Content uploaded by Andrew Metcalf
Author content
All content in this area was uploaded by Andrew Metcalf on Jul 07, 2015
Content may be subject to copyright.
3478 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
Comb-Based RF Photonic Filters Based
on Interferometric Configuration
and Balanced Detection
Hyoung-Jun Kim, Daniel E. Leaird, Senior Member, IEEE, Andrew J. Metcalf, Student Member, IEEE,
and Andrew M. Weiner, Fellow, IEEE
Abstract—We demonstrate a novel technique to improve radio
frequency (RF) performance such as RF gain and noise figure (NF)
for comb-based RF photonic filters. While conventional RF pho-
tonic links use a dual-output modulator and balanced detection,
this RF photonic filter utilizes an interferometric configuration
with double sideband suppressed carrier modulation and balanced
detection. This technique can simultaneously provide filter tunabil-
ity, 6-dB RF gain increase, and noise cancellation. The RF gain and
NF of the RF photonic filter are improved to approximately 0 and
24 dB, respectively. With the improved RF performance, we per-
form the tuning of the filter center frequencies from 2 to 8 GHz
with no baseband filter response (<−38 dB), no RF power fad-
ing, while maintaining good filter shape (sidelobe suppression and
stopband attenuation >32 dB).
Index Terms—Finite impulse response filters, microwave pho-
tonics, optical combs, optical processing, programmable filters,
tunable filters.
I. INTRODUCTION
RADIO frequency (RF) filtering is an essential part of RF
systems used in wireless communication, imaging, and
sensing applications. Recently, with the demand for greater vol-
ume in broadband wireless service as well as increased data
demands in high resolution imaging and sensing applications,
the need for greater RF bandwidth has rapidly increased. In addi-
tion, the increased RF complexity required in applications such
as software defined radio, cognitive radio, and multi-standard
radio, can be simplified by using reconfigurable and multi-
functional RF filters. However, it is difficult to tune traditional
RF filter technologies rapidly over a large RF bandwidth and
even more challenging to reconfigure them for different func-
tionalities. As one example, yttrium iron garnet (YIG) filters
have been widely used in various RF systems and provide wide
tuning range, good selectivity, and good linearity. However, the
tuning speed of the YIG filters is limited to the millisecond
Manuscript received January 15, 2014; revised May 7, 2014 and April 1,
2014; accepted May 11, 2014. Date of publication May 21, 2014; date of current
version September 1, 2014. This work was supported in part by the Office of
the Assistant Secretary of Defense for Research and Engineering under the
National Security Science and Engineering Faculty Fellowship program under
Grant N00244-09-1-0068 from the Naval Postgraduate School. Any opinion,
findings, and conclusions or recommendations expressed in this publication are
those of the authors and do not necessarily reflect the views of the sponsors.
The authors are with the School of Electrical and Computer Engineering,
Purdue University, West Lafayette, IN 47907-2035, USA (e-mail: sjun27@
purdue.edu; leaird@purdue.edu; metcalfa@purdue.edu; amw@purdue.edu).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JLT.2014.2326410
time scale [1]. The introduction of microelectromechanical sys-
tems (MEMS) variable capacitor approaches has led to recent
advances in tunable RF filters [2]–[4]. The fundamental tuning
speed of RF MEMS devices is characterized in [2] as limited
to approximately 1–300 μs. Filters based on solid-state diodes
have fundamentally higher tuning speeds but suffer from in-
creased nonlinearity, loss, and power consumption [2]. In either
of the latter technologies, challenges in controlling the coupled
response of multiple resonances to maintain high filter selec-
tivity practically constrain tuning to speeds much slower than
the fundamental limits. Recent research has begun to go beyond
frequency tuning to focus on reconfigurability. In one exam-
ple, a second-order filter based on coupled evanescent-mode
cavity resonators was used to demonstrate a bandstop-to-all-
pass reconfigurable filter [5]. Reconfiguration was accomplished
via piezoelectric actuation of a flexible copper membrane.
However, research into generally reconfigurable filters remains
in its early stages, and rapid reconfiguration has not been
explored.
RF photonics offers potential to implement filters that over-
come these limitations. One approach utilizes a tapped delay
line scheme [6]. Here, an RF-modulated optical signal is split to
multiple branches which act as filter taps. Each of the branches
has an optical attenuator and an optical delay line used to control
the amplitude and phase of the RF signals, respectively. The split
optical signals are then combined and detected by a photodiode
to produce the RF waveform. The filter transfer function in this
scheme can be characterized by a finite impulse response, al-
lowing for the design of arbitrary amplitude filters. Furthermore,
it is easy to process the wide-band RF signals due to a broad
operational bandwidth of optical components such as optical
attenuators and optical delay lines. However, it is not easy to
implement these schemes with large number of filter taps. Other
approaches utilize multi-wavelength sources [6]–[20]. In these
schemes, RF-modulated multi-wavelength signals are transmit-
ted through a single dispersive fiber or chirped fiber Bragg grat-
ing. Then differential delays between the multi-wavelength sig-
nals (i.e., filter taps) are applied through fiber dispersion. The
amplitudes and delays of the filter taps can be controlled by
adjusting the optical powers of the multi-wavelength signals
and the length of the single dispersive fiber, respectively. Thus,
this scheme also possesses a finite impulse response filter re-
sponse and allows easy scaling of the filter taps by using various
multi-wavelength sources. The above mentioned multi-channel
sources have been realized in various ways, including an array of
0733-8724 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
KIM et al.: COMB-BASED RF PHOTONIC FILTERS BASED ON INTERFEROMETRIC CONFIGURATION AND BALANCED DETECTION 3479
continuous-wave laser sources [7], [8], spectrally sliced broad-
band light sources [9], [10], and optical frequency combs such
as mode-locked lasers [11] and electro-optically (EO) gener-
ated combs [12]–[20]. Among these sources, the EO-generated
combs are very attractive for RF photonic filtering and other RF
photonic signal processing due to their spectral flatness, high
coherence, tunable repetition rate, and good stability [21], [22].
Our group has previously demonstrated reconfigurable RF pho-
tonic filters using an EO-generated comb and pulse shaper in
an interferometric configuration [12]–[16]. Gaussian bandpass
filters with fast (40 ns) tunability of the passband center fre-
quency, high sidelobe suppression radio (>60 dB), and high
stopband attenuation (>70 dB) were demonstrated in [15]. Re-
configurable flat-topped filters were reported in [14]. In [16] and
[18], reconfigurable phase filters were implemented and applied
for compression of wideband RF chirp signals. Using a similar
comb-based RF filter configuration but without the interferome-
ter, rapid (20 ns) bandwidth reconfiguration was demonstrated
in [19], and simultaneous tunable bandpass filtering and down-
conversion was reported in [20]. Thus, such comb-based RF
photonic filters offer a degree of reconfigurability far beyond
what is currently available from other technologies, as well po-
tential for extremely fast tuning and reconfigurability.
To date, however, such RF photonic filters have typically per-
formed poorly in terms of RF gain and noise. In [15] for exam-
ple, the typical RF gain was limited to approximately −40 dB
at 0.5 mA photocurrent. In this and other RF photonic filter
schemes, the output photocurrent can be increased by the use
of erbium-doped fiber amplifiers (EDFAs) to increase the RF
gain. However, the EDFA also amplifies the intensity noise of
the comb sources and generates amplified spontaneous emission
noise, resulting in a very poor noise figure (NF).
Although RF performance metrics such as RF gain and NF
have been studied extensively in research on conventional RF
photonic links [23]–[25], such metrics have seldom been con-
sidered in the context of reconfigurable RF photonic filtering.
Balanced (or differential) detection is a well known approach to
mitigate noise problems in RF photonic links. Demonstrations
such as those reported in [24] and [25] utilized a dual-output
modulator and balanced photodetector (BPD) to suppress com-
mon mode intensity noise while increasing photocurrent by a
factor of two (increasing RF gain by 6 dB). In [24], the reported
RF gain and NF were >17 dB and <6.5 dB across 6–12 GHz,
respectively.
One interesting recent theoretical paper [26] does analyze
the gain, noise, and intermodulation distortion of RF photonic
links extended to include filtering action in a rather general way.
However, the specific tunable RF photonic filter implementation
of interest in the current work, in which a modulator is embed-
ded in one arm of an optical interferometer, does not appear to
be captured within the configuration assumed in [26]. A few re-
cent experimental papers have reported RF photonic filters using
incoherent broadband light sources and BPDs [27]–[30]. How-
ever, the focus of these papers was on filter reconfigurability;
RF gain and NF were not evaluated.
In this paper we seek to demonstrate substantial experimen-
tal improvement in the RF performance of comb-based RF
photonic filters by incorporating balanced photodetection into
the interferometric configuration our group previously intro-
duced to enable tuning in such filters [12]–[16]. Instead of using
a dual-output modulator, we utilize double sideband suppressed
carrier (DSB-SC) modulation in one arm of the interferometer.
This innovative configuration simultaneously provides 6 dB RF
gain increase, while improving noise performance, supporting
tunability, and eliminating unwanted baseband response. Fur-
thermore, we show that optimizing the input split ratio at the
interferometer improves both the effective optical modulation
index and the NF. Here, the interferometer splits the input into
two paths which we refer to as the delay and modulation paths.
As the fractional power split to the delay path decreases, both
RF gain and total output noise are also decreased. However,
because the total output noise decreases more rapidly than the
RF gain, the NF is reduced. This principle is analogous to low-
biasing, a common noise reduction technique in RF photonic
links [23], [24]; to the best of our knowledge, our work is the
first extension of this principle to RF photonic filtering. Another
feature of our work concerns our treatment of the long disper-
sive fiber needed to realize filtering action. Unlike conventional
RF photonic links, where double-sideband modulation leads to
RF fading upon dispersive propagation, in the interferometric
filter scheme, such RF power fading does not occur [12]. How-
ever, one must address the challenge of providing stable and
closely matched dispersive links to each of the photodetectors.
Here we investigate two schemes which achieve simultaneous
dispersion and time delay matching, namely, bidirectional prop-
agation [28], [29] and polarization multiplexing [30]. Our cur-
rent work focuses on RF gain and NF; the important topics
of nonlinear distortion and dynamic range are left for future
study.
A preliminary description of our results was published in
[31]. Here in a substantially expanded discussion, we report the
full theoretical development for the first time and include a new
experiment which achieves further improvement in RF gain and
NF while simultaneously tuning from 2 to 8 GHz.
The remainder of this paper is organized as follows. Section II
describes the concept of this RF photonic filter and derives its
RF performance metrics such as RF gain and NF. In Section III,
we simulate the effect of the input split coefficient of the inter-
ferometer on the RF performance. In Section IV, we experimen-
tally investigate the RF performance improvements of our BPD
scheme compared to the performance obtained using a single
photodetector (SPD). In an initial static filtering experiment, we
obtain RF gain and NF of −10 and 29 dB. The use of a BPD pro-
vides improvements of 6 and 21 dB, respectively, compared to
SPD operation [31]. We also report new data that reveal signifi-
cant contributions to the noise due to Brillouin scattering effects
in the bidirectional propagation geometry employed. Then, in
a new experiment in which we use a lower-VπIM, increase the
output photocurrent, and switch to a polarization multiplexing
geometry, we further improve the RF performance, demonstrat-
ing RF gain and NF of 0 and 24 dB, respectively. While tuning
the filter passband across the 2 to 8 GHz range, the maximum
variation of the RF gain and NF are 1.3 and 2 dB, respectively.
Finally, in Section V, we conclude.
3480 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
Fig. 1. Concept of comb-based microwave photonic filters using the inter-
ferometric configuration with double sideband suppressed carrier and balanced
detection. EDFA: erbium-doped fiber amplifier; VODL: variable optical de-
lay line; IM: intensity modulator; DCF: dispersion compensating fiber; BPD:
balanced photodetector.
II. CONCEPT AND THEORY
A. Concept
Fig. 1 shows a conceptual diagram of RF photonic filters in an
interferometric configuration utilizing DSB-SC and a BPD. The
EO-generated comb, which is nearly flat, is first sent through
an optical pulse shaper. The pulse shaper is used to carve out
a Gaussian-shaped spectrum from the input source in order to
provide a good filter shape in the RF domain. The resulting
Gaussian-shaped comb is amplified by the EDFA and directed
to the interferometer. In the delay path (i.e., upper path), a vari-
able optical delay line is used to tune the center frequency of
the filter passband [12]–[15]; in the modulation path (i.e., lower
path), a DSB-SC IM, biased at a minimum transmission point, is
used [12], [32]. This results in suppression of the optical carriers
and intensity noise in the modulation path. The interferometer
output signals are directed through the dispersion compensating
fiber (DCF) stage and input to the BPD, which acts to suppress
the intensity noise from the delay path. Although two identical
DCFs are pictured in the figure, in experiments we use a sin-
gle DCF configuration with either bidirectional propagation or
polarization multiplexing, discussed in more detail later. Ide-
ally, the intensity noise originating from common mode signal-
spontaneous beat noise can be perfectly suppressed through the
combination of the interferometric configuration with the DSB-
SC and balanced detection. However, the finite extinction ratio
value of the IM causes non-common mode signal-spontaneous
beat noise which is generated by the beating of the optical carri-
ers in one arm with the intensity noise in the other arm. We will
refer to the non-common mode signal-spontaneous beat noise
due to the finite extinction ratio of the IM as excess noise. The
residual common mode noise due to a finite common mode re-
jection ratio of the BPD is negligible when the two DCF paths
are well balanced. The spontaneous-spontaneous beat noise is
also negligible, compared to the signal-spontaneous beat noise.
B. Derivation of RF Performance Metrics
We derive the output photocurrent of the RF signal from
the electric field of the optical frequency comb in order to
describe the filter transfer function and RF gain. Then, the out-
put noise power spectral density (PSD) is derived in order to
describe the NF.
The electric field of the optical frequency comb at the output
of the EDFA can be written as
ecomb(t)=
n
√pnαSgAejωnt+c.c(1)
where pnand ωnare the optical power and angular frequency
of the nth comb line, respectively, where we have used c.c. to
represent the complex conjugate of the first term on the right
side. The angular frequencies satisfy ωn=ω0+nΔω.ω0and
Δωare the angular frequency of the optical carrier and comb
spacing (i.e., repetition rate), respectively. αsis the optical loss
factor of the pulse shaper. gAis the optical gain factor of the
EDFA. An electrical RF signal drives the IM which is biased
at the minimum transmission point. We write the electrical RF
signal at the input of the IM as
νin(t)=Vrf cos(ωRF t)(2)
where Vrf and ωRF are the RF voltage and angular RF fre-
quency, respectively. By taking the small signal approximation,
the electric field at the IM output [33] can be written as
eIM (t)=√αMein(t)
2
×2√ε−jπVrf
2Vπ
ejωRFt−jπVrf
2Vπ
e−jωRFt+c.c
(3)
where ein(t)is the electric field at the IM input. αMand εare the
optical loss factor and extinction ratio of the IM, respectively.
The first term represents the residual optical carriers which are
incompletely suppressed due to the modulator’s finite extinction
ratio. The other terms represent the two optical sidebands. We
write the electric fields at the two outputs of the interferometer
as
eA(t)=
n
√pnαSgA
×
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
α
2αDe−jωnτ−(1−α)εαM
2ejωnt
+jπVrf√(1−α)αM
4√2Vπej(ωn+ωRF )t
+jπVrf√(1−α)αM
4√2Vπej(ωn−ωRF )t
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
+c.c
(4)
eB(t)=
n
√pnαSgA
×
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎢
⎣
jα
2αDe−jωnτ+j(1−α)εαM
2ejωnt
+πVrf√(1−α)αM
4√2Vπej(ωn+ωRF )t
+πVrf√(1−α)αM
4√2Vπej(ωn−ωRF )t
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎥
⎦
+c.c
(5)
where αis the fraction of the power directed toward the upper
arm of the interferometer (the delay path), α:1−αis the input
KIM et al.: COMB-BASED RF PHOTONIC FILTERS BASED ON INTERFEROMETRIC CONFIGURATION AND BALANCED DETECTION 3481
power split ratio of the interferometer, τis the delay difference
between the two paths of the interferometer, and αDis the
optical loss factor of the variable optical delay line. A symmetric
3 dB coupler is assumed to form the interferometer output. After
propagation through the DCF, we write the electric fields at the
two inputs of the BPD as
eBPD−A(t)=
n
√pnαSgAαF
×
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎣
α
2αDe−jωnτ−(1−α)εαM
2ej[ωnt+ψ(ωn)]
+jπVrf√(1−α)αM
4√2Vπej[(ωn+ωRF )t+ψ(ωn+ωRF )]
+jπVrf√(1−α)αM
4√2Vπej[(ωn−ωRF )t+ψ(ωn−ωRF )]
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎦
+c.c
(6)
eBPD−B(t)=
n
√pnαSgAαF
×
⎡
⎢
⎢
⎢
⎢
⎢
⎢
⎣
jα
2αDe−jωnτ+j(1−α)εαM
2ej[ωnt+ψ(ωn)]
+πVrf√(1−α)αM
4√2Vπej[(ωn+ωRF )t+ψ(ωn+ωRF )]
+πVrf√(1−α)αM
4√2Vπej[(ωn−ωRF )t+ψ(ωn−ωRF )]
⎤
⎥
⎥
⎥
⎥
⎥
⎥
⎦
+c.c
(7)
where αFis the optical loss factor of the dispersive media
and ψ(ω)is the quadratic phase introduced by the chromatic
dispersion [34] given by
ψ(ω)=−β(ω)L=ψ0+ψ1(ω−ω0)+ψ2
2(ω−ω0)2(8)
where the dispersion coefficient ψ2relates to the dispersion
parameter D (in ps/nm/km) as
ψ2=Dλ2L
2πc .(9)
The photocurrent at the BPD output is given by
i(t)=κB|eBPD−B(t)|2−κA|eBPD−A(t)|2(10)
where κAand κBare the responsivities of the BPD, which in the
following are taken to be equal (κ=κA=κB), and stands
for averaging over the optical oscillations. We write the output
photocurrents of the RF signal as shown (11), at the bottom of
the next page.
The filter transfer function, giving the ratio of the output RF
voltage to the input RF voltage, can be written as
H(ωRF )∝πκαFαSgAR√α(1−α)αDαM
2Vπ
⎧
⎪
⎨
⎪
⎩
ej[ω0τ+ψ2
2ω2
RF ]
n
pnej[nΔω(ψ2ωRF +τ)]
−e−j[ω0τ+ψ2
2ω2
RF ]
n
pnej[nΔω(ψ2ωRF −τ)] ⎫
⎪
⎬
⎪
⎭
(12)
where R is the impedance of the photodetectors. In this expres-
sion we have omitted phase prefactors. The two terms within
the brackets comprise two different filter passbands, one arising
from each of the modulation sidebands [12]. The filter transfer
functions depend on the Fourier transform of the shaped optical
comb spectrum and are periodic with free spectral range (FSR)
given (in Hz units) by FSR = 1/ψ2Δω=1/T, where T is the
differential delay between the filter taps. The center frequencies
of the two passbands are shifted in opposite directions when the
delay difference between the two paths of the interferometer is
changed. Thus, there is no RF power fading because the two
filter passbands are not overlapped [12]. We write the RF gain
at the center RF frequency of the filter passbands, derived from
(11), as
Grf =Pout
Pin
=πκpSαSgAαFα(1 −α)αDαMR
2Vπ2
.
(13)
where ps=npnis the total comb power. We rewrite the RF
gain in terms of the total output dc photocurrent (IDC), which is
the sum of the photocurrents in each of two photodiodes of the
BPD, and the ratio (η) of the loss factor in the modulation path
of the interferometer to the loss factor in the delay path as
Grf ≈1
4
η
ε[1 + η]2πIDCR
Vπ2(14)
where IDC and ηare given by
IDC ≈psαsgA[ααD+(1−α)εαM]αFκ(15)
and
η=(1 −α)εαM
ααD
.(16)
The total output noise PSD (Nout) [35] can be written as
Nout =(1+Grf)Nth +Nshot +Nex (17)
where Nth =kT represents the thermal noise (k is Boltzmann’s
constant), Nshot =2qI
DCRis the shot noise, and Nex is the
excess noise. As mentioned in the previous section, the excess
noise is dominated by signal-spontaneous beat noise. In our
configuration each optical comb line gives rise to four distinct
beat noise contributions at RF frequency ωRF : the beating of
the optical carrier (ωn) from the delay path with the residual op-
tical noise at each of frequencies (ωn+ωRF and ωn−ωRF )
transmitted through the modulation path, and the beating of
the optical noise at frequencies ωn+ωRF and ωn−ωRF from
the delay path with the residual carrier transmitted through the
modulation path. The excess noise is calculated as
Nex =
n32 ααDαF
2(1−α)εαMαF
2 pnαSgAρnκ2R
!cos2ωnτ+ωRFτ
2+cos
2ωnτ−ωRFτ
2"
(18)
where the first and second terms in parentheses inside the sum-
mation represent the transmission factors associated with up-
per and lower interferometer paths, respectively, including also
the subsequent dispersive fiber propagation, and pnαSgAis the
powerofthenth comb line at the input to the interferometer. ρn
is the PSD of the optical noise in the vicinity of the nth comb line
3482 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
frequency at the output of the EDFA (input to the interferome-
ter) and is assumed to be polarized. One cosine squared term in
(18) arises due to the interferometric combination of the two RF
beat terms involving the optical noise at frequency ωn+ωRF;
the other arises due to the interferometric combination of the
two terms involving the optical noise at frequency ωn−ωRF .
With summation over a large number of comb lines, it can be
shown that the cosine squared terms each average nicely to a
value of approximately 0.5 for most cases. (The exception is
for settings of the interferometer delay τvery close to nπ/Δω,
where nis an integer in this range the cosine squared terms vary
rapidly with τ, even after summation.) Taking each of the cosine
squares as 0.5, the excess noise derived from the electric field
of the optical comb and noise PSD can be written as
Nex ≈8R(καF)2ααD(1 −α)εαM
n
ρnpnαSgA.(19)
We can rewrite the excess noise with IDC and ηas
Nex ≈2RIN η
[1 + η]2I2
DCR(20)
where RIN ≈n4ρnpn/(p2
SαSgA)is the relative intensity
noise, assumed to be dominated by signal-spontaneous beat
noise, of the comb at the output of the EDFA [36]. Finally, the
NF can be calculated by the RF gain and the total output noise
PSD as
NFrf (dB) = 10 log Nout
Grf Nth .(21)
III. SIMULATION
We will now investigate effects of the input split coefficient
(α) on the RF performance, when the EDFA output power or
output photocurrent is fixed. Fig. 2 shows RF gain, path loss
ratio and NF as a function of the input split coefficient αfor the
fixed output photocurrents of 10, 15, and 20 mA. The EDFA
gain is varied to fix the output photocurrents as the input split
coefficient is varied. The simulation parameters are summarized
in Table I. In Fig. 2(a), as the input split coefficient decreases,
RF gain is increased and the interferometer path loss ratio (η)
is increased. When the path loss ratio is 0 dB, the input split
coefficient is 0.008, producing the highest RF gain at all fixed
photocurrents. Then, the RF gain is decreased as the input split
coefficient further decreases. This means that the effective op-
tical modulation index related to the power difference between
the optical carrier and sideband can be improved by adjusting
the path loss ratio. In Fig. 2(b), as the input split coefficient
decreases, the NF is decreased with its lowest value occurring
at the path loss ratio of 0 dB. Although the path loss ratio
of 0 dB produces the highest RF gain, it requires very high
output power of the EDFA. For example, the required EDFA
Fig. 2. (a) RF gain, (b) noise figure as a function of the interferometer split
coefficient (α) for the fixed output photocurrent of 10, 15, and 20 mA.
TABLE I
SIMULATION PARAMETERS
Parameter Value Unit
Half-wave voltage of IM 3 V
Loss of VDL 2.5 dB
Loss of IM 3.5 dB
Extinction ratio of IM 20 dB
Loss of DCF 3.5 dB
Responsivity 0.65 A/W
NF of EDFA 4 dB
RIN of comb source −170 dB/Hz
output power at the path loss ratio of 0 dB and with a fixed
output photocurrent of 20 mA is approximately 39 dBm. In
practice, this high EDFA output power may cause damage of
the optical components placed after the EDFA.
Fig. 3 shows the RF gain, output noise, and NF with respect
to the input split coefficient for fixed EDFA output power. The
EDFA output power is fixed to 31 dBm since the maximum
optical power handling of the IM used for this experiment is
30 dBm. The input split coefficient is varied. In Fig. 3(a), the
RF gain is maximum at the input split coefficient α=0.5 (i.e.,
iRF (t)=πVrfκαF√α(1−α)αDαM
2Vπ
n
pnαSgA⎡
⎣sin ωRF t+ω0τ+ψ1ωRF +ψ2
2ω2
RF +nΔω(ψ2ωRF +τ)
−sin ωRF t−ω0τ+ψ1ωRF −ψ2
2ω2
RF +nΔω(ψ2ωRF −τ) ⎤
⎦.(11)
KIM et al.: COMB-BASED RF PHOTONIC FILTERS BASED ON INTERFEROMETRIC CONFIGURATION AND BALANCED DETECTION 3483
Fig. 3. (a) RF gain, (b) Noise power spectral density, and (c) noise figure as a
function of the interferometer input split coefficient for the fixed erbium-doped
fiber amplifier output power of 31 dBm.
50:50 input split ratio). Then, the RF gain is reduced as the
input split coefficient either increases from 0.5 to 1 or decreases
from 0.5 to 0. In Fig. 3(b), the dominant noise terms (shot noise
and excess noise) are also reduced as the αdecreases from
0.5 towards 0. The RF gain and excess noise in (13) and (19),
respectively, have the same factor of α(1 −α), related with the
input split coefficient. When the excess noise dominates, the NF
remains constant as the input split coefficient is varied. However,
as shown in Fig. 3(b), when αis large, the shot noise dominates.
When the split coefficient αdecreases, the shot noise falls more
quickly than the excess noise and becomes negligible. Thus, for
small α, we may expect excess-noise limited performance. In
Fig. 3(c), as the split coefficient decreases from 0.5 to 0.1, the NF
improves slightly due to shot noise reduction. This improvement
Fig. 4. Experimental configuration A for comparison of two schemes using
the single and balanced photodetectors.
Fig. 5. Spectra of (a) as-generated optical frequency comb and (b) Gaussian-
shaped comb.
is similar to that obtained under low biasing in RF photonic links
[23], [24].
IV. EXPERIMENT
A. Comparison of Filter Using SPD and BPD
Fig. 4 shows the experimental setup for comparison of two
schemes using either SPD or BPD. The optical frequency comb
with 18 GHz repetition rate and nearly flat power spectrum
is generated by cascaded intensity and phase modulation of a
continuous-wave laser [21]. Fig. 5(a) shows the spectrum of the
optical frequency comb; the resolution bandwidth of the optical
spectrum analyzer is 0.02 nm. Then, the amplitude spectrum
of the comb is tailored by a commercial optical pulse shaper
(Finisar WaveShaper 1000S/SP) to make a Gaussian-shaped
comb shown in Fig. 5(b). The Gaussian-shaped comb is ampli-
fied by the EDFA and then split into two delay and modulation
paths through an optical splitter. We use a 10:90 split ratio
achieved with the help of a variable optical attenuator. 10% of
the comb is directed to a fixed delay line; 90% of the comb is
directed to the IM, which is biased at the minimum transmis-
sion point. The IM (EOSPACE AZ-0K5-10) has the half-wave
voltage of 4.1 V, the extinction ratio of 30 dB, and RF band-
width of 10 GHz. The outputs of the fixed delay line and IM
are connected to the inputs of a 2×2 optical coupler having a
coupling ratio of 50:50. The two outputs of the optical coupler
are connected to the BPD through the bidirectional DCF config-
uration which uses a single spool of DCF and two circulators.
The DCF has a dispersion value of −404 ps/nm at 1550 nm. At
the inputs of the BPD, a variable optical delay line and variable
optical attenuator are used. The BPD (Discovery Semiconduc-
tors DSC720-HLPD) is an integrated push-pull device com-
prised of two InGaAs photodetectors (PD-A and PD-B), with
3484 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
Fig. 6. Measured RF gain and calculated filter transfer function as a function
of the RF frequency for two schemes using the single photodetector (PD-A or
PD-B) and balanced photodetector.
responsivities of approximately 0.62 and 0.65 A/W. The RF
bandwidth of the BPD is approximately 16 GHz. The respon-
sivity of PD-A is lower than that of PD-B by approximately 5%,
which corresponds to the common mode rejection ratio of 26 dB.
The measured optical power at the inputs of PD-A and PD-B are
10.2 and 10.0 dBm, where the optical powers are intentionally
mismatched by approximately 5% to achieve a higher common
mode rejection ratio. The total output photocurrent is approx-
imately 13 mA. All RF powers at the output of the BPD are
reduced by a factor of 4 since the BPD has an internal matching
resistor for maximum power transfer to a matched load. All the
numbers shown in this paper for output photocurrent, RF gain,
and noise PSD refer to the values before the internal matching
resistor. In other words, to account for the internal matching
resistor, 6 dB is added to the power measured at the BPD out-
put. To evaluate RF gain, a vector network analyzer is used.
For output noise measurements, an electrical spectrum analyzer
with a low pass filter and a low noise amplifier is used. The
low pass filter (KL Microwaves 6L250-10000/T20000), having
a 3-dB cutoff frequency of 10 GHz, is used to suppress the
18 GHz comb beat note. The low noise amplifier (Miteq AMF-
4D-001180-24-10P) having the RF gain of >30 dB and the NF
of <2.9 dB across 0.1–18 GHz improves the sensitivity of our
noise measurements.
We compare the RF performance of the RF photonic filters
using the SPD and BPD. For the comparison, we use the results
of the scheme using the SPD when only one of the inputs of the
BPD is connected. Fig. 6 shows the RF gain as a function of
the RF frequency for the two schemes using the SPD (PD-A or
PD-B) and BPD. The FSR of the filter is 17.4 GHz. With a de-
lay difference of approximately 10.5 ps between the delay path
and modulation path, the RF passband frequencies are 3.33 and
14.07 GHz, the 3 dB bandwidth is approximately 1.3 GHz, and
the sidelobe suppression is >30 dB. The RF gain at 3.33 GHz is
−16.5 dB for the SPD case (PD-A or PD-B). For the BPD case,
the RF gain at 3.33 GHz is −10.3 dB. Compared to the SPD
cases, the RF gain of the BPD case is improved by 6 dB because
the total output photocurrent is increased by a factor of 2. The
filter shape of the BPD case is compared to that of the calculated
filter transfer function based on (12) and the Gaussian-shaped
comb spectrum shown in Fig. 5(b). At 3.33 GHz, the measured
filter shape is closely matched to the calculated transfer func-
Fig. 7. Noise power spectral density as a function of the RF frequency for
two schemes using the single photodetector (PD-A or PD-B) and balanced
photodetector.
tion. However, the predicted out-of-band suppression (−43 dB)
is approximately 5–10 dB below the simulated value, limited
by noise due to spontaneous Brillouin scattering in the DCF. In
this experiment total optical power of approximately 15 dBm
is injected to the DCF. Since the effective stimulated Brillouin
scattering threshold with the comb is high [35], [37], sponta-
neous Brillouin scattering is dominant and the loss due to the
spontaneous Brillouin scattering is negligible [38], [39]. How-
ever, the scattering affects the stopband attenuation due to the
counterpropagating geometry in the DCF. The RF gain at the
14.07 GHz filter peak for the BPD case interferometer output
signals is approximately 2 dB lower than the peak value of the
calculated filter transfer function. This difference is attributed
to the dependence of the IM half-wave voltage and the BPD
responsivities on the RF frequency.
Fig. 7 shows the output noise PSD as a function of RF fre-
quency for the RF photonic filters using the SPD (PD-A or
PD-B) and BPD. For the SPD cases, the output noise PSDs are
varied in the range of −145.1 to −136.6 dBm/Hz. The domi-
nant noise source is signal-spontaneous beat noise for the SPD
case. For the BPD case, the output noise PSD is approximately
−155 dBm/Hz with a noise suppression of 10 18 dB below
6 GHz. This is close to the shot noise level of approximately
−157 dBm/Hz. Because the noise penalty is approximately
2 dB, the excess noise originating from the finite extinction
ratio of the IM would be approximately −159.5 dBm/Hz if the
suppressed common mode noise is negligible. The NF values
for the SPD and BPD cases are approximately 50 and 29 dB at
the RF center frequency, respectively. As shown in Fig. 7, noise
peaks at the frequencies of approximately 7, 8, and 10 GHz
are generated due to the spontaneous Brillouin scattering in the
DCF which has multiple resonant peaks in its Brillouin gain
spectra [39].
B. Filter Passband Tunability With Enhanced RF Performance
Fig. 8 shows the setup for an experiment which demonstrates
both filter passband tunability and enhanced RF performance.
Compared to the previous experimental setup, some components
have been changed, as has the DCF configuration. First, the fil-
ter taps (i.e., the number of comb lines) are increased through
the use of an upgraded EO-generated comb generator which
KIM et al.: COMB-BASED RF PHOTONIC FILTERS BASED ON INTERFEROMETRIC CONFIGURATION AND BALANCED DETECTION 3485
Fig. 8. Experimental configuration B for further RF performance im-
provement and filter passband tunability. (PBC/PBS: polarization beam
combiner/splitter).
Fig. 9. Optical spectra of (a) EO-generated comb and (b) Gaussian-shaped
comb. (Resolution bandwidth =0.05 nm).
uses three cascaded phase modulators and one IM to generate
60–75 comb lines [22]. The output power of the comb generator
is approximately 17 dBm. Fig. 9(a) and (b) shows the gener-
ated flat optical frequency comb and Gaussian shaped comb,
respectively. The measured relative intensity noise at the out-
put of the EDFA is approximately −152.6 dB/Hz. Second, the
fixed delay line is changed to the variable optical delay line for
tuning the center frequency of the filter passband. Third, an IM
(EOSPACE AZ-1×2-8K8-20) having a lower half-wave voltage
(3 V at 1 GHz), higher optical power handling (1 W), and wider
RF bandwidth (16 GHz) is used to further increase RF perfor-
mance. However, the extinction ratio of the IM is 21 dB which
is worse than that of the previous IM (extinction ratio =30 dB).
In addition, the output photocurrent is increased to 18.2 mA.
Fourth, polarization maintaining fibers and components are used
from the comb source to the interferometer. The EDFA has an
internal polarizer so the amplified spontaneous emission noise
at its output is polarized. Finally, the bidirectional DCF config-
uration was changed to the polarization multiplexing DCF con-
figuration to solve the Brillouin scattering problem. In Fig. 8,
the delay and modulation path signals at the output of the in-
terferometer are orthogonally combined by a polarization beam
combiner. The output signal is transmitted through the DCF to
a polarization beam splitter. The principal axes of the polariza-
tion beam splitter are aligned to have an angle of 45°to the
polarization state of either the delay or modulation path signal
for generation of complementary signals. The complementary
signals are detected by the BPD. In the polarization multiplex-
ing DCF configuration, the effects of the Brillouin scattering
on the filter transfer function and noise peaks is minimized be-
cause the propagation direction in the DCF is unidirectional,
whereas Brillouin scattering occurs in the counterpropagating
Fig. 10. Measured and simulated RF gain at different filter center frequencies
when the center frequencies of lower filter passbands are 2, 4, 6, and 8 GHz.
Solid and dash lines indicate measured and simulated values, respectively.
direction. Therefore, care should be taken to avoid reflection of
the generated Brillouin scattering noise.
Fig. 10 shows the measured and simulated RF gain when
the center frequencies of lower filter passbands are 2, 4, 6, and
8 GHz. Because of the increased number of comb lines, the
3 dB bandwidth is decreased to approximately 770 MHz. The
measured RF gain values at the filter peaks are varied from 0 to
−1.3 dB, which are higher than that of the previous setup due to
the lower half-wave voltage and increased output photocurrent.
The measured RF gain and filter transfer function agree well
with the simulated values. The sidelobe suppression is >32 dB.
The stopband attenuation on the high frequency side of the pri-
mary passband is in the range 40–55 dB. The baseband response
caused by the residual optical carriers in the modulation path
is suppressed by balanced detection. However, some baseband
response remains due to the finite common mode rejection ratio,
and this baseband response varies due to bias drift of the IM.
At baseband, the RF gain is lower than approximately −38 dB.
Fig. 11 shows measured and simulated noise PSD as a function
of the frequency at different filter center frequencies of 2, 4, 6,
and 8 GHz. The estimated shot noise is −155.4 dBm/Hz at the
total output photocurrent of 18.2 mA. Using (17) and (20), the
calculated total output noise and excess noise are −150.3 and
−152 dBm/Hz, respectively. Above 2 GHz, the measured output
noise levels are flat and their average values are very close to
the simulated value. The noise penalty of this setup is approx-
imately 5 dB, which is higher than that of the previous setup
due to the lower extinction ratio of the IM and increased output
photocurrent. As the filter passband is tuned from 2 to 8 GHz,
the measured output noise levels are not changed. Fig. 12 shows
measured and simulated NF as a function of the filter center fre-
quencies. The measured NF values estimated by (21) are from
24 to 26 dB. The difference values between the measured and
3486 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
Fig. 11. Measured and simulated noise power spectral densities as a function
of the frequency at different filter center frequencies of 2, 4, 6, and 8 GHz.
Fig. 12. Measured and simulated noise figure as a function of the filter center
frequency.
simulated NF are less than approximately 1.4 dB for the filter
peaks of 3 to 8 GHz. However, at the filter peak of 2 GHz,
the difference value of 2.6 dB is relatively higher than those of
the other filter peaks due to the low-frequency interferometric
noise shown in Fig. 11. It is attributed to continuous-wave laser
phase- to intensity-noise conversion which takes place in the
DCF-PBS of this scheme [40]–[42]. The low-frequency noise
could be eliminated using a unidirectional DCF configuration
which has two well-matched DCFs. However, from a practical
perspective, the fabrication of two DCFs with precisely the same
delay and dispersion values is difficult, and stabilization may be
required; any mismatches degrade the common mode rejection
ratio for balanced detection. A multi-core fiber or an optical
ribbon could potentially be used because the fiber cores or mul-
tiple fibers are made with a single cladding or ribbon package,
respectively, which should provide for stable matching between
the different cores [43], [44].
The overall RF performance of the RF photonic filter using the
BPD is substantially improved as compared to the conventional
filtering schemes. Table II compares experimental results of the
filtering schemes and simulation results of the conventional link
TABLE II
RF PERFORMANCE COMPARISON
Schemes RF gain(dB) NF(NP)(dB) ∗Vπ(V) ER(dB) IDC (mA)
Filtering [15] −40 – – – 0.9
Experiment-A (This work) SPD −16.5 50 4.1 30 6.5
BPD −10.3 29 (2) 13
Experiment-B (This work) −1.3 02426 (5) 3 21 18.2
∗∗Conventional link [24] −0.42 19.1 (0) 3 – 18.2
∗All values are at 1 GHz.
∗∗ Simulation results with a dual-output modulator and a balanced photodetector.
NP: noise penalty ( =10log(Nout/Nsho t)); ER: extinction ratio.
with the noise reduction technique using a dual-output modula-
tor and a BPD. Compared to [15] and the experiment-A, the RF
gain of the experiment-B is increased to approximately 0 dB due
to the lower-Vπand increased photocurrent. It is comparable to
that of the conventional link at the same Vπand photocurrent.
The NF of the RF photonic filter is improved to approximately
24 dB. However, the noise penalty is degraded from 2 to 5 dB
due to the lower extinction ratio and increased photocurrent. In
the experiment-B, with the extinction ratio of >26 dB, the noise
penalty can be further reduced to approximately 0 dB and thus
the NF of the RF photonic filter would become also close to that
of the conventional link.
V. CONCLUSION
We demonstrate improvements to the RF performance of
comb-based RF photonic filters by using an interferometric
configuration with DSB-SC and balanced detection. Balanced
detection increases the output photocurrent by a factor of two
and thus improves the RF gain by 6 dB. Intensity noise is sup-
pressed both through biasing the intensity modulator inside the
interferometer at the minimum transmission point and through
balanced detection, resulting in the reduction of the noise figure.
In addition, we show how the effective optical modulation index
and NF can be improved by adjusting the input split ratio of the
interferometer. In a first experiment using a balanced photode-
tector (BPD) and a counterpropagating geometry in a dispersive
fiber, we achieve a RF gain of −10.3dBandaNFof29dBat
the filter center frequency. Compared to an identical experiment
but using a single photodiode, the RF gain and NF are im-
proved by approximately 6 and 21 dB, respectively. In a second
experiment using an intensity modulator with lower half-wave
voltage, an increased output photocurrent, and a copropagating
polarization multiplexing geometry in a dispersive fiber, the RF
gain and NF are further improved up to approximately 0 and
24 dB, respectively. In the latter experiment the filter passband
is tuned from 2 to 8 GHz while maintaining roughly constant
RF gain and noise figure, without RF power fading, without
filter baseband response, and with approximately Gaussian RF
filter shape (maintaining sidelobe suppression and stopband at-
tenuation >32 dB). We believe that the RF performance metrics
such as RF gain and NF can be further improved through the
use of a high-power, low-noise continuous-wave laser as a seed
to the electro-optic comb generator and through the use of an
KIM et al.: COMB-BASED RF PHOTONIC FILTERS BASED ON INTERFEROMETRIC CONFIGURATION AND BALANCED DETECTION 3487
intensity modulator having lower half-wave voltage and higher
extinction ratio.
An additional dimension important to pursue for practi-
cal applications is to increase the level of integration. In
another paper to appear in this special issue [45], our group has
demonstrated programmable RF photonic filtering using a comb
source generated via CW-laser pumping of an on-chip silicon
nitride microresonator. Optical pulse shaping has been demon-
strated at the chip-level by several groups in platforms such as
InP, silica, and silicon, e.g., [46]–[49]. Although significant fur-
ther work is needed, such developments suggest the potential
for substantial reduction in the footprint of RF photonic filters
employing optical frequency combs.
ACKNOWLEDGMENT
The authors would like to thank helpful discussions with
J. D. McKinney and V. R. Supradeepa.
REFERENCES
[1] J. Uher and W. J. R. Hoefer, “Tunable microwave and millimeter-wave
band-pass filters,” IEEE Trans. Microw. Theory Tech., vol. 39, no. 4,
pp. 643–653, Apr. 1991.
[2] G. M. Rebeiz, RF MEMS: Theory, Design, and Technology. Hoboken, NJ,
USA: Wiley, 2003.
[3] S. J. Park, I. Reines, C. Patel, and G. M. Rebeiz, “High-Q RF-MEMS 4–6-
GHz tunable evanescent-mode cavity filter,” IEEE Trans. Microw. Theory
Tec h., vol. 58, no. 2, pp. 381–389, Feb. 2010.
[4] X. Liu, L. P. B. Katehi, W. J. Chappell, and D. Peroulis, “High-Q tun-
able microwave cavity resonators and filters using SOI-based RF MEMS
tuners,” IEEE J. Microelectromech. Syst., vol. 19, no. 4, pp. 774–784, Aug.
2010.
[5] E. J. Naglich, J. Lee, D. Peroulis, and W. J. Chappell, “Switchless tunable
bandstop-to-all-pass reconfigurable filter,” IEEE Trans. Microw. Theory
Tec h., vol. 60, no. 5, pp. 1258–1265, May 2012.
[6] J. Capmany, B. Ortega, and D. Pastor, “A tutorial on microwave photonic
filters,” J. Lightw. Technol., vol. 24, no. 1, pp. 201–229, Jan. 2006.
[7] J. Capmany, D. Pastor, and B. Ortega, “New and flexible fiber-optic delay
line filters using chirped Bragg gratings and laser arrays,”IEEE Trans.
Microw. Theory Tech., vol. 47, no. 7, pp. 1321–1327, Jul. 1999.
[8] Y. Dai and J. Yao, “Nonuniformly-spaced photonic microwave delayline
filter,” Opt. Exp., vol. 16, no. 7, pp. 4713–4718, Mar. 2008.
[9] L. Li, X. Yi, T. X. H. Huang, and R. A. Minasian, “Distortion-free spec-
trum sliced microwave photonic signal processor: analysis, design and
implementation,” Opt. Exp., vol. 20, no. 10, pp. 11517–11528, May 2012.
[10] J. Mora, B. Ortega, J. Capmany, J. L. Cruz, M. V. Andres, D. Pastor, and S.
Sales, “Automatic tunable and reconfigurable fiberoptic microwave filters
based on a broadband optical source sliced by uniform ber Bragg gratings,”
Opt. Exp., vol. 10, no. 22, pp. 1291–1298, Nov. 2002.
[11] A. Ortigosa-Blanch, J. Mora, J. Capmany, B. Ortega, and D. Pastor, “Tun-
able radio-frequency photonic filter based on an actively mode locked
fiber laser,” Opt. Lett., vol. 31, no. 6, pp. 709–711, Mar. 2006.
[12] E. Hamidi, R. Wu, V. R. Supradeepa, C. M. Long, D. E. Leaird, and A. M.
Weiner, “Tunable radio frequency photonic filter based on intensity mod-
ulation of optical combs,” in Proc. IEEE Top. Meeting Microw. Photon.,
Oct. 2010, pp. 393–396.
[13] E. Hamidi, D. E. Leaird, and A. M. Weiner, “Tunable programmable
microwave photonic filters based on an optical frequency comb,” IEEE
Trans. Microw. Theory Tech., vol. 58, no. 11, pp. 3269–3278, Nov. 2010.
[14] M. Song, C. M. Long, R. Wu, D. Seo, D. E. Leaird, and A. M. Weiner,
“Reconfigurable and tunable flat-top microwave photonic filters utilizing
optical frequency comb,” IEEE Photon. Technol. Lett., vol. 23, no. 21,
pp. 1618–1620, Nov. 2011.
[15] V. R. Supradeepa, C. M. Long, R. Wu, F. Ferdous, E. Hamidi, D. E.
Leaird, and A. M. Weiner, “Comb-based radio-frequency photonic fil-
ters with rapid tunability and high selectivity,” Nature Photon.,vol.6,
pp. 186–194, 2012.
[16] M. Song, V. Torres-Company, A. J. Metcalf, and A. M. Weiner, “Mul-
titap microwave photonic filters with programmable phase response via
optical frequency comb shaping,” Opt. Lett., vol. 37, pp. 845–847, Mar.
2012.
[17] R. Wu, C. M. Long, D. E. Leaird, and A. M. Weiner, “Directly generated
Gaussian-shaped optical frequency comb for microwave photonic filtering
and picosecond pulse generation,” IEEE Photon. Techn. Lett., vol. 24,
no. 17, pp. 1484–1486, Sep. 2012.
[18] M. Song, V. Torres-Company, R. Wu, A. J. Metcalf, and A. M. Weiner,
“Compression of ultra-long microwave pulses using programmable mi-
crowave photonic phase filtering with >100 complex-coefficient taps,”
Opt. Exp., vol. 22, no. 6, pp. 6329–6338, 2014.
[19] R. Wu, M. Song, D. E. Leaird, and A. M. Weiner, “Comb-based radio-
frequency photonic filtering with 20 ns bandwidth reconfiguration,” Opt.
Lett., vol. 38, no. 15, pp. 2735–2738, 2013.
[20] V.Torres-Company, D. E. Leaird, and A. M. Weiner, “Simultaneous broad-
band microwave downconversion and programmable complex filtering by
optical frequency comb shaping,” Opt. Lett., vol. 37, no. 19, pp. 3993–
3995, 2012.
[21] R. Wu, V. R. Supradeepa, C. M. Long, D. E. Leaird, and A. M. Weiner,
“Generation of very flat optical frequency combs from continuous wave
lasers using cascaded intensity and phase modulators driven by tailored
radio frequency waveforms,” Opt. Lett., vol. 35, no. 19, pp. 3234–3236,
Oct. 2010.
[22] A. J. Metcalf, V. Torres-Company, D. E. Leaird, and A. M. Weiner, “High-
power broadly tunable electro-optic frequency comb generator,” IEEE J.
Sel. Topics Quantum Electron., vol. 19, no. 6, pp. 3500306–3500306,
Nov./Dec. 2013.
[23] C. Cox, E. Ackerman, G. Betts, and J. Prince, “Limits on the per-
formance of RF-over-fiber links and their impact on device design,”
IEEE Trans. Microw. Theory Tech., vol. 54, no. 2, pp. 906–920, Feb.
2006.
[24] E. I. Ackerman, G. E. Betts, W. K. Burns, J. C. Campbell, C. H. Cox,
N. Duan, J. L. Prince, M. D. Regan, and H. V. Roussell, “Signal-to-noise
performance of two analog photonic links using different noise reduction
techniques,” in Proc. IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2007,
pp. 51–54.
[25] J. D. McKinney, M. Godinez, V. J. Urick, S. Thaniyavarn, W. Charczenko,
and K. J. Williams, “Sub-10 dB noise figure in a multiple-GHz analog
optical link,” IEEE Photon. Technol. Lett., vol. 19, no. 7, pp. 465–467,
Apr. 2007.
[26] I. Gasulla and J. Capmany, “Analytical model and figures of merit for
filtered microwave photonic links,” Opt. Exp., vol. 19, no. 20, pp. 19758–
19774, Sep. 2011.
[27] M. Bolea, J. Mora, B. Ortega, and J. Capmany, “Highly chirped single-
bandpass microwave photonic filter with reconfiguration capabilities,”
Opt. Exp., vol. 19, no. 5, pp. 4566–4576, 2011.
[28] X. Xue, X. Zheng, H. Zhang, and B. Zhou, “Highly reconfigurable
microwave photonic single-bandpass filter with complex continuous
time impulse responses,” Opt. Exp., vol. 20, no. 24, pp. 26929–26934,
2012.
[29] X. Xue, X. Zheng, H. Zhang, and B. Zhou, “Spectrum-sliced microwave
photonic filter with an improved dynamic range based on a LiNbO3phase
modulator and balanced detection,” IEEE Photon. Techn. Lett., vol. 24,
no. 9, pp. 775–777, May 2012.
[30] X. Xue, X. Zheng, H. Zhang, and B. Zhou, “Noise reduction by bal-
anced detection in microwave photonic filters based on optical broadband
sources,” in Proc. Conf. Lasers Electro-Opt., 2011, pp. 1–3, Paper CThY3.
[31] H.-J. Kim, D. E. Leaird, and A. M. Weiner, “Improved RF performance of
a comb-based microwave photonic filter using a balanced photodetector,”
in Proc. IEEE Top. Meeting Microw. Photon., Oct. 2013, pp. 80–83.
[32] C. Middleton and R. Desalvo, “High performance microwave photonic
links using double sideband suppressed carrier modulation and balanced
coherent heterodyne detection,” in Proc. IEEE Mil. Commun. Conf., 2009,
pp. 1–6.
[33] V. J. Urick, M. S. Rogge, F. Bucholtz, and K. J. Williams, “The perfor-
mance of analog photonic links employing highly compressed erbium-
doped fiber amplifiers,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 7,
pp. 3141–3145, Jul. 2006.
[34] A. M. Weiner, Ultrafast Optics, 1st ed. New York, NY, USA: Wiley, 2009.
[35] J. D. McKinney and K. J. Williams, “Sampled analog optical links,”
IEEE Trans. Microw. Theory Tech., vol. 57, no. 8, pp. 2093–2099, Aug.
2009.
[36] D. Derickson, Fiber Optic Test and Measurement. Englewood Cliffs, NJ,
USA: Prentice-Hall, 1998.
3488 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 32, NO. 20, OCTOBER 15, 2014
[37] J. D. McKinney, V. J. Urick, and J. Briguglio, “Optical comb sources
for high dynamic-range single-span long-haul analog optical links,” IEEE
Trans. Microw. Theory Tech., vol. 59, no. 12, pp. 3249–3257, Dec. 2011.
[38] A. Kobyakov, M. Sauer, and D. Chowdhury, “Stimulated Brillouin scat-
tering in optical fibers,” Adv. Opt. Photon., vol. 2, no. 1, pp. 1–59, 2010.
[39] A. Yeniay, J. Delavaux, and J. Toulouse, “Spontaneous and stimulated
Brillouin scattering gain spectra in optical fibers,” J. Lightw. Technol.,
vol. 20, no. 8, pp. 1425–1432, Aug. 2002.
[40] W. K. Marshall, B. Crosignani, and A. Yariv, “Laser phase noise to inten-
sity noise conversion by lowest-order group-velocitydispersion in optical
fiber: Exact theory,” Opt. Lett., vol. 25, no. 3, pp. 165–167, Feb. 2000.
[41] J. C. Attard, J. E. Mitchell, C. J. Rasmussen, “Performance analysis of
interferometric noise due to unequally powered interferers in optical net-
works,” J. Lightw. Technol., vol. 23, no. 4, pp. 1692–1703, Apr. 2005.
[42] R. W. Tkach and A. R. Chraplyvy, “Phase noise and linewidth in an
InGaAsP DFB laser,” J. Lightw. Technol., vol. 4, no. 11, pp. 1711–1716,
Nov. 1986.
[43] T. Hayashi, T. Taru, O. Shimakawa, T. Sasaki, and E. Sasaoka, “Low-
crosstalk and low-loss multi-core fiber utilizing fiber bend,” presented
at the Opt. Fiber Commun. Conf. Expo., Los Angeles, CA, USA, 2011,
Paper OWJ3.
[44] G. S. Brockway and M. R. Santana, “Analysis of thermally induced loss
in fiber-optic ribbons,” Bell Syst. Tech. J., vol. 62, no. 4, pp. 993–1018,
Apr. 1983.
[45] X. Xue, Y. Xuan, H.-J. Kim, J. Wang, D. E. Leaird, M. Qi, and A. M.
Weiner, “Programmable single-bandpass photonic RF filter based on Kerr
comb from a microring,” J. Lightw. Technol., vol. PP, no. 99, pp. 1, Mar.
2014.
[46] M. J. R. Heck, P. Munoz, B. W. Tilma, E. A. J. M. Bente, Y. Barbarin, Y.-S.
Oei, R. Notzel, and M. K. Smit, “Design, fabrication and characterization
of an InP-based tunable integrated optical pulse shaper,” IEEE J. Quantum
Electron., vol. 44, no. 4, pp. 370–377, Apr. 2008.
[47] S. Tahvili, S. Latkowski, B. Smalbrugge, X. J. M. Leijtens, P. J. Williams,
M. J. Wale, J. Parra-Cetina,R. Maldonado-Basilio, P. Landais, M. K. Smit,
E. A. J. M. Bente, “InP-based integrated optical pulse shaper: Demonstra-
tion of chirp compensation,” IEEE Photon. Technol. Lett., vol. 25, no. 5,
pp. 450–453, Mar. 2013.
[48] D. J. Geisler, N. K. Fontaine, T. He, R. P. Scott, L. Paraschis, J. P. Heritage,
and S. J. B. Yoo, “Modulation-format agile, reconfigurable Tb/s transmit-
ter based on optical arbitrary waveform generation,” Opt. Exp., vol. 17,
no. 18, pp. 15911–15925, 2009.
[49] M. H. Khan, H. Shen, Y. Xuan, L. Zhao, S. Xiao, D. E. Leaird,
A. M. Weiner,and M. Qi, “Ultrabroad-bandwidth arbitrary radiofrequency
waveform generation with a silicon photonic chip-based spectral shaper,”
Nature Photon., vol. 4, no. 2, pp. 117–122, 2010.
Hyoung-Jun Kim received the B.S. degree in electrical engineering from
Kwangwoon University, Seoul, Korea, in 2005. He received the M.S. and Ph.D.
degrees in electrical engineering from Gwangju Institute of Science and Tech-
nology, Gwangju, Korea, in 2007 and 2011, respectively.
From 2011 to 2012, he was at High Speed Integrated Circuit Laboratory in
GIST, where he was engaged in research on millimeter-wave communication
systems utilizing RF photonics technologies. In 2012, he joined the Ultrafast
Optics group in Purdue University, USA, where he is currently working on re-
configurable RF photonic filters based on optical frequency combs and optical
pulse shaping.
Daniel E. Leaird was born in Muncie, IN, USA, in 1964. He received the B.S.
degree in physics from Ball State University, Muncie, IN, in 1987, and the M.S.
and Ph.D. degrees from the School of Electrical and Computer Engineering,
Purdue University, West Lafayette, IN, in 1996 and 2000, respectively.
He joined Bell Communications Research (Bellcore), Red Bank, NJ, USA,
as a Senior Staff Technologist in 1987, and later advanced to Member of Tech-
nical Staff. From 1987 to 1994, he worked in the Ultrafast Optics and Optical
Signal Processing Research Group, where he was a key team member in re-
search projects in ultrafast optics, such as shaping of short optical pulses using
liquid crystal modulator arrays, investigation of dark soliton propagation in op-
tical fibers, impulsive stimulated Raman scattering in molecular crystals, and
all-optical switching.
Dr. Leaird is currently a Senior Research Scientist and Laboratory Manager
of the Ultrafast Optics and Optical Fiber Communications Laboratory, School
of Electrical and Computer Engineering, Purdue University, where he has been
since 1994. He has coauthored approximately 100 journal articles, 150 confer-
ence proceedings, and has three issued U.S. patents.
He is active in the optics community and professional organizations including
the Optical Society of America and IEEE Photonics Society where he served
as the Chair of the Ultrafast Optics technical committee from 2006–2009 as
well as serving as a Consultant to venture capitalists by performing technical
due diligence. He also serves as a Reviewer for Optics Letters,Optics Express,
Photonics Technology Letters,Applied Optics,andJournal of the Optical Soci-
ety of America B in addition to serving on National Science Foundation review
panels in the SBIR program.
Dr. Leaird received several awards for his work in the ultrafast optics field
including a Purdue Professional Achievement Award, a Magoon Award for out-
standing teaching, an Optical Society of America/New Focus Student Award,
and a Bellcore “Award of Excellence.”
Andrew J. Metcalf received the B.S. degree(summa cum laude) in electrical en-
gineering from the University of Wisconsin-Milwaukee, Milwaukee, WI, USA,
in 2010, and the M.S.E.C.E degree from Purdue University, West Lafayette,
IN, USA, in 2012, where he is currently working toward the Ph.D. degree in
electrical engineering.
From 2008–2010, he worked as a co-op at Harley-Davidson Motor Com-
pany before becoming a Graduate Research Assistant in the Ultrafast Optics
and Optical Fiber Communications group at Purdue. For his undergrad work, he
received the Deans Award for outstanding achievement in electrical and com-
puter engineering. He is a Member of both OSA and IEEE, and serves as a
Reviewer for Optics Letters,Optics Express,andPhotonics Technology Letters.
His research interests include optical pulse shaping, frequency comb generation,
and radio-frequency photonics.
Andrew M. Weiner received the Sc.D. degree in electrical engineering from
the Massachusetts Institute of Technology, Cambridge, MA, USA, in 1984. He
is the Scifres Family Distinguished Professor of Electrical and Computer En-
gineering. In 2008, he was elected to membership in the National Academy of
Engineering and in 2009 was named a Department of Defense National Security
Science and Engineering Faculty Fellow. He recently served a three year term as
the Chair of the National Academy’s U.S. Frontiers of Engineering Meeting; he
is currently serving as an Editor-in-Chief of Optics Express, an all-electronic,
open access journal publishing more than 3000 papers a year emphasizing in-
novations in all aspects of optics and photonics. He joined Bellcore, at that time
a premier telecommunications industry research organization, first as Member
of Technical Staff and later as a Manager of Ultrafast Optics and Optical Signal
Processing Research. He joined Purdue as a Professor in 1992, and has since
graduated more than 30 Ph.D. students. He has also spent sabbaticals at the
Max Born Institute for Nonlinear Optics and Ultrashort Pulse Spectroscopy,
Berlin, Germany, and at JILA, University of Colorado and National Institute of
Standards and Technology, Boulder, CO, USA.
His research interests include ultrafast optics, with a focus on processing
of extremely high speed lightwave signals. He is especially well known for
his pioneering work on programmable generation of arbitrary ultrashort pulse
waveforms, which has found application both in fiber optic networks and in
ultrafast optical science laboratories around the world.
Prof. Weiner is the author of a textbook entitled Ultrafast Optics, has pub-
lished eight book chapters and more than 270 journal articles, and is inventor of
15 U.S. patents. His numerous awards include the Hertz Foundation Doctoral
Thesis Prize in 1984, the Optical Society of America’s Adolph Lomb Medal
in 1990, and R.W. Wood Prize in 2008, the International Commission on Op-
tics Prize in 1997, and the IEEE Photonics Society’s William Streifer Scientific
Achievement Award in 1999 and Quantum Electronics Prize in 2011. At Purdue,
he has been recognized with the inaugural Research Excellence Award from the
Schools of Engineering in 2003, the Provost’s Outstanding Graduate Student
Mentor Award in 2008, and the Herbert Newby McCoy Award for outstanding
contributions to the natural sciences in 2013.
