switching as well
as demultiplexing [3-5]. In addition, Boolean operations such as Exclusive-OR (XOR)
and 2-bit addition have been demonstrated optically using a combination of such
switching devices [6, 7]. Devices have employed both fiber-based and semiconductorbased nonlinear elements. In the former case, the physical nonlinearity is the Kerr
nonlinearity of silica glass. In the latter case, the nonlinearity results from a variety of
ultrafast mechanisms in semiconductor gain media, including carrier heating and spectral
hole burning [8]. Apart from long-haul data transmission, all-optical logic gates might
someday find applications in local area networks to provide certain limited all-optical
functionality. Such functions will only make sense where equivalent electrical solutions
are cumbersome or when a real advantage can be realized by maintaining signals in

optical form. In Wavelength Division Multiplexed (WDM) systems there has been
interest in all-optical functionality for switching and routing, but so far very limited
implementation of these functions [9, 10]. The likelihood of more sophisticated alloptical functions will be higher should Time Division Multiplexed (TDM) or mixed
TDM/WDM systems be implemented since this format lends itself better to a variety of
well-established all-optical switching solutions [1].

2.2 Switching Based on Cross-Phase Modulation

Optical switching using a nonlinear interferometer makes it possible for one optical
signal to control and switch another optical signal through the nonlinear interaction in a
material. The input signal to be switched is split between the arms of the interferometer.
The interferometer is balanced so that, in the absence of a control signal, the input signal
emerges from one output port. The presence of a strong control pulse changes the
refractive index of the medium given by

Æn = n2I,

(2-1)

where Æn is the change in the refractive index of the medium, n2 is the nonlinear
refraction coefficient and I is the intensity of light incident on the medium. A change in
the index adds a phase shift between the two arms of the interferometer, so that the input

signal is switched over to a second output port. This method of switching based on crossphase modulation (XPM) is schematically shown in Figure 2.1.

Control
Pulse

n = n 0 + n 2I
NONLINEAR
MEDIUM

Signal
Pulse

Pout

LINEAR
MEDIUM

Figure 2.1: All-optical switching using XPM in a nonlinear interferometer

A. The Nonlinear Optical Loop Mirror (NOLM)

One way to implement XPM-based switching is to use a Sagnac interferometer, where
one of the output ports also serves as the input port for the signal to be switched. This
configuration is commonly referred to as the Nonlinear Optical Loop Mirror (NOLM)
[11] and is shown in Figure 2.2.

Data AND Control

Data
Control
n2 , L

Figure 2.2: The Nonlinear Optical Loop Mirror (NOLM)

The input coupler splits the input signal pulse into two counter-propagating
pulses, which subsequently combine again at the coupler, each having traveled around the
loop. A strong control pulse is then introduced into the loop as a unidirectional beam. The
nonlinear optical effect of the control pulse is to induce a refractive index change, Æn,
which is experienced fully by a co-propagating signal pulse. This refractive index change
results in a differential phase shift, Æ¯, between the counter-propagating signal pulses as
they arrive back at the input coupler, given by

Æ¯ = k Æn L,

(2-2)

where k is the wavevector and L is the path length over which the induced index change
Æn is effective. The path length is chosen such that complete switching occurs, i.e., the

11
phase shift is π radians. In a fiber interferometer, the physical mechanism is the intrinsic
Kerr nonlinearity of glass [12]. Since the optical nonlinearity in glass is very small (n2 ≈
3×10-20 m2 W-1 for silica), a device control power-length product of typically ~1 W-km is
needed for a phase shift of π. In practice, fiber interferometer path lengths of several
kilometers are needed to keep the average control power to below 100 mW. The long
path lengths make it difficult to build stable and compact devices.

B. The Terahertz Optical Asymmetric Demultiplexer (TOAD)

A Semiconductor Optical Amplifier (SOA) is similar to a semiconductor laser diode,
except that the reflectivity of the end faces is deliberately minimized to suppress lasing.
Thus the SOA acts as a one-pass device for a lightwave with an inversion that is created
by electrical pumping. The conduction and valence bands in a semiconductor can be
modeled as an ensemble of two level atom-like systems, which are coupled through
various scattering mechanisms. These scattering mechanisms which control the evolution
of the two-level systems will be discussed in greater detail in Chapter 3. For photons that
are resonant with the transition energy levels of the states that are inverted, stimulated
emission can occur, i.e., photons at these frequencies see a gain. As the intensity of light
increases, the gain saturates from the depopulation of the conduction band due to
stimulated emission. Associated with this change in the gain due to saturation is a
refractive index change, as described by the Kramer-Kršnig dispersion relations. The
refractive nonlinearity of the SOA is ~108 times larger than an equivalent length of silica

12
fiber. The relaxation time associated with the relaxation of the refractive index to its
equilibrium value is governed by the interband carrier lifetime, and is typically 100-500
picoseconds.
Since the interband carrier lifetimes are very slow, switching at data-rates much
higher than 1 Gbit/s based on inter-band carrier relaxation did not seem possible. This
limitation was overcome by placing the SOA asymmetrically with respect to the center of
the loop in the Sagnac interferometer, as shown in Figure 2.3. This is called the
Semiconductor Laser Amplifier in a Loop Mirror (SLALOM) [13] or Terahertz Optical
Asymmetric Demultiplexer (TOAD) [14]. Since the rise time associated with the change
in refractive index is less than a picosecond [15], it is possible to obtain a switching
window which is shorter than the recovery time limited by the inter-band carrier
relaxation. It has been demonstrated that a window width of ≤ 10 picoseconds can be
created which allows demultiplexing a 50 GHz pulse train down to a base rate of 1 GHz
[16]. By operating the SOA in strong saturation, the carrier relaxation time can be
modified to ~25 picoseconds [17]. The XPM bandwidth also increases with the device
length due to traveling wave effects [18, 19] and demonstrations of wavelength
conversion at 100 Gbit/s using XPM have been reported [20].

Data AND Control

Data
Control

SOA
∆x

Figure 2.3: The Terahertz Optical Asymmetric Demultiplexer (TOAD)

There have been successful demonstrations of different optical functionalities
using configurations based on the NOLM and TOAD. These include demultiplexers [3,
17] and the implementation of simple Boolean functions such as AND, NOT and
Exclusive-OR (XOR) for address recognition [21, 22]. More sophisticated Boolean
functions such as 3-bit adders [23, 24] and parity bit generators [25] have been
demonstrated using a combination of several TOAD-based gates.

14
C. The Ultrafast Nonlinear Interferometer (UNI)

The UNI is a balanced, single-arm interferometer that does not require any external
stabilization of the interferometer arms. The signal pulse that is to be switched is split
into two orthogonal polarization components with a time delay (typically equal to half the
bit period) by passing it through highly birefringent fiber. The two orthogonal pulses then
pass through a SOA and are temporally recombined after passing through a second
birefringent fiber and interfered. The State of Polarization of the signal pulse after
recombining is determined by the induced phase changes from the time-dependent
refractive index changes in the presence of a control pulse (which could be copropagating or counter-propagating) that is aligned temporally with one of the orthogonal
pulses in the SOA. The signal pulse then passes through a fiber polarizer that is adjusted
such that the signal pulse is orthogonal to the polarizer in the presence of the control
pulse and parallel to the polarizer when the control pulse is not present. 100 Gbit/s bitwise switching has been demonstrated using the UNI gate [10]. A schematic of the UNI
gate is shown in Figure 2.4.

Data

PMF

Data AND Control

SOA

Control

Figure 2.4: The Ultrafast Nonlinear Interferometer (UNI)

15
For dual rail logic, such as XOR, the phase of each polarization state can be
accessed and changed independently by two separate control pulses. All-optical XOR
gate based on the UNI has been demonstrated on a 40 GHz clock pulse [26] and at 20
Gbit/s [27]. A 40 GHz all-optical shift-register with an inverter has also been
demonstrated using 2 mm SOAs in the UNI gate configuration [28].

Other important applications of XPM in advanced optical communications
systems include regeneration and wavelength conversion in the optical domain.
Regeneration is essential to remove the errors at the detection-end caused by the
distortion of the signal due to noise, dispersion, and crosstalk. This is called 3R
regeneration, where the signal is reamplified, reshaped and retimed [13]. 3R regeneration
along with wavelength conversion at 80 Gbit/s with error-free operation has been
demonstrated using XPM in a nonlinear Mach-Zehnder Interferometer (MZI) with a SOA
[29]. Interferometric gates based on XPM in SOAs using the Mach-Zehnder or
Michaelson configuration have been integrated on planar lightwave circuits and are
reviewed in [30]. Figure 2.5 shows a schematic of a Mach-Zehnder Interferometer (MZI)
gate to obtain a short switching window by adding a time delay between the control
pulses entering each arm of the interferometer.

16
Control
Data
AND
Control
Data
SOA 1

Control
delay

SOA 2

∆τ

Figure 2.5: Schematic of the Mach-Zehnder Interferometer (MZI) gate that
uses a time delay between the control pulses entering each arm to open a
small switching window

2.3 Switching Based on Cross-Gain Modulation

The Cross-Gain Modulation (XGM) uses an input signal to saturate the gain and thereby
modulate a Continuous Wave (CW) signal at a desired output wavelength. Wavelength
conversion at 100 Gbit/s using XGM in SOAs has been demonstrated [31]. Although
simple to implement, the XGM gate has a number of shortcomings, such as inversion of
the input-control signal and the relatively large chirp of the output signal due to the large
gain modulation. Format conversion from Return to Zero (RZ) to Non-Return to Zero
(NRZ) and vice versa has been demonstrated using XGM gates [32].

2.4 Ultrafast Switching Based on Four-Wave Mixing

The TOAD based gates are limited by the interband carrier lifetimes. Higher speeds of
operation can be obtained in SOAs by nonlinear processes that involve intraband
transitions, such as Four-Wave Mixing (FWM). FWM in active semiconductor gain
media will be investigated in more detail in Chapter 3. In short, the intraband processes
leading to FWM are carrier heating (CH) and spectral hole burning (SHB) [8]. The speed
of CH process is limited by carrier-phonon scattering lifetime and SHB is limited by
carrier-carrier scattering lifetime, i.e., the time required to establish a Fermi-Dirac
distribution of the carrier population. These CH and SHB lifetimes have been measured
experimentally by Zhou et al. [33] and it is believed that carrier-carrier scattering occurs
at a time scale of 50-100 femtoseconds. Thus in principle, optical logic gates based on
FWM in SOAs with speeds of a few terabits per second are possible. FWM in SOAs has
been used for applications such as demultiplexing [34], wavelength conversion [34] and
even dispersion compensation using optical phase conjugation [35]. All-optical AND
gates have been developed and error-free performance at 100 Gbit/s has been
demonstrated [36]. Applications of AND gates also include address recognition of optical
packets [36] and clock-recovery [37].

As OTDM or mixed TDM/WDM systems evolve to higher levels of
sophistication and speed, optical logic gates and circuits built using a combination of
such gates will be used not only for demultiplexing, wavelength conversion and 2R/3R

18
regeneration, but also for header recognition for all-optical packet switched networks to
enable all-optical routing, coding and decoding for error detection and correction and
possibly encryption for added security. Though at present such devices cannot be as
densely integrated on a chip as their electronic equivalents, they offer advantage in terms
of speed and compatibility with state-of-the-art optical transmission systems. With the
development of newer network architectures and protocols, all-optical logic gates will be
used increasingly in tandem with electronics to greatly enhance the capacity and
throughput of future communications networks.

Bibliography
1. D. Cotter, R.J. Manning, K.J. Blow, A.D. Ellis, A.E. Kelly, D. Nesset, I.D. Phillips,
A.J. Poustie, and D.C. Rogers, "Nonlinear Optics for High-Speed Digital Information
Processing," Science, 286, 1523-1528 (1999).
2. V.W.S. Chan, K.L. Hall, E. Modiano, and K.A. Rauschenbach, "Architectures and
Technologies for High-Speed Optical Data Networks," J. Lightwave Technol., 16,
2146-2168 (1998).
3. T. Yamamoto, E. Yoshida, and M. Nakazawa, "Ultrafast Nonlinear Optical Loop
Mirror for Demultiplexing 640 Gbit/s TDM Signals," Electron. Lett., 34, 1013-1014
(1998).
4. K. Uchiyama, H. Takara, S. Kawanishi, T. Morioka, and M. Saruwatari, "Ultrafast
Polarization-Independent All-Optical Switching Using a Polarization Diversity
Scheme in the Nonlinear Optical Loop Mirror," Electron. Lett., 28, 1864-1866
(1992).
5. M. Eiselt, W. Pieper, and H.G. Weber, "All-optical High-Speed Demultiplexing with
a Semiconductor-Laser Amplifier in a Loop Mirror Configuration," Electron. Lett.,
29, 1167-1168 (1993).
6. A.J. Poustie, K.J. Blow, A.E. Kelly, and R.J. Manning, "All-Optical Binary HalfAdder," Opt. Commun., 156, 22-26 (1998).
7 . R.J. Manning, A.D. Ellis, A.J. Poustie, and K.J. Blow, "Semiconductor Laser
Amplifiers for Ultrafast All-Optical Signal Processing," J. Opt. Soc. Am. B, 14, 32043216 (1997).

20
8 . G.P. Agrawal, "Population Pulsation and Nondegenerate Four-Wave Mixing in
Semiconductor Laser Amplifiers," J. Opt. Soc. Am. B, 5, 147-159 (1988).
9. D. Nesset, M.C. Tatham, L.D. Westbrook, and D. Cotter, "Degenerate Wavelength
Operation of an Ultrafast All-Optical AND Gate Using Four Wave Mixing in a
Semiconductor Laser Amplifier," Electron. Lett., 30, 1938-1940 (1994).
10. K.L. Hall, and K.A. Rauschenbach, "100-Gbit/s Bitwise Logic," Opt. Lett., 23, 12711273 (1998).
11. N.J. Doran, and D. Wood, "Nonlinear-Optical Loop Mirror," Opt. Lett., 13, 56-58
(1988).
12. G.P. Agrawal, Nonlinear Fiber Optics, Academic Press, San Diego, CA, 1989.
13. M. Eiselt, W. Pieper, and H.G. Weber, "SLALOM - Semiconductor-Laser Amplifier
in a Loop Mirror," J. Lightwave Technol., 13, 2099-2112 (1995).
14. J.P. Sokoloff, P.R. Prucnal, I. Glesk, and M. Kane, "A Terahertz Optical Asymmetric
Demultiplexer (TOAD)," IEEE Photon. Tech. Lett., 5, 787-790 (1993).
1 5 .J. Mark, and J. Mork, "Subpicosecond Gain Dynamics in InGaAsP Optical
Amplifiers - Experiment and Theory," Appl. Phys. Lett., 61, 2281-2283 (1992).
16. J.P Sokoloff, I. Glesk, P.R. Prucnal, and R.K. Boncek, "Performance of a 50 Gbit/s
Optical-Time Domain Multiplexed System Using a Terahertz Optical Asymmetric
Demultiplexer," IEEE Photon. Tech. Lett., 6, 98-100 (1994).
1 7 .I. Glesk, J.P. Sokoloff, and P.R. Prucnal, "Demonstration of All-Optical
Demultiplexing of TDM Data at 250 Gbit/s," Electron. Lett., 30, 339-341 (1994).

21
18. D.D. Marcenac, A.E. Kelly, D. Nesset, and D.A.O. Davies, "Bandwidth Enhancement
of Wavelength Conversion by Semiconductor Optical Amplifier Cascade," Electron.
Lett., 31, 1442-1443 (1995).
1 9 D.A.O.
Davies, "Small-Signal Analysis of Wavelength Conversion in
Semiconductor-Laser Amplifiers via Gain Saturation," IEEE Photon. Tech. Lett., 7,
617-619 (1995).
20. J. Leuthold, C.H. Joyner, B. Mikkelsen, G. Raybon, J.L. Pleumeekers, B.I. Miller, K.
Dreyer, and C.A. Burrus, "100 Gbit/s All-Optical Wavelength Conversion with
Integrated SOA Delayed-Interference Configuration," Electron. Lett., 36, 1129-1130
(2000).
2 1 .K.L. Hall, and K.A. Rauschenbach, "All-Optical Bit Pattern Generation and
Matching," Electron. Lett., 32, 1214-1215 (1996).
22. M. Jinno, and T. Matsumoto, "Ultrafast All-Optical Logic Operations in a Nonlinear
Sagnac Interferometer with 2 Control Beams," Opt. Lett., 16, 220-222 (1991).
23. A.J. Poustie, K.J. Blow, A.E. Kelly, and R.J. Manning, "All-Optical Binary HalfAdder," Opt. Commun., 156, 22-26 (1998).
24. A.J. Poustie, K.J. Blow, A.E. Kelly, and R.J. Manning, "All-Optical Full-Adder with
Bit Differential Delay," Opt. Commun., 168, 89-93 (1999).
25. A.J. Poustie, K.J. Blow, A.E. Kelly, and R.J. Manning, "All-Optical Parity Checker
with Bit-Differential Delay," Opt. Commun., 162, 37-43 (1999).
26. G. Theophilopoulos, K. Yiannopoulos, M. Kalyvas, C. Bintjas, G. Kalogerakis, H.
Avramopoulos, L. Occhi, L. Schares, G. Guekos, S. Hansmann, and R. Dall' Ara, "40

22
GHz All-Optical XOR with UNI Gate," Paper MB2, OFC '2001, Anaheim, CA,
March 17-22 (2001).
27. C. Bintjas, M. Kalyvas, G. Theophilopoulos, T. Stathopoulos, H. Avramopoulos, L.
Occhi, L. Schares, G. Guekos, S. Hansmann, and R. Dall' Ara, "20 Gb/s All-Optical
XOR with UNI gate," IEEE Photonic. Techol. Lett., 12, 834-836 (2000).
28. K.L. Hall, J.P. Donnelly, S.H. Groves, C.I. Fennelly, R.A. Bailey, and A. Napoleone,
"40 Gbit/s All-Optical Circulating Shift Register with an Inverter," Opt. Lett., 22,
1479-1481 (1997).
29. A.E. Kelly, I.D. Phillips, R.J. Manning, A.D. Ellis, D. Nesset, D.G. Moodie, and R.
Kashyap, "80 Gbit/s All-Optical Regenerative Wavelength Conversion Using
Semiconductor Optical Amplifier based Interferometer," Electron. Lett., 35, 14771478 (1999).
30. K.E. Stubkjaer, "Semiconductor Optical Amplifier-Based All-Optical Gates for HighSpeed Optical Processing," IEEE J. Sel. Top. Quant. Electron., 6, 1428-1435 (2000).
31. A.D. Ellis, A.E. Kelly, D. Nesset, D. Pitcher, D.G. Moodie, and R. Kashyap, "Error
Free 100 Gb/s Wavelength Conversion Using Grating Assisted Cross Gain
Modulation in a 2 mm Long Semiconductor Optical Amplifier," Electron. Lett., 34,
1958-1959 (1998).
32. D. Norte, and A.E. Willner, "Multistage All-Optical WDM-to-TDM-to-WDM and
TDM-to-WDM-to-TDM Data-Format Conversion and Reconversion Through 80 km
of Fiber and Three EDFAs," IEEE Photon. Technol. Lett., 7, 1354-1356 (1995).
33. J. Zhou, N. Park, J.W. Dawson, K.J. Vahala, M.A. Newkirk, and B.I. Miller,
"Terahertz Four-Wave Mixing Spectroscopy for Study of Ultrafast Dynamics in a
Semiconductor Optical Amplifier," Appl. Phys. Lett., 63, 1179-1181 (1993).

34. K. Uchiyama, S. Kawanishi, and M. Saruwatari, "100-Gb/s Multiple-Channel Output
All-Optical OTDM Demultiplexing Using Multichannel Four-Wave Mixing in a
Semiconductor Optical Amplifier," IEEE Photonic. Technol. Lett., 10, 890-892
(1998).
35. D.F. Geraghty, R.B. Lee, M. Verdiell, M. Ziari, A. Mathur, and K.J. Vahala,
"Wavelength Conversion for WDM Communication Systems Using Four-Wave
Mixing in Semiconductor Optical Amplifiers," IEEE J. Sel. Top. Quant. Electron., 3,
1146-1155 (1997).
36. D.D. Marcenac, D. Nesset, A.E. Kelly, M. Brierley, A.D. Ellis, D.G. Moodie, and
C.W. Ford, "40 Gbit/s Transmission over 406 km of NDSF Using Mid-Span Spectral
Inversion by Four-Wave Mixing in a 2 mm Long Semiconductor Optical Amplifier,"
Electron. Lett., 33, 879-880 (1997).
37. D. Cotter, J.K. Lucek, M. Shabeer, K. Smith, D.C. Rogers, D. Nesset, and P.
Gunning, "Self-Routing of 100 Gbit/s Packets Using 6 Bit Keyword Address
Recognition," Electron. Lett., 31, 1475-1476 (1995).
38. O. Kamatani, and S. Kawanishi, "Ultrahigh-speed Clock Recovery with Phase Lock
Loop Based on Four-Wave Mixing in a Traveling-Wave Laser Diode Amplifier," J.
Lightwave Technol., 14, 1757-1767 (1996).

Chapter 3

Four-Wave Mixing in Semiconductor Optical
Amplifiers
________________________________________________________________________

3.1 Introduction

Four-Wave Mixing (FWM) is a third-order nonlinear process in which a polarization is
created in a medium that depends on the product of three electric fields. The induced
polarization leads to the creation of new frequency components of the electric field. In a
semiconductor, the nonlinear polarization can be mediated by resonant interactions of the
electric fields with the carriers in the medium [1]. Since resonant interactions lead to
attenuation of light, the nonlinear products that are created have low intensities. One way
to circumvent this problem is to use population inversion to create a gain in the medium
[2]. A Semiconductor Optical Amplifier (SOA) can provide high gain (typically ~ 30 dB)
which then amplifies the nonlinear mixing products that are created. FWM in SOAs has
been studied extensively for wavelength conversion [3-6] and it has the advantage of
being transparent to bit format. Furthermore, FWM in SOAs is an ultrafast nonlinear
process with bandwidths that are well over a terahertz, since the contributions to the

25
nonlinear susceptibility include intraband carrier dynamics that are limited by the carrierphonon and carrier-carrier scattering lifetimes which occur in the sub-picosecond regime.
In a typical FWM experiment, a strong pump wave at frequency ωp and a probe
wave at ωq are combined and coupled to the waveguide modes of the SOA, which is a
travelling wave amplifier. Dynamic gain and index gratings are formed due to the beating
of the pump and probe waves, at a detuning frequency given by Ω = ω q − ω p . The pump
and the probe waves are subsequently scattered by these gratings which give rise to the
FWM sidebands, as shown in Figure 3.1.

ωp

ωp
ωq

ωq
FWM
FWM

SOA
-Ω-

- Ω -- Ω -- Ω -

Input

Output

Figure 3.1: Generation of the FWM sidebands in a SOA

Consider the scattering process in which the pump and probe form a grating,
which scatters the pump and gives rise to one of the FWM sideband as shown in Figure
3.1. The frequency of this component is ω FWM = ω p − Ω or ω FWM = ω p + Ω . When

ω p > ω q , the FWM signal has a higher frequency than the probe (or is wavelength
downshifted) and when ω p < ω q , the FWM signal has a lower frequency than the probe
(or is wavelength upshifted). In the copropagation FWM geometry, guided waves in the

[(

)]

SOA are given by E j ( z )exp i k j z − ω j t , where j = p,q,c indicate the pump, probe and
converted signal respectively; {Ej(z)} are the field amplitudes and z is the longitudinal
coordinate along the propagation axis of the SOA. Then electric field of the FWM
component generated can be written as

Ec ( z ) = χ ( 3) (Ω) E p2 ( z ) Eq* ( z )ei∆kz ,

(3-1)

where χ ( 3) (Ω) is the third-order nonlinear susceptibility which depends on the detuning
frequency, Ω, between the pump and the probe; and ∆k = 2 k p − kq − kc is the wavevector phase mismatch between the pump, probe, and the converted signal. In a SOA, the
free-space wave-vectors are replaced with the propagation vectors of the waveguide
modes for the pump, probe, and the converted signal.

3.2 Contributions to the FWM Susceptibility

For detuning frequencies less than a few gigahertz, the largest contribution to the FWM
susceptibility is due to carrier density modulation (CDM) which arises from the beating
of the pump and probe waves. The intensity beating due to the pump and the probe lead

27
to a pulsation of the population inversion in the medium. This pulsation of the total
carriers leads to a gain modulation as seen by the traveling waves, which gives rise to the
FWM sidebands. Due to the slow recovery of the carrier density, determined by the
carrier lifetime, τs, which is on the order of several hundred picoseconds, the efficiency of
FWM mediated by CDM drops off for frequency detunings much larger than 10 GHz.
At large detuning frequencies, the gain and index gratings are formed by
intraband processes, such as carrier heating (CH) and Spectral Hole Burning (SHB) [7,
8]. Instead of a pulsation of the total carriers in the conduction and valence bands of the
semiconductor, the occupation probabilities of the individual transition levels are
modulated to form the gain and index gratings that give rise to the FWM sidebands. The
pulsation of the occupation probabilities create distortions in the equilibrium Fermi
distribution functions of the carriers, which relax back to their equilibrium state through
different scattering processes, such as CH and SHB. SHB arises from carrier-carrier
scattering which tend to restore a quasi-equilibrium Fermi distribution function.
Typically, carriers scatter each other at a time scale, τSHB, which is of the order of 50-100
femtoseconds. The quasi-equilibrium Fermi distribution function is characterized by a
temperature which is different from the lattice temperature. The relaxation to the lattice
temperature is through the emission of optical phonons with a time constant due to
carrier-phonon scattering lifetime, τCH, which is of the order of 0.5-1 picoseconds. Thus
the relaxation of the carriers can be explained as a three step process: (1) the carriers first
scatter with each other to create a quasi-equilibrium Fermi distribution which has a
temperature Tx that is different from the lattice temperature (SHB), (2) the temperature of

28
the carriers then relaxes to the lattice temperature through carrier-phonon scattering (CH)
and (3) the carriers then recombine through stimulated recombination.
The total FWM susceptibility due to these contributions can be derived by solving
the density matrix equations for the occupation probabilities and the polarization of the
levels that are resonant with the electric fields launched into the SOA, and in the presence
of the relaxation processes mentioned above. A rigorous derivation of the FWM
susceptibility is presented by Uskov et al. [8] that can be given by the expression

 A eiφ CDM
ACH eiφ CH
A eiφ SHB 
+ SHB
χ ( 3) (Ω) =  CDM
,
 (1 − iΩτ s ) (1 − iΩτ CH )(1 − iΩτ SHB ) (1 − iΩτ SHB ) 

(3-2)

where ACDM, ACH and ASHB are constants that determine the strength of the scattering
process leading to the nonlinearity, and φCDM, φCH and φSHB are the phase factors due to
the scattering processes. The frequency dependence of χ(3)(Ω) was first measured by
Zhou et al. [9]. The observed difference in the conversion efficiencies for positive and
negative detuning frequencies of the pump and the probe was attributed to the
interference between the terms arising from different scattering mechanisms. A typical
FWM response for wavelength downshift measured in a 1.5 mm long bulk SOA is shown
in Figure 3.2.

Normalized Conversion Efficiency (dB)

X X20 dB/dec
X X XX
XXXXXX

-10
-20
-30
-40
-50

XXX20X dB/dec
XXXXXXXXX
XX
XXXXXX 40 dB/dec
XXXXXXXXXX
XXXXX

-60
10

100

1000
Detuning (GHz)

10000

Figure 3.2: Conversion efficiency of the FWM signal measured for
wavelength downshift in a 1.5 mm bulk SOA

3.3 Conversion Efficiency of FWM in a SOA

The FWM conversion efficiency is defined as the ratio of the intensity of the FWM signal
at the output of the SOA to the intensity of the probe wave that is launched into the SOA.
The total conversion efficiency at the output of the SOA can be derived by using a
lumped model as proposed by Zhou et al. [10] and solving the coupled propagation
equations for each frequency component of the travelling waves that can be written as

30
dE p, q ( z ) 1 
g0
(1 − iα ) − α l  E p, q ( z )
dz
2 
P( z ) 
 1 +
Psat 
 


(3-3a)

dEc ( z ) 1 
g0
(1 − iα ) − α l  Ec ( z ) − κ ( z ) E p2 ( z ) Eq* ( z )ei∆kz
dz
2 
P( z ) 
 1 +
Psat 
 


(3-3b)

P( z ) = E p ( z ) + Eq ( z ) + Ec ( z ) ,

(3-3c)

where κ(z) is the FWM coupling constant, g0 is the unsaturated gain coefficient, P(z) is
the total optical power at position z along the propagation axis of the waveguide, Psat is
the saturation power of the SOA, α is the linewidth enhancement factor [11] and αl is the
nonsaturable internal loss per unit length of the waveguide. Zhou et al. showed that κ(z)
could be written as

κ (z) =

1 − iα m 1
g0
⋅ ,
2
P( z )  m =1 1 − iΩτ m Pm
1 +
Psat 

(3-4)

where m=1, 2, 3 denote CDM, CH and SHB. {τ m} and {Pm } are the lifetimes and
saturation powers associated with these mechanisms and {αm} are the ratios between the
real and imaginary parts of the refractive index change induced by these mechanisms.
Solving the coupled equations under moderate saturation, the expression for the

31
conversion efficiency, η, with phase-matching conditions satisfied at the output of the
SOA was derived as

η(Ω) = 10 log10

Pc ( L)
= 3G + 2 I p + 20 log10 ∑ cm ⋅
Pq (0)
1 − iΩτ m
m =1

(3-5)

where L is the length of the SOA, Ip is the input pump power in dBm, G is the saturated
gain of the SOA in dB and is given by

 
 
L 
 
G = exp  ∫ 
− α l  dz 

P( z ) 
 0  1 +
 
Psat 
  
 

(3-6a)

g0
G( dB) = 4.34 ∫
− α l  dz ,

P( z )
 1 +
Psat 
 


(3-6b)

The importance of the length of the SOA in determining the conversion efficiency
of the FWM signal was investigated by Mecozzi et al. [5, 12]. Under the assumption of
perfect phase-matching, an analytical expression for the conversion efficiency was
derived and they showed that for frequencies exceeding 10 GHz, i.e., Ωτ s >> 1, the
conversion efficiency can be written in the form

 Pp (0)  R(Ω)  G0 
G,
η( Ω ) =
ln
2 = 
 P(0)  (1 + εPsat )  G 
Eq (0)
Ec ( L)

(3-7)

where L is the length of the SOA, G0 = G0 ( L) is the unsaturated gain (or small signal
gain) of the SOA while G is the saturated gain of the SOA given by Eqn. (3-6a), ε is the
gain compression coefficient and R(Ω) is the frequency response function of the FWM
process due to the different scattering processes. Since the small signal gain, G0 neglects
saturation effects, it is expressed as

L
G0 = exp  ∫ ( g0 − α l )dz  = exp[( g0 − α l ) L] ,
0

(3-8)

The conversion efficiency strongly depends on the length of the device through
saturation of the SOA. The saturated gain G can be expressed in the same form as Eqn.
(3-8) with an effective length Leff over which transparency (no gain or loss seen by the
travelling waves) is achieved, i.e.,

 Leff
G = exp  ∫ ( g0 − α l )dz  = exp ( g0 − α l ) Leff ,
 0


(3-9)

With these expressions, keeping all other parameters constant, the dependence of the
conversion efficiency versus the length of the SOA is of the form,

η( L) ∝ ( g0 − α l ) L − Leff

η( dB) = C + 10 log10 ( L − Leff )2 ,

(3-10a)
(3-10b)

where C is a constant. High conversion efficiencies can thus be obtained by using longer
SOAs that operate in the gain saturation regime. Leff can be made smaller by strongly
amplifying the input waves before coupling them into the SOA waveguide. Thus the
SOA operates in transparency over most of the propagation length. Using longer SOAs
provides longer interaction length over which FWM can occur leading to higher
conversion efficiencies. At low detuning frequencies, the nonlinearities can be high from
the large conversion efficiencies, gain and longer interaction lengths. Figure 3.3 shows
many cascaded FWM processes that occur in a 1.5 mm SOA when two waves P1 and P2
are launched with a detuning of ~ 0.15 nm (or close to 18 GHz).

34
P2

-10

(3) P1

(3)

(5)

Intensity (dBm)

-20

(7 )

(5)

-30

(9)

(7)

-40

-50
1547

(11)

1547.5

1548

1548.5
Wavelength (nm)

1549

1549.5

1550

Figure 3.3: Many cascaded FWM processes can occur due to high
nonlinearity in a SOA when two waves P1 and P2 are launched

The above expressions are valid when phase matching between the wave-vectors
of the different frequencies is satisfied ( ∆k = 0). For broadband wavelength conversion
(detunings in the terahertz regime), phase matching is not necessarily satisfied and has to
be included in the expressions derived above. The expression for the conversion
efficiency is modified to

 2  ∆kL  
 Sin 
2 
η( L) ∝ ( g0 − α l ) ( L − Leff ) 
  ∆kL  
  2  

(3-11)

where ∆k = 2 k p − kq − kc , and is a function of the detuning frequency, Ω. The effect of
phase matching on the conversion efficiency was measured for a 2.2 mm and a 4 mm
long SOA and is shown in Figure 3.4.

Conversion efficiency (dB)

4 mm SOA

X 2.2 mm SOA

-10

XXXX
XX
XX
XXXX
XXXXXXXX
XXXXX
XXXX
XXXX
XX
XX

-20
-30
-40
-50
10

100

1000
Detuning (GHz)

10000

Figure 3.4: Conversion efficiency versus detuning for a 2.2 mm and 4 mm
long SOA. The SOAs are biased at 1 A

3.4 Optical-Signal-to-Noise-Ratio of the FWM Signal

Due to the high FWM conversion efficiencies possible by strongly saturating the SOA,
FWM is an ideal candidate for wavelength conversion since it has the advantage of being
transparent to bit-format. Another important requirement for implementation of SOAs in

36
real systems is the Optical-Signal-to-Noise-Ratios (OSNR) on the converted signals.
Since the SOA is composed of a gain medium, Amplified Spontaneous Emission (ASE)
can severely degrade the OSNR of the converted signal. OSNR can be increased by
strongly saturating the SOA [13]; thereby the ASE is generated over a very small length
of the device while the mixing occurs in the transparency region. The ASE in a
bandwidth ∆ν is given by the expression

ASE = (G-1) hν ∆ν,

(3-12)

where h is the Planck's constant, ν is the frequency, and G is the saturated gain of the
SOA. From Equations (3-7), (3-10a) and (3-12) the expression for the OSNR can also be
written as

 2  ∆kL  
 Sin 
2 
OSNR( L) ∝ ( g0 − α l ) ( L − Leff ) 
  ∆kL  
  2  

(3-13)

Thus longer devices operating under saturation are key to successful systems
applications. The small-signal gain coefficient can be written as g0 = a0 (n − n0 ) , where n
is the injection carrier density and n0 is the carrier density needed to offset the intrinsic
losses of the SOA waveguide. Higher conversion efficiency and OSNR can be obtained
by increasing the carrier density in the active region of the SOA.

3.5 Broad-band Wavelength Conversion Using Long SOAs

Long SOAs were fabricated to study conversion efficiencies at large detunings and obtain
high OSNR for broad-band wavelength conversion. A 2.2. mm long compressively
strained quantum well SOA was tested under high injection current (1 Amp). Figure 3.5
shows a 40 nm wavelength downshift with a 23.2 dB OSNR in a 0.1 nm Resolution
Bandwidth, which is sufficient for error free detection (Bit Error Rate < 10 -9) on the
converted signal [14].

2.2 mm SOA

23.2 dB S/N
@ 0.1 nm BW
40 nm

Figure 3.5: 40 nm downshift in a 2.2 mm long SOA with 23.12 dB OSNR
for low error-rate detection at 10 Gb/s (from Ref. [14])

3.6 Polarization Dependence of FWM

FWM is sensitive to the states of polarization (SOP) of the electric fields of the pump and
the probe. The FWM susceptibility is a tensor of rank four and the FWM process can be
written as

( 3)
Ei( c ) = χ ijkl
(Ω) E (j p ) Ek( p ) El( q )* .

(3-14)

The susceptibility tensor in a semiconductor can be derived from the matrix elements of
the dipole transitions between the valence and conduction bands. A detailed calculation
by Paiella et al. [15, 16] shows that the FWM susceptibility can be written as


χ
χ ijkl =  ∑ µivc µ cvj   ∑ µ kc' v' µlv' c'  ( χ CDM + χ CH ) +  ∑ µivc ( µ cvj ' µ kc' v + µ cj ' v µ kcv' ) µlv' c'  SHB
 c, v
 c , v , c' , v'
  c' , v'
 2
(3-15)

The Bloch states of the heavy and light hole valence bands in III-V semiconductors can
be written in terms of the total angular momentum eigenstates

3 3
,±
2 2

and

3 1
,± .
2 2

Based on these eigenstates the general solution of the coupled mode equations of FWM
can be written in the form [16]

*
Ei( s ) ( L) = ( E ( p ) (0)) ( E ( q ) (0)) ∑ ( pi Miikk pk qk* ) + Milil l ≠ i pl pi ql* + Milli l ≠ i pl2 qi*  , (3-16)
 k =1

where pi (qi) are the i-th component of the polarization vector of the pump (probe) waves,

Mijkl = χ ijkl Rijkl , where Rijkl is a propagation factor that accounts for the birefringence,
gain anisotropy and wavelength dependence on the gain coefficient and refractive index
of the SOA. The first term arises from the beating of the same components of the pump
and the probe, i.e., the TE (TM) component of the pump can form a grating only the TE
(TM) of the probe. The TE (TM) component of the pump can then be scattered by these
gratings into a FWM signal into the TE (TM) direction. This is true for CDM and CH
processes which are predominant in the sub-terahertz detuning range, and the ordinary
SHB process, which involves the modulation of the occupation probabilities of the
electronic states. For CDM and CH, the formation of the gratings and the subsequent
scattering of the pump into the FWM sideband are two distinct processes which can
involve different transition levels as indicated by the first term in Equation (3-15). In the
terahertz detuning range, where SHB is the dominant mechanism, gratings can also be
formed between the orthogonal TE and TM modes of the pump and the probe through the
modulation of relative coherence between the initial state and the second intermediate
state with opposite spin. These gratings then scatter the pump into FWM sidebands with
orthogonal polarization (given by the other terms in Equation (3-16)). However, in the
sub-terahertz regime, the contribution from the first term in Equation (3-16) dominates
and the other terms can be neglected. In general, the contribution from the last two terms
in Equation (3-16) can be much smaller than the first one due to birefringence in the
waveguide structure.

Bibliography

1. R.W. Boyd, Nonlinear Optics, Academic Press, London, UK, 1992.
2 . J. Reintjes, and L.J. Palumbo, "Phase Conjugation in Saturable Amplifiers by
Degenerate Frequency Mixing," IEEE J. Quantum Electron., QE-18, 1934-1940
(1982).
3. D. Campi, and C. Corriaso, "Wavelength Conversion Technologies," Photonic. Netw.
Commun., 2, 85-95 (2000).
4 . D.F. Geraghty, R.B. Lee, M. Verdiell, M. Ziari, A. Mathur, and K.J. Vahala,
"Wavelength Conversion for WDM Communication Systems using Four-Wave
Mixing in Semiconductor Optical Amplifiers," IEEE J. Sel. Top. Quant. Electron., 3,
1146-1155 (1997).
5. A. D'Ottavi, F. Girardin, L. Graziani, F. Martelli, P. Spano, A. Mecozzi, S. Scotti, R.
Dall 'Ara, J. Eckner, and G. Guekos, "Four-Wave Mixing in Semiconductor Optical
Amplifiers: A Practical Tool for Wavelength Conversion," IEEE J. Sel. Top. Quant.
Electron., 3, 522-528 (1997).
6 . S.J.B. Yoo, "Wavelength Conversion Technologies for WDM Network
Applications," J. Lightwave Technol., 14, 955-966 (1996).
7 . G.P. Agrawal, "Population Pulsation and Nondegenerate Four-Wave Mixing in
Semiconductor Laser Amplifiers," J. Opt. Soc. Am. B, 5, 147-159 (1988).
8. A. Uskov, J. Mork, and J. Mark, "Wave Mixing in Semiconductor Laser Amplifiers
due to Carrier Heating and Spectral-Hole Burning," IEEE J. Quantum Elect., 30,
1769-1781 (1994).

9. J.H. Zhou, N. Park, J.W. Dawson, K.J. Vahala, M.A. Newkirk, and B.I. Miller,
"Terahertz Four-Wave Mixing Spectroscopy for Study of Ultrafast Dynamics in a
Semiconductor Optical Amplifier," Appl. Phys. Lett., 63, 1179-1181 (1993).
10. J.H. Zhou, N. Park, J.W. Dawson, K.J. Vahala, M.A. Newkirk, and B.I. Miller,
"Efficiency of Broad-band Four-Wave Mixing Wavelength Conversion using
Semiconductor Traveling-Wave Amplifiers," IEEE Photonic. Tech. Lett., 6, 50-52
(1994).
11. C.H. Henry, "Theory of the Linewidth of Semiconductor Lasers," IEEE J. Quantum
Electron., QE-18, 259-264 (1982).
1 2 .A. Mecozzi, "Analytical Theory of Four-Wave Mixing in Semiconductor
Amplifiers," Opt. Lett., 19, 892-894 (1994).
13. A. D'Ottavi, P. Spano, G. Hunziker, R. Paiella, R. Dall 'Ara, G. Guekos, and K.J.
Vahala, "Wavelength Conversion at 10 Gb/s by Four-Wave Mixing over a 30-nm
Interval," IEEE Photon. Tech. Lett., 10, 952-954 (1998).
14. K. Vahala, R. Paiella, G. Hunziker, A. Bhardwaj, D. Dougherty, and U. Koren,
"Ultrafast Dynamics in Quantum Well Semiconductor Optical Amplifiers and
Applications," Paper WB1, Topical Meeting on Optical Amplifiers and their
Applications, Vail, Colorado, July 27-29, 1998.
15. R. Paiella, G. Hunziker, U. Koren, and K.J. Vahala, "Polarization-Dependent Optical
Nonlinearities of Multiquantum-Well Laser Amplifiers Studied by Four-Wave
Mixing," IEEE J. Sel. Top. Quant., 3, 529-540 (1997).

42
1 6 . R. Paiella, G. Hunziker, J.H. Zhou, K.J. Vahala, U. Koren, and B.I. Miller,
"Polarization Properties of Four-Wave Mixing in Strained Semiconductor Optical
Amplifiers," IEEE Photon. Tech. Lett., 8, 773-775 (1996).

Chapter 4

Advanced All-Optical Logic on a Spectral Bus
________________________________________________________________________

4.1 The Spectral Data Bus

Data transfer within a computer or between computers and its peripherals is done using a
parallel architecture, where byte-wide (or bit-parallel) transmission is used. A byte, which
is composed of several bits, is transmitted in one time slot using a data bus. Typically the
data-bus is made using a ribbon of electrical cables. Within a computer, the data-bus
forms a computer back-plane, over which data is transferred between different processing
units. The parallel architecture is also more suitable for the transfer of data between
computers, since it eliminates the need for serializer/deserializer circuits which can be
expensive as the processing speeds increase. Transfer of data between processors using
efficient protocols using byte-wide transmission over shared computer back-planes is
essential for Massively Parallel Processing (MPP) [1]. The throughput of future computer
networks for ultra high-speed distributed computing over a cluster of supercomputers will
be limited by the speed with which data can be transferred over the data bus. One way to
increase the transfer rate is to simply replace the electrical wires in the data-bus with
optical fibers to form a multifiber ribbon [2, 3]. However, multifiber ribbons are not

44
practical for long-distance links since they can be very expensive and it is difficult to
correct for bit-skew that can arise between different fibers of the ribbon. One possible
approach is to use serial transmission over the optical fiber instead of byte-wide
transmission. This would require the need for serializer/deserializer circuits at the
transmitting and receiving ends of the optical fiber to maintain the bit-parallel format at
the processing ends. As processor speeds increase, such circuits can be very expensive.
Furthermore, the speed at which these circuits operate determine the "electronic
bottleneck", i.e., the ultimate speed over which the cluster can operate.
The Spectral Data Bus was proposed by Loeb et al. [4] to increase the throughput
of high-speed links between computers using optical fibers while maintaining the bitparallel format for data transfer. Unlike conventional WDM systems, where the data on
different wavelength channels is uncorrelated, the spectral data bus assigns each bit of the
byte to be transmitted on a separate wavelength channel, as shown in Figure 4.1. Thus in
one time slot, the entire byte is transmitted in parallel over different WDM channels. This
approach eliminates the need for expensive serializer/deserializer circuits and multifiber
ribbons. The individual bits can be recovered using all-optical filters, which isolate single
WDM channels. Recent advances in optical device technology enable the integration of
WDM components such as multi-wavelength laser arrays [5-8], wavelength
multiplexers/demultiplexers [9-11] and photodiode arrays [12-14] on a single chip. Thus
the WDM bus replaces a serial link by trading the serializer/deserializer for wavelength
multiplexers/demultiplexers and trading the high-speed circuits for low-speed multiple
circuits in an array. The major limitation to the implementation of the spectral data bus is
the bit-skew that arises from fiber group velocity dispersion between different

45
wavelengths. However, techniques to compensate for dispersion, such as using dispersion
compensating fiber [15], chirped fiber gratings [16], or phase conjugation [17] have been
developed. Consequently, there has been an increased interest in realizing the potential
attributes of the spectral bus [1, 18-19].

E/O interface

O/E interface

λ1

λ2

λ3

Figure 4.1: Schematic of the Spectral Data Bus for byte-wide transmission
(C1, C2, C3 etc. are different WDM channels)

4.2 All-Optical Front-End Processing on the Spectral Bus

There are various front-end signal-processing possibilities associated with the spectral
bus, including byte-wide error correction and detection. A front-end device is defined as
one that is located between the bus and whatever the bus is linked to (presumably a

46
computer through an optoelectronic interface). The throughput of a computer network
can be further increased by moving some of the front-end digital processing functions to
the all-optical domain. Thus, all-optical front-end devices on both the transmitting and
receiving ends of the spectral bus as shown in Figure 4.2 could be realized. One example
of such devices could be to implement coding and decoding for error detection and
correction [20] or encryption for improved security.
In this very limited context, a form of optical logic based on Four-Wave Mixing
(FWM) designed to process data on-the-fly in the spectral form is proposed and studied.
Since FWM involves a third-order nonlinearity, a Non-Degenerate Four-Wave Mixing
(ND-FWM) process can involve a simultaneous interaction between three different
frequencies of light. Thus, up to three bits can be multiplied on the spectral bus using
ND-FWM. So, in principle, a three-bit Boolean operation is possible. ND-FWM on
intensity modulated data offers a simple "AND" operation between three bits and cannot
be easily generalized to implement other Boolean functions. However, if the information
on each bit is encoded in the State of Polarization (SOP) of the electric field rather than
the intensity of the WDM channel, the polarization-selection rules of the ND-FWM
process can be exploited to incorporate more versatile Boolean functions, since the SOP
of the FWM output is sensitive to the SOPs of all the fields involved in the mixing
process.

47
C1

E/O interface

O/E interface

encoder

decoder

front-end

front-end
λ1

λ2

λ3

Spectral Data Bus

Figure 4.2: All-optical logic circuits can be designed for some form of
limited on-the-fly front-end signal processing on a spectral bus (C1, C2,
C3 etc. are different WDM channels)

4.3 Polarization Shift Keying (PolSK)

The idea of encoding binary information on the state of polarization of the electric field
for transmission is not new. Polarization Shift Keying (PolSK) was proposed by
Benedetto et al. [21] for data transmission in long-haul networks, where information is
coded on a set of Stokes parameters that describe the electric field vector while the
amplitude of the electric field is kept constant. In this case, the Stokes parameters would
form a vector that would lie on a PoincarŽ sphere. In general, a mapping between m-ary
(m logic states) coded information with a constellation of m points on the PoincarŽ sphere
can be found [21]. It was further noted that when a polarized lightwave is sent through a

48
single-mode optical fiber, the SOPs of the electric fields are altered due to the presence of
birefringence over the transmission path. However, the fiber birefringence only causes a
rigid rotation of all the vectors on the PoincarŽ sphere and the spatial relationship
between the different vectors is preserved. Thus the birefringence would only cause a
uniform rotation of the entire constellation, assuming that polarization dependent gain or
loss were absent in the system. As a result, the PolSK coded information remains
uncorrupted and can be recovered with the design of suitable receivers that can recover
the Stokes parameters and decode the coded information electronically from the Stokes
parameters [22].
A simple example is binary PolSK (or 2-PolSK), where the "1" and "0" states are
represented by orthogonal linear states of polarization of the electric field, as shown in
Figure 4.3.

Binary “1”

Binary “0”

Figure 4.3: 2-PolSK coding using orthogonal linear states of polarization

49
System demonstrations of data transmission using 2-PolSK are predicted to have lower
power penalties compared to data transmission using intensity modulation [21, 22].
Higher extinction ratios and signal to noise ratios using 2-PolSK over intensity
modulation were experimentally demonstrated using a differential heterodyne detection
scheme [23, 24]. It was also shown that PolSK modulation is less sensitive to phase noise
of the transmitting lasers in comparison with other coherent transmission schemes [25].
Furthermore, since PolSK modulated electric fields have a constant amplitude, the effects
of Self-Phase Modulation (SPM) are absent in comparison to intensity modulation (IMDD systems) where SPM leads to pulse broadening which limits the bit rate over which
information can be transmitted.
The polarization states of the PolSK coded electric field is naturally scrambled in
comparison with IM-DD systems in which system performance can be degraded due to
polarization-hole-burning in EDFAs [26], where the signal experiences a lower gain than
the ASE noise copropagating in the orthogonal direction. PolSK modulation is immune to
polarization-hole-burning effects. Though 2-PolSK would be affected by Polarization
Mode Dispersion (PMD), it has been argued that for ultra long-haul transmission
systems, the pulse broadening due to SPM in IM-DD systems is larger than the
broadening of PolSK coded pulse due to Polarization Mode Dispersion (PMD). Thus,
long-haul data transmission using PolSK modulation has many advantages to be of
practical interest.
Though 2-PolSK transmission over single-mode optical fiber requires special
receivers to recover the Stokes parameters, direct detection can be employed with the use
of Polarization Maintaining (PM) fiber. A PM fiber uses an asymmetric core, which

50
breaks the degeneracy between its orthogonal modes thereby reducing the coupling of
power between the modes. Thus, if the orthogonal linear states of polarization
representing the binary "1" and "0" are aligned along the fast and slow axes (or principal
axes) of the PM fiber, mode coupling between the orthogonal modes is minimized in the
presence of birefringence. The SOPs of the "1" and "0" states would remain strictly
aligned along the principal axes of the PM fiber over the entire transmission length. In
this case, direct detection can be employed by simply inserting a polarizer correctly
aligned to the fast (or slow) axis of the PM fiber before the photodetector.
Though the installation of PM fiber for long-haul transmission (especially transoceanic transmission) seems unlikely in the very near future due to the high costs
involved, PM fiber offers a solution to circumvent PMD effects that occur in a single
mode fiber. The advantages of using 2-PolSK modulation using PM fibers could also be
realized for the different WDM channels of a spectral data bus.

4.4 All-Optical Processing on PolSK Coded Spectral Data

2-PolSK coding over a spectral bus using PM fiber is particularly attractive for
implementing limited front-end all-optical processing based on the polarization selection
rules of FWM for shared computer back-plane applications where the optical power in
each WDM channel is sufficiently high. One possible application is the design of alloptical coding and decoding circuits for error detection and correction. As a simple
example, a circuit that performs error detection and correction on a (3,1) Hamming Code

51
is designed to illustrate the potential offered by the proposed scheme. It is shown that the
PolSK coding and byte-wide format can be maintained using this scheme, which makes it
possible to generalize the scheme to enable more sophisticated Boolean operations
involving higher-level Hamming Codes.

4.5 Coding and Decoding Using the (3,1) Hamming Code

The (3,1) Hamming Code is the simplest of a class of linear Forward Error-Correcting
Codes (FECs) as proposed by R.W. Hamming [27], which allow for single errorcorrection. FECs are based on adding redundancy to the information being transferred.
This can then be used to correct for errors that can occur from the corruption of data
while transmitting over a noisy channel. For serial transmission, the added redundancy
comes with a cost of reduced channel speed while for a parallel architecture involving
byte-wide transmission, the channel speed can be maintained by increasing the word
length and adding more channels (more wavelength channels in case of the spectral bus).
The (3,1) Hamming Code allows for single error correction at a time by
transmitting each data bit with two check-bits with the same value. Thus the 3-bit word is
transmitted as a vector [0,0,0] or [1,1,1]. Single error correction is possible if the
Hamming distance (defined as the number of bits that differ between two legitimate code
words) is greater than or equal to 3 ( 2 m + 1 with m = 1 for one error). Furthermore, the
probability of two errors occurring simultaneously in the 3-bit word is far less than the
probability of one error occurring in the 3-bit word. For example, consider the case when

52
the transmitted word is [1,1,1] and the data-bit is corrupted over the channel so that the
received word is [0,1,1]. It is still possible to extract the information that was transmitted
(i.e., the 3-bit word was a [1,1,1]) by taking a "majority poll" of the bits in the 3-bit word.
The truth table for the (3,1) Hamming code is given in Table 4.1, where C1, C2, and C3
denote the channels that carry the data and check-bits, and EC denotes the error corrected
output. The (3,1) Hamming Code can also be interpreted as a projection of the 2 3 = 8
possible 3-bit words on the two originally transmitted 3-bit words [0,0,0] or [1,1,1] which
can be visualized using the cube shown next to the table in Figure 4.4. Thus a single error
creates a new word which lies on the edge of the cube containing the original code word.
Thus the Hamming distance from the correct code word is one, while its Hamming
distance from the incorrect code word is two.

C1 C2 C3 EC

[1,1,1]

[0,0,0]

Figure 4.4: Truth table and pictorial representation of the (3,1) Hamming Code

The Boolean function for this truth table can be written as in Equation (4-1),

EC = (C1 ∩ C 2) ∪ (C 2 ∩ C3) ∪ (C1 ∩ C3)

(4-1)

where "∩" denotes the Boolean "AND" function and "∪" denotes the Boolean "OR"
function. An implementation of the Boolean function in terms of the standard 2-input
"AND" and "OR" gates is shown in Figure 4.5.

C1
C2
C1
C3

C2
C3

Figure 4.5: Implementation of the Boolean function for the (3,1) Hamming
Code using 2-input "AND" and "OR" gates

We now proceed to design an error-correcting circuit for the (3,1) Hamming Code just
described, based on the polarization properties of ND-FWM.

4.6 Polarization Properties of ND-FWM Between 2-PolSK
Coded WDM Channels
Assuming that the 3-bit word is PolSK coded in a byte-wide format, the front-end
decoder circuit can receive a combination of 23=8 possible inputs, as shown in the table
in Figure 4.4, and should produce a unique output for each case. The polarization
selection rules of the ND-FWM process between three different wavelength channels that
carry 2-PolSK coded information can be used to implement the Boolean function for
error correction. If the bits are placed on spectral channels C1, C2 and C3 the ND-FWM
process creates an error-corrected channel (EC channel) whose electric field is given by
the expression

( 3)
El(ωC1)Em(ωC2)En*(ωC3),
Ek(ωEC=ωC1+ωC2-ωC3) ∝ χ klmn

(4-2)

where ω i, i = EC, C1, C2 and C3, is the angular frequency of the optical wave and (*)
( 3)
denotes complex conjugation. χ klmn
is the third-order nonlinear susceptibility, which is a

tensor of rank four and, as noted, is dependent on the states of polarization of the electric
fields of C1, C2 and C3. The geometry of the FWM process considered in this work is
such that the three input waves are launched into a single transverse mode waveguide
(here a Semiconductor Optical Amplifier) along the same direction of propagation. The
extracted product-wave hence propagates along the direction of incidence.

55
In a bulk SOA, the polarization dependence of the mixing product at ωEC is given
by

eˆEC ∝ (eˆC1 ⋅ eˆC* 3 )eˆC 2 + (eˆC 2 ⋅ eˆC* 3 )eˆC1 ,

(4-3)

where êi , i = EC, C1, C2, and C3, is the unit vector along the direction of the electric
field. This relation is strictly valid only when the unit vectors point along either the TE or
TM axes of the waveguide. The terms in Equation (4-3) can be physically interpreted as
the FWM signal at EC being generated as follows - C3 forms dynamic gain and index
gratings with C1 (or C2). Then, C2 (or C1) scatters off this grating to generate two FWM
side bands, one of them being at ω EC [28]. These processes are diagrammatically
represented in Figure 4.6.

EEC

(eˆC1 ⋅ eˆC* 3 )eˆC 2
EC2

EC1

(a)

EC3*

EEC

(eˆC 2 ⋅ eˆC* 3 )eˆC1
EC1

EC2

EC3*

(b)
Figure 4.6: Diagrammatic representations of the non-degenerate FWM
process where in (a) C3 forms a grating with C1 that scatters off C2 and in
(b) C3 forms a grating with C2 that scatters off C1, to form the FWM
sideband

In a Semiconductor Optical Amplifier (SOA), the unit vector êi , representing each
binary state, is aligned along the TE or TM direction of the waveguide structure to avoid
polarization walk-off of the incident fields arising from birefringence of the waveguide
[29]. The EC signal is generated in one of the following ways:

When all three input electric vectors are parallel (corresponding to identical bits on
each channel, i.e., no errors are present), the electric field of the mixing signal at ωEC
is parallel to the three inputs. This is used, in turn, to generate an output when no
error-correction is necessary.

When one of the electric fields is orthogonal to the other two (corresponding to an
error on that bit), a product wave at ω EC is generated only when C1 and C2 are
orthogonal. In this case, C3 creates a grating with either C1 or C2 (the one whose
polarization is parallel to C3), which scatters energy off the third wavelength to
generate a FWM signal at ω EC that is orthogonal to C3. This property is utilized to
correct for errors. When C3 is orthogonal to both C1 and C2, C3 cannot form a
grating with either C1 or C2 and thus no mixing at ωEC.

These different cases are shown in Figure 4.7,

no
FWM

C3
C2

Figure 4.7: Polarization selection rules for ND-FWM between C1, C2, and C3

4.7 Error-Correcting Circuit for the (3,1) Hamming Code

The error-correcting circuit requires at least three SOAs (only two are required if
retaining the PolSK format on the EC channel is not required) to generate the proper
FWM signal in all possible cases. The circuit is designed in such a way that the FWM
product at ω EC occurs in only one SOA at a time. This is done in order to avoid
interference of the desired FWM signal with additional spurious signals that would
degrade the performance of the circuit. This is accomplished by adding a pre-processing
element before each SOA, as shown in Figure 4.8.

non-correcting arm
Pre-processing
element

SOA 1

INPUT

OUTPUT

Pre-processing
element

SOA 2

correcting arm

Figure 4.8: Layout of the error-correcting circuit showing the "noncorrecting" and the "correcting arm"

59
One of the pre-processing elements is a polarizer with its transmission axis
aligned to either the fast or slow axis of the polarization maintaining (PM) fiber. The
other pre-processing element is a wavelength selective half-wave plate that will be
referred to as the "birefringent element". It acts as a half-wave plate for C3 and a fullwave plate for C1 and C2. The result is that the state of polarization of C3 gets rotated by
90 degrees (and thus inverts the binary state on C3) while that of C1 and C2 remain
almost unchanged.

The output of the circuit for each possible case is as follows:

In the absence of any errors, C1, C2, and C3 are parallel at the input and mixing at
ωEC occurs in the SOA after the polarizer (SOA 1 in Figure 4.8) whose axis coincides
with C1, C2, and C3. The mixing signal is parallel to the input bits and has the same
binary state as the input bits. Thus the output is generated without error correction
and it is for this reason that this arm is called the "non-correcting" arm. (It should be
noted that a polarizer changes PolSK modulation to Amplitude Shift Keying (ASK)
modulation. Hence, if PolSK modulation is to be preserved, two such non-correcting
arms are required, each with a polarizer as a pre-processing element aligned to the
slow and fast axes of the polarization maintaining fiber respectively.) Furthermore,
when C1, C2, and C3 (all being parallel) pass through the birefringent element, which
is the pre-processing element in the other arm, C3 becomes orthogonal to C1 and C2.

60
In this case, no mixing at ωEC takes place in the SOA after the birefringent element
(SOA 2 in Figure 4.8) in accordance with Equation (4-3).

In the presence of an error, C1, C2, and C3 will not all be parallel and thus one or
more of them will not pass through the polarizer. Hence no mixing will occur in SOA
1. There are two possible cases. When the error is on C3, it is orthogonal to both C1
and C2. After passing through the birefringent element, C3 will become parallel to C1
and C2 and the mixing signal in SOA 2 will have the same binary state as C1 and C2.
Thus an error on C3 will be corrected. When the error is on either C1 or C2, C3 will
align with the incorrect bit (since it gets inverted by the birefringent element) and will
form a grating which will scatter off the correct bit to give a mixing signal parallel to
the correct bit. Thus, an error-corrected signal is generated and hence the arm with the
birefringent element as the pre-processing element is called the "correcting arm".

The working of the error-correcting circuit described above is shown for a few different
3-bit received Hamming words in Figure 4.9.

C1 C2

C1 C2 C3

input [111]

input [110]

C1 C2

EC C1 C2 C3

input [101]
EC C1

C2 C3

correcting arm

EC C1 C2 C3

C1 C2

non-correcting arm

correcting arm
C3

Figure 4.9: Working of the error-correcting circuit for different input Hamming words

The Boolean function corresponding to the error correction is equivalent to the
computation of the "CARRY" bit of a 3-bit modulo-2 addition, can be written in terms of
triple-product Boolean operations as

EC = (C1∩C2∩C3) ∪ ( C1∩C2∩C3) ∪ (C1∩ C2 ∩C3) ∪ (C1∩C2∩ C3 ). (4-4)

62
Equation (4-4) can be easily derived by noting for a general Boolean expression, "A",

( A ∪ A ) = 1. Thus each 2-bit "AND" operation in Equation (4-1) can be rewritten as

(C ∩ C ) = (C ∩ C ∩ 1) = C ∩ C ∩ (C ∪ C ) = (C ∩ C ∩ C ) ∪ (C ∩ C ∩ C ),

(4-5)

where i, j, k = 1, 2, 3 and i≠j≠k. Adding the terms and noting that for a general Boolean
expression, "A", (A∪A) = A, Equation (4-4) is obtained.
It can then be seen that the Boolean operation implemented in the "non-correcting
arm" is (C1∩C2∩C3), while that implemented in the "correcting arm" is ( C1∩C2∩C3)
∪ (C1∩ C2 ∩C3) ∪ (C1∩C2∩ C3 ). Thus each triple-product Boolean function can be
associated with the terms in Equation (4-3), or a Feynman diagram shown in Figure 4.7
contributing to the FWM process.

Bibliography

1. L.A. Bergman, A.J. Mendez, and L.S. Lome, "Bit-Parallel Wavelength Links for
High Performance Computer Networks," Proc. SPIE Critical Review of Optical
Science and Technology, Optoelectronic Interconnects and Packaging, R.T. Chen
and P.S. Cuilfoyle, 2R62, 210-226 (1996).
2. N. Fujimoto, A. Ishizuka, H. Rokugawa, and K. Mori, "Skew-Free Parallel Optical
Transmission Systems," J. Lightwave Technol., 16, 1822-1831 (1998).
3. Y. Ota, and R.G. Swartz, "Multi-Channel Optical Data Link (MOD-LINK)," Proc.
OEC, 11D1-5, 42-43 (1990).
4. M.L. Loeb, and G.R. Stilwell, Jr., "High-Speed Data Transmission on an Optical
Fiber Using a Byte-Wide WDM System," J. Lightwave Technol., 6, 1306-1311
(1988).
5 . A.V. Krishnamoorthy, and K.W. Goossen, "Optoelectronic-VLSI: Photonics
Integrated with VLSI Circuits," IEEE J. Sel. Top. Quant. Electron., 4, 899-912
(1998).
6. S.Y. Hu, J. Ko, E.R. Hegblom, and L.A. Coldren, "Multimode WDM Optical Data
Links with Monolithically Integrated Multiple-Channel VCSEL and Photodetector
Arrays," IEEE J. Quantum Elect., 34, 1403-1414 (1998).
7. C.E. Zah, M.R. Amersfoort, B.N. Pathak, F.J. Favire, P.S.D. Lin, N.C. Andreadakis,
A.W. Rajhel, R. Bhat, C. Caneau, M.A. Koza, and J. Gamelin, "Multiwavelength
DFB Laser Arrays with Integrated Combiner and Optical Amplifier for WDM Optical
Networks," IEEE J. Sel. Top. Quant. Electron., 3, 584-597 (1997).

64
8. M. Zirngibl, "Multifrequency Lasers and Applications in WDM Networks," IEEE
Commun. Mag., 36, 39-41 (1998).
9. P. Bernasconi, C. Doerr, C. Dragone, M. Cappuzzo, E. Laskowski, and A. Paunescu,
"Large N x N Waveguide Grating Routers," J. Lightwave Technol., 18, 985-991
(2000).
10. R. Adar, C.H. Henry, C. Dragone, R.C. Kistler, and M.A. Milbrodt, "Broad-band
Array Multiplexers made with Silica Wave-guides on Silicon," J. Lightwave
Technol., 11, 212-219 (1993).
11. E.S. Koteles, "Integrated Planar Waveguide Demultiplexers for High-Density WDM
Applications," Fiber Integrated Opt., 18, 211-244 (1999).
12. M.J. Cohen, "Photodiode Arrays Help Meet Demand for WDM," Laser Focus World,
36, S7−+ (2000).
13. M. Zirngibl, C.H. Joyner, and L.W. Stulz, "WDM Receiver by Monolithic Integration
of an Optical Preamplifier, Wave-guide Grating Router and Photodiode-Array,"
Electron. Lett., 31, 581-582 (1995).
14. C.R. Doerr, M. Zirngibl, C.H. Joyner, L.W. Stulz, and H.M. Presby, "Polarization
Diversity Waveguide Grating Receiver with Integrated Optical Preamplifiers," IEEE
Photonic. Tech. Lett., 9, 85-87 (1997).
15. R.J. Nuyts, Y.K. Park, and P. Gallion, "Dispersion Equalization of a 10 Gb/s
Repeatered Transmission System Using Dispersion Compensating Fibers," J.
Lightwave Technol., 15, 31-42 (1997).

65
16. K.O. Hill, F. Bilodeau, B. Malo, T. Kitagawa, S. Theriault, D.C. Johnson, J. Albert,
and K. Takiguchi, "Chirped in-fiber Bragg Gratings for Compensation of OpticalFiber Dispersion," Opt. Lett., 19, 1314-1316 (1994).
17. A. Yariv, D. Fekete, and D.M. Pepper, "Compensation for Channel Dispersion by
Nonlinear Optical Phase Conjugation," Opt. Lett., 4, 52-54 (1979).
18. G. Jeong, and J.W. Goodman, "Long-Distance Parallel Data Link Using WDM
Transmission with Bit-Skew Compensation," J. Lightwave Technol., 14, 655-660
(1996).
19. C.C. Chen, L.A. Wang, and S.Y. Kuo, "A Wavelength Encoded Multichannel Optical
Bus for Local Area Networks," J. Lightwave Technol., 14, 315-323 (1996).
20. S.K. Shao, and M.S. Kao, "WDM Coding for High-Capacity Lightwave Systems," J.
Lightwave Technol., 12, 137-148 (1994).

21. S. Benedetto, and P. Poggiolini, "Theory of Polarization Shift Keying Modulation,"
IEEE T. Commun., 40, 708-721 (1992).

22. S. Benedetto, R. Gaudino, and P. Poggiolini, "Direct Detection of Optical Digital
Transmission Based on Polarization Shift Keying Modulation," IEEE J. Sel. Area
Comm., 13, 531-542 (1995).

23. R. Calvani, R. Caponi, and F. Cisternino, "Polarization Phase-Shift Keying: A
Coherent Transmission Technique with Differential Heterodyne Detection," Electron.
Lett., 24, 642-643 (1988).

66
24. E. Dietrich, B. Enning, R. Gross, and H. Knupke, "Heterodyne Transmission of a 560
Mbit/s Optical Signal by means of Polarization Shift Keying," Electron. Lett., 23,
421-422 (1987).
25. S. Benedetto, R. Gaudino, and P.Poggiolini, "Performance of Coherent Optical
Polarization Shift Keying Modulation in the Presence of Phase Noise," IEEE T.
Commun., 43, 1603-1612 (1995).
26. V.J. Mazurczyk, and J.L. Zyskind, "Polarization-Dependent Gain in Erbium-Doped
Fiber Amplifiers," IEEE Photonic. Tech. Lett., 6, 616-618 (1994).

27. R.W. Hamming, "Error Detecting and Error Correcting Codes," Bell Syst. Tech. J.,
29, 147-160 (1960).

28. G.P. Agrawal, "Population Pulsation and Nondegenerate Four-Wave Mixing in
Semiconductor Laser Amplifiers," J. Opt. Soc. Am. B, 5, 147-159 (1988).

29. S. Diez, C. Schmidt, R. Ludwig, H.G. Weber, P. Doussiere, and T. Ducellier, “Effect
of Birefringence in a Bulk Semiconductor Optical Amplifier on Four-Wave Mixing,”
IEEE Photonic. Tech. Lett., 10, 212-214 (1998).

Chapter 5

The Error-Correcting Circuit for the (3,1)
Hamming Code
________________________________________________________________________

5.1 Introduction

The error-correcting circuit for the (3,1) Hamming Code described in the previous
chapter consists of two "non-correcting arms" and one "correcting arm". Since the errorcorrecting circuit is a front-end device operating before photodetection, maintaining the
PolSK modulation on the error-corrected (EC) channel is not required. Thus, as discussed
previously, the non-correcting arm for the case when the received 3-bit word is [0,0,0] is
not required. Thus the circuit consists of only one "correcting" and one "non-correcting"
arm and the output on the EC channel is converted to intensity modulation before
filtering the EC channel optically and detecting it with a photodetector. Details of the
construction of the circuit are presented along with results of experimental demonstration.

68
Figure 5.1 shows the details of the experimental setup. The setup consists of three
distinct parts, (i) the transmitter which generates the PolSK modulated 3-bit word, (ii) the
front-end error-correcting circuit preceding the receiver, and (iii) the receiver.

SATURATING
WAVE

ERROR CORRECTING
CIRCUIT
PM-EDFA

DFB
TRANSMITTER
LASER
SOURCES

SOA 1

POLARIZER
PM-EDFA

MODULATOR 1
C3

VARIABLE
DELAY

PM-EDFA

SOA 2

MODULATOR 2

BIREFRINGENT
ELEMENT

POLARIZER

10% TAP

BPF

PREAMPLIFIED
RECEIVER
Bit
ERROR
RATE
TESTER

OPTICAL
SPECTRUM
ANALYZER

Figure 5.1: Schematic of the error-correcting circuit

VARIABLE
ATTENUATOR

69
The PolSK transmitter and the error-correcting circuit are built using polarization
maintaining (PM) components to preserve PolSK modulation and add robustness to the
setup. Besides the Optical-Signal-to-Noise (OSNR) of each channel, another important
characteristic of wavelength channels is the Extinction Ratio (ER) of the States of
Polarization (SOP) of the individual PolSK coded bits. The ER of a PolSK bit is defined
as the ratio of the optical power present along the fast (slow) axis to the optical power
present along the orthogonal axis. For power levels where OSNR of the individual
channels is sufficiently high, the degradation of the ER of the PolSK bits during
propagation will determine the systems performance of the error-correcting circuit. High
ER of each PolSK bit must be maintained to ensure a high contrast between the "1"s and
"0"s for error-free detection. In reality, perfect extinction (power is present purely along
the fast or the slow axis) is not achieved. Furthermore, every PM component degrades the
ER of the light that propagates through it, either due to birefringence effects or due to
imperfect alignment. Thus, it is important to characterize the effect of the PM
components used and the FWM process on the ER of the PolSK bits.

5.2 Characterization of the Extinction Ratio

One way to characterize each component is to measure the ER of light at the output of the
PM component assuming that perfectly polarized light ( ER = ∞ ) was launched at the
input. This is defined as the Extinction Ratio (ER) of the PM component. The ER of a
PM component indicates its ability to preserve the Extinction Ratio of the PolSK bits.

70
Thus the characterization of the ER of each component used or built in the setup is
important.
The Extinction Ratio of a PM component can be measured using a Modular
Polarization Analyzer. This instrument measures the State of Polarization (SOP) of the
iδ

incident light and extracts the Stokes parameters. If Ex eiδ x and Ey e y are the components
of the electric field along the principal axes of a PM fiber, the Stokes parameters [1] are
given as

s0 = Ex2 + Ey2 , s1 = Ex2 − Ey2 , s2 = 2 Ex Ey cos δ x − δ y , s3 = 2 Ex Ey sin δ x − δ y .

(6-1)

Since s02 = s12 + s22 + s32 for a well-defined SOP, the Stokes parameters can be mapped to a
PoincarŽ sphere, where s1, s2 and s3 normalized to s0 are the spatial co-ordinates

(sˆ1, sˆ2 , sˆ3 ) . For purely linear y-polarized light ( ER = ∞ ), the Stokes vector is (-1, 0, 0)
while for purely x-polarized light, the Stokes vector is (1, 0, 0), and the points lie on
opposite ends of the equator of the PoincarŽ sphere. For light with a finite extinction
ratio, any induced change of birefringence in the fiber (e.g., by the application of stress
on the PM fiber) will change the phase factors δ x and δ y resulting in a change of the
Stokes parameters. Thus s0 and s1 remain unchanged and the precession of the Stokes
vector, (sˆ1 , sˆ2 , sˆ3 ) , about axis ŝ1 will form a circle with a radius R equal to
2 Ex Ey

( E + E ) . For a high extinction ratio it then follows that R ≈ 2 ER. Thus,

ER ≈ 2 R , i.e., the ER is inversely proportional to the radius of the circle formed on the

PoincarŽ sphere. A Polarization Analyzer deduces the ER of light after measuring the

71
radius of the circle that is formed on the PoincarŽ sphere while stress is applied to the PM
fiber [2].

5.3 The PolSK Transmitter

Three New Focus external cavity tunable diode lasers are used to generate the
wavelength channels for the 3-bit Hamming word. PolSK modulation is achieved by
coupling the DATA and DATA outputs of a dual output Mach-Zehnder electro-optic
modulator (UTP) along the fast and slow axes of a polarization maintaining fiber, using
an in-fiber polarization beam combiner. Since the DATA and DATA outputs of the
modulator are intensity modulated, energy is present only along the fast or the slow axis
of the PM fiber. To introduce errors, one of the laser sources is connected to modulator 1,
while the remaining two sources are connected to modulator 2. Non-return-to-zero (NRZ)
data streams at 2.5 Gbit/s are generated using a dual output pattern generator. A variable
delay line is further introduced between the pattern generator output and modulator 1 to
shift the data streams temporally with respect to each other. This enables the addition of
random errors on one of the bits of the 3-bit word. High ER on the PolSK bits is obtained
by carefully adjusting the DC bias and the amplitude of the voltage generated from the
pattern generator applied on the modulators. The values are optimized by minimizing the
Bit Error Rates (BER) measured immediately after transmission. PolSK modulation is
converted to Amplitude Modulation by passing the PolSK bits through an in-fiber
polarizer and is detected using a high-speed photodetector. Figure 5.2 shows the PolSK
transmitter.

Figure 5.2: The PolSK transmitter

Figure 5.3 shows the measured Bit Error Rate (BER) versus received power for the
pseudo random PolSK modulated data (27-1 PRBS) at 2.5 Gbit/s on channel C1.

-Log(BER)

12
-44 -42 -40 -38 -36 -34 -32 -30
received power [dBm]

Figure 5.3: BER versus received power (in 0.5 nm Resolution Bandwidth)
for PolSK modulated data at 2.5 Gbit/s measured immediately after
transmission on channel C1

5.4 Building the Error-Correcting Circuit

In this section, the pre-processing elements and the Polarization Maintaining ErbiumDoped Fiber Amplifiers (PM-EDFAs) used in building the error-correcting circuit are
described in detail.

74
A. The Pre-Processing Elements

The pre-processing element in the "non-correcting arm" is a fiber-polarizer whose axis
coincides with one of the principal axes of the PM fiber. The ER of the fiber-polarizer is
greater than 40 dB, or four orders of magnitude.
The pre-processing element in the "correcting arm" is a birefringent element that
acts as a wavelength selective half-wave plate. The birefringent element is prepared by
splicing the principal axes of a polarization maintaining Bow-tie fiber at an angle of 45
degrees on both sides with respect to the principal axes of Polarization Maintaining
Panda fiber used in the rest of the setup, as shown in Figure 5.4. The length of the Bowtie fiber is 75 cm. The birefringent element is temperature controlled to enable tunability
and increased stability.

75 cm Bowtie fiber

Out
45 degree
cross-splice

45 degree
cross-splice

Figure 5.4: Schematic of the birefringent element

The response of the birefringent element was measured after it was inserted in the
correcting arm using the setup shown in Figure 5.5. ASE from a PM-EDFA was
polarized using a fiber-polarizer and launched into the correcting arm. The output of the
"correcting arm" was analyzed using an Optical Spectrum Analyzer (OSA) after passing
it through another fiber-polarizer. The output spectrum is shown in Figure 5.6.

PM-EDFA

polarizer

correcting
arm

OSA

EDFA-ASE

Figure 5.5: Setup to measure the response of the birefringent element

response (dBm)

-30
-40
-50
-60
-70
1543

1545

1547 1549 1551
wavelength (nm)

1553

Figure 5.6: Response of the birefringent element measured by the OSA

B. Polarization Maintaining EDFA (PM-EDFA)

The PM-EDFAs used in the experiment were built with PM components using a
backward pumping scheme as shown in Figure 5.7(a). Six meters of PM Erbium-doped
fiber with an elliptic core from Lucent (R37PM01) was pumped with a high-power (200
mW) 1480 nm laser diode from Sumitomo. Figure 5.7(b) shows the PM-EDFA.

1480 pump
PM-EDF

out
PM isolator

PM-WDM

PM isolator

Figure 5.7(a): Schematic of the backward pumped PM-EDFA

Figure 5.7(b): The PM-EDFA

78
The axes-alignment routines of the PM splicer (Ericsson FSU 900) could not
identify the orientation of the principal axes of the PM Erbium-doped fiber. The
alignment of the principal axes of the PM Erbium-doped fiber with those of the PM fiber
during splicing was done manually. Linearly polarized light with high ER (>40 dB) in the
Panda fiber was coupled into the PM Erbium-doped fiber pumped with the 1480 nm laser
diode. The coupling was done by manually aligning the cores of the two fibers through
the microscope in the splicing machine. The ER of the amplified output was analyzed on
the PoincarŽ sphere by applying stress on the PM Erbium-doped fiber. To enable accurate
measurement of the ER, the amplified output was analyzed in an optical spectrum
analyzer to ensure that the ASE was suppressed and high OSNR (>40 dB in 0.1 nm
Resolution Bandwidth) on the amplified signal. The PM Erbium-doped fiber was
manually rotated while ensuring suppression of ASE and the ER of the amplified output
was maximized by minimizing the radius of the circle on the PoincarŽ sphere traced by
the application of stress on the fiber. The splicing procedure was repeated on the other
end of the PM Erbium-doped fiber.
The ER of the PM-EDFA was then measured by launching linearly polarized light
with high ER (>40 dB) and looking at the circle traced on the PoincarŽ sphere by
applying stress on the fiber at the output of the PM-EDFA as shown in Figure 5.8(a). The
result is shown in Figure 5.8(b), and ER better than 19 dB was obtained for the PMEDFAs. Figure 5.9 shows the gain and noise figure characteristics of the PM-EDFAs
measured for the 1480 nm pump laser biased at 400 mA.

Polarization
Analyzer
polarizer

PM-EDFA

ECL

ER>40 dB

stress

Figure 5.8(a): Setup to measure the ER of the PM-EDFA

Figure 5.8(b): Circle traced on the PoincarŽ sphere shows 19 dB ER for
PM-EDFA

X gain (dB)

Gain and Noise-Figure (dB)

35
30

X XX X X X

noise-figure (dB)

15
10
-35

-30

-25 -20 -15 -10
Input power (dBm)

-5

Figure 5.9: Gain and Noise Figure of the PM-EDFA

A polarizer is introduced after SOA 2 in the correcting arm to convert PolSK
modulation to Amplitude Shift Keying (ASK) modulation. The lengths of the two arms in
the circuit are synchronized to within 20 picoseconds, which is about 1/20-th of a bit
period at 2.5 Gbit/s. This reduces undesired overlap between two temporally adjacent 3bit words.

5.5 Experiment

The FWM process takes place in a Semiconductor Optical Amplifier (SOA) in each arm.
High FWM conversion efficiency is achieved by using 1.5 mm long bulk SOAs from
Optospeed biased at 650 mA. High Optical-Signal-to-Noise Ratio (OSNR) is achieved by
fully saturating the SOAs after preamplifying the input channels [3] using the PMEDFAs. The principal axes of the PM fibers are carefully aligned with the TE and TM
axes of the SOA waveguide to avoid polarization walk-off of the incident fields arising
from the birefringence of the waveguide [4]. The wavelengths for C1 and C2 are 1547.43
nm and 1547.85 nm (birefringent element is almost a full-wave plate) while that of C3 is
1550.47 nm (birefringent element is a half-wave plate). Note that the SOA in the
correcting arm is always saturated since there is always power incident on it. However,
the SOA in the non-correcting arm is not saturated when all the bits are identically "0", as
the polarizer before it does not allow any channel to pass through and be amplified. This
leads to a modulation of Amplified Spontaneous Emission (ASE) from SOA 1, because
its gain recovery time is comparable to the bit rate. Thus an additional CW laser is
coupled into SOA 1, which is called the "saturating-wave laser" (see Figure 5.1). Its
power is adjusted such that it is low compared to the power in the other channels, but
high enough to ensure saturation of SOA 1. Its wavelength is carefully selected at 1558.5
nm so that the additional FWM sidebands that it generates do not interfere with the EC
channel. Figure 5.10 shows the optical spectrum at the output of SOA 1 (in 0.1 nm
Resolution Bandwidth) when the 3-bit word is [1,1,1]. The wavelength channels C1, C2,
and C3 along with the desired FWM process at ωEC are marked in the spectrum.

-5

-10
Intensity (dBm)

-15
-20

-25
-30
-35
-40
-45
-50
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553
Wavelength (nm)

Figure 5.10: Optical spectrum at the output of the SOA (in 0.1 nm
Resolution Bandwidth) showing the wavelengths channels [C1-C3] and
EC (from Ref. [7])

The SOAs are further tested for mode conversion effects [5], which would lead to
a degradation of the extinction ratio of the PolSK signals. TE polarized light is launched
into the SOAs operating under identical conditions as used in the experiment. Using
another polarizer at the output of the SOAs, the TM component is found to be at least 30
dB lower than the TE component, which is of the same order as the extinction ratio of the
polarizers used in the measurement. For TM polarized light launched into the SOAs, the
TE component at the output is also found to be at least 30 dB lower than the TM

83
component. Hence, we conclude that mode conversion is not significant in the devices
used in this experiment.
The output from each arm is combined using a polarization maintaining coupler.
Spurious interference can occur between the two arms due to the presence of power in the
EC channel originating in the arm where FWM is not supposed to occur. This is because
the state of polarization on each channel is not perfectly linear along the fast or the slow
axis of the PM fiber, leading to a residual power in the orthogonal direction. This
interference is minimized by coupling the EC channel in each arm to the orthogonal axes
of the PM fiber. The error-corrected channel is filtered and detected using a pre-amplified
receiver. The "correcting arm" of the error-correcting circuit is shown in Figure 5.11.

Figure 5.11: The Òcorrecting armÓ of the error-correcting circuit

5.6 Results and Discussions

The performance of error-correcting circuit was first tested with CW waves to evaluate
the ability of the components to control and preserve the States of Polarization (SOP) on
the three channels C1, C2, and C3. The SOP of each channel was adjusted by changing
the DC bias on the Mach-Zehnder modulators and the FWM signal generated in each
SOA of the "non-correcting arm" and the "correcting arm". Figure 5.12 shows the optical

85
spectra obtained when the 3-bit word was adjusted to be [1,1,1]. Figure 5.12(a) shows the
FWM signal generated in the "non-correcting arm" while Figure 5.12(b) shows the FWM
signal generated in the "correcting arm".

Intensity (dBm)

C1 C2

-10
-20

-30
-40
-50
-60
1540

1545
1550
Wavelength (nm)

1555

Figure 5.12(a): Optical Spectrum after the SOA (in 0.1 nm Resolution
Bandwidth) in the "non-correcting arm" for the received word [1,1,1]
(after Ref. [6])

Intensity (dBm)

C1 C2
C3

-10
-20

no
FWM

-30
-40
-50
-60
1540

1545
1550
Wavelength (nm)

1555

Figure 5.12(b): Optical Spectrum after the SOA (in 0.1 nm Resolution
Bandwidth) in the "correcting arm" for the received word [1,1,1] (after
Ref. [6])

The results of Figure 5.12 show a good amplitude Extinction Ratio (~20 dB) on
the FWM signal generated in the two arms for the input word being [1,1,1]. Dynamic
performance of the error-correcting circuit was tested at 2.5 Gbit/s. Figure 5.13(a) shows
an 8-bit pattern [10011100] that was encoded on channels C1 and C2 using modulator 1
(see Figure 5.1). The encoded information on channel C3 was inverted by adjusting the
DC bias on modulator 2 and the 8-bit pattern on C3 is shown in Figure 5.13(b). Thus the
data encoded on channel C3 is always erroneous with respect to the data on channels C1
and C2. The 8-bit pattern recovered on the FWM signal after combining the "non-

87
correcting arm" and the "correcting arm" is shown in Figure 5.13(c) and it corresponds to
the pattern on channels C1 and C2. Hence the information coded on the 3-bit Hamming
word is corrected for single errors.

150
100
50
-50
-100
-150

1000

2000

3000

4000

3000

4000

time (ps)

(a)

150
100
50
-50
-100
-150

1000

2000
time (ps)

(b)

150
100
50
-50
-100
-150

1000

2000

3000

4000

time (ps)

(c)

Figure 5.13: Oscilloscope traces of the 8-bit pattern for (a) erroneous data
on channel C3, (b) data on channels C1 and C2, and (c) error-corrected
FWM signal (from Ref. [6])

Figure 5.14(a) shows a 16-bit pattern [1001110011110000] at 2.5 Gbit/s on each
channel. The variable time delay is adjusted so that there is a one-bit delay on C3 relative
to C1 and C2. Thus, C3 is the channel which has occasional errors. The resultant patterns
on ωEC, which are obtained from the non-correcting and correcting arm separately, are
shown in Figure 5.14(b) and 5.14(c), respectively. Figure 5.14(d) shows the pattern on
ωEC after both arms have been combined which is identical to the pattern on C1 and C2.
This shows that the data stream with errors was corrected.

Voltage (linear units)

• C1&C2
••
•••• °°°°°° ••••••••°•°•°••°•°•°•°•°•°•°•°•°°°°°°°°° ••••••°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°•°°°°°°°°
•• °°
• •°°° °° •• °° • °° ••• °°
• °° ° C3
••• •° ° • °° • ° • °°
• •°•° ° •• ° •• ° ••
••• °
• °°
• ° •• °
•• °°• ° • °
•• °°°
••• ° ••• °°
• °°• ° •• °
•• °
• ° • °°
°°•• ° •• °
• ° •• °• °°
•• •°°°
•• ••°°•• °°
••••••°••••°•°•°•°•°••
••••••°°•°•°°°°°°
•°°•°•°•°•°•°•°•°••°°°°°°° •••••••••°°•°•°•°°°°°°
°°°° °

2000

4000
Time (ps)

6000

8000

Voltage (linear units)

(a)

••••••
•••••
•••• ••••
••••• •••••••••••••
• •
•••
••• ••
••
••
••
••
••
••
••
• •• ••
•• •
••
••••••••••••••••••
•••• •••••••
•••••••••••••••••••••••••••••••••••••••••••••
••

2000

4000
Time (ps)

6000

8000

Voltage (linear units)

(b)

•••
•••
•••• ••
•••
•••• •••
• •
•• •
•••••
••
•• •••
•• ••
••
• •••
• •
•• •
• ••• ••• ••
• •
•• ••• • •
• •
•• •
•• ••• • ••
• ••
• •• ••
••
• ••
•••••••••••••••••••••••••••••••••••••••
••••••••••••••••••••••••
•••••••••••• •••••••••••

2000

4000
Time (ps)

(c)

6000

8000

Voltage (linear units)

••••••
••••••••
•••• •••••••••••••••••••••••
•••••••
•••••• •••••••••••
• •
••
•• •
••••
••
••
• ••
••
••
••
•••
••
••
•• •• ••
••
•• • •
•• •
• • •
••• •••
•• • •
••••
••••••
•••••••••••••
••••••••••••••••
•••••••••• ••••••••••••••
••

2000

4000
Time (ps)

6000

8000

(d)

Figure 5.14: Oscilloscope traces of (a) 16-bit patterns on channels C1, C2,
and C3 at 2.5 Gbit/s, (b) EC output from the non-correcting arm, (c) EC
output from the correcting arm, and (d) EC output from both arms
combined (from Ref. [7])

We further demonstrate the dynamic operation of this circuit by modulating C1,
C2, and C3 with a pseudo-random bit stream (27-1 PRBS), with a one bit delay on C3
relative to C1 and C2. In this case the binary state on C3 is complementary to the state on
C1 and C2 approximately 50 percent of the time. Figure 5.15(a) shows the Bit Error Rate
(BER) versus received power (in 0.5 nm Resolution Bandwidth) of the EC channel for
this case. This is compared to the case when there is no errors on C3 relative to C1 and
C2. Detection with a low Bit Error Rate <10-9 is demonstrated despite a 50 percent error
rate on the received word. Similar results are obtained by modulating C1, C2 and C3 with
pseudo-random bit streams with a one bit delay on C1 relative to C3 and C2, in which
case the binary state on C1 is complementary to that on C2 and C3 approximately 50

91
percent of the time. The Bit Error Rate for this is shown in Figure 5.15(b). The slight
degradation after error correction in Figure 5.15(b) compared to Figure 5.15(a) can be
explained by the lowering of FWM efficiency in the correcting arm when C1 and C2 are
orthogonal compared to when C1 and C2 are parallel, as determined by Equation (4-3).
Figure 5.16 shows the eye diagram obtained on the EC channel after error correction.

no errors on C3

° 50% error on C3
-Log(BER)

-44

-42

-40
-38
-36
-34
Received power (dBm)

-32

-30

Figure 5.15(a): BER versus received power (in 0.5 nm Resolution
Bandwidth) at 2.5 Gbit/s for random errors on C3 (from Ref. [7])

no errors on C1

° 50% error on C1
-Log(BER)

°°

-44

-42

-40
-38
-36
-34
Received power (dBm)

-32

-30

Figure 5.15(b): BER versus received power (in 0.5 nm Resolution
Bandwidth) at 2.5 Gbit/s for random errors on C1 (from Ref. [7])

X XXX XXXXXXXX XX XXXXXXXXXXXXXXX XXXXXXXXXX
XXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXX X XXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXX
100 X
XX XXXX X XXXXXXX XX XXX XXXXX XXXXX XXXXXX
XXX XX
XX
XXX XXX
50
XX XXXX
XXXX XXXX
XXXXXX
XXXXX
XXXXXX
XX
XX
XXX
X XXXX
XXX XX
XX X
XX XXXXX
-50
XXXXXXXXX XXXXXX XXXXXX X XXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
-100 X
XXX X X XX XXX X X X XXXXXX X X

Voltage (mV)

150

-150

200

400

600
800
Time (ps)

1000

1200

Figure 5.16: Eye diagram after error-correction on the EC channel

The operation of the logic circuit is further tested when the information on C3 is
severely distorted, so that a Bit Error Rate no better than 30 percent could be achieved on
C3. This is achieved by changing the DC-bias of the Mach-Zehnder modulator. Thus the
information on C3 is ambiguous in that there are no clearly defined binary states on it.
Pseudo-random bit streams on C1 and C2 were detected to be error-free upon
transmission through the circuit. Figure 5.17 shows the Bit Error Rate versus received
power on the EC signal for this case. Once again a low Bit Error Rate <10-9 on the
mixing signal for ambiguous data on C3 and error-free data on C1 and C2 is
demonstrated. This shows that error-correction on certain ill-defined states is also
possible.

-Log(BER)

-44

-42

-40
-38
-36
-34
Received power (dBm)

-32

-30

Figure 5.17: BER versus received power (in 0.5 nm Resolution
Bandwidth) at 2.5 Gbit/s for error correction on ill-defined states with
30% errors on C3 (from Ref. [7])

Bibliography

1. J.D. Jackson, Classical Electrodynamics, John Wiley & Sons, New York, NY, 1998.
2. "Modular Polarization Measurement System, PAT 9000," Operation Manual, Profile
Optische Systeme GmbH, 1998.
3. A. D'Ottavi, P. Spano, G. Hunziker, R. Paiella, R. Dall' Ara, G. Guekos, and K.J.
Vahala, "Wavelength Conversion at 10 Gb/s by Four-Wave Mixing Over a 30-nm
Interval," IEEE Photonic. Tech. Lett., 10, 952-954 (1998).
4. S. Diez, C. Schmidt, R. Ludwig, H.G. Weber, P. Doussiere, and T. Ducellier, "Effect
of Birefringence in a Bulk Semiconductor Optical Amplifier on Four-Wave Mixing,"
IEEE Photonic. Tech. Lett., 10, 212-214 (1998).
5 . H. Soto, D. Erasme, and G. Guekos, ÒCross-Polarization Modulation in
Semiconductor Optical Amplifiers,Ó IEEE Photonic. Tech. Lett., 11, 970-972 (1999).
6. A. Bhardwaj, P.O. Hedekvist, H. Andersson, and K.J. Vahala, "All Optical Front End
Error Correction on a Spectral Data Bus," Paper CWI5, presented at the Conference
on Lasers and Electro-Optics, San Francisco, CA, May 7-12 (2000).

7. A. Bhardwaj, P.O. Hedekvist, and K. Vahala, "All-Optical Logic Circuits Based on
Polarization Properties of Nondegenerate Four-Wave Mixing," J. Opt. Soc. Am. B,
18, 657-665 (2001).

Chapter 6

All-Optical Logic Circuits Based on Polarization
Properties of ND-FWM
________________________________________________________________________

6.1 Introduction

The error-correcting circuit for the (3,1) Hamming code based on the polarization
properties of ND-FWM can be further generalized to implement other Boolean functions,
such as a 3-bit addition. The scheme can also be used to implement more than one
Boolean operation simultaneously in the same set of SOAs using different ND-FWM
processes. Higher level Hamming codes can be implemented using the 3-bit adder
circuits as fundamental building blocks and it will be seen that the circuit design for the
(7,4) Hamming Code exhibits the symmetries of the error-correcting code. Pertinent
issues such as the cascading of such circuits and degradation of the Extinction Ratio (ER)
of the PolSK coded bits after each cascade, are investigated further.

6.2 Generalization to a 3-bit Adder

The Boolean function for error correction using the (3,1) Hamming code on the 3-bit
word [C1, C2, C3] is identical to computing the "CARRY" bit of the modulo-2 3-bit
addition of the bits C1, C2 and C3. If CARRY[C1, C2, C3] and SUM[C1, C2, C3] denote
the "CARRY" and "SUM" bits of the modulo-2 3-bit addition of C1, C2 and C3,

CARRY[C1, C 2, C3] = (C1 ∩ C 2) ∪ (C 2 ∩ C3) ∪ (C1 ∩ C3)

(6-1)

SUM[C1,C 2, C3] = C1 ⊕ C 2 ⊕ C3 ,

(6-2)

where "∪" denotes the Boolean "OR" function, "∩" denotes the Boolean "AND" function
and "⊕" denotes the Boolean function Exclusive-OR (XOR). The "SUM" bit can be
written in terms of the triple product Boolean functions using the identities,

( A ⊕ B) = ( A ∩ B ) ∪ ( A ∩ B) , ( A ∩ B) = A ∪ B , ( A ∪ B) = A ∩ B and ( A ∩ A ) = 0 .
Thus,

SUM = (C1∩C2∩C3) ∪ (C1∩ C2 ∩ C3 ) ∪ ( C1∩C2∩ C3 ) ∪ ( C1∩ C2 ∩C3).

(6-3)

In comparision, the "CARRY" bit is given by Equation (4-4) as

CARRY = (C1∩C2∩C3) ∪ ( C1∩C2∩C3) ∪ (C1∩ C2 ∩C3) ∪ (C1∩C2∩ C3 ).

(6-4)

98
The "CARRY" bit was implemented using the error correcting circuit, where the first
term on the right hand side was generated in the "non-correcting arm" and the remaining
three terms were generated in the "correcting arm" of the circuit. The "SUM" bit
operation can also be implemented by inverting the output of the "correcting arm" in the
circuit, i.e., using a "NOT" function in the "correcting arm" before combining it with the
"non-correcting arm". With PM fiber, the NOT function can be implemented with the use
of a 90 degree cross-splice between the principal axes of two PM fibers.
Figure 6.1 shows how the "SUM" and "CARRY" bits of the 3-bit addition can be
generated in the same set of SOAs. Thus it is possible to implement two different truth
tables related to the modulo-2 3-bit addition in the same circuit.

non-correcting arms

Polarizer
SOA
CARRY-bit
SOA

BPF

Input
Polarizer
SUM-bit

SOA
BPF
Wavelength selective
half-wave plate

cross-splice

correcting arm

Figure 6.1: Schematic of the full 3-bit adder (from Ref. [2])

6.3 Encoding the (7,4) Hamming Code

The (3,1) Hamming Code has a high overhead due to the high redundancy. This overhead
can be reduced by the use of more sophisticated error correcting codes [1] and the study
of these codes has been an area of active research. The (7,4) Hamming Code is the next
higher-level Hamming Code with a lower redundancy than the (3,1) Hamming Code. The
"SUM" bit, C1 ⊕ C2 ⊕ C3, also corresponds to the parity of the three bits C1, C2, and
C3 and can be used to generate parity bits for encoding of other Hamming Codes.
For example, the encoder for the (7,4) Hamming Code takes four input data bits
[D1, D2, D3, D4] and creates three additional parity bits given by

P-(412) = D4 ⊕ D1 ⊕ D2

(6-5a)

P-(423) = D4 ⊕ D2 ⊕ D3

(6-5b)

P-(413) = D4 ⊕ D1 ⊕ D3,

(6-5c)

i.e., the parity bits are ÒSUMÓ bits of the 3-bit additions of D4 with two additional bits
[Di, Dk] (i,k=1,2,3) from the remaining three bits (there is nothing special about D4, it is
just taken for this example). For spectrally placed channels [D1-D4], each ND-FWM of
D4 with [Di, Dk] will occur at a different wavelength channel as shown experimentally in
Figure 6.2 and given by

100
( 3)
Ek(ωP-(412)=ωD1+ωD2-ωD4) ∝ χ klmn
El(ωD1)Em(ωD2)En*(ωD4)

(6-6a)

( 3)
El(ωD2)Em(ωD3)En*(ωD4)
Ek(ωP-(423)=ωD2+ωD3-ωD4) ∝ χ klmn

(6-6b)

( 3)
El(ωD1)Em(ωD3)En*(ωD4).
Ek(ωP-(413)=ωD1+ωD3-ωD4) ∝ χ klmn

(6-6c)

D3
D4

-10

Intensity (dBm)

-20

-30

P-(423)
P-(413)

-40

P-(412)

-50

-60
1535

1540

1545

1550

1555

1560

Wavelength (nm)

Figure 6.2: Optical Spectrum at the output of the SOA (in 0.1 nm
Resolution Bandwidth) in the presence of four input waves (from Ref. [2])

In this case, the 3-bit adder circuit described above can be used as an encoder for
the (7,4) Hamming Code, which simultaneously generates the three parity bits using
different ND-FWM processes. Since D4 is common to all the additions, the pre-

101
processing element in one of the arms should act as a half-wave plate for D1, D2, and D3
and a full-wave plate for D4. The 7-bit word at the output of the encoder will be in a
byte-wide format with the data and the parity bits on separate wavelength channels.
Figure 6.2 shows the different FWM signals arising due to the presence of four
wavelength channels [D1-D4] and the ND-FWM signals generating the parity bits are
marked. Thus three independent logic functions can in principle be implemented in
parallel in one circuit. This is schematically shown in Figure 6.3.

D4
D1
D2
4-bit word
[D1-D4]

D4
D2

P-(412)
SUM

P-(423)

7-bit coded
word

SUM

D3
D4
D1
D3

P-(413)
SUM

Figure 6.3: Generating the parity bits for the (7,4) Hamming Code

6.4 Decoder Circuit for the (7,4) Hamming Code

In a similar fashion, one way to realize a decoder circuit for the (7,4) Hamming code is
by using the 3-bit adder circuits as building blocks and cascading them. The transmitted

102
code word contains the original four bits [D1-D4] and three parity bits [P-(412), P-(423),
P-(413)]. To make a distinction between transmitted and received data, the transmitted
bits are denoted by upper case while the received bits are denoted by lower case. Thus the
transmitted word is [D1, D2, D3, D4, P-(412), P-(423), P-(413)], while the received word
is [d1, d2, d3, d4, p-(412), p-(423), p-(413)]. For the sake of brevity, SUM[A1, A2, A3]
and CARRY[A1, A2, A3] is used to denote the SUM and the CARRY bits resulting from
the modulo-2 addition of the three bits A1, A2 and A3. Thus P-(412) = SUM[D4, D1,
D2].
Since D4 is present in all three parity bits, it is the first bit that is checked for
errors. The following additions are performed on the received bits,

d4-(12) = SUM[p-(412), d1, d2]

(6-7a)

d4-(23) = SUM[p-(423), d2, d3]

(6-7b)

d4-(13) = SUM[p-(413), d1, d3].

(6-7c)

In the absence of any errors (received bit equals transmitted bit), each of the above
additions would equal D4. For example, if d4-(12) is computed in the absence of any
errors, it equals

d4-(12) = SUM[p-(412), d1, d2] = SUM[P-(412), D1, D2]
= (D4 ⊕ D1 ⊕ D2 ⊕ D1 ⊕ D2) = D4,

(6-8)

103
Since CARRY[A1,A2,A3] equals the bit that occurs the largest number of times among
[A1-A3]. This can be used to find the correct transmitted bit D4 from the four bits [d4,
d4-(12), d4-(23), d4-(13)] using the following operations,

D4 = CARRY[CARRY[d4, d4-(12), d4-(23)], CARRY[d4, d4-(23), d4-(13)],
CARRY[d4, d4-(12), d4-(13)]].

(6-9)

This ensures that D4 is correctly generated for all possible cases of the received code
word including those with a single error on any bit. For example, if the error is on d2, i.e.
d2= D2 =D2 ⊕ 1,we obtain

d4-(12)=SUM[P-(412), D1, d2]=(D4 ⊕ D1 ⊕ D2 ⊕ D1 ⊕ D2 ⊕ 1)=D4 ⊕ 1= D4 (6-10a)
d4-(23)=SUM[P-(423), D2, D3]=(D4 ⊕ D2 ⊕ D3 ⊕ D2 ⊕ D3 ⊕ 1)=D4 ⊕ 1= D4 (6-10b)
d4-(13)=SUM[P-(413), D1, D3]= (D4 ⊕ D1 ⊕ D3 ⊕ D1 ⊕ D3) = D4.

(6-10c)

In this case, the right-hand side of Equation (6-9) equals

CARRY[CARRY[D4, D4 , D4 ], CARRY[D4, D4 , D4], CARRY[D4, D4 , D4]]
= CARRY[ D4 , D4, D4] = D4.

(6-11)

104
Similarly, if the error is on d4, i.e. d4= D4 =D4 ⊕ 1, the right-hand side of Equation (6-9)
equals

CARRY[CARRY[ D4 , D4, D4], CARRY[ D4 , D4, D4], CARRY[ D4 , D4, D4]]
= CARRY[D4, D4, D4] = D4,

(6-12)

and if the error is on any one of the parity-bits, say p-(412),i.e., p-(412)=P-(412) ⊕ 1, the
right hand side of Equation (6-9) equals

CARRY[CARRY[D4, D4 , D4], CARRY[D4, D4, D4], CARRY[D4, D4 , D4]]
= CARRY[D4, D4, D4] = D4.

Figure 6.4 shows the block-diagram of a circuit that generates D4.

(6-13)

105
D4 Generator

p-(412)
d1
d2

SUM

d4-(12)

p-(413)
d1
d3

SUM

p-(423)
d2
d3

SUM

CARRY

d4-(13)
d4

CARRY

d4-(23)
d4

CARRY

Figure 6.4: Generator circuit for D4

The other bits [D1-D3] occur symmetrically in the 7-bit word and can be found
using the following additions:

d1-(42) = SUM[p-(412), d2, D4]

(6-14a)

d1-(43) = SUM[p-(413), d3, D4]

(6-14b)

D1 = CARRY[d1, d1-(42), d1-(43)]

(6-14c)

106

d2-(41) = SUM[p-(412), d1, D4]

(6-15a)

d2-(43) = SUM[p-(423), d3, D4]

(6-15b)

D2 = CARRY[d2, d2-(41), d2-(43)]

(6-15c)

d3-(41) = SUM[p-(413), d1, D4]

(6-16a)

d3-(42) = SUM[p-(423), d2, D4]

(6-16b)

D3 = CARRY[d3, d3-(41), d3-(42)].

(6-16c)

It is easy to verify that the bits [D1-D3] are also generated correctly for all possible cases,
where the received 7-bit word has at the most one erroneous bit. Figure 6.5 shows the
block-diagrams of the circuits that generate the bits [D1-D3] once D4 has been generated.

107
D1 Generator

p-(412)
d2
D4

d1-(42)

SUM
d1

p-(413)
d3
D4

CARRY

d1-(43)

SUM

D2 Generator
p-(412)
d1
D4

d2-(41)

SUM
d2

p-(423)
d3
D4

CARRY

d2-(43)

SUM

D3 Generator
p-(413)
d1
D4

d3-(41)

SUM
d3

p-(423)
d2
D4

CARRY

d3-(42)

SUM

Figure 6.5: Generator circuits for bits [D1-D3]

108

6.5 Comments on Generalization

The experimental feasibility of cascading such circuits remains to be investigated. Issues
such as the strength of the nonlinearity present in the devices used will determine the
conversion efficiencies and Optical-Signal-to-Noise Ratios (OSNRs) of the FWM signals
involved. The degradation of the Extinction Ratio (ER) of the PolSK signals due to mode
conversion effects in these elements will affect the performance of these circuits and
should be minimized. The results will be especially important in the design and
implementation of more complex logic gates involving several Boolean operations, e.g.,
the decoding circuit for higher-level linear codes, such as the (7,4) Hamming Code.
The degradation of the ER of the PolSK bits can be modeled by using a Jones
Matrix approach for each component (or cascades of several components). The ER of
each component can be used to calculate an angular offset between the principal axes of
the output with respect to the input of the component. The effect of this offset on the ER
of the input PolSK bit can be calculated and the degraded ER of the PolSK bit can be
obtained at the output. The procedure is repeated to find the ER as the PolSK bits pass
through several components, each characterized by an ER.
Other Boolean functions can be realized by using cascaded FWM processes. One
example is the parity-bit generation of the five bits [C1, C2, C3, C4, C5], which can be
written as

P-(12345) = C1 ⊕ C2 ⊕ C3 ⊕ C4 ⊕ C5 = SUM[C1, C2, SUM[C3, C4, C5]].

(6-17)

109
Thus, the 3-bit adder circuit also generates the parity-bit P-(12345) using two cascaded

χ ( 3) processes (or a fifth order nonlinear process). One way to implement the 5-bit parity
generation is to choose the wavelengths for channels [C1-C5] such that the birefringent
element (pre-processing element) acts as a wavelength selective half-wave plate for two
channels, e.g., C1 and C4 and a full-wave plate for channels C2, C3, and C5. One of the
four different cascaded scattering processes for the 5-bit parity generation is shown in
Figure 6.6.

EP-(12345)

EP-(123)
EC5

EC4*
EC3

EC2

EC1*

Figure 6.6: Two cascaded FWM processes can be used to implement a 5bit parity generator

110
Depending on the strength of the nonlinearity in the device, the scheme can be
generalized to the 7-bit parity generation using a seventh order nonlinear process and so
on. In general, other types of building blocks for implementing other Forward ErrorCorrecting Codes (FECs) could be realized using different configurations and preprocessing elements. A comprehensive study of other types of logic elements is beyond
the scope of this thesis.

6.6 Conclusion

It has been shown that FWM on PolSK coded bits can be used to construct certain higherlevel logic elements without resorting to the standard 2-input gates. Taking the simple
example of the (3,1) Hamming code, on-the-fly error correction on severely distorted data
has been demonstrated. The data is recovered with a Bit Error Rate <10-9. To the best of
our knowledge, this is the first demonstration of a fiber optic logic circuit that performs
signal processing on more than two input channels simultaneously. The bit rate of the
experiment was limited to 2.5 Gbit/s by the bandwidth of the modulators. Since FWM is
an ultrafast nonlinearity, the error correcting circuit can be made to perform at much
higher bit rates. The scheme has been generalized, and it has been shown that a 3-bit
adder can be implemented in a single circuit. It has also been shown that several NDFWM processes can be used to perform different triple-product logic operations
simultaneously and this can simplify the design of the encoder circuit for the (7,4)
Hamming Code. Furthermore, the 3-bit adder circuit can be used as a building block to

111
implement more sophisticated Boolean functions, such as a decoder circuit for the (7,4)
Hamming Code, without resorting to the standard 2-input transistor based logic using
conventional electronic circuit theory. In passing, it should also be mentioned that the
circuits designed for the (7,4) Hamming Code using the 3-bit adders reflect the
symmetries of that error-correcting code. Finally, it should be noted that the error
detection and correction schemes have been merely taken as examples to demonstrate the
potential offered by using the polarization selection rules of ND-FWM processes to
implement ultrafast all-optical logic.

112

Bibliography

1. V. Pless, "Introduction to the Theory of Error-Correcting Codes," John Wiley &
Sons, New York, NY, 1982.

2. A. Bhardwaj, P.O. Hedekvist, and K. Vahala, "All-Optical Logic Circuits based on
Polarization Properties of Nondegenerate Four-Wave Mixing," J. Opt. Soc. Am. B,
18, 657-665 (2001).