Applications of Modern RF Photonics

For a complete listing of titles in the

Artech House Applied Photonics Series,
turn to the back of this book.
Applications of Modern RF Photonics

Preetpaul Singh Devgan

Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the U.S. Library of Congress.

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library.

Cover design by John Gomes

ISBN 13: 978-1-63081-159-4

Supplemental videos for this book can be found online at:

http://us.artechhouse.com/Applications-of-Modern-RF-Photonics-P1932.aspx

© 2018 ARTECH HOUSE

685 Canton Street
Norwood, MA 02062

All rights reserved. Printed and bound in the United States of America. No part of this book
may be reproduced or utilized in any form or by any means, electronic or mechanical, includ-
ing photocopying, recording, or by any information storage and retrieval system, without
permission in writing from the publisher.
All terms mentioned in this book that are known to be trademarks or service marks have
been appropriately capitalized. Artech House cannot attest to the accuracy of this information.
Use of a term in this book should not be regarded as affecting the validity of any trademark
or service mark.

10 9 8 7 6 5 4 3 2 1
For Marie, Alexander, and Louise
Contents

Preface 13

Acknowledgments 15

1 Introduction to Applications of Modern RF Photonics 17

1.1 A Brief Overview of RF Photonic History 18

1.2 RF Photonic Advantages 19

1.3 Analog versus Digital Photonics 22

1.4 Current Needs for RF Photonics 28

1.5 Conclusions 29
References 30

2 Analog Delay Lines 33

2.1 Different Examples of Analog Delay Lines Using RF

Photonics 34

2.2 Definitions of RF Metrics 36

2.3 Different Architectures of RF Photonic Delay Lines 40

7
8 Applications of Modern RF Photonics

2.4 RF Photonic Component Performance 49

2.5 Conclusions 55
References 56

3 Advancements in Analog Delay Line Performance 59

3.1 Performance Improvement Through the Photonic

Components 60

3.2 Improvements in the Photodetector 61

3.3 Improvements in the Optical Fiber 64

3.4 Improvements in the Optical Amplifier 66

3.5 Improvements in the Optical Modulator 69

3.5.1 Off-Quadrature Biasing of the Optical Modulator 69
3.5.2 Low Biasing of the Optical Modulator with Dual
Wavelengths 71
3.5.3 Cancelation of Dispersion Induced Second
Harmonics by Using Dual Wavelengths 74
3.5.4 Single-Sideband Modulation 76
3.5.5 Single-Sideband Modulation to Cancel
Photodetector Nonlinearities 78

3.6 Conclusions 79
References 79

4 Oscillators Utilizing RF Photonics 85

4.1 Need for Oscillators 85

4.2 Phase Noise and Timing Jitter 86

4.3 Optoelectronic Oscillator 88

4.3.1 Multiloop OEO 90
4.3.2 OEO with All-Photonic Gain 93
4.3.3 Clock Synchronization Using an OEO 94

4.4 Oscillators Based on Two Laser Sources 98

 Contents 9

4.5 Conclusions 102

References 102

5 Signal Isolation Utilizing RF Photonics 107

5.1 Need for Signal Separation 107

5.2 Using RF Photonics for Separation of Signals 109

5.3 Finite Impulse Response Filters Using RF Photonics 111

5.4 Isolation of RF Signals Along a Common Path 115

5.5 Conclusions 119

References 119

6 Signal Identification Utilizing RF Photonics 123

6.1 Need for Signal Identification 124

6.2 Using RF Photonics for Spectrum Analysis 125

6.3 Using Photonics Filters for Instantaneous Frequency

Measurement 127

6.4 Using Dispersion for Instantaneous Frequency

Measurement 129

6.5 Combinations of Different Methods for Frequency

Measurement 132

6.6 Using FIR and IIR Filters for Instantaneous

Frequency Measurement 133

6.7 Frequency Measurement with Multimode Photonic

Systems 134

6.8 RF Frequency Identification Using Optical Injection

Locking 138

6.9 Conclusions 142

References 142
10 Applications of Modern RF Photonics

7 Signal Processing Utilizing RF Photonics 145

7.1 Need for Downconversion 146

7.2 Using RF Photonics for Downconversion 148

7.3 Advancements in RF Photonic Downconverters 150

7.4 RF Photonic Analog-to-Digital Conversion 153

7.5 RF Photonics Sampling Combined with Electronic

Quantization 155

7.6 Photonics Sampling and Quantization 158

7.7 Arbitrary-Transmit Waveform Generation Using

RF Photonics 159

7.8 Conclusions 162

References 163

8 Advancements in Integrated RF Photonics 167

8.1 Integrated Photonic Fundamentals 168

8.2 IPCs 170

8.3 Applications of IPCs to RF Photonics 173

8.4 Other Applications in IPCs 175

8.5 Further Work in IPCs for Analog Applications 177

8.6 Conclusions 178

References 178

9 Conclusions 183

9.1 A Brief Review of RF Photonics 184

9.2 Discrete-Based RF Photonic Subsystems 185

9.3 Alternative Systems Using RF Photonics 188

9.4 Future Work in RF Photonics 191

 Contents 11

References 193

List of Acronyms 197

About the Author 203

Index 205
Preface
This book is intended as an overview of the many diverse ways that RF pho-
tonics can be used to address the needs of signal-processing applications.
While many previous volumes have covered the fundamental principles of
RF photonics, this work focuses on the application of RF photonics to ad-
dress specific needs. The book uses a demonstration architecture to highlight
different areas where RF photonics can be applied. The intended audience
includes senior undergraduate students studying engineering and/or physics,
as well as professional researchers and engineers who are interested in using
RF photonics. The book serves as a primer for RF photonic applications,
making it useful for the second half of an introductory class on RF photonics.
While the book presents multiple examples, it is not complete. Any omissions
are not meant as a judgment on technical deficiencies, but are due simply to
a limited amount of time and space.
Chapter 1 begins with a brief overview of the history of RF photonics,
compares digital and analog photonics, and discusses the need for RF pho-
tonic technologies. Chapter 2 then briefly discusses the performance of the
different photonic components that make up an analog delay line, highlight-
ing where RF photonics may offer advantages. Chapter 3 covers multiple
methods for improving the performance of the analog delay line; subsequent-
ly, Chapter 4 presents an overview of RF oscillators using photonic means.
Chapter 5 describes methods for isolating and separating RF signals received
by the demonstration architecture using photonic subsystems. Chapter 6 cov-
ers spectral analysis using RF photonic means. Chapter 7 provides an over-
view of RF photonics-based down conversion, analog-to digital conversion,
and arbitrary waveform generation. Chapter 8 discusses the use of integrated

13
14 Applications of Modern RF Photonics

photonic circuits. Finally, Chapter 9 provides some concluding remarks while

looking at the past, present, and future of RF photonic applications.
It is my hope that this book addresses a gap in the current literature
about the application of RF photonic technologies to real-world needs. This
field will continue to grow, and this book should provide a basis for inspiring
new designs and techniques to meet upcoming challenges.
Acknowledgments
The opinions and views expressed in this book are purely those of the author
and do not reflect the official position of the Air Force Research Laboratory,
the United States Air Force, or the U.S. government.
This book is a private effort, and I would like to recognize the many
people who inspired and supported me in this work. Beginning with my early
studies, I would like to thank Andrès, Jim, Matt, Pradeep, Mitch, Mike, Al,
and all the others who challenged me and pushed me to work harder.
Following my early education, working at Lucent Technologies helped
me to define my future goals and gave me the confidence to believe in my
own abilities, even when I made mistakes. For everything I learned, and for
all the opportunities I was given, I would like to thank Eric, Marty, Tina,
Dan, Dave, Bob, Olga, Peter, Jim, Jeff, Ron, and all the others at Lucent who
mentored and supported me.
During my doctoral education, I was given many opportunities to push
my understanding and gain new experiences. I would like to thank Eric, Paul,
Sarah, Matt, and all the others with whom I was fortunate to work. To Jay
Sharping, Kobi Lasri, and Renyong Tang, thank you for all your help in the
lab, for all your mentoring, and for all the fruitful discussions. To Darwin
Serkland, Gordon Keeler, and Kent Geib, thank you for providing a summer
spent learning new skills and reinforcing basic concepts. To William Kath
and Mary Phillips, thank you for your advice and for serving on my commit-
tee. To Prem Kumar, thank you for giving me a chance and for the freedom
to explore new ideas and grow as an engineer.
At the Naval Research Laboratory, I was able to fully develop a deeper
understanding of the field of RF photonics. Having the ability to work with

15
16 Applications of Modern RF Photonics

great people was always one of the best parts of the job. Thank you to Jason,
Alex, Modesto, Chris, Ross, Meredith, Dave, Jim, Kristina, and the others.
To Frank Bucholtz, thank you for offering me an opportunity. To John Diehl,
Chris Sunderman, and Joe Singley, thank you for all the hard work, support,
and positive discussions—even the heated ones. To Vince Urick, thank you
for being a great technical role model and for all of your support, mentorship,
patience, and commitment to the mission. To Keith Williams, thank you for
giving me the freedom to pursue my interests, while always making sure they
fit into the greater goals.
At the Air Force Research Laboratory, I have been fortunate to continue
working in RF photonics and pursue new ideas and techniques. Having a
supportive management team has helped me to expand my horizons. Thank
you to Attila Szep, Steve Hary, Charles Cerny, Chris Bozada, Jesse Fanning,
and Timothy Johnson. To Matt Casto, thank you for the informative discus-
sions on analog-to-digital converters. To Dale Stevens, thank you for your
unwavering support and help. To Rob Nelson, Ben Griffin, and Dean Brown,
thank you for challenging my preconceived notions and helping me to see
things from a different point of view. To Joe Suelzer, Tom Simpson, and Joe
Haeffner, thank you for all of the technical discussions on optical injection
locking and nonlinear laser dynamics. To Nick Usechak, thank you for the
stimulating conversations on integrated photonics and for welcoming me
into your lab without any hesitation.
Finally, none of this is possible without the support of my family. For
all of their help, patience, encouragement, and love, I am forever grateful. To
Alex and Louise, thank you for understanding and being the best kids I could
ever ask for. To Marie, thank you for everything. I miss you.
1
Introduction to Applications of Modern
RF Photonics
Radio frequency (RF) photonics is an interdisciplinary field that combines
analog electronics with photonic components for the high-speed processing
of RF signals. While the term RF photonics is most commonly used, this
technology area is also known as microwave photonics [1–3] or analog op-
tics [4]. Regardless of the name given, the field is focused on up-converting
continuously time-varying electronic signals into the optical domain, where
high-speed signal processing can occur before downconverting the signal back
to the electronic domain. Signal processing is a general term that incorporates
techniques to address many different applications. These applications cover a
broad range, including the transmission of analog signals, the generation of
low-phase noise clock signals, the isolation and frequency identification of RF
signals, and many others. While these types of applications have traditionally
been served by electronic means, the need to access higher frequencies and
wider bandwidths has driven the search for alternatives. Because photonics
can operate essentially from DC to daylight, it seems an ideal solution to
address these requirements. In this introduction, a brief history covering the
uses of RF photonics is presented. Next, the chapter provides an overview
of the advantages of photonics versus electronics, along with a comparison
between the use and performance metrics of digital and analog photonics.
Finally, the chapter reviews the current needs for RF photonics.

17
18 Applications of Modern RF Photonics

1.1 A Brief Overview of RF Photonic History

The field of RF photonics has seen a significant growth in its exploration
since the beginning of the millennium. While the field has been around since
the 1970s, the adoption of RF photonic technologies has been uneven. The
initial interest in RF photonics started as the use of fiber-optics for com-
munications really began to take off. Using the same components that were
developed for digital transmission, RF photonics investigated different ap-
plications. By the 1980s, the loss of optical fibers in the 1,550-nm operating
window allowed them to be considered as a replacement for traditional cop-
per wires. With the invention of the erbium-doped fiber amplifier (EDFA),
however, the use of fiber-optic cables in long-haul communication networks
quickly spread. While the long-haul communication networks were interest-
ed in the transmission of data over transoceanic distances, there was interest
in sending analog data over optical fiber links for shorter distance applica-
tions. In the 1970s, cable television was being deployed to households in the
United States. The advent of new networks that were only available through
cable TV prompted a significant increase in subscribers. As this growth con-
tinued, the need to meet the demand led to the investigation of analog signals
being transported over fiber-optic cables. One of the main solutions uses hy-
brid fiber coax (HFC) systems to deliver not only cable TV signals, but also
voice [5]. With the optical fiber and amplifiers now in use, the latter part of
the 1980s was devoted to improving the laser sources that could provide the
needed performance for analog video transmission. This work culminated in
the 1990s with analog fiber systems being readily deployed to deliver multiple
TV channels to the user.
However, once digital video technology matured, the need for analog
video transmission began to fade. The availability of high-definition TVs,
along with other factors, helped push the network providers to deliver their
programming in digital formats. While the fiber-optic networks could sup-
port either analog or digital data transmission, the stringent requirements
on the components themselves were relaxed for digital transmission. Digital
transmission also allowed for voice, TV, and internet traffic to be handled all
on the same network without mixing analog and digital formats. This simpli-
fication led to the current connectivity that households now enjoy.
The advancements in RF photonic systems in the 1980s led to RF pho-
tonics being explored for other uses. One of the early applications was pho-
tonic beam-steering for RF phased-array antennas [6]. While fiber-optic ca-
bles’ low loss and immunity to electromagnetic interference are attractive for
phased arrays, their small physical size is also very important. The reduced size
allows for smaller and more tightly configured antenna arrays. For a phased
 Introduction to Applications of Modern RF Photonics 19

array that can be steered up to 90°, the spacing between the elements has to
be less than half of the wavelength of the highest frequency. For example,
an array that operates at 30 GHz or higher requires element spacing on the
order of half a millimeter, which is much larger than the diameter of a coaxial
cable. While projects explored the uses of RF photonics to feed phased arrays
[7], they never realized the full potential of the technology. These results led
to an investment in improving the components that made up the photonic
beam-steering system. One of the key components is the external modulator.
A large amount of time and money was spent to improve the performance
of these devices. In addition, laser and photodetector performance was im-
proved. These investments have now pushed the overall performance of the
systems to the point that RF photonics has seen an explosion in uses.

1.2 RF Photonic Advantages

With the increased performance of photonic components developed over the
last few decades, RF photonics can offer many advantages over their elec-
tronic equivalents for operating at these higher frequencies. One of the easiest
comparisons to make is between the transmission lines. In RF photonics, the
signal is carried over a single-mode fiber-optic cable. The single-mode optical
fiber consists of a core of glass that is typically around 9 microns in diameter.
The small core of glass is then surrounded by a cladding that is 125 microns
in diameter. The fiber-optic cable is used to guide light generated by a laser
from one point to another. The laser light has been modulated by the RF
signals that are to be transported over this cable. The electronic equivalent
is a coaxial cable. This cable consists of an inner conductor made of copper,
which is surrounded by either a dielectric material or air. The cable then has
an outer conductor. The size of the inner conductor and the amount of air
or dielectric depends on the range of frequencies that the cable is designed to
transport. Many different arrangements of connectors for coaxial cables exist.
Table 1.1 lists the various connector types with the corresponding size of their
inner conductors and the maximum frequency they support.
When fiber-optic cables were first named by Kapany [8], the losses as
a function of length were quite large. However when Kao and others [9–11]
were able to remove the water vapor from the glass as it was being drawn, the
losses were able to approach the Rayleigh scattering limit. Figure 1.1 shows
the loss of an optical fiber as a function of wavelength. The loss of optical
fibers decreases as the wavelength increases. Two small peaks occur at 1,240
nm and 1,380 nm; they are both due to water absorption in the glass (silica).
The minimum loss window is centered at 1,550 nm and is flat around 30 nm
20 Applications of Modern RF Photonics

Table 1.1
List of Coaxial Connectors
and Their Maximum
Frequency
RF Coax Maximum
Connector Frequency
SMA 18 GHz
3.5 mm 34 GHz
2.92 mm 40 GHz
2.4 mm 50 GHz
1.85 mm 70 GHz
1.0 mm 110 GHz

Figure 1.1 Loss of fiber-optic cable in decibels per kilometer as a function of optical wave-
length from 1,100 to 1,750 nm. (After: [12].)

of that center. Above 1,620 nm, the infrared absorption of silica dominates,
and the losses quickly rise. In terms of frequencies, the fiber loss is flat over
approximately 3.75 THz. For a laser source operating in the low-loss wave-
length window, the fiber loss is flat for any RF frequencies that are considered
here.
Coaxial cables do not have a flat loss for a given length as a function
of frequency. Looking at a standard coaxial cable, the losses are dominated
by two effects: conductor loss and dielectric loss. The conductor loss has a
 Introduction to Applications of Modern RF Photonics 21

square-root dependence on frequency, while the dielectric loss is linearly de-

pendent on frequency. Figure 1.2 shows a plot of the loss of a fiber-optic
cable and a coaxial cable as a function of frequency. A fiber-optic cable with
a length of 1 kilometer has a loss of approximately 0.2 dB and is flat over
the entire frequency range from 1 MHz to 10 GHz. A coaxial cable with a
length of 10m has a very flat loss as well, which slowly rolls off to 10 dB at
10 GHz [13]. However, the loss of the same coaxial cable with a length of
100m quickly begins to grow as the frequency increases. At a frequency of
800 MHz, the loss is as high as 20 dB, or 100 times lower than the power of
the signal at the input of the cable.
The bandwidth capacity of the optical fibers is also much larger than
coaxial cables. Multiple lasers of different wavelengths can be multiplexed
together and transmitted down a single optical fiber. Wavelength division
multiplexing (WDM) has been used in multiple photonic systems to increase
the data-carrying capacity of a single optical fiber. Since the loss of the fiber
is flat over 30 nm (as seen in Figure 1.1), the loss each wavelength experiences
is the same 0.2 dB/km. Between 40 and 80 separate wavelengths can be mul-
tiplexed, with each wavelength carrying its own data. Since these data streams
are in parallel, the aggregate data rate can be quite high. The data-carrying
capacity of a single fiber-optic cable is on the order of 10 Tbps over tens of
kilometers, compared to a single coaxial cable carrying data on the order of
10 Gbps over tens of meters.

Figure 1.2 Loss of a 1-km fiber-optic cable and 10-m and 100-m coaxial cables in decibels
as a function of frequency from 1 MHz to 10 GHz. (After: [12].)
22 Applications of Modern RF Photonics

While loss and bandwidth capacity are two key advantages of fiber-
optic cables, there are numerous others. The weight per unit length of a co-
axial cable operating out to higher frequencies can be on the order of 20–200
grams/meter, while a fiber-optic cable will weigh 1 gram/meter [13]. Thus
the weight of a 1-km length of fiber-optic cable is 20–200 times less than a
coaxial cable of the same length. A meter length of optical fiber costs on the
order of tens of dollars while a similar length of 1-mm coaxial cable would
cost on the order of a couple of thousands of dollars. Since glass expands less
than copper as a function of temperature, quantified by a thermal coefficient
of delay (TCD), fiber-optic cables are less susceptible to phase changes over
temperature gradients than coaxial cables [14]. Optical fibers are electrically
nonconductive, which makes them immune to electromagnetic interference
and electrically isolates the end points of the fiber link. This can be impor-
tant, as a large electrical discharge, such as a lightning strike, will not propa-
gate down the cable and potentially destroy valuable electronics at the end of
the link. Additionally, the fiber-optic cable will not spark, allowing it to be
used in environments that contain highly flammable liquids, such as a fuel
tank. Finally, fiber-optic cables can be easily concealed. They cannot be lo-
cated using metal detectors, unlike a coaxial cable. Table 1.2 summarizes the
advantages of fiber cables over coaxial cables.

1.3 Analog versus Digital Photonics

The advantages provided by fiber-optic cables are the same for both analog
and digital photonic applications. The question then becomes why RF pho-
tonics did not become as widely adopted as digital photonics starting in the
1990s. The answer lies in the fact that while the fiber-optic cable and the
EDFA met the performance requirements for both of the types of system,
the laser, optical modulator, and photodetector did not provide the perfor-

Table 1.2
Fiber-Optic Cable’s Advantages Over Coaxial Cables
Fiber versus Coax Advantage
Loss (db/100m) 19.98 dB lower (at 800 MHz) [12]
RF bandwidth 1,000x (190 THz versus 110 GHz)
Size (inner diameter) 900x (9 microns versus 1 millimeter)
Cost ($/m) 100x ($1,000 versus $10s)
TCD (ppm/°K) 5–10x (fiber dependent) [13]
Weight (kg) 20–200x less at 1 km [12]
 Introduction to Applications of Modern RF Photonics 23

mance needed for analog systems. In order to understand why, the distinc-
tion between digital and analog signal processing has to be reviewed. Digital
photonic signal-processing applications have been widely explored for many
different commercial needs. The first and most common use is for transmit-
ting large amounts of digital data over transoceanic distances. As opposed to
analog data, digital data consists of discrete voltage levels. For binary digital
data, the discrete voltage levels are denoted as ones and zeros. Typically, a one
is when the voltage is greater than some threshold level, and a zero is when the
voltage is below this threshold. The bit rate is given by how long in time the
voltage stays a one or a zero before it changes. In Figure 1.3(a), an ideal data
stream is shown. The levels of ones and zeros are clearly above and below the
threshold, denoted by a horizontal dashed line. In addition, the transition be-
tween the two levels is instantaneous. Thus, the voltage stays at the intended
level for the entire bit time, as denoted by the vertical dashed lines. However,
a practical data stream does not look like the ideal one. Figure 1.3(b) shows
a more practical data stream. In this case, the voltage levels are not always
the same. While the ones and zeros are above and below the threshold line,
respectively, they can often be very close. A smaller change in levels can lead
to errors, where a one is identified as a zero and vice versa. Additionally, the
transition between levels is not instantaneous in a practical system. A finite
time is needed for the transition to occur. This reduces the time that the volt-
age will stay at the one or zero level. Errors occur when the measurement of
the level is taken during this transition time and the wrong level is chosen.
An actual measurement of the data will yield a bit error rate (BER). Us-
ing a BER tester (BERT) will actually measure the errors of the digital data
stream. By generating a pseudorandom bit stream at the input of the link,
the BERT will recover the bits at the output of the link and compare them to
the original bit stream. The BERT will then count the number of errors and
calculate a BER. This is a measure of how many errors are found in a given
unit of time. As an example, a BER of 1×10-10 means that one error occurs
every 1010 bits. For a bit rate of 10 Gbps, you would on average expect to see
one error every second. So the error rate is a function of the rate of the data
that is being checked. However the BERT does not provide any insight on

Figure 1.3 (a) Ideal binary digital data stream and (b) real binary digital data stream.
24 Applications of Modern RF Photonics

the source of the errors that occur in the link. For this, a visual representation
of the data would be useful.
Using a sampling oscilloscope, an eye diagram can be measured instead.
The eye diagram captures many of the important metrics for digital data by
making a histogram of the measured digital data. An eye diagram gets its
name from the shape of the diagram. The slope of the transitions gives the
open area an oval shape, very much like the shape of a human eye [15]. Fig-
ure 1.4 shows an eye diagram for a practical data stream. The one and zero
levels are not narrow lines but have a finite width to them. This width is a
measure of the amplitude noise on each level. The larger the noise, the wider
the level and closing of the eye diagram. The transitions between levels are
also shown. The noninstantaneous rise and fall time is shown as the slope of
the transitions. The width of the rise and falling is a measure of how much
timing jitter, or uncertainty, there is when the transition occurs. Again, as
the timing jitter increases, the width of the transitions increases and the eye
begins to close. An eye mask, as shown inside the open spaces of the data
stream in Figure 1.4, is a visual measure of the openness of the eye diagram.
Whenever measured bits appear within the eye mask, they will lead to errors
in the received data stream. Looking at the ratio of voltages for the ones and
zeros level as compared to the relative widths of each level, a Q-factor can be
defined. Equation (1.1) shows the Q factor as the difference in mean voltages
between the one (µ1) and zero (µ0), divided by the sum of the standard devia-
tions of the voltages of the two levels (σ1 and σ0).

µ1 − µ0
Q= (1.1)
σ1 + σ 0

Figure 1.4 Eye diagram of a binary digital data stream. (After: [16].)
 Introduction to Applications of Modern RF Photonics 25

Assuming the noise has a Gaussian distribution, the Q factor can be

used to estimate the BER of the data. The relation between the Q factor and
BER is shown in (1.2), where the erfc() function is the complementary error
function. While not as exact as the BERT, the visual information provided
by the eye diagram is very useful for understanding the performance of the
digital system.

1 Q 
BER = erfc  (1.2)
2  2 

For an analog system, the key metrics are different from the digital
equivalent. While a digital link is defined ultimately by the BER, the analog
link performance can be distilled down to four metrics: RF gain, minimum
detectable signal (MDS), compression dynamic range (CDR), and max input
spur-free (MISF) power [17]. The MDS is the input power at which the out-
put power of the system will be above the noise floor plus some margin. The
formal definition of MDS appears in (1.3). The MDS is directly proportional
to the RF noise figure (NFRF­) of the system. The MISF is the maximum in-
put power before any output spurious signal power appears above the noise
floor of the system. The definition of MISF is given in (1.4).

MDS (dBm ) = N th (dBm ) + NFRF (dB) + 10 log10 (BW (Hz ))

(1.3)
+ Margin(dB)

MISF (dBm ) = MDS (dBm ) + SFDR (dB in1Hz ) − Margin(dB) (1.4)

The MISF is directly proportional to the spur-free dynamic range

(SFDR). As an equivalent to the eye diagram for digital systems, a plot of the
analog response of an RF photonic system can be made that visually captures
the metrics for these systems. Such a plot appears in Figure 1.5, which shows
the relationship between the input RF power (X-axis) and output power (Y-
axis) in units of decibels relative to 1 mW or log scale. The noise floor of the
system is measured and displayed as a horizontal line. The power of the noise
floor in relation to the thermal noise limit and the RF gain of the system is the
RF noise figure. The fundamental response for a given operating frequency
of the analog system is represented by a line with a slope of one. As the input
power increases, the output power increases by the same amount. Simply tak-
26 Applications of Modern RF Photonics

Figure 1.5 RF metrics represented by a plot of RF input power versus output power. (After:
[12].)

ing the difference between the input and output powers along this line yields
the RF gain.
As the input power increases, the measured output power will begin
to deviate from the ideal slope of one. The deviation occurs when the sys-
tem enters compression and can no longer provide the same amount of gain.
When the output power is 1 dB less than the output power predicted by the
slope of the line, the 1-dB compression point has been reached. The range
of input powers over which the signal is above the noise floor and equal to
the 1-dB compression point is the CDR. Equivalently, the range of input
powers over which the signal is above the noise floor but the spurious signal,
represented by the third-order intermodulation distortion (IMD_3) response
line in Figure 1.5, is below the noise floor. This range of input powers is the
SFDR. Finally, assuming the fundamental response did not roll over at the
compression point, the line would continue until it crosses the IMD_3 line.
The output power where these two lines cross is the output-referenced inter-
cept point.
Note that digital photonic systems have a lot of advantages. Since
the data is digitized, or transformed into a stream of discrete one and zero
 Introduction to Applications of Modern RF Photonics 27

voltages, the potential exists to correct any errors that occur as the data passes
over long distances. Multiple methods exist for correcting errors in transmis-
sion of digital data. One method is called forward error correction (FEC).
Similar to the idea of a parity bit, FEC utilizes extra bits to create a Reed-Sol-
omon code that can distinguish when errors occur in transmission [18]. Un-
like a parity bit, the codes can actually be used to correct the errors at the end
point, thus earning the name FEC. FEC requires extra overhead in order to
work. For a SONET optical channel denoted as 9.953 Gbps or OC-192, the
FEC-enabled data rate is actually 10.664 Gbps. The higher speed allows for
the extra FEC overhead bits to be incorporated in the SONET network. FEC
has been shown to take raw BERs as high as 10-4 and correct them to BERs of
10-10, which is enough to be used at the end terminal point. The drawback is
that the data has to be processed in order to correct the errors. So it is useful
at the end point where the added processing time is not a hindrance.
Other methods exist to correct errors within the network that require
less processing time. Regenerating the signal is one example of this type of
error correction. Digital photonic networks often use 3R regeneration. 3R re-
generators perform three operations that all start with “R”: retime, reamplify,
and reshape. Retiming corrects timing errors that can occur. It minimizes the
width of the transitions in the eye diagram. Reamplification corrects errors by
increasing the level of the ones, moving it further above the threshold voltage.
Reshaping involves reducing the width of the one’s and zero’s levels. It can
also be used to improve the slope of the transitions between levels. Regenera-
tion has been a key factor in allowing digital data to be corrected before being
further transmitted.
For analog transmission, the only regeneration that can be done is re-
amplify. The ability to retime and reshape the data does not really exist for
analog transmission. This is due to the fact that digital data is known a priori
to entering the network. Since the original signal is digitized into the ones
and zeros, all the information is encoded. For an analog transmission sys-
tem, there is no knowledge of what the signal is before it enters the network.
Therefore, no retiming can be performed as there is no known fundamen-
tal frequency that the signal will follow. In fact, the analog signal may be
purposefully changing frequencies over time. The analog network needs to
be transparent to this effect and simply pass the signal without adding any
other distortions. Due to the more stringent requirements, the lasers, optical
modulators, and photodetectors used in digital applications were not able to
provide the required performance. However, the photonic component per-
formance has now improved in order to meet these requirements. Thus, RF
photonics can now address the needs of current systems.
28 Applications of Modern RF Photonics

1.4 Current Needs for RF Photonics

While the component performance was improved, new applications for RF
photonics were also becoming apparent. One example of the need for operat-
ing at higher frequencies is in wireless communications. Mobile telephones
dominate the frequency bands from 600 MHz to 2.1 GHz in the United
States. The 2.4-GHz band is also used by wireless LAN routers as well as
some cordless telephones. The latest 802.11ac WiFi standard now uses fre-
quencies at 5 GHz. Satellite communications are expanding out to 10 GHz
and higher. As more and more wireless signals take up the available frequency
spectrum, the only available bandwidth will be at higher frequencies. In addi-
tion, higher carrier frequencies will enable faster data rates for wireless devices
in order to meet the needs of users. To meet these needs, communications
will make use of frequencies above 20 GHz. As seen in Figure 1.6, the loss of
a signal that is transmitted through the atmosphere is very low (< 1 dB/km)
from 1 to 100 GHz, except for the rather large absorption peak around 60
GHz [19]. Clearly there is room to grow in the future. While RF photonics
may not be used in the wireless transmission of data, it will be used to con-
nect the wireless towers to the base stations and the base stations to the central
station. In addition, the technology can be used to generate RF carriers at
frequencies much higher than used currently, as well as to separate transmit-
ted and received signals.

Figure 1.6 Atmospheric attenuation in decibels per kilometer as a function of frequency

from 1 to 100 GHz. (After: Devgan [12].)
 Introduction to Applications of Modern RF Photonics 29

While the atmospheric attenuation can be low over a wide frequency

range, wireless signals experience higher attenuation due to the materials used
in buildings. The loss increases as the frequency of the wireless carrier in-
creases. For a large apartment building, a hybrid method may be used to sup-
ply each apartment with access to a wireless signal. From the central station,
a fiber-optic cable can be run to each apartment. The optical cable will feed
an antenna in the apartment, thus creating a local wireless access point. With
a proper choice of radio frequency, the signal from the access point will be
attenuated enough so that it does not interfere with any neighboring apart-
ment’s signal. Thus, each apartment will act as a picocell. An RF photonic
system is best suited to address this application, due to the high-frequency
(HF) radio signals that are used. These specific types of links are often re-
ferred to as radio-over-fiber links [20].
RF photonic technologies have applications for needs other than data
transmission. Another common application is the distribution of RF signals
from one point to another. RF photonic links are currently used in radio
astronomy, which utilizes large dish antennas to collect very faint radio waves
from distant stars. In order to improve the signal-to-noise ratio (SNR) of the
received signals, multiple antennas are used. The antennas are spread out over
multiple kilometer distances from each other. In order to recover the detected
signal, each antenna is connected to a central base station. Due to the large
separation between antennas, RF photonic links are used to connect them
together. This is just one example of antenna remoting. Other systems also
use RF photonic links for antenna remoting. In some cases, the antennas can
be placed in hostile environments. With coaxial cables, the antennas can only
be separated from the central station by about 1,000 feet. This places the
people in the central station in the same environment as the antennas. With
an RF photonic link, the distance can be increased to several kilometers. One
such demonstration has sent signals in the HF band over seven kilometers
[21]. Other needs that have been identified include beamforming in radio
astronomy [22] and true time delay [23]. All of these needs can be addressed
with RF photonic solutions. Later chapters will expand on these and other
applications where RF photonics provides distinct advantages over electronic
solutions.

1.5 Conclusions
Analog systems do offer advantages over digital systems in some cases. When
the frequency of the signal is higher than can be digitized with the required
fidelity, an analog system must be used. Analog systems are also often more
30 Applications of Modern RF Photonics

robust than digital systems. In a measure of the reliability of links for a radio
astronomy system, the analog links had a fourfold increase in the mean time
between failures when compared to digital links [24]. In the digital link, the
analog-to-digital converter (ADC) is placed right at the antenna, exposing
it to the elements. ADCs often failed due to environmental effects, such as
lightning. Analog links have less complex front ends, leading to a less failures.
Clearly, analog systems are useful in certain applications. RF photonic sys-
tems that offer advantages over other solutions are explored in further details
in the following chapters.
The remainder of the book focuses on the various applications of RF
photonics. These include examples of analog delay lines, oscillators, signal
isolation, and filtering. The various applications are reviewed in the context
of an analog architecture that requires these various functions. In addition,
we will develop a demonstration architecture over the course of the book.
The different RF photonic signal-processing applications that can be accom-
plished within the architecture are highlighted in individual chapters. The
initial architecture will consist of an antenna that receives an RF spectrum of
signals and passes those signals through to a final output. Subsequently, we
will explore additional functionality to add to the architecture. While not
all applications will be covered by the architecture, it will provide a concrete
example of how RF photonics can address different applications.

References
[1] Seeds, A. J., and K. J. Williams, “Microwave Photonics,” Journal of Lightwave Technol-
ogy, Vol. 24, No. 12, 2006, pp. 4628–4641.
[2] Capmany, J., and D. Novak, “Microwave Photonics Combines Two Worlds,” Nature
Photonics, Vol. 1, No. 6, 2007, pp. 319–330.
[3] Vilcot, A., B. Cabon, and J. Chazelas (eds.), Microwave Photonics: From Components to
Applications and Systems, Boston, MA: Kluwer Academic Publishers, 2003.
[4] Cox III, C., Analog Optical Links, Cambridge, England: Cambridge University Press,
2004.
[5] Kaminow, I. P., and T. Koch (eds.), Optical Fiber Telecommunications IIIA, Cambridge,
MA: Academic Press, 1997.
[6] Williamson, R., and R. Esman, “RF Photonics,” Journal of Lightwave Technology, Vol.
26, No. 9, 2008, pp. 1145–1153.
[7] Yao, J., “A Tutorial on Microwave Photonics,” IEEE Photonics Society Newsletter, June
2012, pp. 5–12.
 Introduction to Applications of Modern RF Photonics 31

[8] Kapany, N. S., Fiber-optics Principle and Applications, Cambridge, MA: Academic Press,
1967.
[9] Kao, K. C., and T. W. Davies, “Spectrophotometric Studies of Ultra Low Loss Optical
Glasses-I: Single Beam Method,” Journal of Physics E: Scientific Instruments, Vol.1, No.
11, 1968. pp. 1063–1068.
[10] Jones, M. W., and K. C. Kao, “Spectrophotometric Studies of Ultra Low Loss Optical
Glasses-II: Double Beam Method,” Journal of Physics E: Scientific Instruments, Vol. 2,
No. 4, 1969, pp. 331–335.
[11] Wright, C. R., and K. C. Kao, “Spectrophotometric Studies of Ultra Low Loss Optical
Glasses-III: Ellipsometric Determination of Surface Reflectances,” Journal of Physics E:
Scientific Instruments, Vol. 2, No. 7, 1969, pp. 579–583.
[12] Devgan, P., “Review of Radio Frequency Photonics Basics,” AFRL Report, AFRL-RY-
WP-TR-2017-0157, 2017.
[13] MIL-C-17 Attenuation loss and power handling tables.
[14] Dreher, J., Phase Stability of ATA Fiber-optic Cables, Seti Institute, ATA Memo 55,
March 2003.
[15] Agrawal, G. P., Fiber-optic Communication Systems, Hoboken, NJ: John Wiley and
Sons, 1997.
[16] Devgan, P., “High-speed Signal Processing Using Nonlinear Fibers and Optoelectronic
Devices,” Ph.D. dissertation, 2005.
[17] Pozar, D. M., Microwave Engineering, Hoboken, NJ: John Wiley and Sons, 2009.
[18] Chow, M-C., Understanding SONET/SDH Standards and Applications, Holmdel, NJ:
Andan Publisher, 1996.
[19] ITU Recommendation, Attenuation by Atmospheric Gases, ITU-R P.676-10, Aug. 2013.
[20] Berceli, T., and P. Herczfeld, “Microwave Photonics—A Historical Perspective,”
Microwave Theory and Techniques, Vol. 58, No. 11, 2010, pp. 2992–3000.
[21] Urick, V. J., et al., “Field Test on the Feasibility of Remoting HF Antennas with Fiber-
optics,” NRL Memorandum Report, NRL/MR/5652-08-9137, 2008.
[22] van Ardenne, A., et al., “A Dense Aperture Array with Photonic Beamforming For
Radio Astronomy,” International Conference on Electromagnetics in Advanced Applications
(ICEAA), 2014.
[23] Zach, S., and L. Singer, “RF Photonics—Why Should Defense Take Notice?,”
Convention of Electrical and Electronics Engineers in Israel, 2006.
[24] Montebugnoli, S., et al., “Large Antenna Array Remoting Using Radio-Over-Fiber
Techniques for Radio Astronomical Application,” Microwave and Optical Technology
Letters, Vol. 46, No. 1, 2005, pp. 48–54.
2
Analog Delay Lines
One of the first applications where RF photonics has proved to offer advan-
tages is a simple RF link. The link is one of the easiest applications to under-
stand when first discussing RF photonic technologies. The RF photonic link
simply needs to transport a spectrum of signals from one point to another.
One example of an RF photonic link would be connecting an antenna to a
signal-processing system, as shown in Figure 2.1. A laser provides an optical
carrier to a photonic modulator. The RF signal is then upconverted onto the
optical carrier via the photonic modulator and then sent down an optical fi-
ber cable. At the end of the link, a photodetector downconverts the RF signal
to the electrical domain. For this application, the photonic link is designed
to deliver the RF signal at the end point with as little distortion as possible.
When compared to a coaxial cable link, the RF photonic link may not always
be able to match the nonlinear performance. As the coaxial cable is a passive
transmission line, it adds no nonlinearities to the signals. However, the RF
photonic link is an active link. Since the link has an upconversion opera-
tion followed by a downconversion operation, there are nonlinear distortions
added. Often, the choice between a coaxial cable link and an RF photonic
link comes down to needs. RF photonic links are best suited for either long
lengths or ultrawide frequency operations. For short lengths (< 100m) and
low frequencies (< 1 GHz), coaxial cables are more than capable of handling
the needs of most RF transmission links. However, when the length increases
or the frequency exceeds these bounds, the RF photonic link is often the
preferred choice. While the coaxial links do not add nonlinearities, they do

33
34 Applications of Modern RF Photonics

Figure 2.1 Example of an RF photonic link.

suffer from very high losses as either the length or the frequency increases, as
seen in Figure 1.2. One specific application where RF photonic links have an
advantage over coaxial links is analog delay lines.
An analog delay line has many applications, from emulating long-dis-
tance transmission to acting as an analog memory unit. Analog delay lines
meet the thresholds for using RF photonic technologies, as they often have
large frequency bandwidth requirements and require long lengths. In order
to quantify the performance of an analog delay line, the RF metrics need to
be addressed. While the key RF metrics can be distilled down to the MDS
and the MISF, these are based off the RF gain, noise figure (NF), and SFDR.
Phase noise and CDR are two other important RF metrics. To design an RF
photonic delay line, the predicted RF metrics must be calculated. RF photon-
ic links can be broken down into their constituent components, namely the
laser, modulator, optical amplifier, optical fiber, and photodetector. As these
components drive the overall RF performance of the delay line, a translation
between photonic performance and RF performance needs to be defined.
This chapter defines the various RF metrics, both inherently and as a func-
tion of photonic component performance.

2.1 Different Examples of Analog Delay Lines Using RF Photonics

As discussed previously, analog delay lines require different performance met-
rics in order to be usable for their intended applications. A few examples of
the possible uses of analog delay lines can highlight the metrics that need to
be accounted for when designing RF photonic delay lines. In one such ex-
ample, the RF photonic delay line can be used as an analog buffer, and this
is how we will use the analog delay line in our demonstration architecture.
As seen in Figure 2.2, a spectrum of RF signals can be captured from an
antenna and split into two different paths. The first path delivers the signals
to the processing system, which will process the signals. One example of this
type of processor is a thresholder that identifies signals above a certain power
level. Any signal that meets this requirement can then be digitized. However,
 Analog Delay Lines 35

Figure 2.2 Demonstration architecture incorporating an analog buffer.

the original signal may not be able to be digitized after the thresholder as it
may be significantly distorted or attenuated by the thresholding process. If a
copy of the signal was kept, then the unmodified signal can be digitized. By
splitting the spectrum of signals, an original copy is preserved in the analog
buffer. The analog buffer just needs to hold the spectrum long enough for the
processing system to identify the chosen signal and then set up the digitizer
to record the signal. The required processing time can be on the order of 100
µs. An analog buffer would have to hold the spectrum of signals for at least
that length of time. To achieve this buffering time, an analog delay line would
need to be constructed with a delay length of approximately 20 km. A coaxial
delay line would not be able to meet this requirement as the loss would be too
large to properly preserve the original signals. Fortunately, an RF photonic-
based delay line would be able to meet these requirements. In this case, the
key RF metric of interest is the NF of the delay line. Since the signal power
can be conditioned (amplified or attenuated as necessary) at the input of the
delay line, the signals can be prepared to stay below the MISF and above the
MDS. Even if the input power is above the MISF, any spurious signals that
are generated will be ignored as they would not match the frequency of the
signal to be digitized. However, if the delay line has too high of a NF, the
SNR will be sufficiently degraded. The digitizer will then be unable to cap-
ture any useful information. So, in the design of this delay line, the RF NF
can be optimized at the expense of the SFDR.
In another case, the RF photonic delay line can be used as a range emu-
lator. In this case, a user wishes to transmit a single or limited set of signals
over a long distance (> 10 km). Multiple reasons exist for not actually trans-
mitting the signal over that long of a distance, including not having access
to an open area of that distance and the complexity and cost of setting up a
transmitting and receiving station over a significantly long distance. Instead,
an analog delay line can emulate the long distance in a small form factor, al-
lowing the signal fidelity after transmission to be compared with the original
36 Applications of Modern RF Photonics

in the same location. For this application, the delay line needs to add as little
distortion as possible to the signals. Since the signal is known going into the
delay line, the input SNR can be properly set at the input so that after the
delay line the SNR is sufficient for the signal performance to be determined.
Therefore, the NF of the delay line is important, but not as important as the
nonlinear distortion. In this case, the delay line is optimized for SFDR and
the amount of added phase noise is reduced.

2.2 Definitions of RF Metrics

The five main RF metrics that define most RF photonic systems are the RF
gain, RF NF, SFDR, CDR, and the phase noise. With these metrics, the per-
formance of the RF photonic system can be calculated and used to determine
if it is appropriate for the application. The first metric that all RF applications
need to know is the RF gain of the system. The RF gain is a measure of the
RF power at the output of the system relative to the input RF power to the
system. For this discussion, the RF gain will be defined in logarithmic units
of decibels. Note an RF gain of 0 dB is defined as a unity gain, and a negative
RF gain means that the system has loss. The mathematical expression of the
RF gain appears in (2.1a). For a complex RF system that contains multiple
RF components, the total RF gain is simply a sum of the various gains of the
individual components, also shown in (2.1b).

Grf (dB ) = Prf ,out (dBm ) − Prf ,in (dBm ) (2.1a)

n
Gtot (dB ) = ∑Gi (dB ) (2.1b)
i =1

The RF NF is measured as the change in the SNR at the input of the

system in relation to the SNR at the output. Note that the noise at the input
of the system must be thermally limited. If this is not the case, the definition
of NF is no longer true. The mathematical definition of RF NF is presented
in (2.2a), where kB is Boltzmann’s constant, T is the temperature (typically
defined at 290K), and Nout­ is the measured output noise level. Note that the
bandwidth is assumed to be 1 Hz in this definition. The NF of a composite
system can be determined by using the NF of the individual components
that make up the system. The Friis NF cascade equation is shown in (2.2b)
[1]. Looking at (2.2b), the overall system NF is often determined by the first
 Analog Delay Lines 37

component’s NF, as long as the gain of the first component is enough to

compensate the NF of the following components. In most RF systems, a low
noise amplifier is the first component in the system. The low NF, combined
with high RF gain, sets the overall NF performance of the system.

SNR in,th S in ⋅ N out N out

NFrf ≡ = = (2.2a)
SNR out kBT ⋅ S out G rf kBT

 
 NF − 1
m
NF = NF1 + ∑  i −1i  (2.2b)
i=2  
m ≥ 2  ∏G p 
 p =1 

The RF SFDR is a measure of the input powers over which the original
input signal is above the system noise floor while no spurious signals appear.
The spurious signal is a result of nonlinearities in the components of the
system, which leads to either harmonic generation from a single RF input or
intermodulation distortions from the mixing of multiple frequency compo-
nents. A plot of the different spurious signals that are generated from a non-
linear RF system is shown in Figure 2.3. In this case, two different single, CW
frequency tones are input to the RF system. At the output, the original two
signals are observed, as are several other spurious signals. For each of the two
tones, the harmonic spurious signals appear. The second harmonic appears at
twice the original frequencies, followed by the third, and so on. In addition,
since there were two original signals, the intermodulation spurious signals ap-
pear as well. The second-order intermodulation spurious signals appear at the
sum and difference of the original two signals. The third-order intermodula-
tion spurious signals appear at the difference of double one signal’s frequency

Figure 2.3 Diagram of the various nonlinear spurious signals for a two-tone input.
38 Applications of Modern RF Photonics

minus the other signal’s frequency. Often RF systems are limited to a single
octave of bandwidth in order to filter out the harmonic and second-order
intermodulation spurious signals. However, the third-order intermodulation
spurious signals appear within an octave bandwidth and cannot be filtered
out. These types of RF systems are referred to as third-order-limited. One
of the advantages of an RF photonic system is that they can be third-order-
limited while operating over greater than an octave of bandwidth. The reason
why this occurs will be discussed later in Section 2.3.
The SFDR can be calculated from two different data points, the NF
of the system and the output-referenced intercept point (OIP). The OIP is
measured by plotting the output power of the system as a function of the
input power, as seen in Figure 2.4. For the two-tone input, the input power
of each tone is increased by the same amount, and the power of each is kept
equal. The plot of the two-tone output, known as the fundamental response,
is a measure of the gain or loss of the RF system. The spurious signals of
interest are also plotted. In the case of Figure 2.4, the second- and third-
order harmonics as well as the intermodulation distortions are plotted. For
the fundamental response, the line will ideally have a slope of one, while
the second- and third-order spurious signal responses will have slopes of two
and three, respectively, on a logarithmic scale. The OIP is the output power
where the extrapolated fundamental response line crosses that of the spuri-
ous signal response. Note that the second harmonic OIP2H is 6 dB higher
than the second-order intermodulation OIP2IMD, while the third harmonic

Figure 2.4 Second- and third-order nonlinear intercept points for intermodulation and
harmonic distortion.
 Analog Delay Lines 39

OIP3H is 4.8 dB higher than the third-order intermodulation OIP3IMD [2].

Once the OIP is determined, the SFDR is calculated using the output noise.
Just like the NF, an overall OIP can be calculated using the OIP of the indi-
vidual components that make up the RF system. The OIP cascade formula
is presented in (2.3b). From (2.3b), the worst case OIP of the overall system
is often set by the last element in the system, which is the opposite of the
cascade result for the RF NF.
(n −1) n
( ) OIPn 
SFDR Hz (n −1) n = 
 N out 
(2.3a)

−2
 −1 2 
 m −1  m  −1 2 
OIP 2 =  ∑ OIP 2i ∏ G p  + OIP 2m 
 im=≥1 2  p = i +1  
 
(2.3b)
−1
 −1 
 m −1  m  −1 
OIP 3 =  ∑ OIP 3i ∏ G p  + OIP 3m 
 im=≥1 2  p = i +1  
 

The RF CDR is the range of input powers over which the output signal
is above the noise floor of the system and the output power follows the linear
gain of the system. An example of the CDR appears in Figure 2.5; the solid
line is the measured output power while the dashed black line is the ideal
response line with a slope of one. The signal power is above the noise floor
and follows the ideal response until it begins to roll off. When the measured
output power is 1 dB below the ideal response, the input power is referred to
as the 1-dB compression point. The CDR is then calculated using (2.4). In
most RF photonic systems, the CDR is often much larger than the SFDR,
due to the nonlinear response of the upconversion and downconversion pro-
cess that occurs in these systems [3].

CDR (dB ⋅ Hz ) = P1dB (dBm ) + 1 − N out (dBm Hz ) (2.4)

Finally the added phase noise is an important metric for analog delay
lines. Specifically, analog delay lines that are used to emulate the transmis-
sion of a signal over a long length are often concerned with the added phase
noise to the signal. Phase noise is used as a measure of how spectrally pure a
40 Applications of Modern RF Photonics

Figure 2.5 One-dB CDR. (After: [4].)

single RF tone from an oscillator is. While an ideal oscillator will generate a
perfect RF tone at a given frequency, a practical RF tone will have some noise
bandwidth. The phase noise of the tone determines the power of the noise
relative to the power of the carrier as a function of the frequency offset from
the center frequency. Phase noise is measured by taking the single sideband
power spectral density of the signal’s phase fluctuations [5]. The larger the
phase noise, the more the RF signal will deviate from its fundamental fre-
quency. The phase noise is also used to calculate the timing jitter of the RF
signal. Phase noise, which is often used to measure the quality of an oscillator,
is covered in greater detail in Chapter 4.

2.3 Different Architectures of RF Photonic Delay Lines

In order to express the important RF metrics of an analog delay line in terms
of the RF photonic components that make up the delay line, the architec-
ture has to be determined. When determining the type of architecture to
use, the most important decision to be made is the type of modulation that
is used. Multiple ways exist to upconvert the RF spectrum of signals onto
an optical carrier. All RF photonic links require at least a laser, an optical
 Analog Delay Lines 41

fiber, and a photodetector. A choice needs to be made between external and

direct modulation. Direct modulation acts on the laser itself, while external
modulation uses a separate modulator after the laser to upconvert the signal
on to the optical carrier. Direct modulation has been investigated by many
for use in RF photonic links, including delay lines. Figure 2.6, which shows
a directly modulated link along with a plot of the laser power as a function
of drive current, illustrates how the direct modulation occurs. Once above
the threshold current for lasing to occur, the optical power increases as the
drive current increases. The relation between the drive current and the opti-
cal power is known as the slope efficiency of the laser. This relation appears
in (2.5). Picking a bias point where the laser is above threshold, the current
can be modulated by an analog signal, allowing it to be upconverted onto the
power of the optical carrier. One of the advantages of direct modulation is
how simple it is to implement. Unfortunately, though, direct modulation has
a serious disadvantage. The RF gain from a directly modulated link does not
meet most analog delay line applications. The RF gain for a directly modulat-
ed link is directly proportional to the square of the slope efficiency, sL, of the
laser (Grf ∝ s L2 ) as seen in (2.6), where K is a constant related to the link loss
and the responsivity of the photodetector and Zin and Zout are the input and
output impedances, respectively. K is always less than one. Equations (2.7)
and (2.8) show the second and third harmonic OIP for the directly modu-
lated link. The constants b and c are due to the nonlinear slope of the bias

Figure 2.6 Directly modulated link architecture with laser output power as a function of
bias current. (After: [4].)
42 Applications of Modern RF Photonics

curve and have to be measured for each different laser. Since a practical laser
will never be 100% efficient, the slope efficiency will always be less than one.
From the above relation, the RF gain will always be negative in the decibel
scale if the slope efficiency is less than one. Therefore, the RF link will always
introduce loss to the signal. In order to overcome this limitation, the link will
require either optical or electronic amplification. Unfortunately, an amplifier
will add additional noise, degrading the SNR performance of the link. Other
disadvantages of direct modulation include limited frequency response and
added frequency chirp to the optical carrier that can degrade the RF perfor-
mance after traveling through a dispersive fiber medium [6].

I dc
sL ∝ (2.5)
I bias

Z out
Grf = Κs L2 (2.6)
Z in

Κ 4
OIP2H = 2 s L Z out (2.7)
b2

Κ 3
OIP3H = 2 s L Z out (2.8)
c

Having eliminated direct modulation as a choice for the analog delay

line architecture, external modulation is now left. In external modulation,
multiple solutions can be used. Two main types have been investigated: phase
and intensity. Phase modulators encode the analog signal onto the phase of
the optical carrier, while keeping the intensity of the light constant. Phase-
modulated links, as seen in Figure 2.7, have been investigated for multiple
applications [7]. Phase modulation offers several advantages, including re-
moving the need for a DC bias at the modulator as well as having lower
optical loss when compared to other intensity modulators. In fact, the phase-
modulated RF link can provide more RF gain at the output when compared
to an intensity-modulated link [7]. However, the phase-modulated link
comes with a couple of drawbacks. First, the phase modulation has to be
 Analog Delay Lines 43

Figure 2.7 Phase-modulated link architecture with RF response of asymmetric interfer-

ometer detection with a delay of 200 psec. (After: [4].)

converted to intensity before the photodetector in order to recover the RF

signal. While multiple conversion methods exist [8, 9], a common one is the
use of an asymmetric interferometer for demodulation. The asymmetry in
the interferometer used for demodulation places a bandwidth limitation on
the link. This effect is shown in Figure 2.7. When the frequency of the RF
signal is near the peak of the response, the RF gain of the link is maximized.
However, as the frequency changes, the RF gain drops. This is a function of
the asymmetry of the interferometer and has a response that is proportional
to sin2(τΩ/2) , where τ is the time difference between the two arms of the
asymmetric interferometer arms (τ is 200 psec in Figure 2.7) and the Ω is
the angular frequency of the signal. This response places a limit on the fre-
quencies that can be properly sent down an RF link using phase modulation.
Along with the bandwidth limitation, the phase noise of the laser will also
play a role in the performance of the RF link. Like RF phase noise, the opti-
cal phase noise of a laser is often a measure of how far the light from the laser
is from ideal monochromatic light. Since the laser light is passed through an
asymmetric interferometer, any phase noise on the laser will be converted to
intensity noise. This added noise adversely affects the RF noise performance.
44 Applications of Modern RF Photonics

The added noise is shown in Figure 2.8 [10, 11]. Both of these drawbacks
limit the use of phase modulation in RF links.
Two main choices exist for intensity modulation. The first is the elec-
troabsorption modulator (EAM), which directly affects the intensity of the
light as a function of the RF signal that is applied. The EAM has previously
been investigated for uses in RF photonic links [12]. The EAM link, along
with a plot of the output power of an EAM as a function of applied voltage,
appears in Figure 2.9. The EAM has a transfer function as shown in (2.9),
where P­o is the maximum output power of the EAM, V is the applied voltage
to the EAM, Vo is the voltage when the output power is 1/e of the maximum
output power, and α is a constant of the EAM, with typical values between
1–2 [12]. The transfer function demonstrates the nonlinear relationship be-
tween optical power and voltage. From Figure 2.9, the slope efficiency can be
greater than one, overcoming the deficiency of the direct modulation of the
laser. Assuming α = 1, a simplified version of the RF gain with an EAM link
is shown in (2.10), where I­dc is the DC photocurrent generated at the photo-
detector from the incoming optical carrier, and Zin and Zout are the input and
output impedances, respectively. The slope of the EAM is highly nonlinear,
allowing for very strong spurious tones to be created. The OIP of the second
and third harmonics are shown in (2.11) and (2.12), respectively. One draw-
back is that the EAM has a strong second harmonic response, which often
limits the use of this component to single-octave operation. Other drawbacks
to using the EAM are the high insertion loss along with the low optical power
handling. Most EAMs can only accept a few tens of milliwatts of optical

Figure 2.8 The added RIN from a 100-kHz line width after an asymmetric interferometer
having a differential delay of 100 ps. (After: [4].)
 Analog Delay Lines 45

Figure 2.9 EAM link with transfer function of the EAM. (After: [4].)

power before they are permanently damaged. While recent advances have in-
creased the optical power handling up to 100 milliwatts, the insertion loss is
still large. Finally, the EAM requires temperature control as the semiconduc-
tor material is very temperature-dependent.

( )
α
- VV (2.9)
Pout = Po e o

2
I DC
Grf = Z out Z in (2.10)
Vo2

1 2
OIP2H = I dc Z out (2.11)
4

OIP3H = 12I dc2 Z out (2.12)

The other commonly used external intensity modulator is the Mach

Zehnder interferometer-based modulator (MZM). While other modulators
also exist, they often have the same limitations as the modulators listed above.
46 Applications of Modern RF Photonics

For these reasons, the lithium niobate-based (LiNbO3) MZM is the preferred
solution for externally modulated RF photonic applications. The LiNbO3
material requires no external temperature control and can handle well over
100 milliwatts of optical power. The insertion loss can be made as small as 3
dB, and the modulator can be driven with up to a watt of RF power before
failing. Finally, the MZM configuration can be operated at a bias point where
the RF even-order nonlinearities are nulled. For all of these advantages, the
MZM is the chosen modulator for most RF applications.
The sinusoidal transfer function of the MZM, along with the link ar-
chitecture, is shown in Figure 2.10. The MZM operates by splitting the light
between two parallel paths. On each path, the applied RF signal acts through
the electro-optic effect in LiNbO­3 to either retard or advance the light. This
results in a complementary phase modulation of the light in each arm. The
coupler at the end of each path then recombines the lights, allowing them to
interfere with each other. The resulting output from the MZM is an intensity
modulation of the light from the laser. From a traditional RF point of view,
the MZM is simply a mixer, yet one that operates at 192 THz. The optical
carrier from the laser mixes with the RF signal to upconvert the signal into
the optical domain. In the frequency domain, the result is double-sideband
modulation of the optical carrier, as seen in Figure 2.10. The sidebands are
symmetrically separated from the optical carrier by the frequency of the RF
signal. After transmission of the modulated light, the photodetector simply
downconverts the RF signal from the optical domain back to its original

Figure 2.10 MZM link with transfer function of the MZM and optical spectrum of the dual
sideband modulation. (After: [4].)
 Analog Delay Lines 47

frequency. Since the optical carrier arrives at the photodetector along with
the sidebands, no external local oscillator is required to recover the RF signal.
This type of RF photonic link is referred to as an intensity-modulated direct
detection (IMDD) link.
One of the key advantages of the IMDD link using an MZM is the lin-
ear performance. When the DC bias point that provides a phase shift of π/2
is chosen, the even order terms of the nonlinear transfer function becomes
zero. This operating point is often referred to as the quadrature bias point.
One of the drawbacks of using an MZM is keeping the modulator biased
at quadrature. Often the bias point will drift and requires an external bias
control circuit to keep the MZM at the quadrature bias point. Two popular
methods exist for keeping the MZM at the quadrature bias point: optical
power monitoring or RF second harmonic power monitoring.
The optical power monitoring works for a dual-output MZM by tap-
ping off a small amount of the light from each of the optical outputs. The
ratio between the powers is then used in a feedback loop to keep them at the
same ratio. Ideally, the power in both arms would be equal, but for practical
MZMs that do not have an infinite extinction ratio, the ratio has to be deter-
mined for each MZM to remain at quadrature bias. For an MZM that does
not have two optical outputs, the optical power ratio is measured between the
input of the MZM and the output. Either method requires measuring the
losses of the optical taps as well as the loss of the MZM and then setting the
appropriate ratio for the feedback loop to maintain the quadrature operating
point. This type of bias controller is often referred to as ditherless control.
The RF second harmonic power monitoring technique works by in-
troducing a known RF frequency into the MZM. The bias controller then
taps off a small amount of the light from the optical output and measures
the power of the second harmonic of the RF frequency that was injected into
the MZM. The feedback loop acts to minimize the second harmonic power,
which ideally occurs at the quadrature bias point. This type of bias control-
ler is referred to as dither-based control, since a frequency dither or tone is
added to the MZM. The dither bias controller can automatically find the
quadrature bias point without any manual tuning. The disadvantage is that
the addition of an RF tone can potentially corrupt the RF signals that are be-
ing stored in the analog delay line. The choice of bias controller depends on
the application. The RF tone can also be used to improve the RF performance
of the photodetection process, as discussed in Chapter 3.
The resulting RF metrics for this type of link assume that the quadra-
ture bias point is selected. As it turns out the RF gain is also maximized when
the link is operated at the quadrature bias point. The resulting RF metrics are
shown (2.13a)–(2.13c), which assume that shot noise is the dominant noise
48 Applications of Modern RF Photonics

term. In these equations, e is the electron charge, Vp is the modulation effi-

ciency of the MZM, I­dc is the DC photocurrent generated at the photodetec-
tor from the incoming optical carrier, and Zin and Zout are the input and out-
put impedances, respectively. The equations offer a translation between RF
performance and the photonic metrics of the link. From a cursory glance, the
main photonic metrics that improve the RF performance are the Vπ and the
Idc. As an example, the RF gain as a function of DC photocurrent at different
Vπ levels is plotted in Figure 2.11. Clearly, the higher the DC photocurrent,
the higher the RF gain, while the lower the Vπ the higher the RF gain as well.
The MZM not only plays a significant role in determining the RF gain, but
also in the SFDR. From Figure 2.10(a), as the Vp decreases, the linear portion
of the sinusoidal transfer function shrinks. This in turn reduces the SFDR.
The tradeoff in RF gain, NF, and SFDR as a function of Vπ has been dis-
cussed in other works [13]. Knowing that the Vp can only be reduced so far,
the DC photocurrent will give us the best option to improve the RF metrics.

π2 Z in Z out
Grf = I dc2 (2.13a)
V π2

1 2eV π2
NFrf ,shot = (2.13b)
I dc π2kBTZ in

23
 2
SFDRimd 3,shot = I dc2 3   (2.13c)
 
e

Figure 2.11 RF gain as a function of DC photocurrent for different MZM Vp.

 Analog Delay Lines 49

Often there is a question about why the DC photocurrent is the domi-

nant term rather than the RF photocurrent. As seen in Figure 2.10, the ma-
jority of the optical power is contained in the optical carrier of the modulated
laser and not the sidebands. At the photodetector, the optical carrier mixes
with the two sidebands to generate the RF. Thus the optical carrier, which
acts like a local oscillator for the mixing at the photodetector, is the dominant
term for the RF performance. This leads to the DC photocurrent being the
dominant term.

2.4 RF Photonic Component Performance

Having determined the type of modulator to use, the next photonic compo-
nent to discuss is the laser. The laser provides the optical carrier upon which
the RF signal will be upconverted. For an IMDD link, the laser is quantified
by two main factors: the output power and the relative intensity noise (RIN).
The output power is directly proportional to the DC photocurrent, as the
more light that the laser can generate and deliver to the photodetector will
both improve the RF gain and NF of the photonic delay line. While the fact
that higher RF gain comes with higher optical power intuitively makes sense,
the reduction in RF NF is not so obvious. In order to understand this, the re-
lation of shot noise to incoming optical power must be described. Shot noise
arises from the random fluctuations in the photocurrent that come from the
random arrival time of the photons at the photodetector. The fundamental
nature of shot noise is well investigated [14]. For the purpose of this discus-
sion, the important relation to note is that the level of the shot noise in terms
of RIN will decrease with increased DC photocurrent [3]. Thus, the RF NF
will improve as the DC photocurrent increases. While increasing the DC
photocurrent will ideally continuously lower the shot noise level, eventually
the RIN of the laser itself will exceed the shot noise limit. The laser RIN will
set a lower limit on the level of the noise floor in the system. Table 2.1 shows

Table 2.1
Minimum NFs for Different Laser
RIN Levels
Laser RIN Min RF Noise Figure
(dBc/Hz) (dB) at 10 mA
–145 44
–155 34
–165 27
–175 24
50 Applications of Modern RF Photonics

the minimum RF NF for different laser RIN levels. Thus, measuring the
laser’s RIN is very important to predict the RF NF of the photonic link. A
plot of the RIN of two different types of lasers appears in Figure 2.12. Some
lasers have very low RIN over frequencies as high as 10 GHz. Others have
RIN levels that are quite high at lower frequencies. The choice of laser is often
determined by the RIN level at a given frequency and the application’s needs.
The distributed feedback (DFB) semiconductor-based laser in general has
the best noise performance over wide frequency bandwidths and is often the
preferred solution for most systems. Typical DFB lasers can have RIN levels
around –155 dBm/Hz over frequency ranges in the tens of gigahertz. At the
same time, they are often limited to optical powers of 100 mW.
In order to increase the optical power while keeping the RIN low, an
optical amplifier can be used. Not all RF photonic links require an optical
amplifier. A photonic link that requires no optical amplifier is designated as
a passive optical network (PON). PONs have been investigated for short- to
medium-length links [15]. Examples include fiber-to-the home links as well
as some antenna-remoting links. PONs are used when it is impractical to
place an amplifier somewhere in the link. Some uses cannot provide electrical
power for an optical amplifier, while others cannot afford the monetary cost
of an amplifier. In the case of a photonic delay line, amplifiers are required.
Even though the loss of the optical fiber is very low, the required time delay
for the analog memory application requires very long lengths. For a delay of
100 µs, 20 km of optical fiber is necessary. The minimum optical loss would
be 4 dB. From the above equations on RF gain, every 1-dB reduction of DC

Figure 2.12 RIN of two different types of lasers. (After: [4].)

 Analog Delay Lines 51

photocurrent or optical power results in a 2-dB loss of RF power. So in 20

km, the RF power will be decreased by approximately 85%. Therefore the de-
lay line would require optical amplification in order to deliver the appropriate
optical power to the end of the delay line. Since all parts of the delay line are
accessible, one or more optical amplifiers can be placed at appropriate points
in the delay line. The optical gain of the amplifier will then compensate for
the loss of the optical fiber. The drawback is that the optical amplifier will add
additional optical noise.
While optical amplification can be accomplished by multiple means,
the most commonly used amplifier is the EDFA. While many reasons exist to
use EDFAs over other types of optical amplifiers [16], the two most impor-
tant ones are the low optical NF and the extremely slow time response of the
gain medium as compared to the other types of amplifiers. The added optical
noise of the EDFA affects the RF NF of the overall photonic link. However,
the optical NF of the EDFA does not directly translate to the RF NF. A noise
penalty metric was defined in order to address how the optical noise adds to
the RF NF [17]. EDFAs will typically have a noise penalty between 3 and 24
dB. The result of the noise penalty on the RF NF as a function of DC photo-
current is shown in Figure 2.13. Depending on the application and the noise
penalty of the amplifier, the use of an EDFA could provide the necessary
optical power for the analog delay line. Unfortunately, the other photonic
components will limit the optical power.
After the optical amplifier, the next photonic component is the optical
fiber that connects the laser and modulator to the photodetector. As described
in Chapter 1, the loss of the optical fiber is already optimized at a value near
0.2 dB/km. After the loss of the optical fiber, the maximum input power the
optical fiber can accept is the next important parameter for use in photonic

Figure 2.13 RF NF as a function of DC photocurrent at different noise penalties. (After: [4].)

52 Applications of Modern RF Photonics

links. As discussed previously, the various RF metrics improve as the optical

power delivered to the photodetector increases. However, the optical fiber can
only accept a specific amount of optical power before nonlinear optical effects
will occur. Once a threshold power is exceeded, the nonlinear optical effects
will begin to grow exponentially and affect the performance of the photonic
link. The nonlinearity with the lowest threshold power in long lengths of
standard optical fibers is the stimulated Brillouin scattering (SBS) effect. The
SBS effect is well understood and described in detail in other works [18, 19].
For RF photonic applications, the SBS effect is often characterized by an
increase in the reflected optical power at the input of the optical fiber. While
Rayleigh scattering will result in some optical power being reflected back to
the input of the fiber rather than traveling down to the end of the fiber, it
has a linear response as a function of the input optical power. When the SBS
threshold is reached, the reflected power increases exponentially as a function
of input optical power as seen in Figure 2.14. At the same time, the power
received at the end of the fiber no longer increases but flattens as a function
of the input optical power, placing a limit on the amount of optical power
that can be transmitted down the fiber. The threshold power for the SBS
effect is inversely proportional to the length of the optical fiber link. Since
analog delay lines require very long lengths of fiber, the SBS threshold can be
quite low. While the SBS effect limits the amount of optical power that can
be transmitted, it also has another disadvantage. When the SBS threshold is
reached, the added noise becomes quite large. The added noise is shown in
Figure 2.15. The optical noise is shot noise-limited when the optical power
is below the SBS threshold. However, when the SBS threshold is reached, the

Figure 2.14 Output and reflected power from optical fiber as function of input power.
(After: [4].)
 Analog Delay Lines 53

Figure 2.15 Optical noise spectrum above and below the SBS threshold (After: [4].)

added noise level increases significantly. Thus, the SBS effect is a significant
problem that has to be designed around.
Another parameter of the fiber link is the chromatic dispersion. The
dispersion in optical fibers acts like an optical frequency-dependent time de-
lay, which depends on the length of the fiber. Simply put, different optical
frequencies will travel at different speeds through the same length of fiber.
In the case of an IMDD link, the optical carrier and the upper and lower
sidebands (as seen in Figure 2.10) will all arrive at slightly different times. As
the RF frequency increases, the sidebands will spread further apart from each
other, causing them to become out of phase with each other. As the phases
between the sidebands change, the recovered RF will be degraded. This effect
is known as RF fading [20]. The results are shown in Figure 2.16, with the
RF power shown as a function of RF frequency. For a given length of fiber
with a known dispersion constant, the RF power will go from a peak to a
null and back again. Much like the phase-modulated link with asymmetric
interferometer detection, the chromatic dispersion places a frequency limita-
tion on the link.
An additional effect that occurs due to the chromatic dispersion is the
increase in the second-order distortion in the analog photonic link. Normally,
the even-order nonlinear distortion is nulled in an IMDD link with an MZM.
However, in the presence of chromatic dispersion, the even-order distortions
are no longer nulled [21]. The increase in the second harmonic power as a
function of frequency is plotted in Figure 2.17. While the fundamental signal
54 Applications of Modern RF Photonics

Figure 2.16 RF fading as a function of frequency due to chromatic dispersion. (After: [24].)

Figure 2.17 Fundamental and second harmonic response due to chromatic dispersion.
(After: [4].)

power is not degraded noticeably for lower frequencies, the second harmonic
power immediately grows and becomes very significant. Again, this places an
operating frequency limitation on the analog photonic link. The chromatic
dispersion of the optical fiber link is the cause of both of these deleterious ef-
fects. Chapter 3 discusses methods for mitigating this problem.
 Analog Delay Lines 55

Finally, the last photonic component is the photodetector itself. The

first metric of importance regarding the photodetector is the responsivity.
If the slope efficiency of the laser is a measure of the optical power emitted
as a function of the drive current, the responsivity of the photodetector is a
measure of the photocurrent generated as a function of the optical power in-
put to the device. The higher the responsivity, the more photocurrent that is
generated for a given optical power. Just like with the laser, the responsivity is
usually less than one. Also just like the laser, the slope is not completely linear.
This leads to the photodetector adding spurious signals at the output. This
nonlinearity is often overlooked, as the modulator will often add stronger
spurs than the photodetector. However, recent work has shown when pho-
tonic links can be limited by the photodetector nonlinearities [22].
The other metrics of interest for the photodetector are the optical pow-
er handling and the frequency response. Often the power handling and the
frequency response cannot be simultaneously optimized. A common reason
for this is due to the dimensions of the device. In order to accept higher op-
tical powers, the device often requires a larger absorption region. However,
the larger absorption region means that the device has longer transit times
for the generated electrons to be swept out of the device. This reduces the
frequency response. Newer devices can overcome these limitations, allowing
for both higher power handling as well as increased frequency response [23].
These photodetectors can be very useful for many RF applications, as they
can meet the required frequency and power handling necessary for delay line
applications.

2.5 Conclusions
The analog delay line serves an important purpose in the demonstration ar-
chitecture. It provides a means for storing the original RF spectrum for future
processing needs. Due to the required time delays, an analog delay line is an
ideal application for an RF photonic link. While many different versions of a
photonic link can be used, the external intensity modulation combined with
direct detection is the preferred option. Having determined the type of pho-
tonic link that will be used, the RF metrics as a function of the photonic com-
ponent performance can be defined. This chapter has described the differ-
ent photonic components metrics in relation to optimizing the performance
of the photonic link. Chapter 3 covers advanced techniques for overcoming
the individual component limitations. These advanced techniques allow for
high-performance RF photonic links to be realized.
56 Applications of Modern RF Photonics

References
[1] Friis, H. T., “Noise Figures of Radio Receivers,” Proceedings of the IRE, Vol. 32, No. 7,
1944, pp. 419–422.
[2] Pozar, D. M., Microwave Engineering, Hoboken, NJ: John Wiley and Sons, 1998.
[3] Urick, V. J., et al., “Long-Haul Analog Photonics,” Journal of Lightwave Technology,
Vol. 29, No. 8, 2011, pp. 1182–1205.
[4] Devgan, P., “Review of Radio Frequency Photonics Basics,” AFRL Report, AFRL-RY-
WP-TR-2017-0157, 2017.
[5] Navid, R., et al., “Close-In Phase Noise in Electrical Oscillators,” Proc. SPIE Symposium
Fluctuations and Noise, 2004, pp. 27–37.
[6] Cox, C. H., et al., “Limits on the Performance of RF-Over-Fiber Links and Their Im-
pact on Device Design,” Transactions on Microwave Theory and Techniques, Vol. 54, No.
2, 2006, pp. 906–920.
[7] Urick, V. J., et al., “Phase modulation with Interferometric Detection as an Alternative
to Intensity Modulation with Direct Detection for Analog-Photonic Links,” Transac-
tions on Microwave Theory and Techniques, Vol. 55, No. 9, 2007, pp. 1978–1985.
[8] Chi, H., X. Zou, and J. Yao, “Analytical models for Phase-Modulation-Based Micro-
wave Photonic Systems with Phase Modulation to Intensity Modulation Conversion
Using a Dispersive Device,” Journal of Lightwave Technology, Vol. 27, No. 5, 2009, pp.
511–521.
[9] Chain, H.-C., et al., “Optical Millimeter-Wave Generation and Transmission Without
Carrier Suppression for Single- and Multiband Wireless Over Fiber Applications,” Jour-
nal of Lightwave Technology, Vol. 28, No.16, 2010, pp. 2230–2237.
[10] Salehi, M. R., and B. Cabon, “Theoretical and Experimental Analysis of Influence of
Phase-to-Intensity Noise Conversion in Interferometric Systems,” Journal of Lightwave
Technology, Vol. 22, No. 6, 2004, pp. 1510–1518.
[11] Tkach, R., and A. Chraplyvy, “Phase Noise and Linewidth in an InGaAsP DFB
Laser,” Journal of Lightwave Technology, Vol. 4, No. 11, 1986, pp. 1711–1716.
[12] Cox III, C., Analog Optical Links, Cambridge, England: Cambridge University Press,
2004.
[13] Bucholtz, F., et al., “Graphical Approach for Evaluating Performance Limitations in
Externally Modulated Analog Photonic Links,” Transactions on Microwave Theory and
Techniques, Vol. 56, No. 1, 2008, pp. 242–247.
[14] Haus, H, A., Electromagnetic Noise and Quantum Optical Measurements, New York, NY:
Springer, 2012.
[15] Lam, C., Passive Optical Networks Principles and Practice, Cambridge, MA: Academic
Press, 2007.
[16] Desurvire, E., Erbium Doped Fiber Amplifiers Principles and Applications, New York,
NY: John Wiley and Sons, 2002.
 Analog Delay Lines 57

[17] Urick, V. J., F. Bucholtz, and K. J. Williams, “Noise Penalty of Highly Saturated
Erbium-Doped Fiber Amplifiers in Analog Links,” Photonics Technology Letters, Vol. 18,
No. 6, 2006, pp. 749–751.
[18] Kittel, C., Introduction to Solid State Physics, New York, NY: John Wiley and Sons,
1996.
[19] Mao, X. P., et al., “Stimulated Brillouin Threshold Dependence on Fiber Type And
Uniformity,” Photonics Technology Letters, Vol. 4, No. 1, 1992, pp. 66–69.
[20] Schmuck, H., “Comparison of Optical Millimeter-Wave System Concepts with Regard
to Chromatic Dispersion,” Electronics Letters, Vol. 31, No. 21, 1995, pp. 1848–1849.
[21] Meslener, G., “Chromatic Dispersion Induced Distortion of Modulated Monochromatic
Light Employing Direct Detection,” Journal of Quantum Electronics, Vol. 20, No. 10,
1984, pp. 1208–1216.
[22] Williams, K. J., “Nonlinear Mechanisms in Microwave Photodetectors Operated with
High Intrinsic Region Electric Fields,” Applied Physical Letters, Vol. 65, No. 10, 1994,
pp. 1219–1221.
[23] Williams, K. J., and R. D. Esman, “Design Considerations for High-Current
Photodetectors,” Journal of Lightwave Technology, Vol. 17, No. 8, 1999, pp. 1443–1454.
3
Advancements in Analog Delay Line
Performance
Chapter 2 identifies the analog buffer as one of the applications where RF
photonics can provide an advantage over other solutions. Specifically, an RF
photonic link can be used as an analog delay line, allowing for the storage of a
wide spectrum of signals for up to hundreds of microseconds. In fact, Chap-
ter 2 makes the case for using an IMDD link as the preferred solution for the
delay line and defines the RF metrics for this architecture as a function of the
photonic components’ performance.
Given the definition of IMDD link performance, the only way to im-
prove the performance is to increase the optical power, decrease the electro-
optic conversion efficiency (measured by the Vp), or reduce the optical noise
from the laser (and optical amplifier if used). However, these actions all re-
quire improving the individual photonic component performance. Perhaps
there are other means to improve the performance without changing the in-
dividual photonic components? This chapter presents advanced techniques to
improve the RF performance of the analog delay line without fundamentally
changing the individual components. In doing so, this chapter provides RF
photonic link designers with additional tools to use when building a potential
system. While the focus is on the analog delay line, the techniques presented
here will also improve any RF photonic link.

59
60 Applications of Modern RF Photonics

3.1 Performance Improvement Through the Photonic Components

The RF metrics are given in the logarithmic scale in (3.1a)–(3.1d), where Vπ
is the half-wave voltage of the MZM in units of volts, Idc is the DC photocur-
rent in units of milliamps, and Zin and Zout are the input and output imped-
ance, respectively. The assumptions for these expressions are that the MZM is
biased at the quadrature bias point, that the photodiode is completely linear
in its response, and that the input and output impedances of the MZM and
the photodetector are 50Ω. If the shot noise is the dominant source, the RF
metrics can be further simplified. Equation (3.2a) is the shot noise contribu-
tion in logarithmic scale. Substituting (3.2a) into (3.1b)–(3.1d) yields the
simplified forms shown in (3.2b) and (3.2c), with e as the electron charge.
From (3.1a), the RF gain is dependent on two parameters, the DC photo-
current (Idc­) from the photodetector and the Vπ of the MZM. Similarly, the
RF noise figure is dependent on at least one of the two parameters. With the
above assumptions, one simple way to improve the RF metrics is simply to
increase the optical power delivered to the photodetector. Examples of RF
photonic links with improved RF gain and NF have been previously dem-
onstrated [1–3]. The increased optical power can be accomplished by using
multiple lasers as well as using high-power optical amplification to increase
the total power delivered to the photodetector. While the increased optical
power will improve the RF metrics, limitations exist on how much improve-
ment can be realized.

G rf [dB] = −22 + 20 log (I dc [mA ]) − 20 log (V π [ V ]) (3.1a)

NFrf [dB] = 174 − Grf + N total [ dBm Hz ] (3.1b)

OIP3 [dBm ] = −13 + 20 log (I dc [mA ]) (3.1c)

SFDR 3 [dB] = 2 3 (OIP3 − N out ) (3.1d)

N shot [ dBm Hz ] = 10 log10 (0.5eI dc Z out )

(3.2a)
= −174 + 10 log10 (I dc [mA ])
 Advancements in Analog Delay Line Performance 61

NFrf [dB] = 22 − 10 log (I dc [mA ]) + 20 log (V π [ V ]) (3.2b)

SFDR 3 [dB] = 2 3 161 + 10 log (I dc [mA ]) (3.2c)

3.2 Improvements in the Photodetector

One of the first limitations is on the photodetector itself. Photodetectors
with a frequency response above 20 GHz can operate linearly when the in-
put optical power is kept at 5–10 mW or lower. Beyond this optical power,
the typical photodetector will no longer operate with a linear response and
will often be destroyed. A common failure mechanisms is overheating due to
the intense amount of light being focused on the small active region of the
device. To overcome this problem, new designs for photodetectors have been
developed. One method involves the use of a gradient index (GRIN) lens
before the photodiode to distribute the optical power more uniformly over
the active region [4]. This allows the photodetector to handle more optical
power before the heating problem destroys the device. Other photodetectors
with a frequency response above 20 GHz and operating input powers in the
100-mW range have been shown [5–8]. Using a high-power handling pho-
todetector, an RF link with RF NFs below 10 dB at frequencies as high as 10
GHz being achieved [9].
While photodetectors are often modeled as linear devices, they do have
nonlinear responses that can adversely affect the RF metrics of an analog de-
lay line. One way that photodetectors can adversely affect analog delay lines
is in the phase noise performance. Recall from Chapter 2 that some analog
delay lines can be used to model long-distance transmission and that the
added phase noise needs to be minimized. Previous work in distributing and
generating low-phase noise clock signals using photonics have shown that
the photodetector can be the limiting component, due to intensity-to-phase
noise conversion in the photodetector [10–12]. In order to minimize this
effect, a method has been proposed to optically saturate the photodetector
[13]. By compressing the photodiode, the intensity-to-phase response can be
clamped and thus the added phase noise can be reduced. The results of the
phase noise performance when the photodiode is saturated or not appears
in Figure 3.1. By correctly choosing the saturation point of the photodiode,
the phase noise is improved by over 10 dB in the 10 kHz–1 MHz frequency
62 Applications of Modern RF Photonics

Figure 3.1 Phase noise of a photodetector in and out of saturation. (After: [14].)

offset range. When used in a 14-km long delay line, the overall timing jitter
is within 0.3 fs (femtoseconds) of the case where there is no delay line. The
trade-off for saturating the photodetector is a reduction in available RF power
from the output of the device. A method for overcoming this drawback is ad-
dressed in Section 3.5.2.
Along with the added phase noise, the response of the photodetector
does add second- and higher-order nonlinear distortions. The nonlinearity
originates from the voltage dependent responsivity of the photodetector and
has been studied [15]. Work has been done to make photodetectors with bet-
ter linearity, with reports of measured output intercept points of the second
harmonic (OIP22H) of 50 dBm [15, 16]. However these are noncommercial,
specialty photodetectors, and the measurements were limited to the 1–2-
GHz frequency range. These effects can limit the overall RF performance of
the analog delay line and have to be considered in designing the delay line.
The required nonlinear performance for the photodetector has been investi-
gated previously [17]. In order to overcome the third-order nonlinearities due
to the photodetector, predistortion techniques have been used at the modu-
lator [18–20]. Often, these involve added electronic components and come
at the cost of a reduction in the power of the fundamental RF frequency. A
photonic configuration for canceling the second-order distortion of the pho-
todiode is discussed in detail later in Section 3.5.5.
 Advancements in Analog Delay Line Performance 63

While work continues to address the deficiencies of single photodetec-

tors, the analog delay line performance can be improved through the use of
balanced detection. Balanced photodetection can provide several benefits. If
one photodetector can handle a certain amount of optical power, then two
photodetectors can handle twice as much optical power. For example, the
increase in RF gain can be 6 dB, as has been shown in phase-modulated links
combined with balanced detection [21]. While balanced photodetection is
common for phase-modulated links, it can be used in intensity-modulated
links as well. Since the MZM intensity modulator can be made with two op-
tical outputs, each output can be then connected down an independent fiber
to the inputs of the balanced photodetector. The resulting link would provide
the same increased RF gain as the phase-modulated link. Demonstrations of
links using parallel optical fiber paths have been shown in the lab [22]. Note
that the two fiber-optic paths have to be length-matched to less than a single
wavelength of the highest frequency at which the link will operate. While
such a link could be hard to realize in real-world environments, there have
been demonstrations using multiple parallel links combined with balanced
photodetection [23]. Further studies have also shown that colocated optical
fibers can have relatively small phase changes over large temperature ranges,
especially if they are buried [24]. For the analog delay line, the temperature
and fiber lengths can be properly controlled to make use of the benefits of
balanced detection.
Balanced photodetection can also provide a way to reduce the impact of
laser RIN on the RF metrics. Note that the two optical outputs of the MZM
are 180° out of phase with each other. When the two outputs are connected
to the balanced photodetector, the photocurrent generated from the indi-
vidual photodetectors is also 180° out of phase with each other. The result
is that the recovered RF signal now adds coherently. On the other hand, the
laser noise is in-phase on the two MZM outputs. Therefore, the noise will
not add in the balanced photodetector but will instead be subtracted. This is
known as common-source RIN rejection [25]. So not only does the RF gain
increase, the noise is simultaneously suppressed. The amount of noise that is
suppressed depends on the source and how well matched the two paths are.
For analog delay lines, this type of architecture can help bypass the limits on
RF NF that the RIN of the laser imposes. Since the RIN is from a common
source in both paths, it can be suppressed. Balanced photodetection has been
well investigated for improving RF NF [26].
In addition, the use of balanced photodiodes can cancel the even-order
distortion due to the individual photodiodes [22]. However, each photodiode
will react differently and the cancelation will not be optimized when the bias
voltage of each photodiode is the same. In practice, each photodiode’s bias
64 Applications of Modern RF Photonics

is set to optimally cancel the even-order distortion. The optimal bias voltage
depends on the analog delay line architecture. Using the dither tone from
the MZM bias controller, the bias voltages of each of the photodiodes can be
automatically tuned to minimize the even-order distortion due to the photo-
diodes. A demonstration using a microcontroller to optimize the bias voltages
showed an improvement of 8.3 dB in the second harmonic distortion, while
the fundamental RF power, third harmonic distortion, and RF NF are not
degraded [27].
The system works by using a microcontroller to scan the optimal bias
voltage in order to minimize the second harmonic of the dither tone. Figure
3.2 shows the power of the second harmonic as the bias voltage changes. After
the minimum second harmonic power is found, the microcontroller sets the
correct bias voltages for the photodiodes. Using the equation for the required
OIP2 in order to remain third-order limited OIP2req(dBm) = 47 + 70/3 *
log10 (Idc(mA)) [27], the achieved OIP2 can provide third-order limited per-
formance for photocurrents up to Idc = 33.5 mA.

3.3 Improvements in the Optical Fiber

With multiple methods to overcome the limitations of the photodetector, the
next component that needs to be overcome is the optical fiber. Due to the
nonlinear SBS effect, incoming light above the SBS threshold will limit the
light that can be sent down the fiber, reflecting any extra light back to the
input. The dispersion in the optical fiber will also add impairments to the RF

Figure 3.2 Improvement in second harmonic changing the bias voltage. (After: [14].)
 Advancements in Analog Delay Line Performance 65

signal. To combat both of these problems, multiple solutions exist. A disper-

sion compensation fiber module can be used at the end of the optical fiber
to correct the impairments to the original signal. Such modules are common
in long-haul digital fiber-optic networks [28–31]. However, they add loss
and do not address the SBS effect. To increase the SBS threshold in optical
fibers, different approaches have been introduced, ranging from changing the
temperature of the fiber [32], introducing strain [33], and adding another
waveguide to decouple the acoustic traveling wave from the optical light [34].
Changing the temperature of the fiber can be done in a laboratory environ-
ment, but it is not very practical for use in real-world applications. Fabricat-
ing custom fibers increases the cost and may only be produced in limited
lengths.
One method that has been explored to increase the amount of power
before triggering the SBS effect is the use of alternating pairs of fiber spans
with different SBS spectra [35, 36]. Unfortunately, the increase is only around
2 dB as compared to a single fiber type. To increase the input power even fur-
ther, optical isolators can be placed between each alternating pair. The isola-
tors act to make each alternating pair of fibers independent from each other.
Moreover, by choosing two optical fibers with opposite dispersion signs, the
overall dispersion of the fiber link can be made close to zero. As shown in Fig-
ure 3.3, an example link shows the use of an optical fiber with a dispersion of
–4ps/(nm•km) and another optical fiber with a dispersion of +8ps/(nm•km).
By using a length of fiber with the lower dispersion that is twice the length
of the higher dispersion fiber, the overall fiber will have a dispersion close
to zero. The combination of fibers with the isolators also increases the SBS
threshold. As seen in Figure 3.4(a), the alternating fiber with optical isolators
increases the SBS threshold by 6 dB when compared to a single optical fiber
span. In tandem, the optical power that can be transmitted down the fiber
is increased by 6 dB when compared to a single fiber, as illustrated in Figure
3.4(b). Thus, the alternating fiber span addresses both the dispersion and the
SBS issues. The improvements do come at a cost. Overall fiber losses increase
by adding the isolators and splicing the different types of fiber together. In
addition, the configuration cannot be used when access to the fiber is not
available.

Figure 3.3 Optical fiber span with isolators in-between. (After: [14].)
66 Applications of Modern RF Photonics

Figure 3.4 (a) Measure of the reflected optical power below and above the SBS threshold
and (b) transmitted optical power below and above the SBS threshold. (After: [14].)

3.4 Improvements in the Optical Amplifier

The optical amplifier is the next component that needs to be addressed. The
optical amplifier can take the optical power from a low-power laser and in-
crease it significantly. The increase in power comes at a cost. One detriment
 Advancements in Analog Delay Line Performance 67

that is often overlooked is the addition of spurious signals. Optical amplifica-

tion systems that have very fast response times can have their gain modulated
by the RF signal that is present on the incoming optical light that is to be
amplified. While these effects can be useful for other applications [37, 38],
they seriously degrade the performance of analog systems. Semiconductor- or
nonlinear-based optical amplification (e.g., four wave mixing) has response
times ranging from picoseconds to femtoseconds. The EDFA, on the other
hand, has response times on the order of milliseconds, making it ideal for
analog photonic links [39]. The EDFA is the preferred optical amplifier for
RF photonic applications.
Another cost to using an optical amplifier is the added noise. The added
noise from the EDFA will degrade the RF metrics of the photonic link. While
the EDFA’s optical NF is well understood [40], the effect the optical NF has
on the RF metrics is not as clearly defined. In order to address this issue, the
RF noise penalty metric was developed [41]. The noise penalty of the EDFA
is measured using a similar method to measuring the RIN of a laser. The pow-
er spectral density is recorded and compared to the shot noise limit for the
operating photocurrent. The difference is then called the noise penalty. An
example measurement is shown in Figure 3.5. The noise penalty will affect
the RF metrics as shown in (3.3a)–(3.3c), where j1dB is the RF phase shift in
the MZM which drives the link’s output power 1 dB into compression and
RINsh­ is the shot noise relative to the photocurrent. The RINsh is the domi-
nant noise term for photocurrents between 1 and 100 mA. For a delay line of
100 µs or more, multiple EDFAs could be used in the link. To calculate the
overall noise penalty from multiple EDFAs, a cascaded noise penalty formula
was derived [42]. The cascaded NF depends on the gain of each EDFA, the

Figure 3.5 Noise penalty of an EDFA. (After: [14].)

68 Applications of Modern RF Photonics

noise penalty of each EDFA, and the loss between each EDFA. An example
link is shown in Figure 3.6, along with the cascaded noise penalty equations
in (3.3a)–(3.3c). Equation (3.4) has the overall noise penalty as a function of
lo, the loss before the photodetector, Ln; the optical loss of each length of fiber,
Gn; the gain of each EDFA; and NPn, the noise penalty of each EDFA. The
measured noise penalty of the link was shown to be in very good agreement
with the predicted noise penalty from the cascaded noise penalty equation.

 (RIN sh )V π2 
NFrf, amp = NP ⋅ NFrf, sh = 2  NP (3.3a)
 π kBTZ in 

2
CDR sh j1dB
CDRamp = = (3.3b)
NP 2(RIN sh )(NP )

23
SFDR sh  4 
SFDRamp = =
 (RIN sh )(NP )
23 (3.3c)
NP

2q
RIN sh = (3.3d)
I dc

NPc ≈ 1 + l o [L 2 • • •Gn (NPEDFA, 1 − 1) + .....

+LnGn (NPEDFA, − 1) + (NPEDFA, − 1)] . (3.4)
n −1 n

With a method for determining the amount of noise penalty the EDFA
will introduce to the RF photonic link, the next step is to determine the
best operating point for the EDFA. Previous work has shown that an EDFA

Figure 3.6 Noise penalty link. (After: [14].)

 Advancements in Analog Delay Line Performance 69

operating in saturation will reduce the noise penalty significantly, due to the
reduction in spontaneous emission [43]. The EDFA will enter saturation as
the optical power at its input increases. The noise penalty of the EDFA will
decrease as the amplifier goes further into compression. An example of the
noise penalty of an EDFA as a function of input power is shown in Figure
3.7. Note that the trade-off is that an EDFA in compression will not provide
as much optical gain as when it is operating in the linear regime.

3.5 Improvements in the Optical Modulator

Having addressed the EDFA, optical fiber, and photodetector, the next step
is to look at the optical modulator. The key metrics are the insertion loss and
the modulator efficiency (Vp). The insertion loss will limit the amount of
optical power that appears at the output of the modulator. This in turn will
reduce the amount of photocurrent at the output of the analog delay line. In
addition, the reduced output power will make it harder to saturate the EDFA
if it is placed after the MZM. So the insertion loss is just as important as the
Vp of the modulator. The Vp is dependent on both the material that is used
along with how well the RF electric field is matched to the optical field [44].
Since both the Vp and insertion loss of the device are set at the time of manu-
facture, it cannot be changed by the analog delay line designer. Instead, the
MZM can be used in different ways to overcome the Vp and insertion loss.

3.5.1 Off-Quadrature Biasing of the Optical Modulator

Another method for operating the MZM to improve the RF metrics is to
change the DC operating bias point. As stated previously, the quadrature bias

Figure 3.7 Noise penalty as a function of the input power to the EDFA. (After: [14].)
70 Applications of Modern RF Photonics

point is the preferred operating point for an MZM intensity-modulated link.

For an unamplified link, quadrature-biasing results in the highest RF gain as
well as providing the conditions for canceling the even-order distortions after
photodetection. However, in a link with an MZM followed by optical ampli-
fication, the quadrature bias may no longer provide the highest RF gain. In
fact, the highest RF gain occurs at a point below the quadrature bias point.
This is often referred to as low-biasing of the MZM. The exact operating
point is different for each MZM, as it depends on the extinction ratio and Vπ
of the MZM along with the gain and Po, the output saturation power of the
EDFA. The relation between RF gain and Idc and Vπ appears in (3.5), where
R is the responsivity of the photodetector.

π2 sin 2 ( jdc ) Z in Z out

Grf = I dc2 2
  I  
g
(3.5)
V π 1 + 
2 dc
1 − cos ( jdc ) 
  RPo .max  


The increased RF gain for operating at the low bias point comes from
the improved optical gain of the amplifier out of compression. As seen in
Figure 3.8, when the MZM is operated at quadrature bias, the unmodulated
optical carrier is much stronger than the sidebands that carry the RF signal.
After passing through the EDFA, the strong optical carrier compresses the
EDFA and steals the majority of the optical gain, keeping the modulation
depth low. On the other hand, when the MZM is operated at the low bias
point, the optical carrier is suppressed to be about the same power as the
sidebands. The EDFA is no longer in compression, and the available optical

Figure 3.8 Optical carrier and sidebands before and after EDFA at quadrature and low
bias.
 Advancements in Analog Delay Line Performance 71

gain is higher. After passing through the EDFA, the carrier (along with the
sidebands) experiences an increase in the modulation depth. This in turn
increases the RF gain of the link. The gain improvement is referred to as
all-photonic gain, as it does not depend on any electronic amplification to
achieve higher RF power.
Multiple advantages come with operating the MZM at the low bias
point. Previous work has shown that both the RF gain and the RF NF can be
improved when compared to a quadrature bias MZM at the same photocur-
rent [45]. The improvement in the RF gain is shown in Figure 3.9. As the bias
voltage goes below quadrature (Vbias = 0.5*Vπ), the modulation index, along
with the RF gain, increases. It peaks and then quickly drops as the carrier be-
comes completely suppressed. The increase in RF gain is 10 dB as compared
to quadrature bias. Note that since the optical carrier is suppressed, the DC
photocurrent is much lower than at the quadrature bias point. While the
photocurrent is lower, the higher modulation index may still lead to compres-
sion of the photodetector. Additionally, by suppressing the optical carrier, the
delay line can operate below the SBS threshold of the optical fiber while still
providing the necessary RF performance.

3.5.2 Low Biasing of the Optical Modulator with Dual Wavelengths

As mentioned previously, operating a photodiode in saturation is beneficial
for certain analog delay lines. Compressing the photodiode clamps the in-
tensity-to-phase response and reduces the added phase noise. Unfortunately,
running the photodiode in compression limits the RF power available at the

Figure 3.9 RF gain as a function of DC bias voltage relative to Vp . (After: [14].)

72 Applications of Modern RF Photonics

output of the analog delay line. To overcome this limitation, an RF amplifier

can be used to reach the desired power. Unfortunately, the flicker noise of the
RF amplifier will place a lower bound on the phase noise of the signal [46].
By using a low-biased MZM followed by and EDFA, the all-photonic gain
can provide the necessary RF output power, even with the photodetector in
saturation. Combining the all-photonic gain with a saturated photodetector
provides an analog delay line with low phase noise and higher RF output
power.
Using a low-biased MZM followed by an EDFA does come at a cost.
Since the MZM is no longer operating at the quadrature bias point, the even-
order distortion is no longer canceled at the photodiode. The resulting even-
order distortion can actually be quite large and limits these types of delay
lines to single-octave operation. Another disadvantage comes from the fact
that the EDFA is no longer operating in the saturation regime. While this
provides more gain to the sidebands, it also increases the RF noise penalty
associated with the EDFA, as mentioned previously. The added noise penalty
can be detrimental to the overall RF NF of the delay line. The EDFA has to
be carefully selected in order to ensure that the added noise does not over-
come the improvement from low biasing.
In order to overcome some of these limitations, an advanced form of
low biasing was developed [47]. Other demonstrations have been made with
multiple wavelengths in order to improve the RF metrics of photonic links
[48]. These other demonstrations used only a single-output MZM. An ad-
vanced system is shown in Figure 3.10. The technique uses a dual-output
MZM with two wavelength inputs, one operating at low bias and the other
operating at high bias. Since the MZM has a wavelength-dependent electro-
optic coefficient, the transfer curve of the modulator as a function of DC
bias voltage is different for various wavelength inputs. Thus, for a given DC
bias voltage applied to the MZM, different wavelength inputs will operate at
different DC phase shifts. The result is illustrated in Figure 3.11. By using a
dual-output MZM, each wavelength can then be low-biased at their respec-
tive output. The first transfer curve is for one wavelength while the other

Figure 3.10 RF subsystem using two low-biased wavelengths to remove even order dis-
tortion. (After: [14]).
 Advancements in Analog Delay Line Performance 73

Figure 3.11 Operating two wavelengths each at low bias. (After: [14]).

transfer curve is for a second independent wavelength. By wavelength multi-

plexing the low-biased first wavelength from one arm and the complementary
low-biased second wavelength from the other arm, one can transmit two low-
biased wavelengths to the optical amplifier. The amplifier output is then sent
to a wavelength demultiplexer and the separated wavelengths are each sent
to a photodiode input. The outputs of each photodiode are then subtracted
in order to recover the RF input. Combining the low-bias outputs through
an optical amplifier can simultaneously improve the various RF metrics as
compared to a quadrature-biased MZM through the same optical amplifier.

( ) ( )
I output 1 jdc,λ1 ,t ∝ 1 − cos jdc,λ1
(3.6a)
(
+ sin jdc,λ1 ) sin (Ω t ) − cos ( j ) cos (2Ω t )
rf dc,λ1 rf

( ) ( )
I output 2 jdc,λ2 ,t ∝ 1 + cos jdc,λ2
(3.6b)
(
− sin jdc,λ2 ) sin (Ω t ) + cos ( j ) cos (2Ω t )
rf dc,λ2 rf

jdc , λ2 = − π − jdc , λ1 (3.6c)

2q
RIN sh = (3.6d)
I dc

Equations (3.6a) and (3.6b) show the photocurrent for each wave-
length at the output of the demultiplexer, with jdc as the DC bias point for
each wavelength. By setting the correct phase bias condition between the
two wavelengths (3.6c), the even order distortion will be cancelled while the
74 Applications of Modern RF Photonics

improved fundamental is preserved (3.6d). The advantage of the dual output

MZM method is it requires a smaller wavelength separation for phase bias
jdc greater than 0.25π output MZM method. For a specific instantiation of
this technique, the wavelength separation was only 5 nm as compared to 10
nm in the single output case. With two wavelengths, twice the optical power
is input to the EDFA. This pushes the EDFA further into compression, re-
ducing the noise penalty. The dual wavelength, low biased setup improves
the RF gain, noise figure and spur free dynamic range when compared to a
quadrature biased link operating at the same DC photocurrent. In addition,
since the optical carriers are suppressed, the link length can be increased by a
factor of five before the SBS threshold is reached. The overall improvements
in the performance of the RF metrics are shown in Table 3.1 as compared to
an MZM link biased at quadrature and operating at the same photocurrent.
Thus the use of multiple wavelengths can improve all the RF metrics.
This technique does have some drawbacks. It requires two lasers, which
adds to the complexity of the system. It also requires path matching of the
two outputs from the MZM before they are combined at the EDFA. Finally,�
while the system uses balanced detection, the two lasers are not common so
the noise will not be cancelled. Also, even though the EDFA is common, the
filtered noise is not from the same part of the spectrum and thus is also not
common. Nevertheless, this technique points to the use of multiple wave-
length sources as a means to overcome some of the penalties of using non-
quadrature biased modulation.

3.5.3 Cancelation of Dispersion Induced Second Harmonics by Using Dual

Wavelengths
The dual laser system with a single output has also been shown to compensate
for the second harmonic distortion that is introduced by dispersion in optical
fibers. Chromatic dispersion has been shown to limit the frequency band-
width of fiber links, with a typical 50 km link having a 3-dB bandwidth of ~7
GHz [49]. While this limit is significant, the dispersion-induced second har-
monic is even more restricting. The dispersion-induced second harmonic can
limit the analog performance bandwidth as low as 3 GHz for a 50-km link
[50]. Clearly, this limitation is significant for long-haul microwave frequency
analog links. While the alternating fiber pair configuration can be used to
remove the dispersion penalty, the availability of the right fiber types may not
be possible. This technique can be used for these situations.
The technique of using a dual-input MZM with two separate wave-
length inputs, which cancels the chromatic dispersion-induced second
 Advancements in Analog Delay Line Performance 75

Table 3.1
RF Metric Improvements
from the Two Laser Low Bias
System versus Quadrature
Bias at the Same Photocurrent
Improvments in
Link Metrics
RF Gain 9 dB
RF NF 5.4 dB
RF SFDR 3.6 dB
Length 5x increase

harmonic, appears in Figure 3.12. Note that the architecture is almost the
same as the two laser system in Table 3.1, with two exceptions. First, there is
no EDFA, and second, the lasers are not set for low bias but for quadrature
bias. In this application, the unamplified quadrature-biased performance is
sufficient. Again using the MZM’s wavelength-dependent electro-optic coef-
ficient, two wavelengths are chosen to operate at the same quadrature point
of their respective transfer curves. The output of the MZM is then passed
through a length of fiber with high chromatic dispersion. After the fiber, a
WDM separates the two wavelengths, and each is sent to a photodiode input.
The outputs of each photodiode are then subtracted in order to recover the
RF input. The results show a cancelation of the second harmonic while also
increasing the RF power of the fundamental signal by 6 dB as compared to a
single wavelength.
Equations (3.7a) and (3.7b) show the proportional received photocur-
rent due to an optical carrier modulated by an RF signal via a MZM biased
at quadrature and after passing through a fiber with nonzero chromatic dis-
persion. The equations have been truncated to the DC (first term), funda-
mental (second term), and the second harmonic (third term) terms for an
optical carrier that we have designated as λ­1 and λ2, respectively. Using the

Figure 3.12 System for canceling the second-order distortion due to chromatic dispersion.
(After: [14].)
76 Applications of Modern RF Photonics

small-signal approximation of OIP2req(dBm) = 47 + 70/3 * log10(Idc(mA)),

with jrf as the phase shift in the MZM arms due to the RF signal yields

1
I λ1 (Ω,t ) = I DC + I DC jrf cos( θ 2 )sin (Ωt − θ1 )
2
2 (3.7a)
I DC jrf
+ cos (2 θ 2 ) cos (2Ωt − 2 θ1 )
4

1
I λ2 (Ω,t ) = I DC − I DC jrf cos( θ 2 + j∆λ )sin ( Ωt − θ1 + j∆λ )
2
2 (3.7b)
I DC jrf
+ cos ( 2 θ 2 + 2 j∆λ ) cos ( 2Ωt − 2 θ1 + 2 j∆λ )
4

where θ1 = β1ΩL, θ2 = β2Ω2L, β1 is the phase velocity of each optical carrier,

β2 is the group velocity dispersion of each optical carrier, Ω is the angular RF
frequency, L is the length of the fiber between the output of the MZM and
the photodiode, jrf is the phase shift due to the RF drive voltage relative to
the RF voltage required for p is the phase shift, and j∆λ is an additional phase
term to account for the wavelength dependence of the dispersion. Note the
necessary condition that the loss in the MZM and the fiber, the photodiode
responsivity, and the input power all be the same for both wavelengths. In
order for the cancelation to work, the proper delay must also be set at the end
of the fiber before subtracting the two photocurrents such that j∆λ can be set
to zero. With this condition met, the subtraction of the photocurrents yields
the following:

1
I λ1 − I λ2 = 2I DC jrf sin (Ωt − θ1 ) cos( θ 2 ) (3.7c)
2

Looking at (3.7c), we can see two results: First, the second harmonic
term has been canceled, and second, the fundamental term is doubled.

3.5.4 Single-Sideband Modulation

While the MZM can be used to cancel the second harmonic distortion in-
troduced by dispersion, other modulation formats can be used to overcome
dispersion in optical fibers. Single-sideband (SSB) modulation has been iden-
 Advancements in Analog Delay Line Performance 77

tified as a means to overcome the RF fading penalties associated with disper-

sion [51]. SSB modulation places the RF sideband on only one side of the
carrier, rather than the more common dual-sideband modulation. Multiple
methods have been demonstrated to generate SSB optical signals, includ-
ing optical filtering of dual-sideband modulation [52], using a 90° RF hy-
brid with a single push-pull MZM [53] and using a single electrode MZM
with bidirectional optical inputs [54]. Despite the advantages of using SSB
modulation, the use of SSB is limited in RF photonic applications. One of
the common misconceptions is that the RF output power of SSB modula-
tion will be less than DSB modulation, as there is only one less sideband to
mix with the optical carrier. However, work has been done to show that the
SSB does not degrade the RF performance when compared to other types of
DSB modulation from an MZM [55]. Specifically a dual-drive MZM us-
ing a 90° hybrid or a 180° hybrid or with the RF signal connected to just
one arm of the MZM are compared. The various RF gains for each type of
configuration are given in (3.8a)–(3.8d) using the small signal approxima-
( )
tion of J n ( jrf ) ≈ jnrf 2n n ! , with jrf as the phase shift in the MZM arms
due to the RF signal. In (3.8), the R is the responsivity of the photodetector,
αMZM is the insertion loss of the MZM, Plaser is the power of the laser, and
∆jdc is the difference between the DC bias points for the two electrodes in
the MZM. From this analysis, the SSB modulation from a 90° hybrid MZM
has the same OIP3 performance as the three other cases. The SSB link does
provide 3 dB less RF output power when compared to a DSB link using a
180° hybrid, but the SSB link has the same RF output power as the push-
pull MZM and single-arm MZM. The SSB modulation increases the optical
power in the sideband that is not nulled, making up for the loss of the other
optical sideband.

2 2
jrf2
PRF ,Fund ,ss ≈ (R αMZM Plaser ) sin ( ∆jdc ) Z out , (3.8a)
8
jrf2
PRF ,Fund ,ss ≈ (R αMZM Plaser )2 sin 2 ( ∆jdc ) Z out (3.8b)
4

2 2
jrf2
PRF ,Fund ,ss ≈ (R αMZM Plaser ) sin ( jdc 1 ) Z out . (3.8c)
8

2 2
jrf2
PRF ,Fund ,ss ≈ (R αMZM Plaser ) sin ( jdc 1 ) Z out (3.8d)
8
78 Applications of Modern RF Photonics

3.5.5 Single-Sideband Modulation to Cancel Photodetector Nonlinearities

As mentioned previously, the second-order nonlinearity from the photodetec-
tor can become a limiting factor. In-depth studies have shown that the non-
linear photodetector response will be limited by the second-order distortion
[17]. While specialty photodetectors have been demonstrated with improved
nonlinear performance, it is desirable to improve the performance of com-
mercially available photodiodes. One method uses a dual-output MZM at
quadrature bias with balanced detection to cancel the photodetector-induced
second-order nonlinearity in the HF band [25]. Phase-matching two paral-
lel links operating in the multigigahertz range and over long fiber lengths
becomes a significant challenge. Another method incorporates two MZMs
inside a third overall MZM structure known as the dual-parallel MZM (DP-
MZM). This type of modulator allows for the generation of SSB-modulated
signals. Using two discrete MZMs has been investigated to linearize unwant-
ed nonlinear distortions due to the individual MZMs [56]. Recent work has
also demonstrated high-performance links using an integrated DP-MZM
[57–59]. A recent demonstration uses the SSB modulation from a DP-MZM
to cancel the unwanted second-harmonic distortion from the photodetec-
(
tor. Using the small-signal approximation of J n ( jrf ) ≈ jnrf 2n n ! , with jrf )
as the phase shift in the MZM arms due to the RF signal, (3.9) shows that
the generated photocurrent from the DP-MZM at the photodetector can be
written as

2 + cos( jdc 3 ) − cos( jdc 1 ) − cos( jdc 2 ) − cos( jdc 1 − jdc 3 )

I DC ≈ 2 z  ,
 − cos( jdc 2 + jdc 3 ) + cos( jdc 1 − jdc 2 − jdc 3 ) 
(sin( j ) + sin( j − j ))sin(Ω t ) 
 dc 1 dc 1 dc 3 rf

I Fund (t ) ≈ 2 jrf z  +(sin( jdc 2 ) + sin( jdc 2 + jdc 3 ))cos(Ωrf t )  ,
  (3.9)
 + 2 sin( jdc 1 − jdc 2 − jdc 3 ) sin(Ωrf t + φrf )
 
 −(cos( jdc 1 ) + cos( jdc 1 − jdc 3 ))cos(2Ωrf t ) 
1  
I Second (t ) ≈ jrf 2 z  +(cos( jdc 2 ) + cos( jdc 2 + jdc 3 ))cos(2Ωrf t ) ,
2  
 +2 cos( jdc 1 − jdc 2 − jdc 3 )cos(2Ωrf t + 2 φrf )

where z = RaMZMPlaserG0/16,Go is the optical gain from the EDFA after the
DP-MZM output to the photodetector, is the photodetector responsivity,
and jrf = 3π/4. From (3.9), the phase between the fundamental and the
 Advancements in Analog Delay Line Performance 79

second harmonic can be adjusted by simply changing jdc1,2,3. In order to

cancel the photodetector-induced second harmonic, the phase of the second
harmonic photocurrent from the DP-MZM is set to be 180° with relation to
the second harmonic photocurrent generated from the photodetector by the
fundamental optical power, which is shown in (3.10).
2
 jrf z 
I PD (Popt , in ) = −a 2  σ comb  2 cos(2Ωrf t + 2 φcomb + 2 φdc ) (3.10)
 ℜ 

The system uses the various bias control points of the DP-MZM to
find the correct second harmonic. An experimental demonstration showed
an improvement in the OIP22H of 55.3 dBm. In addition, the third-order
nonlinear distortion is shown to remain the same. However, the cancelation
of the second harmonic comes at the cost of a reduction in the fundamental
power of 6.3 dB.

3.6 Conclusions
Various advanced techniques have been demonstrated to improve the perfor-
mance of the analog delay line. Various photonic components can be used to
improve the overall RF metrics of the delay line. The photodiode nonlinearity
can be improved with better component design. However, the nonlinearities
also can be overcome by using different modulation formats. In addition, the
limitations of the optical fibers can be overcome with the use of different fiber
types and isolators. The noise of the EDFA can be characterized and properly
controlled by properly designing the analog delay line to minimize the added
noise from the noise penalty of each EDFA. Finally, the MZM can be used at
different bias points to improve the RF performance. Chapter 4 develops the
use of RF photonics for oscillators in the demonstration architecture.

References
[1] Funk, E. E., V. J. Urick, and F. Bucholtz, “High Dynamic Range, Long Haul (>100
km) Radio over Fiber,” in Microwave Photonics (C. H. Lee, (ed.), Boca Raton, FL: CRC
Press, 2007, pp. 185–212.
[2] Ackerman, E. I., et al., “RF-over-Fiber Links with Very Low Noise Figure,” Journal of
Lightwave Technology, Vol. 26, No. 15, 2008, pp.2441–2448.
[3] Karim, A., and J. Devenport, “Noise Figure Reduction in Externally Modulated Analog
Fiber-Optic Links,” Photonics Technology Letters, Vol. 19, No. 5, 2007, pp. 312–314.
80 Applications of Modern RF Photonics

[4] Joshi, A., S. Datta, and D. Becker, “GRIN Lens Coupled Top-Illuminated Highly
Linear InGaAs Photodiodes,” Photonics Technology Letters, Vol. 20, No. 17, 2008, pp.
1500–1502.
[5] Cox, C. H., et al. “Limits on the Performance of RF-over-Fiber Links and Their Impact
on Device Design,” Transactions on Microwave Theory and Techniques, Vol. 54, No. 2,
2006, pp. 906–920.
[6] Williams, K. J., and R. D. Esman, “Design Considerations for High-Current Photode-
tectors,” Journal of Lightwave Technology, Vol. 17, No. 8, 1999, pp. 1443–1454.
[7] Li, Z., et al., “High-Saturation-Current Modified Uni-Traveling-Carrier Photodiode
with Cliff Layer,” Journal of Quantum Electronics, Vol. 46, No. 5, 2010, pp. 626–632.
[8] Lin, L. Y., et al., “High-Power High-Speed Photodetectors-Design, Analysis, and Ex-
perimental Demonstration,” Transactions on Microwave Theory and Techniques, Vol. 45,
No. 8, 1997, pp. 1320–1331.
[9] Urick, V. J., et al., “Fiber-Optic Links with All-Photonic RF Gain and Low RF Noise
Figure,” International Microwave Symposium Digest, Baltimore, MD, 2011.
[10] Williams, K. J., R. D. Esman, and M. Dagenais, “Nonlinearities in p-i-n Microwave
Photodetectors,” Journal of Lightwave Technology, Vol. 14, No. 1, 1996, pp. 84–96.
[11] Taylor, J., et al., “Characterization of Power-to-Phase Conversion in High-Speed P-I-N
Photodiodes,” Photonics Journal, Vol. 3, No. 1, 2011, pp. 140–151.
[12] Tulchinsky, D. A., and K. J. Williams, “Excess Amplitude and Excess Phase Noise of
RF Photodiodes Operated in Compression,” Photonics Technology Letters, Vol. 17, No.
3, 2005, pp. 654–656.
[13] Eliyahu, D., D. Seidel, and L. Maleki, “RF Amplitude and Phase-Noise Reduction of
an Optical Link and an Opto-Electronic Oscillator,” Transactions on Microwave Theory
and Techniques, Vol. 56, No. 2, 2008, pp. 449–456.
[14] Devgan, P., “Review of Improvements in Radio Frequency Photonics,” AFRL Report,
AFRL-RY-WP-TR-2017-0156, 2017.
[15] Hastings, A. S., D. A. Tulchinsky, and K. J. Williams, “Photodetector Nonlinearities
Due to Voltage-Dependent Responsivity,” Photonics Technology Letters, Vol. 21, No. 21,
2009, pp. 1642–1644.
[16] McKinney, J. D., et al., “Measurement of Photodiode Harmonic Distortion Using
Optical Comb Sources And High-Resolution Optical Filtering,” Conference on Lasers
and Electro-Optics, 2009.
[17] Urick, V. J., et al., “Wideband Analog Photonic Links: Some Performance Limits and
Considerations for Multioctave Implementations,” International Society for Optics and
Photonics OPTO, 2012, pp. 825904–825904.
[18] Ackerman, E. I., “Broadband Linearization of a Mach-Zehnder Electrooptic
Modulator,” Transactions on Microwave Theory and Techniques, Vol. 47, No. 12, 1999,
pp. 2271–2279.
 Advancements in Analog Delay Line Performance 81

[19] Ismail, T., et al., “High-Dynamic-Range Wireless-over-Fiber Link Using Feedforward

Linearization,” Journal of Lightwave Technology, Vol. 25, No. 11, 2007, pp. 3274–3282.
[20] O’Connor, S. R., T. R. Clark, Jr., and D. Novak, “Wideband Adaptive Feedforward
Photonic Link,” Journal of Lightwave Technology, Vol. 26, No. 15, 2008, pp.2810–2816.
[21] Urick, V. J., et al., “Phase Modulation with Interferometric Detection as an Alternative
to Intensity Modulation with Direct Detection for Analog-Photonic Links,” Transactions
on Microwave Theory and Techniques, Vol. 55, No. 9, 2007, pp.1978–1985.
[22] Hastings, A., et al., “Suppression of Even-Order Photodiode Nonlinearities in
Multioctave Photonic Links,” Journal of Lightwave Technology, Vol. 26, No. 15, 2008,
pp. 2557–2562.
[23] Urick, V. J., et al., “Photodiode Linearity Requirements for Radio-Frequency Photonics
and Demonstration of Increased Performance Using Photodiode Arrays,” Proc. Topical
Meeting Microwave Photonics, Gold Coast, Australia, 2008, pp.86–89.
[24] Roman, J. E., et al., “Optical Fiber Cables for Synchronous Remoting of Numerous
Transmitters/Receivers,” Photonics Technology Letters, Vol. 10, No. 4, 1998, pp. 591–
593.
[25] Urick, V. J., et al., “Field Test on the Feasibility of Remoting HF Antennas with Fiber
Optics,” NRL Memorandum Report, NRL/MR/5652-08-9137, 2008.
[26] Klamkin, J., et al., “Unitraveling-Carrier Balanced Photodiode with Tunable MMI
Coupler for Optimization of Source Laser RIN Suppression,” Proc. Topical Meeting
Microwave Photonics, Valencia, Spain, 2009.
[27] Theys, M. D., et al., “Using a Microcontroller to Optimize the Bias Voltage of Balanced
Photodiodes to Minimize Even-Order Distortion in Analog Fiber-Optic Links,” Proc.
Optical Fiber Communication Conference, 2010.
[28] Grüner-Nielsen, L., et al., “Dispersion Compensating Fibers,” Optical Fiber Technology,
Vol. 6, No. 2, 2000, pp. 164–180.
[29] Akasaka, Y., et al., “Dispersion Compensating Optical Fiber for Wavelength Multiplex
Transmission and Method Using Same,” United States Patent No. 5,568,583, 1996.
[30] Vengsarkar, A. and J. Wagener, “Dispersion Compensating Optical Fiber and
Communication System Comprising Same,” United States Patent No. 5,802,234,
1998.
[31] Onishi, M., and H. Kanamori, “Optical Transmission System with Dispersion
Compensating Optical Fiber,” United States Patent No. 5,555,340, 1996.
[32] Liu, A., “Suppressing Stimulated Brillouin Scattering in Fiber Amplifiers Using
Nonuniform Fiber and Temperature Gradient,” Optics Express, Vol 15, No. 3, 2007,
pp. 977–984.
[33] Parker, T. R., et al., “Temperature and Strain Dependence of the Power Level and
Frequency of Spontaneous Brillouin Scattering in Optical Fibers,” Optics Letters,
Vol. 22, No. 11, 1997, pp. 787–789.
82 Applications of Modern RF Photonics

[34] Dragic, P. D., et al., “Optical Fiber with an Acoustic Guiding Layer for Stimulated
Brillouin Scattering Suppression,” Proc. Conference on Lasers and Electro-Optics, 2005.
[35] Mao, X. P., et al., “Stimulated Brillouin Threshold Dependence on Fiber Type and
Uniformity,” Photonics Technology Letters, Vol. 4, No. 1, 1992, pp. 66–69.
[36] De Oliveira, C. A. S., et al. “Stimulated Brillouin Scattering in Cascaded Fibers of
Different Brillouin Frequency Shifts,” Journal of Optical Society America B, Vol. 10, No.
6, 1993, pp. 969–972.
[37] Devgan, P. S., et al., “SOA-Based Regenerative Amplification of Phase Noise Degraded
DPSK Signals,” Proc. Optical Fiber Communication Conference, 2005.
[38] Devgan, P. S., et al., “Highly Efficient Multichannel Wavelength Conversion of DPSK
Signals,” Journal of Lightwave Technology, Vol. 24, No. 10, 2006, pp.3677–3682.
[39] Becker, P. C., N. A. Olsson, and J. R. Simpson, Erbium-Doped Fiber Amplifiers
Fundamentals and Technology, San Diego, CA: Academic Press, 1999.
[40] Haus, H. A., “The Noise Figure of Optical Amplifiers,” Photonics Technology Letters,
Vol. 10, No. 11, 1998, pp. 1602–1604.
[41] Urick, V. J., F. Bucholtz, and K. J. Williams, “Noise Penalty of Highly Saturated
Erbium-Doped Fiber Amplifiers in Analog Links,” Photonics Technology Letters, Vol. 18,
No. 6, 2006, pp. 749–751.
[42] Devgan, P. S., et al., “Cascaded Noise Penalty For Amplified Long-Haul Analog Fiber-
Optic Links,” Transactions on Microwave Theory and Techniques, Vol. 55, No. 9, 2007,
pp. 1973–1977.
[43] Urick, V. J., et al., “The Performance of Analog Photonic Links Employing Highly
Compressed Erbium-Doped Fiber Amplifiers,” Transactions on Microwave Theory and
Techniques, Vol. 54, No. 7, 2006, pp. 3141–3145.
[44] Wooten, E. L., et al. “A Review of Lithium Niobate Modulators for Fiber-Optic
Communications Systems,” Journal of Selected Topics in Quantum Electronics, Vol. 6,
No. 1, 2000, pp. 69–82.
[45] Urick, V. J., et al.,“ Analysis of an Analog Fiber-Optic Link Employing a Low-Biased
Mach–Zehnder Modulator Followed by an Erbium-Doped Fiber Amplifier,” Journal of
Lightwave Technology, Vol. 27, No. 12, 2009, pp. 2013–2019.
[46] Foreman, S. M., et al. “Remote Transfer of Ultrastable Frequency References Via Fiber
Networks,” Review of Scientific Instruments, Vol. 78, No. 2, 2007, 021101.
[47] Devgan, P. S., et al. “Even-Order Harmonic Cancellation for Off-Quadrature
Biased MZM with Improved RF Metrics Using Dual Wavelength Inputs and Dual
Outputs,” Optics Express, Vol. 17, No. 11, 2009, pp. 9028–9039.
[48] Rollins, D., “Linearized Optical Link Using a Single MZM and Two Optical Carriers,”
United States Patent No. 7,079,780, 2006.
[49] Schmuck, H., “Comparison of Optical Millimetre-Wave System Concepts with Regard
to Chromatic Dispersion,” Electronics Letters, Vol. 31, No. 21, 1995, pp. 1848–1848.
 Advancements in Analog Delay Line Performance 83

[50] Corral, J. L., J. Marti, and J. M. Fuster, “General Expressions for IM/DD Dispersive
Analog Optical Links with External Modulation or Optical Up-Conversion in a
Mach-Zehnder Electrooptical Modulator,” Transactions on Microwave Theory and
Techniques, Vol. 49, No.10, 2001, pp. 1968–1976.
[51] Hraimel, B., et al., “Optical Single-Sideband Modulation with Tunable Optical Carrier
to Sideband Ratio in Radio Over Fiber Systems,” Journal of Lightwave Technology, Vol.
29, No. 5, 2011, pp. 775–781.
[52] Park, J., W. V. Sorin, and K. Y. Lau, “Elimination of the Fibre Chromatic Dispersion
Penalty on 1550 nm millimetre-Wave Optical Transmission,” Electronics Letters, Vol.
33, No. 6, 1997, pp. 512–513.
[53] Smith, G. H., D. Novak, and Z. Ahmed, “Technique for Optical SSB Generation to
Overcome Dispersion Penalties in Fibre-Radio Systems,” Electronics Letters, Vol. 33,
No. 1, 1997, pp. 74–75.
[54] Loayssa, A., D. Benito, and M. J. Garde, “Single-Sideband Suppressed-Carrier
Modulation Using a Single-Electrode Electrooptic Modulator,” Photonics Technology
Letters, Vol. 13, No. 8, 2001, pp. 869–871.
[55] Devgan, P. S., et al., “Cancellation of Photodiode-Induced Second Harmonic Distortion
Using Single Side Band Modulation from a Dual Parallel Mach-Zehnder,” Optics
Express, Vol. 20, No. 24, 2012, pp. 27163–27173.
[56] Korotky, S. K., and R. M. de Ridder, “Dual Parallel Modulation Schemes for
Low-Distortion Analog Optical Transmission,” Journal on Selected Areas in
Communications, Vol. 8, No. 7, 1990, pp. 1377–1381.
[57] Zhu, G., W. Liu, and H. R. Fetterman, “A Broadband Linearized Coherent Analog
Fiber-Optic Link Employing Dual Parallel Mach–Zehnder Modulators,” Photonics
Technology Letters, Vol. 21, No. 21, 2009, pp. 1627–1629.
[58] Li, S., et al., “Highly Linear Radio-Over-Fiber System Incorporating a Single-Drive
Dual-Parallel Mach–Zehnder Modulator,” Photonics Technology Letters, Vol. 22, No. 24,
2010, pp. 1775–1777.
[59] Kawanishi, T., and M. Izutsu, “Linear Single-Sideband Modulation for High-SNR
Wavelength Conversion,” Photonics Technology Letters, Vol. 16, No. 6, 2004, pp. 1534–
1536.
4
Oscillators Utilizing RF Photonics
Returning to the demonstration architecture for using RF photonics, the em-
phasis of this chapter is on oscillators. The oscillator is used for many differ-
ent applications. The oscillator offers a common clock signal to all the other
subsystems in the architecture. It can be used for downconverting the RF
signal or for triggering a sampling function, such as an ADC. The oscillator
plays a significant role in the RF architecture and can be implemented using
RF photonic techniques. This chapter discusses the need for high-quality and
fast-tuning oscillators. In addition, the chapter reviews various techniques for
making oscillators based on RF photonic technologies, with an emphasis on
the optoelectronic oscillator and the beating of two laser sources to generate
an RF tone.

4.1 Need for Oscillators

One of the common signal-processing requirements many varied applications
share is the need for very precise timing. Doppler radars require very low
phase noise clock sources in order to minimize the uncertainty of objects that
are being tracked [1]. As optical digital networks have been shown to trans-
mit an aggregate of 10 Tbps of data [2], they require low phase noise clocks
for multiplexing the data streams on the entire network. Other applications
requiring low phase noise clocks include sampling for ADCs [3, 4], clock re-
covery [5], and pulse sources [6]. In computers, on-chip optical interconnects
are being investigated to overcome the clock skew and speed limitations that
copper interconnects currently exhibit [7, 8]. In addition, as the processing

85
86 Applications of Modern RF Photonics

speeds of the chips increase beyond 1 GHz, the requirement on the phase
noise of the data clock becomes more stringent. Thus, a photonic low phase
noise clock that synchronizes the data bus on the chip becomes very impor-
tant. On the opposite end, optical networks are used to send voice, video,
and internet data of 10 Gbps or higher over thousands of kilometers [9]. The
ability to synchronize these signals for proper routing and processing is essen-
tial to meet the ever-increasing demand for bandwidth. Again the low phase
noise photonic clock becomes a key part of the network. A final example of
the importance of low phase noise photonic clocks is in their use in ADCs.
The sample and hold circuit of the ADC depends on a very low jitter clock
to operate correctly. Any error in the timing of the clock can lead to degraded
SNR for the ADC [10], where ENOB is the effective number of bits.

 1 
SNR (dB ) = 20 log10   (4.1a)
 2 πf rf t jitter 

SNR (dB ) = 6.02 ∗ ENOB + 1.78 (4.1b)

Equation (4.1a) defines the dependence of the SNR of the ADC as a

function of both the timing jitter of the sampling clock (tjitter) and the fre-
quency of the input sinusoidal signal to the ADC (frf ). Equation (4.1b) then
shows the relation of SNR to the number of bits in the ADC [11]. Then the
( )
following relationship can be derived t jitter ≈ 1 f samp 2N bits + 2 , where fsamp =
2*frf is the Nyquist sampling frequency of the ADC. Plotting this relation-
ship in Figure. 4.1 shows the required timing jitter for a given ENOB and
sampling rate of the ADC. As the ADCs have increased to sampling rates of 1
GHz and higher, the requirement to get even 10 ENOB forces the timing jit-
ter to be on the order of tens of femtoseconds. ADCs place the most stringent
requirements on the timing jitter of the clock and are one of the strongest
drivers for low phase noise sources.

4.2 Phase Noise and Timing Jitter

Before discussing examples of photonic-based oscillators, a brief discussion
of phase noise and its relationship to timing jitter is required. In character-
izing oscillators of any type, the common measurement that is made of the
frequency fluctuations is called the phase noise measurement. At least two
formal definitions of phase noise have been presented in the literature, which
 Oscillators Utilizing RF Photonics 87

Figure 4.1 Required timing jitter for an ADC at a sampling rate between 1 and 40 Gbps for
8, 10, 12, 13, and 14 ENOB resolution. (After: [12].)

can lead to some confusion [13]. In this discussion, the accepted definition,
denoted as L(f ), is the single-sideband power spectral density (PSD). The
PSD is normalized to the power of the RF signal and has units of decibels
relative to carrier power per one hertz bandwidth.
In general, the PSD will have contributions from both amplitude and
phase noise. However, the measurement system can be configured to separate
the amplitude noise and the phase noise contributions to the PSD. One ex-
ample uses a phase detector, which converts any phase fluctuations between
two input signals into a voltage that can be measured [14]. Other methods
also exist to measure only the phase noise contribution to the PSD. In all of
these cases, the amplitude noise does not contribute to the measured PSD. In
this case, the timing jitter can be extracted directly from the measured PSD
using the following relation

f max
σn = 2∫ ( f ) df (4.2)
f min

where σn is the root-mean-squared (rms) total noise, and fmax­ and fmin are the
maximum and minimum frequency offsets that the phase noise is integrated
over. Since the rms amplitude noise (σa) contribution is negligible, the rms
timing jitter (σj) can be obtained by the following relationship sj = σn/(2πfrf)
where frf is the RF frequency of the clock. However, if the measurement sys-
tem is not set to ignore the amplitude contribution, further work is needed
to extract the timing jitter contribution from the rms total noise. When the
higher harmonics of the RF signal can be measured, it has been shown previ-
88 Applications of Modern RF Photonics

ously [15] that the rms amplitude noise and rms timing jitter components
can be separated from the overall rms noise by the following relationship
( )
2
σ n2 = n 2 2 πf rf σ 2j + σ a2 where n is the n-th harmonic of the RF signal being
measured. The noise PSD of both the timing jitter and the amplitude noise
is plotted in Figure 4.2. Note that the amplitude noise spectra is arbitrarily
plotted broader than the timing jitter noise in order to differentiate the two
sources of noise and is not necessarily representative of a real system. As the
noise of each harmonic is measured, it will increase quadratically with the
timing jitter noise while the amplitude noise contribution is constant. Thus,
if one can measure the noise of a sufficient number of harmonics and then
fit the results to a curve, the timing jitter and amplitude noise contributions
can be determined.

4.3 Optoelectronic Oscillator

One of the most used examples of RF photonics for oscillators is the opto-
elect (OEO). The OEO was first proposed by Yao and Maleki [16–18] in
1996. The OEO is similar to optoelectronic feedback circuits demonstrated
by Neyer and Voges [19] in 1982, by Nakazawa, et al. [20] in 1984, and by
Lewis [21] in 1992.

Figure 4.2 Theoretical noise PSD for multiple harmonics of the RF frequency showing the
timing jitter and amplitude noise contributions. (After: [12].)
 Oscillators Utilizing RF Photonics 89

Figure 4.3 Diagram of a basic OEO (where PD stands for photodiode and opt. mod. stands
for optical modulator). (After: [12].)

As seen in Figure 4.3, the basic OEO starts with a continuous wave
(CW) laser feeding an optical modulator. The optical output of the modula-
tor is passed through a length of optical fiber into a photodiode. The recov-
ered electrical signal is amplified and sent through an electronic bandpass
filter. The output of the filter is connected to the RF input of the modulator
in order to complete the optoelectronic cavity. When the gain within the
cavity is greater than the loss, the OEO will begin to oscillate. The electronic
bandpass filter selects the frequency of oscillation by attenuating the other
free-running modes of the cavity.
The OEO differs from the previous optoelectronic circuits by including
a very low loss fiber-optical delay, which is essential to creating a cavity with
a very high Q factor. In the case of the OEO, The Q factor is defined as the
ratio of the stored energy in the cavity over the loss of the cavity. With opti-
cal fibers having a loss of 0.2 dB/km, the cavity can store a large amount of
energy with very little loss. Thus, the Q factor of the OEO can easily achieve
the level of 108 or higher. This translates to a 10-GHz clock signal with a
phase noise of –140 dBc/Hz at 10 kHz offset [23]. Unfortunately the phase
noise of the OEO cannot be continuously improved as the length of fiber in-
creases. Lasri et al. [22] demonstrated the improvement in the timing jitter of
an OEO having an inverse square root dependence on the fiber length. The
limitations in that case were due to the random fiber length variations due to
small fluctuations in the environment around the fiber span.
90 Applications of Modern RF Photonics

Figure 4.4 Theoretical phase noise plot showing the different influences of noise on the
PSD at frequency offsets from 1 Hz to 1 MHz from the carrier. (After: [12].)

To show the general shape of the PSD, a typical theoretical phase noise
plot of an OEO is presented in Figure 4.4. The oscillator’s performance is
determined by different sources of noise at different frequency offsets [13]. In
the 1–10-Hz frequency offset range, the noise is dominated by environmen-
tal fluctuations, caused by temperature variations and acoustic vibrations in
the fiber-based cavity, and has a slope of f  –4. In the 10-Hz–1-kHz frequency
offset range, the noise is dominated by flicker phase noise from the RF ampli-
fication stage necessary for most photonics-based oscillators to oscillate and
has a slope of f  –3. The next frequency range of 1 kHz–30 kHz is dominated
by white phase noise and has a slope of f  –2. Finally, at frequency offsets of 30
kHz and higher, the phase noise has a flat noise response that is due to the
adjacent nonoscillating side mode of the OEO.
One of the advantages of the OEO is the high performance achieved
with the use of commercially available parts that are inexpensive. Once the
first OEO was demonstrated, multiple demonstrations were presented to im-
prove on the initial design. The following sections present some examples of
improved OEOs.

4.3.1 Multiloop OEO

One of the disadvantages of the initial OEO design was the limitation of the
electrical bandpass filter to suppress unwanted side modes. The fiber length
must be maximized in order to achieve as high a Q factor as possible. As the
fiber length increases, the spacing between the cavity modes decreases. For
example, a 3-km length of fiber will yield a cavity mode spacing of approxi-
mately 67 kHz [24]. A high-quality electrical bandpass filter at 10 GHz has a
 Oscillators Utilizing RF Photonics 91

3-dB bandwidth of 10 MHz [23]. Thus, many side modes will pass through
the electrical bandpass filter and adversely affect the phase noise. One method
to reduce this problem was to introduce a second fiber length into the OEO
cavity [25]. An example of this type of OEO is shown in Figure 4.5. The sec-
ond loop of the OEO has its own set of cavity modes. As long as the length
of the second loop is not a harmonic multiple of the first loop, the cavity
modes will not overlap with each other, as seen below the OEO in Figure 4.5.
However, the modes from each loop that are closest to each other will lock
and suppress the other cavity modes. This can be seen in Figure 4.6, which
shows the single-loop phase noise spectrum with the side modes along with
the dual-loop spectrum with the side modes suppressed below –110 dBc/Hz.
The trade-off of the system is that the phase noise is now an average of noise
of the two loops individually, not the phase noise of just the longer loop. In
addition, since both loops support the side modes, they are not completely

Figure 4.5 Diagram of a dual-loop OEO along with the cavity modes within the filter pass-
band for the two fiber delays. (PD: photodiode; mod: optical modulator; and EBF: bandpass
filter.) (After: [12].)
92 Applications of Modern RF Photonics

Figure 4.6 Measured phase noise of the single-loop and dual-loop OEO showing the non-
oscillating side mode suppression below -110 dBc/Hz. (After: [12].)

eliminated, but merely suppressed. Nevertheless, many demonstrations have

been made with the multiloop OEO in order to suppress the side modes
[26–28].
In order to address the shortcomings of the multiloop OEO, a dual-
cavity OEO was conceived and demonstrated by Zhou and Blasche [29], in
which a master-slave configuration was used. In this demonstration, a short-
loop OEO, known as the slave, is injection-locked by a longer loop OEO,
known as the master. An example of the master-slave OEO appears in Fig-
ure 4.7. Equivalent systems using electronic oscillators have shown that the
slave oscillator will have the phase noise performance of the master oscillator.
In this case, the overall system has the widely spaced cavity modes of the

Figure 4.7 Diagram of a master-slave OEO. (DFB: distributed feedback laser; PD: photodi-
ode; MZM: Mach Zehnder modulator; and EBF: electrical bandpass filter.)
 Oscillators Utilizing RF Photonics 93

short-loop OEO while having the phase noise performance of the longer loop
OEO. The demonstrated phase noise of the generated 10-GHz RF signal is
–130 dBc/Hz at an offset of 1 kHz [29]. The nonoscillating side modes are
less than –140 dBc/Hz, which is lower than the side modes in the dual-loop
configurations shown previously [25]. This configuration avoids the averag-
ing of the two fiber loop’s phase noise, with the added complexity of having
two entirely separate OEO cavities. Further work has been done to analyze
this system more closely and further demonstrations have been made [30,
31].

4.3.2 OEO with All-Photonic Gain

One of the limiting noise sources for the OEO is the RF amplifier necessary
for the OEO to oscillate. Even though the loss of the optical fiber in the cav-
ity is very low, both the RF-to-optical conversion at the optical modulator
and the optical to RF conversion at the photodetector have loss. Therefore
the electronic RF amplifier is necessary to overcome both of these losses.
However, like any amplifier, it adds additional phase noise to the OEO, as
shown in Figure 4.4. While the amplifier is used in the OEO, it will become
compressed as it provides unity gain for the cavity to meet the oscillation
condition. In compression, amplitude fluctuations will become suppressed,
but the phase fluctuations will increase. This leads to the added phase noise.
While a lot of work has been done to reduce the phase noise introduced by the
electronic RF amplifiers used in OEOs, they are expensive and not wideband.
Therefore, each OEO would need its own special RF amplifier to operate at a
given frequency, limiting the otherwise potentially wideband operation of an
OEO. If the OEO can oscillate utilizing only photonic gain, the wideband
operation can be preserved. One method for providing all-photonic gain is
the low-biased MZM followed by an optical amplifier. As discussed in Chap-
ter 3, the low-biased MZM can provide increased RF gain without the need
of an electronic amplifier. A 10-GHz OEO using all-photonic gain has been
previously demonstrated [32, 33]. In this demonstration, the RF gain has to
match the loss in the OEO cavity, which is 13 dB. Looking at Figure 4.8, this
would require a photocurrent of approximately 60 mA, which is well above
the power-handling capabilities of the photodetector. So a different method
is required to meet the oscillation condition. By using an optical amplifier
after the MZM and lowering the DC bias of the MZM, the RF gain of the
OEO can be increased to a level that matches the loss of the OEO cavity. One
advantage of the all-photonic OEO is that the RF power is increased while
the photocurrent is kept at 6 mA, a level that the photodetector can handle.
Looking at Figure 4.8, the RF gain at 6 mA for the all-photonic gain is 21.5
94 Applications of Modern RF Photonics

Figure 4.8 Theoretical plot of the RF gain for a single pass in the OEO as a function of DC
photocurrent. (After: [12].)

dB higher than in the quadrature bias case. A disadvantage of the all-photonic

gain OEO is that the second harmonic of the RF signal becomes quite large.
However, in the case of the OEO, the electrical bandpass filter will suppress
the second harmonic and make it negligible. The all-photonic gain OEO also
showed an improvement in the phase noise as compared to the same OEO
used with an electrical RF amplifier. The phase noise spectrum shows that the
noise is no longer dominated by the flicker phase noise in the 100 Hz–1 kHz
frequency offset range. In addition, the overall phase noise spectrum of the
all-photonic OEO is lower than in the electronic RF amplifier OEO, in part
due to the lower RF noise figure of the overall system. Thus, the all-photonic
gain OEO has several advantages over electronic RF amplifier OEOs.

4.3.3 Clock Synchronization Using an OEO

The OEO can also be used for clock synchronization. As discussed previously,
the outputs of two oscillators can be synchronized by injection locking. The
theory of injection locking is a well-studied subject—after being first real-
ized by Huygens, who discovered in the late 1800s that the pendulum clocks
on the wall of his bedroom would become synchronized if they were placed
within a certain distance of each other [34]. The locking mechanism occurred
through mechanical vibrations transmitted through the wall. As seen in the
 Oscillators Utilizing RF Photonics 95

movie at the website found on page iv of this book, the two pendulums are
out of sync. As the mechanical vibrations increase, the pendulums become
synchronous. Since then, injection locking has been observed in many differ-
ent systems. In fact, the injection locking process is what allows the multiloop
OEO to combine the RF signal between the different loops to yield a single
low-noise RF signal. In order to use the OEO for clock recovery, the OEO
can be injection-locked by an external signal. The output of the OEO will
then become synchronized with the incoming RF signal. Before presenting
examples of clock synchronization, we will review the basics of the injection
locking process.
The injection locking process for electronic oscillators was first formally
described by Adler [35]. One oscillator is termed as the injecting oscillator,
also called the master oscillator, as it is the source of the signal, and the other
oscillator is the free-running oscillator, or the slave oscillator, as it is the oscil-
lator to be locked to the injecting oscillator. As long as the frequency of the
RF signal is within the locking range of the slave oscillator, the slave oscillator
will become both frequency- and phase-locked to the incoming RF signal.
Adler developed mathematical expressions for the locking phenomena be-
tween the electronic oscillators and determined the relationship for locking
as a function of frequency detuning between the two oscillators. While the
initial work involved electronic oscillators, the theory can be expanded to any
oscillator, including optical oscillators such as lasers.
As mentioned above, the OEO clock synchronization performance can
be measured by looking at the phase noise of the OEO when it is injection-
locked. Following a similar derivation as that shown in [36], the phase noise
of the slave OEO can be described in relationship to the phase noise of the
injected signal. To begin, the frequency-locking range is defined as ωlock =
2ηωBW, where η = Ainj/Afree, ωBW = ωfree/2Q , Afree is the free-running OEO’s
signal amplitude, Ainj­ is the injected signal’s amplitude, Q is the quality factor
of the OEO, and ωfree = 2πf free is the free-running OEO’s angular frequency.
These two components give some insight into the injection locking process
of the OEO. First, the higher the Q of the OEO, the narrower the range of
frequencies over which the OEO can be injection-locked. Second, the stron-
ger the injected signal amplitude (or power) is compared to the free-running
OEO, the wider the locking range is. The locking range can be adjusted by
controlling these two parameters for use in a given application.
Now the phase noise transfer can be discussed. First, inj(ω), free(ω),
and OEO(ω) are defined as the single-sideband phase noise for the injected
signal, the free-running slave OEO RF signal, and the injection-locked OEO
oscillator, respectively. Then looking at the case where the signal is injected
into the OEO cavity, the following can be written [37]
96 Applications of Modern RF Photonics

2
 2ω 
 OEO ( ω) = Ptransfer ( ω) inj ( ω) + Ptransfer ( ω) free ( ω) 
 ωlock 
1 (4.3)
Ptransfer ( ω) = 2
 2ω 
1+ 
 ωlock 

where Ptransfer (ω) represents the phase noise transfer function of the oscillator
and has the form of a lowpass filter.
Figure 4.9 shows the calculated PSD of the phase noise, measured in
units of decibels relative to carrier power per one hertz bandwidth, as a func-
tion of offset frequency from the carrier for an injection-locked OEO. The
frequency-offset range is between 100 Hz and 10 MHz. The phase noise of
the free-running OEO and the injected signal both follow an ideal 1/f 2 rela-
tionship (dominated by white noise),­ with the free-running OEO phase noise
approximately 20 dB lower than that of the injected signal. For three differ-
ent locking ranges, the phase noise of the injection-locked OEO follows that
of the injected signal within the various locking ranges. Outside the locking
range, the phase noise decreases to the level of the free-running OEO phase
noise, following the slope of Ptransfer(w). With this description in place, some
different demonstrations of clock synchronization can be presented.
Using an OEO, clock synchronization was demonstrated by measur-
ing the phase noise of the OEO before and after injecting the cavity with an
RF signal [38]. The phase noise results are plotted in Figure 4.10. Like the
theoretical plot in Figure 4.9, the phase noise of the injected signal and the

Figure 4.9 Theoretical plot of the phase noise of an injection-locked OEO along with the
injected signal and free-running OEO’s phase noise. (After: [12].)
 Oscillators Utilizing RF Photonics 97

Figure 4.10 Measured plot of the phase noise of an injection-locked OEO along with the
injected signal and free-running OEO’s phase noise. (After: [12].)

free-running OEO are separate, and the phase noise of the injection-locked
OEO follows the injected signal until about 1 kHz offset and then decreases
to the phase noise of the free-running OEO. Note that the measured phase
noise does decrease with a –20 dB/decade slope as predicted by the theory
between the (–85 dBc at 1 kHz offset to –105 kHz at 10 kHz offset). In
order to increase the locking range, the injected optical power into the OEO
is increased by approximately 12 dB while keeping everything else the same.
With the higher injected power, the locking range increases just less than 20
times, which again matches the theoretical prediction.
The clock synchronization can also be measured in the time and fre-
quency domain. In this case, the RF output of the OEO is split to both
the input of an electrical spectrum analyzer (for frequency domain measure-
ments) and the trigger input of a high-speed sampling scope (for time domain
measurements). A portion of the injected signal is connected to the RF input
of the high-speed sampling scope. The frequency and time domain plots ap-
pear in Figure 4.11 for three different cases. In the first case, the injected
signal’s frequency is outside the locking range. The frequency measurement
shows the free-running signal and the injected signal, along with beat notes
between the two. The time measurement shows random noise as the trigger
signal is not phase-locked to the input signal of the scope. The second case
shows the free-running OEO and the injected signal almost phase-locked.
Again the frequency domain measurement shows beat notes, and the time
98 Applications of Modern RF Photonics

Figure 4.11 Measured plot of the frequency and time domain for the OEO when the inject-
ed signal is outside the locking range, just outside the locking range, and injection-locked.
(After: [12].)

domain measurement is still noisy. The third case shows when the OEO and
the injected signal are phase-locked. Now, the frequency measurement shows
a single tone, and the time domain measurement clearly shows the injected
sinusoidal signal. This demonstration shows that the OEO-generated signal
is locked to the injected signal and can be used for synchronization purposes.
Other examples of injection-locked OEOs exist using different modu-
lators, higher frequencies, and different injected signals [39–41]. Work con-
tinues to use these systems for clock synchronization for various applications.
While OEOs were originally used for clock signal generation and syn-
chronization, they have also been investigated for many other signal-pro-
cessing applications [12]. OEOs have very good phase noise performance.
Unfortunately, the trade-off is the ability to be tunable. The very nature of
the OEO limits its ability to be tuned very quickly in frequency. Some ap-
plications may require an oscillator that can be tuned very quickly.

4.4 Oscillators Based on Two Laser Sources

Beyond OEOs, other methods exist for generating clock signals using pho-
tonics. One of the most common methods involves the beat signal between
 Oscillators Utilizing RF Photonics 99

two lasers. Starting with two laser sources that operate at two different optical
frequencies denoted by ω1 and ω2, the two-laser-generated optical waves can
be mixed at a single photodetector. Assuming the polarization state of the
two lasers are the same at the photodetector and that the difference between
the two optical frequencies is less than the RF bandwidth of the photodetec-
tor, the generated RF photocurrent (IRF) can be expressed as IRF ∝ g[cos((w1
– w2)t)cos(q1(t) – q2(t)) – sin((w1 – w2)t)sin(q1(t) – q2(t)), where γ is a pro-
portional factor related to the power of the two lasers and the responsivity of
the photodetector θ(t)1,2 and are the time-varying phases of the two optical
waves, respectively. From the previous equation, the RF photocurrent will
have an oscillating frequency equal to the wavelength spacing between the
two lasers. However, the lasers have to be phase-locked to each other. If the
phases between the lasers are the same, then the resulting photocurrent is
just IRF ∝ g[cos((w1 – w2)t)]. If the lasers have a phase difference between
them, the mixing process due to photodetection will change the resulting RF
photocurrent. Over time, the amplitude will fluctuate, as the phases of the
two lasers change from a maximum to a minimum. In order to maximize the
RF photocurrent, various phase-locking techniques have been explored. An
optical phase lock loop (OPLL) uses an external oscillator to mix with the
generated RF signal to create an error signal that is then filtered and used to
tune one of the lasers [42]. However, the delay of the loop often has to be less
than a nanosecond in order to compensate the phase drifts [43]. Examples of
OPLLs that can meet these requirements have been shown [44].
Another method for locking the phases of two lasers is through the op-
tical injection locking process. Much like the injection locking of the OEO
discussed previously, the optical injection locking between two lasers allows
for the injecting or master laser to impose its phase coherence upon the in-
jected or slave laser. The process is faster than any electronic feedback loop
and can create very stable RF sinusoidal signals. An example of an optical
injection lock setup is shown in Figure 4.12. The output of the master laser
is injected into a slave laser, whose optical isolator has been removed. The
master laser imprints its phase variations on the slave laser, locking the slave
laser output to the master laser output. By keeping the master laser optical
frequency lower than the slave laser, the slave laser will be stably locked to
the master laser. The output of the slave laser, containing both the slave and
master outputs is then connected to the photodetector. Advanced versions
of this type of injection locking have been made, including using multiple
slave lasers [45], as well as modulating the master laser to create sidebands
that can be used to lock the slave laser [46]. The optical injection-locked
system can also be used in an OEO. While most OEO demonstrations use a
laser combined with a modulator, the laser can be directly modulated inside
100 Applications of Modern RF Photonics

Figure 4.12 Setup for using a master laser to optically injection lock a slave laser in order
to generate an RF signal with a frequency equal to the difference between each of the
lasers. (After: [47].)

the OEO cavity. In order to increase the oscillating frequency and improve
performance, an optically injected OEO (OIL-OEO) has been demonstrat-
ed [48]. In this configuration, a slave laser in the OEO cavity is optically
injection-locked by a master laser in the stable locking regime. The output of
the slave laser is then detected by a photodiode, and the resulting RF signal is
then filtered and fed back to the RF input of the slave laser. This allows the
OEO to generate a 20-GHz RF signal without the need for any additional
modulator. The resulting phase noise is just above –110 dBc/Hz at a 1-kHz
offset. Other OEOs have also been demonstrated using optical injection
locking of lasers to generate the RF tone [49]. Finally the combination of
both the OPLL technique and optical injection locking has been shown to
create stable oscillators [50].
One of the advantages to the two-laser system for RF signal generation
is the ability to tune the frequency of the signal continuously over a large
range and with high speed. While OEOs have been shown that can be tuned
in frequency [51, 52], the OEO will only allow frequencies that match one of
the cavity modes in the OEO to oscillate. The OEO cavity can be made long,
allowing the cavity modes to be spaced very close to one another. However
these closely spaced modes will still not allow true continuous tuning. The
two-laser system, on the other hand, does not depend on an external cav-
ity and can be continuously tuned over a wide frequency range. The master
slave can be tuned either by changing the temperature or drive current of the
laser. The temperature will change the length of the optical cavity, tuning the
optical frequency of the master laser output. However, temperature tuning is
very slow (on the order of seconds) and is not useful for most applications.
 Oscillators Utilizing RF Photonics 101

Tuning the pump current of the master laser has a much faster response. Yet,
changing the drive current will also affect other parameters of the master
and/or slave laser. Specifically, the optical power changes. In order to quan-
tify the changes, studies of the dynamics of the optical injection system have
been made [53, 54]. One interesting region that has been investigated is the
period-one oscillation region. As mentioned previously, when the master la-
ser’s optical frequency is lower than the slave laser’s optical frequency, stable
injection locking is achieved. On the other hand, when the master laser’s
optical frequency is higher than the slave laser’s frequency, a Hopf bifurcation
can be crossed, and the period-one oscillation region can be entered. When
in the period-one regime, the slave laser’s relaxation oscillations will become
undamped, allowing the slave laser’s frequency to shift to lower frequencies
[55]. The optical spectrum showing the shift in the slave laser’s wavelength
is shown in Figure 4.13. This region is very sensitive to the master laser’s
operating frequency and power. A result of operating in this region is that a
very fast tunable oscillator can be realized. Some demonstrations have shown
that the RF frequency can be tuned continuously over a range of 5 GHz in
approximately 5 ns [56]. This operation appears in Figure 4.14. The use of
the period-one oscillation has strong potential for making ultra-fast, continu-
ously widely tunable oscillators.

Figure 4.13 Optical spectrum of slave laser when the master laser is off (free-run slave)
and when the master is on (injected slave). The period-one locking forces the slave to red-
shift.
102 Applications of Modern RF Photonics

Figure 4.14 Demonstration of switching speed of photonically based oscillator. (After:

[47].)

4.5 Conclusions
RF photonic subsystems can be developed to act as oscillators in the demon-
stration architecture. OEOs can provide very precise clock signals for mul-
tiple applications. The demonstrated phase noise performance of an OEO
can meet or exceed the performance of other electronic oscillators [28]. The
OEO can also be used to synchronize a clock to an incoming RF signal, al-
lowing the signal to be captured on a sampling scope that requires a trigger
that matches the frequency of the RF input. Beyond OEOs, the use of two
phase-locked lasers can also be used to generate RF clock signals. The two-
laser system can be tuned continuously over a large band of frequencies. The
use of injection locking of two lasers has also been shown to generate new
oscillating tones due to nonlinear dynamics in the laser. Specifically the pe-
riod-one state has been shown to generate RF signals over a frequency range
of 10 GHz, at very high speeds. The ultrafast tunable oscillator based on the
period-one dynamics can be used as a local oscillator for downconverting
signals or simply to scan through a large spectrum of RF bandwidth. Chapter
5 covers more signal-processing applications using RF photonics, including
the oscillators described in this chapter.

References
[1] Vig, J., “Military Applications of High Accuracy Frequency Standards and Clocks,”
Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 40, No. 5, 1993,
pp. 522–527.
[2] Fukuchi, K., et al. “10.92-Tb/s (273 x 40-Gb/s) Triple-Band/Ultra-Dense WDM Op-
tical-Repeatered Transmission Experiment,” Optical Fiber Communication Conference,
2001.
[3] Shoop, B., Photonic Analog-to-Digital Conversion, New York, NY: Springer Verlag,
2001.
 Oscillators Utilizing RF Photonics 103

[4] Da Dalt, N., et al., “On the Jitter Requirements of the Sampling Clock for Analog-
to-Digital Converters,” Transactions on Circuits and Systems I: Fundamental Theory and
Applications, Vol. 49, No. 9, 2002, pp.1354–1360.
[5] Sartorius, B., “All-Optical Clock Recovery for 3R Optical Regeneration,” Optical Fiber
Communication Conference, 2001.
[6] Clark, T. R., et al., “Phase Noise Measurements of Ultrastable 10 GHz Harmonically
Mode Locked Fibre Laser,” Electronics Letters, Vol. 35, No. 9, 1999, pp. 720–721.
[7] Mule, A. V., et al., “Electrical and Optical Clock Distribution Networks for Gigascale
Microprocessors,” Transactions on Very Large Scale Integration (VLSI) Systems, Vol. 10,
No. 5, 2002, pp. 582–594.
[8] Miller, D. A. B. “Rationale and Challenges for Optical Interconnects to Electronic
Chips,” Proceedings of the IEEE, Vol. 88, No. 6, 2000, pp. 728–749.
[9] Harris, B., and B. Trubey, “10G-Enabled Optical Network Architecture Directions for
Video, Voice and Data: An MSO Perspective,” Optical Fiber Communication Confer-
ence, 2006.
[10] Redmayne, D., E. Trelewicz, and A. Smith, “Understanding the Effect of Clock Jitter
on High Speed ADCs,” Design Note 1013, Linear Technology Corporation, 2006,
http://cds.linear.com/docs/en/design-note/dn1013f.pdf.
[11] Kester, W., Analog-Digital Conversion, Analog Devices, 2004. ISBN13: 978-0-91655-
027-1.
[12] Devgan, P., “A Review of Optoelectronic Oscillators for High Speed Signal Processing
Applications,” International Scholarly Research Notices (ISRN) Electronics, Vol. 2013,
Article ID 401969, 2013.
[13] Navid, R., et al., “Close-In Phase Noise in Electrical Oscillators,” Proc. SPIE Symposium
on Fluctuations and Noise, 2004.
[14] Gheen, K., “Phase Noise Measurement Methods and Techniques,” Keysight Webcast,
July 2012.
[15] Von der Linde, D. “Characterization of the Noise in Continuously Operating Mode-
Locked Lasers,” Applied Physics B, Vol. 39, No. 4, 1986, pp. 201–217.
[16] Yao, X. S., and L. Maleki, “Optoelectronic oscillator for photonic systems,” Journal of
Quantum Electronics, Vol. 32, No. 7, 1996, pp. 1141-1149.
[17] Yao, X. S., and L. Maleki, “Converting Light into Spectrally Pure Microwave
Oscillation,” Optics Letters, Vol. 21, No. 7, 1996, pp. 483–485.
[18] Yao, X. S., and L. Maleki, “Optoelectronic Microwave Oscillator,” JOSA B, Vol. 13,
No. 8, 1996, pp. 1725–1735.
[19] Neyer, A. and E. Voges, “High-Frequency Electrooptic Oscillator Using an Integrated
Interferometer,” Applied Physical Letters, Vol. 40, No. 1, 1982, pp. 6–8.
104 Applications of Modern RF Photonics

[20] Nakazawa, M., T. Nakashima, and M. Tokuda, “An Optoelectronic Self-Oscillatory

Circuit with an Optical Fiber Delayed Feedback and Its Injection Locking Technique,”
Journal of Lightwave Technology, Vol. 2. No. 5, 1984, pp. 719–730.
[21] Lewis, M. F., “Novel RF Oscillator Using Optical Components,” Electronics Letters,
Vol. 28, No. 1, 1992, pp. 31–32.
[22] Lasri, J., et al., “Self-Starting Optoelectronic Oscillator for Generating Ultra-Low-Jitter
High-Rate (10 GHz or Higher) Optical Pulses,” Optics Express, Vol. 11, No. 12, 2003,
pp. 1430–1435.
[23] Romisch, S., et al., “Performance Evaluation of an Optoelectronic Oscillator,” Transactions
on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 47, No. 5, 2000, pp. 1159–
1165.
[24] K&L Filter Data Sheet, www.klmicrowave.com.
[25] Yao, X. S., et al., “Dual-Loop Opto-Electronic Oscillator,” Proc. of Frequency Control
Symposium, 1998.
[26] Yao, X. S. and L. Maleki, “Multiloop Optoelectronic Oscillator,” Journal of Quantum
Electronics, Vol. 36, No. 1, 2000, pp. 79–84.
[27] Shumakher, E., and G. Eisenstein, “A Novel Multiloop Optoelectronic Oscillator,”
Photonics Technology Letters, Vol. 20, No. 22, 2008, pp. 1881–1883.
[28] Banky, T., B. Horvath, and T. Berceli, “Optimum Configuration of Multiloop
Optoelectronic Oscillators,” JOSA B, Vol. 23, No. 7, 2006, pp. 1371–1380.
[29] Zhou, W., and G. Blasche, “Injection-Locked Dual Opto-Electronic Oscillator with
Ultra-Low Phase Noise and Ultra-Low Spurious Level,” Transactions on Microwave
Theory and Techniques, Vol. 53, No. 3, 2005, pp. 929–932.
[30] Okusaga, O., et al., “Experimental and Simulation Study of Dual Injection-Locked
OEOs,” Proc. of Frequency Control Symposium, 2009.
[31] Okusaga, O., et al., “Spurious Mode Reduction in Dual Injection-Locked Optoelectronic
Oscillators,” Optics Express, Vol. 19, No. 7, 2011, pp. 5839–5854.
[32] Devgan, P., et al., “A Low-Jitter Master-Slave Optoelectronic Oscillator Employing All-
Photonic Gain,” Proc. of International Topics on Microwave Photonics, 2007.
[33] Devgan, P. S., et al., “Improvement in the Phase Noise of a 10 GHz Optoelectronic
Oscillator Using All-Photonic Gain,” Journal of Lightwave Technology, Vol. 27, No. 15,
2009, pp. 3189–3193.
[34] Siegman, A. E., Lasers, South Orange, NJ: University Science Books, 1986.
[35] Adler, R., “A Study of Locking Phenomena in Oscillators,” Proc. of the IEEE, Vol. 61,
1973, pp. 1380–1385.
[36] Lasri, J., and G. Eisenstein, “Phase Dynamics of a Timing Extraction System Based on an
Optically Injection-Locked Self-Oscillating Bipolar Heterojunction Phototransistor,”
Journal of Lightwave Technology, Vol. 20, No. 11, 2002, pp. 1924–1932.
 Oscillators Utilizing RF Photonics 105

[37] Devgan, P., “High Speed Signal Processing Using Nonlinear Fibers and Optoelectronic
Devices,” Ph.D. Dissertation, Northwestern University, 2006.
[38] Lasri, J., et al., “Ultra-Low Timing Jitter 40Gb/s Clock Recovery Using a Self-Starting
Optoelectronic Oscillator,” Photonics Technology Letters, Vol. 16, No. 1, 2004, pp. 263–
265.
[39] Pan, S. and J. Yao, “Optical Clock Recovery Using a Polarization-Modulator-Based
Frequency-Doubling Optoelectronic Oscillator,” Journal of Lightwave Technology, Vol.
27, No. 16, 2009, pp. 3531–3539.
[40] Tsuchida, H., and M. Suzuki, “40-Gbps Optical Clock Recovery Using an Injection-
Locked Optoelectronic Oscillator,” Photonics Technology Letters,” Vol. 17, No. 1, 2005,
pp. 211–213.
[41] Tsuchida, H., “Subharmonic Optoelectronic Oscillator,” Photonics Technology Letters,
Vol. 20, No. 17, 2008, pp. 1509–1511.
[42] Langley, L., et al., “Packaged Semiconductor Laser Optical Phase-Locked Loop
(OPLL) for Photonic Generation, Processing and Transmission Of Microwave Signals,”
Transactions on Microwave Theory and Techniques, Vol. 47, No. 7, 1999, pp. 1257–
1264.
[43] Seeds, A., and K. Williams, “Microwave Photonics,” Journal of Lightwave Technology,
Vol. 24, No. 12, 2006, pp. 4628–4641.
[44] Gliese, U., et al., “A Wideband Heterodyne Optical Phase-Locked Loop for the
Generation of 3–18 GHz Microwave Carriers,” Photonics Technology Letters, Vol. 4, No.
8, 1992, pp. 936–938.
[45] Goldberg, L., et al., “Microwave Signal Generation with Injection Locked Laser
Diodes,” Electronics Letters, Vol. 19, No. 13, 1983, pp. 491–493.
[46] Schneider, G., et al., “Radiofrequency Signal-Generation System with Over Seven
Octaves of Continuous Tuning,” Nature Photonics, Vol. 7, No. 2, 2013, pp. 118–122.
[47] Devgan, P., “Signal Processing Utilizing Radio Frequency Photonics,” AFRL Report,
AFRL-RY-WP-TR-2017-0172, 2017
[48] Sung, H.-K., et al., “Optoelectronic Oscillators Using Direct-Modulated Semiconductor
Lasers Under Strong Optical Injection,” Journal of Selected Topics in Quantum Electronics,
Vol. 15, No. 3, 2009, pp. 572–577.
[49] Haji, M., et al., “High Frequency Optoelectronic Oscillators Based on the Optical
Feedback of Semiconductor Mode-Locked Laser Diodes,” Optics Express, Vol. 20, No.
3, 2012, pp. 3268–3274.
[50] Johansson, L., and A. Seeds, “Millimeter-Wave Modulated Optical Signal Generation
with High Spectral Purity and Wide Locking Bandwidth Using a Fiber-Integrated
Optical Phase-Lock Loop,” Photonics Technology Letters, Vol. 12, No. 6, 2000, pp. 690–
693.
[51] Tang, Z., et al., “Tunable Optoelectronic Oscillator Based on a Polarization Modulator
and a Chirped FBG,” Photonics Technology Letters, Vol. 24, No. 17, 2012, pp. 1487–
1489.
106 Applications of Modern RF Photonics

[52] Li, W., and J. Yao, “An Optically Tunable Optoelectronic Oscillator,” Journal of
Lightwave Technology, Vol. 28, No. 18, 2010, pp. 2640–2645.
[53] Simpson, T., “Mapping the Nonlinear Dynamics of a Distributed Feedback
Semiconductor Laser Subject to External Optical Injection,” Optics Communications,
Vol. 215, No. 1, 2002, 2002, pp. 135–151.
[54] Wieczorek, S., et al., “The Dynamical Complexity Of Optically Injected Semiconductor
Lasers,” Physics Reports, Vol. 416, No. 1, 2005, pp. 1–128.
[55] Hwang, S., J. Liu, and J. White, “Characteristics of Period-One Oscillations in
Semiconductor Lasers Subject to Optical Injection,” Journal of Selected Topics in
Quantum Electronics, Vol. 10, No. 5, 2004, pp. 974–981.
[56] Usechak, N., J. Suelzer, and J. Haefner, “High-Speed Wideband Voltage-Controlled
Oscillator via an Injection-Locked Laser,” Photonics Technology Letters, Vol. 29, No. 13,
2017, pp. 1132–1135.
5
Signal Isolation Utilizing RF Photonics
In Chapters 1–3, the demonstration architecture has only considered pro-
cessing incoming RF signals from the antenna. However, the system can also
be used to transmit RF signals. One possible architecture could include two
antennas, one that is used to transmit signal and another used to receive. A
more elegant solution, however, would use a single antenna for both trans-
mission and reception. To accomplish both transmit and receive functions
with a single antenna, the transmission path must be isolated from the receive
path. This is accomplished with an RF circulator, which is connected right
after the antenna, as shown in Figure 5.1. The three-port circulator device
isolates the transmit path to the antenna from the receive path. On the receive
side, the architecture will often need to be able to sense only a certain part of
the overall spectrum. Being able to filter out specific signals or sets of signals
can be useful. The receive path has a new subsystem, which can separate
multiple received signals before processing. RF photonics can provide solu-
tions for both the RF circulator and filters. The photonic-based filters can be
widely tunable and have narrow frequency bandwidths. The photonic-based
circulator can provide high isolation between transmit and receive paths. This
chapter covers isolation and separation of RF signals, providing examples of
how RF photonic applications can address these needs.

5.1 Need for Signal Separation

The RF spectrum is increasingly being filled with new signals. While tradi-
tional amateur transoceanic radio communications operated in the HF band

107
108 Applications of Modern RF Photonics

Figure 5.1 Demonstration architecture with RF circulator and RF signal separation sub-
systems.

(known as HAM radio) and analog broadcast television used the very high
frequency (VHF) and lower ultrahigh frequency (UHF) bands, other signals
began to use these frequency bands as well. Frequencies in the VHF band
(118–136 MHz) are used for air traffic control as well as for civil emergency
radio communications (138–144 MHz) [1]. Extending into the UHF band,
the 400-MHz frequency is used for time and frequency standard transmis-
sion to satellites. Cordless landline telephones use the 900-MHz frequency
band but have extended into the 2.5- and 5-GHz bands. Equivalently, the
use of home Wi-Fi routers has also made use of these same frequency bands
for connecting computers, tablets, and smart phones to the internet. The
1.2- and 1.5–1.6-GHz frequencies are used for commercial GPS signals.
The 1.675-GHz frequency is used for meteorological satellite data, while the
2.7–2.9 GHz band is used for airport surveillance radars. Looking beyond
the UHF band, the 4.2–4.4-GHz frequency band is used for airborne radar
altimeters. The 9–9.5-GHz range is used for precision approach radars as
well as airborne weather radars. Speed guns operate at 10.525 GHz, as well
as 24.25 GHz. The 15.7–16.2-GHz band is used for airport surface detec-
tion equipment, and the automotive industry uses the 76–77-GHz band for
collision and road radars [1]. In general, the NTIC long-range spectrum [2]
use plan predicts that the 2–18-GHz frequency spectrum will continue to
be used with increases in use from 8.5 to 10 GHz and 14 to 15.5 GHz. At
higher frequencies, the increase will be seen in the 31–36-GHz band for 5G
communication as well as the 77-GHz range for automotive radar and road
sensing. As the congestion of signals in these frequency bands increases, better
isolation will be needed. To accomplish this, the first consideration would be
 Signal Isolation Utilizing RF Photonics 109

the use of RF filters, specifically widely tunable filters that also have narrower
passbands than those of traditional electronic filters.

5.2 Using RF Photonics for Separation of Signals

As seen in Figure 5.1, the receive path of the architecture has a new subsystem
for signal separation. The subsystem permits certain signals to continue to the
processing block while removing others. In essence the subsystem acts as an
RF filter. The easiest example of an RF filter using photonics is simply the
use of an optical filter. When using an optical filter, the modulation format
has to be considered. Since most modulation formats are dual-sideband, the
bandwidth of the optical filter will affect the RF bandwidth. An example is
shown in Figure 5.2. Assuming an optical filter with a given bandwidth, the
dual-sideband modulation will only pass through RF signals of one-half the
filter bandwidth. As seen in Figure 5.2(a), the optical carrier ideally sits in the
middle of the optical filter, so that the sidebands see the same filter response.
The use of dual-sideband modulation can be a benefit as it allows an optical
filter of twice the needed bandwidth to be used. On the other hand, if the full
bandwidth of the optical filter is necessary, single-sideband modulation can
be used, as seen in Figure 5.2(b). In either case, the modulation format will
play a role in the bandwidth of the optical filter that can be used. Note that
most optical filters are based on a cavity and thus have a periodic structure in
their response. The passband will repeat at a frequency spacing known as the
free spectral range (FSR) [3]. Often, the narrower the passband, the narrower
the FSR. The FSR can actually limit the wavelength range over which the
filter will be usable and has to be taken into consideration.
The most common metric for the filter is the quality factor, also known
as the Q-factor. The Q factor is defined as Q = Fc/∆F , where Fc is the center

Figure 5.2 Optical filter with either (a) dual-sideband modulation (left) or (b) single-side-
band modulation (right). (After: [4].)
110 Applications of Modern RF Photonics

frequency of the filter, and ∆F is the full width at half-maximum bandwidth

of the filter. For typical electronic-based filters, Q factors between 1,000 and
10,000 are easily achievable. At a center frequency of 10 GHz, a bandwidth
between 100 and 1,000 MHz can be realized. Often, these can be too wide
for the chosen operation, and narrower bandwidth filters are required. Opti-
cal filters can provide much higher Q factors than electronic solutions.
Optical filters have been realized in many different ways, including us-
ing SBS [5], ring resonators [6], and thin-film materials [7]. Nevertheless,
here we focus on the fiber Bragg grating filter. The Bragg grating filter is
realized in optical fibers by writing a grating onto a section of optical fiber.
The grating is usually written on the optical fiber. Changes in the refractive
index of the fiber create a periodic grating structure in the fiber. Light of a
certain wavelength will constructively add up as a reflection off of the grating.
Other wavelengths will not constructively reflect and will pass through the
grating as seen in Figure 5.3. This allows the filter to act as either a bandpass
(in reflection) or a notch filter (in transmission). The wavelength that reflects
meets the condition Λ = l/2 where λ is the wavelength of the reflected light
and Λ is the period of the grating. Fiber Bragg gratings are commercially
available, with bandwidths ranging from 50 MHz to 50 GHz. One of the
drawbacks to these types of filters are their inability to easily tune the center
frequency. In order to change the center frequency, the spacing between the
grating lines of the Bragg filter has to be changed. Changing the spacing can
be accomplished by either stressing the fiber or changing the temperature
of the fiber [8]. In either case, the idea is to physically stretch or compress
the fiber so the spacing of the grating lines change. This in turn changes the
wavelength that will reflect at the grating as given by the relationship above.
Stretching the fiber is a slow process and can only be done over a limited
amount before causing damage to the fiber. In order to overcome this limita-
tion, other filter techniques have been investigated.

Figure 5.3 A fiber Bragg grating with a broadband optical source at the input. The trans-
mitted response acts like a notch filter while the reflected response is a bandpass filter.
(After: [4].)
 Signal Isolation Utilizing RF Photonics 111

5.3 Finite Impulse Response Filters Using RF Photonics

A more advanced method for generating an RF filter using RF photonics is
the use of delay lines. Unlike the delay lines used for analog storage as de-
scribed in Chapter 4, the delay lines are used to create time delays between
the same RF signal. This creates the equivalent of a finite impulse response
(FIR) filter [9]. Before discussing the optical implementation, a brief discus-
sion of the FIR filter is warranted. For linear, time-invariant systems, the
output of the system y(t) can be calculated by taking a continuous time con-
∞
volution y (t ) = ∫ h ( τ )s (t − τ ) d τ evaluated from time t = –∞ to ∞ between
−∞
the input signal s(t) and the impulse response of the system h(t). The convo-
lution operation can also be done in the discrete time domain. Discrete time
signals are time-sampled versions of continuous time signals. Figure 5.4(a)
shows a continuous time varying signal, and Figure 5.4(b) shows the discrete
time version, where T is the sample time period. Discrete time signal process-
ing is used in computers to perform operations, such as filtering signals. The
equivalent convolution in the digital domain is the discrete convolution sum
∞
y[n ] = ∑ m =−∞ h[m ]s[n − m ] [10]. In this case the integral is replaced with a
summation of the input signal s(t) sampled at T, yielding s[n], where n is
the total number of samples. The discrete impulse response is denoted as
h[n], which is also a sampled version of h(t) at a sample rate of T. When the
number of samples is infinite, the system is said to have an infinite impulse
response (IIR). Since practical implementations cannot be infinite, a finite
number of samples is used. This is now called a FIR system. As suggested ear-
lier, the simplest example of signal processing is providing a filtered response

Figure 5.4 (a) A continuous time plot of the sinc function and (b) a sampled version of the
same sinc function in (a). (After: [4].)
112 Applications of Modern RF Photonics

of an input signal. One of the easiest examples of filtering is a low-pass filter,

in which only signals lower than a cutoff frequency pass through without
modification, while signals of a higher frequency are forced to zero. In order
to realize a practical implementation of this type of filter, a FIR filter can be
N −1
used. The mathematical expression is given here y (t ) = ∑ m = 0 h[m ]s[t − mT ]
[9]. The FIR filter is simply the discrete convolution sum of the sampled
impulse response of a given filter shape with multiple time-delayed versions
of the signal. An example is shown in Figure 5.5. A sinc function in the time
domain will have a low-pass filter response in the frequency domain. The sinc
function is defined as sinc(t) = sin(πt)/(πt). The sampled version of the sinc
function is shown in Figure 5.5(a). The sampled points of the sinc are used as
filter coefficients h[m] in the FIR convolution sum. The number of samples is
often referred to as the number of taps in the FIR filter. When the input is an
impulse, the resulting frequency response is plotted in Figure 5.5(b), with N
= 40. The shape of the filter is that of a low-pass filter. Yet, it does not match
the ideal response. Clearly, this is because only a finite sample of the sinc
function was used. Regardless, this is a way to realize a filter response with a
finite number of elements or taps.
In order to realize an FIR filter in practice, three operations need to
be performed. First, the incoming signal needs to be replicated N-number
of times. Each of the replicas then needs to be time-delayed and attenuated.
Finally, the N-number of time-delayed signals has to be summed together.
This accomplishes in practice the convolution sum of the FIR filter. Photonic
implementations of an FIR filter have the advantage of providing time delays
that can be very low-loss as discussed previously. Photonics also provides the
use of multiple wavelengths that can be modulated by the same RF signals

Figure 5.5 (a) A sampled sinc function with 40 samples and (b) the resulting low-pass filter
response with an impulse input using the sinc coefficients from (a).
 Signal Isolation Utilizing RF Photonics 113

and then split to multiple paths. A photonic-based FIR filter can implement
this in a couple of different ways.
Figure 5.6 illustrates the first example. The output of an optical source,
consisting of multiple, distinct wavelength, optical carriers, is connected to
an optical modulator to imprint the RF waveform onto each of the wave-
lengths. The output of the optical modulator is then connected to a wave-
length demultiplexer. The demultiplexer separates the different wavelengths
into parallel paths. Each of the paths is then attenuated and passed through
a delay line. Each of the delay lines is one time period, T, longer than the
one before it. At the end, the different paths are then recombined using a
wavelength multiplexer, whose output is connected to a photodiode. The
photodiode then takes the different wavelengths and sums up the RF signals.
This accomplishes the convolution sum of the FIR filter. The disadvantage of
this architecture is that each of the fiber delay lengths has to be maintained
relative to the other delays. External environmental perturbations could cause
the lengths of the different fibers to change independent of each other, essen-
tially changing the sample time of each tap of the filter. A phase shifter can be
added after each attenuator to correct environmental perturbations, as seen
in Figure 5.6. On the other hand, an alternative architecture can address this
issue. In Figure 5.7, the same multiwavelength optical source and modulator
is used. Instead of separating the wavelengths first, they all pass through the
same optical fiber. The fiber has multiple Bragg gratings written into it, like
the one shown in the Section 5.2. Each of the Bragg gratings is set to reflect
a specific wavelength of the optical source. Each of the Bragg gratings is also
spaced by a length that yields a delay of T/2. In this way, the first wavelength
sees a delay of T, the second wavelength sees a delay of 2T, the third wave-
length sees a delay of 3T, and so on. All of the reflected wavelengths are then
collected onto a single photodiode for the addition operation. To weight each
of the signals with the correct impulse response coefficient, the reflectivity of
the Bragg grating can be changed [11], or each wavelength’s optical power

Figure 5.6 An RF photonic implementation of an FIR filter using multiwavelength source

and different time delays. (After: [4].)
114 Applications of Modern RF Photonics

Figure 5.7 An RF photonic implementation of an FIR filter using a multiwavelength source

and different Bragg gratings. (After: [4].)

can be set at the optical modulator. The single-path implementation of the

FIR filter provides a single path for all the wavelengths, overcoming the par-
allel path architecture’s sensitivity to external influences. Still, this advantage
comes at the expense of not being able to change the weights of each time
delay as easily as in the parallel path architecture.
There are a few caveats that must be discussed when discussing the pho-
tonic implementation of an FIR filter. Ideally, the FIR filter acts to add up the
different time-delayed and amplitude-weighted versions of the RF waveform.
However, the photonic implementation of the FIR filter still has the optical
carrier. The optical carrier will also see the time delays and weights and will
interfere at the photodetector, degrading the performance of the FIR filter in
the RF domain. Thus, the use of either a multiwavelength source, as seen in
Figures 5.6 and 5.7, or a broadband optical source is necessary to achieve the
desired FIR filter response [9]. Another consideration is the weighting func-
tion that is to be implemented for the FIR filter. Looking at the sampled sinc
function in Figure 5.5(a), the samples consist of both positive and negative
numbers. In practice, photonic implementations of the FIR filters cannot
provide negative weights to the time delays. Work has been shown to allow
for negative weights [12], including using phase shifts in the modulator to
implement negative weights [13], using differential photodetection [14] and
other methods [15–17]. Unfortunately, these require more complex setups
than the simple implementations shown here. One last consideration is that
the FIR filter is a periodic filter. The response of the filter will repeat with a
separation denoted by the FSR. This is a consequence of the sampling nature
of the FIR filter implementation. The FSR is related to the sample time pe-
riod T by the simple relation FSR = 2π/T. Multiple implementations have
been investigated to generate filter responses that do not repeat (i.e., single-
passband filters), including cascading filters of different FSRs that have a
single passband in common [18], or a combination of an asymmetric Mach-
Zehnder interferometer (AMZI) and an optical modulator to select only one
mode of the AMZI [19].
 Signal Isolation Utilizing RF Photonics 115

The FIR filters have been shown to be very flexible and adaptable. The
weights at each of the time delays can be set by variable optical attenuators.
Therefore, the weights can be dynamically changed. This will allow an FIR
filter to go from a low-pass response to a high-pass response simply by flip-
ping every other coefficient in the filter weights (also known as spectral inver-
sion) [20]. In addition, the center frequency of the FIR filter response can be
tuned. Tuning is often achieved by the use of a tunable laser source combined
with a chirped fiber Bragg grating [21, 22]. In either case the change in wave-
length will result in different time delays for each of the delay lines, resulting
in different center frequencies for the filter. Since the wavelength of the laser
can be tuned very quickly, the FIR filter can be tuned at much faster speeds
than the fiber Bragg grating described earlier. Many other methods have also
been investigated for making tunable frequency FIR filters [23–25].

5.4 Isolation of RF Signals Along a Common Path

Clearly, there are many examples of separating RF signals using photonic
methods. Different filter architectures can be used to cut out certain signals
from a larger spectrum, allowing them to be processed without being affected
by neighboring signals. Beyond separation of signals, another operation that
is of use is the isolation of transmit and receive signals at a single antenna.
Multiple applications require the capability to both transmit and receive from
the same antenna. Examples include radar, radio communication, and wire-
less routers. One method commonly used is a time-gating method. Time
gating means that the antenna is set for a certain time in transmission mode
and then switched to reception mode. High-speed T/R, or transmit/receive,
switches have been used in time-gating architectures [26]. While this works
for some applications, the time-gating method will have dead times, where
signals of interest are not being received. For certain applications, the dead
time is unacceptable.
To overcome this limitation, an RF circulator can be used to isolate the
transmitted and received signals from the same antenna. Figure 5.8 shows a
close-up of the RF circulator pictured in Figure 5.1. As mentioned previously,
the RF circulator is typically a three-port device, where signals to be transmit-
ted enter port 1 and are passed onto port 2 with very low loss (~1 dB). Port
2 is connected to the antenna, where signals can both be transmitted and re-
ceived simultaneously, without any time gating. The signals that are received
enter port 2 and are passed to port 3 with low loss (~1–2 dB). The losses be-
tween ports 1 and 3 are very high (> 20 dB), as well as the loss for any signal
entering port 2 and leaking out to port 1 (> 25 dB). RF circulators provide
116 Applications of Modern RF Photonics

Figure 5.8 An RF circulator connected to an antenna to transmit and receive simultane-

ously.

very good isolation between the receive and transmit paths. Unfortunately,
ferrite-based RF circulator technology cannot easily cover wide frequency
ranges while maintaining high isolation. Wideband RF circulators have been
demonstrated that provide an isolation of 10 dB at frequencies above 5 GHz
to just under 20 GHz [27]. At higher frequencies, the isolation begins to
degrade while the insertion loss increases. Methods have been investigated
to improve the isolation by using either multiple circulators [28] or matched
pairs of circulators with antennas [29].
In order to provide larger isolation of the transmit and receive paths
from a single antenna over a wide frequency range, RF photonics can pro-
vide a solution. Combining a bidirectional signal interface using an optical
modulator [30] with single-sideband modulation multiplexing [31] allows
for high isolation between two counterpropagating RF signals traveling on
a common path [32]. As shown in Figure 5.9, the configuration acts like a
three-port RF circulator. The RF signal for transmission is input to port 1.

Figure 5.9 An RF photonic-based circulator for isolating transmit and receive paths. (Af-
ter: [32].)
 Signal Isolation Utilizing RF Photonics 117

The transmit signal is then split by a 90° hybrid onto the two RF waveguide
electrodes in the MZM, with the zero-degree output on the top waveguide
electrode and the 90° output on the bottom waveguide electrode. Note that
the RF signal from port 1 counterpropagates with respect to the optical field,
which is input to the MZM from the laser at the optical input of the MZM.
The split signals then propagate to another 90° hybrid, which is placed in a
complementary arrangement to the first hybrid. For this case, the 90° input
of the second hybrid is connected to the top RF electrode waveguide, and the
zero-degree input is connected to the bottom RF electrode waveguide. The
recombined RF signal is then transmitted out port 2 through the antenna.
The antenna can also receive signals at the same time. Thus, on port 2, an-
other RF signal can be received. The received signal is split by the second 90°
hybrid, with the 90° output on the top RF electrode waveguide and the zero-
degree output on the bottom RF electrode waveguide. Note that the received
RF signal will copropagate with the optical field of the laser in the MZM and
counterpropagate to the transmitted RF signal. The recombined received sig-
nal can be transmitted out port 1. Both RF signals will be upconverted onto
the optical carrier frequency of the laser. With the correct DC bias chosen for
each electrical waveguide, the resulting optical output will be single sideband-
modulated. Since the transmit RF signal passes through the complementary
90° hybrid as compared to the received RF signal, the transmit RF signal will
appear on the opposite side of the optical carrier as compared to the received
RF signal. The optical output of the MZM is then passed through an optical
filter in order to remove the unwanted transmit RF signal’s sideband, while
preserving the sideband with the received RF signal. The output of the filter
is connected to a photodetector where the recovered RF signal is then output
at port 3.
A mathematical description of the single-sideband modulation is shown
in (5.1). The MZM has a transfer function given by

1 i j1 (t ) i j2 (t )
E out (t ) = e −e  E in (t ) (5.1)
2 

In the case of the RF signal entering port 2, the phase modulation on

each arm of the MZM after the 90° hybrid can be written as ϕ1(t) = ϕdc1
+ϕrf 1 sin(Ωrf t) and ϕ2(t) = ϕdc2 +ϕrf 2 cos(Ωrf t), with ϕdc1,2 = π(Vdc1,2/Vπ,dc1,2)
and ϕrf 1,2 = π(Vrf 1,2/Vp,rf 1,2(Ωrf )).Now using the Jacobi Anger expansion
∞ ∞
e iz cos θ = ∑ n =−∞ i n J n (z )e in θ and e iz sin θ = ∑ n =−∞ J n (z )e in θ the following can
be written
118 Applications of Modern RF Photonics

E in e i ωot i jdc 1
E carrier (t ) = e J 0 ( jrf 1 ) − e i jdc 2 J 0 ( jrf 2 ) ,
2 
i ωot −i Ωrf t
E in e (5.2)
E usb , fund (t ) =  −e i jdc 1 J 1( jrf 1 ) − ie i jdc 2 J 1 ( jrf 2 ) ,
2  
i ωot +i Ωrf t
E in e
E lsb , fund (t ) = e i jdc 1 J 1 ( jrf 1 ) − ie i jdc 2 J 1 ( jrf 2 ) ,
2  

In this case, the RF signal is the same on both electrical waveguides

so ϕrf 1 = ϕrf 2. For the condition of single-sideband operation, there are two
cases. For the upper sideband to be nulled, ϕdc1 = –π/2 + ϕdc2. For the lower
sideband to be nulled, ϕdc1 = π/2 + ϕdc2. Now for the RF signal input to
port 2, the hybrid is in the complementary configuration so ϕ1(t) = ϕdc1
+ϕrf 1cos(Ωrf t) and ϕ2(t) = ϕdc2 +ϕrf 2sin(Ωrf t). In this configuration, the re-
sulting fields are written as

E in e i ωot i jdc 1
E carrier (t ) = e J 0 ( jrf 1 ) − e i jdc 2 J 0 ( jrf 2 ) ,
2 
i ωot −i Ωrf t
E in e
E usb , fund (t ) = ie i jdc 1 J 1( jrf 1 ) + e i jdc 2 J 1( jrf 2 ) , (5.3)
2  
i ωot +i Ωrf t
E in e
E lsb , fund (t ) = ie i jdc 1 J 1( jrf 1 ) − e i jdc 2 J 1( jrf 2 ) ,
2  

Now for the single-sideband condition, the upper sideband is nulled

when ϕdc1 = π/2 +ϕdc2, and the lower sideband is nulled when ϕdc1 = –π/2
+ ϕdc2. From this one can see that the RF signals will always appear on op-
posite sides of the optical carrier when the bias conditions are set for single-
sideband modulation. This allows for the RF signals to be separated in the
optical spectrum from each other. Then, using an optical filter will allow for
one sideband to be removed while the carrier and the other sideband will pass
on to the photodetector to recover the RF signal at port 3.
The RF circulator using photonics has many advantages compared to
the ferrite-based circulator, beyond just the increased RF bandwidth. Because
the transmit signal is removed by the optical filter before the detector, the RF
isolation between port 1 and port 3 can be infinite in theory. The insertion
loss between port 1 and port 2 can be relatively low, as the signals are split
and then recombined between the two ports. The MZM has been shown to
 Signal Isolation Utilizing RF Photonics 119

handle RF powers as high as 20W when used in a bidirectional application

[30]. However, there are a few drawbacks to this solution. Since most of the
ferrite-based RF circulators are passive devices, they only add loss between
ports. However, the photonic-based RF circulator is an active device and thus
adds noise and distortion. The system from port 2 to port 3 is essentially an
RF photonic link. Therefore it suffers from the drawbacks of a traditional RF
photonic link. The loss, NF, and linearity of the system from port 2 to port 3
will be determined by the laser RIN and optical power, as well as the modula-
tion efficiency (Vπ) of the MZM, as shown in Chapter 2. The application will
decide what performance is needed for the receive link.

5.5 Conclusions
RF signal separation and isolation can play a significant role in different ap-
plications. As the RF spectrum continues to be exploited for various uses, the
number of signals will increase over time. The ability to separate signals of in-
terest from unwanted signals will continue to be a necessary function for most
architectures. Beyond simply separating signals, the FIR filters presented in
this chapter can be used to recover specific types of signals. By knowing the
specific encoding of certain signals, the FIR system can be used as a matched
filter to recover and decode these signals. In a congested RF environment,
these types of encoded signals will be useful to maximize spectral efficiency.
The cellular phone networks already use these types of codes to enable mul-
tiple users to access the service in a limited frequency bandwidth [33]. In ad-
dition, being able to both transmit and receive signals from a single antenna
is also of great benefit. Often, physical space or cost limits a system to only a
single antenna. While T/R switches can be used to time-gate the transmit and
receive signals, the ultimate solution would allow for simultaneous use of the
antenna for both transmission and reception. The RF circulator provides the
capability to both transmit and receive RF signals by providing isolation be-
tween the transmit and receive paths. While the traditional ferrite-based RF
circulators can provide isolation of signals around 10 dB, RF photonic solu-
tions have shown isolation between 20 and 40 dB over equivalent operational
frequencies. Clearly, the RF photonic circulator can overcome the limitations
of traditional circulator technologies.

References
[1] Stine, J., and D. Portigal, “Spectrum 101: An Introduction to Spectrum Management,”
MTR 04W0000048, MITRE, 2004.
120 Applications of Modern RF Photonics

[2] U.S. Spectrum Requirements, Projections and Trends, Ch. 4, www.ntia.doc.gov/legacy/

osmhome/EPS/openness/sp_rqmnts/radar4/html.
[3] Hecht, E., Optics, Boston, MA: Addison-Wesley, 1998.
[4] Devgan, P., “Signal Identification and Isolation Utilizing Radio Frequency Photonics,”
AFRL Report, AFRL-RY-WP-TR-2017-0158, 2017.
[5] Tanemura, T., Y. Takushima, and K. Kikuchi, “Narrowband Optical Filter, with a
Variable Transmission Spectrum, Using Stimulated Brillouin Scattering in Optical Fi-
ber,” Optics Letters, Vol. 27, No. 17, 2002, pp. 1552–1554.
[6] Barwicz, T., et al., “Microring-Resonator-Based Add-Drop Filters in SiN: Fabrication
and Analysis,” Optics Express, Vol. 12 No. 7, 2004, pp. 1437–1442.
[7] MacLeod, H. A., Thin-Film Optical Filters, Boca Raton, FL: CRC Press, 2001.
[8] Hill, K. O., and G. Meltz, “Fiber Bragg Grating Technology Fundamentals and Over-
view,” Journal of Lightwave Technology, Vol. 15, No. 8, 1997, pp. 1263–1276.
[9] Capmany, J., B. Ortega, and D. Pastor, “A Tutorial on Microwave Photonic Filters,”
Journal of Lightwave Technology, Vol. 24, No. 1, 2006, pp. 201–229.
[10] Oppenheim, A. V., and W. Schafer, Discrete Time Signal Processing, Upper Saddle River,
NJ: Prentice Hall, 1989.
[11] Yao, J., “A Tutorial on Microwave Photonics,” IEEE Photonics Society Newsletter, 2012.
[12] Minasian, R., E. Chan, and X. Yi, “Microwave Photonic Signal Processing,” Optics
Express, Vol. 21, No. 19, 2013, pp. 22918–22936.
[13] Capmany, J., et al., “Microwave Photonic Filters with Negative Coefficients Based on
Phase Inversion in an Electro-optic Modulator,” Optics Letters, Vol. 28, No. 16, 2003,
pp. 1415–1417.
[14] Sales, S., et al., “Experimental Demonstration of Fibre-Optic Delay Line Filters with
Negative Coefficients,” Electronics Letters, Vol. 31, No. 13, 1995, pp. 1095–1096.
[15] Coppinger, F., et al., “All-Optical RF Filter Using Amplitude Inversion in a
Semiconductor Optical Amplifier,” Transactions on Microwave Theory and Techniques,
Vol. 45, No. 8, 1997, pp. 1473–1477.
[16] You, N., and R. A. Minasian, “Synthesis of WDM Grating-Based Optical Microwave
Filter with Arbitrary Impulse Response,” in Proc. of International Topical Meeting on
Microwave Photonics, 1999.
[17] Loayssa, A., et al., “Demonstration of Incoherent Microwave Photonic Filters with All-
Optical Complex Coefficients,” Photonics Technology Letters, Vol. 18, No. 16, 2006, pp.
1744–1746.
[18] You N., and R. Minasian, “High-Q Optical Microwave Filter,” Electronics Letters,
Vol. 35, No. 24, 1999, pp. 2125–2126.
[19] Capmany, J., et al., “High-Q Microwave Photonic Filter with a Tuned Modulator,” Optics
Letters, Vol. 30, No. 17, 2005, pp. 2299–2301.
 Signal Isolation Utilizing RF Photonics 121

[20] Smith, S., The Scientists and Engineers Guide to Digital Signal Processing, 1997, www.
DSPguide.com.
[21] Zhang, W., et al., “Fibre Optic Radio Frequency Notch Filter with Linear And
Continuous Tuning by Using a Chirped Fibre Grating,” Electronics Letters, Vol. 34, No.
18, 1998, pp. 1770–1772.
[22] Hunter, D. A., and R. Minasian, “Tunable Transversal Filter Based on Chirped
Gratings,” Electronics Letters, Vol. 31, No. 25, 1995, pp. 2205–2207.
[23] Yan, Y., and J. Yao, “A Tunable Photonic Microwave Filter with a Complex Coefficient
Using an Optical RF Phase Shifter,” Photonics Technology Letters, Vol. 19, No. 19, 2007,
pp. 1472–1474.
[24] Frankel, M., and R. Esman, “Fiber-Optic Tunable Microwave Transversal Filter,”
Photonics Technology Letters, Vol. 7, No. 2, 1995, pp. 191–193.
[25] Mora, J., et al., “Photonic Microwave Tunable Single-Bandpass Filter Based on a Mach-
Zehnder Interferometer,” Journal of Lightwave Technology, Vol. 24, No. 7, 2006, pp.
2500–2509.
[26] Skolnik, M., Radar Handbook, New York, NY: McGraw-Hill Education, 2008.
[27] Schloemann, E. F., “Circulators for Microwave And Millimeter-Wave Integrated
Circuits,” Proceedings of the IEEE, Vol. 76, No. 2, 1988, pp. 188–200.
[28] Ho, T. Q., et al., “Wideband High Isolation Circulator Network,” United States Patent
No. 5,815,803, 1998.
[29] Ho, T. Q., “Matched Pair Circulator Antenna Isolation Circuit,” United States Patent
No. 6,313,713, 2001.
[30] Ackerman, E. I., and C. H. Cox. “Optimization of Photonic Transmit/Receive Module
Performance,” Proc. of International Topical Meeting on Microwave Photonics, 2009.
[31] Narashima, A., et al., “Tandem Single Sideband Modulation Scheme for Doubling
Spectral Efficiency of Analogue Signals,” Electronics Letters, Vol. 36, No. 13, 2000, pp.
1135–1136.
[32] Devgan, P., “Isolation of RF Signals Using Optical Single Side Band Modulation
Combined with Optical Filtering,” United States Patent No. 9,240,842, 2016.
[33] Schulze, H., and C. Lueders, Theory and Applications of OFDM and CDMA: Wideband
Wireless Communications, Hoboken, NJ: John Wiley & Sons, 2005.
6
Signal Identification Utilizing RF
Photonics
Over the previous chapters, the demonstration architecture has evolved to in-
corporate an RF circulator after the antenna, a remoting link for the antenna,
and oscillators for signal generation. The next subsystem to be considered is
one that allows for the identification of signals using RF photonics. A specific
example is determining the center frequency of signals that are received at the
antenna. As seen in Figure 6.1, the subsystem takes the RF signals from the
antenna and then identifies the center frequencies. The information is then
passed to the photonic filters and oscillators to assist with later processing of
signals. Since the architecture is designed to have a very wideband capability,
multiple signals will be received at the input of the system. Some of these sig-
nals will be of interest to the receiver, while others are meant to be ignored. In
order to find the signals that are of interest, the center frequency of the vari-
ous signals needs to be determined. The photonic filter can then be tuned to
the center frequency in order to pass than signal and suppress the unwanted
ones. In the case of the oscillator, it can be tuned to the center frequency for
clock synchronization purposes. This chapter reviews methods for identifying
the center frequency. In addition, the chapter presents a couple of methods
for not only identifying the center frequency but also selectively amplifying
chosen signals while suppressing other signals.

123
124 Applications of Modern RF Photonics

Figure 6.1 Demonstration architecture with signal identification block added.

6.1 Need for Signal Identification

The identification of the frequency of received RF signals is important in
many fields, including radio astronomy. Radio astronomers often look for
molecular and atomic spectral resonances in order to determine where celes-
tial bodies are and what they consist of. The molecular frequencies cover a
large range, including hydrogen at 1.4 GHz, methanol at 12.17 GHz, am-
monia at 23.8 GHz, and carbon monoxide at 806 GHz [1]. Another appli-
cation would be for a survey of the local RF environment. As mentioned in
Chapter 5, the RF spectrum is becoming more and more crowded. Having
the ability to identify what signals exist in the surrounding area can be useful
for diagnosing interference as well as policing unauthorized use of protected
frequency bands.
Ideally, a wideband frequency analyzer could be used to detect and
identify the center frequencies of the received signals. In traditional electronic
systems, a swept spectrum analyzer can be used for determining the frequency
of the incoming signals [2]. Unfortunately, this electronic solution suffers
from two main deficiencies: slow scanning speeds and limited bandwidths.
As seen in Figure 6.2, the spectrum analyzer operates by using a tunable local
oscillator connected to a mixer. The signals that pass through the mixer are
downconverted to a frequency at the center of a bandpass filter. The filtered
signal is then detected and displayed. The displayed frequency matches the
center frequency of the tunable local oscillator. While the spectrum analyzer
is fairly simple to implement, the use of a tunable local oscillator combined
with a mixer for downconversion limits the frequency range of the spectrum
analyzer. In addition, the slow tuning speed of the local oscillator can al-
low for signals to be missed. Photonic methods for determining the center
 Signal Identification Utilizing RF Photonics 125

Figure 6.2 Block diagram of a scanning electrical spectrum analyzer (LPF: low-pass filter;
LO: local oscillator; VCO: voltage-controlled oscillator; IF Amp: intermediate frequency am-
plifier; BPF: bandpass filter; and Det: envelope detector. (After: [3].)

frequency of the various signals coming into the system can overcome the
electronic deficiencies.

6.2 Using RF Photonics for Spectrum Analysis

A photonic-based spectrum analyzer seeks to replace a traditional electronic
spectrum analyzer by providing increased frequency range. One of the most
straightforward implementations of a photonics-based spectrum analyzer is
to replace the electronic components in the architecture shown in Figure 6.2
with photonic components [4]. As seen in Figure 6.3, the incoming RF sig-
nals are upconverted into the optical domain by use of an optical modulator,
which takes the place of the mixer in the electronic implementation. The
optical signal is then passed through a scanning Fabry-Perot (FP) filter, with
the filter now scanning rather than the local oscillator in the electronic ver-

Figure 6.3 Block diagram of a scanning electrical spectrum analyzer (MZM: Mach-
Zehnder modulator; Amp: optical amplifier; TFPF: tunable FP filter; and Det: optical photo-
detector). (After: [3].)
126 Applications of Modern RF Photonics

sion. The resulting output is then sent to a photodetector. As the FP filter

is scanned, the optical sidebands due to the modulated RF signals will meet
the resonance condition of the FP and pass through the filter with maximum
transmission. If the scanning filter is synchronized to the output of the pho-
todetector, a display of power as a function of RF frequency can be obtained.
One of the advantages of the photonic architecture is that it does not require
a downconverter, as the photonic components have more than enough band-
width for the RF frequencies [4]. This makes the system scalable to higher
frequencies without the need for additional electronics. However the system
still has the disadvantage of being a scanning system and not instantaneously
capturing wideband spectrum.
To overcome the disadvantage of scanning filter systems, another type
of photonic spectrum analyzer has been investigated. Originally intended for
use as a form of photonic-based memory, rare Earth-doped crystals can also
be used as a spectrum analyzer [5]. Through a process known as spectral hole
burning (SHB), the absorption of the crystal at a wavelength that is resonant
with the atomic transitions can be modified. In one of the first demonstra-
tions of this effect, an experimental setup was developed to show the feasibil-
ity of the spectrum analyzer. As seen in Figure 6.4, two laser beams arriving at
different angles to the surface are used to write an absorption grating on the
crystal. A third laser beam that contains the RF information is then input to
the crystal from the opposite side of the first two beams. The absorption grat-
ing set up in the crystal by the first two beams then deflects the beam with the
RF information onto a photodiode array. By precisely maintaining the power
of the two optical beams that form the absorption grating, the deflection of
the beam with the RF signals will be precisely mapped to the photodiode
array. The results showed that the spectrum analyzer could resolve two differ-
ent frequencies at a minimum resolution of 1.5 MHz [5]. While the original
demonstration was limited by the photonic components that were used, a
follow-on demonstration was able to improve the resolution to 500 kHz [6].

Figure 6.4 Spectrum analyzer based on a SHB crystal. Different RF frequencies are de-
flected onto different parts of the detector array. (After: [3].)
 Signal Identification Utilizing RF Photonics 127

One of the advantages of this type of spectrum analyzer is the removal of a

scanning filter. Since the bandwidth of the crystal is on the order of 100 GHz,
the system can instantaneously identify the center frequencies of signals over
that bandwidth, assuming that the RF signals can be modulated onto the
optical beam before entering the crystal. Such a wide bandwidth operation
would be very useful to radio astronomers [6]. Unfortunately the system has
some disadvantages. The lifetime of excitation of the ions in the crystal is
quite short and requires the crystal to be cooled on the order of 5° Kelvin in
order to provide the minimum resolution. Also, the crystal has to have the
grating constantly updated by the two optical beams. Finally the system oper-
ates at the resonance of the ions in the crystal, which in this case is 793 nm.
Note that this does not match with the low-loss window of the optical fibers
that are being used in the rest of the demonstration architecture. One possible
solution is to use an optical frequency doubler in order to get the wavelength
of the laser to match the crystal resonance. This does increase the complexity
of the system, however. Nevertheless, these systems do show some promise to
be able to perform spectrum analysis over a large instantaneous bandwidth.

6.3 Using Photonics Filters for Instantaneous Frequency

Measurement
Other methods for determining the center frequency of RF signals have also
been investigated. Most of them have focused on removing the scanning fil-
ter limitation and providing wide instantaneous bandwidth operation. One
method for determining the center frequency involves the use of fixed opti-
cal filters. Unlike the scanning FP filter described in the previous section,
the fixed optical filter has a sinusoidal response that can be combined with
two lasers in order to determine the frequency of the RF signal [7]. The
wavelength of one laser is set at the null of the sinusoidal response while the
other laser is set at the peak of the filter response. This allows the generated
sidebands to appear on complementary slopes of the filter response, as seen
in Figure 6.5. The optical power of each pair of sidebands is separated and
detected. A ratio of the powers from the two photodetectors, referred to as
the amplitude comparison function (ACF), can then be used to determine
the frequency of the RF signal. The transfer function of the filter for the two
different laser wavelengths can be written as

  ω 
H laser 1( ω) = 1 + cos  RF   (6.1a)
  FSR  
128 Applications of Modern RF Photonics

Figure 6.5 Two-laser system using an optical filter to identify RF center frequency, with
the frequency response of the filter plotted above (PM MUX: polarization-maintaining
wavelength multiplexer; DEMUX: wavelength demultiplexer; PD: photodiode; and Elec.
Proc.: electronic processor. (After: [3].)

  ω 
H laser 2 ( ω) = 1 − cos  RF   (6.1b)
  FSR  

where ΩRF is the RF angular frequency and the FSR is the free spectral range
of the filter. The ACF can be expressed as the ratio of the two transfer func-
tions, which can then be described as
2
  ωRF  
ACF = tan   (6.2)
  2 ∗ FSR  

A plot of the ratio in Figure 6.6 shows that the ACF monotonically
increases as a function of frequency. Thus, the frequency can be recovered
by knowing the ACF response. An alternative approach, as shown in Figure
6.7, exchanges the two lasers for two filters [8]. The resulting response is
equivalent to the two-laser case, with the advantage that the cost and power
consumption of two filters is less than that of two lasers. The setups provide
instantaneous measurement of the RF frequency with a good match to the
theoretical performance. The disadvantage of these systems stems from their
inability to measure more than one CW RF signal at a time. Also, these sys-
tems require the suppression of the optical carrier. If the optical carrier is not
sufficiently reduced, it introduces a noise floor that limits the sensitivity of
the overall system. Finally, the overall bandwidth of the system is limited to
fmax = FSR/2 of the sinusoidal filter. Beyond this frequency, the system cannot
disambiguate the frequency between the lower and upper half of the FSR of
the filter. Despite these limitations, a similar system has been shown to mea-
sure the frequency of both a CW and pulsed RF signal at the same time [9].
Work continues to improve the performance of these systems.
 Signal Identification Utilizing RF Photonics 129

Figure 6.6 The ACF as a function of RF frequency for a two-laser system using an optical
filter. (After: [3].)

Figure 6.7 Two complementary filter systems for identifying the RF center frequency, with
the frequency response of the two filters overlaid on each other (CW: CW laser source; PD:
photodiode; and Elec. Proc.: electronic processor). (After: [3].)

6.4 Using Dispersion for Instantaneous Frequency Measurement

While using optical filters is one way to create the ACF response, another
method is to use dispersion. As described in Chapter 2, dispersion in optical
fibers leads to RF fading as a function of RF frequency when dual-sideband
modulation is used. While often an unwanted response, the power fading can
be seen as a narrow filter function. Making use of this function can provide an
ACF similar to the one in the previous section. Since the dispersion is like a
single filter that is wavelength-dependent, an initial demonstration was made
with two lasers, as presented in Figure 6.8 [10]. In this case, the ACF will
present a slope as shown in Figure 6.9 and described by
130 Applications of Modern RF Photonics

Figure 6.8 Two-laser system utilizing dispersion for identifying the RF center frequency
(PM MUX: polarization maintaining wavelength multiplexer; DEMUX: wavelength demulti-
plexer; CW: CW laser source; PD: photodiode; and Elec. Proc.: electronic processor). (After:
[3].)

Figure 6.9 Power as a function of RF frequency for the two different wavelengths (dashed
lines) after passing through the dispersive medium. The ratio of the responses is shown in
the solid line. (After: [3].)

ACF =
(
cos 2 πDL λ12 f c)
2

(6.3)
cos 2 ( πDL λ f c )
2
2
2

where D is the dispersion (ps/(nm•km)), L is the length of the dispersive me-

dium, c is the speed of light, f is the RF frequency, and λ1,2 is the wavelength
of the two lasers, respectively.
The two dashed lines are the RF response for each of the two lasers
individually. By taking the ratio of the responses in the electrical processing
unit, the solid black line ACF can be generated. The advantage of this system
is the high slope of the ACF, which allows for a small change in frequency to
result in a large change in amplitude. The disadvantage is that the system still
 Signal Identification Utilizing RF Photonics 131

cannot recover the frequency of more than one signal at a time. In addition,
the demonstration was limited to a frequency range of 7–10.5 GHz for the
chosen dispersion of each laser.
As an improvement, another version of the dispersion-based system was
developed to be able to measure simultaneous signals [11]. Figure 6.10 shows
an example system with a single laser followed by a dispersive element and
a photodetector. In this case, the dispersion is actually used to perform a
frequency-to-time mapping of the RF-generated dual optical sidebands. The
system takes advantage of the fact that the time delay through a dispersive
medium can have a linear response as a function of wavelength. Using this
capability, each optical sideband will arrive at the photodetector at a different
time, as seen in Figure 6.10.The difference in time between the arrival of the
first sideband and its corresponding twin is directly proportional to twice the
RF frequency. Therefore, measuring the time delay and knowing the slope of
the time delay through the dispersive medium will yield the corresponding
RF frequencies. In the demonstration, two RF signals of 20 and 40 GHz were
detected and the frequency recovered through the frequency-to-time mapping
process [11]. However, the demonstration was limited to frequency separa-
tions of 12.5 GHz. The advantage of this system is that multiple RF signals
can be modulated onto the optical carrier of the laser, and the frequencies can

Figure 6.10 Single-laser system using dispersion to map frequency to time. Taking the
received optical power at the photodiode and measuring the time difference between
changes will yield the frequency of the RF signals. (After: [3].)
132 Applications of Modern RF Photonics

be detected. The system is limited by the resolution of the timing system used
to differentiate between signals, as well as any deviation from linear that the
dispersion slope may have. Regardless, this time-based solution offers capa-
bilities beyond some of the other systems previously described.

6.5 Combinations of Different Methods for Frequency

Measurement
A combination of the scanning local oscillator and dispersion has also been
investigated for IFM. These systems can identify the center frequencies of
multiple RF signals [12, 13]. Again, making use of the power fading as a
function of frequency in a dispersive medium, a tunable laser can be used to
sweep through the dispersive medium. The result can be thought of as a scan-
ning filter as seen in Figure 6.11. By tracking the tuning laser wavelength, the
power measurement can be translated into a frequency and power measure-
ment of the RF signals using a digital processor. In the first demonstration
[12], the ability to successfully resolve two different frequencies at least 15
MHz apart was shown. The measured power of the RF signal was also shown
to be within 0.1 dB of the actual power. Still, the maximum measured fre-
quency was limited to 18 GHz due to photonic components used. Another

Figure 6.11 Using a tunable laser along with a fixed laser to move the RF power notch as
a function of frequency. (After: [3].)
 Signal Identification Utilizing RF Photonics 133

demonstration was shown to expand the frequency operation to 24 GHz

while also demonstrating a sensitivity of –48 dBm [13].

6.6 Using FIR and IIR Filters for Instantaneous Frequency

Measurement
The previous demonstrations have been interested in using optical filters,
dispersion, or some combination of the two in order to recover the center
frequency of the RF signals. Another method to determine the frequency is
to use either FIR or IIR filters. Looking back at Chapter 5, FIR and IIR filters
can be realized by using photonic means. In one demonstration, a combina-
tion of FIR and IIR filters can be used to identify the center frequency of an
RF signal [14]. The system is presented in Figure 6.12, where the FIR filter
is generated using photonic means, and the IIR filter is an electronic imple-
mentation. Note that the gain in the overall loop is set below threshold for
oscillation to prevent the cavity from becoming an RF source. The combina-
tion of the two filter responses yields a transfer response of

Figure 6.12 Setup combining an FIR and IIR filter to identify the RF signal, along with the
power as a function of RF frequency for the setup. (After: [3].)
134 Applications of Modern RF Photonics

H ( ω) =
(
L feedback 1 + e j ωRFT )
(6.4)
e j ωRFT − G 2

where ΩRF is the RF angular frequency, Lfeedback is the loss in the opto-elec-
tronic feedback loop, G is the gain in the loop, and T is the time delay in the
loop. Note that the time delay of the IIR feedback is set to match the time
delay of the two arms of the FIR filter [14]. For a given set of parameters the
detected power increases as a function of frequency, which is shown in Figure
6.12. The response is similar to demonstrations using dispersive methods.
The FSR, as well as the Q factor, of the overall filter response is limited by
the long length of lossy RF cable used in the system. In order to remove the
limitation of the IIR filter, a photonic implementation was used [15]. As
discussed in Chapter 5, the photonic IIR filter implementation requires a
broadband light source in order to remove any coherence effects. The long
RF cable was replaced with an optical fiber loop with an optical amplifier.
While the loss of the optical fiber is much lower than the RF cable, the length
was not much shorter. Therefore, the FSR was still shown around 42 MHz.
Again, the gain in the optical loop was set below threshold to make sure there
would be no oscillatory behavior. Still, the input power of the RF signals is
limited to a range between –7 and –16 dBm, severely reducing the dynamic
range of the measurement system.

6.7 Frequency Measurement with Multimode Photonic Systems

As mentioned previously, the use of a fiber loop can provide advantages in
terms of the filter response with a much smaller form factor. The dispersion
demonstrations require long lengths of fiber or specialty devices to give the
proper response. The fiber loop can provide a similar response with a much
smaller length of fiber. Yet, the previous demonstrations were limited in the
sensitivity to the RF signal power at the input of the system. In order to ad-
dress these deficiencies, a different photonic system can be used. Specifically,
a multimode OEO can provide enhanced sensitivity as well as selectivity.
As seen in previous chapters, the OEO can be used as a low-noise oscil-
lator that can generate RF signals. However, the OEO does have applications
in analog systems beyond the ones described previously. One such applica-
tion is the identification of low-power RF signals in a cluttered environment.
Imagine a wide frequency spectrum containing multiple RF signals, all with
varying powers as shown in Figure 6.13. In order to identify a subset of the
RF signals from the others, signals of interest need to be amplified while
 Signal Identification Utilizing RF Photonics 135

Figure 6.13 RF signals covering 1–10 GHz of bandwidth from an antenna before and after
selective amplification.

suppressing the others below some chosen threshold level. A narrowband

electronic filter combined with an RF amplifier could accomplish this task.
Such a system is hard to realize at higher frequencies, is not widely tunable,
and has limited sensitivity to low-power signals. An OEO however can han-
dle this task. Specifically, the injection locking process in an OEO can be
used to amplify certain RF signals while suppressing the other ones. Injection
locking on an OEO occurs when an external RF signal is injected into the
OEO cavity, as seen in previous demonstrations [16]. The frequency of the
signal generated by the OEO will become locked to the frequency of the in-
jected RF signal. In a traditional OEO, the range of frequencies to which the
OEO can be locked is limited by the electronic bandpass filter used inside the
OEO cavity. For the case of signal identification and amplification, the elec-
tronic bandpass filter is removed from the OEO cavity, and the power inside
the OEO is set just below threshold. The setup is shown in Figure 6.14. Such
an OEO is referred to as a multimode OEO (MM-OEO). When an external
signal is injected into the OEO cavity, two results can occur. If the injected
signal’s frequency matches one of the cavity modes of the MM-OEO, it will
cause the OEO cavity to oscillate, and the signal will see gain. If the injected
signal’s frequency does not match one of the cavity modes, the signal will see

Figure 6.14 Setup of a MM-OEO. (After: [3].)

136 Applications of Modern RF Photonics

loss and be suppressed. This process is depicted in the movie at the website
found on page iv of this book, where an injected signal is swept from one
mode to an adjacent one, and the resulting gain and loss is observed. The first
demonstration of an MM-OEO used an electronic RF amplifier to show the
loss and gain profile over multiple cavity modes [17]. Because the RF amplifi-
er adds noise to the system, the work was extended to an MM-OEO that used
all-photonic gain [18]. The all-photonic gain uses the method of a low-biased
MZM followed by an optical amplifier, as discussed in the previous chapters.
Due to the long length of fiber in the EDFA used for the all-photonic gain,
the MM-OEO cavity modes are spaced about 5.2 MHz apart, with each
cavity mode having a gain bandwidth of 1 MHz. The MM-OEO has been
shown to provide gain for RF signals at frequencies as high as 6 GHz. In
order to characterize the MM-OEO further, the sensitivity and compression
dynamic range are measured. The sensitivity of the MM-OEO is as low as -83
dBm, while the maximum input power into the MM-OEO before the gain
becomes compressed is –11 dBm. This yields a compression dynamic range
of 72 dB. In order to select a small subset of cavity modes from the MM-
OEO, the optical output of the MM-OEO was passed through an integrated
silicon FP filter (FPF). The FPF has a 3-dB bandwidth of 4.5 GHz, and after
passing the optical signal from the MM-OEO through the FPF, the filtered
RF gain is only positive at frequencies of 3 GHz or lower, as opposed to 6
GHz without the FPF. Thus, an optical filter can be used to identify the RF
frequency of the MM-OEO. The advantage of the integrated silicon FPF is
the ability to fabricate up to 100 filters on a single chip. This scalability in a
small size can be advantageous in systems wanting to identify RF signals of
multiple gigahertz in bandwidth.
While the MM-OEO described above shows promise for use in selec-
tive amplification of RF signals, it is desirable to increase the mode spacing of
the MM-OEO cavity, which requires shortening the length of the cavity. As
mentioned previously, the EDFA adds an equivalent of approximately 10m
of fiber into the cavity. In order to overcome this limitation while still using
all-photonic gain in the MM-OEO cavity, the EDFA was replaced with a
semiconductor optical amplifier (SOA). This device uses a semiconductor
gain medium that can be on the order of a centimeter in length, plus the
fiber pigtails in order to connect it to the other fiber components in the
MM-OEO. Such a length savings can significantly improve the cavity mode
spacing. In this new configuration, an MM-OEO was demonstrated having
a mode spacing of 100 MHz, which is a 20-times improvement over the
previous demonstration [19]. While the mode gain bandwidth increases also
from 1 to 11 MHz, the overall selectivity of this MM-OEO is better than the
previous demonstration. Another advantage of increasing the mode spacing
 Signal Identification Utilizing RF Photonics 137

is that it allows the MM-OEO to be used with two lasers. By proper selec-
tion of the DC bias of the MZM, the two lasers will allow the MM-OEO to
switch between complementary modes in the MM-OEO cavity. The modes
are complementary in that in one condition, a given frequency will see gain
(loss), while in the other condition, the same frequency will now see loss
(gain). As seen in Figure 6.15, with one laser on, one set of modes have gain
in the MM-OEO. However when both lasers are on, the complementary
set of cavity modes now see gain. This allows the MM-OEO to selectively
amplify more frequencies with the same cavity than is possible with just one
laser. Another improvement in the system comes from the sensitivity of this
MM-OEO. With the increased mode spacing, the integrated noise of the
MM-OEO can be measured over 140 MHz while only incorporating one
cavity mode. With a measured gain of 10 dB, input sensitivity is calculated
to be –78.4 dBm. When compared to thermal noise over the same 140-MHz
bandwidth (–92.6 dBm), the MM-OEO is only 14.2 dB above the thermal
limit. Thus the MM-OEO can be used to amplify very low-power RF signals
for detection.
The MM-OEO has been shown to offer a method for selectively ampli-
fying low-power RF signals. The system can have very good input sensitivity,
high CDR, and large selectivity. Further improvements of the MM-OEO
can be made, the first being to further shorten the cavity in order to get
the mode spacing to 1 GHz, while increasing the overall bandwidth to 10
GHz and higher. Such improvements can make the MM-OEO attractive

Figure 6.15 The two-laser MM-OEO showing selective amplification of complementary

modes when either one or both lasers are on. (After: [20].)
138 Applications of Modern RF Photonics

to applications such as identification of multiple RF signals in a cluttered

environment.

6.8 RF Frequency Identification Using Optical Injection Locking

While the MM-OEO has shown promise in terms of being able to select out
and identify signals over a wide range of RF frequencies it still is limited to a
fairly closely spaced set of modes. In addition, the suppression of unwanted
signals was only 25 dB compared to the signal of interest. Finally the tuning
of the modes of the cavity is achieved by changing the length of the OEO cav-
ity. While using a fiber stretcher can tune the modes of the OEO cavity, the
speed of tuning is limited. Another method for amplifying signals involves
the use of optical injection locking of two separate lasers [21]. Resembling the
optical injection locking process for making a fast tunable oscillator, the two-
laser system can be used to discriminate and identify RF signals. The system
works by amplifying the selected RF signal while completely suppressing all
other RF signals, regardless of their power. By injecting the output of a master
laser that has been phase-modulated with RF signals into a slave laser, selected
RF signals will see gain while completely suppressing other unwanted signals.
The system can be frequency-tuned very quickly with very high extinction ra-
tio between gain and loss. The system is different than other injection locking
methods [22] in a couple of different ways: First, the slave laser is multimode,
and second, the slave laser is kept below threshold in order to give gain to the
appropriate sideband. Also, this system is designed for fast-tunable, RF signal
discrimination, which is different from other demonstrations. Finally, most
demonstrations use a MZM instead of a phase modulator.
The system for RF signal discrimination is illustrated in Figure 6.16.
The optical output of the master laser connects to the optical input of a phase
modulator. The RF input of the phase modulator receives the signals, which
are illustrated to demonstrate the different operation. The optical output of
the phase modulator is injected into the optical output of the multimode
slave laser via an optical circulator (port 1-2). The mixed signal generated
inside the slave laser is output to an optical filter in order to select one of
the modes from the slave laser via the third port of the optical circulator.
The filtered optical signal is connected to the input of a photodetector. The
resulting RF signal from the photodetector is measured. The drive current
of the multimode slave laser is set below threshold in order to provide gain
to the incoming RF-modulated signal from the master laser. In one case, the
wavelength of the master laser is set to the wavelength of one of the modes of
the slave laser minus the RF frequency of interest. Under this condition, the
 Signal Identification Utilizing RF Photonics 139

Figure 6.16 Optical injection locking method for selectively amplifying RF signals along
with identifying them. (After: [21].)

wavelength of the upper sideband generated by the RF signal on the master

laser will match the wavelength of one of the modes of the slave laser, while
the lower sideband does not match a nonoscillating mode of the slave laser.
In this case, the upper sideband will see gain while the lower sideband will see
loss. After the photodetector, the resulting RF signal will see gain as compared
to the RF signal at the phase modulator input. Note that the system can be set
to amplify the lower sideband with the same result. In the other case, when
the RF signal sidebands do not match the wavelength of any of the modes
of the slave laser, there will be no RF signal present at the output of the pho-
todetector. This is inherent to the phase modulation process as long as the
sidebands’ power levels are equal. Thus, the system will completely suppress
the RF signal if it does not match the appropriate conditions.
When the optical sideband is set to the wavelength of one of the modes
of the slave laser, the resulting sideband is higher than the other sideband,
which sees optical loss. This allows the otherwise balanced signal to be de-
tected at the photodetector. A measurement has been made with an RF sig-
nal at 14.18 GHz input to the phase modulator. The input power is –40
dBm, and the measured output power is –39 dBm, providing a gain of 1
dB. When the RF signal frequency is moved off of the slave laser’s operating
wavelength, there is no observable signal. In the optical spectrum, the side-
mode is amplified to the same level as the carrier. When the signal is moved
off the operating point, the sideband is much lower than the optical carrier.
At the point that the sideband is not amplified, there is no observable signal
at the output of the photodetector. The system will also detect RF signals
that are modulated by other data. For example, the RF signal at 14.18 GHz
has been modulated with a 500-kHz pulse. The detected RF signal is ampli-
fied along with the 500-kHz sidebands, allowing the data to be recovered.
140 Applications of Modern RF Photonics

The additional sidebands are higher frequency components due to the pulse
shape. Note that any modulation can be amplified as long as it is within the
gain bandwidth of the system. By changing the bias current of the slave laser,
the wavelength changes, consequently changing the RF frequency that will be
amplified. In this case, the current can be changed on the order of ~100 nsec.
The RF frequency to be amplified can be tuned over a frequency range of >
1 GHz in that time frame.
A mathematical description of the phase modulation is shown in (6.5).
The phase modulator has a transfer function given by

E out (t ) = αPM Gopt e i j(t )E in (t ) (6.5)

where αPM is the optical insertion loss of the phase modulator, Gopt is opti-
cal gain/loss before the photodetector, ϕ(t) = ϕrf sin(Ωrf t) and ϕrf = π(Vrf /
∞
Vπ,rf(Ωrf)). Now using the Jacobi Anger expansion e iz sin θ = ∑ n =−∞ J n (z )e in θ
the following can be written:

E carrier (t ) = αPM Gopt E in e i ωot  J 0 ( jrf ) ,

i ωot −i Ωrf t
E usb , fund (t ) = αPM Gopt E in e  − J 1( jrf ) , (6.6)
 
i ωot +i Ωrf t
E lsb , fund (t ) = αPM Gopt E in e  J 1( jrf ) ,
 

where Ωo is the optical frequency of the laser and Jn is the Bessel function of
the nth order. At the photodetector, the carrier field beats with both the up-
per and lower sidebands in order to generate an optical photocurrent. In the
case of phase modulation, the resulting mixing terms are given below:

iΩ t
E carrier × E ∗usb , fund = αPM PlaserGopt  J 0 ( jrf )  − J 1( jrf ) e rf ,
−i Ω t
E ∗carrier × E usb , fund = αPM Plaser Gopt  J 0 ( jrf )  − J 1( jrf ) e rf ,
−i Ω t
(6.7)
E carrier × E ∗lsb , fund = αPM Plaser Gopt  J 0 ( jrf )  J 1( jrf ) e rf ,
iΩ t
E ∗carrier × E lsb , fund = αPM PlaserGopt  J 0 ( jrf )  J 1( jrf ) e rf ,

where E in E in∗ = Plaser . By inspection, when the four terms are added together
they cancel and no RF photocurrent will be generated. So as long as the two
sidebands are equal in power, the RF signal will be completely suppressed at
the output of the photodetector. However, when the phase-modulated opti-
 Signal Identification Utilizing RF Photonics 141

cal signal is injection-locked into the slave laser, one sideband can be made
higher than the other sideband. For this case we make the upper sideband
larger than the lower sideband. We can represent that by making Gopt dif-
ferent for the upper sideband than for the lower sideband. Including this in
(6.7), we now determine that the terms do not add to zero but give the fol-
lowing result for the photocurrent:

G
I PD = R αPM Plaser  opt ,usb G
 opt ,lsb

(
− 1 J 0 ( jrf ) J 1( jrf )cos Ωrf t ,

) (6.8)

where there is a different Gopt for the lower and upper sideband and R is the
responsivity of the photodetector. Since we are interested in low-power RF
signals, we can take the small signal approximation for the Bessel functions
2
and find the RMS output RF power given by Prf = 1 2I PD Z out , with Zout is
the output impedance of 50Ω. The small signal RF power is then given by
the following:

2 2
1 2 G opt ,usb  2 Vrf
Prf ,ss ,out = (R αPM Plaser )  Gopt ,lsb − 1 π V 2 jrf Z out .
(6.9)
4  π,rf

Now the RF gain is simply the ratio of output RF power to the input RF
power, which is given by Prf,in = Vrf /(2Zin) with Zin as the input impedance.
The RF gain is given by the following in both linear and decibel units.

2
Gopt ,usb  2
(R αPM Plaser )
2
 Gopt ,lsb − 1 π Z out Z in
Grf = ,(linear)
4V π2,rf
(6.10)
Grf = −22.1 + 20 log10 (ℜαPM Plaser ) − 20 log10 (V π,rf )
G
+20 log10 ( opt ,usb G − 1),(dB ).
opt ,lsb

The system can see positive gain when the ratio of sidebands is enough
to compensate for the Vphe of the modulator. This mathematical description
demonstrates the use of the system to either completely suppress an RF signal
or to provide gain to an RF signal.
This method provides an RF identifier with fast tuning speeds that can
provide gain to specific RF signals while completely suppressing other RF
signals. The fast tuning speed is useful for reconfigurable systems that need
142 Applications of Modern RF Photonics

to change the RF discriminator’s frequency very quickly. The use of injec-

tion-locked lasers improves the mode spacing when compared to MM-OEO
demonstrations. The ability to provide gain without the use of an electronic
amplifier increases the frequency range over which the system works when
compared to systems that require electronic gain. The system is also more
compact when compared to the MM-OEO demonstration.

6.9 Conclusions
A review of photonic methods for determining the center frequency of RF
signals entering the demonstration architecture has been performed. The first
method is a photonic-based spectrum analyzer. While a scanning photonic
analyzer has wider bandwidths and does not require a downconverter, it still
suffers from relatively slow sweeping speeds. A more advanced photonic spec-
trum analyzer, based on SHB in a rare Earth-doped crystal, can instanta-
neously capture the entire frequency band without sweeping. However, the
crystal requires constant updates from the writing laser and needs to be cooled
to 5K. Beyond spectrum analyzer-based methods, the use of optical filters has
been used to identify the frequency of an RF signal. While initial demonstra-
tions were limited to only one RF signal, more recent methods have shown
the capability to identify the frequency of multiple RF signals, including ones
that are continuous as well as pulsed. Other methods have used the frequen-
cy-dependent loss as a function of RF frequency in a dispersive medium to
identify the RF signal. Another subset uses a combination of photonic-based
FIR and IIR filtering responses to identify the frequency of the signal. Finally,
the use of multimode photonic systems can be used to identify the frequency,
while suppressing unwanted signals and offering the ability to be tuned very
quickly. Most, if not all of the methods, work by using a transfer function
to convert the frequency of the incoming RF signal into a change in optical
power. By making a unique transfer function, the signal’s frequencies can be
identified. The identification of the frequency of the incoming signal will
be important for Chapter 7, where signal-processing techniques using pho-
tonics are introduced. Many of these require some knowledge of the center
frequency in order to operate on the signals.

References
[1] Keysight Technologies, Spectrum Analysis Basics, Application Note 150, www.key-
sight.com.
 Signal Identification Utilizing RF Photonics 143

[2] http://www.setileague.org/articles/protectd.htm.
[3] Devgan, P., “Signal Identification and Isolation Utilizing Radio Frequency Photonics,”
AFRL Report, AFRL-RY-WP-TR-2017-0158, 2017.
[4] Winnall, S. T., and A. C. Lindsay, “A Fabry-Perot Scanning Receiver For Microwave
Signal Processing,” Transactions on Microwave Theory and Techniques, Vol. 47, No. 7,
1999, pp. 1385–1390.
[5] Ménager, L., et al., “Demonstration of a Radio-Frequency Spectrum Analyzer Based on
Spectral Hole Burning,” Optics Letters, Vol. 26, No. 16, 2001, pp. 1245–1247.
[6] Lavielle, V., et al., “Wideband Versatile Radio-Frequency Spectrum Analyzer,” Optics
Letters, Vol. 28, No. 6, 2003, pp. 384–386.
[7] Chi, H., X. Zou and J. Yao, “An Approach to the Measurement of Microwave Fre-
quency Based on Optical Power Monitoring,” Photonics Technology Letters, Vol. 20, No.
14, 2008, pp.1249–1251.
[8] Zou, X., H. Chi and J. Yao, “Microwave frequency Measurement Based on Optical
Power Monitoring Using a Complementary Optical Filter Pair,” Transactions on Micro-
wave Theory and Techniques, Vol. 57, No. 2,2009, pp. 505–511.
[9] Lu, B., et al., “Photonic Frequency Measurement and Signal Separation for Pulsed/CW
Microwave Signal,” Photonics Technology Letters, Vol. 25, No. 5, 2013, pp. 500–503.
[10] Nyugyen, L. and D. Hunter, “A Photonic Technique for Microwave Frequency
Measurement,” Photonics Technology Letters, Vol. 18, No. 10, 2006, pp. 1188–1190.
[11] Nyugyen L., “Microwave Photonic Technique for Frequency Measurement of
Simultaneous Signals,” Photonics Technology Letters, Vol. 21, No. 10, 2009, pp. 642–
644.
[12] Vidal, B., T. Mengual, and J. Marti, “Photonic Technique for the Measurement of
Frequency and Power of Multiple Microwave Signals,” Transactions on Microwave
Theory and Techniques, Vol. 58, No. 11, 2010, pp. 3103–3108.
[13] Wang, Y., et al., “Photonic Approach for Microwave Spectral Analysis Based on Fourier
Cosine Transform,” Optics Letters, Vol. 36, No. 19, 2011, pp. 3897–3899.
[14] Zhou, J., et al., “Instantaneous Microwave Frequency Measurement Using a Photonic
Microwave Filter with an Infinite Impulse Response,” Photonics Technology Letters, Vol.
22, No. 10, 2010, pp. 682–684.
[15] Niu, J., et al., “Instantaneous Microwave Frequency Measurement Based on Amplified
Fiber-Optic Recirculating Delay Loop and Broadband Incoherent Light Source,” Journal
of Lightwave Technology, Vol. 29, No. 1, 2011, pp. 78–84.
[16] Lasri, J., et al., “Ultra-Low Timing Jitter 40Gb/s Clock Recovery Using a Self-Starting
Optoelectronic Oscillator,” Photonics Technology Letters, Vol.16, No. 1, 2004, pp. 263–
265.
[17] Urick, V. J., et al., “Channelisation of Radio-Frequency Signals Using Optoelectronic
Oscillator,” Electronics Letters, Vol. 45, No. 24, 2009, pp. 1242–1244.
144 Applications of Modern RF Photonics

[18] Devgan, P. S., et al., “Detecting Low-Power RF Signals Using a Multimode Optoelectronic
Oscillator and Integrated Optical Filter,” Photonics Technology Letters, Vol. 22, No. 3,
2010, pp. 152–154.
[19] Devgan, P. S., V. J. Urick, and K. J. Williams, “Detection of Low-Power RF Signals
Using a Two Laser Multimode Optoelectronic Oscillator,” Photonics Technology
Letters, Vol. 24, No. 10, 2012, pp. 857–859.
[20] Devgan, P., V. J. Urick, and K. J. Williams, “Multimode Optoelectronic Oscillator,”
United States Patent 8,824,901, 2014.
[21] Devgan P. and N. Usechak, “Discrimination of Low Power RF Signals Using Phase
Modulation with Injection Locked Lasers,” United States Patent 9,435,839, 2016.
[22] Hong, C., et al., “Single-Sideband Modulation Based on an Injection-Locked DFB
Laser in Radio-Over-Fiber Systems,” Photonics Technology Letters, Vol.22, No. 7, 2010,
pp. 462–464.
7
Signal Processing Utilizing RF
Photonics
Up to this point, the demonstration architecture has been incorporating dif-
ferent subsystems linking the antenna to the signal processing block. These
subsystems are designed to identify and separate specific signals for further
processing. Often signal processing is done in the digital domain, where the
advantages provided by high-speed computing can be leveraged. However,
the limitations of the bandwidth of digital systems still leave a need for ana-
log signal-processing applications. Analog signal-processing systems are often
used to prepare signals for later digital analysis. Examples of analog signal
processing include downconversion, phase corrections, and even analog-
to-digital conversion. All of these applications can be accomplished by RF
photonic means. Along with the signal-processing block, the demonstration
architecture also contains a transmit block that is connected to the antenna.
Transmission often requires the generation of unique, wideband waveforms.
RF photonic methods for generating these types of waveforms can be useful
for the transmission block. In Figure 7.1, the processing and transmission
blocks of the demonstration architecture are expended to show a frequency
downconverter, an ADC, and an arbitrary waveform generator. This chapter
reviews downconversion of chosen RF signals. In addition, the chapter covers
analog-to-digital conversion and waveform generation parts.

145
146 Applications of Modern RF Photonics

Figure 7.1 Demonstration architecture processing and transmit blocks, expanded view.

7.1 Need for Downconversion

Most traditional RF systems require a downconversion stage before the signals
can be digitized. As mentioned previously, the SNR of the ADC is directly
related to the ENOB it can produce. For example, an ADC with an ENOB
of 8 has an SNR of 50 dB (see [4.1b]). The SNR of the ADC will increase
by 6 dB for every extra bit. However, as the input analog bandwidth of the
ADC increases, the ENOB decreases, due to various limitations that include
quantization error as well as timing jitter on the sampling clock. In order
to digitize the RF signals with the highest fidelity, a balance must be struck
between the ENOB and analog bandwidth of an ADC. For most wideband
applications, an ADC bandwidth of 500 MHz can operate with an ENOB of
around 10. Note that a bandwidth of 500 MHz requires a sampling rate of at
least 1 GHz in order to satisfy the Nyquist relation. Still, the center frequency
of the RF signals received by the wideband system can be much higher than
the ADC bandwidth. Thus, a downconverter is needed to shift the frequency
into the operational range of the ADC.
An RF downconverter consists of a mixer, along with amplification and
filtering. An example appears in Figure 7.2. The mixer is a three-port device
that takes in the RF signal at the input, mixes it with a local oscillator (LO)
 Signal Processing Utilizing RF Photonics 147

Figure 7.2 Block diagram of a downconverter (IFF: intermediate frequency filter; LO: local
oscillator; VCO: voltage-controlled oscillator; and IF Amp: intermediate frequency ampli-
fier). (After: [1].)

and produces an intermediate frequency (IF) at the output. The IF output

is determined by the frequency of the RF signal and the LO. For example,
an RF of 10 GHz will mix with an LO of 9.5 GHz to generate IF outputs at
19.5 GHz (sum frequency) and 0.5 GHz (difference frequency), along with
higher-order mixing terms. The IF is usually limited to match the analog
bandwidth of the ADC by using low-pass filters that only allow the IF of
interest to pass while suppressing the unwanted higher-order mixing terms.
The mixing process also introduces loss that is measured as the conversion ef-
ficiency. To compensate for a conversion efficiency less than 100%, amplifiers
are used to increase the power of the IF that appears at the output. Since the
downconverter uses a mixer and an amplifier, it has a noise figure and SFDR
that are often specified by the manufacturer. In addition, the phase noise
of the LO can be mixed with the incoming RF signal which will degrade
the generated IF. While the downconverter does not supply the LO, it does
specify the maximum phase noise allowed at the LO input before the IF per-
formance is degraded beyond the intrinsic noise of the downconverter. One
additional performance metric of the downconverter is the isolation. While
the downconverter would ideally allow neither the LO nor RF signal to leak
out the IF output, in practice either of the signals can appear. The isolation
is a measure of the suppression of the unwanted signals at the IF output. The
isolation can vary over a large range, with most having an isolation of 20–30
dB [2].
148 Applications of Modern RF Photonics

7.2 Using RF Photonics for Downconversion

A photonic-based downconverter seeks to replace a traditional electronic
downconverter by providing increased isolation and frequency range, while
improving the SFDR. One of the methods for using photonics for downcon-
version is replacing the various electronic components with photonic ones.
An example of a photonic downconverter appears in Figure 7.3. The gain and
filter stages can be easily replicated in the optical domain. The mixer is the
piece that requires more analysis. One of the most straightforward ways to re-
alize the mixing function is by using two cascaded optical modulators placed
in a series configuration [3]. The generated optical sidebands are shown in
Figure 7.3. The incoming RF signal has already been upconverted into the
optical domain by an optical modulator at the antenna. A second modulator
will use the LO as the input to the RF port. The resulting optical sidebands
are present at the output of the modulator. The optical filter will then select
out the pair of sidebands that match the desired IF. Then, the optical gain can
be used to increase the conversion efficiency of the downconverter. Finally,
the chosen LO and RF sidebands will be converted at the photodetector and
sent to the electronic ADC for digitization.
The photonic downconverter offers some unique advantages. The opti-
cal modulator at the antenna acts like an upconverter, moving the RF sig-
nals into the optical domain. The same thing happens at the modulator in
the downconverter for the LO. Since the LO and the RF are in the optical
domain, they can only be detected when they beat with the optical carrier.
Proper use of the optical filter will suppress the optical carrier and only al-
low the LO and RF optical sidebands to be presented to the photodetector.
Ideally, the output will only be the IF, which is the difference in frequency
between the LO and the RF. The result makes the isolation of the downcon-
verter essentially infinite, as the LO and RF frequencies do not appear at the
output of the downconverter. The other advantage of the optical filter is the
ability to suppress the image frequency. The LO can be set near one edge of
the filter, and the image frequency of the RF signal to be downconverted will
be outside the passband of the filter. This is also true for the image frequency
on the other side of the optical carrier. Another advantage of the photonic
downconverter is the wide bandwidth operation. The photonic modulator
can be used with an LO frequency as high as 100 GHz. The use of photonic-
based oscillators, as described in Chapter 4, can generate HF LOs with low
phase noise. Therefore, the photonic downconverter can operate at frequen-
cies much higher than most conventional electronic solutions.
The photonic downconverter does have some limitations. As discussed
in Chapter 2, the RF gain of a photonic link is directly related to the DC
 Signal Processing Utilizing RF Photonics

Figure 7.3 Block diagram of a serial cascaded modulator photonic RF downconverter (EOM: electro-optic modulator; OC: optical carrier; OA: optical
amplifier; and PD: photodiode). (After: [1].)
149
150 Applications of Modern RF Photonics

photocurrent squared. The DC photocurrent is directly proportional to the

optical power that is received by the photodetector. In the serial modulator
downconverter configuration, the optical carrier and all other sidebands are
filtered out, significantly reducing the total optical power at the photodetec-
tor. Therefore, the conversion efficiency of the downconverter will be lower.
While using an optical amplifier can help improve the conversion efficiency,
it will still be low. In addition the optical amplifier will add noise that will
also cause the RF noise figure to be higher. Finally, the optical modulator has
its own nonlinear transfer function, which can reduce the SFDR of the down
converted signals.

7.3 Advancements in RF Photonic Downconverters

In order to address some of the deficiencies of the RF photonic downconvert-
er, different architectures have been developed. In one example, the DC bias
of the modulator used for the LO has been optimized to cancel the distor-
tion generated from the modulator at the antenna [4]. The results show that
either the NF or the SFDR can be improved but not both at the same time.
In another configuration, the intensity modulators are replaced with phase
modulators [5]. The advantage of this configuration is the lower optical inser-
tion loss for each of the modulators as well as the lack of a bias control circuit
to hold the modulator at the appropriate operating point. The lower optical
losses result in a relative increase in the RF gain of 2.6 dB when compared
to an intensity-modulated photonic link. The link when set for the optimal
SFDR performance demonstrates an increase of 10.5 dB. This comes at a
reduction of the conversion efficiency of 13.6 dB.
Beyond using modulators in series-cascaded architecture, a parallel
modulator scheme can be employed for downconversion [6]. As seen in Fig-
ure 7.4, the laser output is split along two parallel paths. The upper path con-
tains the modulator at the antenna, while the lower path contains the modu-
lator that is connected to the LO. The two outputs of the modulators are then
filtered to get the LO and RF sidebands. The two arms are then combined
at a 50/50 coupler, and the two outputs are connected to a balanced pair of
photodiodes. This architecture allows for cancelation of common mode noise
from the laser, which can improve the RF NF. The separate paths also provide
more optical power at the photodiodes, which allows for a higher conversion
efficiency. In fact, a comparison of the parallel and series architectures for
downconversion summarizes the improvements of the parallel architecture
over the cascaded series configuration [7]. The parallel architecture allows
for suppressed carrier operation of the modulators, allowing for improved
 Signal Processing Utilizing RF Photonics

Figure 7.4 Block diagram of a parallel modulator photonic RF downconverter (EOM: electro-optic modulator; OC: optical carrier; OA: optical ampli-
151

fier; and PD: photodiode). (After: [1].)

152 Applications of Modern RF Photonics

performance. The measured results show that the parallel architecture offers
an improvement in the SFDR of 14 dB while offering a conversion efficiency
4 dB higher. The comparison does note that the drawbacks of using the par-
allel setup. The system looks like a large interferometer. It requires the two
paths to be matched on optical scales. Any variation in the two paths, due
to either temperature or vibration, will cause the system to become unstable
and result in added noise. An active control to compensate for changes in the
path lengths must be implemented. Thus, the added improvements come at
an added complexity as well.
In order to overcome the deficiencies of the parallel architecture while
maintaining the improved performance, the use of dual-parallel MZMs (DP-
MZMs) has been proposed [8]. The DPMZM has been described in previous
chapters and appears as an integrated form of the parallel-down converting
setup. The DPMZM also has a phase control in one of the parallel arms
of larger interferometer structure. The use of this structure has been shown
to have a conversion efficiency of 7.8 dB, which is a 23.7-dB improvement
over a series-cascaded downconverter operating at similar conditions. The
DPMZM’s susceptibility to environmental fluctuations is much lower than
a fiber-based approach, because in the former, the paths are very short and
confined to a single chip. The structure does have some trade-offs. First, the
structure does not have balanced outputs to cancel the common mode noise.
The demonstration does show a higher noise floor when compared to the
serial cascade setup. Second, the rejection of the LO is not infinite. While it
is 45 dB, it is not as high as in the other optical configurations. Finally, since
the parallel structure is integrated into a single modulator, the RF and the LO
have to be colocated.
In the demonstration architecture, the RF is received at the antenna
while the LO is generated at the processing block. The LO can be remoted to
the antenna site, but this leads to added complexity. To address the LO gen-
eration at the DPMZM, a demonstration utilizing an OEO has been made
[9]. The DPMZM is incorporated into an OEO feedback cavity. The RF sig-
nal is still connected to an antenna, but the LO port is connected to the OEO
feedback. A low phase noise LO can be generated right at the DPMZM,
bypassing the need to remote an LO to the antenna. Unfortunately, the OEO
cavity is not widely and continuously tunable, making it ideal for downcon-
verting fixed-frequency RF signals but not able to downconvert other fre-
quencies. Clearly, different setups exist for downconverting RF signals. The
ideal one often depends on the application and the required performance. In
most cases, the trade-off comes between the RF NF and SFDR. Optimizing
one will often sacrifice the performance of the other.
 Signal Processing Utilizing RF Photonics 153

7.4 RF Photonic Analog-to-Digital Conversion

While the RF photonic downconverter has great utility for the demonstra-
tion architecture it is still limited by the electronic ADC. For an ADC, there
are often three important noise factors that limit the SNR performance of
the digitizer. The first is the noise floor of the ADC, which is a combination
of the thermal noise limit for a given bandwidth and the added noise of the
ADC itself. The second is the aperture jitter, which is due to the noise in the
sample and hold circuit of the ADC. Finally there is the clock jitter, which
is due to the sampling clock. Previous work often assumed that the jitter
was white noise and uniform. However, it has been shown that this does
not hold true, as the phase-locked loop-based synthesizers used as the clock
source shape the noise by the loop transfer function [10]. Because of this, the
clock jitter is the dominant noise source for high sampling frequencies [11].
One of the reasons for this is that the aperture’s jitter-dependent SNR can be
improved by oversampling, whereas the SNR due to the sampling jitter can-
not. With this limitation in mind, recent advancements in ADC technology
have led to wider analog bandwidths being available. One of the methods
for increasing the bandwidth of the ADCs involves taking lower bandwidth
ADCs and parallelizing them in order to increase the overall bandwidth. The
bandwidth of the composite ADC improves by N times the bandwidth of an
individual ADC, where N is the number of ADCs used in the parallel struc-
ture. The use of the parallel ADC comes with some added complexities. The
total SNR performance of the parallel ADC is shown in (7.1) [12].

 
 
 
 1 
 − 1.78

) ( ) 
1  2 πf bw σ aj
2 
ENOBtotal = 
(  (7.1)
2
6.02   2 πf bw σ tj + 
 N 
 1  
+  
  N × SNRth  
 
 

where fbw is the 3-dB bandwidth of the individual ADCs, SNRth is the ther-
mal-limited SNR, stj is the timing jitter of the sampling clock, and saj is the
aperture jitter. The SNR improves as a function of N, while the aperture
jitter also improves as a function of N. However, the sampling jitter does
154 Applications of Modern RF Photonics

not improve as the number of ADCs increases, as the same clock is used for
all of the ADCs. Since the clock is the same, the timing jitter is correlated
between the various ADCs and does not scale with N. As seen in Figure 7.5,
the timing jitter will limit the performance of the parallel ADC structure as
the input bandwidth increases. Changing the phase of the incoming signal to
each of the parallel ADCs has been shown to improve the overall performance
by only 3 dB or 0.5 ENOB [12]. Another improvement would be to make
an independent clock for each of the ADCs in parallel. While this would al-
low for some decorrelation of the noise, it yields some other problems. For
example, clock skew between the various clocks will create strong peaks at
harmonic multiples of the sampling clock frequency [13]. This will limit the
overall usable bandwidth of the composite ADC.
Note that the above analysis examines SNR improvements without
considering the nonlinear dynamic range. By using a parallel structure, the
ideal system would provide an improvement of N to the dynamic range as
compared to a single ADC. This assumes that the nonlinear responses of
the ADCs are completely decorrelated. Previous work has shown that this
is not the case [14]. While predistorting the signal before each ADC and
then undoing the response has been shown to improve 2x2 spurs, the third-
order intermodulation shows no improvement. Parallel ADC structures show
promise for improving the overall bandwidth and ENOB while making use
of lower-sampling speed ADCs. Photonic techniques may be able to improve
some of the limitations of the electronic solutions.

Figure 7.5 ENOB as a function of frequency for a single (N = 1) ADC, for two (N = 1), and
four (N = 4) parallel-configured ADCs. The parallel ADCs improve the ENOB at low frequen-
cies but become limited at higher frequencies due to the timing jitter of the sampling clock.
(After: [1].)
 Signal Processing Utilizing RF Photonics 155

RF photonics brings two important features for ADCs: wide bandwidth

combined with low sampling jitter. In an in-depth review paper [15], four
classes of photonic ADCs are identified: photonic-assisted, photonic-quan-
tized, photonic-sampled, and photonic-sampled and -quantized. Each class is
described with examples [16–19]. The main difference between each class is
how RF photonics are utilized while covering the advantages and disadvan-
tages of the various implementations. For this discussion, the focus will be
placed on photonic sampled as well as photonic sampled and quantized. The
photonic-sampled case uses techniques in the photonic domain to sample the
incoming RF signal so that one or multiple electronic ADCs can digitize the
signal. This class is one of the most mature and has the highest performance
to date. It is also the most closely related to the parallel electronic structure.
On the other hand, photonic sampling and quantization is the most ambi-
tious as it seeks to bypass any electronic ADC and perform all functions be-
fore finally using photodetection to create the bits. It is also the least mature
and offers the most potential growth.

7.5 RF Photonics Sampling Combined with Electronic

Quantization
The use of photonics for the sampling function takes advantage of the wide
bandwidth operation as well as the low timing jitter. An RF photonic sam-
pling scheme is shown in Figure 7.6. The use of mode-locked lasers has been
able to generate a high repetition rate with very low pulse-to-pulse jitter
[17]. The optical pulses are then modulated by the RF signal to be sampled.
The modulated optical pulses can then be time-demultiplexed and sent to
multiple ADCs [20]. The configuration has the advantages of the electronic

Figure 7.6 Block diagram of a photonic ADC using optical sampling and electronic quanti-
zation with time-division multiplexing using an optical switch. (After: [1].)
156 Applications of Modern RF Photonics

parallel ADC while improving the sampling clock jitter. However, the setup
requires strict calibration to ensure that the time-demultiplexing is properly
maintained. The use of optical switches for the demultiplexing also adds extra
loss—and requires a time synchronization between the laser and the switch.
All of these complexities can degrade the ENOB performance of the pho-
tonic ADC [15]. An improvement in the system uses multiple wavelength
pulse sources, as seen in Figure 7.7. The use of a WDM reduces the losses and
simplifies the timing synchronization [21].
In either case, the laser plays a significant role in how well the photonic
ADC operates. Since the sampling function is done in the photonic domain,
the laser pulse width, timing jitter, and amplitude variations will play directly
into the overall performance. The ENOB as a function of pulse width and
sampling frequency is given by [22]

 
4
ENOB pw = log 2   (7.2)

( ) 
2
 π f samp τ pw 6

where fsamp is the sampling frequency and τpw is the sampling optical pulse
width. The timing jitter is given by the Walden relation [15, 23]

 1 
ENOBtj = log 2   (7.3)
 πf samp σ tj 6 

where σtj is the timing jitter. Finally the amplitude variations have to be less
than the quantization error for the least significant bit as given by [15]

Figure 7.7 Block diagram of a photonic ADC using optical sampling and electronic quanti-
zation with wavelength-division multiplexing using an optical demux. (After: [1].)
 Signal Processing Utilizing RF Photonics 157

 1 
ENOBamp = log 2   (7.4)
 σ amp 6 

where σamp is the amplitude ripple. All of these relations have to be consid-
ered to optimize the ENOB performance for a given sampling frequency.
Finally, the laser clock signal has to be recovered so that it can be used to
synchronize the clock of the electronic ADCs in order to realize the improved
performance.
Another method is to use a time stretch method. In this case the op-
tical source is one that can generate a chirped waveform as seen in Figure
7.8. Ideally the waveform has been modified such that each of the individual
wavelengths in the pulse has been continuously spread in time [24]. The
waveform can then be passed through an intensity modulator, which will im-
part the RF signal onto the amplitude of the waveform. Then a second stage
will time-stretch the waveform. By proper configuration of the second stage,
a wavelength demultiplexer can be used to create overlapping time blocks of
the RF signal. The separated signals can then be quantized by a set of paral-
lel electronic ADCs. The resulting digital data then has to be interleaved to
recover the original RF signal.
The time-stretch photonic ADC has considerations similar to those of
the optical pulse-sampled system. The optical paths have to be well matched
in both length and loss. However, it has been noted that the overlap in sig-
nals across the ADCs can help simplify the calibration needed to coordinate
the multiple paths [25]. In addition, the use of dispersion to both create the
chirped waveform as well as to time-stretch the RF-modulated optical signal
places limits on the bandwidth of the system [24]. As described in Chapter
3, the use of dual-sideband modulation places a limit on the frequency as
dispersion will result in RF fading. While single-sideband techniques have
been demonstrated to overcome these issues, they can limit the frequency
bandwidth of the system.

Figure 7.8 Block diagram of a time-stretch photonic ADC using optical sampling and elec-
tronic quantization with wavelength-division multiplexing using an optical demux. (After:
[2].)
158 Applications of Modern RF Photonics

7.6 Photonics Sampling and Quantization

Photonic ADCs that include photonic quantization. The quantization that
occurs in the previous discussion uses electronic solutions that are limited by
the comparators that are used in the electronic ADCs. Instead, Taylor [19]
proposed to use an array of modulators to perform the quantization function.
As seen in Figure 7.9, each of the modulators is identical except for the length
of the RF electrodes. By setting the electrodes of one of the modulators to half
the length of the other, a two-bit quantizer can be realized, as seen in the inset
of Figure 7.9. However, the system requires that not only the length of the
electrodes be half, but that the half-wave voltage, or Vp, be half as well [15,
19]. So if the first modulator has a Vp of 6V, the second one must have a volt-
age of 3V. This requirement has limited the demonstration of this configura-
tion to only 4 bits. Other demonstrations have been designed to improve the
performance of this initial system. Instead of a parallel array, serially cascaded
intensity [26] and phase modulators [27] have been demonstrated. Also, a
parallel array of modulators with equal-length electrodes has been proposed
[28]. The quantization is achieved by operating each modulator at a different
DC bias point in order to create an intensity-dependent response. Polariza-
tion has also been used to create a photonic quantizer [29].
Beyond the used of multiple modulators, other methods for photonic
quantization have been explored. Nonlinear optical responses can be exploit-
ed to perform a thresholding effect on the RF signal for quantization. The
thresholder will take the intensity of the optical signal and compare it to a set
quantization level. If the intensity is greater than the level, it will register as a
1. Otherwise it will register as a 0. While many examples of nonlinear optical
responses can be used, one method uses a saturable absorber to perform the
thresholding operation [30]. As seen in Figure 7.10, the saturable absorber
will block the transmission of light when the input optical power is below
a certain level. However, when the optical power increases, the saturable

Figure 7.9 Block diagram of a photonic ADC using optical quantization with two bits of
resolution. (After: [1].)
 Signal Processing Utilizing RF Photonics 159

Figure 7.10 Transmission response as a function of input power for a saturable absorber
(left) and a saturable absorber in an optical cavity (right). (After: [1].)

absorber will begin to allow the optical power to pass with lower loss. Unfor-
tunately, the slope of the response is too shallow for use as a thresholder. In
order to improve the slope, the saturable absorber can be placed in an optical
cavity to increase the slope. Ideally, the cavity will allow the incoming light
to pass through the cavity multiple times, causing the response to become
sharper, as seen in Figure 7.10. The increase in the response comes with an
increased overall transmission loss, limiting the optical power for detection
that will be used to derive the bits for the ADC. Other nonlinear optical
responses have been investigated, including four-wave mixing [31] and non-
linear Sagnac interferometers [32].

7.7 Arbitrary-Transmit Waveform Generation Using RF Photonics

The demonstration architecture incorporates a circulator in order to allow the
antenna to be used for both transmission and reception. To generate simple
RF sinusoidal signals, the oscillators described in Chapter 4 can be used as
very HF signal sources. They can either have very low phase noise or can be
tuned over very wide frequency ranges. In fact a signal source has been dem-
onstrated that can generate signals up to 100 GHz [33]. On the other hand,
some applications may require more complex types of waveforms. Examples
include communications, medical imaging, and test and measurement sys-
tems, along with predistortion of signals to compensate for the nonuniform
impulse response of the antenna itself [34]. All of these can be accomplished
with photonic-based arbitrary waveform generation. This section reviews a
few different methods for arbitrary waveform generation.
160 Applications of Modern RF Photonics

A review of arbitrary waveform generation using photonics highlights

some of the earlier works that utilized a spatial light modulator (SLM) in a
free-space configuration [35]. As seen in Figure 7.11, an optical source that
has a large optical bandwidth is input to the SLM. The SLM then acts as a
spectral filter or shaper. It modifies the amplitude and phase of the different
parts of the spectrum. The light is then passed down a medium with a linear
dispersion profile. The dispersion is used to convert the wavelength of the
optical spectrum into the time domain. Note that the use of dispersion is
similar to the time-stretch process described in the photonic ADC section,
Section 7.5. Finally, the optical signal is input to a photodetector, and the
resulting time-varying signal can be sent to the antenna. This system is a
combination of wavelength-to-space mapping, followed by wavelength-to-
time mapping techniques to generate the required waveform. The SLM can
be implemented using many different techniques [36]. The use of a liquid-
crystal array has been used most frequently, due to its abilities to be quickly
reconfigured and to control both the phase and amplitude of the light at each
input to the array.
This system has been shown to generate very wideband arbitrary wave-
forms. Unfortunately, the use of a free-space system can be difficult to stabi-
lize. An advancement of the previous demonstration is to replace the gratings
with a fiber-based arrayed-waveguide grating (AWG) to spread the optical
spectrum before the SLM [37]. The AWG can have the various channels of
the spectrum travel different lengths along with individual attenuators, allow-
ing for control of the phase and the amplitude as in the other configuration.

Figure 7.11 Photonic arbitrary waveform generation using a broadband source with a
spectral shaper followed by wavelength-to-time mapping using dispersion. (After: [1].)
 Signal Processing Utilizing RF Photonics 161

The advantages of this configuration are greater stability with a smaller foot-
print. On the other hand, the resolution of a typical AWG is around 12
GHz, while a free-space grating can have a resolution of a few gigahertz. This
resolution will limit the performance of the system in terms of the waveform
that is generated.
Beyond the use of an AWG or an SLM for the spectral shaper, other
methods have been suggested. In one case, a linearly chirped Fiber-Bragg
grating (FBG) can be used to perform the proper modification of both the
phase and amplitude [38]. In fact, the linearly chirped grating can be used
for both the spectral shaping and the wavelength-to-time mapping [39]. Un-
fortunately, in both of these cases, the grating cannot be easily changed, thus
making the system unable to be dynamically configured for new waveform
generation.
Another improvement on the previous demonstrations involves the op-
tical source used. The previous methods depend on either a supercontinuum
source or a mode-locked laser that generates very short optical pulses. In ei-
ther case, much of the power will be lost as it is removed by the SLM or
AWG. In order to generate a strong RF signal after detection, the optical
power will have to be quite high before entering the system. In addition, the
optical spectrum must ideally be flat in amplitude and have the phase of the
optical wavelengths be a fixed relationship. Another source that has been sug-
gested for use in waveform generation is an optical frequency comb source
[40]. The optical comb source provides some advantages. The comb lines can
be set to match the channel spacing of the grating or the AWG in order to
perform phase and amplitude control on each of the comb teeth individu-
ally. This will allow for no additional optical power to be lost. The resulting
signal will then be used at the photodetector to generate an RF signal with
the desired waveform.
Other methods for generating arbitrary waveforms exist beyond that
of the spectral filtering followed by wavelength-to-time transform. In some
cases, an optical pulse can be time-multiplexed to generate time-delayed ver-
sions of itself. When the pulses are detected at a photodetector and the elec-
trical signal is passed through a low-pass filter, the generated RF waveform
will be a smooth, continuously time-varying RF signal [41]. Another method
uses an optical modulator combined with higher-order dispersion effects to
generate higher frequency-chirped RF waveforms from a low-frequency RF
signal at the modulator input [42]. Finally photonic-based microwave filters
can be designed as matched filters in order to generate specialty waveforms
for communication applications [43].
162 Applications of Modern RF Photonics

7.8 Conclusions
This chapter expands the processing-block demonstration architecture to in-
clude three different systems: a photonics-based RF downconverter, an ADC,
and an arbitrary-waveform generator. Once a signal has been identified for
further investigation, the RF photonic downconverter can be tuned to the
signal frequency and mix the signal down for digitization by a low sampling
frequency electronic ADC. As the signals received by the demonstration ar-
chitecture can have very high center frequencies, a wideband photonic RF
downconverter will be well suited for this application. Multiple methods for
performing downconversion of RF signals are presented, each with their own
advantages and disadvantages. The choice of which one to use depends on the
overall application and need.
The photonic-based ADC seeks to overcome the limitations of the
traditional electronic-based ADC. As discussed previously, one of the main
limiters for wideband ADCs is the required sampling clock timing jitter.
Photonics has long been shown to provide sampling sources with timing jit-
ter on the order of tens of femtoseconds. When combined with arrays of
low-bandwidth quantizers, demonstrations have shown sampling of wide-
band signals with a high effective number of bits. These hybrid photonic-
electronic ADCs have become mature in their performance. On the other
hand, fully photonic-based sampling and quantization remains a field with
room to grow. Photonic quantization has been shown to be achievable, but
with only a limited number of bits. Work continues to explore advancements
in this area and may lead to future developments that make photonic ADCs
with much higher performance.
The photonic-based arbitrary-waveform generator can be used to cre-
ate unique waveforms for transmission from the demonstration architecture.
Multiple implementations have been shown for generating waveforms with
very wide bandwidth. Many of them use a spectral filter or shaper in order
to operate on individual wavelengths of a broad spectrum optical source.
By separating the individual components of the broader optical spectrum,
individually modifying the phase and amplitude of those components, and
then recombining the components, a new spectrum can be generated. Con-
verting the spectrum to a time-varying signal and then photodetecting the
optical signal generates the required RF signal. While some methods have
used fixed spectral filters, others have incorporated reconfigurable amplitude
and phase control, allowing the RF signal to be dynamically modified. All of
these subsystems allow the demonstration architecture to measure, record,
and transmit wideband signals.
 Signal Processing Utilizing RF Photonics 163

References
[1] Devgan, P., “Signal Processing Utilizing Radio Frequency Photonics,” AFRL Report,
AFRL-RY-WP-TR-2017-0172, 2017.
[2] Marki, F., and C. Marki, “Mixer Basics Primer,” Marki Microwave Inc., Application
Note, 2010.
[3] Gopalakrishnan, G., W. Burns, and C. Bulmer, “Microwave-Optical Mixing in LiN-
bO3 Modulators,” Transactions on Microwave Theory and Techniques, Vol. 41, No. 12,
1993, pp. 2383–2391.
[4] Karim, A., and J. Davenport, “High Dynamic Range Microwave Photonic Links for RF
Signal Transport and RF-IF Conversion,” Journal of Lightwave Technology, Vol. 26, No.
15, 2008, pp. 2718–2724.
[5] Pagan, V., B. Haas, and T. Murphy, “Linearized Electrooptic Microwave Downcon-
verter Using Phase Modulation and Optical Filtering,” Optics Express, Vol. 19, No. 2,
2011, pp. 883–895.
[6] Burns, W., G. Gopalakrishnan, and R. Moeller, “Multioctave Operation of Low-Biased
Modulators by Balanced Detection,” Photonics Technology Letters, Vol. 8, No. 1, 1996,
pp. 130–132.
[7] Gallo, J. and J. Godshall, “Comparison of Series and Parallel Optical Modulators for
Microwave Down-Conversion,” Photonics Technology Letters, Vol. 10, No. 11, 1998, pp.
1623–1625.
[8] Chan, E., and R. Minasian, “Microwave Photonic Downconverter with High Conver-
sion Efficiency,” Journal of Lightwave Technology, Vol. 30, No. 23, 2012, pp. 3580–3585.
[9] Tang, Z., F. Zhang and S. Pan, “Photonic Microwave Downconverter Based on an Op-
toelectronic Oscillator Using a Single Dual-Drive Mach-Zehnder Modulator,” Optics
Express, Vol. 22, No. 1, 2014, pp. 305–310.
[10] Da Dalt, N., et al., “On the Jitter Requirements of the Sampling Clock for Analog-
to-Digital Converters,” Transactions on Circuits and Systems I: Fundamental Theory and
Applications, Vol. 49, No. 9, 2002, pp.1354–1360.
[11] Löhning, M., and G. Fettweis, “The E�ffects of Aperture Jitter and Clock Jitter in
Wideband ADCs,” Computer Standards & Interfaces, Vol. 29, No. 1, 2007, pp. 11–18.
[12] Lauritzen, K., S. Talisa, and M. Peckerar, “Impact of Decorrelation Techniques on
Sampling Noise in Radio-Frequency Applications,” Transactions on Instrumentation and
Measurement, Vol. 59, No. 9, 2010, pp. 2272–2279.
[13] Kobayashi, H., et al., “Aperture Jitter Effects in Wideband Sampling Systems,” Proc. of
Instrumentation and Measurement Technology Conference, 1999.
[14] Howard, L., and D. Rabideau, “Correlation of Nonlinear Distortion in Digital Phased
Arrays: Measurement and Mitigation,” Microwave Symposium Digest, 2002.
[15] Valley, G., “Photonic Analog-to-Digital Converters,” Optics Express, Vol. 15, No. 5,
2007, pp. 1955–1982.
164 Applications of Modern RF Photonics

[16] Jacobs, E., et al., “Optically Clocked Track-And-Hold for High-Speed High-Resolution
Analog-to-Digital Conversion,” International Topical Meeting on Microwave Photonics,
2004.
[17] DePriest, C., et al., “High-Quality Photonic Sampling Streams from A Semiconductor
Diode Ring Laser,” Journal of Quantum Electronics, Vol. 38, No. 4, 2002, pp. 380–389.
[18] Zmuda, H., “Analog-to-Digital Conversion Using High-Speed Photonic Processing,”
SPIE International Symposium on Optical Science and Technology, 2001, pp. 84–95.
[19] Taylor, H., “An Optical Analog-to-Digital Converter—Design and Analysis,” Journal of
Quantum Electronics, Vol. 15, No. 4, 1979, pp. 210–216.
[20] Bell, J., et al., “A/D Conversion of Microwave Signals Using a Hybrid Optical/
Electronic Technique,” SPIE Optical Technology for Microwave Applications, Vol. 1476,
1991, pp. 326–329.
[21] Juodawlkis, P., et al., “Optically Sampled Analog-to-Digital Converters,” Transactions
on Microwave Theory and Techniques, Vol. 49, No. 10, 2001, pp. 1840–1853.
[22] Valley, G., J. Hurrell, and G. Sefler, “Photonic Analog-to-Digital Converters:
Fundamental and Practical Limits,” Proc. of SPIE, Vol. 5618, 2004, pp. 96–106.
[23] Walden, R., “Analog-to-Digital Converter Survey and Analysis,” Journal on Selected
Areas in Communications, Vol. 17, No. 4, 1999, pp. 539–550.
[24] Han, Y. and B. Jalali, “Photonic Time-Stretched Analog-to-Digital Converter:
Fundamental Concepts and Practical Considerations,” Journal of Lightwave Technology,
Vol. 21, No. 12, 2003, pp. 3085–3103.
[25] Han, Y., and B. Jalali, “Continuous-Time Time-Stretched Analog-To-Digital Converter
Array Implemented Using Virtual Time Gating,” Transactions on Circuits and Systems I:
Regular Papers, Vol. 52, No. 8, 2005, pp. 1502–1507.
[26] Jalali, B., and X. Yie, “Optical Folding-Flash Analog-to-Digital Converter with Analog
Encoding,” Optics Letters, Vol. 20, No. 18, 1995, pp. 1901–1903.
[27] Currie, M., T. Clark, and P. Matthews, “Photonic Analog-to-Digital Conversion by
Distributed Phase Modulation,” Photonics Technology Letters, Vol. 12, No. 12, 2000, pp.
1689–1691.
[28] Chi, H., and J. Yao, “A Photonic Analog to Digital Conversion Scheme Using Mach
Zehnder Modulators with Identical Half Wave Voltages,” Optics Express, Vol. 16, No. 2,
2008, pp. 567–572.
[29] Li, W., et al., “All Optical Analog to Digital Conversion Based on Polarization-
Differential Interference and Phase Modulation,” Photonics Technology Letters, Vol. 19,
No. 8, 2007, pp. 625–627.
[30] Hayduk, M., R. Bussjager, and M. Getbehead, “Photonic Analog to Digital Conversion
Techniques Using Semiconductor Saturable Absorbers,” Proc. of SPIE, Vol. 4042, 2000,
pp. 54–60.
[31] Oda S., A. Maruta, and K. Kitayama, “All-Optical Quantization Scheme Based on
Fiber Nonlinearity,” Photonics Technology Letters, Vol. 16, No. 2, 2004, pp. 587–589.
 Signal Processing Utilizing RF Photonics 165

[32] Kitayama, K., et al., “Photonic Analog-to-Digital Conversion,” Digest of the LEOS
Summer Topical Meetings, 2005.
[33] Schneider, G., et al., “Radiofrequency Signal-Generation System with Over Seven
Octaves of Continuous Tuning,” Nature Photonics, Vol. 7, No. 2, 2013, pp. 118–122.
[34] McKinney, J. and A. Weiner, “Compensation of the Effects of Antenna Dispersion on
UWB Waveforms Via Optical Pulse Shaping Techniques,” Transactions on Microwave
Theory and Techniques, Vol. 54, No. 4, 2006, pp. 1681–1685.
[35] Yao, J., “Photonic Generation of Microwave Arbitrary Waveforms,” Optics
Communications, Vol. 284, No. 15, 2011, pp. 3723–3736.
[36] Goswami, D., “Optical Pulse Shaping Approaches to Coherent Control,” Physics
Reports, Vol. 374, No. 6, 2003, pp. 385–481.
[37] Fontaine, N., et al., “Compact 10 GHz Loopback Arrayed Waveguide Grating for
High Fidelity Optical Arbitrary Waveform Generation,” Optics Letters, Vol. 33, No. 15,
2008, pp. 1714–1716.
[38] Wang, C. and J. Yao, “Photonic Generation of Chirped Microwave Pulses Using
Superimposed Chirped Fiber Bragg Gratings,” Photonics Technology Letters, Vol. 20,
No. 11, 2008, pp. 882–884.
[39] Wang, C. and J. Yao, “Simultaneous Optical Spectral Shaping and Wavelength-to-
Time Mapping for Photonic Microwave Arbitrary Waveform Generation,” Photonics
Technology Letters, Vol. 21, No. 12, 2009, pp. 793–795.
[40] Cundiff, S., and A. Weiner, “Optical Arbitrary Waveform Generation,” Nature
Photonics, Vol. 4, No. 11, 2010, pp. 760–766.
[41] Leaird, D. and A. Weiner, “Femtosecond Direct Space-to-Time Pulse Shaping in an
Integrated-Optic Configuration,” Optics Letters, Vol. 29, No. 13, 2004, pp. 1551–
1553.
[42] Azana, J., et al., “Reconfigurable Generation of High-Repetition-Rate Optical Pulse
Sequences Based on Time-Domain Phase-Only Filtering,” Optics Letters, Vol. 30, No.
23, 2005, pp. 3228–3230.
[43] Bolea, M., et al., “Optical UWB Pulse Generator Using an N Tap Microwave Photonic
Filter and Phase Inversion Adaptable to Different Pulse Modulation Formats,” Optics
Express, Vol. 17, No. 7, 2009, pp. 5023–5032.
8
Advancements in Integrated RF
Photonics
The RF photonic subsystems that have been covered previously for use in a
demonstration architecture have all been-based on discrete components that
are connected by fiber-optic cables. As discussed in previous chapters, the per-
formance of these components directly relates to the expected performance
of the subsystems, from the relative intensity noise and optical power of the
laser, to the electro-optic efficiency as a function of the RF frequency of the
modulator and finally to the optical power handling and RF bandwidth of the
photodetector. These and other discrete photonic components have been well
developed by both commercial and private industries. In fact, advancements
in these devices over the last 20 years has allowed for the previously described
subsystems to meet the needs of various RF applications. Unfortunately, the
components still suffer from drawbacks. One of the most important is the
need for fiber-optic connections between the components. Most components
come from commercial vendors with fiber-optic cable connections of at least
one meter in length, and often as long as 2 or 3m. Figure 8.1 shows a pho-
tonic component with a 2-m length of fiber attached. For a simple link con-
sisting of a laser, modulator, and photodetector, there can be as much as 12m
of optical fibers connecting all the components. The length can be advanta-
geous when the laser has to be physically placed away from the modulator, or
the optical output of the modulator has to be routed to the front panel of a
large box before it can be connected to a fiber-optic cable. However, the push
has been to smaller and more integrated subsystems. Thus, the extra length of
fiber consumes valuable space and often leads to complex routing inside the

167
168 Applications of Modern RF Photonics

Figure 8.1 Photonic component with a fiber pigtail of 2m. (After: [1].)

subsystem box in order to minimize the space used. Another issue with the
fiber-optic connections comes with their sensitivity to the local environment.
When the laser output is split into two or more parallel paths, any thermal or
vibration changes will be imparted onto the optical fiber. Often these changes
cause the fiber to either stretch or contract, changing the path length that
the light has to travel. If these changes are not matched or canceled out, they
will impart fluctuations on the RF signal that is detected at the output of the
subsystem, oftentimes degrading the fidelity of the signal.
A solution to these issues exists in integration. Being able to realize these
components in an integrated photonic circuit can provide much smaller sizes
while also making the subsystem impervious to environmental fluctuations.
Integration of photonics supports both digital and analog applications. Inte-
gration does come with challenges. While discrete photonic components are
mature, their integrated counterparts are still being improved. This chapter
reviews integrated photonic components for RF photonics, along with sub-
systems that can benefit from the use of these circuits.

8.1 Integrated Photonic Fundamentals

When discussing integrated photonics, some basic definitions have to be set.
Photonic integrated circuit (PIC) is a term that is often used in this area. For
the purpose of this discussion, PICs are defined as photonic components that
are monolithically integrated onto the same substrate. An example would be
a laser integrated with an optical amplifier on a III-V semiconductor material
platform. This provides a single component that can be used in a larger sys-
tem. However, these PICs often require external electronic circuitry to power
 Advancements in Integrated RF Photonics 169

and control these components. In the above example, the laser and amplifier
PIC are packaged and used as discrete, fiber-coupled components. The laser
current source, along with the temperature controller, is separate from the
PIC. On the other hand, integrated photonic circuits (IPCs) cover the area
where multiple photonic and HF electronic components are integrated into
a single circuit. For example, an IPC would only have an external power sup-
ply. The rest of this chapter focuses on IPCs and their use for RF photonic
applications.
One of the ongoing debates in IPC is the choice of material upon which
to fabricate the components. III-V-based semiconductor materials, such as
indium phosphide, have been shown to support the photonic components
necessary for making IPCs. The laser, modulator, and photodiode can all be
fabricated on a single chip. In addition, HF modulators and photodiodes
have been demonstrated in semiconductor materials [2, 3]. However the
use of III-V materials comes with some caveats. The optical loss for III-V
waveguides is larger than it is for silicon [4]. Also semiconductor modulators
have been shown to have strong temperature-dependent performance [5].
Silicon-based integrated photonics (or silicon photonics for short) have been
put forward as an alternative to semiconductor-based implementations. Be-
ing CMOS-compatible leverages the large foundry infrastructure that already
exists for electronic integrated circuits. Along with this advantage, the optical
propagation loss in silicon photonics has very low losses at 1,550 nm when
compared to other material substrates. Germanium-based photodiodes with
HF bandwidths have also been successfully fabricated on silicon [6]. Unfor-
tunately, the use of silicon photonics comes with its own set of drawbacks.
Unlike semiconductor-based IPCs, lasers at 1,550 nm are not easily achiev-
able in silicon. Raman-based laser sources in silicon have been demonstrated,
but require another optical sources to pump the laser [7]. The silicon pho-
tonic modulators that have been demonstrated depend on changing the free-
charge carrier density in a PN diode configuration [8, 9]. This often leads
to limits on the linearity performance of the modulator as well as the upper
frequency operation.
Due to its low loss, silicon photonics appears best suited for use in IPCs
that are mostly passive. These IPCs do not require a laser or optical amplifica-
tion. Examples of these types of circuits are optical filters, switches, attenu-
ators, wavelength multiplexers and demultiplexers, and optoelectronic feed-
back. Semiconductor-based IPCs provide more utility to active applications.
A few examples include RF links, waveform generation, and RF oscillators
based on optical injection locking. Unfortunately, no one material system
seems to meet the needs of all applications.
170 Applications of Modern RF Photonics

The ideal solution would use heterogeneous integration in order to

achieve the best performance. A heterogeneously integrated photonic circuit
would be able to combine the HF operation of CMOS-based electronic cir-
cuits with the optical components afforded by semiconductor technology.
The silicon waveguides and filters would also provide very low loss.

8.2 IPCs
Since RF photonic subsystems have been built on the low loss of optical
fibers, silicon waveguides are preferred for these applications. Therefore, the
use of IPCs will focus on technologies that can be used with these low-loss
waveguides. The first component will be the laser. A germanium-based laser
on silicon has been demonstrated with optical power on the order of a mil-
liwatt [10]. For RF applications, the laser needs to deliver high optical power
with low noise. Fortunately, the preferred discrete optical source is a semicon-
ductor-based DFB laser. Work has focused on integrating the semiconductor
laser with silicon waveguides. Direct bonding of the laser die to the silicon
wafer has been demonstrated [11], with other methods focusing on using
other types of adhesion [12, 13]. The bonding method has to take into ac-
count the different thermal expansion properties of the materials, which can
lead to cracking or separation. In addition, the optical coupling between the
laser and the silicon waveguide has to be properly designed in order to make
the losses as small as possible.
The next component is the optical modulator. For analog applications,
a high electro-optic conversion efficiency along with a low optical insertion
loss will be desired. While a semiconductor-based modulator can be bonded
with the silicon waveguides in a similar fashion as the laser, the lithium nio-
bate (LiNbO­3) MZM has been the preferred modulator for RF photonics.
Thin film LiNbO­3 (TFLN) has been developed in order to integrate it with
silicon [14].The TFLN can be bonded to the silicon using an intermediate
layer such as benzocyclobutene (BCB) that has low losses and will mitigate
the thermal expansion difference between the two materials [15]. TFLN on
silicon has been shown to have an electro-optic conversion efficiency and
insertion loss close to bulk lithium niobate used in discrete modulators [16].
The optical amplifier is the next integrated component for discussion.
Like the laser, the optical gain, along with a minimum amount of noise, is im-
portant. Again, semiconductor optical amplifiers can be bonded to the silicon
waveguides. Still, the semiconductor-based amplifiers have two drawbacks
when used for analog applications. The noise penalty for the semiconductor
amplifier is around 8–26 dB, which is higher than the EDFA that is used
 Advancements in Integrated RF Photonics 171

in most analog applications. In addition, the semiconductor optical ampli-

fier gain has a fast response time, leading to the gain being unintentionally
modulated by the RF signal. This additional modulation can degrade the lin-
ear performance of the overall system. While an EDFA cannot be integrated,
work has been done on doping the silicon waveguides with erbium [17]. The
erbium-based amplifier will require a pump laser while the semiconductor
amplifier can be electrically driven. The choice of amplifier will often be de-
termined by the needs of the system.
For analog subsystems, integrated optical filters with high Q factors
along with low insertion loss are needed, along with the ability to be tuned
over a wide frequency range. Just as filters used in the discrete systems were
based on FBGs, similar types of filters can be realized in the integrated pho-
tonic domain. Bragg gratings can be used in a FP configuration to provide
an integrated filter, as seen in Figure 8.2. However, the optical insertion loss
can be quite large for these devices. Another alternative is the ring resonator
structure as seen in Figure 8.3. The ring resonator can act as both a bandpass
and notch filter, much like the FBG [18, 19]. When light of a certain wave-
length is coupled to the ring structure through the lower waveguide, it will
simply pass through if it does not meet the resonant condition of the ring.
The resonant wavelength is directly related to an integer multiple of the cir-
cumference of the ring. If the wavelength of the light does match the resonant
condition, it will couple into the ring. The light will then couple out the
upper waveguide. The upper waveguide will act like the bandpass while the
lower waveguide will act like the notch filter. Ring resonators lend themselves
to using multiple cascaded copies in order to improve the overall Q factor
[20]. The resonant wavelength can also be tuned, similar to a FBG. The two
methods are either heating [21] or straining by using a liquid-crystal material
as the cladding for the ring resonator structure [22]. These types of filters can
meet the various requirements of RF photonic subsystems.

Figure 8.2 Integrated photonic FP filter. (After: [1].)

172 Applications of Modern RF Photonics

Figure 8.3 Integrated photonic ring resonator filter. (After: [1].)

Finally, the photodetector is the last component that needs to be real-

ized for integrated RF photonic applications. The photodetector has to be
able to handle high optical powers while maintaining a linear response over
a wide frequency operating range. These types of photodetectors have been
made in discrete forms and their performance continues to improve [23, 24].
One of the relevant improvements has occurred in how the light enters the
device. Previous demonstrations have relied upon a surface normal interface
to allow the light to reach the absorption region of the photodiode struc-
ture. As an improvement upon this device, photodetectors have started to
use a waveguide interface [25]. In this case, the photodetector can be placed
right above the waveguide, allowing the light to evanescently couple into the
absorption region. The pursuit of these types of photodetectors can allow
for higher frequency response and make it easier to integrate with silicon
waveguides. In fact, high-power, HF photodetectors have been designed to
integrate with silicon waveguides [26].
All of these heterogeneously integrated components should be able to
provide the required performance for RF photonic subsystems. The main
obstacle right now is the actual fabrication of these circuits in a single foundry
setting. The integration will most likely require some form of 3-D structure,
where the various components can be made separately and then stacked and
bonded together. The key piece will then be the interconnections that link
the different layers, allowing both the electrical and optical signals to move to
the appropriate layer. The loss of these interconnects will be very important,
as the analog performance strongly depends on the optical power that can be
 Advancements in Integrated RF Photonics 173

delivered to the photodetector. Work continues to improve performance in

this area [27, 28].

8.3 Applications of IPCs to RF Photonics

Previous chapters have demonstrated RF photonic applications using discrete
components. IPCs can be used in place of these systems. Starting with the full
demonstration architecture shown in Figure 6.1, the various subsystems will
be analyzed to see how IPCs can be used to replace the discrete component
subsystems.
The RF circulator in the demonstration architecture is based on a
MZM that is configured to isolate the transmitted signal at the antenna from
the received one. The single-sideband modulation in the MZM requires the
use of two 90° hybrids. The discrete hybrids can be quite large in size and are
limited in the range of frequencies over which they operate. In an IPC, the
hybrids can be made on the same chip as the MZM and optical filter. The re-
sulting chip will be much smaller than its discrete version and can potentially
be designed to have much wider frequency operation. In addition, the elec-
tronic bias control circuit needed to hold the MZM at the proper DC bias
point can be included in the chip. Finally the laser can be also integrated onto
the same chip. The RF circulator IPC could then be packaged in the same di-
mensions as a traditional electronic circulator, with the only difference being
that the output port will be a fiber connection rather than an RF connection.
The next subsystem is the signal identification block. The spectrum
analyzer examples from Chapter 7 can be realized in an IPC. The mixture
of photonic and electronics would be a good candidate for this type of im-
plementation. However, the same drawback of scanning the filter will still
limit the utility of the circuit. Using the rare Earth-doped crystal removes
the scanning drawback. Unfortunately, the crystal requires an external cool-
ing system, potentially removing the advantages of miniaturization. Other
methods of separating the different optical sidebands and using an array of
photodiodes would seem better suited for IPCs. Just like the discrete version,
the drawback is the separation that is achievable in the integrated filter.
The other frequency-identification demonstrations depend on passing
the signal through either a dispersive element or a filter. Optical filters in
IPCs can be accomplished either using Bragg gratings or ring resonators. The
dispersive elements can be made in the waveguides as well. One method is
to simply modify the waveguide design in order to induce a strong dispersive
effect. Along the same lines, the waveguide can be doped with materials that
will also increase the dispersion. Unfortunately, both ways may also increase
174 Applications of Modern RF Photonics

the optical loss. The other method for signal identification is the MM-OEO.
Unlike most OEOs, the MM-OEO tries to make the feedback loop as short
as possible. This type of system seems well suited to an IPC, as it requires
both photonic and electronic components that need to be placed as close
together as possible. Along with the capability to amplify the signal, it is a
strong candidate for an IPC implementation.
The next subsystem is the signal-separation block. The integrated opti-
cal filters can provide similar performance to their discrete counterparts. Still,
the finite and infinite impulse response filters may be implemented in an
IPC. The wavelength multiplexer and demultiplexer can be realized in silicon
waveguides with very low loss [29]. The short delays can be implemented as
well along with the amplitude control. The electronic circuit that will control
each of the amplitude weights can also be integrated on the same chip. In
fact other demonstrations have already been shown in integrated forms of the
finite-impulse response filter [30].
The next block to be considered for implementation in an IPC is the
oscillator. The MM-OEO has already been considered a good candidate for
realization in an IPC. The standard OEO, however, requires a long delay line
in order to achieve the low phase noise performance. One solution would be
to integrate all parts of the OEO, but allow an output and input to the circuit
that would allow for a long fiber delay to be added. Unfortunately, coupling
the light out and back into the circuit can lead to extra loss, degrading the
overall performance of the OEO. One way to avoid this loss is to create a cav-
ity on the chip itself that can mimic the long fiber-delay line. An OEO with
a whispering gallery mode optical resonator has been demonstrated that can
produce high-quality, low phase noise RF signals without the need for a long
fiber-delay line [31]. As for oscillators based on optical injection locking of
two lasers, these setups require an integrated optical circulator. Further dis-
cussion on the optical circulator will be reserved for Section 8.5.
The other subsystems that can be integrated are the RF downconverter,
waveform generator, and photonic ADC. The RF downconverter depends
on the modulator, amplifier, filter, and photodetector. For a parallel imple-
mentation of the downconverter, the advantage of the integrated version is
the environmental stability that comes from miniaturization onto a common
platform. In addition, the downconverter can be integrated with either an
electronic- or photonic-based oscillator on the same chip. As for the wave-
form generator and the ADC, they both share the need for a pulsed laser
source. The use of an integrated ring resonator for generating an optical pulse
stream has been demonstrated by multiple groups [32, 33]. For the ADC,
an IPC can provide all the required parts. For example, the system shown in
Figure 7.6 can be realized. The electronics required for the clock recovery,
 Advancements in Integrated RF Photonics 175

and the array of electronic ADCs, can be combined with the optical modula-
tor, switch, and array of photodetectors to realize the full subsystem. For the
waveform generator, the one piece that is missing is the SLM. Fortunately,
this function can be accomplished by a wavelength demultiplexer along with
a phase and amplitude control piece. This is similar to the FIR filter block
shown in Figure 5.6.

8.4 Other Applications in IPCs

The demonstration architecture utilizes a single antenna for both reception
and transmission of signals at the same time. A single RF aperture is used to
show how the RF photonic circulator can separate the bidirectional signal
paths. However, some modern RF systems do not use a single antenna. In-
stead an array of antennas is used. One of the advantages of such a configura-
tion is the overall RF gain for this type of aperture. An array of antennas can
have N times the gain of a single antenna, where N is the number of elements
in the array. For the increase to be realized, the signals from each of the ele-
ments of the array have to add coherently. This is known as beamforming
the received signal. In order to beamform the signal, the phase of the signal
from each element has to be controlled independently. In addition, to avoid
unwanted grating lobes in the antenna pattern, the spacing between elements
has to be no more than half a wavelength of the highest frequency that will be
received. For signals above 30 GHz, the spacing can be on the order of milli-
meters. Placing a discrete photonic modulator at each element will not physi-
cally fit in such a tight spacing. An IPC-based modulator should be able to fit.
In addition, the IPC can have a low-noise amplifier integrated between the
antenna element and the modulator. Making multiple copies of the photonic
circuit will allow each element to be upconverted into the optical domain.
In order to address the need for phase control, different methods can
be implemented. An electronic phase shifter can be integrated into each of
the photonic circuits behind each element. Unfortunately, the electrical phase
shifter will only provide the correct phase shift for a small bandwidth of sig-
nals. Since the system is meant to be wideband, another method must be
used. A photonic-based phase shifter has been shown to provide a flat phase
shift over a large bandwidth of frequencies [34]. The phase change can be
accomplished with a change in RF power of less than 0.2 dB over the entire
frequency range. In Figure 8.4, the phase of a set of signals over 4 GHz of
bandwidth can be tuned from +90° to –90°. Another method to ensure that
the signals add coherently is to use a true time-delay element. In true time
delay, the physical path that the signal travels over is changed. Often this
176 Applications of Modern RF Photonics

Figure 8.4 Phase shift from 90° to –90° degrees over a 1–12-GHz operating frequency
range along with power change over the entire range of phase change.

involves the use of delay lines that can be switched in and out in order to
get the correct amount of delay. While this type of delay line is very short,
implementing them in an electronic circuit can be very lossy and require am-
plifiers, which, in turn, add noise. A photonic circuit version would be able
to accomplish the same task but with less loss and potentially without the
need for additional amplification. Figure 8.5 shows an example of a delay-line
architecture along with switches.
With phase control, the array of antennas can now be steered to beam-
form signals that arrive at a certain angle to the front of the array. When the
phase of each of the elements of the array is set correctly, a signal received
from a specified angle will constructively add, while signals arriving from oth-
er directions will not. By changing the phases at each element, the array can
be steered to different receive angles. However, the array can only be steered
to one angle at a time. In order to form more than one beam at the same
time, multiple phase shifters are needed at each element. The signal from
each element would have to then be split to the multiple phase shifters and
then recombined along separate paths. Such a configuration would be quite
large. One of the advantages of using photonics is the ability to beamform
multiple signals at the same time. Instead of using one laser, multiple lasers
at different wavelengths can be combined through a wavelength multiplexer
into the same modulator. Then each wavelength will pass through the same
photonic delay lines. Due to dispersion, each of the wavelengths will see a
different time delay. The different time delays mean that each wavelength
 Advancements in Integrated RF Photonics 177

Figure 8.5 Block diagram of a true time-delay circuit using switches for variable delays.

will beamform a different angle from the array. Thus, the photonic circuit can
form multiple beams at the same time.

8.5 Further Work in IPCs for Analog Applications

IPC technology has reached the point that it can now be considered to ad-
dress the needs of RF photonic subsystems. However, a couple of points still
need to be addressed. The first is the lack of an integrated optical circulator
and isolator. Note that an optical isolator is simply an optical circulator with
the third port terminated, so this discussion will cover both components.
Discrete optical circulators depend on a ferromagnetic material in order to
isolate the three ports. Integration of ferromagnetic materials onto an inte-
grated platform has proven to be troublesome. Work has been focused on
ferromagnetic materials such as cerium ytterbium garnet (Ce:YiG) to try and
integrate with silicon platforms for isolators [35]. Recent work has shown
integrated devices that can provide isolation as high as 20 dB [36] as well as
dynamically tunable circulators [37]. While these demonstrations are promis-
ing, they have to be made repeatable enough to be transferred into a foundry
process for high-yield fabrication.
This leads to the second point. For all the work that has been covered in
IPCs, there is a limited amount of access for designing and fabricating these
chips in small quantities. The electronic industry addressed this problem with
services such as the metal oxide semiconductor implementation service (MO-
SIS) [38]. The service allows for multiple users to design into a shared wafer.
These multiproject wafers (MPWs) allow users to fabricate prototype circuits
without having to buy an entire wafer. The resulting cost reduction allows
groups such as universities and small businesses access to foundries. The IPC
area needs a similar service. The European Union is addressing this need with
the Joint European Platform for Photonic Integration of Components and
Circuits (JePPIX) [39]. JePPIX is looking to bring together indium phos-
phide and silicon nitride platforms together under one service. MPWs would
be available for access to members. The ePIXfab service is also a consortium
providing silicon photonic design access and contracting out the fabrication
without actually having its own foundry (known as a fabless service) [40].
178 Applications of Modern RF Photonics

Also, the United States recently set up a manufacturing institute in IPCs

called AIM Photonics [41]. While both services are looking to integrate both
traditional electronic circuits along with photonic circuits, the main custom-
ers appear to be digital data communications. However, just as RF photonics
took the digital communications discrete components and refined them for
analog applications, the RF IPCs may also have to leverage the larger market
provided by digital data applications.

8.6 Conclusions
IPCs have the potential to meet the needs of current and future analog appli-
cations. The current advantage of these circuits is a reduction in size and iso-
lation from environmental fluctuations. By heterogeneously integrating dif-
ferent material structures onto silicon waveguides, the best performance can
be achieved. This chapter reviews the various components that are required
for IPCs as well as the subsystems that could be realized. While demonstra-
tions have been made that show the promise of these photonic circuits, they
have not been made accessible to general users. The AIM Photonics institute
appears to be addressing this issue in the United States. Much like the evo-
lution of discrete components, IPCs will most likely be pushed forward by
digital communication requirements first. The use of photonic circuits for
communications between mainframes in large server farms may lead to a
high-yield silicon photonic process. In turn, this will reduce the cost of these
components and allow for a reduction in costs for the analog version of these
components, along with further advancements. In the end, the same path as
discrete photonic development may be followed for IPCs. Namely, digital
applications will allow for analog applications to be realized at a cheaper cost.
However, improvements in performance will still need to be realized.

References
[1] Devgan, P., “Signal Processing Utilizing Radio Frequency Photonics,” AFRL Report,
AFRL-RY-WP-TR-2017-0172, 2017.
[2] Irmscher, S., R. Lewen, and U. Eriksson, “InP-InGaAsP High-Speed Traveling-Wave
Electroabsorption Modulators with Integrated Termination Resistors,” Photonics Tech-
nology Letters, Vol. 14, No. 7, 2002, pp. 923–925.
[3] Bach, H. G., et al., “InP-Based Waveguide-Integrated Photodetector with 100-GHz
Bandwidth,” Journal of Selected Topics in Quantum Electronics, Vol. 10, No. 4, 2004, pp.
668–672.
 Advancements in Integrated RF Photonics 179

[4] Bowers, J., “Heterogeneous silicon III–V Photonic Integrated Circuits,” Proc. of Optical
Fiber Communications Conference and Exhibition, Los Angeles, CA, 2015.
[5] Boss, M., D. Evans, and E. Jacobs, “Performance of a 40-GHz RF Photonic Balanced
Link Using a Polarization Modulator,” Proc. of Avionics, Fiber-Optics and Photonics Tech-
nology Conference, San Diego, CA, 2013.
[6] Famà, S., et al., “High Performance Germanium-on-Silicon Detectors for Optical
Communications,” Applied Physical Letters, Vol. 81, No. 4, 2002, p. 586.
[7] Boyraz, O., and B. Jalali, “Demonstration of a Silicon Raman Laser,” Optics Express,
Vol. 12, No. 21, 2004, pp. 5269–5273.
[8] Liu, A., et al., “High-Speed Optical Modulation-based on Carrier Depletion in A Sili-
con Waveguide,” Optics Express, Vol. 15, No. 2, 2007, pp. 660–668.
[9] Timurdogan, E., et al., “An Ultralow Power Athermal Silicon Modulator,” Nature Com-
munications, Vol. 5, No. 4008, 2014, pp. 1–11.
[10] Camacho-Aguilera, R. E., et al., “An Electrically Pumped Germanium Laser,” Optics
Express, Vol. 20, No. 10, 2012, pp. 11316–11320.
[11] Liang, D., and J. E. Bowers, “Recent Progress in Lasers on Silicon,” Nature Photonics,
Vol. 4, No. 8, 2010, pp. 511–517.
[12] Roelkens, G., et al., “Adhesive Bonding of InP∕InGaAsP Dies to Processed Silicon-
on-Insulator Wafers Using DVS-bis-benzocyclobutene,” Journal of the Electrochemical
Society, Vol. 153, No. 12, 2006, pp. G1015–G1019.
[13] Kostrzewa, M., et al., “InP Dies Transferred onto Silicon Substrate for Optical
Interconnects,” application. Sensors and Actuators A: Physical, Vol. 125, No. 2, 2006,
p. 411–414.
[14] Toney, J., Lithium Niobate Photonics, Norwood, MA: Artech House, 2015.
[15] Guarino, A., et al., “Electro–optically Tunable Microring Resonators in Lithium
Niobate,” Nature Photonics, Vol. 1, No. 7, 2007, pp. 407–410.
[16] Rao, A., et al., “High-Performance and Linear Thin-Film Lithium Niobate Mach–
Zehnder Modulators on Silicon up to 50 GHz,” Optics Letters, Vol. 41, No. 24, 2016,
pp. 5700–5703.
[17] Kik, P. G., and Polman, A., “Erbium-Doped Optical-Waveguide Amplifiers on Silicon,”
MRS Bulletin, Vol. 23, No. 4, 1998, pp. 48–54.
[18] Rabus, D. G., Integrated Ring Resonators, Berlin Heidelberg: Springer-Verlag, 2007.
[19] Little, B. E., et al., “Ultra-Compact Si/SiO2 Microring Resonator Optical Channel
Dropping Filters,” Photonics Technology Letters, Vol. 10, No. 4, 1998, pp. 549–551.
[20] Xu, Q., et al., “Cascaded Silicon Micro-Ring Modulators for WDM Optical
Interconnection,” Optics Express, Vol. 14, No. 20, 2006, pp. 9431–9436.
[21] Chen, L., N. Sherwood-Droz, and M. Lipson, “Compact Bandwidth-Tunable
Microring Resonators,” Optics Letters, Vol. 32, No. 22, 2007, pp. 3361–3363.
180 Applications of Modern RF Photonics

[22] De Cort, W., et al., “Tuning of Silicon-on-Insulator Ring Resonators with Liquid
Crystal Cladding Using the Longitudinal Field Component,” Optics Letters, Vol. 34,
No. 13, 2009, pp. 2054–2056.
[23] Li, Q., et al., “High-Power Flip-Chip Bonded Photodiode with 110 GHz Bandwidth,”
Journal of Lightwave Technology, Vol. 34, No. 9, 2016, pp. 2139–2144.
[24] Beling, A., and J. C. Campbell, “InP-based High-Speed Photodetectors,” Journal of
Lightwave Technology, Vol. 27, No. 3, 2009, pp. 343–355.
[25] Shi, Jin-Wei, et al., “Extremely High Saturation Current-Bandwidth Product
Performance of a Near-Ballistic Uni-Traveling-Carrier Photodiode with a Flip-Chip
Bonding Structure,” Journal of Quantum Electronics, Vol. 46, No. 1, 2010, pp. 80–86.
[26] Arakawa, Y., et al., “Silicon Photonics for Next Generation System Integration
Platform,” IEEE Communications Magazine, Vol. 51, No. 3, 2013, pp. 72–77.
[27] Parekh, M. S., P. A. Thadesar, and M. S. Bakir, “Electrical, Optical And Fluidic Through-
Silicon Vias for Silicon Interposer Applications,” Proc. of Electronic Components and
Technology Conference, 2011.
[28] Killge S., et al., “Optical Through-Silicon Vias,” in 3D Stacked Chips (I. Elfadel, G.
Fettweis, eds.), New York, NY: Springer, 2016 pp. 21–234.
[29] Trinh, P. D., et al., “Silicon-on-Insulator (SOI) Phased-Array Wavelength Multi/
Demultiplexer With Extremely Low-Polarization Sensitivity,” Photonics Technology
Letters, Vol. 9, No. 7, 1997, pp. 940–942.
[30] Chen, H.-W., et al., “Integrated Microwave Photonic Filter on a Hybrid Silicon
Platform,” Transactions on Microwave Theory and Techniques, Vol. 58, No. 11, 2010,
pp. 3213–3219.
[31] Matsko, A., et al., “Whispering Gallery Mode-based Optoelectronic Microwave
Oscillator,” Journal of Modern Optics, Vol. 50, No. 15-17, 2003, pp. 2523–2542.
[32] Kippenberg, T. J., R. Holzwarth, and S. A. Diddams. “Microresonator-Based Optical
Frequency Combs,” Science, Vol. 332, No. 6029, 2011, pp. 555–559.
[33] Ferdous, F., et al., “Spectral Line-By-Line Pulse Shaping of On-Chip Microresonator
Frequency Combs,” Nature Photonics, Vol. 5, No. 12, 2011, pp. 770–776.
[34] Chan, E. H. W., and R. A. Minasian, “Photonic RF Phase Shifter and Tunable Photonic
RF Notch Filter,” Journal of Lightwave Technology, Vol. 24, No. , 2006, pp. 2676–2682.
[35] Bi, L., et al., “On-Chip Optical Isolation in Monolithically Integrated Nonreciprocal
Optical Resonators,” Nature Photonics, Vol. 5, No. 12, 2011, pp. 758–762.
[36] Tien, M. C., et al., “Silicon Ring Isolators with Bonded Nonreciprocal Magneto-Optic
Garnets,” Optics Express, Vol. 19, No. 12, 2011, pp. 11740–11745.
[37] Huang, D., et al., “Dynamically Reconfigurable Integrated Optical Circulators,”
Optica, Vol. 4, No. 1, 2017, pp. 23–30.
[38] https://www.mosis.com/what-is-mosis .
[39] http://www.jeppix.eu/welcome/.
 Advancements in Integrated RF Photonics 181

[40] http://epixfab.eu/about/about-epixfab/.
[41] http://www.aimphotonics.com/.
9
Conclusions
The field of RF photonics has progressed significantly in the past 20 years.
After the explosion in the use of photonics for digital communications in the
late 1980s to late 1990s, a strong push was made to incorporate photonics
into analog applications, such as cable TV distribution. Unfortunately, the
same photonic components that were able to provide longer reach and higher
bandwidth for digital applications were not able to compete with electronic
solutions in RF applications. The high RF loss and NF associated with early
RF photonic links kept them from being widely used. The optical fiber had
already reached its optimal loss performance of just under 0.2 dB/km, while
the optical amplifier could provide significant optical gain with an associated
optical NF around 3 dB. On the other hand, the laser, modulator, and photo-
diode were identified as the weak components for use in analog applications.
Most digital networks are designed to avoid nonlinear effects that occur when
a significant amount of optical power propagates down tens of kilometers of
optical fiber. Since the networks use wavelength-division multiplexing, each
individual laser’s power is kept at 1 mW or less so that the aggregate optical
power does not exceed the nonlinear threshold. In addition, the modulator
for digital applications is not optimized for linearity or insertion loss, as the
digital data can be corrected in postprocessing and the optical power is pur-
posefully kept low. Finally, the low power of the laser reduces the requirement
on the power-handling of the photodiode.

183
184 Applications of Modern RF Photonics

9.1 A Brief Review of RF Photonics

Having identified the issue, these three components—the laser, the modu-
lator, and the photodiode—were the subject of a large amount of research
and engineering aiming to improve their performance. The RF photonic
requirements for these components are more stringent, requiring advance-
ments in their performance. Starting with the laser, the RIN has been im-
proved significantly. Where a RIN of –145 dBc/Hz would be acceptable for
digital applications, RIN on the order of –155 dBc/Hz is now realizable in
commercially available parts [1] with –165 dBc/Hz available from specially
designed devices [2]. Along with these noise results, the output power needs
to be maximized. Commercially available lasers at 100 mWs of output power
in the 1,550-nm range can achieve the above noise levels, while research has
produced lasers with higher output powers available with low noise. Finally,
the laser linewidth needs to be minimized. While not as important in the
intensity-modulated direct detection link used in the demonstration architec-
ture, the linewidth does impact potential use with phase modulation. Com-
mercially available lasers with linewidths on the order of 1 kHz are available
[3] while specialty lasers can achieve subhertz linewidths [4]. However, these
types of lasers are only available at 20 mW and do not have the lowest RIN.
Often these trade-offs, along with others, play a role in the choice of modula-
tion format.
After the laser, the modulator is a key component. Extending the RF
bandwidth is necessary as digital communications has not gone beyond 40
Gbps on one polarization of a single optical carrier. Commercially available
phase modulators at 100 GHz are now available [5, 6], with Mach Zehnder
interferometer-based intensity modulators also being available at similar
frequencies. Research work showing modulation out to 300 GHz [7] has
also been demonstrated. Along with the extended frequency operation, the
electro-optic efficiency (directly related to the Vp) and the insertion loss has
to be optimized. Specialty devices have been shown with low Vp [8] and an
insertion loss of around 3 dB.
Finally, the photodetector continues to have its performance increased.
Like the modulator, the extended bandwidth is necessary to meet the re-
quirements of the RF photonic systems. Commercially available components
out to 20 GHz are readily available with high responsivity (optoelectronic
efficiency) [9]. Achieving higher power-handling and higher frequency has
required new designs. While research has demonstrated devices with both
high power and ultrawide frequency [10, 11], commercially available devices
are only now starting to compete [12]. Again, note that the high power and
ultrawide frequency often come at the expense of reduced responsivity.
 Conclusions 185

While the optical amplifier is also an important component, the perfor-

mance has already been optimized. The erbium-doped fiber amplifier already
has an optical NF near the theoretical limit of 3 dB, and the noise penalty can
be optimized. In terms of compactness, the EDFA can be made in very small
packages [13]. While amplifiers using semiconductor or nonlinear optical ef-
fects continue to be explored, the EDFA is still the chosen amplifier.
The optical filter has been well explored in terms of the use of a FBG.
Commercially available devices with bandwidths on the order of 50 MHz are
available [14]. While their tunability is limited, their optical insertion loss
and their high Q, makes them preferred devices for RF applications.

9.2 Discrete-Based RF Photonic Subsystems

The transmission of analog signals over a fiber-optic link is one of the first
and most basic applications of RF photonics. As described in Chapter 2,
there are multiple ways to realize an RF photonic point-to-point link. Com-
mercial vendors provide both MZM-based photonic links [15, 16] and di-
rect laser modulation-based ones [17]. An extension of a simple RF photonic
link, vendors also provide analog delay lines, or analog memory units [18].
Due to a need for higher-fidelity delay lines, research work continues to seek
improvement in the RF performance of the photonic delay lines [19]. The
demonstrated performance allows for delay lines that can operate up to 40
GHz with a dynamic range on the order of 103‑107 dB•Hz2/3 [20], while
other delay lines have been demonstrated with a NF less than 10 dB and with
a positive gain of almost 40 dB [21]. Clearly, delay lines can be made with
different applications in mind.
Appearing just after the antenna, the RF photonic-based circulator is
a subsystem that serves multiple purposes. It isolates the transmission and
reception path, allowing the antenna to serve two functions at once. The
circulator also takes the received signals and upconverts them into the optical
domain. By doing so, the signals can now be operated on by other photonic-
based systems. Work in photonic-based RF circulators has focused on dif-
ferent needs. In some cases, the emphasis is on a single antenna aperture,
isolated from other sources of interference. Here, the circulator has to han-
dle very high RF powers on the transmission side. Also, the received signals
can be very low-power, requiring the circulator to have a low NF. In other
cases, the antenna could be part of a larger array of antennas, reducing the
required power at each aperture. However, the various antennas can couple in
strong unwanted signals. Under this condition, the circulator has to be able
to operate over a large range of input power without having intermodulation
186 Applications of Modern RF Photonics

distortion products appearing in the receive path. The SFDR of the circula-
tor is the important RF metric for this application. Research work continues
on these RF photonic-based circulators [22] with possible paths to providing
commercial products [23].
The signal separation subsystem depends strongly on optical filter tech-
nology. Commercial vendors are strong in both FBG [24] as well as in thin
film-based filters [25]. AWGs are another optical filter technology that can
be used for signal separation, with at least one commercial vendor providing
solutions [26]. With the use of an AWG, the finite impulse response filter can
be realized. In order to overcome the coherence problems, a multiwavelength
source has to be used. One method is to use a bank of lasers, each operating at
their own independent wavelength. Banks of lasers are available commercially
[27], while research on a single source that can provide multiple independent
optical carriers is ongoing [28]. The infinite impulse response filter can also
be realized by the multimode OEO [29].
The photonic RF oscillator has been investigated thoroughly. The OEO
has been around since its invention in 1996. Commercial OEO systems are
now also available for purchase from commercial vendors [30]. Neverthe-
less, research work continues on OEOs, with one of the main areas focused
on increasing the tuning range of the frequency of the signal from the OEO
[31].The other photonic oscillator highlighted in Chapter 8 is based on the
optical injection locking of two lasers. Research work in this area is ongoing
with hybrid systems that combine the OEO feedback with the flexibility of
the optical injection locking scheme showing promise [32]. Beyond these two
types of photonic RF oscillators, optical frequency combs can also generate
low phase noise RF signals. In fact, the carrier envelope phase-locked opti-
cal frequency comb source can have very low phase noise [33]. The system
takes advantage of a mode-locked laser to generate very short optical pulses.
The optical spectrum associated with these very short pulses can cover over
an octave in frequency. By taking a spectral comb line, frequency-doubling
it, and then beating the doubled signal with the spectral comb line of the
same frequency, an error signal can be generated. The error signal can then
be used to correct the envelope offset. The result is a very stable frequency
source that can be used as an RF clock source. The carrier envelope phase-
locked frequency comb source is now available for purchase through multiple
vendors [34, 35].
The signal identification block depends on a form of spectrum analysis.
The use of SHB in a crystal can capture a wideband RF spectrum [36]. The
RF spectrum is upconverted onto an optical carrier and then written into the
rare Earth-doped crystal. A second laser is then used to read out the stored RF
spectrum, allowing the center frequency of any signal within the bandwidth
 Conclusions 187

to be recovered. A commercial company is developing a crystal-based spec-

trum analyzer for both frequency identification and analog data processing
[37]. Research work in methods using dispersion for frequency identification
is also ongoing. One other method involves the use of a recirculating loop
[38]. An RF signal can be input to a long fiber loop with dispersion. After
each trip through the loop, the RF sidebands will see different time delays.
When enough delay is accumulated, the time-delayed sidebands are autocor-
related at a detector. The autocorrelation signal is then processed and the
center frequency is determined.
Another method for spectrum analysis is channelization. A channel-
izer takes a wideband RF spectrum and decimates it into smaller parts or
channels. Each channel can be processed in parallel. RF photonics lends
itself to channelization by the use of optical filters that can be arrayed to
split the wideband spectrum into the individual channels. An example of an
RF photonic-based channelizer has been previously demonstrated [39]. The
RF spectrum is modulated onto an optical carrier. The upconverted optical
signals are then passed through an optical filter bank. After the filter bank,
each of the channels is then downconverted to an intermediate frequency and
digitized for processing. Many examples of RF photonic channelizers have
been investigated, including using an optical frequency comb [40] as well as
a parallel set of chirped filters that chops up a larger spectral bandwidth to
smaller 2-GHz channels [41].
The processing block of the demonstration architecture contains both
the downconversion subsystem and the ADC. Both of these technologies con-
tinue to be explored in multiple architectures. Downconversion techniques
are now incorporating multiple sidebands in order to spectrally fold higher-
order Nyquist bands to the first Nyquist band [42]. This method is similar to
the electronic Nyquist folding receiver [43]. Other methods have used opti-
cal pulses in order to downconvert higher-frequency RF signals [44]. New
methods continue to be explored, taking advantage of new advancements in
optical combs as well as short pulses.
The use of stable optical pulses in photonic ADCs is also pushing the
performance of these systems. While a large amount of work is still being
done on improving sampling jitter in the area of research [45], there have
been demonstrations of photonic sampling being used in real-world systems
[46]. However, photonic-based analog-to-digital conversion is still based on
discrete components. Both analog-to-digital and downconverter systems will
benefit from IPC advancements, perhaps finally allowing them to compete
with purely electronic solutions.
188 Applications of Modern RF Photonics

9.3 Alternative Systems Using RF Photonics

The demonstration architecture presented throughout the book is just one
example in which different RF photonics subsystems can be used. Multiple
other examples exist of system architectures that incorporate different types
of RF photonic subsystems. RF photonic links are the most commonly used
for different applications. In addition to their use in antenna remoting and
analog delay lines, RF photonic links can be used for other applications. One
example is the use of an RF photonic link for in-situ antenna calibration [47].
A large antenna array often needs to be calibrated after long operational peri-
ods. To provide a calibration signal to each of the antenna elements, long cop-
per coax cables are used. As these metal cables block or distort any incoming
signals, the antenna system has to be taken out of service while the calibration
is done. However, a fiber-optic cable can be used to provide the RF signal to
the antenna elements without interfering with any incoming signal. The key
element in this setup is an unbiased photodiode at end of each fiber-optic
link. The photodiode can convert the optical signal back to the RF domain
to excite the antenna. Since the photodiode is unbiased, it can only handle a
certain amount of optical power, and the RF conversion is very lossy. Fortu-
nately, the photodiode can be placed very close to the antenna element, com-
pensating for the large loss. Investigations into the nonlinear performance
of the unbiased photodiode and its utility for antenna calibration have been
done [48, 49]. While not a traditional use of an RF photonic link, it is one
that makes use of the components developed for RF photonic applications.
Another example is a photonically fed array of tightly coupled antenna
elements. Tightly coupled arrays have been shown to have ultrawideband RF
performance [50]. However, the elements are often closely spaced and require
RF feeds that can operate over a large bandwidth. On top of this, the im-
pedance of the tightly coupled antenna elements may not be easily matched
with a traditional coaxial cable. Any significant mismatch in the impedance
will result in reflections and reduce the radiated power from the antenna. To
overcome these limitations, an RF photonic feed has been demonstrated with
a tightly coupled array [51]. A fiber-optic cable is connected to a photodiode
behind each of the elements. The photodiodes are biased in this case, as op-
posed to the antenna calibration example presented above. The photodiode
converts the optical signal into an RF signal that will then be radiated out
of the antenna. The photodiode has very low resistance, allowing an exter-
nal impedance to be used to match the impedance of the element. Previous
demonstrations used a ultrawide frequency photodiode connected to single
antenna to send digital data on a 120-GHz RF carrier [52]. Using an array
ultrawide frequency photodiode with moderately high RF output can make
 Conclusions 189

use of the ultrawideband capability of the tightly coupled array [53]. A widely
tunable RF signal can be generated in the photonic domain, by beating two
optical carriers at the photodiode. Along with the wideband capability, a pho-
tonically fed antenna can provide the advantages of lower weight and size,
while also providing immunity to electromagnetic interference.
While the previous two examples focused on radiating RF signals over
large areas or distances, the same technology can be used for much smaller
scale applications. A photoconductive probe can be used to perform time do-
main reflectometry on integrated electronic circuits [54]. As the circuits’ node
size becomes smaller and smaller, small defects can reduce the performance.
To detect these small defects, very short time transitions are needed. The
advantage of using RF photonics is the availability of short optical pulses.
The short optical pulses can resolve neighboring defects with a ten times im-
provement when compared to electronic time domain reflectometry. The use
of specially designed photoconductive probes provides advantages over other
demonstrations [55]. To measure the circuit under test, the optical pulses
are used as an input to the injection photoconductive probe. The resulting
terahertz radiation is passed out the probe tip into the test circuit. Then, a
detection probe tip is used. Using a copy of the optical pulses allows for the
coherent detection of the radiation at the detection probe tip. The detection
probe will sample any reflections from circuit defects. Along similar lines, a
ultrawide frequency photodiode can be used to radiate millimeter or terahertz
radiation through a sample [56]. Note that the sample can be an integrated
circuit or a biological sample. The input to the photodiode is often generated
by the beating of two laser sources at the photodiode. After passing through
the sample, the radiation is mixed down with a LO, also generated by the
beating of two lasers. A lock-in amplifier is used to recover both amplitude
and phase information. The key components of this system are an oscillator
based on photonic sources and a ultrawide frequency photodiode.
While many other RF photonic link demonstrations exist, other sys-
tems have also been demonstrated. An RF photonic-based coherent radar
system has been developed and used in a real-world situation [57]. At the
heart of the photonic radar system is a low-jitter optical pulse source. The
pulse source can be used to create different waveforms for transmission from
the radar. The carrier frequency of the photonically generated can as high as
40 GHz, without the need for RF upconversion. On the receive side of the
radar, the same optical pulse source can be used as the optical sampler for a
photonic ADC. Again, the system requires no downconversion on the receive
end. The system has even been used in real-world environments in order to
track commercial airplanes. The track data from the photonic radar agreed
very well with other data sources, confirming that the radar operated as
190 Applications of Modern RF Photonics

expected. When compared to a traditional electronic-based radar system, the

main advantages of the photonic-based system are higher carrier frequency
and lower timing jitter. The main weakness in the photonic system is a lower
dynamic range on the receiver as compared to the electronic equivalent. A
review of the radar performance has been performed independently [58]. The
review article notes that the system may be limited in the range over which
the radar can accurately operate. Improvements in the long-term stability of
the optical pulse source can lead to better range performance. In spite of these
limitations, the article also notes that the radar demonstration shows great
promise in leading to a larger acceptance of RF photonic-based systems in
real-world applications.
Other applications of RF photonics have been pursued using subsys-
tems in nontraditional ways. An OEO can be driven into a chaotic state
[59]. The optoelectronic oscillator is similar to the MM-OEO demonstrated
in Chapter 5. However, where the MM-OEO is kept below threshold, the
chaos OEO is set to operate above threshold. The experimental and theoreti-
cal work was able to map the different operating states of the OEO. When
the OEO was driven into a specific chaotic state, a flat, broadband power
spectrum was generated. As noted in the article, the chaotic behavior of the
OEO is different than other chaotic systems. This investigation could lead to
fundamentally new observations in chaos.
RF photonic systems can also perform Fourier processing on signals
without the need for digitization. The use of optics for Fourier transforma-
tion and processing is well understood [60]. Arbitrary waveform generation
is one example of an RF photonic application that makes use of an optical
Fourier transform. While an early demonstration of waveform generation
used a free-space setup involving two lenses and a spatial mask [61], more
recent examples use other methods. Another example is the rare Earth-doped
crystal-based spectrum analyzer. The spectrum analyzer takes the time do-
main signals and stores the energy in the frequency domain. The Fourier
transform allows the spectrum analyzer to capture wideband signals instan-
taneously [62]. Finally demonstrations have shown the capability to do frac-
tional Fourier transforms (FrFTs) on RF signals [63]. The FrFT can be useful
for recovering information on chirped RF waveforms. Performing this opera-
tion in the photonic domain reduces the latency associated with this type of
processing.
Finally RF photonics can be used for advanced sensors. Two examples
are a millimeter-wave radiometer and photonic-based gyroscopes. A radiom-
eter is often used to measure the amount of radiation that is emitted from a
blackbody source [64]. A basic radiometer consists of an antenna, a detector,
and an integrator. A photonic implementation of a radiometer allows the
 Conclusions 191

system to measure noise powers at millimeter-wave frequencies or higher,

beyond what is capable of using purely electronic means [65, 66]. In fact,
such a system could be used as a passive imager [67]. At least one commercial
company has even begun to use such imaging technologies for different ap-
plications [68].
As a second example, a gyroscope is used to measure the angular veloc-
ity of an object. The photonic gyroscope has been used for many years in
many different applications [69]. The photonic implementation of the gyro-
scope has provided higher sensitivity and stability than other demonstrations.
As the limitations on the performance of these gyroscopes have been inves-
tigated, the need for high-power-handling photodiodes, low-noise lasers,
and low-loss optical loops has been identified [69]. All of these components
are being developed to meet other RF photonic needs. The performance of
photonic gyroscopes will increase as the RF photonic components improve.
Thus, photonic gyroscopes can also be considered a part of the RF photonic
application space. Clearly, many examples of RF photonic applications exist.
While many are covered here, the list is by no means complete. Those that
are left out are not due to any technical deficiency but merely limited space.

9.4 Future Work in RF Photonics

New applications will continue to be developed as both the needs of future
systems increase in frequency and instantaneous bandwidth and the compo-
nent performance improves. Wireless communications has become one of the
new drivers for technological innovations. With the introduction of the smart
phone, wireless data usage continues to increase. As opposed to traditional
long-haul communications, where each household needs access to high-speed
data, wireless communications provides individual users access to the same
amount of data bandwidth. In order to meet this demand, wireless providers
now need to utilize higher-frequency carriers. Fifth-generation systems are
already considering the use of carrier frequencies in the 30-GHz or higher
range in order to meet the data demand, as well as to access a part of the spec-
trum that is less congested. Other commercial and military applications are
also looking to exploit the millimeter-wave region of the spectrum.
Looking to the near future of RF photonics, three key developments
appear to be the driving factors in new discrete component-based RF pho-
tonic applications: the optical frequency comb, the high-Q optical cavity,
and the high-power, ultrawide frequency photodiode. The optical fiber and
the amplifier used in discrete systems are mature and already are well near
the practical limits in terms of performance. While electro-optic modulator
192 Applications of Modern RF Photonics

performance has room for improvement, the electro-optic conversion effi-

ciency trades RF gain and NF improvements for dynamic range or vice versa.
Besides, many electronic and photonic means exist for improving the overall
performance of the modulator without changing the electro-optic conversion
efficiency.
Advancements in the optical frequency comb have utility in many of
the RF photonic applications previously discussed. In the same way that
digital photonic networks make use of WDM with multiple optical carriers,
RF photonics will benefit from optical frequency combs. From oscillators to
downconversion to waveform generation to ADCs, all of these can benefit
from the optical frequency comb. The ability to replace multiple lasers with a
single source that can generate coherent optical lines is not only cost-saving,
but can lead to new signal-processing techniques.
The high-Q photonic cavity will allow for many different applications.
The oscillator, signal separation, and signal identification systems can make
use of a high-Q cavity. An OEO with a high-Q photonic cavity will provide
low-phase noise RF sources without the need for long fiber lengths. The re-
moval of the long fiber cavity will reduce the sensitivity of the system to en-
vironmental fluctuations and increase the nonoscillating sidemode frequency
spacing. The high-Q cavity will also provide higher resolution for spectrum
analysis, channelization, and the identification of the center frequency of RF
signals.
Finally, the ultrawide frequency, high-power photodiode will benefit
the transmission of RF signals. As mentioned above, wireless technologies
will need to be able to transmit higher-frequency RF signals. The ability to
feed an antenna with a higher-frequency signal at the top of a tall cellular
tower will require photonics. By placing a photodiode as the feed to the an-
tenna, the system will be able to accommodate these signals. Also, the ul-
trawide frequency photodiodes will also be able to nondestructively probe
integrated circuits as well as biological samples.
The far future of RF photonics rests in IPCs. The reduction in size
will allow these subsystems to become more ubiquitous—and less expensive.
IPCs are also finally beginning to break out to more real-world uses. Multiple
foundries exist around the world where a user can design a circuit using stan-
dard process design kits (PDKs). In fact, software for designing into multiple
foundries that use both silicon and semiconductor materials is available [70].
The ability to design into these various foundries opens new opportunities
for IPCs.
The application space for RF photonics will continue to grow in ways
that cannot be completely predicted. What is known for sure is the continued
need to access higher frequencies and utilize larger bandwidths. RF photonics
 Conclusions 193

is well suited to address these needs and will continue to improve its perfor-
mance and reliability. Along with continued work in IPCs, the next 20 years
should see a large expansion in the use of RF photonic technologies.

References
[1] https://goochandhousego.com/product-categories/dfb-lasers-modules/.
[2] http://mortonphotonics.com/ultra_low_noise_lasers.
[3] http://www.rio-lasers.com/_products/planex.html.
[4] http://www.menlosystems.com/products/.
[5] http://www.eospace.com/.
[6] http://www.phasesensitiveinc.com/components/optical-modulators/.
[7] Macario, J., et al., “Full Spectrum Millimeter-Wave Modulation,” Optics Express, Vol.
20, No. 21, 2012, pp. 23623–23629.
[8] Juodawlkis, P. W., et al., “InGaAsP/InP Quantum-Well Electrorefractive Modulators
with Subvolt Vpi,” Proc. of SPIE, Vol. 5435, 2004, pp. 53–63.
[9] http://www.discoverysemi.com/
[10] Li, Z., et al., “High-Power High-Linearity Flip-Chip Bonded Modified Unitraveling
Carrier Photodiode,” Optics Express, Vol. 19, No. 26, 2011, pp. B385–B390.
[11] Ito, H., et al., “High-Speed and High-Output InP-InGaAs Unitraveling-Carrier
Photodiodes,” Journal of Selected Topics in Quantum Electronics, Vol. 10, No. 4, 2004,
pp. 709–727.
[12] https://www.finisar.com/optical-components
[13] https://www.nuphoton.com/product-category/amplifiers/pulsed-amplifiers/edfa-cw/.
[14] http://www.teraxion.com/en/tfn.
[15] http://www.pharad.com/rf-photonic-transceivers.html.
[16] http://www.photonicsinc.com/photonic_links.html.
[17] http://linphotonics.com/products/.
[18] http://www.b2bphotonics.com/products/.
[19] Diehl, J. F., et al., “Microwave Photonic Delay Line Signal Processing,” Applied Optics,
Vol. 54, No. 31, 2015, pp. F35–F41.
[20] Urick, V. J., et al., “Design and Performance Of Ka-Band Fiber-Optic Delay Lines,”
NRL Memorandum Report, NRL/MR/5650-12-9456, 2012.
[21] Singley, J. M., et al., “Design and Performance of a 560-microsecond Ku-Band Binary
Fiber-Optic Delay Line,” NRL Memorandum Report, NRL/MR/5650-15-9545, 2015.
194 Applications of Modern RF Photonics

[22] Yap, D., J. H. Schaffner, and O. M. Efimov. “Tunable Photonic RF Circulator for
Simultaneous Transmit and Receive,” United States Patent No. 9,250,452, 2016.
[23] http://www.photonicsinc.com/tirx.html.
[24] http://www.ofsoptics.com/fiber-bragg-gratings.html
[25] https://www.newport.com/c/fiber-optic-components.
[26] https://www.ntt-electronics.com/en/products/photonics/awg_mul_d.html.
[27] https://www.thorlabs.com/
[28] http://www.ramphotonics.com/products/.
[29] Devgan, P., V. J. Urick, and K. J. Williams, “Multimode Optoelectronic Oscillator,”
United States Patent 8,824,901, 2014.
[30] http://www.oewaves.com/products.
[31] Li, W., and J. Yao, “A Wideband Frequency Tunable Optoelectronic Oscillator
Incorporating a Tunable Microwave Photonic Filter Based on Phase-Modulation
to Intensity-Modulation Conversion Using a Phase-Shifted Fiber Bragg Grating,”
Transactions on Microwave Theory and Techniques, Vol. 60, No. 6, 2012, pp. 1735–
1742.
[32] Suelzer, J. S., et al., “Tunable, Low-Phase-Noise Microwave Signals from an Optically
Injected Semiconductor Laser with Opto-Electronic Feedback,” Optics Letters, Vol. 42,
No. 16, 2017, pp. 3181–3184.
[33] Jones, D., et al., “Carrier Envelope Phase Control of Femtosecond Mode-Locked Lasers
and Direct Optical Frequency Synthesis,” Physics Faculty Contributions, 86, 2000.
[34] http://www.menlosystems.com/.
[35] http://www.toptica.com/products/frequency-combs/.
[36] Schlottau, F., et al., “Spectral Hole Burning for Wideband, High-Resolution Radio-
Frequency Spectrum Analysis,” Optics Letters, Vol. 30, No. 22, 2005, pp. 3003–3005.
[37] https://www.s2corporation.com/.
[38] Stead, M. R., W. Zhou, and M.-C. Li, “Using Dispersion in a Fiber-Optic Loop to
Perform Time Domain Analogue RF Signal Auto-Correlation,” Proc. of Optical Fiber
Communication Conference and Exposition, 2011.
[39] Rogge, M. S., V. J. Urick, and F. Bucholtz, “Analysis of an Optical Channelization
Technique for Microwave Applications,” NRL Memorandum Report, NRL/MR/5652-
07-9061, 2007.
[40] Xie, X., et al., “Broadband Photonic Radio-Frequency Channelization Based on a 39-
GHz Optical Frequency Comb,” Photonics Technology Letters, Vol. 24, No. 8, 2012, pp.
661–663.
[41] Hunter, D. B., L. G. Edvell, and M. A. Englund. “Wideband Microwave Photonic
Channelised Receiver,” Proc. of International Topical Meeting on Microwave Photonics,
2005.
 Conclusions 195

[42] Agarwal, A., et al., “Photonically-Enabled RF Front-End for Wideband Flexible Down-
Conversion,” Optics Express, Vol. 25, No. 7, 2017, pp. 7338–7348.
[43] Fudge, G. L., et al., “A Nyquist Folding Analog-to-Information Receiver,” Proc. of
Signals, Systems and Computers, 2008.
[44] McKinney, J. D., and V. J. Urick, “Radio-Frequency Down-Conversion Via Sampled
Analog Optical Links,” NRL Memorandum Report, NRL/MR/5650-10-9275, 2010.
[45] Khilo, A., et al., “Photonic ADC: Overcoming the Bottleneck of Electronic Jitter,”
Optics Express, Vol. 20 No. 4, 2012, pp. 4454–4469.
[46] www.dasphotonics.com.
[47] McDermitt, C. S., et al., “Performance of 16-Channel, Photonic, Phased-Array Antenna
Calibration System,” Electronics Letters, Vol. 45, No. 24, 2009, pp. 1249–1250.
[48] Godinez, M. E., et al., “RF Characterization of Zero-Biased Photodiodes,” Journal of
Lightwave Technology, Vol. 26, No. 24, 2008, pp. 3829–3834.
[49] Diehl, J. F., et al., “The Performance of a Fiber-Optic Link Using Unbiased Balanced
Photodiodes for Antenna Array Calibration,” Transactions on Microwave Theory and
Techniques, Vol. 60, No. 1, 2012, pp. 195–200.
[50] Tzanidis, I., K. Sertel, and J. L. Volakis, “UWB Low-Profile Tightly Coupled Dipole
Array with Integrated Balun and Edge Terminations,” Transactions on Antennas and
Propagation, Vol. 61, No. 6, 2013, pp. 3017–3025.
[51] Shi, S., et al., “Ultrawideband Optically Fed Tightly Coupled Phased Array,” Journal of
Lightwave Technology, Vol. 33, No. 23, 2015, pp. 4781–4790.
[52] Minotani, T., A. Hirata, and T. Nagatsuma, “A Broadband 120-GHz Schottky-Diode
Receiver for 10-Gbit/s Wireless Links,” IEICE Transactions on Electronics, Vol. 86 No.
8, 2003, pp. 1501–1505.
[53] Cross, A. S., et al., “High-Power Flip-Chip Mounted Photodiode Array,” Optics Express,
Vol. 21, No. 8, 2013, pp. 9967–9973.
[54] Nagel, M., A. Michalski, and H. Kurz, “Contact-Free Fault Location And Imaging
with On-Chip Terahertz Time-Domain Reflectometry,” Optics Express, Vol. 19, No. 13,
2011, pp. 12509–12514.
[55] Cai, Y., et al., “Electro Optical Terahertz Pulse Reflectometry—An Innovative Fault
Isolation Tool,” Proc. of Electronic Components and Technology Conference, 2010.
[56] Nagatsuma, T., et al., “Millimeter-Wave and Terahertz-Wave Applications Enabled By
Photonics,” Journal of Quantum Electronics, Vol. 52, No. 1, 2016, pp. 1–12.
[57] Ghelfi, P., et al., “A Fully Photonics-Based Coherent Radar System,” Nature, Vol. 507,
No. 7492, 2014, pp. 341–345.
[58] McKinney, J., “Technology: Photonics Illuminates the Future of Radar,” Nature, Vol.
507, No. 7492, 2014, pp. 310–312.
[59] Callan, K. E., et al., “Broadband Chaos Generated by an Optoelectronic Oscillator,”
Physical Review Letters, Vol. 104, No. 11, 2010, pp. 113901–4.
196 Applications of Modern RF Photonics

[60] Goodman, J. W., Introduction to Fourier optics, Englewood, CO: Roberts and Company,
2005.
[61] Weiner, A. M., “Femtosecond Pulse Shaping Using Spatial Light Modulators,” Review
of Scientific Instruments, Vol. 71, No. 5, 2000, pp. 1929–1960.
[62] Gorju, G., et al., “10-GHz Bandwidth RF Spectral Analyzer with MHz Resolution
Based on Spectral Hole Burning in Tm/sup 3+: YAG,” Photonics Technology Letters, Vol.
17, No. 11, 2005, pp. 2385–2387.
[63] Schnébelin, C. and H. G. de Chatellus, “Agile Photonic Fractional Fourier
Transformation of Optical and RF Signals,” Optica, Vol. 4, No. 8, 2017, pp. 907910.
[64] www.millitech.com/pdfs/Radiometer.pdf.
[65] Zhang, Y., “Application of Millimeter-Wave Photonics Technology in Passive
Millimeter-Wave Imaging,” Proc. of SPIE, Vol. 7854, 2010, pp. 785436B1–785436B6.
[66] Schuetz, C. A., et al., “Radiometric Millimeter-Wave Detection Via Optical
Upconversion and Carrier Suppression,” Transactions on Microwave Theory and
Techniques, Vol. 53, No. 5, 2005, pp. 1732–1738.
[67] Schuetz, C. A., et al., “Sparse Aperture Millimeter-Wave Imaging Using Optical
Detection and Correlation Techniques,” Proc. of SPIE, Vol. 6548, 2007, pp. 65480B1–
65480B8, 2007.
[68] http://www.phasesensitiveinc.com/systems/video-rate-mmw-imager/.
[69] Ciminelli, C., et al., “Photonic Technologies For Angular Velocity Sensing,” Advances
in Optics and Photonics, Vol. 2, No. 3, 2010, pp. 370–404.
[70] http://www.phoenixbv.com/.
List of Acronyms
3R reamplify, retime, and reshape

ACF amplitude comparison function

ADC analog-to-digital converter

AIM Photonics American Institute for Manufacturing Photonics

AMP amplifier

AMZ asymmetric interferometer

AWG arrayed-waveguide grating

BER bit error rate

BERT bit error rate tester

BPF bandpass filter

CDR compression dynamic range

CMOS complementary metal oxide semiconductor

CW continuous wave

197
198 Applications of Modern RF Photonics

dB decibel

dBc decibel relative to carrier

dBm decibel relative to 1 mW

DC direct current

DEMUX demultiplexer

DET detector

DFB distributed feedback

DPMZM dual-parallel Mach Zehnder modulator

EAM electro-absorption modulator

EBF electronic bandpass filter

EDFA erbium-doped fiber amplifier

ENOB effective number of bits

EOM electro-optic modulator

FBG fiber-Bragg grating

FEC forward error correction

FIR finite impulse response

FP Fabry-Perot

FrFT fractional Fourier transform

Gbps gigabits per second

GRIN gradient index

HF high-frequency
 List of Acronyms 199

IF intermediate-frequency

IF AMP intermediate-frequency amplifier

IFM instantaneous-frequency measurement

IIR infinite-impulse response

IMD_2 second-order intermodulation

IMD_3 third-order intermodulation

IMDD intensity-modulated direct detection

JePPIX Joint European Platform for Photonic Integration of

Components and Circuits

km kilometer

LiNbO3 lithium niobate

LO local oscillator

LPF low-pass filter

MDS minimum detectable signal

MISF maximum-input spur-free

MM-OEO multimode optoelectronic oscillator

MOD modulator

MOSIS metal oxide semiconductor implementation service

MPW multiproject wafer

MUX multiplexer

mW milliwatt
200 Applications of Modern RF Photonics

MZM Mach-Zehnder modulator

NF noise figure

OA optical amplifier

OC-192 optical channel 192 (9.953 Gbps)

OEO optoelectronic oscillator

PD photodetector

PDK process design kit

PIN positive-intrinsic-negative

PN positive-negative

PON passive optical network

PSD power spectral density

RF radio frequency

RIN relative intensity noise

SBS stimulated Brillouin scattering

SFDR spur-free dynamic range

SHB spectral hole burning

SLM spatial light modulator

SNR signal-to-noise ratio

SONET synchronous optical network

SSB single-sideband

Tbps terabits per second

 List of Acronyms 201

TBPF tunable bandpass filter

TFLN thin-film lithium niobate

THz terahertz

UHF ultrahigh frequency

VCO voltage-controlled oscillator

VHF very high frequency

WDM wavelength demultiplexer/division multiplex

About the Author
Preetpaul Devgan is currently a senior electrical engineer at the Air Force
Research Laboratory (AFRL). Prior to joining AFRL, he worked at the Naval
Research Laboratory, after receiving his Ph.D. from Northwestern University.
Prior to that, he worked at Lucent Technologies in the optical networking
group, which he joined after receiving his B.S. and M.S. from Purdue Uni-
versity. His current research interests include RF photonic systems for signal-
processing applications, specifically optoelectronic oscillators, novel modula-
tion formats, frequency conversion, and identification.
Dr. Devgan is a senior member of IEEE and the winner of the 2016
IEEE Dayton Section Photonics Society award. Previously, he received tech-
nology transfer awards from both the NRL and the AFRL Sensors direc-
torate. Devgan attended Northwestern University under a National Science
Foundation IGERT fellowship, where he was recognized as a finalist for the
2005 OSA New Focus/Bookham Student award. Prior to that, he held the
office of president of the student chapter of Eta Kappa Nu at Purdue, during
which time the chapter was recognized with the best student chapter award.
Dr. Devgan has published over 60 peer-reviewed journal and conference pub-
lications (including multiple invited papers) and holds four patents.

203
Index
All-photonic gain, 93–94 parallel, 156
Amplitude comparison function (ACF) photonic, block diagram, 155, 156, 158
defined, 127 photonic quantization, 158–59
expression as ratio, 128 RF photonic, 153–55
frequency recovery and, 128 sampling for, 85
as function of RF frequency, 129 sampling jitter, 155
high slope of, 130 time-stretch, block diagram, 157
slope illustration, 129–30 timing jitter and, 154
Amplitude noise Application space, 192–93
PSD and, 87 Arbitrary-transient waveform generation,
spectra, 88 159–61
Analog delay lines Arrayed-waveguide grating (AWG), 160, 161
advancements in performance, 59–79 Asymmetric Mach-Zehnder interferometer
architectures of, 40–49 (AMZI), 114
component performance, 49–55 Atmospheric attenuation, 28, 29
conclusion, 55
demonstration architecture, 35 Balanced photodetectors, 63
overview, 3–4 Bandwidth capacity, 21–22
as range emulator, 35 BER tester (BERT), 23
RF metrics, 36–40 Bias voltage, 64
Analog photonics, 22–27 Binary digital data streams
Analog-to-digital converters (ADCs) eye diagram, 24
on antenna, 30 ideal versus real, 23
bandwidth, 155 Bit error rate (BER), 23, 25
clock synchronization, 157 Bit rate, 23
discrete-based RF photonic subsystems, Bragg gratings, 110, 113
187
noise floor, 153 Cancel the even-order distortion, 63–64
nonlinear responses of, 154 Channelization, 187
number used in parallel structure, 153

205
206 Applications of Modern RF Photonics

Chirped fiber Bragg grating, 115 RF photonics for downconversion,

Chromatic dispersion 148–50
defined,, 53 subsystem, 187
fundamental and second harmonic Dual-input MZM, 74
response due to, 54 Dual-laser system, 74
in limiting frequency bandwidth, 74 Dual-loop OEO
Clock skew, 85 diagram, 91
Clock synchronization phase noise, 92
analog-to-digital converters (ADCs), Dual-parallel MZM (DP-MZM)
157 bias control points, 79
demonstration, 96 defined, 78
OEO for, 94–98 in OEO feedback cavity, 152
performance, 95 phase control, 152
Coaxial cables phase of second harmonic
fiber-optic cables advantages over, 22 photocurrent, 79
loss, 20–21 SSB modulation from, 78
Coaxial connectors, maximum frequency, Dual-sideband modulation, 109
20
Compression dynamic range (CDR) Effective number of bits (ENOB), 154, 156
defined, 26, 39 Electroabsorption modulator (EAM)
as key metric, 25 defined, 44
one-dB, 40 drawbacks, 44–45
SFDR and, 39 link with transfer function, 45
Continuous wave (CW) laser, 89 Electronic quantization, 155–57
Erbium-doped fiber amplifier (EDFA), 18,
DC photocurrent 22, 51
RF gain and, 60 best operating point, 68
RIN as function of, 51 as chosen amplifier, 185
Digital photonics, 22–27 in compression, 69
Direct modulation, 41–42 noise penalty, 67–68
Discrete-based RF photonic subsystems operating in saturation, 68–69
ADCs, 187 optical carrier/sidebands before/after,
circulator, 185–86 70
downconversion, 187 response times, 67
oscillator, 186 External modulation
signal identification, 186–87 intensity, 44–48
signal separation, 186 phase, 42–44
Dispersion Eye diagram, 24
chromatic, 53, 54, 74
for instantaneous frequency Fiber-Bragg grating (FBG), 110, 161
measurement, 129–32 Fiber-optic cables
in optical fiber, 64–65 analog versus digital photonics, 22–27
single-laser system with, 131 bandwidth capacity, 21–22
two-laser system with, 130 coaxial cables versus, 22
Distributed feed (DFB) semiconductor- losses, 19–20
based laser, 50 for phased arrays, 18
Downconversion Fifth-generation systems, 191
need for, 146–47
 Index 207

Finite impulse response (FIR) filters fabrication, 172

degrading performance of, 114 further work in analog applications,
discrete convolution sum, 112 177–78
as flexible and adaptable, 115 MM-OEO and, 174
implementation, 113, 114 ongoing debate, 169
for instantaneous frequency optical amplifier, 170–71
measurement, 133–34 optical filters, 171
realization of, 112 photodetector, 172
RF photonics and, 111–15 RF circulator, 173
Forward error correction (FEC), 27 semiconductor-based, 169
Fourier processing, 190 use of, 170
Free spectral range (FSR), 109 Integrated photonic FP filter, 171
Frequency-chirped RF waveforms, 161 Integrated photonic ring resonator filter,
Frequency measurement 172
combinations of different methods, Integrated RF photonics
132–33 advancements in, 167–78
dispersion for, 129–32 conclusion, 178
FIR and IIR filters for, 133–34 fundamentals, 168–70
with multimode photonic systems, IPCs, 169–78
134–42 overview, 167–68
photonic filters for, 127–29 PICs, 168–69
Future work, 191–93 Intensity-modulated direct detection
(IMDD) link
Gradient index (GRIN), 61 defined, 47
laser quantification, 49
Harmonic distortion, 38 performance, 59
High-frequency (HF) radio signals, 29 as preferred solution, 59
High-Q photonic cavity, 192 Intensity modulation
Hybrid fiber coax (HFC), 18 EAM, 44–45
MZM, 45–48
Intermodulation, 37, 38
Infinite impulse response (IIR), 111,
133–34
Injection-locked OEO Joint European Platform for Photonic
defined, 95–96 Integration of Components and
frequency and time domain plot, 98 Circuits (JePPIX), 177
measured plot of phase noise, 97
phase noise plot, 96 Lasers
See also Optoelectronic oscillator bias point and, 41
(OEO) DFB semiconductor-based, 50
Instantaneous frequency measurement as link requirement, 40
dispersion for, 129–32 output power, 41, 49
FIR and IIR filters for, 133–34 relative intensity noise (RIN), 49–50,
photonic filters for, 127–29 51
Integrated photonic circuits (IPCs) slope efficiency, 42
applications (other), 175–77 tunable, 132
applications (RF photonics), 173–75 Low-biased MZM, 71–74
components of, 170–72 Low phase noise photonic clocks, 86
defined, 169
208 Applications of Modern RF Photonics

Mach Zehnder interferometer-based Off-quadrature biasing, 69–71

modulator (MZM) Optical amplifier
defined, 45–46 increase in power, 66–67
dual-input, 74 IPC, 170–71
dual-output, 72–73 noise and, 67
dual-parallel (DP-MZM), 78–79, 152 performance improvements, 66–69
IMDD link, 47 See also Erbium-doped fiber amplifier
in improving RF metrics, 69 (EDFA)
link with transfer function, 46 Optical comb source, 161
low-biased, 71–74 Optical fiber
operating point, 70 chromatic dispersion, 53, 74
optical input, 117 dispersion in, 64–65
optical output, 117 optical power, 52
photonic links, 185 output and reflected power, 52
as preferred solution, 46, 70 parallel paths, 63
push-pull, 77 performance improvements, 64–66
quadrature-biased, 73 span with isolators in-between, 65
in RF gain, 48 Optical filters
single-output, 72 Bragg grating, 110
sinusoidal transfer function, 46 complementary systems, 129
transfer function, 117 IPC, 171–72
wavelength-dependent coefficient, 72, realization of, 110
75 with SSB and DSB, 109
Master-slave OEO, 93 two-laser system with, 128
Max input spur-free (MISF) power, 25, 35 Optically injected OEO (OIL-OEO), 100
Metal oxide semiconductor implementation Optical modulator
service (MO-SIS), 177 low biasing with dual wavelengths,
Millimeter-wave radiometer, 190–91 71–74
Minimum detectable signal (MDS), 25 off-quadrature biasing of, 69–71
Multiloop OEO, 90–93 performance improvements, 69–79
Multimode OEO (MM-OEO) single sideband modulation, 76–79
defined, 134, 135 Optical power
first demonstration, 136 increasing, 60
injection-locking method, 139 photodetector and, 55
IPCs and, 174 reflected, 52, 66
maximum input power, 136 single-sideband (SSB) modulation
new configuration, 136–37 and, 77
phase modulator, 140 threshold, 52
in selective amplification, 136, 137 Optoelectronic oscillator (OEO)
selectivity, 137 advantages of, 90
sensitivity, 136 with all-photonic gain, 93–94
setup, 135 basic, diagram, 89
two-laser, 137 for clock recovery, 95
See also Optoelectronic oscillator clock synchronization using, 94–98
(OEO) continuous wave (CW) laser, 89
defined, 88
Noise penalty link, 68 driven into chaotic state, 190
Noise transfer, 95 free-running, 95, 96, 97
 Index 209

frequency and time domain plot, 98 as measure, 39–40

initial design disadvantages, 90 optoelectronic oscillator (OEO), 89
injection-locked, 95–98 photodetectors, 62
low loss fiber-optical delay, 89 PSD and, 87
master-slave, 93 timing jitter and, 86–88
multiloop, 90–93 use of, 40
multimode (MM-OEO), 134–42 Phase shifters, 113, 175
optically injected (OIL-OEO), 100 Photoconductive probe, 189
phase noise, 89 Photodetectors
Q factor, 89 balanced, 63
RF amplifiers in, 93 cancel the even-order distortion and,
short-loop, 92–93 63–64
signal amplitude, 95 nonlinearities cancellation, 78–79
timing jitter, 89 nonlinear responses, 61
Organization, this book, 13–14 optical power, 55
Oscillators performance improvements in, 61–64
common clock signal, 85 phase noise, 61
discrete-based RF photonic response of, 62
subsystems, 186 responsivity, 55
need for, 85–86 second-order distortion, 62
RF photonics and, 85–102 specialty, 78
two laser sources, 98–102 Photonic-based gyroscopes, 190–91
See also Optoelectronic oscillator Photonic-based phase shifter, 175
(OEO) Photonic downconverter
Output intercept point (OIP), 38–39 advancements in, 150–52
advantages, 148
Parallel modulator photonic RF architecture development, 150
downconverter, 151 block diagram, 149
Passive optical networks (PONs), 50 defined, 148
Performance improvements limitations, 148–50
analog delay line, 59 parallel modulator, 151
conclusion, 79 See also Downconversion
in optical amplifier, 66–69 Photonic fed array, 188–89
in optical fiber, 64–66 Photonic integrated circuits (PICs), 168–69
in optical modulator, 69–79 Photonics sampling
in photodetector, 61 combined with electronic
photonic components and, 60–61 quantization, 155–57
Phase control, 175–76 quantization and, 158–59
Phase modulation Power spectral density (PSD), 87, 88
advantages, 42 Push-pull MZM, 77
defined, 42
drawbacks, 42–43 Q factor, 25, 89
link architecture, 43 Quadrature-biased MZM, 69–71, 73
Phase modulator, 140 Quantization, photonics sampling and,
Phase noise 158–59
clock, 85
defined, 39, 86–87 Radio frequency (RF) photonics. See RF
dual-loop OEO, 92 photonics
210 Applications of Modern RF Photonics

Rayleigh scattering limit, 19 minimum, 49–50

Reflected optical power, 52, 66 performance of system, 37
Relative intensity noise (RIN) RF photonic ADCs, 153–55
defined, 49 RF photonic-based coherent radar system,
as function of DC photocurrent, 51 189–90
minimum NF for, 49–50 RF photonic links
for two different types of lasers, 50 component requirements, 40–41
Responsivity, 55 directly modulated, 41
RF circulators example of, 34
advantages with photonics, 118–19 IMDD, 47
discrete-based RF photonic MZM-based, 185
subsystems, 185–86 phase modulated, 43–44
illustrated, 115 for in-situ antenna calibration, 188
IPC, 173 use of, 33
photonic-based, 116–19 RF photonics
in signal isolation, 115 advanced sensors, 190–91
in subsystem, 108 advantage of, 19–22
wideband, 116 alternative systems using, 188–91
RF downconverters applications of IPCs to, 173–75
advancements in, 150–52 application space, 192–93
block diagram, 147 arbitrary-transient waveform
components of, 146–47 generation using, 159–61
photonics-based, 148–50 brief overview of, 184–85
RF fading brief overview of history, 18–19
defined, 53 conclusions, 183–93
as function of frequency, 54 current needs for, 28–29
single-sideband (SSB) modulation defined, 17
and, 76–77 discrete-based subsystems, 185–87
RF gain for downconversion, 148–50
DC photocurrent and, 60 fiber core, 19
defined, 36 field progression, 183
directly modulated link, 41 FIR filters using, 111–15
as function of DC bias voltage, 71 future work in, 191–93
as output RF power to input RF growth of, 18
power, 141 historical advancements in, 18–19
for single pass in OEO, 94 integrated, 167–78
RF metrics introduction to applications, 17
definitions of, 36–40 MZM role in, 48
improvements from two laser low bias oscillators utilizing, 85–102
system, 75 for separation of signals, 109–10
in logarithmic scale, 60 signal identification with, 123–42
performance improvement, 60 signal isolation with, 107–19
plot, 26 signal processing with, 145
types of, 25 for spectrum analysis, 125–27
RF NF RF signal separation subsystem, 108
cascaded, 67–68 RF spectrum, 107
defined, 36
 Index 211

Sampling jitter, 155 RF photonics for downconversion,

Scanning electrical spectrum analyzer, 125 148–50
Second-order distortion sampling combined with electronic
chromatic dispersion and, 53 quantization, 155–57
dual laser system and, 74 Signal-to-noise (SNR) ratio, 29, 36, 86
photodetectors, 62 Single-sideband (SSB) modulation
system for canceling, 75 in canceling photodetector
Semiconductor-based IPCs, 169 nonlinearities, 78–79
Separation of signals optical filter with, 109
discrete-based RF photonic optical power and, 77
subsystems, 186 RF fading and, 76–77
RF photonics for, 109–10 RF sideband and, 77
subsystem, 108 SONET, 27
Short-loop OEO, 92–93 Spatial light modulator (SLM), 160, 161
Signal identification Spectral hole burning (SHB) spectrum
amplitude comparison function analyzers, 126
(ACF), 127–29 Spectrum analysis, 125–27, 187
combination method of frequency Spectrum analyzers
measurement, 132–33 channelization and, 187
conclusion, 142 removal of scanning filter and, 127
demonstration architecture, 124 scanning electrical, 125
discrete-based RF photonic spectral hole burning (SHB), 126
subsystems, 186–87 Spur-free dynamic range (SFDR)
instantaneous frequency measurement, calculation, 38
127–32, 133–34 defined, 37
need for, 124–25 MISF and, 25
photonic filters, 127–29 range of input powers, 26
with RF photonics, 123–42 RF NF and, 35
spectrum analysis and, 125–27 Stimulated Brillouin scattering (SBS)
subsystem, 123 effect, 52, 65
Signal isolation spectra, 65
along common path, 115–19 threshold, 52, 53, 64
conclusion, 119
demonstration architecture, 108 Temperature tuning, 100–101
need for, 107–9 TFLN, 170
with RF photonics, 107–19 Thermal coefficient of delay (TCD), 22
Signal processing Third-order limited systems, 38
ADC, 153–55 Time delays, 176
advancements in downconverters, 150 Time grating, 115
analog applications, 145 Timing jitter
arbitrary-transient waveform analog-to-digital converters (ADCs)
generation, 159–61 and, 154
conclusion, 162 optoelectronic oscillator (OEO), 89
defined, 17 phase noise and, 86–88
need for downconversion, 146 True time delay, 29
photonics sampling and quantization, True time-delay circuit, 177
158–59 Tunable lasers, 132
with RF photonics, 145–62 Two-laser MM-OEO, 137
212 Applications of Modern RF Photonics

Two laser oscillators Ultrahigh frequency (UHF), 108

beat signal, 98–99
defined, 98–99 Very high frequency (VHF), 108
locking phases, 99 Voltage levels, 23
master laser and slave laser, 100–101
optically injected OEO (OIL-OEO), Waveform generation, 161
100 Wavelength division multiplexing (WDM),
polarization state, 99 21, 156, 192
setup, 100 Wideband frequency analyzer, 124
switching speed, 102
temperature tuning, 100–101
Two-laser system
with dispersion, 130
with optical filter, 128