Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Reconfigurable and flexible high-speed optical signal processing and spectrally shared optical subsystems
(USC Thesis Other)
Reconfigurable and flexible high-speed optical signal processing and spectrally shared optical subsystems
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Reconfigurable and Flexible High-Speed Optical Signal
Processing and Spectrally shared Optical Subsystems
by
Fatemeh Alishahi
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ELECTRICAL ENGINEERING)
December 2021
Copyright 2021 Fatemeh Alishahi
ii
Dedication
To my lovely parents, and my dear sister Marzieh
who always love and support me
unconditionally.
iii
Acknowledgments
I would like especially thank Professor Alan E. Willner who was a great advisor and teacher. I
would like to thank Prof. Willner because of his patient, supports, invaluable technical and
scholarly discussions and because of all his professional and priceless life-long lessons.
I would like to thank the dissertation committee Professor Stephan W. Haas, Professor Wei Wu
and Professor Jonathan L. Habif for their supports, invaluable feedback, insightful comments, and
contributions. I am also very thankful to Professor Todd Brun, Professor Wei Wu, Professor Keith
Jenkins, and Professor Stephan W. Haas for their guidance during my qualification exam. Also, I
am grateful to Dr. Youichi Akasaka for his involved discussions and advice. I would like to thank
the amazing and helpful staffs of the electrical engineering department at USC, Diane Demetras,
Gerrielyn Ramos, and Corine Wong.
My warmest thanks go to the many members of the Optical Communications Lab (OCLab) at
USC for all these years of insightful discussions and collaboration. I would like to Dr. Yinwen
Cao, Dr. Nisar Ahmed, Dr. Bishara Shamee, Dr. Yan Yan, Dr. Dr. Yongxiong Ren, Dr. Morteza
Ziyadi, Dr. Guodong Xie, Dr. Changjing Bao, Dr. Ahmed Almaiman, Dr. Peicheng Liao, Dr.
Amirhossein Mohajerin-Ariaei, Dr. Long Li, Dr. Zhe Zhao, Dr. Cong Liu and Dr. Ahmad
Fallahpour for valuable discussions and their help. I also want to thank Amir Minoofar, Kaiheng
Zou, Kai Pang, Haoqian Song, Zhe Wang, Runzhou Zhang, Karapet Manukyan, Huibin Zou and
Hao Song for all their valuable help.
I have been blessed with great friends Haleh Akrami, Sepideh Azarnoosh, Kaveh Rezaei
Moghadam, Emi Tomoyose, Amir Mohajerin-Ariaei, Shiva Navabi and Arash Fayazi who helped
me and supported me during my PhD at USC.
iv
Finally, I would love to express my deepest gratitude to my parents and my sister whom without
their love, support, encouragement, and patience I could not reach at this stage.
v
Table of Contents
Dedication ................................................................................................................... ii
Acknowledgments ..................................................................................................... iii
List of Figures ........................................................................................................... vii
Abstract xiii
Chapter 1 Introduction on Signal Processing in Optical Domain ......................... 1
1.1 Nonlinear Optical Processes ........................................................................... 1
1.1.1 Degenrate Four-Wave Mixing in χ
(3)
Materials ................................... 2
1.1.2 Three Wave Mixing in χ
(2)
Materials ................................................... 3
1.1.3 Materials and Devices .......................................................................... 4
1.2 Basic Enabling Operations for Optical Signal Processing ............................. 4
1.2.1 Wavelength Conversion ....................................................................... 4
1.2.2 Wavelength Multicasting using cSFG-DFG ........................................ 5
1.2.3 Optical Multiplexing using cSFG-DFG ............................................... 7
1.2.4 Optical Delays using Conversion-Dispersion ...................................... 8
Chapter 2 Tunable All-Optical WDM Channel Selection using Optical
Parametric Amplification ...................................................................... 10
2.1 Introduction .................................................................................................. 10
2.2 Concept ......................................................................................................... 11
2.3 Experimental Setup ...................................................................................... 12
2.4 Results .......................................................................................................... 12
2.7 Conclusion .................................................................................................... 15
Chapter 3 Optical Buffer based on Discrete Time Delays based on
a Fiber Loop with an Internal Frequency Shifter ............................... 16
3.1 Introduction .................................................................................................. 16
3.2 Concept ......................................................................................................... 16
3.3 Experimental Setup ...................................................................................... 17
3.4 Results .......................................................................................................... 18
3.4 Conclusion .................................................................................................... 21
Chapter 4 Optical Mitigation of Interchannel Interference (ICI) for
Multiple Spectrally Overlapped WDM Channels using
Nonlinear Wave Mixing ........................................................................ 22
4.1 Introduction .................................................................................................. 22
4.2 Concept ......................................................................................................... 23
4.3 Experimental Setup ...................................................................................... 28
4.4 Experimental Results .................................................................................... 31
vi
4.5 Conclusion .................................................................................................... 34
Chapter 5 Optical Generation of Nyquist WDM/TDM Channels with
Sinc-shaped Temporal Pulse Trains using Microresonator-
based Kerr Frequency Comb ................................................................ 36
5.1 Introduction .................................................................................................. 36
5.2 Concept of optical generation of WDM Nyquist channels .......................... 38
5.5 Experimental Setup ...................................................................................... 39
5.3 Experimental Results .................................................................................... 41
5.4 TDM of Nyquist Sinc-Shaped Channels ...................................................... 46
5.5 Concept of TDM of Nyquist Sinc-Shaped Channels ................................... 47
5.6 Experimental setup for TDM of Nyquist Channels ...................................... 49
5.7 Experimental results for TDM of Nyquist Channels .................................... 51
5.8 Conclusion .................................................................................................... 53
Chapter 6 Remotely Controlled and Powered Tunable Optical 2-4
Taps Correlator of a 20-100 Gbit/s QPSK Channel Based
on Cascaded MZIs and Laser-Delivered Control Signals .................. 55
6.1 Introduction .................................................................................................. 55
6.2 Theoretical study of cascaded MZIs as a tapped delay line ......................... 58
6.3 Concept of remote correlator based on cascaded MZIs ............................... 67
6.4 Experimental Setup and Results ................................................................... 68
6.5 Conclusion .................................................................................................... 73
Chapter 7 Remotely Controlled and Monitored Tunable Nonlinear
Optical Correlator Based on Temperature-controlled
Nonlinear Wave Mixing ........................................................................ 75
7.1 Introduction .................................................................................................. 75
7.2 Concept ......................................................................................................... 76
7.3 Experimental Setup ...................................................................................... 78
7.3 Experimental Results .................................................................................... 79
7.3 Conclusion .................................................................................................... 83
References ................................................................................................................. 84
vii
List of Figures
Figure 1.1 Degenerate four-wave mixing (FWM) schemes for
generation of phase conjugate signal copy. ZDW: zero dispersion
wavelength. ................................................................................................. 3
Figure 1.2 (a) Cascaded sum and difference frequency generations
(cSFG-DFG) and (b) Second harmonic generation and DFG
(cSHF-DFG) for wavelength conversion in a PPLN device. QPM:
quasi-phase matching. ................................................................................. 3
Figure 1.3 (a) Wavelength conversion in a PPLN waveguide, (b) pump
configurations, (c) amplitude and phase of the generated signals
in the cSFG-DFG processes [18]. ............................................................... 5
Figure 1.4 Various configurations for N-fold signal multicasting using
multi-pumps. ............................................................................................... 6
Figure 1.5 Illustration of signal multicasting of a signal onto multiple
frequencies using coherent frequency comb [18]. ...................................... 7
Figure 1.6 Optical coherent multiplexing of coherent signals using
optical frequency combs [18]. ..................................................................... 8
Figure 1.7 Tunable Conversion/dispersion based optical delay. .......................... 9
Figure 2.1 Conceptual Schematics of the WDM channel filtering based
on the phase sensitive parametric amplification. (b): Concept of
nonlinear interactions in a PS-FOPA interactions. ................................... 11
Figure 2.2 Experimental setup for tunable all-optical filter using PS-
FOPA and the spectra at different location of the filter. ........................... 12
Figure 2.3 The constellation diagrams of the input and output of phase
quantizer for different levels of phase noise. ............................................ 13
Figure 2.4 Output power spectra for the selection of middle channel
(a), Left channel (b), right channel. System performance for the
selection of right channel (a), Left channel (b), middle channel. ............. 14
Figure 3.1 The concept of the optical loop buffer. ............................................. 17
viii
Figure 3.2 Experimental setup. PC: Polarization Controller, PD:
Photodetector, MZM: Mach-Zehnder Modulator. .................................... 18
Figure 3.3 (a): Spectrum before LCoS. (b): Spectrum after LCoS.
Experimental setup. ................................................................................... 19
Figure 3.4 (a): The magnitude of S21 for 1
st
copy (blue), 2
nd
copy (red)
and open loop operation (dashed red). (b): Phase of S21 for 1
st
copy (blue), 2
nd
copy (dashed red). (c): Third-order distortion of
the system for 2
nd
copy. ............................................................................. 19
Figure 3.5 Delays for first, second and third copies. .......................................... 20
Figure 3.6 Constellation diagrams and BER curves for the different
copies of a QPSK data at the output. ........................................................ 21
Figure 4.1 Conceptual block diagram of optical interchannel crosstalk
mitigation method. The scheme consists of three main blocks: (i)
Conjugate copies generation in PPLN-1 waveguide, (ii)
Wavelength conversion in PPLN-2 which coherently mixes
original channels with amplitude/phase adjusted conjugate
copies (iii) Wavelength conversion in PPLN-3 which coherently
mixes original channels with amplitude/phase and delay adjusted
conjugate copies. ....................................................................................... 24
Figure 4.2 (a) Experimental setup. PPLN: periodically poled lithium
niobate, PC: polarization controller, LCoS: liquid crystal on
silicon. (b) Optical spectrum of generated signal conjugate copies
for 20 Gbaud overlapped QPSK channels at point A and optical
spectrum of ICI mitigated odd channels 1, 3, 5, and 7 after the
last PPLN waveguide at point B. .............................................................. 30
Figure 4.3 Experimentally recorded signal constellation diagrams of
channels 1, 3, and 6 with (w.) and without (w/o) optical ICI
mitigation method for 20 Gbaud overlapped QPSK signals and at
channel spacings of 17.5 GHz, 20 GHz, and 25 GHz. .............................. 32
Figure 4.4 BER measurements with (w.) and without (w/o) optical ICI
compensation method for QPSK overlapped channels and for ∆f
=17.5 GHz ................................................................................................. 32
Figure 4.5 Experimentally recorded signal constellation diagrams of
channels 1, 3, and 6 with (w.) and without (w/o) optical ICI
ix
mitigation method for 20 Gbaud overlapped 16-QAM signals and
at different channel spacings of 17.5 GHz and 20 GHz. ........................... 33
Figure 4.6 BER measurements with (w.) and without (w/o) optical ICI
compensation method for 20 Gbaud 16-QAM overlapped
channels and for ∆f =17.5 GHz. ................................................................ 34
Figure 5.1 Concept of WDM Nyquist channels with sinc-shaped pulse
trains using a Kerr frequency comb. ....................................................... 39
Figure 5.2 Experimental setup; PC: Polarization Controller, PS: Phase
Shifter; Att: Attenuator; EDFA: Erbium-Doped Fiber Amplifier;
OSA: Optical Spectrum Analyzer; OSC: Oscilloscope; BERT:
Bit Error Rate Tester. Measured spectra at points (1), (2) and (3)
of the setup are shown. .............................................................................. 41
Figure 5.3 Measured sinc-shaped pulse trains at the outputs of (a): the
first MZM driven by f1=9 GHz and (b): the second MZM driven
by f2=3 GHz recorded by a sampling scope. ............................................. 43
Figure 5.4 (a): Four Kerr comb lines (FSR=71.7 GHz). (b): Four
corresponding sub-combs. (c-f): Sinc pulses observed in OSC
(black line) and calculated by Eq. (5.1) (dashed red line). ....................... 43
Figure 5.5 (a): Experimental output spectrum of nine Nyquist WDM
channels using a Kerr comb (FSR=192.0020 GHz). (b):
Measured sinc-shaped pulse trains of different channels. ......................... 44
Figure 5.6 The BER for nine 6 Gbit/s OOK-modulated WDM channels,
with sinc-shaped pulse trains as carriers. .................................................. 46
Figure 5.7 Concept of TDM of Nyquist channels via Kerr comb and a
PPLN waveguide. PPLN: Periodically-Poled-Lithium-Niobate,
SFG: Sum Frequency Generation, DFG: Difference Frequency
Generation, TDM: Time Division Multiplexing, MZM: Mach-
Zehnder Modulator, RF: Radio Frequency ............................................... 49
Figure 5.8 . Experimental setup; PC: Polarization controller, PS: Phase
shifter. Spectrum at points (a), (b) and (c) are shown. AFG:
arbitrary function generator; ECDL: external cavity diode laser;
ASE: amplified spontaneous emission; Att: attenuator; BPF:
bandpass filter; PC: polarization controller; EDFA: erbium-
doped fiber amplifier; TOF: tunable optical filter; LF: lensed
x
fiber; FBG: fibber Bragg grating; Res.: resolution; OSA: optical
spectrum analyser. ..................................................................................... 50
Figure 5.9 Spectrum at the output of PPLN. Three OOK-modulated
channels are located at 1551.1 nm. 1552.5 nm and 1553.9 nm and
the converted channel is located at 1542.2 nm. ........................................ 52
Figure 5.10 (a): Eye diagrams of the three OOK-modulated channels.
(b) Bit error rate versus optical signal to noise ratio of each OOK
modulated channels. .................................................................................. 53
Figure 6.1 Concept of cascaded MZI with 3 MZI acting as a 4-tap TDL.
Each MZI includes a waveguide acting as a delay line with delay
t, a phase-shifter with phase-shift f and an attenuator with
attenuation h. The MZIs are connected to each other by shared
directional couplers with coupling coefficient k. Under non-ideal
conditions, the phase-shifters can drift over have a limited phase
coverage. MZI: Mach-Zehnder interferometer, PS: phase-shifter.
................................................................................................................... 60
Figure 6.2 Transfer function (b) extinction ratio (c) passband ripple and
(d) phase linearity for different phase shifter bias drift when only
MZI-1 (green), only MZI-3 (yellow) and all MZIs (red) have bias
drifts. ......................................................................................................... 62
Figure 6.3 (a)Transfer function (b) extinction ratio (c) passband ripple
and (d) phase linearity when only MZI-1 (green), only MZI-3
(yellow) and all MZIs (red) have phase shifters with limited
phase coverages. ........................................................................................ 63
Figure 6.4 (a) Passband ripple (blue) and phase linearity (red) for
different waveguide losses when only MZI-1 (solid), only MZI-
3 (dashed) and all MZIs (dotted) are lossy. (b) EVM for different
waveguide losses when only MZI-1 (grey), only MZI-3 (cyan)
and all MZIs (red) are lossy. (c) Passband ripple (blue) and phase
linearity(red) for different waveguide refractive index errors. (d)
EVM for different waveguide refractive index errors when only
MZI-1 (grey), only MZI-3 (cyan) and all MZIs (red) have
refractive index error. ................................................................................ 65
Figure 6.5 (a) Fundamental harmonic versus input power for an
incoming microwave signal for when all MZIs are idea (green
circles), when MZI-1 has an attenuation of 10 dB/m (dashed
xi
green), when MZI-1 has bias drift of 300 mV (dotted-dashed light
green). Third harmonic power versus input power for when all
MZIs are idea (red circles), when MZI-1 has an attenuation of 10
dB/m (dashed red), when MZI-1 has bias drift of 300 mV (dotted-
dashed light red). (b) SFDR for different scenarios. ................................. 67
Figure 6.6 (a) Concept of a remotely controlled tunable optical
correlator (b) Basics of the 2 and 4-taps QPSK correlator using
single MZI and two cascaded MZIs. ............................................................ 68
Figure 6.7 Experimental setup and power spectra at points A and B. ................ 69
Figure 6.8 (a) PD array voltage/current vs optical power (b)
Backscattered power suppression measured at the input of SMF
vs phase modulator RF drive. Backscattered power spectrum
with and without phase modulation for input link power of 17
dBm (c), 20.5 dBm (c1) and 26.5 dBm (c2). ............................................ 70
Figure 6.9 (a) Measured Brillouin backscattered power vs link input
power (b) Measured link output power vs link input power(c)
Phase-shift of the MZI vs link output power. ........................................... 72
Figure 6.10 Input and output constellation diagrams and target patterns
for the 2-Tap (a) and 4-Tap (b) correlator and different baud-
rates. .......................................................................................................... 73
Figure 7.1 (a) Concept of a QPSK correlator using the nonlinear wave
mixing (a). Concept of a remotely/controlled and monitored
nonlinear QPSK modulator (b). Correlated signal generation
inside a PPLN for an adjusted QPM and for the blue/red shift of
QPM wavelength (c). TEC: Thermo-electric controller. ........................... 77
Figure 7.2 Experimental setup. MLL: Mode Lock Laser , DLI: Delay
Line Interferometer, HNLF: Highly Nonlinear Fiber, BPF: Band
Pass Filter, PC: Polarization Controller, EDFA: Erbium Doped
Fiber Amplifier, DCF: Dispersion Compensating Fiber, PPLN: :
Periodically Poled Lithium Niobate. ......................................................... 79
Figure 7.3 Rayleigh and Brillouin scattering from the link vs power
delivered by the pump and comb lines. ..................................................... 80
Figure 7.4 Output spectra of the PPLN for different temperature
settings. ..................................................................................................... 81
xii
Figure 7.5 Power delivered to the node and power gain after monitoring
and tuning the pump power vs temperature drift for a QPSK at
10 Gbaud (a) and 15 Gbaud (b). ............................................................... 82
Figure 7.6 Monitored powers vs temperature drift before and after
pump power amplification for a QPSK at 15 GBaud (a) and 10
GBaud (b). ................................................................................................. 82
Figure 7.7 EVM vs temperature drift before and after pump power
amplification for a QPSK at 15 GBaud (a) and 10 GBaud (b).
Constellation diagrams of the correlated signals (c). ................................ 83
xiii
Abstract
Technology has enabled people in all walks of life to generate, store, and communicate
enormous amounts of data. Recent technological advances in high-speed backbone data networks,
together with the growing trend of bandwidth demanding applications have created a need for
higher capacities in signal transmission and signal processing. The bandwidth-hungry applications
such as cloud computing, photos and video sharing, data storage systems and recent technological
advances in high-speed data networks have created a demand for higher speeds in data processing
and transmission. Lately, system capacity increases have been realized by coherent technologies
and advanced modulation formats [1–10].
High-data-rate all-optical signal processing has been one of the main research goals in
photonics. Signal processing using nonlinear optics has been of great interest due to its inherent
ultrafast THz bandwidth and its phase-preserving nature [1,11-14]. A key feature of optical signal
processing is that optical techniques do not need to “touch” or switch every individual “bit,” as
electronic transistors do. That is; optical approach may operate at the line rate of the data and the
latency o the process can be very small. Optical amplifiers, for instance, can amplify Tb/s signals
without touching the signal at the bit level. Another example is basic wavelength conversion using
a laser pump and a nonlinear device, where the data information can be transferred from one carrier
wavelength to another at a very fast speed (nonlinearities have femtosecond response times) as
optical signals fly through the device [3, 4]. Importantly, advances in materials and devices which
have resulted in devices with higher nonlinearities and higher efficiencies, and photonic integrated
circuits (PICs) technologies, are the key for any practical utilization of optical signal processing in
xiv
the future [15-17]. Additionally, if signals are already in the optical domain, it might be beneficial
to avoid inefficient optical-to-electrical-to-optical (OEO) conversion by doing optical signal
processing. Also, multiple signal processing functions can be achieved for multiple channels
within the same optical device such that multiple electronic devices can be replaced with a single
optical device. And finally, it is possible to exploit and simultaneously manipulate multiple
dimensions of the optical wave, e.g., amplitude, phase, wavelength, polarization, and space, to
provide more degrees of freedom.
This Ph.D. thesis explores the potential of ultra-highspeed and reconfigurable optical sub-
systems to function as different processing units through of a huge amount of data. The optical
signal processing units being explored here has the advantage that they can be employed in flexible
networks which hosts channels with shared spectral regions. This dissertation demonstrates optical
systems that avoid redundant OEO conversion and can control the processing units locally either
at the transmitter site or at a remote location from it throughout the optical network. Therefore,
subsystems introduced throughout this thesis can support in-line signal processing for high baud
rate signal. By utilizing various forms of photonic nonlinear interactions different functions
including wavelength selection, optical buffering, optical Nyquist WDM and TDM channel
generation and transmission, optical inter-channel-interference mitigation for multiple spectrally
overlapped channels and remote pattern recognition of digital data using different optical
subsystems are demonstrated.
1
Chapter 1 Introduction on Signal Processing in Optical
Domain
High speed optical processing is now an essential part of almost all fields of study. Also,
enormous bandwidth of fiber optic has led to the increase in the data traffic. On the other hand,
optical data processing can leverage electronic signal processing by ability to process Tbytes of
data and rapidly identify key patterns and features. This is while the electronic processing is limited
to Gigabytes features extraction with lower speed.
This chapter introduces the basic building blocks and nonlinear processes which become the
building blocks for optical signal processing sub-systems introduced and implemented in the
upcoming chapters. We will particularly discuss four-wave mixing and three wave mixing
interactions inside nonlinear materials. Also, the implementation of tunable delays in optical
systems is introduced. As an immediate implication of the abovementioned topics, we will also
introduce optical multicasting and coherent addition.
1.1 Nonlinear Optical Processes
Nonlinear processes in materials with either second or third order nonlinearity have ultra-fast
sub-ps response times. Hence they can be employed for implementation in various wave mixing
and coherent manipulations of signals with very high speed and ultra-wide bandwidth [1,18]. In
general, “optical wave mixing” is a process of nonlinear interaction of different optical waves with
the same or different wavelengths which results in the generation of an optical wave at new
wavelength [1]. The conservation efficiency of the generated wave is related to the phase matching
conditions [1,18,19,20]. In this chapter we will review both second order, χ
(2)
and third order, χ
(3)
2
nonlinear processes. One major implication of the nonlinear interactions arising because of the
third order nonlinearity is the interaction between three waves which is known as four-wave
mixing (FWM) [18,19]. On the other hand, the nonlinear elements possessing the χ
(2)
nonlinearity
can host the mixing between two waves which can result in variety of interactions such as second
harmonic generation (SHG), sum frequency generation (SFG), difference frequency generation
(DFG). Moreover the abovementioned second-order nonlinear interactions can happen
simultaneously which yields combinations of nonlinear interactions such as SFG-DFG and SHG-
DFG [1,18,21].
1.1.1 Degenrate Four-Wave Mixing in χ
(3)
Materials
Third order nonlinear meterial can host the Four-wave mixing (FWM) interaction. In this
process, three input waves mix under the phase-matching conditions in a χ
(3)
nonlinear medium to
generate a fourth wave at a new wavelength. Two versions of such interaction, known as
degenerate and non-degenerate FWM are depicted in Fig. 1.1 [1].
In the degenerate FWM, a continuous wave pump at frequency fpump and a data signal with
frequency fsignal are interacting inside a χ
(3)
nonlinear material, like a highly nonlinear fiber
(HNLF) [1]. If the pump is located around the zero-dispersion-wavelength (ZDW) of the nonlinear
device, the phase matching conditions are then satisfied, and the frequency of the newly created
wave follows:
𝑓
!"#$
= 2𝑓
%&'%
−𝑓
()*#+,
(1.1)
3
Figure 1.1 Degenerate four-wave mixing (FWM) schemes for generation of phase conjugate signal copy.
ZDW: zero dispersion wavelength.
1.1.2 Three Wave Mixing in χ
(2)
Materials
Inside a χ
(2)
nonlinear material and under phase-matching conditions two waves at frequencies,
f1 and f2 can mix to generate a new signal at the sum frequency fSFG= f1 + f2 and difference
frequency fDFG= f1 - f2 ,. Also, a single wave at frequency f1 can interact with itself to produce the
second harmonic term with fSFG= 2f1 [18].
In practice, it is desirable to keep the generated signals in the same frequency band as the input
signals. To do so a cascade of SFG and DFG (cSFG-DFG) or SHG and DFG (cSHG-DFG)
processes [18] are used. These interactions are schematically depicted in Figures 1.2 (a) and (b).
Figure 1.2 (a) Cascaded sum and difference frequency generations (cSFG-DFG) and (b) Second harmonic
generation and DFG (cSHF-DFG) for wavelength conversion in a PPLN device. QPM: quasi-phase matching.
sig
f
conv
f
ZDW
pump
f
*
Degenerate FWM
*
*
Non-Degenerate FWM
ZDW
c
(2)
SFG
f
dummy
f
pump
f
SFG
QPM
c
(2)
DFG
f
signal
f
idler
Cascaded SFG and DFG
QPM: Quasi-phase matching
c
(2)
SHG
f
pump f
SFG
QPM
c
(2)
DFG
f
signal
f
idler
*
Cascaded SHG and DFG
4
1.1.3 Materials and Devices
In this thesis, highly nonlinear fibers (HNLF) are used for the degenerate FWM nonlinear
process. A particular variant of the degenerate FWM, namely, parametric amplification is
discussed in the next chapter. Optical fibers with dramatically small effective cross section can be
used as efficient mixing devices. Also, the dispersion profile of optical fibers can be engineered to
exhibit a flat dispersion [18]. However, one of the draw-backs of nonlinear wave mixing process
for optical signal processing tasks in χ
(3)
optical devices is that multiple interfering and redundant
mixing terms is generated because of the wide bandwidth of the nonlinear interaction [18].
For the χ
(2)
nonlinear elements, periodically poled lithium niobate (PPLN) devices can be used.
PPLNs can provide relatively low propagation loss thanks to their centimeter-long length of the
waveguiding structure. However, it should be noted that for the nonlinear interactions to take place
effectively inside the PPLN, all the interacting waves have to be placed symmetrically around the
quasi-phase matching (QPM) wavelength of the PPLN device [1].
1.2 Basic Enabling Operations for Optical Signal Processing
In this section, basic operations as fundamental blocks for optical signal processing system will
be reviewed which will frequently be used in the following chapters.
1.2.1 Wavelength Conversion
Wavelength conversion is a useful phenomenon which is abundantly used in optical signal
processing [22-27]. It can be achieved via FWM in a HNLF or silicon waveguide or through a
cascade of sum frequency generation followed by difference frequency generation (cSFG-DFG)
in a PPLN waveguide. As shown in Fig. 1.3, in the cSFG-DFG wavelength conversion process,
5
the input signal (at 𝑓
()*#+,
) and a continuous wave pump laser (at 𝑓
%&'%
) are symmetrically placed
around the quasi-phase matching (QPM) frequency of the PPLN waveguide [18]. The latter is a
characteristic of PPLN structure. As a result, the phase matching condition for SFG is satisfied and
the signal will be first copied to the frequency 𝑓
()*#+,
+𝑓
%&'%
= 2𝑓
-./
[18]. This new SFG signal
can also be used in another nonlinear process to generate a new signal copy at a frequency close
to the original frequency. To achieve this, a dummy laser at wavelength fdummy can also be sent into
the PPLN waveguide where it can interact with the SFG signal and produce a copy at 𝑓
!0123412
=
𝑓
()*#+,
+𝑓
%&'%
–𝑓
5&''6
through the DFG process [18]. According to phase matching condition
the electrical field of the signal copy is proportional to 𝐸
()*#+,
𝐸
%&'%
𝐸
5&''6
∗
, in which (.)
∗
denotes the complex conjugate [18].
Figure 1.3 (a) Wavelength conversion in a PPLN waveguide, (b) pump configurations, (c) amplitude and phase of
the generated signals in the cSFG-DFG processes [18].
1.2.2 Wavelength Multicasting using cSFG-DFG
Optical wavelength multicasting similar to wavelength conversion could be realized by using
nonlinearities to create multiple copies of the input data signal at different wavelengths [28,29].
c
(2)
SFG
f
signal
f
dummy
f
SFG
QPM
c
(2)
DFG
f
cSFG-DFG
f
pump
…
Signal
Copy
Original
Signal
Electrical Fields
(Phase Matching)
Frequencies
(Energy Conservation)
Comb-Based Lasers
Phase Noise
(φ
SFG
= φ
signal
= φ
pump
≡ φ
comb
)
Sum Frequency
Generation (SFG)
E
SFG
~ E
signal
E
pump
φ
SFG
~ φ
signal
+ φ
pump
f
SFG
= f
signal
+ f
pump
φ
SFG
~ 2φ
comb
Difference Frequency
Generation (DFG)
E
DFG
~ E
SFG
(E
dummy
)*
φ
DFG
~ φ
SFG
– φ
dummy
f
DFG
= f
SFG
– f
dummy
φ
DFG
~ φ
comb
Cascaded SFG-DFG
E
cSFG-DFG
~ E
signal
E
pump
(E
dummy
)*
φ
cSFG-DFG
~ φ
signal
+ φ
pump
– φ
dummy
f
cSFG-DFG
= f
signal
+ f
pump
– f
dummy
φ
cSFG-DFG
~ φ
comb
Dummy
Signal
Signal Copy
PPLN Waveguide
Pump
(a) (b)
(c)
6
There are different methods based on different materials, nonlinear processes, numbers of pumps
and pump configurations for multicasting [1]. Figure 1.7 shows conceptual spectra for two
multicasting techniques [1,30]. In Figure 1.7, to generate N signal copies, the input data signal is
sent to a nonlinear device with N discrete pumps to create N copies of the input signal in a
degenerate FWM process. In the literature there are many examples of optical multicasting. For
example, the results of multicasting 16-QAM signals in a PPLN waveguide were reported in [29].
Moreover, similar to the wavelength conversion, various configurations can provide either phase
conjugated or non-phase conjugated copies.
Figure 1.4 Various configurations for N-fold signal multicasting using multi-pumps.
In another technique, signal multicasting can be realized by utilizing a frequency comb. The
input signal will be copied onto multiple wavelengths in a process known as wavelength
multicasting. Wavelength multicasting can be achieved using FWM in a HNLF or the cascaded
χ
(2)
::χ
(2)
processes of sum frequency generation followed by difference frequency generation in a
PPLN waveguide. In the cSFG-DFG process, the input signal (at fs) and a continuous wave (CW)
pump at fP1 are symmetrically placed around the quasi-phase matching (QPM) frequency of the
PPLN waveguide. As a result, the phase matching condition for SFG is satisfied and the signal
Requires N probe pumps for N-
fold multicasting
f
Signal
Requires (N+1)/2 pumps for N-
fold multicasting
Signal
f
Multi-Pump multicasting configurations
7
will be first copied to the frequency fSIG+fP1=2fQPM. The new signal can also be used in a nonlinear
process to generate multiple new copies at wavelengths close to the original frequency. To achieve
this, an optical frequency combs (frequency lines at fDi’s) can also be sent into the PPLN
waveguide where each of them will interact with the SFG signal and produce multiple copies at
fCi=fSIG+fP1–fDi through the DFG process. Therefore, each copy (fCi) and its corresponding comb
finger (fDi) will also be symmetric around the QPM frequency. Thus, a set of signal replicas can
be obtained. The power of each copy is proportional to the power of its corresponding comb finger.
Moreover, the phase of each copy is the summation of its corresponding comb finger and original
signal. Since all comb fingers are coherent, all signal replicas are also phase locked [18].
Figure 1.5 Illustration of signal multicasting of a signal onto multiple frequencies using coherent frequency comb
[18].
1.2.3 Optical Multiplexing using cSFG-DFG
There have been different techniques for optical multiplexing of WDM data channels into a
single channel. These include but not limited to the use of XPM in HNLFs [31], FWM in HNLFs
and waveguides [22,32], and cSFG-DFG in PPLN devices [33].
In the coherent multiplexing, optical frequency comb as well as the same cSFG-DFG process
Optical
Frequency
Comb
Nonlinear
Medium
Amplitude
Programmable
Filter
Input
Signal
Coherent
Copies
α
1
α
2
α
3
l
f
f
C3
f
C2
f
C1
f
D1
f
D2
f
D3
f
s
QPM
f
8
that was used to multicast the signal to multiple copies can be used to multiplex multiple coherent
signals to a single signal. The phase locked signal and their comb fingers are sent into a PPLN
waveguide along with a CW pump laser (at fP2≈fP1). Therefore, every signal mixes with its
corresponding phase-manipulated comb fingers through the SFG process and gives rise to a new
signal at frequency fCi+fDi=2fQPM. The multiplexed signal at 2fQPM will then interact with the added
pump laser at fP2 via the DFG process and create a multiplexed signal at original frequency fSIG=
2fQPM-fP2. The coherency of original signals as well as comb fingers is necessary to realize coherent
addition [18].
Figure 1.6 Optical coherent multiplexing of coherent signals using optical frequency combs [18].
1.2.4 Optical Delays using Conversion-Dispersion
The wavelength-dependent speed of light in a dispersive medium makes different spectral
channel to be delayed differently and as a result chromatic-dispersion-based delays to be obtained.
The essential of the optical delay based on a combination of wavelength conversion which was
described in the previous section and optical delays based on a dispersive element is shown in Fig.
1.7 [1,18].
Optical
Frequency
Comb
Nonlinear
Medium
Amplitude
Programmable
Filter
Coherent
Multiplexed
signal
α
1
α
2
α
3
l
f
f
C3
f
C2
f
C1
f
D1
f
D2
f
D3
f
s
QPM
f f
C3
f
C2
f
C1
BPF
f
f
s
Coherent
signals
9
Figure 1.7 Tunable Conversion/dispersion based optical delay.
The technique shown in Fig. 1.7 consists of the two main blocks of (1) wavelength conversion
and (2) dispersion induced delays. In the wavelength conversion, the original signal is wavelength
converted using optical nonlinear effects, such as FWM in an HNLF or cSFG-DFG in a PPLN
waveguide as were explained in the previous section. In the Dispersion Induced Delay block, the
wavelength converted copy of the signal is sent into a highly dispersive element. Each wavelength
will be relatively copied as the result. Another nonlinear element can be used to wavelength-
convert the signal again and transfer it to its original wavelength. However it is very important to
keep the dispersion of the element low to avoid any distortion caused by the intra-channel
dispersion [1,18].
Dispersive
Element
t
Δτ
Δτ
Wavelength
Conversion #2
λ
in
λ
1
λ
2
λ
1
λ
2 λ
in
λ
in
t
t t
Wavelength
Conversion #1
Relative Group Delay (s)
λ
1
λ
2
Δλ
Δτ
Wavelength dependent delay via chromatic dispersion Δτ = D Δλ
10
Chapter 2 Tunable All-Optical WDM Channel Selection
using Optical Parametric Amplification
2.1 Introduction
Tunable optical filtering plays an essential part in optical links due to the ever-increasing
demand for channel selection of wavelength division multiplexing (WDM) of multi-level
amplitude and phase encoded data [37]. An optimum tunable optical filter strives to keep certain
potential advantages such as wavelength tunability, bandwidth tunability, high extinction ratio and
favorably optical gain. However, though the filtering takes place in optical domain, most of the
available technology for tuning relies on mechanical rather optical techniques [38]. Aside from the
fact that mechanical tuning is prone to noise, it also dramatically decreases the speed of filter
reconfigurability. Also, as a passive element, optical filters are lossy elements and unable to
provide gain.
The ability to combine all the potential advantages of optical tunable filter using nonlinear
interactions is intriguing enough, since the nonlinearity has been known to have low additive noise
and potential higher speeds [39]. Hence, utilizing nonlinear interactions, high speed reconfigurable
optical filters are promising. Phase-sensitive fiber optic parametric amplifiers (PS-FOPA) with
their unique phase-sensitive amplification ability and ultralow noise figure, are among the most
efficient nonlinear interactions which can potentially provide high speed reconfigurability [40].
To date, there are quite few works on PS-FOPA performance as wavelength selective elements
[41-42]. Here, we experimentally demonstrate PS-FOPA in a special configuration to realize
tunable optical filters. The filter is not only wavelength tunable but is also bandwidth using the
11
pump wavelength and pump phase as the optical knob. Using PS-FOPA amplification, the filter
can also provide optical amplification for the selected channels. The extinction ratio between 13dB
to16 dB has been achieved and the system shows the net gain of around 6dB and almost zero
OSNR penalty.
2.2 Concept
The conceptual schematic in Fig. 2.1(a) describes the fundamentals of our PSA based filtering.
Multiple channels of WDM are at the input. At the output either of WDM channels of input data
stream can be selected by amplifying that channel while de-amplifying the others. The selection
process takes place all-optically by tuning the wavelength and the phase of the pump of a PS-
FOPA. The conceptual diagram in Fig. 2.1(b) shows the interaction inside a cascaded PS-FOPA
which calls for phase tunability. PS-FOPAs are based on the Kerr-induced four-wave mixing
(FWM) process occurring between one or two pump wave(s), a signal and an idler wave. The
phase mismatch between pump(s), signal and idler dictates the efficiency and direction of the
energy transfer between pump(s) and signal/idler. In the cascaded structure shown in Fig. 2.1(b)
the phase mismatch can be controlled by the wavelength detuning of pump and signal/idler due to
the strong dispersion of the connecting SMF between two highly nonlinear fibers (HNLF).
Figure 2.1 Conceptual Schematics of the WDM channel filtering based on the phase sensitive parametric
amplification. (b): Concept of nonlinear interactions in a PS-FOPA interactions.
(b)
(a)
HNLF2(PS-FOPA) SMF HNLF1(Copier)
P S I P S I P S I P S
Tuning fP
and lP
l
S1
l
S2
l
S3
f
P3 , l
P3
f
P2 , l
P2
f
P1 , l
P1
12
2.3 Experimental Setup
The Experimental setup is shown in Fig.2.2. Three WDM channels at wavelengths 1562.83 nm,
1563.5 nm and 1564.12 nm, carrying QPSK modulation are combined with a phased modulated
cw pump. All the waves are then launched to the 500m-long HNLF1 which acts as a copier to
generate idler waves as depicted in the spectrum observed from 1% tap at point A. The waves are
then injected into an optical processor which is basically a phase shifter for the pump. Optical
processor also equalizes the three channels before entering 500 m of HNLF2 (see the spectrum at
point B). Before entering the HNLF2, pump is strongly amplified by a 700 m-long Raman
amplifier. The output spectrum at point C shows the quasi-periodic trend of PSA that mainly takes
place inside HNLF2. After passing through an arrayed waveguide grating (AWG) pump is filtered
out and three channels are routed to a coherent receiver locked to a Local Oscillator (LO).
Figure 2.2 Experimental setup for tunable all-optical filter using PS-FOPA and the spectra at different location of
the filter.
2.4 Results
13
Fig.2.3(a) and (b) shows the output spectra of PS-FOPAs. As shown, PS-FOPA have acute
wavelength dependent output spectra [41]. Instructive or destructive interferences between pump
signals and generated idlers leave quasi periodic bumps and dips-typically narrower than 1nm- in
the output spectra [41-42]. The position/repetition frequency of these bumps and dips can be
changed by changing the pump wavelength. For instance, Fig. 2.3(a) shows the period reduction
by almost 50% around 1564.5 nm when pump wavelength is 6 nm larger. Also, Fig. 2.3(b)
demonstrates the interchange of bumps and dips around 1563.5nm having the pump wavelength
changed by 1nm.For the pump wavelength of 1555.4 nm, the middle channel modulated at 20
Gbaud symbol rate, is appropriately positioned at the center of a bump as depicted in the output
spectra of Fig. 2.3(b) while the two sideband channels are placed at the adjacent dips. Therefore,
the middle channel is selected from the others with an extinction ratio of nearly 13 dB. However,
the selection of channels at 1562.85 nm or 1564.12 nm needs a different procedure since they are
symmetrically positioned either at dips or bumps. Here, the pump wavelength is first tuned at
1554.4 nm so the two channels are equally placed at two bumps and the middle channel is in a dip
(Fig. 2.3(b)).
Figure 2.3 The constellation diagrams of the input and output of phase quantizer for different levels of phase noise.
1,567 1,566 1,565 1,564 1,563
-12
-8
-4
0
4
8
Wavelength (nm)
Gain (dB)
Pump wavelength of 1557nm
Pump wavelength of 1551nm
(b)
1562 1563 1564 1565 1566 1567
Wavelength (nm)
Power (a.u)
pump wavelength of 1554.4nm
pump wavelength of 1555.4nm
(a)
14
To cancel out either wavelengths of 1562.85 nm or 1564.12 nm the pump phase is adjusted to a
specific value by optical processor. At this specific pump, the channel at 1564,12 nm is suppressed
by 16 dB lower than channel at 1562.85 nm. Interestingly, by changing the pump phase by almost
90
o
from its previous value it is possible to have the opposite situation and select the channel at
1564.12 nm and suppress the one at 1562.85 nm (see Fig. 2.4(c)). This time the extinction ratio is
around 13 dB.
Also, two different symbol rates of 20 GB/s and 25 GB/s are considered and the BERs are
captured in the coherent receiver. These results are compared against the Back-to-Back (B2B)
results when the LO is tightly locked to the desired channel of the transmitter. Almost no OSNR
penalty is observed. It is important to mention that our proposed PSA produces 6dB overall gain
for each channel. The gain can be increased if high power EDFAs and pulsed pump are used [42].
Also, the filtering scheme used here can be extended to larger number of channels when farther
channels are killed out by tuning their phases.
Figure 2.4 Output power spectra for the selection of middle channel (a), Left channel (b), right channel. System
performance for the selection of right channel (a), Left channel (b), middle channel.
1545 1550 1555 1560 1565
-60
-40
-20
0
Wavelength (nm)
Power (dBm)
~13dB
1545 1550 1555 1560 1565
-60
-40
-20
0
Wavelength (nm)
Power (dBm)
~14dB ~16dB
1545 1550 1555 1560 1565
-60
-40
-20
0
Wavelength (nm)
Power (dBm)
~13dB
8 9 10 11 12 13
10
-5
10
-4
10
-3
10
-2
10
-1
Recieved OSNR (dB)
Bit Error Rate
B2B measurement of 20GB data at 1564.12nm
PSA selection of 20GB data at 1564.12nm
8 9 10 11 12 13 14
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Recieved OSNR (dB)
Bit Error Rate
B2B measurement of 20GB data at 1562.83nm
PSA selection of 20GB data at1562.83nm
8 9 10 11 12
10
-4
10
-3
10
-2
10
-1
Recieved OSNR (dB)
Bit Error Rate
B2B measurement of 20Gbuad data at 1563.5nm
PSA selection of 20Gbuad data at 1563.5nm
B2B measurement of 25Gbuad data at 1563.5nm
PSA selection of 25Gbuad data at 1563.5nm
15
2.7 Conclusion
In conclusion, we experimentally implemented a phase-sensitive amplifier to act as a tunable
optical filter. The filter is not only wavelength tunable but is also bandwidth-tunable using the
pump wavelength and pump phase as an optical knob. Using phase sensitive amplification, the
filter can also provide optical amplification for the selected channels. The extinction ratio between
13 dB to16 dB has been achieved and the system shows the net gain of around 6 dB and almost
zero OSNR penalty.
16
Chapter 3 Optical Buffer based on Discrete Time
Delays based on a Fiber Loop with an Internal Frequency
Shifter
3.1 Introduction
There are various applications for the ability to delay a data channel and choose among different
delay values [43,44]. Applications include microwave photonics for analog signals and data
synchronization and network management for digital channels [43-45]. The basic function that is
desired is the ability to create multiple copies of a data channel and then delay and select each
version differentially. By incorporating an internal frequency shifter into a relatively short fiber
loop, a recent report has shown that multiple differentially delayed pulses can be accessed at the
exit [45]. However, the report did not show data being transmitted and did not show continuous
data stream operation without time gaps in the analog or digital data channel.
In this chapter, a fiber loop with an internal frequency shifter is utilized to create and access
multiple delayed copies of a data channel. An optical delay of ~165 nsec is achieved avoiding the
need for long lengths of fibers. The linear phase frequency response of the system is verified for
analog signals and the SFDR is measured. As for digital data, the BER and constellation diagrams
of 10 Gbaud QPSK signals are obtained.
3.2 Concept
The conceptual block diagram of a discrete-time-based optical buffer is shown in Fig. 3.1. An
incoming light at the frequency f0 is modulated by a data signal and it is coupled to an optical fiber
17
loop at time t0. The light coupled to the loop continues to circulate through the loop till the overall
propagation loss prohibits any effective circulation. At each circulation the light experiences a
delay (DT) proportional to the loop length, and a shift in frequency (Df) as the result of an internal
frequency shifter. Therefore, different delayed copies of the input are preserved at equidistance
frequencies at the only output port of the loop (Fig. 3.1). It should be noticed as the different
delayed copies of the signal is stored in different spectral slots, then the structure is compact
compared to the previous designs for the optical buffer which used fiber banks for separating
delayed copies [43, 44].
Figure 3.1 The concept of the optical loop buffer.
3.3 Experimental Setup
The experimental setup of the optical loop with an internal frequency shifter is shown in Fig.
3.2. A cw laser at carrier frequency of 191.500 THz is intensity/phase modulated by an incoming
signal. Then light passes through a 50/50 coupler. Through the port 2 of the 50/50 coupler, light
goes to a liquid crystal on silicon (LCoS) filter, a pre-amplifier, a 1-nm tunable filter and a
photodetector (PD). The port 3 of 50/50 coupler directs the modulated laser to a dual parallel MZM
I/Q modulator [46]. With appropriate bias of the I/Q modulator, single sideband modulation (SSB)
of the carrier using an RF tone at frequency Df = 15 GHz is achieved, realizing the optical
f
0
+2Df
f
0
Frequency
Shifter (Df)
Delay
(DT)
f
0
+nDf
nDT
DT
f
0
+Df
t
t 0
t
0
+DT
t
0
+nDT
f
f+Df
(a)
(c)
1 2
3 4
Signal (Analog/Digital)
(b)
(d)
18
frequency shifter. Hence, the carrier suppression of around 31 dB and sideband suppression of
almost 40 dB are obtained. After passing through the I/Q modulator, light goes to a low noise
EDFA to adjust the power. After EDFA an isolator is placed to suppress any back-reflections. A
polarization controller (PC3) aligns the light to the axis of a polarization stabilizer. The latter filters
out the fast variations in the polarization. The output of the polarization stabilizer is then connected
to the port 4 of the 50/50 coupler.
Figure 3.2 Experimental setup. PC: Polarization Controller, PD: Photodetector, MZM: Mach-Zehnder Modulator.
3.4 Results
The LCoS filter can be adjusted to select any of the generated copies. The output spectra at port
2 of the 50/50 coupler is measured in Fig. 3.3 (a). As can be seen all the copies have been
effectively generated at one side of the carrier: longer wavelengths. The first copy is at the
frequency of the incoming light. Fig. 3.3 (b) shows the spectrum at the output of LCoS, for two
different adjustments corresponding to first copy and second copy selections.
f
0
+2Df
f
0
Frequency
Shifter (Df)
Delay
(DT)
f
0
+nDf
nDT
DT
f
0
+Df
t
t 0
t
0
+DT
t
0
+nDT
f
f+Df
(a)
(c)
1 2
3 4
Signal (Analog/Digital)
(b)
(d)
19
Figure 3.3 (a): Spectrum before LCoS. (b): Spectrum after LCoS. Experimental setup.
The system performance is measured using vector network analyzer (VNA). The magnitude
and phase of the transfer function S21 of the system for both first and second copies are plotted in
Fig. 3.4 (a) and Fig.3.4 (b). The nearly linear phase responses for both copies are observed. To
further evaluate the distortion of the system on analog signals, two tones at frequencies 10 GHz
and 10.5 GHz is used to modulate the input MZM, and the third harmonic intermodulation
distortion (IMD3) is measured. The result after PD and for the second copy selection is plotted in
Fig. 3.4 (c). A 29 dB dynamic range is observed. For this copy, the SFDR is found to be around
70 dB.Hz
2/3
.
Figure 3.4 (a): The magnitude of S21 for 1
st
copy (blue), 2
nd
copy (red) and open loop operation (dashed red). (b):
Phase of S21 for 1
st
copy (blue), 2
nd
copy (dashed red). (c): Third-order distortion of the system for 2
nd
copy.
f
0
+2Df
f
0
Frequency
Shifter (Df)
Delay
(DT)
f
0
+nDf
nDT
DT
f
0
+Df
t
t 0
t
0
+DT
t
0
+nDT
f
f+Df
(a)
(c)
1 2
3 4
Signal (Analog/Digital)
(b)
(d)
1565
191.25 191.3 191.35 191.4 191.45 191.5 191.55 191.6 191.65
-70
-60
-50
-40
-30
-20
-10
0
Frequency (THz)
Power (dBm)
Power(10dB/div)
1564 1564.5
nm
1565
Before
LCoS
1564 1564.2 1564.4 1564.6 1564.8 1565 1565.2
-50
-40
-30
-20
-10
0
Wavelength (nm)
Power(dBm)
Power(10dB/div) 1564 1564.5 1565nm
After
LCoS
(a)
(b)
-120
-100
-80
-60
-40
-20
Power(dBm)
2f1-f2 f1=10GHz f2=10.5GHz
2f2-f1
(c)
~29dB
(a) (b)
|S
21
| (dB)
0 2.5 5 7.5 10 12.5 15
-180
-90
0
90
180
Frequency (GHz)
Phase (degree)
0 2 4 6 8 10 12
-70
-60
-50
-40
-30
-20
Frequency (GHz)
20
Also, a pseudo-random bit pattern (PRBS) of 2
7
-1 bits with a clock of 500 MHz is used to
modulate the incoming light. An isolated pulse in the pattern is tracked to measure the amount of
delay between the first, second and third copies. According to Fig. 3.5, a delay of ~165 nsec is
observed.
Figure 3.5 Delays for first, second and third copies.
The system performance for digital data is assessed by modulating the incoming light with 10
Gbaud QPSK signals. The bit error rate and the constellation diagrams for first, second, third and
fourth copies are plotted in Fig. 3.6.
0 75 150 225 300 375
Third copy Second copy First copy
~165nsec
~165nse
c
Time(nsec)
21
Figure 3.6 Constellation diagrams and BER curves for the different copies of a QPSK data at the output.
3.4 Conclusion
To conclude, a fiber loop with an internal frequency shifter is utilized to create and access
multiple delayed copies of a data channel. An optical delay of ~165 nsec is achieved avoiding the
need for long lengths of fibers. The linear phase frequency response of the system is verified for
analog signals and the SFDR is measured. As for digital data, the BER and constellation diagrams
of 10 Gbaud QPSK signals are obtained.
-log
10
(BER)
OSNR (dB)
10 Gbaud QPSK channel
2
2.5
3
3.5
4
4.5
5
10 15 20 25
1st copy with delay 0
2nd copy with ΔT delay
3rd copy with 2ΔT delay
4th copy with 3ΔT delay
1
st
copy
No delay
2
nd
copy
delay ΔT
3
rd
copy
delay 2ΔT
4
th
copy
delay 3ΔT
22
Chapter 4 Optical Mitigation of Interchannel
Interference (ICI) for Multiple Spectrally Overlapped
WDM Channels using Nonlinear Wave Mixing
4.1 Introduction
Maximizing spectral efficiency, defined in terms of bits/sec/Hz being transmitted within an
available wavelength range of optical communication bandwidth, is a significant yet challenging
task [47,48]. Approaches to increase spectral efficiency include (i) reducing the guard band
between adjacent data channels and (ii) spectrally overlapping of the data channels [49,50].
However, these methods typically give rise to increased interchannel interference (ICI), thereby
requiring effective compensation techniques to recover data.
There have been reports of different approaches to reduce ICI in spectrally overlapped
wavelength division multiplexed (WDM) systems using electronic digital signal processing (DSP)
[51-58]. Typical DSP schemes for ICI reduction include the individual detection of each
wavelength channel across a WDM system [59-61]. Common digital multichannel ICI
compensation algorithms use the received crosstalk information to estimate the channel spacing
and reduce the crosstalk of each channel [62-67]. The physical implementation of the DSP
algorithm for ICI compensation usually requires a complex detection scheme that relies on
multiple synchronized receivers or a single receiver with high bandwidth [62, 63].
Alternatively, interchannel crosstalk can be mitigated prior to detection using optical
techniques, in which multichannel detection and channel spacing estimation are not necessarily
required for ICI compensation of a single target channel [64, 65]. In [64], an optical approach
23
based on optical multicasting, complex tailoring, and multiplexing for ICI mitigation of overlapped
channels of data carrying quadrature-phase-shift-keyed (QPSK) modulation is introduced and its
potential for adding and dropping optical QPSK channels is explored. In [65], a system of
overlapped channels carrying QPSK, and quadrature amplitude modulation (QAM) is optically
ICI compensated, and the ability of a scheme for ICI compensation of a hybrid overlapped system
with channels carrying different modulation formats is demonstrated. This paper considers ways
to further extend the approach of optical ICI compensation such that multiple spectrally overlapped
WDM channels can be recovered simultaneously.
Here, we optically mitigate [66] the interchannel crosstalk of multiple spectrally overlapped
channels of a WDM system within an individual element operating on multiple channels
simultaneously. The ICI mitigation takes place in three stages of periodically poled lithium niobate
(PPLN) waveguides. In the first stage, the optical conjugates of the WDM channels are constructed
using a set of concurrent nonlinear processes. The conjugate copies of the signals are separated
into two groups of even and odd channels. The amplitudes and phases of each channel in each
group are adjusted and coherently mixed with their adjacent crosstalk channels in the second-stage
PPLN to mitigate the ICIs. The conjugate copies are delayed, and in a third stage PPLN, these
copies are mixed with the crosstalk signals to further decrease the ICI level. We experimentally
demonstrate the proposed scheme using seven spectrally overlapped 20 Gbaud QPSK or 16-QAM
channels. A nearly 4-dB OSNR gain is achieved for QPSK data channels at a BER of 10
-3
. For
the 16-QAM channels the error vector magnitudes (EVMs) are reduced by almost 28% for a
channel spacing of 17.5 GHz.
4.2 Concept
24
Figure 4.1 shows the conceptual block diagram of the proposed optical ICI mitigation method
for multiple spectrally overlapped channels. As depicted in Fig. 4.1, the incoming overlapped
WDM channels are injected into a PPLN waveguide (PPLN-1) which is pumped by a CW laser at
the quasi-phase matching (QPM) wavelength [67]. Inside PPLN-1, through concurrent sum
frequency generation (SFG) and difference frequency generation (DFG), the WDM channels is
wavelength-converted. That is, the conjugate copies of the WDM channels at the symmetrical
spectral location with respect to QPM wavelength is generated. This is depicted in “conjugate
copies generation in PPLN-1” block of the scheme in Fig. 4.1.
Figure 4.1 Conceptual block diagram of optical interchannel crosstalk mitigation method. The scheme consists of
three main blocks: (i) Conjugate copies generation in PPLN-1 waveguide, (ii) Wavelength conversion in PPLN-2
which coherently mixes original channels with amplitude/phase adjusted conjugate copies (iii) Wavelength
conversion in PPLN-3 which coherently mixes original channels with amplitude/phase and delay adjusted
conjugate copies.
After conjugate copies generation block, the signals are sent to a two-output-port optical
programmable filter based on liquid crystal on silicon (LCoS) technology. Odd and even channels
Conjugate copies generation in
PPLN-1
Port-1 Port-2
? 2
? 2
?
SHG+ DFG Mixing
?
2
LCoS filter
Even and
conjugate of
odd channels
?
2
?
2
?
?
2
?
?
2
?
2
SHG+ DFG Mixing
LCoS filter
Odd and
conjugate of
even channels
Optical filter Optical filter
Even and
conjugate of
odd channels
Odd and
conjugate of
even channels
LCoS filter
SHG+ DFG Mixing
Wavelength conversion in PPLN-2
Wavelength conversion in PPLN-3
ICI-mitigated
even channels
ICI-mitigated
odd channels
Overlapped WDM
channels
PPLN-2
PPLN-3
PPLN-1
25
are separated at the outputs of the LCoS filter and directed to different paths. In the path following
port 1 of LCoS filter, the ICI of even channels is mitigated and in the path after port 2 of LCoS
filter, the ICI mitigation of odd channels is performed. In the following, we only explain the ICI
mitigation for even channels. Similar explanation is true for the ICI mitigation of the odd channels,
if the terms “odd” and “even” are exchanged.
At port 1 of the LCoS filter, even channels of the initial signal and odd channels of the conjugate
copied signal are passed. Simultaneously, the amplitudes and phases of the conjugate copies of
odd channels are adjusted such that crosstalk suppression to be performed in the next stage is
maximized. Next stage is composed of another PPLN waveguide (PPLN-2), in which through the
wavelength conversion, the original even channels and the twice-wavelength-converted odd
channels are coherently added.
Next, through another LCoS filter, even channels in the original wavelength region and the odd
channels of conjugate copies are selected. The amplitudes, phases and delays of the conjugate
copies are adjusted into the LCoS filter to further mitigate the ICI of even channels, through the
wavelength conversion in the third PPLN (PPLN-3) [65].
Since the input WDM channels, the pump, and the conjugate copies remain throughout the
wavelength conversion processes, there is no need to precisely adjust the frequency spacing among
channels, and coherent addition of the input signal and the wavelength converted signal becomes
possible. Also, we should mention that we have used PPLN in our experiments for three main
reasons: First because of the small size of the PPLN, there would be low latency or walk-off for
the propagating signals compared to highly nonlinear fibers (HNLFs). Second, high conversion
efficiency of PPLN potentially provides efficient nonlinear mixing of the signals. Third, since
26
PPLN has a χ
(2)
-type nonlinear response, the possible nonlinear mixing processes are second
harmonic generation (SHG), SFG and DFG which were used in our scheme. In an c
(3)
medium
such as an HNLF, more undesirable mixing terms are produced. Therefore, PPLNs can be
potentially used for the ICI mitigation of WDM channels with less crosstalk terms.
The mathematical representation of the proposed scheme is as follows. Consider three adjacent
signals: Si-1, Si, and Si+1. Waveform Y represents the spectral combination of these three signals
with a channel spacing of Δf and is defined as follows:
𝑌(𝑓) = 𝑆
)38
(𝑓−∆𝑓 )+𝑆
)
(𝑓)+𝑆
)98
(𝑓+∆𝑓 ) (4.1)
Without loss of generality, we assume that i is an even number. In this case i-1 and i+1 are odd
and the first and third terms in the right side of eq. 4.1 are interference terms. In PPLN-1, waveform
Y
*
is produced at symmetrical wavelength position with respect to the CW pump as schematically
plotted in the “conjugate copy generation” block of Fig.4.1. Note that “*” denotes the complex
conjugate. The two-output LCoS filter in the “Wavelength conversion in PPLN-2” block selects
Si, Si-1* and Si+1*. The amplitudes and phases of Si-1* and Si+1* are adjusted through complex taps;
ci-1 and ci+1 imposed by LCoS filter. The signals at port 1, compose the adjusted signals Xi-1 and
Xi+1 and Yi as follows:
𝑌
)
(𝑓) ∝ 𝛼
)
(𝑓)𝑆
)38
(𝑓−∆𝑓)+𝑆
)
(𝑓)+𝛼
)
(𝑓)𝑆
)98
(𝑓+∆𝑓 ) (4.2)
𝑋
)38
(𝑓) ∝ 𝑐
)38
𝑆
)38
∗
(𝑓−∆𝑓)+𝑐
)38
𝛼
)38
∗
(𝑓)𝑆
)
∗
(𝑓) (4.3)
𝑋
)98
(𝑓) ∝ 𝑐
)98
𝛼
)98
∗
(𝑓)𝑆
)
∗
(𝑓)+𝑐
)98
𝑆
)98
∗
(𝑓+∆𝑓) (4.4)
27
In equations (4.2-4.4), αj
*
(f), (j = i-1, i+1) denotes the filtering response of the optical
programmable filter, centered at the central frequency of the signal Sj
*
. Also, αi(f) represents the
filtering response of the optical programmable filter, centered at the signal Si. Since the channels
are overlapped, the filtering responses affect the neighboring channels. Inside the “Wavelength
conversion in PPLN-2” block, signal Yi is mixed with signals Xi-1
*
and Xi+1
*
. The following terms
will be obtained:
𝑌
3
(𝑓) = 𝛾
)38
𝑆
)38
(𝑓−∆𝑓)+𝛾
)
𝑆
)
(𝑓)+𝛾
)98
𝑆
)98
(𝑓+∆𝑓) (4.5)
In eq. 4.5, gi-1(f) = ai(f)+ ci-1
*
, gi(f) = 1+ ci-1
*
ai-1 (f) + ci+1
*
ai+1 (f) and gi+1(f) = ai(f)+ ci+1
*
. The
goal is to reduce the value of the crosstalk terms Si-1 and Si+1 through adjusting the coefficients, ci-
1 and ci+1. It should be noted that if the crosstalk terms are cancelled out completely, the tails of
the main signal to survive would also be suppressed. Therefore, ci-1 and ci+1 are adjusted to
minimize the coefficients of the crosstalk terms Si-1 and Si+1 and maximize the coefficient of Si.
Also, eq. 4.5 is similar to equations derived in [64] and will reduce to those equations if only one
crosstalk term is considered.
Furthermore, as demonstrated in the schematic spectrum inside the “Wavelength conversion in
PPLN-3” block of Fig. 4.1, signals gi-1
*
Si-1* and gi+1
*
Si+1* are selected at the output of the second
LCoS filter inside the block. The amplitudes and phases of gi-1
*
Si-1* and gi+1
*
Si+1* are adjusted with
complex taps c
’
i-1 and c
’
i+1. Relative delays; τi-1 and τi+1, are also imposed on gi-1
*
Si-1* and gi+1
*
Si+1*
by the LCoS filter. The adjusted signals are then described by:
28
𝑋
)38
:
(𝑓) ∝ 𝑐
)38
:
𝑒
;<=>?
!"#
𝛾
)38
∗
𝑆
)38
∗
(𝑓−∆𝑓)+𝑐
)38
:
𝑒
;<=>?
!"#
𝛾
)38
∗
(𝑓)𝛼
)38
(𝑓)𝑆
)
∗
(𝑓) (4.6)
𝑋
)98
:
(𝑓) ∝ 𝑐
)98
:
𝑒
;<=>?
!$#
𝛾
)98
∗
𝑆
)98
∗
(𝑓+∆𝑓)+𝑐
)98
:
𝑒
;<=>?
!$#
𝛾
)98
∗
(𝑓)𝛼
)98
(𝑓)𝑆
)
∗
(𝑓) (4.7)
In the PPLN-3, 𝑌
3
(𝑓) is coherently mixed with signals 𝑋
)98
:
∗
and 𝑋
)38
:
∗
. The resulting spectrum
in the original spectral region as schematically depicted in “Wavelength conversion in PPLN-3”
block of Fig. 4.1, would be:
Ψ(𝑓) = 𝛽
)38
𝛾
)38
𝑆
)38
(𝑓−∆𝑓)+𝛽
)
(𝑓)𝛾
)
(𝑓)𝑆
)
(𝑓)+𝛽
)98
𝛾
)98
𝑆
)98
(𝑓+∆𝑓) (4.8)
In which, bi-1=1+ c
’
i-1
*
e
-j2pfti-1
, bi(f) = 1+ c
’
i-1
*
e
-j2pfti-1
ai-1
*
(f) + c
’
i+1
*
e
-j2pfti+1
ai+1
*
(f) and bi+1= 1+
c
’
i+1
*
e
-j2pfti+1
. To mitigate the ICIs on channel Si, the values of cj, cj', and τj (j = i-1 or i+1) should
be adjusted to have 𝛽
)38
𝛾
)38
and 𝛽
)98
𝛾
)98
relatively smaller compared to the coefficient of Si, that
is, 𝛽
)
(𝑓)𝛾
)
(𝑓). In this case, the spectrum Ψ(𝑓) yields the ICI mitigation of channel Si. Similar to
[57, 61, 64], the non-deterministic nature of signals due to the data and noise has not taken into
account in the derivation of the above equations. Note that, the optical signal to noise ratio of the
channels can be decreased by propagating through optical devices and nonlinear elements.
However, the focus of this paper is on the scenarios for which the main cause of signal distortion
is the ICI of neighboring channels. An overall improvement in signal quality is obtained by the
mitigation of those ICIs through the proposed scheme.
4.3 Experimental Setup
29
Figure 4.2 shows the experimental setup to demonstrate the proof of concept for the proposed
optical ICI mitigation scheme. Seven tunable narrow-linewidth laser sources are clustered in two
groups to emulate the odd and even channels. The even and odd channels are created by two I/Q
modulators with independent data streams. A pseudo-random bit pattern (PRBS) generator with
pattern length of 2
15
-1 and not-return-to-zero (NRZ) pulses are used to generate the data streams.
To overlap the WDM channels, the optical frequencies of these seven channels are chosen so that
their difference, Δf, is smaller than the baud rate of the data. The polarization of each signal
channel is independently tuned to be aligned with the principal axis of the MZM modulator. The
odd and even channels are combined using a 50/50 coupler. Optical signals are amplified using
two stages of erbium-doped fiber amplifiers (EDFAs) and mixed with a CW pump laser at ~1540
nm. Before being combined, the CW pump laser is amplified to ~22 dBm using an EDFA followed
by a tunable 1 nm filter. The optical signals and the CW pump are sent into the first PPLN
waveguide to generate the conjugate copies of the signals. The QPM wavelength of the PPLN
waveguide is temperature tuned and stabilized around the CW pump wavelength. This allows for
the maximum conversion efficiency of the SHG and DFG processes inside the PPLN. In Fig. 4.2,
the spectrum at point A shows the seven original channels, their generated conjugate copies, and
the CW pump.
The signals, the conjugate copies, and the pump are sent to a two-output ports spatial light
modulator (SLM) filter based on LCoS technology, in which the odd and even conjugate copies
based on target channel selections for ICI mitigation of either odd or even channels are selected,
and the amplitudes and phases of the signals and the conjugates are adjusted. Note that to mitigate
the ICIs of even (odd) channels, the odd (even) signals from the generated conjugate copies and
30
the original even (odd) channels are selected. The adjusted signals, the conjugates, and the CW
pump are amplified to ~21 dBm and sent into a second PPLN waveguide with a similar QPM
wavelength as the first PPLN waveguide. In this second waveguide, the signals are mixed with
amplitude- and phase-adjusted crosstalk channels to reduce the ICIs. By using another LCoS filter,
the conjugate copies’ amplitudes and phases are adjusted. In this second filter, the conjugate copies
are further delayed and sent to another PPLN waveguide along the original target channels for ICI
mitigation. Inside this PPLN waveguide, the signals and delayed variants of the crosstalk
neighboring signals are mixed to further mitigate the ICIs. In each of these steps, it is not necessary
to estimate the channel spacing because the pump and signals are preserved throughout the
nonlinear processes. Channel spacing remains unchanged throughout each nonlinear interaction.
The ICI mitigated channels are filtered and sent into a coherent detector to record the constellation
diagrams and measure the BER. For offline DSP, we have used frequency offset and phase noise
compensation. We avoided using adaptive equalization in DSP algorithms to be able to observe
the performance of the proposed scheme for optical ICI mitigation.
Figure 4.2 (a) Experimental setup. PPLN: periodically poled lithium niobate, PC: polarization controller, LCoS: liquid
crystal on silicon. (b) Optical spectrum of generated signal conjugate copies for 20 Gbaud overlapped QPSK channels
at point A and optical spectrum of ICI mitigated odd channels 1, 3, 5, and 7 after the last PPLN waveguide at point B.
Ch5
Ch2
5 nm 3 nm 5 nm
Data1
Mod.
BPF
PC
CW Pump
( 1540 nm)
1 nm
5 nm
Mod.
Data2
PPLN
Ch3
Ch7
Ch4
Coherent
Receiver
LCoS
Filter
13 nm
PPLN
LCoS
Filter
PC
EDFA
13 nm
PPLN
Ch1
Ch6
A
B
0 10 0 20 0 300 400 50 0 600 700 80 0 90 0 1000
-7 0
-6 0
-5 0
-4 0
-3 0
-2 0
-1 0
0
0 100 200 30 0 400 50 0 600 70 0 800 900 100 0
-8 0
-7 0
-6 0
-5 0
-4 0
-3 0
-2 0
-1 0
Pump
Conjugate copies
of 7 channels
7 overlapped
Channels
ICI mitigated
channels 1,3,5,7
1540 1536 1544
1540 1536 1544
Wavelength (nm)
Conjugate
Copies
Pump
B
A
Wavelength (nm)
P
o
w
e
r
(
1
0
d
B
/
d
i
v
)
P
o
w
e
r
(
1
0
d
B
/
d
i
v
)
(a)
(b)
31
4.4 Experimental Results
The seven lasers of the previous experimental setup are first modulated by two independent
electrical QPSK data streams. Figure 4.3 shows the constellation diagrams of channels 1, 3, and 6.
The constellations are measured (i) without optical ICI mitigation (back to back) and (ii) with
optical ICI mitigation. To achieve the constellation diagrams for the ICI-mitigated signals the
coefficients 𝑐
)
and 𝑐
)
:
and delays are manually tuned by monitoring the received error vector
magnitude (EVM).
The channels are modulated with 20 Gbaud signals, and experiments are run for three different
values for channel spacing: 17.5 GHz, 20 GHz, and 25 GHz. The ICI mitigation method provides
negligible benefit when the channel spacing is larger than the baud rate of the signals, which is
acceptable because the ICI effect there is insignificant (∆𝑓=25 GHz in Fig. 4.3). When the channel
spacing is equal to or less than the signal baud rate, the ICI is significant, and the ICI mitigation
on the proposed method becomes noticeable (∆𝑓=20 GHz and 17.5 GHz in Fig. 4.3).
For all three channel spacings, the ICI mitigations for the odd channel (channel 3) shows similar
performance as for the even channel (channel 6). Note that both channels 3 and 6 incur two
crosstalk terms from two neighboring channels. The slightly lower EVMs for channel 1, which
incurs just one interference term, can be attributed to lower power of this channel as demonstrated
in the spectrum of Fig. 4.2(b).
32
Figure 4.3 Experimentally recorded signal constellation diagrams of channels 1, 3, and 6 with (w.) and without (w/o)
optical ICI mitigation method for 20 Gbaud overlapped QPSK signals and at channel spacings of 17.5 GHz, 20 GHz,
and 25 GHz.
Figure 4.4 shows the BER versus optical signal-to-noise ratio (OSNR) results for channels 1, 3,
and 6 carrying 20 Gbaud QPSK signals and with the channel spacing of 17.5 GHz. For a QPSK
channel with channel spacing of 17.5 GHz, the required OSNR to achieve a BER of 10
-3
is reduced
by ~4 dB after optical ICI mitigation.
Figure 4.4 BER measurements with (w.) and without (w/o) optical ICI compensation method for QPSK overlapped
channels and for ∆f =17.5 GHz
Channel-6 Channel-1 Channel-3 Channel-6
Channel spacing=17.5 GHz
Channel spacing= 20 GHz
Channel-1 Channel-3
w/o optical ICI
mitigation
w. optical ICI
mitigation
Channel spacing= 25 GHz
Channel-6 Channel-1 Channel-3
EVM=37.8% EVM=38.3% EVM=38.5% EVM=35.7% EVM=36.1% EVM=36.6%
EVM=25.8% EVM=26.9% EVM=26.2% EVM=27.9% EVM=27.7% EVM=28.2%
EVM=20.3% EVM=20.8% EVM=21.1%
EVM=24.1% EVM=24.7% EVM=24.8%
-log
10
(BER)
OSNR (dB)
20 Gbaud QPSK, ∆f=17.5 GHz
2
2.5
3
3.5
4
4.5
5
10 15 20 25
ch1 w/o ICI Comp.
ch1 w. ICI Comp.
ch3 w/o ICI Comp.
ch3 w. ICI Comp.
ch6 w/o ICI Comp.
ch6 w. ICI Comp.
33
To further investigate the performance of the proposed ICI mitigation scheme for WDM
channels, a different format of the modulation is considered. The set of even and odd channels are
now modulated with 16-QAM data. Figure 4.5 shows the constellation diagrams of channels 1, 3,
and 6 with and without the optical ICI mitigation method under channel spacings of 17.5 GHz and
20 GHz. The constellation diagrams for a channel spacing of 25 GHz are not shown here because
the ICI mitigation is again insignificant for a Df larger than baud rate. Like QPSK constellation
diagrams of the overlapped channels, the EVMs for 16-QAM signals are also generally reduced
which shows possible modulation transparency of the proposed ICI mitigation scheme. The EVMs
for all 16-QAM channels of the WDM system with channel spacing of 17.5 GHz are reduced by
almost 28%. Figure 4.6 shows the BER results of channels 1, 3, and 6. In this case, the channel
spacing is 17.5 GHz.
Figure 4.5 Experimentally recorded signal constellation diagrams of channels 1, 3, and 6 with (w.) and without (w/o)
optical ICI mitigation method for 20 Gbaud overlapped 16-QAM signals and at different channel spacings of 17.5
GHz and 20 GHz.
Channel-6 Channel-1 Channel-3 Channel-6
Channel spacing=17.5 GHz
Channel spacing= 20 GHz
Channel-1 Channel-3
w/o optical ICI
mitigation
w. optical ICI
mitigation
EVM=18.1% EVM=18.2% EVM=18.4%
EVM=13.3% EVM=13.6% EVM=13.3% EVM=12.7% EVM=12.9% EVM=13.8%
EVM=15.1% EVM=15.5% EVM=15.8%
34
The 16-QAM channels are much more prone to the destructive effect of ICI than are the QPSK
channels. Therefore, the BER versus OSNR curves of the QAM signals without ICI compensation
fail to retain their linear trends, for a BER value around 10
-2.5
and up. Again, a similar performance
for ICI mitigation of odd and even channels is observed.
Figure 4.6 BER measurements with (w.) and without (w/o) optical ICI compensation method for 20 Gbaud 16-QAM
overlapped channels and for ∆f =17.5 GHz.
4.5 Conclusion
To conclude, A method for optical mitigation of the ICI of multiple spectrally overlapped data
channels is experimentally demonstrated. The method is based on a cascade of conjugate wave
generations along with phase, amplitude, and delay adjustments. For ICI mitigation using this
method, individual detection and channel spacing estimation is not required, and the ICI of all
channels can be mitigated simultaneously. The system performance is assessed for multiple
spectrally overlapped 20 Gbaud QPSK and 16-QAM data channels. The similar performance of
the method for both QPSK and 16-QAM channels shows the potential modulation transparency of
2
2.5
3
3.5
19 21 23 25 27
ch1 w/o ICI Comp.
ch1 w. ICI Comp.
ch3 w/o ICI Comp.
ch3 w. ICI Comp.
ch6 w/o ICI Comp.
ch6 w. ICI Comp.
-log
10
(BER)
OSNR (dB)
20 Gbaud 16QAM, ∆f=17.5 GHz
35
the scheme. The BERs are measured for 20 Gbaud signals and under different channel overlapping
(spacing) conditions inducing different ICIs. After optical ICI mitigation, a reduction of almost 4
dB is obtained in the required OSNR to achieve a BER of 10
-3
for 20 Gbaud QPSK signals with a
channel spacing of 17.5 GHz. The optical ICI compensation scheme has also been used for an
overlapped WDM system of a 20 Gbaud 16-QAM signals with channel spacings of 17.5 GHz and
20GHz. The EVMs for all 16-QAM channels of a WDM system with channel spacing of 17.5
GHz, are reduced by almost 28%.
36
Chapter 5 Optical Generation of Nyquist WDM/TDM
Channels with Sinc-shaped Temporal Pulse Trains using
Microresonator-based Kerr Frequency Comb
5.1 Introduction
Nyquist pulses are a family of waveforms which theoretically have zero inter-symbol
interference (ISI) since the pulses have minima in the center of neighboring time slots [68-70].
Among Nyquist waveforms, the temporal sinc-shaped pulse has an ideal rectangular spectrum and
zero roll-off factor [70]. Such waveforms are interesting since they produce a compact data channel
bandwidth for high spectral efficiency and yet with low ISI [71].
A single optical sinc-shaped Nyquist temporal pulse train can be generated using a single laser
source in which an external modulator is driven by a sinc-shaped electronic signal from an arbitrary
waveform generator (AWG) [72,73]. A single sinc-shaped pulse train can also be generated
optically using a spatial modulator to shape a train of Gaussian pulses into a train of sinc-shaped
pulses [71].
Another approach [74] was reported for optically generating a single sinc-shaped pulse train
in which a single laser line is externally amplitude modulated. This modulation produces additional
coherent lines around the original laser line such that an overall rectangular Nyquist spectral shape
is achieved. Recently, it was shown that (i) externally amplitude modulation of a Kerr frequency
comb can insert coherent lines into the comb [75] and (ii) a single superchannel is generated using
the line insertion into a Kerr frequency comb [76]. The inserted lines produce coherent subcarriers
that are mutually orthogonal with sinc shapes in the frequency domain [76].
37
In this chapter, we demonstrate the simultaneous and reconfigurable optical generation of
multiple Nyquist-shaped WDM channels having temporal sinc-shaped carrier pulse trains [77].
The channels are generated through the insertion of coherent lines around the spectral lines of a
microresonator-based Kerr optical frequency comb [77-80] using cascaded continuous-wave
(CW) amplitude modulations. Throughout the process of generating the sinc-shaped pulse trains,
the coherency and frequency-locking of the comb lines remain intact. The Nyquist-shaped WDM
channels exhibit fairly rectangular spectra and therefore can be placed close to each other with
almost zero guard-band. We use nine Kerr frequency comb lines and insert sub-groups of uniform
and coherent lines to generate 9 WDM channels. The deviation from ideal Nyquist pulses is
between 4.2%-6.1% for the 9 channels at a repetition rate of 6 GHz. For the 9 channels at a
repetition rate of 2 GHz, the deviation from theory is found to be between 2%-4.5%. Each WDM
channel is modulated with a 6 Gbit/s on-off keying (OOK) data. We also demonstrate the
reconfigurability of this approach by varying the number of channels, generating pulse trains with
repetition rates of 110 ps and 333 ps, different temporal widths of 74 ps and 37 ps, and two or eight
zero-crossings in one period.
Also, an application, we experimentally demonstrate the generation and time division
multiplexing (TDM) of Nyquist sinc-shaped pulses from a single microresonator-based Kerr
frequency comb [81]. The Nyquist sinc-shaped pulses are constructed using amplitude modulation
of Kerr comb spectral lines. Pulse trains are delayed proportionally so peak of each pulse is placed
at the zero-crossing of another pulse. Thanks to the minimum ISI property of Nyquist sinc pulse
trains the data for each channel of a TDM system is recovered and the system performance is
evaluated in terms of bit-error-rate (BER).
38
5.2 Concept of optical generation of WDM Nyquist channels
The concept of optical generation of WDM Nyquist sinc-shaped pulse trains using the Kerr
frequency comb is shown in Fig. 5.1. A CW light is injected into the Kerr frequency generation
block consisting of a silicon nitride (Si3N4) microresonator. Subsequently, the flat coherent line
insertion block converts the CW WDM channels to the rectangular-shaped WDM channels. This
block optically synthesizes a series of spectrally uniform lines referred to as sub-comb around each
line of the Kerr frequency comb. The flat coherent line insertion block is made of a chain of Mach-
Zehnder Modulators (MZMs) driven by frequency-locked radio frequency (RF) signals. The bias
and the RF voltage of each MZM are adjusted to produce the first sidebands at the same level with
the carrier and simultaneously suppress the higher order sidebands [74]. Therefore, a single MZM
produces three lines for any comb line. Consequently, if n MZMs are used, a total number of N =
3
n
uniform
lines would be inserted for each Kerr comb line. Each set of N uniform lines forms a
rectangular-shaped sub-comb. In order to insert equidistance lines, the modulating frequency of
each modulator is set to 1/3 of the modulating frequency of its previous modulator. The inserted
lines are placed at distance Df away from each other, which is equal to the modulating frequency
of the last MZM. Also, the delay between the paths of each RF signal is adjusted to have an equal
phase for all RF signals. This results in a uniform phase across the optical lines of each sub-comb.
39
Figure 5.1 Concept of WDM Nyquist channels with sinc-shaped pulse trains using a Kerr frequency comb.
Using straightforward calculations [74], it can be shown that each sub-comb consisting of N
uniform and phase-locked spectral lines, separated by Df, represents the Fourier transform of a
time-waveform S(t) such that [74]:
𝑆(𝑡) = sin(𝜋𝑁Δ𝑓𝑡)/[𝑁𝑠𝑖𝑛(𝜋Δ𝑓𝑡)] (5.1)
The above equation shows that 𝑆(𝑡) is a periodic function with a temporal period of 1/Df.
Within one period, the waveform 𝑆(𝑡) resembles a time-unlimited sinc Nyquist pulse with zero
roll-off factor. The waveform 𝑆(𝑡) has N-1 zero-crossing points in each period. The pulse duration
between zero-crossings is equal to 2/(NDf). Zero-crossing points in 𝑆(𝑡) allow for zero ISI
transmission and the rectangular-shaped spectrum ensures the maximum spectral efficiency
associated with Nyquist pulses with a zero roll-off factor [71].
5.5 Experimental Setup
Figure 5.2 shows the experimental setup for optical generation of WDM rectangular channels
with sinc-shaped carriers using the microresonator-based Kerr frequency comb. In the “Kerr
frequency comb generation” block, a micro-ring resonator produces a frequency comb to obtain
Subcomb2
Subcomb3
MZM MZM MZM
MZM MZM
Flat coherent line insertion
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟏
𝐟
𝟐
𝐟
𝟐
……….
Comb line 1 Comb line 2 Comb line 3
CW WDM channels
𝑵=𝟑
𝑵=𝟑 𝑵=𝟑 𝑵=𝟗 𝑵=𝟗 𝑵=𝟗
Laser
Integrator
Subcomb1
Time
Kerr frequency
comb generation
Subcomb1 Subcomb2 Subcomb3
….. ….. ….. …..
….. …..
Rectangular WDM channels
𝑵=𝟑
𝒏
𝑵=𝟑
𝒏
𝑵=𝟑
𝒏
𝐟
𝐧
𝐟
𝟏 𝐟
𝟐
40
Kerr comb-lines. The resonator is fabricated using a Si3N4 waveguide with a cross section of 1.5
μm × 0.9 μm. The quality (Q) factor of the resonator and the resonance bandwidth are 1.3×10
6
and
150 MHz, respectively. The ring is pumped with a CW light from an external cavity diode laser
(ECDL) at approximately 1555.0 nm, which is in the anomalous dispersion region of the ring. The
pump laser is amplified to 4.2 W. The amplified spontaneous emission of the amplifier is filtered
out using a filter with a bandwidth of 0.4 nm and the output is coupled to the chip through a lensed
fiber. The low phase-noise Kerr comb state is obtained when the pump wavelength is red-detuned
[82]. The comb lines are then directed to a tunable sharp filter in which up to nine coherent
selected. Next, the selected comb lines enter the “sinc-shaped pulse generation” block which
comprises two cascaded MZMs. The biases of two MZMs are properly adjusted to obtain uniform
line insertion. The modulating frequencies of the two MZMs, f1 and f2, are also frequency-locked
such that f2 = f1/3. This ensures the insertion of equidistance lines. The phase shifter in the path of
RF signal of second MZM compensates the residual phase difference between the two RF signals.
The roll-of-factor of the generated pulses can be tuned by adjusting the biases of the MZMs and
the RF phase shift. At the output of first MZM, each comb line produces three flat and coherent
lines. These lines generate a sinc-shaped pulse train with a repetition rate of f1. Subsequently, at
output of the second MZM, each comb line generates nine coherent lines which represent a sinc-
shaped pulse train with a repetition rate f2.
41
Figure 5.2 Experimental setup; PC: Polarization Controller, PS: Phase Shifter; Att: Attenuator; EDFA: Erbium-
Doped Fiber Amplifier; OSA: Optical Spectrum Analyzer; OSC: Oscilloscope; BERT: Bit Error Rate Tester.
Measured spectra at points (1), (2) and (3) of the setup are shown.
These sinc-shaped pulse trains are then directed to the “sinc-shaped pulses encoding” block.
The sinc-shaped pulse trains are OOK-modulated with a pseudo random bit sequence (PRBS) of
length 2
31
-1 and rate f2 bps. Subsequently, the modulated pulses are sent to a dispersion-
compensating fiber (DCF) followed by a programmable amplitude/phase optical filter based on
Liquid Crystal on Silicon (LCoS) technology to adjust the delay and de-correlate the data pattern
of each sinc pulse train. The output of the “sinc-shaped pulses encoding” block is sent to the
“detection” block. In one branch of the “detection” block, a photodetector (PD) generates the
output temporal waveform. Another branch of the detection block measures the bit error rate
(BER).
5.3 Experimental Results
(1)
Wavelength(nm)
(2)
1544 1546 1548
Power(dBm)
(3)
1 nm
12 nm
Tunable
Sharp filter
12 nm
PC
EDFA
LCoS filter OSC
BERT
Bias
Bias
Synchro
𝐟
𝟏
𝐟
𝟐
Kerr frequency comb generation
Sinc-shaped pulse generation
Detection
Microresonator
Intensity Mod. 12 nm
DCF
12 nm
PD
Sinc-shaped pulses encoding
(1)
(3)
(2)
MZM MZM
MZM MZM
MZM
MZM
PS
PRBS
ECDL
1500 1550 1600
Power(dBm)
20
-20
-60
Power(dBm)
-10
-30
-50
1535 1545 1555
Wavelength(nm)
Wavelength(nm)
(1)
ATT
OSA
-14
-24
-34
42
To show the tunability of the generated sinc-shaped pulse trains, a single comb line is first
selected. The first MZM driven by a modulating frequency of f1=9 GHz is bias-controlled to
generate a sub-comb with three flat lines separated by 9 GHz and occupying a total bandwidth of
3×9=27 GHz. The resulting sinc-shaped pulse train is shown in Fig. 5.3(a). The sinc-shaped pulse
train has a temporal period of ≈ 110 ps (corresponding to a repetition rate of 9 GHz) and a pulse
duration that is 2/(27 GHz) ≈ 74 ps wide. Furthermore, at the second MZM driven by a modulating
frequency of f2 = 3 GHz, the 27 GHz bandwidth of the pulse train in Fig. 5.3 (a) will be populated
by a total number of nine spectral lines. Figure 5.3(b) shows the sinc-shaped pulse train at the
output of second MZM. The second MZM reshapes the pulse train further and elongates its
temporal period to 1/(3 GHz) ≈333 ps by adding multiple consecutive zero-crossings while
preserving the pulse duration at 74 ps.
Figures 5.4(a) and 5.4(b) show four selected Kerr comb-lines of a microresonator with a free
spectral range (FSR) of 71.728 GHz and their corresponding generated sub-combs after second
MZM, respectively. The slight power difference of sub-combs can be due to initial power
difference between comb lines and the dispersive response of the MZMs.
43
Figure 5.3 Measured sinc-shaped pulse trains at the outputs of (a): the first MZM driven by f1=9 GHz and (b): the
second MZM driven by f2=3 GHz recorded by a sampling scope.
Figure 5.4 (a): Four Kerr comb lines (FSR=71.7 GHz). (b): Four corresponding sub-combs. (c-f): Sinc pulses
observed in OSC (black line) and calculated by Eq. (5.1) (dashed red line).
40.3 40.4 40.5 40.6 40.7
0
0.2
0.4
0.6
0.8
1
Time(nsec)
Power (a. u.)
40.3 40.4 40.5 40.6 40.7
0
0.2
0.4
0.6
0.8
1
Time(nsec)
Power (a. u.)
(a) (b)
40.3 40.4 40.5 40.6 40.7
0
0.2
0.4
0.6
0.8
1
Time(nsec)
Power (a. u.)
40.3 40.4 40.5 40.6 40.7
0
0.2
0.4
0.6
0.8
1
Time(nsec)
Power (a. u.)
(a) (b)
74 ps
110 ps
333 ps
74 ps
Df = f1
Df = f2
N = 3
N = 9
40.3 40.4 40.5 40.6 40.7 40.3 40.4 40.5 40.6 40.7
Time (ns) Time (ns)
1
0.8
0.6
0.4
0.2
0
Normalized Power
1
0.8
0.6
0.4
0.2
0
(a)
(b)
Pulse train 2
1533.5 1534.5 1535.5
Line1
Line2 Line3
Line4
-20
-40
-60
Power(dBm)
Sub-comb1
Sub-comb2
Sub-comb3
Sub-comb4
-25
-35
-45
Power(dBm)
(a) (b)
Normalized power
(f)
(d)
(e)
(c)
-1 -0.5 0 0.5 1
1533.5 1534.5 1535.5
1
0.5
0
1
0.5
0
1
0.5
0
Wavelength(nm) Wavelength(nm)
Pulse train 1
1
0.5
0
Normalized power
Pulse train 3
Normalized time Normalized time
Pulse train 4
-1 -0.5 0 0.5 1
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
44
For the generated sub-combs, the modulating frequencies of f1=21 GHz and f2=7 GHz are
chosen. Each sub-comb is composed of nine flat spectral lines separated by 7 GHz, extending over
a bandwidth of 63 GHz. The generated sinc-shaped pulse train for each of the four WDM channels
is shown in Fig. 5.4(c-f). The horizontal and vertical axes are normalized to 1/f2 and the maximum
power, respectively. Also, the experimental results are compared to the theoretical sinc-shaped
pulse trains generated by Eq. (1). Both theoretical and experimental results are plotted in
Fig. 5.4(c-f).
Figure 5.5 (a): Experimental output spectrum of nine Nyquist WDM channels using a Kerr comb (FSR=192.0020
GHz). (b): Measured sinc-shaped pulse trains of different channels.
1536 1540 1544 1548
-10
-30
-50
Power (dBm)
CH1
CH2
CH4
CH5
CH6
CH8
CH9
CH3
CH7
(a)
54GHz
CH4
CH8
CH3 CH2
CH9
CH6
CH5
Power (a.u)
CH7
(b)
CH1
167psec
38 38.7
37 psec
38 38.7
Time(nsec)
38 38.7
Time(nsec)
38 38.7
Time(nsec)
Power (a.u) Power (a.u)
38 38.7 38 38.7
38 38.7 38 38.7 38 38.7
Wavelength(nm)
45
The experimental setup is also employed to dual-modulate nine Kerr comb lines with an FSR
of 192.0020 GHz. The modulating frequencies of f1 = 18 GHz and f2 = 6 GHz are selected for the
first and second modulator, respectively. The resulting spectrum after the second MZM shows nine
WDM channels as depicted in Fig. 5.5(a). Each WDM channel represents a rectangular-shaped
sub-comb. Each sub-comb consists of nine lines with a frequency separation of f2 = 6 GHz,
covering a wavelength span of 54 GHz. For each sub-comb, a sinc-shaped pulse train exhibiting
eight zero-crossings during one temporal period is obtained. These sinc-shaped pulse trains are
depicted in Fig. 5.5(b). The temporal period of all these channels is equal to 1/f2 ≈ 167 ps and their
pulse duration is 2/(9f2) ≈ 37 ps.
In order to evaluate the quality of the generated sinc-shaped pulses, we compare the measured
pulse trains of each channel to the theoretical formula of Eq. (1). For each of the nine channels of
Fig. 5.5(b), we calculate the normalized root-mean-square error as compared to the theoretical
formula given in Eq. (1), with zero roll-off factor. It should be noted that the repetition rate, Df, of
the theoretical pulse train is equal to the second modulating frequency f2. The deviation from zero
roll-off factor slightly varies with the channel number. To show the potential capability of using
generated pulse trains for carrying data, the sinc-shaped pulse trains of Fig. 5.5 are OOK-
modulated with a data at 6 Gbit/s. The BER curves versus optical signal to noise ratio (OSNR) are
plotted in Fig. 5.6. We believe that the difference in the BER curves can be primarily attributed to
the unequal power levels for the WDM channels as seen in Fig. 5.6 (a); we believe this issue and
can be alleviated using equalization techniques in the generator.
46
Figure 5.6 The BER for nine 6 Gbit/s OOK-modulated WDM channels, with sinc-shaped pulse trains as carriers.
5.4 TDM of Nyquist Sinc-Shaped Channels
Optimal use of system bandwidth has become important in high-performance, spectral efficient
optical communication systems [83]. One specific example that is gaining in importance for the
bandwidth-efficient optical communication is the use of Nyquist data channels, for which the time
waveform is the sinc shaped pulse train trains. For these channles: (i) the channel spectrum is
relatively rectangular (ii) the spectral inefficient guardbands between channels can be minimized
[84,85] and (iii) minimum inter-symbol interference (ISI) can be achieved when Nyquist pulses
are used as time channels in a Time Division Multiplexing (TDM) system. Generating Nyquist
pulse trains that can be data encoded is a key enabler for TDM systems carrying Nyquist pulse
trains. Typical TDM systems would require a Nyquist transmitter for each time channel, which
normally generates sinc pulses by an electronic arbitrary waveform generator [86,87]. In this
manner, Nyquist pulses can be generated using Nyquist filtering or pulse shaping. Nyquist filtering
takes advantage of a sinc-shaped impulse response with multiple zero-crossing and is often
implemented in electrical domain, which limits the speed the limited bandwidth of electronics.
2
4
6
8
10
18 19 20 21 22 23 24
OSNR (dB)
-Log(BER)
2
4
6
8
10
18 19 20 21 22 23 24
CH1 CH2 CH3
CH4 CH5 CH6
CH7 CH8 CH9
47
Optical generation and time multiplexing of sinc-shaped Nyquist pulse trains can have its own
exclusive advantages. Reconfigurability in terms of number of time channels and tunability in
terms of the buadrate of modulating data are two important advantages of optical generation and
time multiplexing of Nyquist pulses. It has been reported in [88,89] that pulse shaping of a mode
locked laser output can lead to optical Nyquist pulse generation. The baudrate of the data carried
by sinc-shaped pulse generated in this method is then limited to mode-locked laser repetition rate,
which is very hard to be continuously tunable in wide ranges especially in active designs [90,91].
Also, different mode locked lasers should be used to practically imitate a system with multiple
access which results in more physical ports and bulkier (de)multiplexing scheme.
It was recently reported that the proper filtering of several equal-power lines from a micro-
resonator Kerr comb can produce a source of Nyquist pulses for multiple wavelengths [92,93]. In
this paper we use these optically generated Nyquist pulse trains as the time channels of a TDM
system. In this manner, high quality sinc-shaped pulse trains with fairly rectangular spectrum, are
optically generated and time-multiplexed to achieve higher bitrates while minimum bandwidth is
occupied.
5.5 Concept of TDM of Nyquist Sinc-Shaped Channels
The concept of optical TDM of Nyquist sinc-shaped pulse trains using the Kerr frequency comb
and PPLN is demonstrated in Fig.23. As it is shown in Fig.5.7, a CW light is injected into the Kerr
frequency generation block which is a silicon nitride microresonator [94]. A programmable
phase/amplitude filter which is based on liquid crystal on silicon (LCoS) technology selects several
comb lines and partitions the comb lines in two segments and directs each segment to one of its
output ports. At one port half of the comb lines directly goes to a PPLN waveguide. At another
48
port, the other comb lines pass through optical Nyquist pulse generation block which all-optically
synthesizes a series of spectrally rectangular shaped lines around each coherent synchronized RF
tones. A uniform set of strictly phase/frequency-locked spectral lines represents a particular
temporal pulse train which is sinc-shaped and periodic. The mathematical equation of such pulse
trains is given as [74]:
𝑆𝑖𝑛𝑐
%
(𝑡) = ∑ 𝑠𝑖𝑛𝑐(𝑁Δ𝑓
@AB
@A3B
(𝑡−
@
C>
)) (5.2)
In which N is the number of spectral lines and Df is the frequency separation between each two
lines. The sinc-shaped pulse train of eq. 1 has a period of P = 1/Df and N-1 zero-crossing points at
ti = i/NDf for i={1,2,…,N-1}, in each period.
The RF tones of MZMs have a phase/amplitude specific relation and the sinc-shaped pulse
train repetition rate is equal to the RF drive of the last MZM [92]. If n MZMs are used, each
rectangular shaped channel represents a sinc-shaped Nyquist pulse train with a main lobe and
fixed, equidistance n-1 zero crossings in one period. Therefore, each pulse train can be perfectly
time-multiplexed with n-1 sinc-shaped pulse trains, if other pulses are placed at zero crossings. To
obtain this, each consecutive is relatively delayed by a zero-crossing spacing. This specific delay
is obtained in phase/delay adjustment block. The schematic of Fig. 23 shows the situation when
three comb lines are selected at each port of the input LCoS filter. Using just one MZM in the
block of optical Nyquist pulse generation, each comb line is converted to Nyquist sinc-shaped
pulse trains with two zeroes in one period which are 1/3 of period apart. The three sinc pulse trains
are proportionally delayed in the next block to generate the TDM-like waveform in time. Since
49
each pulse trains at the output of delay adjustment block occupies different wavelength bands,
nonlinear mixing inside the PPLN is employed to convert all of them to one single wavelength.
Inside the PPLN, sinc pulse trains are coherently mixed with a comb lines coming from port 2 of
LCoS filter and another CW pump. Through the cascaded processes of sum frequency generation
(SFG) followed by difference frequency generation (DFG), all channels are wavelength-converted
to a wavelength which is symmetrical with respect to the CW pump wavelength. As the result, the
TDM waveform of sinc pulse trains will occupy a bandwidth exactly equal to the pulse train
bandwidth.
Figure 5.7 Concept of TDM of Nyquist channels via Kerr comb and a PPLN waveguide. PPLN: Periodically-
Poled-Lithium-Niobate, SFG: Sum Frequency Generation, DFG: Difference Frequency Generation, TDM: Time
Division Multiplexing, MZM: Mach-Zehnder Modulator, RF: Radio Frequency
5.6 Experimental setup for TDM of Nyquist Channels
50
The experimental setup is shown in Fig. 5.8. A CW light is amplified to 4.2W using a high
power EDFA and coupled into a silicon nitride microresonator with the FSR of 192.0020 GHz.
The low phase-noise Kerr comb state is obtained when the red-detuned pump wavelength [95].
Figure 5.8 . Experimental setup; PC: Polarization controller, PS: Phase shifter. Spectrum at points (a), (b) and (c)
are shown. AFG: arbitrary function generator; ECDL: external cavity diode laser; ASE: amplified spontaneous
emission; Att: attenuator; BPF: bandpass filter; PC: polarization controller; EDFA: erbium-doped fiber amplifier;
TOF: tunable optical filter; LF: lensed fiber; FBG: fibber Bragg grating; Res.: resolution; OSA: optical spectrum
analyser.
An even number of comb lines are selected by a two-port LCoS filter. Half of the lines are
directed to one port and the other half emerges at the other output port. At port 1, the comb lines
enter the Nyquist generation block which contains two MZMs driven at RF frequencies of f1 and
f2. For the generation of Nyquist pulse trains with just three lines in spectrum the first MZM is
easily bypassed. The generated Nyquist pulse trains are then directed to an intensity modulator to
be OOK modulated with a pseudo-random bit pattern (PRBS) at f2 bits/s rate.
1 nm
PC
EDFA
LCoS
filter
Microresonator
MZM
12 nm
DCF
12 nm 12 nm
MZM MZM
Bias
Bias
Synchronized
f1
f2
PS
OSC
OSA
(a)
LCoS
filter
Comb generation
Port-1
Bit pattern
generator
Port-2
PPLN
OSC
Offline
Processing
Optical Nyquist TDM
Port 2 of LCoS
Port 1 of LCoS
1500 1550 1600
-60
-20
20
1470 1625
P
o
w
e
r
(
4
0
d
B
/
d
i
v
)
1542 1544 1546 1548
P
o
w
e
r
(
2
0
d
B
/
d
i
v
)
Wavelength(nm)
1546 1550 1554
Wavelength(nm)
P
o
w
e
r
(
1
0
d
B
/
d
i
v
)
1551 1553 1555
Wavelength(nm)
(a)
(b)
(b)
(c)
(c)
51
The modulated Nyquist pulses are entering a DCF followed by an SLM filter based on (LCoS)
Technology. Different Nyquist WDM channels are delayed proportionally in a DCF. The delay of
each channel is further tuned through the phase/delay adjustment in an SLM filter. The output of
SLM filter is amplified by a high power EDFA and combined with the amplified output of port 2
of LCoS filter and a CW pump at 1558 nm to enter the PPLN. The quasi-phase matching (QPM)
wavelength is temperature-tuned and stabilized at around 1550 nm to produce the highest
conversion efficiency. The wavelength converted channel at the output of PPLN is filtered and
directed to the detection unit which is composed of real time sampling and offline digital signal
processing (DSP).
5.7 Experimental results for TDM of Nyquist Channels
For proof of concept, we used OOK modulation of three time-multiplexed channels. Since only
three pulses are multiplexed the sinc pulses should have three lines in their spectrum and therefore
only one MZM is required. Six comb lines are first selected at the output of Kerr microresonator
comb (Fig. 5.8). Three lines out of these six lines are used for generation of three Nyquist sinc-
shaped pulse trains of three lines. The Nyquist pulses are OOK-modulated with a PRBS of length
2
15
-1 with a clock which is synchronized with the MZM RF drive. The choice of length of pattern
have been solely made because of memory limitations. Three Nyquist pulse trains are injected into
the PPLN along with the other three comb lines to convert the three pulse trains to a single
wavelength (Fig. 5.9). The maximum conversion efficiency is around 15 dB.
To show the efficiency of wavelength conversion, the extracted eye diagrams at the output of
the PPLN for each single channel and for the bit rate of 10 GHz which are shown in Fig. 5.10 (a).
52
Figure 5.9 Spectrum at the output of PPLN. Three OOK-modulated channels are located at 1551.1 nm. 1552.5
nm and 1553.9 nm and the converted channel is located at 1542.2 nm.
As can be seen the eyes for each channel is shifted in time by 33.3 psec which is the zero-
crossing spacing. The top envelopes of the eye diagrams reconstruct the waveform of a sinc-shaped
pulse train with almost 2% discrepancy. For another bitrate of 5 GHz the experiment is carried out
the three TDM channels are extracted from real time sampling scope data. Offline DSP has
performed on the TDM waveforms to detect the data of all three channels, which has resulted in
the zero-error detection. The bit error rate versus OSNR for the bit rate of 5 GHz is plotted in Fig.
5.10 (b). Because of pattern length the minimum detectable error is almost 3e-5. The thumbnails
in Fig. 5.10 (b) show the eye diagrams of the TDM waveforms for each OSNR measurement point.
It should be mentioned although we used OOK modulation for the proof of concept, there is no
limitation on the modulation format in the proposed scheme.
1542 1546 1550 1554 1558
-60
-40
-20
0
Wavelength(nm)
Power(dBm)
Aux Pump
WDM Nyquist
channels
Wavelength
conversion Pumps
TDM Nyquist
Channel:
CH1+Ch2+Ch3
Ch1
Ch2
Ch3
53
Figure 5.10 (a): Eye diagrams of the three OOK-modulated channels. (b) Bit error rate versus optical signal to
noise ratio of each OOK modulated channels.
5.8 Conclusion
To conclude, we experimentally demonstrate the (1) optical generation and (2) time
multiplexing of Nyquist sinc-shaped pulses from a single microresonator-based Kerr frequency
comb. In this method, the number of multiplexed channels can be easily changed by changing the
number of selected comb lines. Also, the bitrate of the multiplexed channels is controlled by the
RF derive of the optical Nyquist pule generation module. Using a nonlinear waveguide staged the
time multiplexed channels are wavelength-converted to a single channel by an efficiency of less
than 15 dB. Different bitrates for OOK modulated pulse trains are demonstrated. The data for each
channel of a TDM system composing three channels modulated with the OOK at 5 GHz is
recovered and the system performance is evaluated in terms of bit-error-rate (BER).
54
Also, we experimentally demonstrate the generation and time multiplexing of Nyquist sinc-
shaped pulses from a single microresonator-based Kerr frequency comb. The Nyquist sinc-shaped
pulses are constructed using amplitude modulation of Kerr comb spectral lines. Pulse trains are
delayed proportionally so peak of each pulse is placed at the zero-crossing of another pulse. Sinc-
shaped pulse trains are nonlinearly mixed inside a periodically poled lithium niobate (PPLN)
waveguide to generate the time division multiplexed (TDM) of Nyquist channels. The number of
multiplexed channels can be easily changed by changing the number of selected comb lines. Also,
different bitrates for OOK modulated pulse trains are demonstrated. The optically generated
Nyquist pulses show below 2% discrepancy with theory. Thanks to the minimum ISI property of
Nyquist sinc pulse trains the data for each channel of a TDM system is recovered and the system
performance is evaluated in terms of bit-error-rate (BER).
55
Chapter 6 Remotely Controlled and Powered Tunable
Optical 2-4 Taps Correlator of a 20-100 Gbit/s QPSK
Channel Based on Cascaded MZIs and Laser-Delivered
Control Signals
6.1 Introduction
Tapped-delay-lines (TDL) are basic building blocks for digital signal processing (DSP).
Several key functions such as equalization, correlation, and filtering can be achieved by different
TDL architectures having different values for taps’ complex coefficients and delays [95-101].
Despite the mature electronic technology for TDLs [98,101,102], there are still cases in which it
may be desirable to perform the signal processing in the optical domain, specifically: (i) to avoid
inefficient optical-to-electrical-to-optical conversion if the signal is already in the optical domain
[103,104], and (ii) if the available optical devices can operate at higher speeds than the electronic
ones [103-105].
The optical tapped-delay-line (OTDL) is a digital filter that can act on either digital or analog
signals and can be realized using different architectures [106, 107]. A possible approach to
achieving an optical TDL is to use a cascade of Mach-Zehnder interferometers (MZI), such that
each MZI contributes to a delayed and complex-weighted “tap” in the processing function [108-
121]. An optical TDL based on a cascade of MZIs can be more widely designed and fabricated
given the increase in availability of foundries that can produce high-quality photonic integrated
circuits (PIC) [108,112].
56
A key question is the dependance of signal processing performance and its degradation due to
various non-idealities of a fabricated that the signal-processing performance can degrade due to
the non-ideality of an OTDL made of a cascade of the fabricated MZIs. These non-idealities can
include: (a) the bias point of any given MZI can drift, especially with temperature, (b) the MZIs
phase shift may not cover a full 2p cycle, (c) the refractive index of each waveguide may vary as
the result of doping level non-uniformity which can produce an undesired phase shift, and (d)
residual surface roughness can produce excess losses that alter the various tapped weights
[108,113-115]. Therefore, it may be valuable for PIC designers and fabricators to have guidelines
as to how close to ideal should the various device parameters be targeted.
Importantly, optical-signal-processing nodes may be located at various locations in a network
[116]. However, many locations can be vulnerable to failure due to intentional or unintentional
degradations. One key operational challenge is the potential for power to be interrupted at the
signal-processing site [117]. If there is no local power, the target pattern as tailored by the electrical
bias voltages on the different MZIs could drift, degrade, or be non-existent [118].
A laudable goal might be to enable an architecture that allows for an optical correlator that has
no local power yet is tunably controlled (i.e., target pattern) and powered from a remote location.
One tool that can be leveraged to achieve this remotely controlled and powered correlator might
be the ability to power electrical elements from the output of a series of photodiodes that are driven
from remotely located laser beams [119].
In the first part of this chapter, we are looking for the trends and dependencies of the system
level performance of a chain of cascaded MZIs acting as an OTDL due to the MZI structural
imperfections [120]. We simulate the above non-ideal conditions for an MZI-based optical TDL
57
and find various performance dependencies. We have investigated the consequences where the
optical phase shift between the arms of MZIs does not cover a full 2p cycle. Moreover, we
considered the effect of positive and negative bias drifts of the phase-shifters on the filtering
performance of the chain of MZIs. The excess waveguide loss and refractive index offset from its
nominal value are additionally considered to mimic the effect of fabrication imperfections. The
effects of these structural imperfections are considered on both analog and digital signals quality.
These effects are quantified in terms of error vector magnitude (EVM) increase and spurious free
dynamic range (SFDR) reduction, for analog and digital signals respectively. It has been found
that a 500-mV bias drift of MZIs can reduce the filter extinction ratio by 20 dB and leave a ~2 dB
passband ripple. The imperfections can cause a 13 dB reduction in SFDR for third harmonic
distortion (THD). Moreover, the excess loss of waveguides and refractive- index-error increase
the EVM by ~ 50%, for a 20 GBaud QPSK modulated signal.
In the second part of the chapter, we experimentally demonstrate remotely controlled and
powered tunable optical 2-4 taps correlator of a 20-100 Gbits/s quadrature-phase-shift-keyed
(QPSK) channel based on laser-delivered biases. A remote correlator based on a cascade of MZIs
has been phase-controlled through down-conversion of the light power sent from a distant location
via an optical fiber link. A photodiode (PD) array has been utilized in photovoltaic mode to deal
with the the unavailability of a local power at the correlator site. When the optical power that is
sent to the link increases, the stimulated Brillouin scattering (SBS) component of backscattered
light from the fiber becomes dominant which rapidly saturates the amount of power delivered to
the correlator.
58
This insufficiency of power hampers the correlator to function on target patterns with larger
phase-shifts. This issue is further addressed by monitoring and managing the SBS and also by
adding an extra laser for carrying the optical power. In this manner, the power delivered through
the link is boosted by ~13 dB when the SBS is suppressed. Another ~ 6dB gain in the delivered
power is obtained by adding a second laser. By having the PD array sufficiently powered, the
correlator is shown to locate different target patterns at different baud-rates from 10 to 50 Gbaud.
Also, by adding an extra array of PDs, the correlator is shown to be able to identify target patterns
with a longer length of 4 symbols.
6.2 Theoretical study of cascaded MZIs as a tapped delay line
A generic tapped delay line transfer function can be represented by a finite impulse
response (FIR) filter with the time domain input/output relation as:
𝑦(𝑡) = ∑ ℎ
#
𝑥(𝑡−𝑛𝑇)
D38
#AE
(6.1)
In the above equation, N is the number of taps or filter order, hn is the complex tap
coefficients and T is the inverse of finite spectral range (FSR) of the FIR filter. Since all FIR
transfer functions that differ by a constant phase term are identical, without loss of generality we
can assume h0 is real. Eq. 6.1 can be realized by a pair of 1-by-N splitter and N-by-1 combiner and
N parallel arms encompassing phase-shifters, attenuators and delay lines. The delay of the k
th
arm
is kT. For the attenuation hk, and phase shift of fk for the k
th
arm, one should have hk = hke
jfk
. In
this case the k
th
tap is realized by the k
th
arm of the structure.
59
Moreover, the generic optical TDL of eq. 6.1 can be realized by a chain of MZIs in series
as depicted in Fig.6.1. An MZI is composed of one long and one short arm. The longer arm of each
MZI includes one delay line, a phase-shifter and an attenuator. The MZIs are connected to each
other by shared 2 by 2 directional couplers as shown in Fig. 6.1. Assuming each MZI has a similar
amount of delay, then a series of N cascaded MZIs represents an (N+1)-tapped TDL. The k
th
tap
is realized by all possible paths for light to propagate into the chain of MZIs while passing through
exactly k-1 out of N longer arms. In this case, the first tap is built up by propagating through all
the shorter arms and tap N+1 is obtained by the light passing through longer arms of all MZIs.
Complex tap coefficients are realized via phase-shifters, embedded in MZI arms and direction
couplers coupling coefficients.
The chain of MZIs as depicted in Fig. 6.1 is an example of a 4-tapped optical FIR filter
which is represented by eq. 1 with N=4. All MZIs of Fig. 6.1 have the same delay, T. This amount
of delay is obtained by placing a slab waveguide with a proper length on the longer arms of each
MZI. Different tap coefficients of eq. 1 are realized by setting the amount of phase shift of MZI
phase-shifters (f1, f2, f3), attenuation of each arm (h1,h2,h3), coupling coefficients of the
directional couplers (k1, k2 k3, k4), according to the following equations.
ℎ
E
= (1−𝑘
8
)𝑘
<
𝑘
F
(1−𝑘
G
) (6.2a)
ℎ
F
∗
= 𝑘
8
𝑘
<
𝑘
F
𝑘
G
𝜂
8
𝜂
<
𝜂
F
𝑒
; I
#
9;I
%
9;I
&
(6.2b)
ℎ
8
∗
= 𝑘
8
𝜂
8
(1−𝑘
<
)𝑘
F
(1−𝑘
G
)𝑒
; I
#
+
(1−𝑘
8
) (1−𝑘
<
) (1−𝑘
F
)(1−𝑘
G
)𝜂
<
𝑒
; I
%
+
(1−𝑘
8
) 𝑘
<
(1−𝑘
F
)𝑘
G
𝜂
F
𝑒
; I
&
(6.2c)
ℎ
<
∗
=𝑘
8
𝑘
<
(1−𝑘
F
)(1−𝑘
G
)𝜂
8
𝜂
<
𝑒
; I
#
9;I
%
+
60
(1−𝑘
8
)(1−𝑘
<
)𝑘
F
𝑘
G
𝜂
F
𝜂
<
𝑒𝑥𝑝𝑒
; I
%
9;I
&
+
𝑘
8
(1−𝑘
<
) (1−𝑘
F
)𝑘
G
𝜂
8
𝜂
F
𝑒
; I
#
9;I
&
(6.2d)
As can be seen in Fig. 6.1, a chain of three MZIs acts as a 4-tap tapped delay line through
providing four disjoint optical paths. Two Paths resulting to tap1 and tap4 are plotted in Fig. 6.1
which include all shorter and longer arms, respectively. However, the abovementioned
relationships can become perturbed if different structural non-idealities of MZIs are taken into
account. Under non-ideal conditions, the bias voltage of a phase-shifter can drift over time or
phase-shifter may have a phase coverage of less than 2p cycle. These two situations are depicted
in Fig. 6.1. The bias drift can happen at any settings such as linear regime of electro-optic phase-
shifter (around zero phase shift) or at the saturation point (around p phase shift). Moreover, the
refractive index of the slab waveguides can have an offset from the nominal value due to
fabrication imperfection or MZI arms can have excessive loss as the result of the residual surface
roughness. All these variations can affect the complex taps coefficients and delays of the OTDL
and eventually alter the filtering performance.
Figure 6.1 Concept of cascaded MZI with 3 MZI acting as a 4-tap TDL. Each MZI includes a waveguide acting as
a delay line with delay t, a phase-shifter with phase-shift f and an attenuator with attenuation h. The MZIs are
connected to each other by shared directional couplers with coupling coefficient k. Under non-ideal conditions, the
phase-shifters can drift over have a limited phase coverage. MZI: Mach-Zehnder interferometer, PS: phase-shifter.
Bias control
MZI array on chip
X(t)
y(t)
MZI Phase shifter (PS)
!
!
=Total delay
N = number of MZIs +1
!
!
"
!
#
!
$
!
"
"
#
"
$
"
"
#
#
#
$
#
MZI 1 MZI 2 MZI 3
!
"
!!
!
$
Tap 1
Tap 4
Cascaded MZI chain
X(t)
y(t)
"
"
= Attenuations
Input/output relation Transfer function
Bias (V)
Phase shift
phase
shifter bias
drift
Bias (V)
Phase shift
phase shifter
limited phase
coverage
1.5p
2p
61
A cascade of three MZIs, namely, MZI-1, MZI-2 and MZI-3 with embedded phase shifters and
4 directional couplers; k1, k2, k3 and k4, is considered. The values for delays and coupler coefficients
are as following; (t1=25 psec, t2 = 50 psec, t3 =50 psec) and (k1=0.5, k2=0.2, k3=0.2, k4=0.04). The
Vp of each MZI is equal to 2V. The MZI-3 is biased at null and hence its phase shifter (PS) is set
to shift the phase of the incoming light by p. The MZI-1 and MZI-2 are biased at 0V. The TDL
configured in this way, is an FIR filter with a flat-top transfer function and a fairly linear phase
over the passband. The free spurious range (FSR) of this filter is related to the shortest delay and
is 1/25 psec = 40 GHz. Three metrics; (1) extinction ratio (2) passband ripple and (3) phase
linearity have been considered. The extinction ratio is defined as the difference between power
levels at the passband and out-of-band powers. The passband ripple is the power difference in the
passband and phase linearity is the largest deviation of phase in the passband region from an
interpolated linear line. The radio frequency (RF) transfer function of the cascade of 3 MZIs as
described above shows and extinction ratio of > 25 dB and a <1 dB passband ripple. The simulation
results of Fig. 6.2(a) show the effect of positive and negative voltage drifts on the amplitude of the
transfer function. Two different cases for the bias drift have been considered; a) when only the PS
of one of the MZIs (MZI-1 or MZI-3) has a bias drift and b) when all PSs within MZIs drift
simultaneously. Three metrics, (1) extinction ratio, (2) passband ripple and (3) phase linearity, are
calculated and are shown in Figs. 6.2(b-d). It is evident that, the extinction ratio shrinks abruptly
by an increment or decrement in the bias voltage drift while the ripple and phase deviation increase
more gradually. Moreover, all three metrics degrades more when MZI-1 drifts compared to when
MZI-3 has a positive or negative drift. In other words, the effect of the non-ideality in the first MZI
of the chain is more prominent than the effect of the last MZI. Figure 6.3 shows the effect of the
62
limited PS coverage for the chain of MZIs. Again, three situations in which (1) just the PS of the
first MZI has a limited phase coverage, (2) only last MZI has a PS with limited phase coverage
and (3) all PSs have limited, and identical phase coverages are considered. Based on the transfer
functions of the Fig. 6.3(a), the three filtering performance metrics are calculated and are plotted
in Fig. 6.3(b-d). As can be seen, the non-ideality in MZI-1 produces larger extinction ratio loss,
passband ripple and phase deviation than the non-ideality of other MZIs.
Figure 6.2 Transfer function (b) extinction ratio (c) passband ripple and (d) phase linearity for different phase
shifter bias drift when only MZI-1 (green), only MZI-3 (yellow) and all MZIs (red) have bias drifts.
(a)
(b)
(c) (d)
63
Figure 6.3 (a)Transfer function (b) extinction ratio (c) passband ripple and (d) phase linearity when only MZI-1
(green), only MZI-3 (yellow) and all MZIs (red) have phase shifters with limited phase coverages.
Aside from the non-ideality of the phase shifter, the waveguiding structure of the MZI arms can
be non-ideal as well. To investigate such effects, the previously mentioned MZI chain is
numerically simulated again with ideal PSs but with a loss greater than zero. Figure 6.4(a) shows
the passband ripple and phase linearity for different values of waveguide losses as large as 12
dB/cm. Again, the three situations when only the first and last MZI has an excessive loss and when
all MZIs have the same loss values are considered. The passband ripple remains below 1 dB while
the phase can deviate from linear line by up to 2 degrees. Again, as observed in Fig. 6.2, the first
MZI has a noticeably larger effect than the last MZI (MZI-3). This can be attributed to propagation
of the error in the chain of MZIs. To study the effect of this imperfection on digital signals a
quadrate-phase-shift-keying (QPSK) signal at a speed of 20 Gbaud is passed through the filter.
The center wavelength of the signal is located at the central frequency of the FIR filter. The EVM
(a)
(b)
(c)
(d)
64
of the output signal for all three situations of Fig. 6.4(a) is calculated for different waveguide
losses. The insets show the simulated eye diagrams of the in-phase component of the QPSK signal
at different loss values of the 2, 6, 10 dB/cm for the situation when all MZIs have lossy
waveguides. As can be seen the EVM increases with loss almost linearly and the effect for MZI-1
is more obvious than MZI-3. When all MZIs have lossy waveguides the largest increase in the
EVM is observed. The trend of EVM increase with waveguide loss can be interpreted as the result
of passband alteration of the filter.
Additionally, the refractive index of the waveguide can have an offset with respect to its
nominal value as a result of dopant impurities. We have incorporated this to our simulation and
have simulated our chain of three MZIs with waveguides with different refractive indices errors as
large as 6e-4. The result of passband error and phase linearity is plotted in Fig. 6.4(c), when all
MZIs have such a faulty waveguide. The passband ripple remains below 1 dB but interestingly the
phase can deviate from the linear line by up to 12 degrees which is the largest phase deviation
observed so far. The effect on the digital 20 Gbaud QPSK is plotted in Fig. 6.4(d). The EVM
generally increases as the non-ideality strengthens. However, as seen in Fig.6.4(d), when the
refractive index error is larger than 6e-4, the modulation format changes. This can be due to the
high deviation of the phase of transfer function (~10 degree) from linear trend as depicted in Fig.
6.4(c). This phase distortion can cause the tapped QPSK copies of the signal to deteriorate the
modulation format of the incoming QPSK.
65
Figure 6.4 (a) Passband ripple (blue) and phase linearity (red) for different waveguide losses when only MZI-1
(solid), only MZI-3 (dashed) and all MZIs (dotted) are lossy. (b) EVM for different waveguide losses when only
MZI-1 (grey), only MZI-3 (cyan) and all MZIs (red) are lossy. (c) Passband ripple (blue) and phase
linearity(red) for different waveguide refractive index errors. (d) EVM for different waveguide refractive index
errors when only MZI-1 (grey), only MZI-3 (cyan) and all MZIs (red) have refractive index error.
Finally, to investigate the extent to which the analog signals can be affected by the non-ideality
in the structure of MZIs in the presence of material nonlinearity, a nonlinear coefficient of 9e-18
m
2
/W and a two-photon absorption of 6e-11 is considered for all MZIs of the chain. An input
microwave tone at 20 GHz is launched to the MZI chains with a varying power levels ranging
from -120 dBm to 40 dBm. The power of this tone and its generated third harmonic is numerically
obtained through simulation and are plotted in Fig. 6.5(a). The SFDR can be obtained as the
difference of these powers at 1-dB noise floor. Three situations for ideal MZI, MZI-1 with lossy
waveguide and MZI-1 with a PS with bias drift are shown for example. The SFDRs obtained
through different curves similar to Fig. 6.5(a) are calculated for different scenarios. These results
(a) (b)
(c)
(d)
66
are show in Fig. 6.5(b). Overall, the MZI chain imperfections can cause reduction of SFDR which
means the distortion of microwave signals while the waveguide structure is nonlinear.
However, different imperfection would decrease the SFDR to different extent. As it can be seen
from Fig. 6.5(b), a bias drift of <500mV for MZI-3 would cause almost no SFDR change. This is
also true when the phase shifters of all MZIs will have a phase coverage of less 2p. It should be
mentioned that Fig. 2 and Fig. 3 also show that bias drifts of MZI-3 and limited phase coverage of
all MZIs are not very effective in altering the pass-band ripple and phase linearity criteria. On the
other hand, when MZI-3 has a loss larger than 12 dB/cm, a reduction of SFDR almost to 3dB is
obtained, while an error in the refractive index as large as of 3e-4 reduces the SFDR by an amount
of ~5 dB. These imperfections clearly have reduced the SFDR by an amount that it can noticeably
reduce the amount of microwave peak power in the presence of nonlinearity. Moreover, when the
MZI-1 has a voltage drift of more than >300 mV, a refractive index error of less than 5e-4 or an
excessive loss of <12 dB/cm, the SFDR can be reduced even further by 8dB-10dB. As the result
the dynamic ranges for a microwave tone for these imperfections are limited. It should be noted
all changes in this regime involves the imperfection of the first MZI of the chain, that is, MZI-1.
It should be mentioned that it was also observed in Fig. 2 that the imperfection of the first MZI is
more crucial than the imperfection of other MZIs. The largest reduction in the SFDR which is
around 13 dB happens for the MZI-1 having a voltage drift of more than 300 MV or when all MZIs
have the similar drift of > 300 mV. This shows the most important imperfection regarding the
reduction of the SFDR is the voltage bias drift. Interestingly it has been found that the effect of the
drift in the first MZI is almost similar to the effect of all MZIs having the same amount of voltage
drift. Overall, observations such as Fig. 6.5(b) shows that the effect of first MZI is the chain is
67
larger in terms of the reduction of the SFDR and bias drift can be the primary reason for large
SFDR reduction.
Figure 6.5 (a) Fundamental harmonic versus input power for an incoming microwave signal for when all MZIs
are idea (green circles), when MZI-1 has an attenuation of 10 dB/m (dashed green), when MZI-1 has bias drift of 300
mV (dotted-dashed light green). Third harmonic power versus input power for when all MZIs are idea (red circles),
when MZI-1 has an attenuation of 10 dB/m (dashed red), when MZI-1 has bias drift of 300 mV (dotted-dashed light
red). (b) SFDR for different scenarios.
6.3 Concept of remote correlator based on cascaded MZIs
The conceptual block diagram of a remotely controlled tunable optical correlator is shown in
Fig. 6.6. A transmitter sends a QPSK-modulated data channel along with multiple continuous wave
(cw) laser wavelengths to an optical link in order to deliver power to a remote correlator with no
access to local power. At the correlator site, the power of lasers is captured through arrays of PDs
operating in photovoltaic mode. The powered PDs provide different bias values for the phase-
shifters of a correlator in order to enable it to recognize different target patterns within the
incoming QPSK signal. The amount of power delivered via the link can specify the target pattern
of interest. The backscattering is measured and is minimized through a power management unit in
Noise floor 1 HZ
Third harmonic
Fundamental
harmonic
SFDR
(a)
<300 mV drift for MZI-1
Ideal
<600 mV drift for MZI-3
MZI-3 loss <12dB/cm
All PSs not covering "#
>300 mV drift for all MZIs
MZI-1 ref. index err.>3e-4
MZI-1 ref. index err.<5e-4
MZI-1 ref. index err. <2e-4
>300 mV drift for MZI-1
All MZIs loss>6 dB/cm
MZI-1 loss~12 dB/cm
93.88
93.84
93.78
89.11
88.34
85.26
84.68
83.76
83.21
80.98
80.46
90.27
THD SFDR (dB.Hz
!/#
)
(b)
Different scenarios
1 2 3 4 5 6 7 8 9 10 11 12
(b)
<300 mV drift for MZI-1
Ideal
<600 mV drift for MZI-3
MZI-3 loss <12dB/cm
All PSs not covering "#
>300 mV drift for all MZIs
MZI-1 ref. index err.>3e-4
MZI-1 ref. index err.<5e-4
MZI-1 ref. index err. <2e-4
>300 mV drift for MZI-1
All MZIs loss>6 dB/cm
MZI-1 loss~12 dB/cm
93.9
93.8
93.8
89.1
88.3
85.3
84.7
83.8
83.2
81
80.5
90.2
THD SFDR (dB.Hz
!/#
) Different scenarios
1 2 3 4 5 6 7 8 9 10 11 12
68
the transmitter. Based on the phase-shift of the taps of the correlator, different laser sources may
be switched on or off at the transmitter.
Figure 6.6 (a) Concept of a remotely controlled tunable optical correlator (b) Basics of the 2 and 4-taps QPSK correlator
using single MZI and two cascaded MZIs.
The correlator is composed of MZI(s) and embedded phase-shifter(s). An MZI with an
embedded phase-shifter is equivalent to a 2-Tap tapped-delay line which coherently mixes the
delayed copies of the incoming QPSK signal with complex coefficient. By considering that a single
MZI has a delay t equal to an integer factor of a symbol time and the MZI phase-shift is p/2 or p,
then a 2-taps correlator for QPSK signal results in a 9-QAM signal as schematically plotted in Fig.
6.6(b). When the phase shift of the MZI (taps) matches the phases of two consecutive symbols, a
peak at the output is observed which corresponds to the constellation points at the top right corner.
Likewise, by adding another MZI, a 4-tap correlator is obtained since signal can traverse in four
different paths as shown in Fig.6.6(b). This results to a 25-QAM signals at the output.
6.4 Experimental Setup and Results
λ
1
λ
n
Data
Combiner
Splitter
Link Backscattering
Power
management
control
λ
s
Target
pattern
Optical Link
Transmitter
Remote Terminal
λ
1,
λ
2,…,
λ
n
Photo-voltaic
OE conversion
λ
2
…
…..
Optical
correlator/
N-tap tapped
delay line
QPSK
(𝑵+𝟏)
𝟐
QAM
Optical signal processing
node
Target pattern
phase
conjugate
(a)
λ
s
(b)
𝝋
𝟏
𝜏
MZI 1
Tap 1
Tap 2
1
st
delayed QPSK QPSK
𝐀
∗
.𝐒
𝐢𝐧
𝐭−t + 𝐁
∗
.𝐒
𝐢𝐧
𝒕 = 𝐒
𝐨𝐮𝐭
𝐭
j×[Pattern]
-1×[Pattern]
-
j×[Pattern]
[Pattern]
Tap 4
Tap 2
𝝋
𝟏
𝜏
MZI 1 MZI 2
Tap 1
𝜏 𝜏
Tap 3
𝝋
𝟐
3rd delayed 1st delayed 2
nd
delayed
𝐀
∗
.𝐒
𝐢𝐧
𝐭−𝟑t ++𝐃
∗
.𝐒
𝐢𝐧
𝐭−𝟐t + 𝐁
∗
.𝐒
𝐢𝐧
𝐭−t + 𝐂
∗
.𝐒
𝐢𝐧
𝐭 = 𝐒
𝐨𝐮𝐭
𝐭
QPSK
2-Tap MZI-based correlator 4-Tap-MZI based correlator
Target
Pattern:
[AB]
Target
Pattern:
[ADBC]
69
The experimental setup is shown in Fig. 6.7. A laser at ls = 1549.5 nm is modulated with 10,
20, 40 and 50 Gbaud data streams and sent to a 7.8 Km long single mode fiber (SMF). The lasers
at l1=1544.5 nm and l2 =1546.1 nm are phase modulated by a 2
31
-1 pseudo-random bit sequence
(PRBS), amplified by a tunable optical amplifier and are sent to the SMF. A circulator at the input
of the SMF captures the backscattered light. The backscattered power is monitored in order to tune
the RF drive of the phase modulator to minimize this power. At the output of fiber, the multiplexed
wavelengths are separated. The data channel injects to a series cascade of MZIs; MZI-1 with a free
spectral range (FSR) of 10 GHz and MZI-2 with an FSR of ~5 GHz. The waves at l1 and l2 drive
two arrays of 15 PDs in series and operating in photovoltaic mode. The voltage and current built-
up by the PDs drive the MZIs. The voltage-current versus the input optical power for one the PD
arrays is plotted in Fig. 6.8 (a) for two different resistor loadings.
Figure 6.7 Experimental setup and power spectra at points A and B.
SMF
7.8 km
l
s
QPSK
1 nm
Dual-drive
MZM
OSA
50/50
99/1
50/50
OSA
Amplitude
adjustment
EDFA
5 nm
Optical Power
2^31-1 PRBS
(900 MHz)
Phase
modulator
30dB
50/50
EDFA
PC
PC
A
B
5 nm
1 nm
VR VR
Rx
MZI-1 (10 GHZ)
B
MZI-2 (5 GHZ)
A
l
2
l
1
Photo-
voltaic
mode
l
s
l
1
l
2
l
1
l
2
Wavelength (nm)
Power
Power
70
Figure 6.8 (a) PD array voltage/current vs optical power (b) Backscattered power suppression measured at the
input of SMF vs phase modulator RF drive. Backscattered power spectrum with and without phase modulation
for input link power of 17 dBm (c), 20.5 dBm (c1) and 26.5 dBm (c2).
The backscattered light from the cw power wavelengths can limit the amount of the power
delivered to the PD arrays despite increasing the input power (Fig.6.9 (b)). As shown in the
captured spectra of Fig. 6.8 (c), the backscattered power has three basic components, that is; linear
Rayleigh and nonlinear Brillouin stokes and anti-stokes. While the linear Rayleigh component
dominates the most of lower power regime, at higher power values, the Stokes component of SBS
become dominant surpassing the Rayleigh level at some point. The phase modulation of laser light
at λ1 and l2 with a proper pulse amplitude can noticeably suppress this component. Figure 6.8(b)
shows the amount of total backscattered power suppression with respect to the phase modulator
PRBS pulse amplitude. As can be seen in Fig. 6.9 (a) and (b), phase modulation with a tuned RF
drive can suppress the Stokes component of SBS and therefore increase the delivered power by 13
(b)
Backscattered Power
Suppression (dB)
(c)
Anti-Stokes
(Spontaneous BS)
Stokes BS
Rayleigh
Rayleigh level
~ 30 dB Stimulated BS
W/phase modulation
W/O phase modulation
(d)
(a)
(b)
(c)
(c1)
(c2)
71
dB. Further increasing of the optical power, saturates the output power again at an increased level
of 21 dBm (Fig.6.9 (b)). This corresponds to the point that the Stokes component reaches the
Rayleigh level which according to Fig.6.9 (a) occurs at an input power of ~ 26 dBm. As shown in
Fig.6.9 (b), by adding another phase modulated laser, a total of 6 dB boost in delivered power is
observed. Figure 6.9 (c) shows different power intervals within which the MZI can be derived by
one or two lasers and with or without phase-modulation to obtain a specific amount of phase-shift.
The margins of these intervals correspond to the power at which the Stokes component of reaches
the Rayleigh level. As evident from Fig.6.9 (c), a total power of ~18.5 dBm from a single laser
source is required for the PD arrays to generate a phase shift of p/2 on the MZI arm. A total power
of ~ 23 dBm using two lasers is required to generate a phase-shift of p.
72
Figure 6.9 (a) Measured Brillouin backscattered power vs link input power (b) Measured link output power vs
link input power(c) Phase-shift of the MZI vs link output power.
Figure 6.10 shows the output constellation diagrams of the 2 and 4-taps correlator with input
QPSK signals. The results are obtained for different baud-rates and target patterns using one or
two MZIs. For the 4-Tap correlator, target pattern of [p/4, 3p/4, p/4, 3p/4] (corresponds to a phase-
shift of p/2 for MZI-1 and 0 for MZI-2) is correlated with a 10-Gbaud QPSK signal. The
corresponding symbol series and the location of the target pattern is shown in Fig. 6.10. The input
power for the PD array is ~ 17 dBm which produces a voltage of 0.8 V and a current of 0.15 mA.
For 20 Gbaud, the two MZIs still function as a 4-tap correlator, however there will be a symbol
Dual power
wavelengths:
w/phase modulation
Single power wavelengths:
w/phase modulation
Single/dual
power
wavelengths:
w/o phase
modulation
Rayleigh Level
Single power
wavelength
w/phase modulation
Dual power
wavelengths:
w/phase
modulation
Single/dual power waves:
w/o phase modulation
(b)
(a)
(c)
Power boost
caused by phase
modulation
Power boost by adding
the second laser
13 dB
6 dB
73
neglected between each target pattern symbols. This is also true for the results of the 2-tap
correlator using a single MZI (FSR = 10 GHz) and QPSKs at 40 and 50 Gbaud. Here, the correlator
looks for target patterns of [p/4, 3p/4] and [p/4, 5p/4] where three or four symbols are neglected
for the 40 and 50 Gbaud systems, respectively. For the pattern [p/4, 3p/4] a single laser is used
which produces a voltage of 1.1 V and a current of 0.18 mA. For [p/4, 5p/4] two lasers are used
to deliver a power of 25.5 dBm and generate the voltage and current equal to 1.8V and 0.23 mA.
Figure 6.10 Input and output constellation diagrams and target patterns for the 2-Tap (a) and 4-Tap (b) correlator
and different baud-rates.
6.5 Conclusion
We simulated the OTDL operated under the non-ideal condition and find various performance
dependencies. For a generic tapped delay line composed of parallel taps it has been found that
7p/4
5p/4
3p/4
p/4
0
QPSK phase
630 640 650 670 680
Symbol
4-Tap correlation; Target pattern =[p/4, 3p/4, p/4,3p/4 ] input
10 Gbaud 20 Gbaud
EVM=13.5
EVM=7.8 Eye-I
Eye-Q
input
4-Tap correlation; Target pattern =[p/4,×,3p/4,×,p/4,×,3p/4 ]
EVM=14
EVM=8.5
Eye-I
Eye-Q
40 50 60 70 80
Symbol
7p/4
5p/4
3p/4
p/4
0
QPSK phase
Target
Pattern
Target
Pattern
2-Tap correlation Input
40 Gbaud 50 Gbaud
EVM=15.4%
Input
Target pattern=[p/4,×,×,×,×,3p/4]
EVM=18.5
EVM=15%
Target pattern=[p/4,×,×,×,3p/4]
Target pattern =[p/4,×,×,×,5p/4]
Target pattern =[p/4,×,×,×, ×,5p/4]
EVM=17.1%
7p/4
5p/4
3p/4
p/4
0
QPSK phase
340 350 360 370 380
Symbol
Target
EVM=16.6%
7p/4
5p/4
3p/4
p/4
0
QPSK phase
320 350 360 370 360
Symbol
Target
EVM=14.9%
7p/4
5p/4
3p/4
p/4
QPSK phase
8700 880 890 900 910
Symbol
7p/4
5p/4
3p/4
p/4
QPSK phase
8700 880 890 900 910
Symbol
Target
Target
(a)
(b)
74
removing taps destroys the transfer function drastically. For the MZI-based TDL, it has been found
that a 500-mV bias drift of MZIs can reduce the filter extinction ratio by 20 dB and leave a ~2 dB
passband ripple. The imperfections can cause a 13 dB reduction in SFDR for THD. Also, the
excessive loss of waveguides and refractive- index-error increase the EVM, at most by ~ 50%, for
a 20 GBaud QPSK modulated signal. A modulation format change for a refractive index change
of 6e-4 has been observed. Also, it has been found that a bias drift results to a phase shift swing
around 0 degree, and it is more effective than the one around p phase shift. We have observed that
a non-ideality in the first MZI marks a bigger performance degradation than the other MZIs of the
chain.
We have also experimentally demonstrated an enabling architecture for remotely controlled
optical correlators based on both linear and nonlinear elements. The correlators are places ~7.8
Km apart from the transmitter and have no access to the local optical power. Firstly, a remote
linear correlator based on a cascade of MZIs has been phase-controlled through down-conversion
of the light power sent from a distant location via an optical fiber link. The issue of power loss
because of link nonlinear effects is further addressed by monitoring and managing the
backscattered power. In this manner, the power delivered through the link is boosted by ~13 dB.
Another ~ 6dB gain in the delivered power is obtained by adding a second laser. The linear
correlator is shown to be able to locate different target patterns within incoming QPSK signals at
different baud-rates from 10 to 50 Gbaud.
75
Chapter 7 Remotely Controlled and Monitored Tunable
Nonlinear Optical Correlator Based on Temperature-
controlled Nonlinear Wave Mixing
7.1 Introduction
Optical signal processing has the potential for operation at the line rate and avoiding inefficient
optical-to-electrical-to-optical conversion [121,122]. A key building block of digital signal
processing is a tapped-delay-line (TDL) [121]. Such TDLs have been demonstrated in the optical
domain using “linear” components (e.g., Mach-Zehnder interferometers) and “nonlinear”
waveguides [122-125]. One possible advantage of the nonlinear approach is the potential for more
readily utilizing the wavelength domain [121].
Specifically, optical TDLs (OTDLs) can be used to perform optical correlation, such that a
high-speed data stream is compared to a target pattern and producing a correlation output; if the
output exceeds a threshold, there is a pattern “match” [125]. Importantly, the data and target pattern
can be encoded in both the amplitude and phase, increasing the number of bits per symbol that can
be compared [125]. An OTDL using a periodically-poled-lithium-niobate (PPLN) waveguide has
been used to provide the above- mentioned function. A quadrature-phase-shift-keying (QPSK)
data signal is multicasted into several copies representing the number of taps, each copy is
differentially delayed and given a specific complex weight, and finally multiplexed into a single
correlated output wavelength using a pump wave [125]. However, correlation nodes may be
located at various locations in a network at a distance from the transmitter and without the access
to optical sources [126]. In this case, the signal copies and pump wavelengths required for mixing
76
are sent from a distant location through an optical fiber link. This can be challenging due to
nonlinear effects of the fiber link. Moreover, ensuring proper mixing, tunability and monitoring of
operation can also be difficult.
In this chapter, we experimentally demonstrate a tunable optical correlator for a 10-15 Gbaud
QPSK data signal using nonlinear wave mixing at a remotely controlled node. A high-power pump
is phase-modulated in order to overcome the link backscattering effect and sent along with the
signal copies to a remote correlator node which is at ~7.8 km far from the transmitter. The waves
generated at the output of PPLN are sent back to the transmitter for the monitoring of operation.
The power of the pump is set accordingly. We have shown a power boost of more than 3 dB for
the correlated signal at a temperature drift of 2°C. After remote control and monitoring, improved
constellation diagrams with lower error vector magnitudes (EVMs) for different buadrates and
target patterns in a temperature drift range of <2°C are obtained. The link backscattering is also
mitigated by ~ 7dB.
7.2 Concept
The concept of remote control and monitoring of a QPSK correlator is plotted in Fig. 7.1. We
have used a PPLN waveguide as an N-tap correlator node, in which a correlation signal can be
obtained by a sum-frequency-generation (SFG) and difference- frequency-generation (DFG)
between N signal copies at ls1,..,lsn, N comb lines at lc1,..,lcn and a high-power pump at lp. When
the PPLN waveguide is thermally controlled, a correlated signal appears at the idler wavelength
of pump wavelength with respect to the quasi-phase matching (QPM) wavelength of the PPLN
[125]. Signal copies are obtained by modulating N comb lines and the target pattern is imprinted
77
on another N comb lines using a wave-shaper. Signal copies, comb lines and the high-power pump
are sent via an optical link which is ~7.8 km long.
As a result, these waves are prone to the power loss because of the link Brillouin scattering
(BS) which eventually affects the efficiency of nonlinear interactions inside PPLN. This effect and
can be controlled at the transmitter by phase-modulating of high-power pump at the speed of Df.
The temperature drift of the PPLN, can shift the QPM wavelength and reduce the power of the
correlated signal leading to difficulties in identifying the target pattern. This effect can be
monitored and controlled at the transmitter side by measuring the power of generated sidebands
(SBs) at lMSB1 and lMSB2 and then accordingly powering-up the pump. The two SBs at lMSB1 and
lMSB2 are generated when there is a blue or red shift of the QPM wavelength, respectively. The
two SBs along with the converted signal at lo are sent back to the transmitter for the monitoring
and controlling of the temperature drift effect.
Figure 7.1 (a) Concept of a QPSK correlator using the nonlinear wave mixing (a). Concept of a remotely/controlled
and monitored nonlinear QPSK modulator (b). Correlated signal generation inside a PPLN for an adjusted QPM and
for the blue/red shift of QPM wavelength (c). TEC: Thermo-electric controller.
Link~7.8 km
PPLN
90%
10%
Remote Node
Filtering
Nonlinear correlator
!
!"#$
, !
!"#%
, !
&
Target pattern [A,B,..]
Transmitter
Tunable
filtering
!
!
!
"#, ..
!
"&
!
'#, ..
!
'&
Signal copies
1. Phase
modulation at
(Df)
Pump power management
2. Monitoring
output/
sidebands
Frequency comb
Wave-shaper
delay
/amplitude/
phase
λ
Target pattern
Signal copies
… …
Df
P
!
!
!
"#, …
!
"&
!
'#, ………..
!
'&
Pump
MUX
!
!"#$
!
!"#%
!
&
Thermo-electric
controller (TEC)
SB1 SB2
!
!"#$
!
!"#% !
&
Correlation
wave
!
&
Nonlinear
wave mixing
Target
Pattern
=[AB]
["
∗
#
∗
]
"- QAM
1
st
QPSK 2
nd
QPSK
#
∗
.%
*+
&−( + +
∗
.%
*+
, =%
,-.
&
j×[Pattern]
-1×[Pattern]
-
j×[Pattern
]
[Pattern]
A B
C D
I
Q
Multicasting
/ delay
1
st
QPSK
2
nd
QPSK
Nonlinear QPSK correlator
Threshold
Matched
Not
Matched
t
(a)
Correlation
Output
%
,-.
&
%
*+
,
(b)
λ
SFG
DFG
… …
Blue
Shift SB1
λ
SFG
DFG
… …
Red
shift
SB2
(c1)
(c2)
λ
Correlation
Output
Pump Target pattern (comb lines) Delayed Signal copies
SFG
DFG
… …
Df QPM
P
78
7.3 Experimental Setup
Figure 7.2 shows the experimental setup of the remote nonlinear QPSK correlator. A mode-
locked laser with 10GHz repetition rate and 2ps pulse width generates a coherent frequency comb
with 10-GHz frequency spacing. The optical short pulse is passed through a delay line
interferometer (DLI) with free spectral range (FSR) of 20-GHz to increase the frequency spacing
of optical comb tp 20-GHz. A flat and broad spectrum is generated when the comb is passed
through a highly nonlinear fiber (HNLF). A programmable filter based on Liquid Crystal on
Silicon (LCoS-1) is used to select and write complex weights on comb fingers and separate them
into signal path and pump path. For each path, two comb fingers with a spacing of 100 GHz are
selected. After pre-amplification, a nested Mach-Zehnder modulator generates the 20- Gb/s optical
QPSK data on comb fingers. Then the signals are passed through another programmable LCoS
filter to fine-tune the delay on signals and balance their relative amplitude and phase. All the
signals are amplified in an Erbium-doped fiber amplifier (EDFA) and are combined with an
auxiliary pump at lp=1555.1 nm before being sent through a ~480 m DCF to introduce one
symbol-time relative delay between two adjacent signals. A single mode fiber (SMF) with a length
of ~ 7 km is served then as the optical fiber. All the signals and pumps are sent through a 4-cm-
long PPLN waveguide to create the correlator output signal with the spectrum shown in the figure.
The QPM wavelength of the PPLN waveguide is temperature tuned to the wavelength of ~1549.5
nm at 43.8°C, which is at the center of the tapped signals and pump fingers frequency.
A thermoelectric controller (TEC) sets this fixed temperature. The output signal is then filtered
and sent to the coherent receiver to be analyzed. A 10% of this output is amplified and then directed
79
to the transmitter side through two circulators at the input and output of the link for the monitoring
of the effect of temperature effect.
Figure 7.2 Experimental setup. MLL: Mode Lock Laser, DLI: Delay Line Interferometer, HNLF: Highly Nonlinear
Fiber, BPF: Band Pass Filter, PC: Polarization Controller, EDFA: Erbium Doped Fiber Amplifier, DCF:
Dispersion Compensating Fiber, PPLN: : Periodically Poled Lithium Niobate.
7.3 Experimental Results
Figure 7.3 shows the effect of phase modulation on the Brillouin scattering (BS) suppression
for the pump and the comb lines. The two main components of fiber backscattering; linear
Rayleigh and nonlinear BS are plotted. It is desirable to keep the nonlinear BS well below the
Rayleigh scattering. It has been found that a phase modulation of the pump with a speed of 500
MHz, allows a pump power as large as 22 dBm to be delivered to the remote node. The comb lines
used for the target pattern show an almost > 6 dB margin with the Rayleigh limit thanks to the
coherency of the lines which keeps the BS inefficient. Figure 7.4 shows the spectra at the PPLN
QPSK
MLL
DLI
HNLF
LCoS
Filter-
1
Coherent
Receiver
LCoS
Filter-
2
PPLN
Flat 20GHz Comb Generation
SMF+DCF
450 m
BPF
PC EDFA
15 nm 13 nm
27
dBm
15 nm
13 nm
1 nm
!
!
~ 7 km
1 nm
1 nm
TEC
Power
monitoring
80
output for a two-taps correlator at different temperatures, which include the correlated signal along
with the two generated SBs (SB1 and SB2). The QPM wavelength is ~1549.5 nm at the reference
temperature of 43.8°C. Both positive and negative temperature drifts drop the correlated signal
power. However, the former is responsible for SB2 generation while the latter generates the SB1.
To compensate for the correlation signal power drop, the pump power at the transmitter side is
amplified to deliver ~ 17 dBm for a 2°C drift. This boosts the conversion by around 3 dB according
to Fig.7.5 for two QPSK signals at 10/15 Gbaud.
Figure 7.3 Rayleigh and Brillouin scattering from the link vs power delivered by the pump and comb lines.
In Fig.7.6(a) and (b) the monitored power of the correlation signal, SB1 and SB2 versus the
temperature drift before and after tuning the pump power for compensating the correlation power
loss are plotted for two different baudrates. As can be seen, a blue QPM shift (negative drift)
increases the SB1 while a red QPM shift (positive drift) increases the SB2. In both directions the
correlation signal power drops. The pump power amplification at the transmitter brings back the
correlated signal power to its original value, however, it strengthens the SB1 and SB2 as well. This
can affect the quality of the digital signal as can be seen in Fig. 7.7(a) and (b) in which the EVMs
10 12 14 16 18 20 22 24
Power delivered by the link (dBm)
Link backscatterd power
(5 dB/div)
Rayleigh; cw pump
Brillouin; cw pump
Rayleigh; comb lines
Brillouin;comb lines
Rayleigh; mod. pump
Brillouin;mod. pump
81
are plotted for QPSKs at 15 Gbaud and 10 Gbaud respectively. While the EVM has been generally
decreased after power compensation, the amount of EVM improvement starts to diminish at larger
temperature drifts due to the strengthening of the sidebands by an amplified pump power. The
constellation diagrams of the generated 9-QAM at the output of a 2-taps correlator for different
temperatures, baudrates and target patterns are shown in Fig. 7.7(c) which indicates the
improvement in EVM and successful identification of target pattern at the right corner symbol of
the 9-QAM after the monitoring.
Figure 7.4 Output spectra of the PPLN for different temperature settings.
SB1
SB2
SB2
SB1
1544 1546 1548 1550 1552 1554
Wavelength (nm)
-70
-60
-50
-40
-30
-20
Power (dBm)
Temp. drift = -0.8°
Temp. drift = 0.8°
Temp. drift = 0°
Pump Comb
line1
Comb
line2
Correlated
signal
Signal
copies
Correlated
signal
82
Figure 7.5 Power delivered to the node and power gain after monitoring and tuning the pump power vs temperature
drift for a QPSK at 10 Gbaud (a) and 15 Gbaud (b).
Figure 7.6 Monitored powers vs temperature drift before and after pump power amplification for a QPSK at 15
GBaud (a) and 10 GBaud (b).
-2 -1 0 1 2
13
14
15
16
17
Power delivered to the
node(dBm)
0
1
2
3
4
Correlated power gain
(dB)
-2 -1 0 1 2
Temperature drift (°C)
13
14
15
16
17
18
Power delivered to the
node (dBm)
-1
0
1
2
3
4
Correlated power gain
(dB)
(a)
(b)
-2 -1 0 1 2
Temperature drift (°C)
-2 -1 0 1 2
Temperature drift (°C)
Power (5 dB/div)
Correlated signal (w/monitoring)
Correlated signal(w/o monitoring)
SB1 (w/o monitoring)
SB2 (w/o monitoing)
SB1 (w/monitoing)
SB2 (w/monitoing)
-2 -1 0 1 2
Temperature drift (°C)
Power (5 dB/div)
Red QPM shift
Blue QPM shi0
Blue QPM shift Red QPM shift
10Gbd
15Gbd
(a)
(b)
83
Figure 7.7 EVM vs temperature drift before and after pump power amplification for a QPSK at 15 GBaud (a) and
10 GBaud (b). Constellation diagrams of the correlated signals (c).
7.3 Conclusion
We have experimentally demonstrated a tunable optical nonlinear correlator for a 10-15 Gbaud
QPSK data signal using temperature-controlled nonlinear wave mixing at a remote node. A high-
power pump is phase-modulated in order to overcome the link backscattering effect and sent along
with the signal copies to a remote correlator node a through optical fiber link. The waves generated
at the output of correlator are sent back to the transmitter for the monitoring of operation. We have
shown a power boost of more than 3 dB for the correlated signal at a temperature drift of 2°C.
After remote control and monitoring, improved constellation diagrams with lower error vector
magnitudes (EVMs) for a temperature drift range of < 2°C are obtained. The link backscattering
is also mitigated by ~ 7dB.
-2 -1 0 1 2
Temperature drift (°C)
16
18
20
22
EVM (%)
w/o power monitoring
w/ power monitoring
-2 -1 0 1 2
Temperature drift (°C)
14
16
18
20
EVM (%)
w/o power monitoring
w/ power monitoring
Target pattern
= [p/4, 3p/4]
10 Gbaud 12.5 Gbaud
EVM=14.8%
T= 43.2°C
EVM=14.5%
w/o monitoring w/ monitoring
EVM=16.9%
EVM=14.8%
w/o monitoring w/ monitoring
T= 44.9°C
EVM=18% EVM=16.2%
w/o monitoring w/ monitoring
Target pattern
= [p/4, 5p/4]
T= 44.2 °C T= 45.2 °C T= 42.7 °C
EVM= 16.3 % EVM=16.1%
w/o monitoring w/ monitoring
EVM=17.3% EVM=16.6%
w/o monitoring w/ monitoring
EVM= 19.2 % EVM=18.1%
w/o monitoring w/ monitoring
T= 44.2 °C T= 45.6 °C
T= 42.7 °C
EVM= 16.6 % EVM=16.2%
w/o monitoring w/ monitoring
EVM=19%
EVM=17.3%
w/o monitoring w/ monitoring
EVM= 23 % EVM=22.14%
w/o monitoring w/ monitoring
15 Gbaud
T= 44.7°C
Target pattern
=[ p/4, p/4]
10Gbd
15Gbd
(a)
(b)
(c)
84
References
1. A. E. Willner, S. Khaleghi, M. R. Chitgarha, and O. F. Yilmaz, "All-optical signal
processing," J. Light. Technol. 32, 660–680 (2014).
2. G. P. (Govind P. . Agrawal, Nonlinear Fiber Optics (Elsevier Science, 2013).
3. G. Li, "Recent advances in coherent optical communication," Adv. Opt. Photonics 1, 279
(2009).
4. P. Winzer, "Beyond 100G ethernet," IEEE Commun. Mag. 48, 26–30 (2010).
5. E. Ip, A. P. T. Lau, D. J. F. Barros, and J. M. Kahn, "Coherent detection in optical fiber
systems," Opt. Express 16, 753 (2008).
6. Y. Koizumi, K. Toyoda, M. Yoshida, and M. Nakazawa, "1024 QAM (60 Gbit/s) single-
carrier coherent optical transmission over 150 km," Opt. Express 20, 12508 (2012).
7. A. H. Gnauck, P. J. Winzer, A. Konczykowska, F. Jorge, J.-Y. Dupuy, M. Riet, G.
Charlet, B. Zhu, and D. W. Peckham, "Generation and Transmission of 21.4-Gbaud PDM
64-QAM Using a Novel High-Power DAC Driving a Single I/Q Modulator," J. Light.
Technol. 30, 532–536 (2012).
8. R. Schmogrow, D. Hillerkuss, M. Dreschmann, M. Huebner, M. Winter, J. Meyer, B.
Nebendahl, C. Koos, J. Becker, W. Freude, and J. Leuthold, "Real-Time Software-
Defined Multiformat Transmitter Generating 64QAM at 28 GBd," IEEE Photonics
Technol. Lett. 22, 1601–1603 (2010).
9. M. R. Chitgarha, S. Khaleghi, Z. Bakhtiari, M. Ziyadi, O. Gerstel, L. Paraschis, C.
Langrock, M. M. Fejer, and A. E. Willner, "Demonstration of reconfigurable optical
generation of higher-order modulation formats up to 64 QAM using optical nonlinearity,"
Opt. Lett. 38, 3350 (2013).
10. S. Khaleghi, M. R. Chitgarha, O. F. Yilmaz, M. Tur, M. W. Haney, C. Langrock, M. M.
Fejer, and A. E. Willner, "Reconfigurable optical quadrature amplitude modulation
converter/encoder using a tunable complex coefficient optical tapped delay line.," Opt.
Lett. 38, 1600–2 (2013).
11. J. Leuthold, C. Koos, and W. Freude, "Nonlinear silicon photonics," Nat. Photonics 4,
535–544 (2010).
12. G. Steinmeyer, D. H. Sutter, L. Gallmann, N. Matuschek, and U. Keller, "Frontiers in
Ultrashort Pulse Generation: Pushing the Limits in Linear and Nonlinear Optics.," Science
286, 1507–1512 (1999).
13. S. Khaleghi, O. F. Yilmaz, M. R. Chitgarha, M. Tur, N. Ahmed, S. R. Nuccio, I. M. Fazal,
Xiaoxia Wu, M. W. Haney, C. Langrock, M. M. Fejer, and A. E. Willner, "High-Speed
Correlation and Equalization Using a Continuously Tunable All-Optical Tapped Delay
Line," IEEE Photonics J. 4, 1220–1235 (2012).
85
14. S. J. B. Yoo, "Wavelength conversion technologies for WDM network applications," J.
Light. Technol. 14, 955–966 (1996).
15. J. Kakande, R. Slavík, F. Parmigiani, A. Bogris, D. Syvridis, L. Grüner-Nielsen, R.
Phelan, P. Petropoulos, and D. J. Richardson, "Multilevel quantization of optical phase in
a novel coherent parametric mixer architecture," Nat. Photonics 5, 748–752 (2011).
16. H. Hu, E. Palushani, M. Galili, H. C. H. Mulvad, A. Clausen, L. K. Oxenløwe, and P.
Jeppesen, "640 Gbit/s and 128 Tbit/s polarisation insensitive all optical wavelength
conversion," Opt. Express 18, 9961 (2010).
17. A. Bogoni, X. Wu, S. R. Nuccio, J. Wang, Z. Bakhtiari, and A. E. Willner, "Photonic 640-
Gb/s Reconfigurable OTDM Add–Drop Multiplexer Based on Pump Depletion in a Single
PPLN Waveguide," IEEE J. Sel. Top. Quantum Electron. 18, 709–716 (2012).
18. Amirhossein Mohajerin Ariaei, "Reconfigurable High-Speed Processing and Noise
Mitigation of Optical Data." PhD diss., University of Southern California, (2018).
19. G. Agrawal, "Nonlinear Fiber Optics," New York 467 (2001).
20. M. Adams, "Optical waves in crystals," IEEE J. Quantum Electron. 20, 1294 (1984).
21. A. E. Willner, C. Langrock, J. E. McGeehan, M. M. Fejer, and S. Kumar, "All-Optical
Signal Processing Using χ
(2)
Nonlinearities in Guided-Wave Devices," J. Light. Technol.
Vol. 24, Issue 7, pp. 2579- 24, 2579 (2006).
22. B. J. Eggleton, B. Luther-Davies, and K. Richardson, "Chalcogenide photonics," Nat.
Photonics 5, 141–148 (2011).
23. I. Liberal, A. M. Mahmoud, Y. Li, B. Edwards, and N. Engheta, "Photonic doping of
epsilon-near-zero media," Science (80-. ). 355, 1058–1062 (2017).
24. S. Takasaka, Y. Taniguchi, M. Takahashi, J. Hiroishi, M. Tadakuma, and R. Sugizaki,
"Wideband parametric processing with 1-dB bandwidth of 40nm using dispersion stable
PM-HNLF," in European Conference on Optical Communication, ECOC (2014).
25. P. J. Winzer and R. J. Essiambre, "Advanced modulation formats for high-capacity optical
transport networks," J. Light. Technol. 24, 4711–4728 (2006).
26. S. J. Savory, "Digital filters for coherent optical receivers," Opt. Express 16, 804 (2008).
27. N. R. Newbury, "Searching for applications with a fine-tooth comb," Nat. Photonics 5,
186–188 (2011).
28. G. Contestabile, M. Presi, and E. Ciaramella, "Multiple wavelength conversion for WDM
multicasting by FWM in an SOA," IEEE Photonics Technol. Lett. 16, 1775–1777 (2004).
29. A. Bogoni, A. Malacarne, G. Berrettini, G. Meloni, L. Potì, and N. Sambo, "Optical
Multicasting of 16QAM Signals in Periodically-Poled Lithium Niobate
Waveguide," J. Light. Technol. Vol. 31, Issue 11, pp. 1797-1803 31, 1797–1803 (2013).
86
30. A. E. Willner, O. F. Yilmaz, J. Wang, X. Wu, A. Bogoni, L. Zhang, and S. R. Nuccio,
"Optically efficient nonlinear signal processing," IEEE J. Sel. Top. Quantum Electron. 17,
320–322 (2011).
31. B. Olsson and D. J. Blumenthal, "WDM to OTDM multiplexing using an ultrafast all-
optical wavelength converter," IEEE Photonics Technol. Lett. 13, 1005–1007 (2001).
32. E. J. M. Verdurmen, G. D. Khoe, A. M. J. Koonen, and H. De Waardt, "All-optical data
format conversion from WDM to OTDM based on FWM," Microw. Opt. Technol. Lett.
48, 992–994 (2006).
33. M. R. Chitgarha, S. Khaleghi, M. Ziyadi, A. Almaiman, A. Mohajerin-Ariaei, O. Gerstel,
L. Paraschis, C. Langrock, M. M. Fejer, J. Touch, and A. E. Willner, "Demonstration of
tunable optical generation of higher-order modulation formats using nonlinearities and
coherent frequency comb," Opt. Lett. 39, 4915–4918 (2014).
34. J. Kakande, A. Bogris, R. Slavik, F. Parmigiani, D. Syvridis, P. Petropoulos, D.
Richardson, M. Westlund, and M. Sköld, "QPSK Phase and Amplitude Regeneration at 56
Gbaud in a Novel Idler-Free Non-Degenerate Phase Sensitive Amplifier," in Optical Fiber
Communication Conference/National Fiber Optic Engineers Conference 2011, OSA
Technical Digest (CD) (Optical Society of America, 2011), pp. OMT4-OMT4.
35. M. Ziyadi, A. Mohajerin-Ariaei, A. Almaiman, Y. Cao, M. R. Chitgarha, L. Paraschis, M.
Tur, C. Langrock, M. M. Fejer, J. D. Touch, and A. E. Willner, "Optical channel de-
aggregation of quadrature-phase-shift-keying and eight-phase-shift-keying data using
mapping onto constellation axes," Opt. Lett. 40, 4899 (2015).
36. G.-W. Lu, T. Miyazaki, K. Mishina, S. M. Nissanka, A. Maruta, S. Mitani, K. Ishida, K.
Shimizu, T. Hatta, and K. Kitayama, "Optical phase erasure based on FWM in HNLF
enabling format conversion from 320-Gb/s RZ- DQPSK to 160-Gb/s RZ-DPSK," (n.d.).
37. H. Hu, R. M. Jopson, A. H. Gnauck, M. Dinu, S. Chandrasekhar, C. Xie, and S. Randel.
"Parametric amplification, wavelength conversion, and phase conjugation of a 2.048-Tbit/s
WDM PDM 16-QAM signal." J. Lightw. Technol. 33, 1286-1291 (2015).
38. Y. Liang, M. Lu, and Sh. Chu. "Narrow-band wavelength tunable filter based on
asymmetric double layer metallic grating." Opt. Express 23, 14434-14445 (2015).
39. L. Yan, A. E. Willner, X. Wu, A. Yi, A. Bogoni, Z-Y. Chen, and H-Y. Jiang. "All-optical
signal processing for ultrahigh speed optical systems and networks." J. lightw. Technol. 30,
3760-3770 (2012).
40. Z. Tong, A. Bogris, C. Lundström, C. J. McKinstrie, M. Vasilyev, M. Karlsson, and P. A.
Andrekson. "Modeling and measurement of the noise figure of a cascaded non-degenerate
phase-sensitive parametric amplifier." Opt. Express 18, 14820-14835 (2010).
87
41. T. Renyong, L. Jacob, and S. D. Preetpaul. "Gain characteristics of a frequency
nondegenerate phase-sensitive fiberoptic parametric amplifier with phase sel~ stabilized
input." Opt. Express 12, 10483-10493 (2005).
42. F. Alishahi, A. Vedadi, M. A. Soto, A. Denisov, K. Mehrany, L. Thévenaz, and C-S. Brès.
"Distributed measurement of signal power evolution in a phase sensitive parametric
amplifier." In optical fiber communication conference, Optical Society of America 2014,
paper We-12.
43. A. Almaiman, M. R. Chitgarha, W. Daab, M. Ziyadi, A. Mohajerin-Ariaei, S. Khaleghi, M.
Tur, A. E. Willner, "Experimental demonstration of robustness and accuracy of a DLI-
based OSNR monitor under changes in the transmitter and link for different modulation
formats and baud rates." Opt. letters 40, 2012-2015 (2015).
44. A. Mohajerin-Ariaei, A. Almaiman, M. Ziyadi, Y. Cao, A. Fallahpour, C. Bao, F. Alishahi
et al. "Analog performance of multiple, discretely tunable time delays based on a frequency
comb and a chromatic dispersion element." In CLEO: Science and Innovations, Optical
Society of America 2016, paper Th-13.
45. Jia, Shi, Jinlong Yu, Ju Wang, Wenrui Wang, Qiong Wu, Gangbin Huang, and Enze Yang.
"A novel optoelectronic oscillator based on wavelength multiplexing." IEEE Photon.
Technol. Letters 27, 213-216 (2014).
46. Prosyk, Kelvin, Tino Brast, Marko Gruner, Michael Hamacher, Detlef Hoffmann, Ron
Millett, and Karl-Otto Velthaus. "Tunable InP-based optical IQ modulator for 160 Gb/s."
In European Conference and Exposition on Optical Communications, Optical Society of
America 2011, paper Th-13.
47. N. Sambo, P. Castoldi, A. D’Errico, E. Riccardi, A. Pagano, M. S. Moreolo, J. M. Fàbrega,
D. Rafique, A. Napoli, S. Frigerio, E. H. Salas, G. Zervas, M. Nölle, J. K. Fischer, A. Lord,
and J.P.F.-P Gimenez, “Next Generation Sliceable Bandwidth Variable Transponders,”
IEEE Commun. Mag. 53, 163-171 (2015).
48. P. S. Khodashenas, J. M. R. Moscoso, B. Shariati, D. M. Marom, D. Klonidis, and I.
Tomkos, “Investigation of Spectrum Granularity for Performance Optimization of Flexible
Nyquist-WDM-Based Optical Networks,” J. Lightw. Technol. 33, 4767-4774 (2015).
49. C. Liu, J. Pan, T. Detwiler, A. Stark, Y.-T. Hsueh, G.-K. Chang, and S. E. Ralph, “Joint
Digital Signal Processing for Superchannel Coherent Optical Communication Systems,”
Opt. Express 21, 8342–8356 (2013).
88
50. K. Igarashi, T. Tsuritani, I. Morita, Y. Tsuchida, K. Maeda, M. Tadakuma, T. Saito, K.
Watanabe, K. Imamura, R. Sugizaki and M. Suzuki, “Super-Nyquist-WDM Transmission
over 7,326-km Seven-core Fiber with Capacity-distance Product of 1.03 Exabit/s·km,” Opt.
Express 22, 1220-1228 (2014).
51. M. Xiang, S. Fu, M. Tang, H. Tang, P. Shum, and D. Liu, “Nyquist WDM Superchannel
using Offset-16QAM and Receiver-side Digital Spectral Shaping,” Opt. Express 22, 17448-
17457 (2014).
52. J. Pan, C. Liu, T. Detwiler, A. J. Stark, Y. T. Hsueh and S. E. Ralph, “Inter-Channel
Crosstalk Cancellation for Nyquist-WDM Superchannel Applications,” J. Lightw. Technol.
30, 3993-3999 (2012).
53. S. Chandrasekhar and Xiang Liu, “Experimental Investigation on the Performance of
Closely Spaced Multi-carrier PDM-QPSK with Digital Coherent Detection,” Opt. Express
17, 21350-21361 (2009).
54. M. Sato, R. Maher, D. Lavery, K. Shi, B. C. Thomsen, and P. Bayvel, “Frequency Diversity
MIMO Detection for Dual-carrier DP-16QAM Transmission.” In European Conference on
Optical Communication (ECOC), 2014, Paper Th.2.5.1.
55. E. Al-Rawachy, R. P. Giddings, and J. M. Tang, “Experimental Demonstration of a DSP-
based Cross-channel Interference Cancellation Technique for Application in Digital Filter
Multiple Access PONs,” Opt. Express 25, 3850-3862 (2017).
56. T. Zeng, “Superchannel Transmission System based on Multi-channel Equalization,” Opt.
Express 21, 14799-14807 (2013).
57. S. Yamamoto, K. Saito, F. Hamaoka, T. Matsuda, A. Naka, and H. Maeda, “Characteristics
Investigation of High-Speed Multi-Carrier Transmission using MIMO-Based Crosstalk
Compensation in Homodyne Detection Scheme,” J. Lightwave Technol. 34, 2824-2832
(2016).
58. F. Lehmann, P. Ramantanis and Y. Frignac, “Joint Channel Estimation, Interference
Mitigation, and Decoding for WDM Coherent Optical Communications," J. of Opt. Comm.
Netw. 6, 315-325 (2014).
59. M. Sato, R. Maher, D. Lavery, K. Shi, B. C. Thomsen, and P. Bayvel. "Frequency Diversity
MIMO Detection for DP-QAM Transmission.” J. Lightw. Technol. 33, 1388-1394 (2015).
89
60. Z. Xiao, Songnian Fu, Shuchang Yao, Ming Tang, Perry Shum, and Deming Liu. “ICI
Mitigation for Dual-carrier Superchannel Transmission based on m-PSK and m-QAM
Formats.” J. Lightw. Technol. 34, 5526-5533 (2016).
61. S. Yamamoto, K. Saito, A. Naka, and H. Maeda, “Compatibility between Nonlinear
Compensation and Crosstalk Compensation using MIMO Processing in Super-high-density
multi-carrier transmission system.” In European Conference on Optical Communication
(ECOC), 2015 paper M.1-3.
62. N. Kaneda, T. Pfau, and J. Lee. “Frequency Diversity MIMO Detection for Coherent
Optical Transmission.” In European Conference and Exhibition on Optical Communication
(ECOC), 2013, paper Th3-3.
63. C. Sung, F. de Hoog, Z. Chen, P. Cheng, and D. C. Popescu. “Interference Mitigation Based
on Bayesian Compressive Sensing for Wireless Localization Systems in Unlicensed
Band.” Trans. on Veh. Technol. 66, 7038-7049 (2017).
64. Y. Cao, M. Ziyadi, A. Mohajerin-Ariaei, A. Almaiman, P. Liao, C. Bao, F. Alishahi, A.
Falahpour, B. Shamee, J.-Y. Yang, Y. Akasaka, M. Sekiya, M. Tur, C. Langrock, M. Fejer,
J. Touch, and A. E. Willner, “Reconfigurable Optical Inter-channel Interference Mitigation
for Spectrally Overlapped QPSK Signals using Nonlinear Wave Mixing in Cascaded PPLN
Waveguides,” Opt. Lett. 41, 3233-3236 (2016).
65. A. Mohajerin Ariaei, M. Ziyadi, Y. Cao, A. Almaiman, F. Alishahi, A. Fallahpour, C. Bao,
P. Liao, B. Shamee, J. Touch, M. Tur, C. Langrock, M. Fejer, and A. Willner,
“Demonstration of Tunable Mitigation of Interchannel Interference of Spectrally
Overlapped 16-QAM/QPSK Data Channels using Wave Mixing of Delayed Copies,” in
Optical Fiber Communication Conference (OFC), 2016, paper Th3J.5.
66. A. Mohajerin-Ariaei, F. Alishahi, A. Fallahpour, Y. Cao, A. Almaiman, C. Bao, P. Liao et
al. “Optical Mitigation of Inter-Channel Crosstalk for Multiple Spectrally Overlapped 40-
Gbit/s QPSK WDM Channels using Nonlinear Wave Mixing.” In Optical Fiber
Communications Conference and Exposition (OFC), 2018, Paper W3E.1.
67. A. E. Willner, S. Khaleghi, M. R. Chitgarha, and O. F. Yilmaz, "All-optical Signal
Processing," J. Lightw. Technol. 32, 660-680 (2014).
68. G. Bosco, A. Carena, V. Curri, P. Poggiolini, and F. Forghieri. "Performance limits of
Nyquist-WDM and CO-OFDM in high-speed PM-QPSK systems." IEEE Photon. Technol.
Letters 22, 1129-1131 (2010).
90
69. D. Hillerkuss, “Single-Laser 32.5 Tbit/s Nyquist WDM transmission.” J. Opt. Comm. Netw.
4, 715–723 (2012).
70. K. Harako, D. Suzuki, T. Hirooka, and M. Nakazawa. "Roll-off factor dependence of
Nyquist pulse transmission." Opt. Express 24, 21986-21994 (2014).
71. T. Hirooka, P. Ruan, P. Guan, and M. Nakazawa. "Highly dispersion-tolerant 160 Gbaud
optical Nyquist pulse TDM transmission over 525 km." Opt. Express 20, 15001-15007
(2012).
72. R. Schmogrow, M. Winter, M. Meyer, D. Hillerkuss, S. Wolf, B. Baeuerle, A. Ludwig et
al. "Real-time Nyquist pulse generation beyond 100 Gbit/s and its relation to OFDM." Opt.
Express 20, 317-337 (2012).
73. R. Schmogrow, D. Hillerkuss, S. Wolf, B. Bäuerle, M. Winter, P. Kleinow, B. Nebendahl
et al. "512QAM Nyquist sinc-pulse transmission at 54 Gbit/s in an optical bandwidth of 3
GHz." Opt. Express 20, 6439-6447 (2012).
74. M. A. Soto, M. Alem, M. A. Shoaie, A. Vedadi, C-S. Brès, L. Thévenaz, and T. Schneider.
"Optical sinc-shaped Nyquist pulses of exceptional quality." Nat. Communications 4, 1-9
(2013).
75. C. Bao, P. Liao, A. Kordts, M. Karpov, M. H. P. Pfeiffer, L. Zhang, Y. Cao et al. "Tunable
insertion of multiple lines into a Kerr frequency comb using electro-optical
modulators." Opt. letters 42, 3765-3768 (2017).
76. Y. Geng, X. Huang, W. Cui, Y. Ling, B. Xu, J. Zhang, X. Yi et al. "Terabit optical OFDM
superchannel transmission via coherent carriers of a hybrid chip-scale soliton frequency
comb." Opt. letters 43, 2406-2409 (2018).
77. F. Alishahi, P. Liao, A. Mohajerin-Ariaei, A. Fallahpour, A. Almaiman, Y. Cao, A. Kordts,
M. Karpov, M. H. P. Pfeiffer, T. J. Kippenberg, and A. E. Willner, "Tunable Insertion of
Uniform-Amplitude Multiple Coherent Lines into a Kerr Frequency Comb Using Nyquist
Pulse Generation," in Conference on Lasers and Electro-Optics, Optical Society of
America, 2018, paper SM1D.2.
78. T. J. Kippenberg, R. Holzwarth, and S. Diddams, "Microresonator-based optical frequency
combs," Science 332, 555-559 (2011).
79. P. Marin-Palomo, J. N. Kemal, M. Karpov, A. Kordts, J. Pfeifle, M. H. Pfeiffer, P. Trocha,
S. Wolf, V. Brasch, M. H. Anderson, R. Rosenberger, K. Vijayan, W. Freude, T. J.
91
Kippenberg, and C. Koos, "Microresonator-based solitons for massively parallel coherent
optical communications," Nature 546, 274-279 (2017).
80. A. Fallahpour, M. Ziyadi, A. Mohajerin-Ariaei, Y. Cao, A. Almaiman, F. Alishahi, C. Bao
et al. "Experimental demonstration of tunable optical de-aggregation of each of multiple
wavelength 16-QAM channels into two 4-PAM channels." In Optical Fiber
Communications Conference and Exhibition (OFC), 2017 paper M23-2
81. F. Alishahi, A. Fallahpour, K. Zou, Y. Cao, A. Kordts, M. Karpov, M. H. P. Pfeiffer et al.
"Experimental Generation and Time Multiplexing of Data-Carrying Nyquist Sinc Shaped
Channels from a Single Microresonator-based Kerr Frequency Comb." In Optical Fiber
Communication Conference, Optical Society of America, 2019, paper W3I-2.
82. T. Herr, V. Brasch, J. D. Jost, C. Y. Wang, N. M. Kondratiev, M. L. Gorodetsky, and T. J.
Kippenberg. "Temporal solitons in optical microresonators." Nat. Photonics 8, 145-149
(2014).
83. D. S. Millar, T. Fehenberger, T. Koike-Akino, K. Kojima, and K. Parsons. "Distribution
matching for high spectral efficiency optical communication with multiset partitions." J.
Lightw. Technol. 37, 517-523 (2019).
84. H. Nyquist, Certain topics in telegraph transmission theory. Trans. Am. Inst. Electr. Eng.
vol. 47, 617–644, (1928).
85. D. Hillerkuss, “Single-Laser 32.5 Tbit/s Nyquist WDM transmission”, J. Opt. Comm.
Netw. 4, 715–723 (2012).
86. G. Bosco, A. Carena, V. Curri, P. Poggiolini, and F. Forghieri, “Performance limits of
Nyquist-WDM and CO-OFDM in high-speed PM-QPSK systems”. IEEE Phot. Technol.
Lett. 22, 1129–1131 (2010).
87. R. Schmogrow, M. Winter, M. Meyer, D. Hillerkuss, S. Wolf, B. Baeuerle, A. Ludwig et
al. "Real-time Nyquist pulse generation beyond 100 Gbit/s and its relation to OFDM." Opt.
Express 20, 317-337 (2012).
88. M. Nakazawa, T. Hirooka, P. Ruan, and P. Guan. "Ultrahigh-speed “orthogonal” TDM
transmission with an optical Nyquist pulse train." Opt. Express 20, 1129-1140 (2010).
89. T. Hirooka, P. Ruan, P. Guan, and M. Nakazawa. "Highly dispersion-tolerant 160 Gbaud
optical Nyquist pulse TDM transmission over 525 km." Opt. Express 20, 15001-15007
(2014).
92
90. S. Zhang, F. Lu, X. Dong, P. Shum, X. Yang, X. Zhou, Y. Gong, and C. Lu. "Passive mode
locking at harmonics of the free spectral range of the intracavity filter in a fiber ring
laser." Opt. letters 30, 2852-2854 (2005).
91. O. Schröder, T. D. Vo, and B. J. Eggleton, "Repetition-rate-selective, wavelength-tunable
mode-locked laser at up to 640 GHz," Opt. Letters 34, 3902-3904 (2009).
92. F. Alishahi, A. Fallahpour, A. Mohajerin-Ariaei, Y. Cao, A. Kordts, M. H. P. Pfeiffer, M.
Karpov et al. "Reconfigurable optical generation of nine Nyquist WDM channels with sinc-
shaped temporal pulse trains using a single microresonator-based Kerr frequency
comb." Opt. letters 44, 1852-1855 (2019).
93. F. Alishahi, P. Liao, A. Mohajerin-Ariaei, A. Fallahpour, A. Almaiman, Y. Cao, A. Kordts
et al. "Tunable Insertion of Uniform-Amplitude Multiple Coherent Lines into a Kerr
Frequency Comb Using Nyquist Pulse Generation." In CLEO: Science and Innovations,
Optical Society of America, 2018, paper SM1D-2.
94. T. J. Kippenberg, R. Holzwarth, and S. Diddams, "Microresonator-based optical frequency
combs," Science 332, 555-559 (2011).
95. P. Del’Haye, A. Schliesser, O. Arcizet, T. Wilken, R. Holzwarth, and T. Kippenberg,
“Optical frequency comb generation from a monolithic microresonator,” Nature 450, 1214–
1217 (2007).
96. Compton, R. T. " The relationship between tapped delay-line and FFT processing in
adaptive arrays." IEEE Trans. on Ant. Propag. 36, 15-26 (1988)
97. Almaiman, A., H. Song, A. Minoofar, R. Zhang, X. Su, K. Zou, K. Pang et al. "Experimental
Demonstration of a Data Correlation and Data Equalization using a Tunable Optical
Tapped-Delay-Line Using the Spatial Domain and Modal-Dependent Delays." In CLEO:
Science and Innovations, pp. STu3O-3. Optical Society of America, 2020.
98. Bülow, Henning, Fred Buchali, and Axel Klekamp. "Electronic dispersion
compensation." J. Lightw. Technol. 26, 158-167 (2008).
99. Jackson, Kenneth P., Steven A. Newton, Behzad Moslehi, Moshe Tur, C. Chapin Cutler,
Josepgh W. Goodman, and Herbert J. Shaw. "Optical fiber delay-line signal
processing." IEEE Trans. Microw. Theory Techn. 33, 193-210 (1985)
100. Norton, D., S. Johns, C. Keefer, and R. Soref. "Tunable microwave filtering using high
dispersion fiber time delays." IEEE Photon. Technol. Lett. 6, 831-832(1994).
93
101. Ramisch, Roland, Gerhard R. Olbrich, and Peter Russer. "A tapped-delay-line
superconductive chirp filter in shielded microstrip." IEEE Photon. Technol. Lett. 39,:1575-
1581 (1991).
102. Dudek, Piotr, Stanislaw Szczepanski, and John V. Hatfield. "A high-resolution CMOS time-
to-digital converter utilizing a Vernier delay line." IEEE J. Sol. State Cir. 35, 240-247
(2000).
103. Koos, C., L. Jacome, C. Poulton, J. Leuthold, and W. Freude. "Nonlinear silicon-on-
insulator waveguides for all-optical signal processing." Opt. Express 15, 5976-5990 (2007).
104. Das, Pankaj K. Optical signal processing: fundamentals. Springer Science & Business
Media, 2012.
105. Willner, Alan E., Salman Khaleghi, Mohammad Reza Chitgarha, and Omer Faruk Yilmaz.
"All-optical signal processing." J. Lightw. Technol. 32, 660-680 (2013).
106. Ziyadi, Morteza, Mohammad Reza Chitgarha, Amirhossein Mohajerin-Ariaei, Salman
Khaleghi, Ahmed Almaiman, Yinwen Cao, Moshe J. Willner et al. "Optical Nyquist
channel generation using a comb-based tunable optical tapped-delay-line." Opt. Lett. 39,
6585-6588 (2014).
107. Hu, Jianqi, Camille-Sophie Brès, and Chen-Bin Huang. "Talbot effect on orbital angular
momentum beams: azimuthal intensity repetition-rate multiplication." Opt. Lett. 43, 4033-
4036 (2018).
108. Deng, Qingzhong, Lu Liu, Rui Zhang, Xinbai Li, Jurgen Michel, and Zhiping Zhou.
"Athermal and flat-topped silicon Mach-Zehnder filters." Opt. Express 24, 29577-29582
(2016).
109. Kim, Dae Woong, Assia Barkai, Richard Jones, Nomi Elek, Hat Nguyen, and Ansheng Liu.
"Silicon-on-insulator eight-channel optical multiplexer based on a cascade of asymmetric
Mach–Zehnder interferometers." Opt. Lett. 33, 530-532 (2008).
110. Horst, Folkert, William MJ Green, Solomon Assefa, Steven M. Shank, Yurii A. Vlasov, and
Bert Jan Offrein. "Cascaded Mach-Zehnder wavelength filters in silicon photonics for low
loss and flat pass-band WDM (de-) multiplexing." Opt. Express 21, 11652-11658 (2013).
111. Van Campenhout, Joris, William MJ Green, Solomon Assefa, and Yurii A. Vlasov. "Low-
power, 2× 2 silicon electro-optic switch with 110-nm bandwidth for broadband
reconfigurable optical networks." Opt. Express 17, 24020-24029 (2009).
94
112. Piekarek, Mateusz, Damien Bonneau, Shigehito Miki, Taro Yamashita, Mikio Fujiwara,
Masahide Sasaki, Hirotaka Terai et al. "High-extinction ratio integrated photonic filters for
silicon quantum photonics." Opt. Lett. 42, 815-818 (2017).
113. Bui, Lam Anh. "Characterization of optical Mach–Zehnder modulator using DC
measurements." Opt. Lett. 45, 6346-6349 (2020).
114. Ribeiro, Antonio, Sibert Declercq, Umar Khan, Mi Wang, Lukas Van Iseghem, and Wim
Bogaerts. "Column-row addressing of thermo-optic phase shifters for controlling large
silicon photonic circuits." IEEE J. Sel. Top. Quantum Elect. 26, 1-8 (2020).
115. Shang, Hongpeng, Degui Sun, Peng Yu, Bin Wang, Ting Yu, Tiancheng Li, and Huilin
Jiang. "Investigation for sidewall roughness caused optical scattering loss of silicon-on-
insulator waveguides with confocal laser scanning microscopy." Coatings 10, 236 (2020).
116. Teng, Min, et al. "Miniaturized silicon photonics devices for integrated optical signal
processors." J. Light. Technol. 38, 6-17 (2020).
117. Aikawa, Yohei. "Optical comparator for 4-bit and 6-bit QPSK-modulated signals by using
Optical delayed interferometer." IEEE Photon. J. 10, 1-11 (2018).
118. Koos, Christian, Lenin Jacome, Christopher Poulton, Jeurg Leuthold, and Wolfgang
Freude. "Nonlinear silicon-on-insulator waveguides for all-optical signal processing." Opt.
Express 15, 5976-5990 (2007).
119. Kim, Dae Woong, Assia Barkai, Richard Jones, Nomi Elek, Hat Nguyen, and Ansheng
Liu. "Silicon-on-insulator eight-channel optical multiplexer based on a cascade of
asymmetric Mach–Zehnder interferometers." Opt. Lett. 33, 530-532 (2008).
120. Alishahi, Fatemeh, Ahmad Fallahpour, Kaiheng Zou, Amir Minoofar, Huibin Zou, Cong
Liu, Jonathan Habif, Moshe Tur, Alan E. Willner." Study of Optical Signal Processing
Performance Dependency on Non-Ideal MZI Operation in a Tapped-Delay-Line."
Conference on Laser and Electrooptics (CLEO), (2021) papar JTu3A.52
121. Bogoni, Antonella, and Alan Willner. "Photonic Signal Processing for Logic and
Computation." In All-Optical Signal Processing, pp. 157-183. Springer, Cham, 2015.
122. Chitgarha, Mohammad Reza, Salman Khaleghi, Omer F. Yilmaz, Moshe Tur, Michael W.
Haney, Carsten Langrock, Martin M. Fejer, and Alan E. Willner. "Coherent correlator and
equalizer using a reconfigurable all-optical tapped delay line." Opt. Lett. 38, 2271-2273
(2013).
95
123. Rasras, Mahmoud S., Inuk Kang, Mihaela Dinu, James Jaques, Niloy Dutta, Alfonso
Piccirilli, Mark A. Cappuzzo et al. "A programmable 8-bit optical correlator filter for
optical bit pattern recognition." IEEE Photon. Technol. Lett. 20, 694-696 (2009).
124. Aikawa, Yohei. "Optical comparison operation for 8-Bit QPSK-modulated signal by using
serially-cascaded delay line interferometer." App. Sciences 8, 1440 (2018)
125. Ruffini, Marco, Lena Wosinska, Mohand Achouche, Jiajia Chen, Nick Doran, Farsheed
Farjady, Julio Montalvo et al. "DISCUS: An end-to-end solution for ubiquitous broadband
optical access." IEEE Comm. Mag. 52, S24-S32 (2014).
126. Mukherjee, Biswanath. "WDM optical communication networks: progress and
challenges." IEEE J. on Sel. Areas Comm. 18, 1810-1824 (2000).
Abstract (if available)
Abstract
Technology has enabled people in all walks of life to generate, store, and communicate enormous amounts of data. Recent technological advances in high-speed backbone data networks, together with the growing trend of bandwidth demanding applications have created a need for higher capacities in signal transmission and signal processing. The bandwidth-hungry applications such as cloud computing, photos and video sharing, data storage systems and recent technological advances in high-speed data networks have created a demand for higher speeds in data processing and transmission. Lately, system capacity increases have been realized by coherent technologies and advanced modulation formats [1–10]. ❧ High-data-rate all-optical signal processing has been one of the main research goals in photonics. Signal processing using nonlinear optics has been of great interest due to its inherent ultrafast THz bandwidth and its phase-preserving nature [1,11-14]. A key feature of optical signal processing is that optical techniques do not need to “touch” or switch every individual “bit,” as electronic transistors do. That is; optical approach may operate at the line rate of the data and the latency o the process can be very small. Optical amplifiers, for instance, can amplify Tb/s signals without touching the signal at the bit level. Another example is basic wavelength conversion using a laser pump and a nonlinear device, where the data information can be transferred from one carrier wavelength to another at a very fast speed (nonlinearities have femtosecond response times) as optical signals fly through the device [3, 4]. Importantly, advances in materials and devices which have resulted in devices with higher nonlinearities and higher efficiencies, and photonic integrated circuits (PICs) technologies, are the key for any practical utilization of optical signal processing in the future [15-17]. Additionally, if signals are already in the optical domain, it might be beneficial to avoid inefficient optical-to-electrical-to-optical (OEO) conversion by doing optical signal processing. Also, multiple signal processing functions can be achieved for multiple channels within the same optical device such that multiple electronic devices can be replaced with a single optical device. And finally, it is possible to exploit and simultaneously manipulate multiple dimensions of the optical wave, e.g., amplitude, phase, wavelength, polarization, and space, to provide more degrees of freedom. ❧ This Ph.D. thesis explores the potential of ultra-highspeed and reconfigurable optical subsystems to function as different processing units through of a huge amount of data. The optical signal processing units being explored here has the advantage that they can be employed in flexible networks which hosts channels with shared spectral regions. This dissertation demonstrates optical systems that avoid redundant OEO conversion and can control the processing units locally either at the transmitter site or at a remote location from it throughout the optical network. Therefore, subsystems introduced throughout this thesis can support in-line signal processing for high baud rate signal. By utilizing various forms of photonic nonlinear interactions different functions including wavelength selection, optical buffering, optical Nyquist WDM and TDM channel generation and transmission, optical inter-channel-interference mitigation for multiple spectrally overlapped channels and remote pattern recognition of digital data using different optical subsystems are demonstrated.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Reconfigurable high-speed processing and noise mitigation of optical data
PDF
All-optical signal processing toward reconfigurable optical networks
PDF
High-speed and reconfigurable all-optical signal processing for phase and amplitude modulated signals
PDF
Reconfigurable high speed optical signal processing for optical communication and modulation format manipulation
PDF
Reconfigurable high‐speed optical signal processing and high‐capacity optical transmitter
PDF
Nonlinear optical signal processing for high-speed, spectrally efficient fiber optic systems and networks
PDF
Optical signal processing for enabling high-speed, highly spectrally efficient and high capacity optical systems
PDF
Reconfigurable optical signal processing for efficient spectrum utilization in high-speed optical communication systems
PDF
Optical signal processing for high-speed, reconfigurable fiber optic networks
PDF
Optical wave mixing for tunable delays and high‐speed signal processing
PDF
Microresonator-based Kerr frequency comb for high-speed optical communications and signal processing
PDF
Orbital angular momentum based spatially multiplexed optical and millimeter-wave communications
PDF
optical communication systems and space-time wave packet generation using frequency combs and spatial modes
PDF
Spectrally efficient waveforms for coherent optical fiber transmission
PDF
Integrated silicon waveguides and specialty optical fibers for optical communications system applications
PDF
Silicon photonics integrated circuits for analog and digital optical signal processing
PDF
Optical communications, optical ranging, and optical signal processing using tunable and reconfigurable technologies
PDF
Optical simulation and development of novel whispering gallery mode microresonators
PDF
High capacity optical and RF/mm-wave communications utilizing orbital angular momentum multiplexing
PDF
Applications of all optical signal processing for advanced optical modulation formats
Asset Metadata
Creator
Alishahi, Fatemeh
(author)
Core Title
Reconfigurable and flexible high-speed optical signal processing and spectrally shared optical subsystems
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Electrical Engineering
Degree Conferral Date
2021-12
Publication Date
11/03/2021
Defense Date
09/30/2021
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
coherent communications,coherent optical systems,nonlinear optics,OAI-PMH Harvest,optical signal processing,time division multiplexing systems,wavelength division multiplexing systems
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Willner, Alan (
committee chair
), Haas, Stephan (
committee member
), Habif, Jonathan (
committee member
), Wu, Wei (
committee member
)
Creator Email
alishahi.fateme@gmail.com,falishah@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC16351961
Unique identifier
UC16351961
Legacy Identifier
etd-AlishahiFa-10200
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Alishahi, Fatemeh
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Repository Email
cisadmin@lib.usc.edu
Tags
coherent communications
coherent optical systems
nonlinear optics
optical signal processing
time division multiplexing systems
wavelength division multiplexing systems