You are on page 1of 24

Source: Electronic Instrument Handbook

Chapter

Lightwave Signal Analysis


23
Waguih Ishak
Agilent Technologies
Palo Alto, California

23.1 Introduction
High-capacity digital transmission systems and analog-modulated microwave
frequency systems based on fiber-optic (FO) systems have emerged as very
competitive to conventional communication systems. A variety of lightwave
components have been developed to support these high-speed systems. Most
notable among these components are semiconductor lasers, specifically single-
frequency distributed feedback (DFB) lasers, and high-speed pin photode-
tectors.
The design of the communications systems depends heavily on the
characteristics of the lightwave signal propagating through the various
components of the system. A lightwave signal is an electromagnetic wave in the
wavelength range between 200 and 2000 nm, i.e., a frequency range between
150 and 1500 THz. This signal is generated by the laser in the transmission
section of the system. It is critical to the success of the system to use high-
quality lasers (high power, large modulation bandwidth, low noise, etc.). In
addition, as this signal propagates through the system, its characteristics (power,
polarization, etc.) change. Accurate methods of measuring the characteristics
(power, modulation speed, spectrum, linewidth, state of polarization, etc.) of a
lightwave signal or, alternatively, analyzing the lightwave signal are therefore
essential for the design of high-capacity fiber-optic transmission systems.

23.2 Light Sources


The term laser stands for “light amplification by stimulated emission of
radiation.” We focus the discussion in this section on semiconductor laser diodes

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com) 23.1


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.2 Chapter Twenty-Three

Figure 23.1 Spontaneous emission from semiconductor lasers.

(LDs) which are now being routinely used in FO systems because of their small
size, high efficiency and reliability, and excellent control of wavelength, power,
and spectral characteristics. Because the light output from a laser is coherent,
LDs are used for applications requiring high data rates (in excess of 100 Mbit/
s). On the other hand, the incoherent nature of the output signal from an LED
limits their applications to data rates less than 100 Mbit/s. While the structure
and the fabrication process of LEDs and LDs are very similar, their light output
characteristics are quite different.

23.2.1 Light-emitting diodes (LEDs)


LEDs produce light with a wide spectral width, and when used in FO
communications systems, they can be modulated at frequencies up to 100 MHz.
LEDs have the advantages of low temperature sensitivity and no sensitivity to
back reflections. Furthermore, LEDs produce incoherent light output which is
not sensitive to optical interference from reflections.
LEDs generate light by spontaneous emission. This occurs when an electron
in a high-energy conduction band changes to a low-energy valence band, as
shown in Fig. 23.1. The energy lost by the electron, the bandgap energy E is
g
released as a photon, the entity of light. The energy of the released photon is
equal to the energy lost by the electron, and the wavelength of the emitted
photon is a function of its energy. Because different materials have different
orbital states which determine the energy levels of the various electrons, the
wavelength of the emitted photon is determined by the material used to make
the LED.
The wavelength of the emitted photon is given by

(23.1)

where h=Planck’s constant=6.62×10-34 W/s2


c=speed of light=2.998×108 m/s
Eg=bandgap of material, joules
For a semiconductor material with Eg=0.9 eV, the wavelength of the emitted
photons will be about 1.38 µm. The most commonly used materials for LEDs are

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.3

gallium arsenide (GaAs) with Eg=1.42 eV and gallium phosphide (GaP) with
Eg=2.24 eV. By adding other materials, such as aluminum or indium, to the
GaAs or the GaP, it is possible to tailor the bandgap energy to achieve any
wavelength in the 0.5- to 2.0-µm range.
With appropriate n and p doping, the above materials can be used to form a
simple pn diode which can function as an LED. By forward biasing the pn junction
of the diode, conduction-band electrons are generated. For a better confinement
of the output optical power, a double heterostructure (DH) LED is used. In a
DH-LED, the junction is formed by dissimilar semiconductor materials with
different bandgap energy and refractive index values and the free charges are
confined to recombine in a narrow, well-defined semiconductor layer (called the
active layer).
The spectrum of an LED has a broad distribution of wavelength centered
around a wavelength calculated from the above equation. The spectral width is
often specified as the full width at half maximum (FWHM) of the distribution.
Typical values for the FWHM range from 20 to 100 nm.

23.2.2 LED parameters


Total power. The summation of the power at each trace point, normalized by
the ratio of the trace point spacing and resolution bandwidth. This normalization
is required because the spectrum of the LED is continuous, rather than
containing discrete spectral components (as a laser does).

(23.2)

Mean (FWHM). These wavelengths represent the center of mass of the trace
points. The power and wavelength of each trace point are used to calculate the
mean (FWHM) wavelength.

(23.3)

Sigma. An rms calculation of the spectral width of the LED based on a gaussian
distribution. The power and wavelength of each trace point are used to calculate
sigma.

(23.4)

FWHM (Full Width at Half Maximum). Describes the spectral width of the half-
power points of the LED, assuming a continuous, gaussian power distribution.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.4 Chapter Twenty-Three

The half-power points are those where the power spectral density is one-half
that of the peak amplitude.

(23.5)

3-dB width. Used to describe the spectral width of the LED based on the
separation of the two wavelengths that each have power spectral density equal
to one-half the peak power spectral density. The 3-dB width is determined by
finding the peak of the LED spectrum and dropping down 3 dB on each side.

Mean (3 dB). The wavelength that is the average of the two wavelengths
determined in the 3-dB width measurement.

Peak wavelength. The wavelength at which the peak of the LED’s spectrum
occurs.

Density (1 nm). The power spectral density (normalized to a 1-nm bandwidth)


of the LED at the peak wavelength.

Distribution trace. A trace can be displayed that is based on the total power,
power distribution, and mean wavelength of the LED. This trace has a gaussian
spectral distribution and represents a gaussian approximation to the measured
spectrum.

23.2.3 Fabry-Perot (FP) laser diodes


Lasers are capable of producing high output powers and directional beams. When
used in fiberoptic communication systems, semiconductor lasers can be
modulated at rates up to about 10 GHz. However, lasers are sensitive to high
temperature and back reflections. Additionally, the coherent emitted light is
sensitive to optical interference from reflections. Two different types of
semiconductor lasers are commonly used in FO communications systems: Fabry-
Perot (FP) and distributed feedback (DFB) lasers.
The Fabry-Perot laser is different from an LED in that it generates light by
stimulated emission, where photons trigger additional electron-hole recom-
binations, resulting in additional photons as shown in Fig. 23.2. A stimulated

Figure 23.2 Stimulated emission from semiconductor lasers.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.5

photon travels in the same direction and has the same wavelength and phase as
the photon that triggered its generation. Stimulated emission can be thought of
as amplification of light. As one photon passes through the region of holes and
conduction band electrons, additional photons are generated. If the material is
long enough, enough photons can be generated to produce a significant amount
of power at a single wavelength.
An easier way to build up power is to place a reflective mirror at each end of
the region where the photons multiply so that they can travel back and forth
between the two mirrors, building up the number of photons with each trip.
These mirrors form the resonator needed for the operation of the laser.
Additionally, in order for the laser action to occur, a greater number of conduction
band electrons than valence band electrons must be present. Called population
inversion, this is achieved by forcing a high current density into the active region
of the laser diode.
The possible wavelengths produced by the resonator are given by

(23.6)

where m=integer
c=speed of light
l=length of resonator
n=refractive index of laser cavity

The mode spacing, which is the separation between the different wavelengths,
is determined as

(23.7)

23.2.4 FP laser parameters


Total power. The summation of the power in each of the displayed spectral
components or modes that satisfy the peak excursion criteria (defined below).

Mean wavelength. Represents the center of mass of the spectral components on


screen. The power and wavelength of each spectral component are used to
calculate the mean wavelength.

(23.8)

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.6 Chapter Twenty-Three

Sigma. An rms calculation of the spectral width of the FP laser based on a


gaussian distribution.

(23.9)

FWHM. Describes the spectral width of the half-power points of the FP laser,
assuming a continuous, gaussian power distribution. The half-power points are
those where the power spectral density is one-half that of the peak amplitude.

(23.10)

Mode spacing. The average wavelength spacing between the individual spectral
components of the FP laser.

Peak amplitude. The power level of the peak spectral component of the FP laser.

Peak wavelength. This is the wavelength at which the peak spectral component
of the FP laser occurs.

Peak excursion. The peak excursion value (in dB) can be set by the user and is
used to determine which on-screen responses are accepted as discrete spectral
responses.

23.2.5 Distributed feedback (DFB) lasers


DFB lasers are similar to FP lasers, except that all but one of their spectral
components are significantly reduced. Because its spectrum has only one line,
the spectral width of the DFB laser is much less that of an FP laser. This greatly
reduces the effect of chromatic dispersion in fiberoptic systems, allowing for
greater transmission bandwidths.
The distributed feedback laser utilizes a grating, a series of corrugated
ridges, along the active layer of the semiconductor, as shown in Fig. 23.3.
Rather than using just the two reflecting surfaces at the ends of the diode, as a
Fabry-Perot laser does, the distributed feedback laser uses each ridge of the
corrugation as a reflective surface. At the resonant wavelength, all reflections
from the different ridges add in phase. By having much smaller spacings
between the resonator elements, compared to the Fabry-Perot laser, the
possible resonant wavelengths are much farther apart in wavelength, and only
one resonant wavelength is in the region of laser gain. This results in the
single laser wavelength.
The ends of the diode still act as a resonator, however, and produce the

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.7

Figure 23.3 Distributed feedback laser.

lower-amplitude side modes. Ideally, the dimensions are selected so that the
end reflections add in phase with the grating reflections. In this case the main
mode will occur at a wavelength halfway between the two adjacent side modes;
any deviation is called a mode offset. Mode offset is measured as the difference
between the main-mode wavelength and the average wavelength of the two
adjacent side modes.
The amplitude of the largest side mode is typically between 30 and 50 dB
lower than the main spectral output of the laser. Because side modes are so
close to the main mode (typically between 0.5 and 1 nm) the dynamic range of
an optical spectrum analyzer determines its ability to measure them. Dynamic
range is specified at offsets of 0.5 and 1.0 nm from a large response. Hewlett-
Packard optical spectrum analyzers, for example, specify a dynamic range of -
55 dBc at offsets of 0.5 nm and greater and –60 dBc at offsets of 1.0 nm and
greater. This indicates the amplitude level of side modes that can be detected at
the given offsets.

23.2.6 DFB laser parameters


Peak wavelength. The wavelength at which the main spectral component of the
DFB laser occurs.

Side-mode suppression ratio (SMSR). The amplitude difference between the


main spectral component and the largest side mode.

Mode offset. Wavelength separation (in nanometers) between the main spectral
component and the SMSR mode.

Peak amplitude. The power level of the main spectral component of the DFB
laser.

Stop band. Wavelength spacing between the upper and lower side modes
adjacent to the main mode.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.8 Chapter Twenty-Three

Center offset. Indicates how well the main mode is centered in the stop band.
This value equals the wavelength of the main spectral component minus the
mean of the upper and lower stop band component wavelengths.

Bandwidth. Measures the displayed bandwidth of the main spectral component


of the DFB laser. The amplitude level, relative to the peak, that is used to measure
the bandwidth can be set by the user.

Peak excursion. The peak excursion value (in dB) can be set by the user and is
used to determine which three on-screen responses will be accepted as discrete
spectral responses. To be counted, the trace must rise and then fall by at least
the peak excursion value about a given spectral component. Setting the value
too high will result in failure to count small responses near the noise floor.

23.3 Optical Power Meters


Power is the most basic and common measurement for optical signals. It is
commonly measured using a power meter utilizing a photodetector with
appropriate bias circuits. The strongest concern in power measurements is
accuracy.

23.3.1 Types of optical power meters


If the signal to be measured is a parallel beam of light, a large-area (about 5 mm
in diameter) photodetector is usually used. The large area of the photo-detector
can capture the several modes (from a fiber cladding, for example) composing
the optical signal. However, it is important to know the wavelength of the
signal before an accurate estimate of the power can be made. This is because
photodetectors exhibit a wavelength-dependent responsivity. The
disadvantage of this method is the presence of a large dark current (current
generated in the photodetector with no signal present) for the large-area
photodetector.
Another method of measuring optical power is to focus the optical signal into
a tiny spot (less than 1 mm) on a small-area photodetector. The advantages are
higher sensitivity and lower cost. However, it is difficult to calibrate such a
system, and small-area photodetectors usually result in nonuniform signal
distribution.
Wavelength-independent detectors (such as pyroelectric radiometers and
thermopiles) can be used with parallel beam signals to measure the optical power.
This method is indirect and is based on measuring the temperature change of
the detector as the optical signal falls onto it. One major advantage is the
wavelength insensitivity, but the sensitivity is inferior relative to the
photodetector-based methods.
A problem common to all methods is reflection. Multiple reflections from

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.9

Figure 23.4 Power meter block diagram.

surfaces within the power meter can cause severe inaccuracies, and commercial
power meters must provide means of eliminating or correcting for this effect. A
simple yet effective method to solve this problem is to tilt the photo-detector
surface relative to the normal to the plane of incidence and to use antireflection
coatings as shown in Fig. 23.4. An antireflection coating usually consists of several
layers of thin dielectric films. The thicknesses and the dielectric constants of
the various films are adjusted so that reflections from the films will add
constructively in the transmission direction (the direction of the incident optical
signal) and destructively in the reflection direction, thus eliminating reflected
signals from reaching the device under test.

23.3.2 Accuracy of optical power meters

A general-purpose power meter can measure both absolute and relative optical
power. The critical specification in absolute optical power measurements is the
uncertainty of the displayed power. The more accurate the measurements of
power, the higher the confidence in component verification during incoming
inspection, production, installation, and maintenance. In these areas, a more
accurate power meter will increase yield and reduce cost. In order to achieve
accurate absolute power measurements which are traceable to national
standard laboratories, the detector noise and the postamplifier drift must be
minimized. This can be achieved by mounting the photodetector and the
transimpedance amplifier on a thermoelectric cooler inside a hermetically
sealed package. By keeping the temperature of the photodetector constant, the
error caused by the variation of the quantum efficiency of the photodetector is
almost eliminated. This results in a low-noise power measurement. In addition,
by cooling the detector, the noise equivalent power is kept to a minimum.
Furthermore, by keeping the temperature of the amplifier constant, the drift is
minimized.
Relative power measurement is useful in determining the insertion loss of
optical components. For example, in determining the insertion loss of fiber
connectors, the light out of one connector is taken as the reference; then the
connectors are mated and the measurement is taken at the far end of the second

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.10 Chapter Twenty-Three

fiber. In order to detect very small changes in loss (for example, when testing
very low loss components), very high power meter resolution is required. State-
of-the-art power meters have 0.001 dB resolution. The key to achieving high
resolution is the design of the photodetector and its associated circuity (bias
and amplifier).
On the other extreme, a high dynamic range power meter is needed to measure
high losses (for example, in characterizing long fibers). The linearity of the
detection circuit determines the limit on the dynamic range. It is crucial to
make sure that the gain of the postamplifier chain is matched to avoid
nonlinearities.

23.3.3 Specifications of optical power meters


Wavelength range. The range of optical wavelength over which the rest of the
specifications are valid. This parameter is dependent upon the type of
photodetector used inside the power meter. For silicon detectors, the wavelength
range is between 450 and 1000 nm. For indium gallium arsenide detectors, the
wavelength range is 800 to 1700 nm.

Power range. The minimum and maximum detectable power levels. Typical
range from –100 to +3 dBm.

Resolution. The display resolution in dB or watts. Typically 0.001 dB or 10 pW.

Noise. The internal noise generated by the power meter. Typically 1 to 50 pW.

23.4 Lightwave Signal Analyzers


Lightwave signal analyzers are useful for measuring the optical signal strength
and distortion, modulation depth and bandwidth, intensity noise, and
susceptibility to reflected light. However, these analyzers are quite different
from the optical spectrum analyzers discussed later in this chapter. The optical
spectrum analyzers show the spectral distribution of average optical power and
are useful for observing the modes of multimode optical sources or the side-lobe
rejection of single-mode sources. Their measurement resolution is typically 0.1
nm or about 18 GHz at a wavelength of 1300 nm. On the other hand, the
lightwave signal analyzer displays the total average power and the modulation
spectrum but provides no information about the wavelength of the optical signal.
This distinction is shown in Fig. 23.5.

23.4.1 Block diagram


A simplified block diagram for a high-speed lightwave signal analyzer is
shown in Fig. 23.6. The input lightwave signal passes through an optical

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.11

Figure 23.5 Differences between a


lightwave signal analyzer and optical
spectrum analyzer.

attenuator to prevent overloading of the front end. The high-speed photode-


tector converts the optical signal to an electrical signal. The time-varying
component of the photocurrent, which represents the demodulated signal, is
fed through the preamplifier to the input of a conventional microwave
spectrum analyzer. The dc portion of the photocurrent is fed to a power meter
circuit.

23.4.2 Measured parameters


A key parameter in any lightwave system is the modulation bandwidth of the
optical source of the system. Current-modulated semiconductor lasers have
bandwidths approaching 20 GHz. A high-speed lightwave signal analyzer can
measure the intensity modulation frequency response of these lasers provided
the photodetector and the microwave spectrum analyzer (shown in Fig. 23.6)
have wider bandwidth than the device under test. Another key parameter, for a
number of reasons, is the laser noise spectrum. It obviously impacts the signal-
to-noise ratio in a fiber-optic transmission system. Furthermore, it can be shown
that the intensity noise spectrum has the same general shape as the intensity

Figure 23.6 Lightwave signal analyzer block diagram.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.12 Chapter Twenty-Three

Figure 23.7 Relative intensity noise of a semiconductor laser.

modulation response and can be used as an indicator of potential modulation


bandwidth. Again, a high-speed lightwave signal analyzer can be very useful in
measuring the laser noise spectrum.
The laser intensity noise spectrum can be generally affected by both the
magnitude and the polarization of the optical power that is fed back to the
laser. This is called reflection-induced noise and is typically caused by reflections
from optical connectors. This reflected power upsets the dynamic equilibrium of
the lasing process and typically increases the amplitude of the intensity noise.
It can also induce a ripple on the spectrum. The lightwave signal analyzer is
ideal for measuring the laser intensity noise. It should be mentioned here that
while it is possible to use an optical time domain reflectometer to measure the
magnitude of an optical reflection, the lightwave signal analyzer is the only
instrument that can measure the effect of these reflections on the noise
characteristics of the laser under test.
Perhaps one of the most important parameters of laser sources to be measured
by a lightwave signal analyzer is the relative intensity noise (RIN). It is the
ratio of the optical noise power to the average optical power and is an indication
of the maximum possible signal-to-noise ratio in a lightwave system, where the
dominant noise source is the laser intensity noise. Figure 23.7 shows an example
of the relative intensity noise measurement of a high-speed semiconductor laser.

23.5 Optical Spectrum Analyzers


Optical spectrum analysis is the measurement of the optical power as a
function of the wavelength. This is an important measurement for optical
sources such as lasers and LEDs. The spectral width of a light source is an

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.13

important parameter in fiber-optic communications systems because of


chromatic dispersion, which occurs in the fiber and limits the modulation
bandwidth of the system, as described earlier. The effect of chromatic
dispersion can be seen in the time domain as pulse broadening of a digital
waveform. Since chromatic dispersion is a function of the spectral width of the
light source, narrow spectral widths are desirable for high-speed comm-
unications systems.
Because there are many different types of light sources, with varying spectra,
there is a need to have an instrument capable of measuring the spectrum of
single-mode lasers (such as a DFB or a DBR laser) or a multimode laser (such
as a FP laser). This instrument is called an optical spectrum analyzer. It is
probably the most common instrument used by researchers and manufacturing
engineers who are involved in optical measurements.
Optical spectrum analyzers (OSAs) can be divided into three categories:
diffraction-grating–based and two interferometer–based architectures, the
Fabry-Perot and Michelson interferometer–based OSAs. Diffraction-grating–
based OSAs are capable of measuring spectra of lasers and LEDs. The resolution
of these instruments is variable, typically ranging from 0.1 to 10 nm. Fabry-
Perot (FP) interferometer–based OSAs have a fixed, narrow resolution, typically
specified in frequency, between 100 MHz and 10 GHz (i.e., 0.01 to 0.1 nm at the
1.55-µm wavelength range). This narrow resolution allows them to be used in
measuring laser chirp but can limit their measurement spans much more than
the diffraction-grating–based OSAs. Michelsoninterferometer–based OSAs
display the resolution by calculating the Fourier transform of a measured
interference pattern.

23.5.1 Block diagram


The basic block diagram of an optical spectrum analyzer is shown in Fig. 23.8.
The incoming optical signal passes through a wavelength-tunable optical filter

Figure 23.8 Simplified block diagram of an optical spectrum analyzer.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.14 Chapter Twenty-Three

Figure 23.9 Fabry-Perot–based optical spectrum analyzer.

(monochromator or interferometer) which resolves the individual spectral


components. The photodetector then converts the optical signal into an electric
signal proportional to the incident optical power. An exception to this description
is the Michelson interferometer, which is not actually an optical filter.
The current from the photodetector is then converted to a voltage by the
transimpedance amplifier and then digitized using an A/D converter. The
signal is applied to the display as the vertical data and a ramp generator
determines the horizontal location of the trace as it sweeps from left to right.
This ramp is used to tune the optical filter so that its resonance wavelength is
proportional to the horizontal position. A trace of optical power vs. wavelength
results. The displayed width of each of the modes of the optical signal under
test is a function of the spectral resolution of the optical filter. Therefore, the
resolution of the OSA is mainly determined by the characteristics of the
tunable optical filter.

23.5.2 FP interferometer–based optical spectrum analyzers


The FP interferometer, shown in Fig. 23.9, consists of two highly reflective,
parallel mirrors that act as a resonant cavity which filters the incoming optical
signal. Because the spacing between the mirrors is essentially fixed, the
resolution of an FP interferometer–based OSA is typically fixed. The wavelength
of the cavity is varied by slightly changing the spacing between the mirrors
from its nominal value by a very small amount.
The advantage of the Fabry-Perot interferometer is its very narrow spectral
resolution, which allows it to measure laser chirp. The major disadvantage is
that at any one position multiple wavelengths will be passed by the filter. (The
spacing between these responses is called the free spectral range.) This problem
can be solved by placing a monochromator (a wavelength-tunable narrowband
optical filter) in cascade with the Fabry-Perot interferometer to filter out all

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.15

Figure 23.10 Michelson-interferometer–based optical spectrum


analyzer.

power outside the interferometer’s free spectral range about the wavelength of
interest.

23.5.3 Michelson-interferometer–based optical spectrum analyzers


The Michelson interferometer, shown in Fig. 23.10, is based on creating an
interference pattern between the signal and a delayed version of itself. The
power of this interference pattern is measured for a range of delay values. The
resulting waveform is the autocorrelation function of the input signal. This
enables the Michelson-interferometer–based spectrum analyzer to make direct
measurements of coherence length, as well as very accurate wavelength
measurements. Other types of optical spectrum analyzers cannot make direct
coherence-length measurements.
To determine the power spectra of the input signal, a Fourier transform is
performed on the autocorrelation waveform. Because no real filtering occurs,
Michelson-interferometer–based optical spectrum analyzers cannot be put in a
span of zero nanometers, which would be useful for viewing the power at a
given wavelength as a function of time. This type of analyzer also tends to
have less dynamic range than diffraction-grating–based optical spectrum
analyzers.

23.5.4 Diffraction-grating–based optical spectrum analyzers


Monochromators. The most common optical spectrum analyzers use
monochromators as the tunable optical filter. In the monochromator, a diffraction
grating (a mirror with finely spaced corrugated lines on the surface) separates
the different wavelengths of light. The result is similar to that achieved with a

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.16 Chapter Twenty-Three

Figure 23.11 Prism-based optical spectrum analyzer.

prism. Figure 23.11 shows what a prism-based optical spectrum analyzer might
look like. The prism separates the different wavelengths of light, and only the
wavelength that passes through the aperture reaches the photodetector. The
angle of the prism determines the wavelength to which the optical spectrum
analyzer is tuned, and the size of the aperture determines the wavelength
resolution. Diffraction gratings are used instead of prisms because they provide
a greater separation of wavelengths, with less attenuation. This allows for better
wavelength resolution.
A diffraction grating is a mirror with grooves on its surface, as shown in Fig.
23.12. The spacing between grooves is extremely narrow, approximately equal
to the wavelengths of interest. When a parallel light beam strikes the diffraction
grating, the light is reflected in a number of directions.
The first reflection is called the zero-order beam (m=0), and it reflects in the
same direction as it would if the diffraction grating were replaced by a plane
mirror. This beam is not separated into different wavelengths and is not used
by the optical spectrum analyzer.
The first-order beam (m=1) is created by the constructive interference of
reflections off each groove. For constructive interference to occur, the pathlength
difference between reflections from adjacent grooves must equal one wavelength.

Figure 23.12 Diffraction grating principle of operation.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.17

Figure 23.13 Diffraction-grating–based optical spectrum analyzer.

If the input light contains more than one wavelength component, the beam will
have some angular dispersion; that is, the reflection angle for each wavelength
must be different in order to satisfy the requirement that the pathlength
difference of adjacent grooves is equal to one wavelength. Thus the optical
spectrum analyzer separates different wavelengths of light.
For the second-order beam (m=2), the path-length difference from adjacent
grooves equals two wavelengths. A three-wavelength difference defines the third-
order beam, and so on. Optical spectrum analyzers utilize multiple-order beams
to cover their full wavelength range with narrow resolution.
Figure 23.13 shows the operation of a diffraction-grating–based optical
spectrum analyzer. As with the prism-based analyzer, the diffracted light passes
through an aperture to the photodetector. As the diffraction grating rotates, the
instrument sweeps a range of wavelengths, allowing the diffracted light—the
particular wavelength depends on the position of the diffraction grating—to
pass through to the aperture. This technique allows the coverage of a wide
wavelength range.
Diffraction-grating–based optical spectrum analyzers contain either a single
monochromator, a double monochromator, or a double-pass monochromator.
Figure 23.14 shows a single-monochromator–based instrument. In these
instruments, a diffraction grating is used to separate the different wavelengths
of light. The second concave mirror focuses the desired wavelength of light at

Figure 23.14 Single-pass grating-based


optical spectrum analyzer.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.18 Chapter Twenty-Three

the aperture. The aperture width is variable and is used to determine the
wavelength resolution of the instrument.

Double monochromators. Double monochromators, as shown in Fig. 23.15, are


sometimes used to improve on the dynamic range of single-monochromator
systems. Double monochromators are equivalent to a pair of sweeping filters.
While this technique improves dynamic range, double monochromators typically
have reduced span widths due to the limitations of monochromator-to-
monochromator tuning match; double monochromators also have degraded
sensitivity due to losses in the monochromators.
The double-pass monochromator provides the dynamic-range advantage of
the double monochromator and the sensitivity and size advantages of the single
monochromator. Figure 23.15 shows the double-pass monochromator. The first
pass through the double-pass monochromator is similar to conventional single
monochromator systems. In Fig. 23.15, the input beam (1) is collimated by the
optical element and dispersed by the diffraction grating. This results in a spatial
distribution of the light, based on wavelength. The diffraction grating is
positioned such that the desired wavelength (2) passes through the aperture.
The width of the aperture determines the bandwidth of wavelengths allowed to
pass to the detector. Various apertures are available to provide resolution
bandwidths of 0.08 and 0.1 to 10 nm in a 1, 2, 5 sequence. In a single-
monochromator instrument, a large photodetector behind the aperture would
detect the filtered signal.
The system shown in Fig. 23.15 is unique in that the filtered light (3) is sent
through the collimating element and diffraction grating for a second time. During
this second pass through the monochromator, the dispersion process is reversed.
This creates an exact replica of the input signal, filtered by the aperture. The
small resultant image (4) allows the light to be focused onto a fiber which carries
the signal to the detector. This fiber acts as a second aperture in the system.
The implementation of this second pass results in the high sensitivity of a single
monochromator, the high dynamic range of a double monochromator, as well as
polarization insensitivity (due to the half-wave plate).

23.5.5 Specifications of optical spectrum analyzers

Wavelength tuning. This parameter is controlled by the rotation of the diffraction


grating. Each angle of the diffraction grating causes a corresponding wavelength
of light to be focused directly at the center of the aperture. The diffraction-
grating angle must be precisely controlled and very repeatable over time to
assure accurate tuning.

Wavelength repeatability. This parameter specifies the wavelength tuning drift


in a 1-min period. This is specified with the OSA in a continuous sweep mode
and with no changes made to the tuning.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.19

Figure 23.15 Double-pass grating-based optical spectrum analyzer. (a) Schematic drawing; (b)
implementation in instrumentation.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.20 Chapter Twenty-Three

Wavelength resolution bandwidth. This parameter determines the ability of


the OSA to display two signals closely spaced in wavelength as two distinct
responses. It is determined by the bandwidth of the optical filter and is defined
as the filter bandwidth at the half-power level, referred to as the FWHM. Typical
values for the wavelength resolution bandwidth vary from 0.08 to 5 nm.

Dynamic range. This parameter is commonly specified at 0.5- or 1.0-nm offsets


from the main response. A -60 dB dynamic range specification at 1 nm and
greater indicates that the OSA’s response to a purely monochromatic signal
will be -60 dBc or less at offsets of 1.0 nm or greater.

Sensitivity. This parameter is defined as the minimum detectable signal or,


more specifically, 6 times the rms noise level of the instrument. This is one
difference between OSAs and rf and microwave spectrum analyzers, whose
sensitivities are defined as the average noise level.

Tuning speed. This parameter depends on the sensitivity scale used and the
area of the photodetector.

Polarization insensitivity. This parameter represents how much reflection loss


results when light signals of different polarization states fall onto the diffraction
grating.

23.6 Linewidth Measurements


In high-speed fiber-optic communications systems, and to minimize the
transmission penalties resulting from the dispersion in long fiber links, high-
quality lasers are needed. These lasers should operate in a single longitudinal
mode (i.e., single-frequency oscillation) with minimal dynamic linewidth
broadening (i.e., frequency chirp) under modulation. In coherent and
wavelength division multiplexing communications systems, the lasing
linewidth becomes an important determinant of system performance. It is
therefore important to be able to measure the linewidth of single-frequency
lasers accurately.
A simple block diagram capable of measuring the linewidth of a wide variety
of laser sources is shown in Fig. 23.16 and is called a delayed self-homodyne

Figure 23.16 Delayed self-homodyne linewidth measurement system.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.21

measurement system. A frequency discriminator is used to convert the optical


phase (or frequency) deviation of the signal under test into intensity variation,
which can be detected using the high-speed lightwave signal analyzer. The
frequency discriminator is essentially an unbalanced fiber-optic interferometer
with a directional coupler at the input. The directional coupler splits the
incoming optical signal into two equal parts. The two signals then travel along
separate fiber paths where they experience a differential delay. The two signals
are then recombined using another directional coupler at the output of the
interferometer. Since the fiber-optic interferometer does not preserve the
polarization state, a polarization controller is added to one arm of the
interferometer. By adjusting the polarization controller, it is possible to
maximize the interference between the two signals at the output of the
interferometer.
The minimum linewidth which can be measured using the circuit shown in
Fig. 23.16 is dependent on the amount of delay in the lower arm of the
discriminator. For a 730-m-long fiber (corresponding to about 3.5 µs of delay),
the minimum linewidth will be about 225 kHz, far below typical linewidths of
semiconductor lasers.
To measure the linewidth of a single-frequency laser, such as a DFB laser, an
isolator is used to couple the laser to the discriminator. The isolator is used to
reduce the perturbations of the laser by the reflections from the fiber or the
optical interfaces within the interferometer. After passing through the isolator,
the signal is then applied to the discriminator. The two signals arriving at the
output of the interferometer (inside the discriminator) are uncorrelated and
are recombined by the directional coupler. Therefore, the output signal is the
sum of two signals with equal linewidth and center frequency. It is possible to
prove that this output signal is the autocorrelation function of the two signals
(in the two arms of the interferometer) and for gaussian line shapes, this
autocorrelation function has a linewidth equal to times that of the original signal.
By measuring the linewidth of the output signal (on the screen of the lightwave
signal analyzer) and dividing it by , the linewidth of the signal under test can be
determined.

23.7 Lightwave Polarization Analyzers


Polarization is a fundamental property of light that dramatically affects the
performance of many optical components. An optical isolator, for example, would
be impossible to design and manufacture without paying attention to the various
polarization conditions within the isolator assembly. The same is true for other
polarization-sensitive optical components such as polarization-maintaining fiber,
lasers, beam splitters, modulators, interferometers, retardation plates, and
polarizers and polarization adjusters. It is important to note, however, that
polarization effects may be undesired or desired. For example, simple movement
of fiber-optic cables often changes optical power. Or, for systems which require
interference, polarization is a key parameter to control.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.22 Chapter Twenty-Three

Figure 23.17 Poincaré sphere.

The electric field of a lightwave signal can be resolved into two components:
horizontal and vertical. In polarized light, the field components combine
according to their relative magnitude and phase and create characteristic
polarization states that can be grouped into linear, circular, and elliptical states
of polarization. The state of polarization can be graphically described as a point
on a sphere, called a Poincaré sphere as shown in Fig. 23.17. On a Poincaré
sphere, all linear states exist on the equator of the sphere; right-hand circular
is on top, left-hand circular is on bottom, and all other locations are elliptical. A
coordinate on a Poincaré sphere is a vector, called a Stokes vector, which has
four elements called the Stokes parameters. These parameters are a numerical
method of describing the state of polarization (SOP). The degree of polarization
(DOP) is the ratio of polarized light to the total power. Many polarization-
dependent properties are affected by the DOP.

23.7.1 Block diagram

A block diagram of a lightwave system capable of measuring the SOP and DOP
of a lightwave signal is shown in Fig. 23.18. The incoming light is divided into
four paths and each path is processed separately. They are each passed through
polarizing and retarding elements, and their respective intensities are measured.
These signals are processed by a computer to arrive at the average optical power
and the horizontal, 45°, and circular polarization components of the incoming
signal, as defined by the Stokes parameters. From these four components, the
SOP of the incoming signal is computed.

23.7.2 Polarization-dependent loss, PDL

Two of the most important measurements which can be done using a polarization
meter are polarization-dependent loss (PDL) and polarization-mode dispersion
(PMD). Many optical components such as couplers, isolators, fiber amplifiers,

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

Lightwave Signal Analysis 23.23

Figure 23.18 A block diagram for a polarization measurement system.

and connectors will have loss or gain, which is a function of the state of
polarization incident to them. Their PDL, sometimes called polarization
sensitivity, is a parameter which must be measured before such a component
can be utilized in an actual system. By varying the state of polarization of an
input signal, the polarization meter will then measure the output power from
the device under test as a function of the input power and input state of
polarization. The input state of polarization should be varied in such a way to
cover most of the Poincaré sphere to make sure that worst-case performance is
determined.

23.7.3 Polarization-mode dispersion, PMD


In very long, high-speed fiber-optic transmission systems, PMD can be the
limiting factor because light travelihg through a device in different polarization
mode experiences different propagation delays. In any component exhibiting
PMD, two polarization states exist that represent the fastest and slowest

Figure 23.19 Polarization-mode dispersion measurement block diagram.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.
Lightwave Signal Analysis

23.24 Chapter Twenty-Three

propagation. By determining those two states, the PMD value can be calculated.
One way of measuring PMD is to change the wavelength of the source incident
on the device under test and monitor the output state of polarization. Because
of the delay difference due to PMD, as the wavelength of an optical signal is
varied, a fixed input state of polarization is transformed into a continuous range
of output polarizations. These output states are displayed as an arc on the
Poincaré sphere. The delay difference is directly proportional to the measured
arc angle as shown in Fig. 23.19.

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)


Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.