10 August 2022
POLITECNICO DI TORINO
Repository ISTITUZIONALE
Narrow linewidth CW laser phase noisecharacterization methods for coherenttransmission system applications /
Camatel, S; Ferrero, Valter. - In: JOURNAL OF LIGHTWAVE TECHNOLOGY. - ISSN 0733-8724. - STAMPA. - 26:n.
17(2008), pp. 3048-3055. [10.1109/JLT.2008.925046]
Original
Narrow linewidth CW laser phase noisecharacterization methods for coherenttransmission system
applications.
Publisher:
Published
DOI:10.1109/JLT.2008.925046
Terms of use:
openAccess
Publisher copyright
(Article begins on next page)
This article is made available under terms and conditions as specified in the corresponding bibliographic description in
the repository
Availability:
This version is available at: 11583/1675733 since:
IEEE
3048 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 17, SEPTEMBER 1, 2008
Narrow Linewidth CW Laser Phase Noise
Characterization Methods for Coherent
Transmission System Applications
Stefano Camatel, Member, IEEE, and Valter Ferrero, Member, IEEE
Abstract—Several techniques for phase noise PSD measurement
of continuous wave (CW) lasers to be used in coherent transmis-
sion systems are analyzed. Between them, we evaluate two novel
techniques. The first employs a homodyne optical phase-locked
loop, while the second uses a signal source analyzer. Experimental
results obtained by these two methods are compared with clas-
sical linewidth measurement methods like self-heterodyne and
Michelson interferometer. Limits and accuracy of each method are
discussed. Furthermore, the comparison shows that, for coherent
transmission system applications, only a subset of the analyzed
methods is useful for laser phase noise characterization.
Index Terms—Coherent optical systems, homodyne detection,
phase-locked loops, phase measurement, phase noise.
I. INTRODUCTION
R
ECEIVER sensitivity limit in coherent optical communi-
cations is mainly affected by semiconductor laser phase
fluctuations [1], [2]. Also, sensors based on optical fiber inter-
ferometer systems have a sensitivity limited by phase noise [3].
For these reasons, many works focused on phase noise charac-
teristics of semiconductor lasers have been published [4], [5].
Measurement methods for laser linewidth characterization were
proposed in the past years; most of them are based on interfer-
ometer techniques [6]–[9].
Here, we analyze several phase noise measurement tech-
niques for testing lasers to be used in optical coherent communi-
cations. Common methods based on Michelson interferometer
(MI) and delayed self-heterodyne (DSH) measurements are
compared with two new techniques. The first novel method
[10] is able to retrieve the power spectral density (PSD) of the
overall phase noise produced by two lasers, a source laser and a
local oscillator (LO) laser. The combined phase noise can then
be used for the design and performance estimation of a coherent
transmission system. By the way, the CW source laser phase
noise PSD could be obtained by this method, if the used LO
Manuscript received September 5, 2007; revised March 20, 2008. Current
version published December 19, 2008. This work was supported by the BONE-
project (“Building the Future Optical Network in Europe”), a Network of Ex-
cellence funded by the European Commission through the 7th ICT-Framework
Programme, Italian Ministry of University and Research (MIUR), STORiCo
Project PRIN 2005.
S. Camatel was with the PhotonLab, Istituto Superiore Mario Boella, 10138
Torino, Italy. He is now with Nokia Siemens Networks (e-mail: stefano.ca-
matel@nsn.com).
V. Ferrero is with the PhotonLab, Dipartimento di Elettronica, Politecnico di
Torino, 10129 Torino, Italy (e-mail: valter.ferrero@polito.it).
Digital Object Identifier 10.1109/JLT.2008.925046
laser is affected by a negligible phase noise. Our measurement
technique is based on an optical phase-locked loop (OPLL)
which can be described by a linear model. In this paper, we
show experimental results obtained by using an OPLL based
on sub-carrier modulation (SC-OPLL) [10]. This way, we are
able to characterize CW lasers phase noise; optical oscillators
with direct frequency modulation are not required.
The second novel technique is based on a signal source ana-
lyzer (SSA) designed for phase noise characterization of radio
frequency (RF) oscillators. The optical signal generated by the
laser under test is converted into an RF signal through a self-het-
erodyne architecture; the phase noise of the resulting RF signals
is then characterized by a signal source analyzer and postpro-
cessed in order to obtain the PSD of the laser phase noise.
In order to compare the analyzed methods from the exper-
imental point of view, phase noise of two different optical
sources has been characterized by using the examined measure-
ment techniques, and the experimental results are presented in
this paper.
Section II introduces the notation definitions and expressions
needed to describe laser phase noise. The measurement tech-
nique based on an OPLL is described in Section III, while Sec-
tion IV introduces the signal source analyzer method. Section V
presents experimental results obtained by using a Michelson in-
terferometer and Section VI shows the linewidth measurements
performed through a self-heterodyne technique.
II. L
ASER PHASE NOISE MODEL
The electric field of an unmodulated optical signal emitted by
a single-mode semiconductor laser is
(1)
where
is the amplitude of the electric field and is a
random process that represents the phase noise.
Phase noise models usually consider three contributions to
the phase noise PSD: white, flicker and random walk noises.
The single-sided phase noise PSD
can then be expressed
as
(2)
where
is the Lorentzian spectral linewidth of the laser,
and are constants that give the strength of flicker frequency
noise and random walk frequency noise, respectively.
0733-8724/$25.00 © 2008 IEEE
CAMATEL AND FERRERO: NARROW LINEWIDTH CW LASER PHASE NOISE CHARACTERIZATION METHODS FOR COHERENT 3049
The relaxation frequency noise is not taken into account here
because it appears at very high frequencies, which are outside
the bandwidth of interest for optical coherent transmission sys-
tems.
In order to fit the experimental data obtained in the following
sections, a nonlinear least squares method was employed. The
nonlinear model used for such a fitting is based on a derivation
of (2)
(3)
where the function
is the single-sided phase noise PSD
measured in dBc and
.
III. OPLL M
EASUREMENT
METHOD
A. Principle of Operation
The novel measurement technique has been introduced the
first time in [10], where the flicker noise and random walk con-
tribution were not kept in account. The operation principle is
based on a linear OPLL [see Fig. 1(a)] that can be described
by a linear model. An exhaustive study of such a model was
presented in [1] and will be used as the starting point of the
following treatment. The source laser is not modulated and the
phase-lock to data crosstalk will not be taken into account. The
signal power level at the OPLL input is set, in order to have
shot noise and amplitude electrical noises negligible. This way,
the overall phase noise of source and local oscillator lasers is
the only contribution that will be considered. The fundamental
equation that allows evaluating the phase noise PSD is
(4)
where
is the Fourier transform of the phase error signal,
is a constant coefficient, is the phase noise Fourier
transform, and
is the PLL closed loop transfer func-
tion.
, as defined in [1], depends on photodiode respon-
sivity, transimpedance gain, received signal and local oscillator
powers. In the experimental setup, shown in Fig. 1(a), the elec-
trical spectrum analyzer (ESA) measures the phase error signal
power spectrum
expressed as
(5)
where
is the phase noise lasers PSD. The power
spectrum is related to the measurement performed by the spec-
trum analyzer of Fig. 1 through the equation
(6)
where
is the spectrum analyzer noise equivalent bandwidth
and
is the spectrum analyzer input impedance.
The PLL transfer function and the constant
can be mea-
sured in the experimental setup of Fig. 1(b), where the network
analyzer returns the following response:
(7)
Fig. 1. Setups for measurement of the phase error signal spectrum (a) and the
OPLL frequency response (b) necessary for the estimation of the phase noise
PSD through the OPLL method.
In (7), is the voltage that has to be applied to the phase
modulator in order to get a phase deviation of
radians. From
(7), it is possible to calculate the multiplication between the
squared constant
and the PLL transfer function factor
, which can be substituted in (5) obtaining the fol-
lowing formula:
(8)
Equation (8) returns the PSD of the phase noise lasers given
the measurement results obtained by the experiments shown in
Fig. 1.
B. Experimental Results
The previously described technique was implemented for
the characterization of two couples of external cavity tunable
lasers. The first couple includes two Agilent 81640A, while the
second couple consists of two Anritsu MG9638A. Declared
linewidths are lower than 100 kHz for the Agilent model, and
700 kHz for the Anritsu model. The OPLL employed for phase
noise measurement is an SC-OPLL [10]. The signal power at
the photodiode input was set to
16 dBm, while the overall LO
power was
3 dBm. The photodiode has responsivity equal
to 800 V/W. The optical voltage controlled oscillator (VCO)
includes a 10-GHz LiNbO
intensity modulator and a 6-GHz
electrical VCO.
For the couple of Agilent 81640A, the loop filter is a first
order active filter, whose time constants are
s and
s. Such time constants were chosen in order to
get a second order PLL transfer function with natural frequency
kHz and damping factor . The value of
kHz is the lowest natural frequency that allows OPLL
locking; thus, it affects the evaluation of
for low fre-
quency values. Indeed, the
estimation is accurate as long
as the electrical noise power level is negligible with respect to
the spectrum
of the phase error signal. usu-
ally has a bell-like shape centered around
and the white am-
plitude noise cannot be neglected for frequency values far from
3050 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 26, NO. 17, SEPTEMBER 1, 2008
Fig. 2. Measured phase noise PSD of Agilent 81640A tunable laser obtained
by OPLL method. A theoretical PSD with
=7
:
8
kHz,
k
=9
1
10
Hz
and
k
=0
Hz
is superimposed.
. Therefore, cannot be accurately estimated for fre-
quency values much lower or much higher than
. Such a fact
limits the frequency range on which
can be correctly
evaluated. The upper limit could be overcome by repeating the
measurement procedure for higher OPLL natural frequencies.
Anyway, the highest OPLL natural frequency that can be set
depends on the OPLL loop delay. Our SC-OPLL was affected
by a 15-ns feedback loop delay and 8 MHz is the maximum nat-
ural frequency for which SC-OPLL can still lock (see [10]). By
the way, we were able to measure
and estimate
on an acceptable range, so we performed the proposed measure-
ment just with
kHz.
The measurement setups shown in Fig. 1 were performed.
The electrical power spectrum was experimentally character-
ized by the ESA with repeated acquisitions on consecutive fre-
quency intervals. For each frequency interval ESA was set dif-
ferently:
kHz for kHz kHz,
kHz for kHz MHz, kHz for
MHz MHz, and MHz for MHz
MHz. Video bandwidth was always set one percent of
. The electrical spectrum analyzer of Fig. 1(a) was set with
a resolution bandwidth equal to 1 kHz and a video bandwidth of
100 Hz. The network analyzer of Fig. 1(b) generates a signal of
4 dBm and drives a LiNbO
phase modulator with .
As previously anticipated, the measurements of
and
allowed the phase noise PSD evaluation of two Agilent
81640A external cavity tunable lasers. Fig. 2 shows the average
phase noise PSD of the two Agilent 81640A lasers obtained by
half
.
A theoretical phase noise PSD
was computed applying
(2); a nonlinear least squares method was applied in order to fit
the measured curve of Fig. 2. Due to the limits of such measure-
ment method, previously described, we were not able to take
valid experimental data for high values of the phase noise PSD
at low frequencies. Experimental data were only taken in a fre-
quency range where random walk noise has not got any signifi-
cant effect on the measured curve; so we imposed
Hz .
The resulting fitted coefficients were: linewidth
kHz
6.2%, flicker noise coefficient Hz . The in-
dicated percentages specify the confidence bounds defined with
Fig. 3. Measured phase noise PSD of Anritsu MG9638A tunable laser obtained
by OPLL method. A theoretical PSD with
=
4
:
9
kHz,
k
=1
:
35
1
10
Hz
and
k
=10
Hz is superimposed.
a 95% level of certainty. The
estimate is not as accurate as
the linewidth value; a better estimation requires the measure-
ment of the phase noise PSD at lower frequencies.
For the couple of Anritsu MG9638A, the loop filter is still a
first order active filter, whose time constants are
s and
s, corresponding to natural frequency MHz
and damping factor
. According to the considerations
previously made, a higher
was set due to a higher phase noise
PSD of the Anritsu lasers at lower frequencies. Fig. 3 shows
the average phase noise PSD of the two lasers obtained by half
.
Fig. 3 shows also a theoretical phase noise PSD
com-
puted applying the same procedure previously described, ob-
taining the linewidth
kHz 6.4%, the flicker noise
coefficient
Hz . For the same con-
siderations made for the Agilent laser we imposed
Hz .
With respect to Agilent lasers, the phase noise PSD curve of the
Anritsu MG9638A is steeper at low frequencies. Even if Anritsu
and Agilent laser are characterized by almost the same amount
of white frequency noise, the behavior of the phase noise PSD
at low frequencies is much different due to flicker and random
walk contributions.
IV. RF S
IGNAL SOURCE ANALYZER
METHOD
A. Heterodyne Characterization
This method is based on the characterization of the electrical
signal obtained by the beating of two laser signals. Phase noise
characterization should actually be performed by RF instru-
ments designed for the analysis of electrical signal sources.
Actually, lasers are less stable in frequency than RF sources;
so we excluded the use of instruments whose measurement
is based on a simple electrical phase locked loop which fails
to measure phase noise of relatively drifty and noisy signal
sources. We tried to use an Agilent E5052 which employs a
heterodyne discriminator method in order to measure relatively
large phase noise of unstable signal sources. Even if such
instrument is more tolerant to frequency drifting signals, it was
not able measure lasers phase noise. Thus, we will not present
CAMATEL AND FERRERO: NARROW LINEWIDTH CW LASER PHASE NOISE CHARACTERIZATION METHODS FOR COHERENT 3051
Fig. 4. Heterodyne setup for the measurement of the combined phase noise
PSD of two lasers based on a signal source analyzer.Narrow Linewidth CW
Laser Phase Noise Characterization Methods for Coherent
any measurement result got with this method but its description
is useful in order to understand the next measurement method.
The experimental setup is illustrated in Fig. 4: both lasers
are tuned appropriately and their optical frequencies are kept
constant during the measurement. The polarization controller
is used to align lasers’ polarization state at the photodetector
input. The coupler combines the two optical fields, obtaining
half inputs total power to each output port. The upper output port
is connected to a photodiode that detects the interference beat
tone between the two lasers, converting it into an electrical tone.
Note that lasers frequency must be tuned in order to have the
mixing product in the photodetector and signal source analyzer
electrical bandwidth. The lasers wavelength tuning operation
is monitored by using an OSA. The resulting heterodyne beat
signal is described by
(9)
where the resulting phase noise is given by
(10)
The resulting signal is a sine wave in the RF domain where the
overall electrical phase noise
is due to the lasers optical
phase noise contributions. Such a electrical noise can be char-
acterized employing the phase noise PSD measurement tech-
niques for RF oscillators, i.e. by using a RF signal source an-
alyzer. Actually such a measurement returns the overall phase
noise generated by both lasers of Fig. 4 being the two processes
and uncorrelated. If we are interested in laser 1
phase noise PSD contribution only, it could be evaluated em-
ploying a laser 2 (Local Oscillator) with phase noise PSD neg-
ligible with respect to the one of laser 1, so the resulting phase
noise may be approximated as
.
Theexperimentalsetupwas builtusinganAgilentE5052signal
source analyzer. Such instrument is able to characterize the phase
noise of RF signals if oscillator frequencyis sufficiently stable. In
fact, it is able to track frequency variations in a range lower than
3 MHz. Unfortunately, frequency stability of the heterodyne beat
signal was much worse; the instrument was not able to track the
input signal and measurement could not be performed.
In order to solve the problem of frequency stability, the exper-
imental setup of Fig. 4 was modified and a self-heterodyne mea-
surement was performed as described in the following section.
B. Correlated Delayed Self-Heterodyne Characterization
This method is obtained by substituting the heterodyne archi-
tecture of Fig. 4 with the delayed self-heterodyne one as shown
Fig. 5. Delayed self-heterodyne setup for the phase noise PSD measurement
of a laser based on a signal source analyzer.
in Fig. 5. Correlated delayed self-heterodyne technique employs
just one laser for RF beat signal generation. The optical source
signal is split into two paths by the first splitter. The optical fre-
quency of one arm is offset with respect to the other of
. Un-
like the traditional delayed self-heterodyne technique described
in Section VI, here the delay
of one path must be much lower
than the coherence time of the source laser in order to solve
the frequency instability problem that affects the previously de-
scribed method. Thus, the two combining beams are not statis-
tically independent. The beat tone produced is displaced from
0Hzto
thanks to the frequency shift. A signal source an-
alyzer measures the RF beat tone phase noise PSD, which is
broadened by the laser linewidth. Actually, phase noise infor-
mation is translated from optical frequencies to RF frequencies
where electronics instrumentation operates.
The signal at the photodetector output is given by
(11)
where the resulting phase noise is given by
(12)
Note that (11) does not include any term regarding the phase
noise introduced by the frequency shifter. Since the laser phase
noise is much higher than the electrical oscillator in the fre-
quency shifter, this assumption is easily satisfied. Anyway, the
amplitude of
is lower as delay decreases and can
not be set to too small values.
From (12), the relation between the Fourier transforms of
and is derived as
(13)
A signal source analyzer characterizes the phase noise PSD
. The phase noise PSD is related to the measured
by the following equation:
(14)
In order to verify the right setting of the delay
, the measured
must be much higher than the signal phase noise PSD
used for the frequency shifter operations
The experimental setup of Fig. 5 was built using an acousto-
optic modulator driven by a 27-MHz electrical oscillator. An
Agilent E5052 signal source analyzer characterized the elec-
trical signal phase noise PSD. The delay
was set to 6 ns. By
using the measured
, laser phase noise PSD was com-
puted by (14). The Agilent 81640A tunable laser was charac-
terized and the result is plotted in Fig. 6. A fitting curve was
