Chapter 13 - Optical performance monitoring for coherent ...sb.uta.cl/ebooks/Optical Performance Monitoring/3-s2.0...Configuration of coherent receiver with balanced detector. 13.2

CHAPTER

Optical performance monitoringfor coherent optical systems
13
Yan Tang*, Xingwen Yi{, William Shieh**Department of Electrical and Electronic Engineering, The University of Melbourne, Victoria, Australia

{University of Electronic Science and Technology of China, Chengdu, China

13.1 HISTORICAL ASPECT OF COHERENT OPTICAL SYSTEMSMost commercial optical communication systems use intensity modulation/direct detection (IM/DD)

schemes in which the intensity of semiconductor lasers is modulated to carry the information and

the optical signal is detected directly by a photodiode. In contrast to the IM/DD scheme, a coherent

optical communication system detects the transmitted signal using homodyne or heterodyne detec-

tion schemes. It not only transmits information by modulating the intensity of the optical carrier,

but also the phase or the polarization. Coherent optical communication systems were extensively

studied in 1980s.1–3 Compared with direct detection, coherent detection offers the following

advantages:

1. Improved receiver sensitivity. With sufficient local oscillator (LO) power, the shot-noise-limited

receiver sensitivity can be achieved using coherent detection.

2. Improved spectral efficiency enables high-capacity transmissions, which is particularly attractive

for high-speed transmission systems.

However, with the invention of erbium-doped fiber amplifiers (EDFAs), direct detection systems

could achieve receiver sensitivity within a few decibels of coherent receivers, which made the

shot-noise-limited receiver sensitivity of the coherent receiver less attractive. In addition, the techni-

cal difficulties of coherent detection make it less practical. Coherent detection requires sophisticated

manipulation and processing of phase and polarization. Since the state of polarization (SOP) of the

incoming optical signal is scrambled in the fiber, a dynamic polarization controller is needed to

match the SOP of the signal and LO. The dynamic polarization controller is usually a bulky and

expensive device.4 The difficulty in stable locking of the carrier phase drift also prevents practical

application of the coherent detection. Consequently, further research of coherent optical communica-

tions had been almost abandoned for nearly 10 years.

Coherent detection has resurged to attract great interest in recent years, which is highlighted by

remarkable theoretical and experimental demonstrations from various groups around the world.5–9

The drive behind using coherent communication techniques nowadays is twofold. First, current

coherent detection systems are heavily entrenched in silicon-based digital signal processing (DSP).

By taking advantage of high-speed DSP, both polarization and phase management can be easily

© 2010 Elsevier Inc. All rights reserved.

Doi: 10.1016/B978-0-12-374950-5.00013-4351

352 CHAPTER 13 OPM for coherent optical systems

realized, and thus a free running laser can be used as a local oscillator. Optical coherent detection in

conjunction with high-speed DSP has the potential to increase spectral efficiency and the ability to

compensate linear transmission impairments such as chromatic dispersion (CD) and polarization

mode dispersion (PMD) in the electrical domain.10–12 Second, in contrast to the optical network sys-

tem that was dominated by a low-speed, point-to-point, and single-channel system a decade ago,

modern optical communication systems have advanced to massive wavelength-division multiplexing

(WDM) and reconfigurable optical networks with a transmission speed per channel approaching

100 Gb/s. The primary aim of coherent communications has shifted toward supporting these

high-speed dynamic networks by simplifying network installation, monitoring, and maintenance.

Besides, orthogonal frequency-division multiplexing (OFDM), which has emerged and thrived in

the radio frequency (RF) domain during the past decade, has gradually encroached into the optical

domain. Compared to the single-carrier coherent detection system, the combination of OFDM and

coherent detection not only realizes a robust dispersion transmission, but also brings the benefits

of computation efficiency and easiness of channel and phase estimation.

As optical fiber communication systems evolve toward more advanced all-optical fiber networks,

optical performance monitoring (OPM) has become increasingly important. Coherent communica-

tion brings OPM the following advantages:

1. With rapid advances in high-speed DSP and receiver electronic equalization, the optical channel

response can be accurately monitored with coherent detection.

2. High-frequency resolution enables the precise identification of optical power, optical signal-to-

noise ratio (OSNR), data rate, and modulation format of the signal in the optical domain, which

cannot be achieved in conventional IM/DD systems.

3. Coherent detection linearly converts the optical field into the RF field. Since RF technology is

more mature and accessible than optical technology, extending signal processing from the optical

domain into the RF domain greatly enhances the system’s functionality and flexibility.22

In this chapter we review recent developments on OPM techniques based on coherent detection,

especially in multicarrier optical coherent detection systems. We first introduce the basic theory of

optical coherent detection for both single-carrier and multicarrier systems in Section 13.2. Section

13.3 reviews various coherent detection–based OPM techniques that are categorized according to

whether OPM is accomplished with or without receiver electrical equalization. In Section 13.4 we

introduce the concept and theory of optical channel estimation (OCE), namely, OPM is achieved

via extraction of various optical parameters from the estimated OCE. Section 13.5 reports the recent

progress in OPM for coherent optical OFDM (CO-OFDM) systems in terms of simulation and

experimental demonstration.

13.2 SINGLE-CARRIER AND MULTICARRIER COHERENT OPTICAL SYSTEMSConventional single-carrier systems use one frequency to carry all the data. Multicarrier systems use

multicarrier modulation (MCM) schemes by which the transmitted data stream is divided into several

parallel lower–bit rate subcarriers. OFDM is a special form of MCM in the sense that the subcarriers

are partially overlapped in the frequency domain, but yet orthogonal to each other. In the context

of optical communications, CO-OFDM is a multicarrier optical system that combines OFDM

35313.2 Single-carrier and multicarrier coherent optical systems

techniques and optical coherent detection. We present the principle of coherent detection in Section

13.2.1, and then give a brief introduction to single-carrier systems in Section 13.2.2. Section 13.2.3

focuses on CO-OFDM systems in terms of architecture and signal processing. Finally, Section

13.2.4 compares single-carrier and multicarrier CO-OFDM systems.

13.2.1 Principle of coherent detectionThe basic idea behind coherent detection is to mix the received signal with another continuous light-

wave (CW) emitted from the LO before feeding into the photodetector. The optical field of the

received optical signal ES (only single polarization is considered) can be expressed as

EsðtÞ ¼ As� exp jostþ jfsð Þ; (13.1)

where As is the complex amplitude, and os and fs are, respectively, the angular frequency and the

phase of the input optical signal. The optical field of the LO can be expressed as

ELO ¼ ALO� exp joLOtþ jfLOð Þ; (13.2)

where ALO, oLO, and fLO are the complex amplitude, angular frequency, and phase of the LO,

respectively. For coherent detection, a balanced receiver is generally used to reject the common

mode component, such as suppressing the DC component and minimizing the laser relative intensity

noise (RIN). Figure 13.1 shows the block diagram of coherent detection with a balanced receiver.

The optical signal and the LO are mixed with a 3-dB coupler that adds a 180� phase shift to either

the signal or the LO field and splits into two equal parts that are detected by two photodetectors.

When the signal and LO share the same polarization, the output of the balanced receiver can be

given as

IðtÞ ¼ IþðtÞ � I�ðtÞ ¼ 2RffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPsðtÞPLO

pcos oIFtþ fsðtÞ � fLOðtÞð Þ; (13.3)

where oIF ¼ os � oLO; Ps and PLO are, respectively, the power of the signal and LO; fLO(t) is thephase of LO; and R is the responsivity of the photodetector.

Dependent on whether the LO frequency, oLO, is set close to the signal frequency, the coherent

receiver can be divided into homodyne receiver and heterodyne receiver. The OSNR sensitivity of

ES

ELO

PhotodetectorPolarization controller Laser

I+(t)

I–(t)

I(t)

E1

E2

Balanced receiver

3-dBcoupler

FIGURE 13.1

Configuration of coherent receiver with balanced detector.


the homodyne receiver is the same as the heterodyne counterpart. The homodyne receiver requires an

optical hybrid to recover in-phase (I) and quadrature (Q) components of the signal, and thus needs

twice as many optical components. But the heterodyne receiver needs an image rejection filter and

twice as much photodetector bandwidth. Both homodyne and heterodyne receivers need a polariza-

tion controller to align the SOPs of LO and signal, or preferably, a polarization diversity detector,8

which will be discussed below. Without a polarization controller, because of the random changes

resulted from the birefringence of the fiber, the polarization of the incoming signal is usually misa-

ligned to the SOP of the LO. Thus, the receiver sensitivity of coherent detection is highly dependent

on the SOP of the incoming signal in the absence of a polarization controller. A coherent receiver

with polarization-diverse architecture can be employed to solve this problem. The schematic of a

phase- and polarization-diverse receiver is shown in Figure 13.2. The incoming signal with arbitrary

SOP is divided into two orthogonal linear polarization components with a polarization beam splitter

(PBS) and fed into two homodyne I/Q receivers. A DSP module performs the functionalities such as

analog-to-digital conversion, polarization alignment, and so on. Thus, coherent detection allows us to

obtain full information on the optical signal—that is, signal amplitude, signal phase, and signal SOP.

13.2.2 Single-carrier coherent optical systemsFigure 13.3(a) shows the communication system architecture of a single-carrier coherent optical

system. At the transmitter end, the single-carrier system employs a relatively “conventional” and sim-

pler architecture, where discrete digital modulation is fed into the two arms of the QPSK (optical I/Q)

Es

ELO

DSP

90�optical hybrid

90�optical hybrid IQx

IIx

IIy

IQy

PBS

PBS

Esx

Esy

ELOx

ELOy

I/Q receiver I

I/Q receiver II

FIGURE 13.2

Configuration of phase and polarization-diversity receiver.

Laser

I

Q

QPSK modulator

p/2

Phase andpolarization

diversity receiver

ADC/DSP

(a)

Analog to digital conversion

Clock extract and retiming

Normalization and orthogonalization

Digital filtering

Frequency estimation and carrier recovery

Symbol estimation and FEC

Signals from phase and polarizationdiverse receiver

(b)

DS

P

FIGURE 13.3

(a) Basic single-carrier coherent optical system. (b) Block diagram of digital signal processing.

355

13.2

Single-carrie

randmultic

arrie

rcoherentopticalsyste

ms


modulator. At the receiver, the incoming signal is first detected with the phase- and polarization-

diversity coherent receiver to obtain the full information of the optical signal. Then the down-

converted electrical signal is processed by the DSP module with the process functions shown in

Figure 13.3(b).13 The analog-to-digital converter (ADC) and clock extraction sub-blocks are used

to digitize the received signal and synchronize the four channels with an integer number of sam-

ples per symbol. The normalization and orthogonalization block is used to compensate for imper-

fections in the 90� hybrid and the variation in responsivity of the four photodiodes.13 Digital

filtering is introduced to compensate for transmission impairments. Generally speaking, the func-

tionality of digital filtering can be categorized13 as compensation for (1) polarization-independent

impairments such as chromatic dispersion and (2) polarization-dependent effects such polarization

rotations and PMD dispersion. Additional nonlinear filtering can be implemented if nonlinear

impairments such as self-phase modulation or nonlinear phase noise are to be compensated. After

the digital filtering block, the phase and frequency mismatch between the incoming signal and the

local oscillator is then compensated. One possible phase estimation algorithm for the M-ary PSK-

modulated signal is to take the Mth power of the digital signal to remove the phase modulation.

Finally, the data can be obtained after the symbol estimation and forward error correction (FEC).

13.2.3 Coherent optical OFDM systemsOFDM is a special form of a broader class of MCM. An important property is that the subcarriers are

partly overlapped since each of their subcarrier frequencies is orthogonal with every other subcarrier

frequency. OFDM has been widely studied in mobile communications to combat hostile frequency-

selective fading and has been incorporated into various wireless network standards. CO-OFDM is an

optical equivalent of RF-OFDM that combines the technique of “optical coherent detection” and

“OFDM.”14 The coherent detection brings OFDM a much-needed linearity in RF-to-optical (RTO)

up-conversion and optical-to-RF (OTR) down-conversion. OFDM brings coherent system computa-

tion efficiency and ease of channel and phase estimation. Currently, many CO-OFDM experiments

using offline signal processing8,9,15,16 have been demonstrated. The complementary metal-oxide

semiconductor (CMOS), application-specific integrated circuit (ASIC) chips recently demonstrated

for single-carrier coherent systems17,18 signify that current silicon speeds can support 40-Gb/s OFDM

transmission systems. A real-time CO-OFDM system with a 3-Gb/s data rate19 has been demon-

strated recently. Because of its superior scalability with the bit rate of transmission systems,

CO-OFDM is well-positioned to be an attractive modulation format choice for the next generation

of ultra-high-speed optical networks.

13.2.3.1 Principle of OFDMEach subcarrier of an MCM signal can be represented as a complex wave,

sk ¼ CkðtÞej2pf tk ; (13.4)

where k ¼ 1, 2,. . ., Nsc, Ck(t) represents complex data at the kth subcarrier. For the sake of simplicity,

the pulse shape term is dropped. The MCM-transmitted signal s(t) is the combination of all subcar-

riers, which can be written as

sðtÞ ¼ 1

Nsc

XNsc

k¼1

Cke2pfk t: (13.5)


The OFDM signal employs an overlapped yet orthogonal signal set to reduce bandwidth consump-

tion. It can be seen that if the condition

fk � ft ¼ m1

Ts(13.6)

is satisfied, then the two subcarriers are orthogonal to each other over one OFDM symbol period Ts.In reality, a number of OFDM symbols are grouped together as an OFDM frame, which might also

include additional so-called pilot symbols for synchronization and channel estimation. For an OFDM

frame with multiple OFDM symbols, the output is the summation of multiple Equation (13.5), which

can be expressed as

sðtÞ ¼X1i¼�1

XNsc

k¼1

CkiSk t� iTsð Þ; (13.7)

j2pfk t
SkðtÞ ¼ PðtÞe ; (13.8)
and

PðtÞ ¼ 1 0 < t < Tsð Þ0 t � 0; t > Tsð Þ ;

�(13.9)

where cki is the ith information symbol at the kth subcarrier, Sk is the waveform for the kth subcarrier,

Nsc is the number of subcarriers, fk is the frequency of the subcarrier, and Ts is the symbol period.

The detected information symbol c0ki is given by

c0ki ¼ðTs0

r t� iTsð Þs�kdt ¼¼ðTs0

r t� iTsð Þexp �j2pfktð Þdt: (13.10)

The modulation/demodulation of the OFDM signal can be performed by inverse discrete Fourier

transform/discrete Fourier transform (IDFT/DFT) of the input/output information. The corresponding

architecture of the RF-OFDM transmitter/receiver using IDFT/DFT and digital-to-analog/analog-to-

digital converter (DAC/ADC) is shown in Figure 13.4.

13.2.3.2 Cyclic prefix for OFDMA cyclic prefix is created to prevent intersymbol interference (ISI) when an OFDM signal is trans-

mitted in a dispersive channel. As shown in Figure 13.5, the cyclic prefix (CP) is essentially an iden-

tical copy of the last portion of the OFDM symbol appended before the OFDM symbol. This CP

preserves the orthogonality of the subcarriers and prevents ISI between successive OFDM symbols.

The condition for ISI-free OFDM transmission is given by

tG < DG; (13.11)

where tG is the time delay of the OFDM symbol introduced by CD and PMD.

13.2.3.3 CO-OFDM system architectureA linear transformation in modulation, transmission, and demodulation is the most critical require-

ment for implementing OFDM in the optical domain. A generic CO-OFDM system uses direct

up-/down-conversion (DC) architecture, shown in Figure 13.6, which can be divided into five

Identical copy

ΔGguard

interval

Ts, OFDM symbol period

ts, observation period

FIGURE 13.5

Time-domain OFDM signal for one complete OFDM symbol with cyclic prefix.

Data Subcarriersymbolmapper

S/P IDFT GI

D/A LPF

D/A LPF

I

Q

RF OFDM transmitter

LPF

LPF

A/D

A/D

DFTData

symboldecision

P/SData

RF OFDM receiver

• DFT window synchronization• Frequency window offset compensation• Subcarrier recovery

FIGURE 13.4

Block diagram of RF OFDM transmitter/receiver.


functional blocks, including (1) the RF-OFDM transmitter, (2) the RTO up-converter, (3) the optical

channel, (4) the OTR down-converter, and (5) the RF-OFDM receiver.14

The architecture of RF-OFDM transmitter/receiver is shown in Figure 13.4, and the functional-

ities of the RF-OFDM transmitter include:

1. Converting the digital data from serial to parallel into a “block” of bits.

2. Mapping the information symbol onto the two-dimensional complex signal.

Data

–

–

Data

PD1

PD2

PD3

PD4

LD2

RF-to-optical up-converter

I

Q

I

Q

Optical-to-RF down-converter

RF-OFDMtransmitter

RF-OFDM receiver

MZM

MZM 90°

LD1

Optical link

90°

FIGURE 13.6

CO-OFDM system with direct up-/down-conversion.


3. Performing IDFT of the signal to obtain the time-domain OFDM signal.

4. Inserting pilot symbols and a guard interval for receiver processing, and performing DAC to

generate real-time signal SB(t), which can be shown in Equations (13.7)–(13.9).

The RTO up-converter uses an optical I/Q modulator that comprises two MZMs with a 90� phaseoffset to up-convert the real/imaginary parts of the real-time analog-baseband OFDM signal from the

RF domain directly to the optical domain. The directly up-converted optical OFDM signal can be

expressed as

EðtÞ ¼ eðjoLD1tþjfLD1Þ � SBðtÞ; (13.12)

where oLD1 and fLD1 are, respectively, the angular frequency and phase of the transmitter laser. The

up-converted signal E(t) transmits through the optical medium with impulse response h(t), and the

received optical signal becomes

EðtÞ ¼ eðjoLD1tþjfLD1Þ � SBðtÞ�hðtÞ; (13.13)

where � stands for convolution.

The optical OFDM signal is then fed into the OTR down-converter, and uses two pairs of bal-

anced receivers to perform I/Q detection optically. The output signal of the OTR down-converter is

rðtÞ ¼ eðjoo f f tþjDfÞr0ðtÞ; r0ðtÞ ¼ SBðtÞ�hðtÞ: (13.14)

In the RF-OFDM receiver, the down-converted near-DC-OFDM signal is first sampled with an

ADC. Then the signal needs to go through three sophisticated levels of synchronizations before


the symbol decision can be made. A detailed description of receiver signal processing can be found

in the next section. The three levels of synchronizations are:

1. DFT window synchronization

2. Frequency synchronization

3. Subcarrier recovery

Assuming successful completion of DFT window synchronization and frequency synchronization,

the RF-OFDM signal after DFT of the sampled value of Equation (13.14) becomes

rki ¼ efi hkicki þ nki; (13.15)

where rki is the received information symbol, fi is the OFDM symbol phase or common phase error

(CPE), hki is the frequency domain channel transfer function, and nki is the random noise. After the

phase and channel estimation, an estimated value of cki, cki, is given by the zero-forcing method as

cki ¼ h�kijhkij2

e�ifigki: (13.16)

cki is used for the symbol decision to recover the transmitter value cki, which is subsequently mapped

back to the original transmitted digital bits.

13.2.4 Comparison of single-carrier and multicarrier coherent optical systemsFor the single-carrier and multicarrier system, there are two conspicuous differences. First, the single-

carrier system usually employs a relatively “conventional” and simpler architecture. In contrast, for

the CO-OFDM architecture, the electronic DSP module and DAC are required for complex OFDM

signal generation at the transmit end. The OFDM transmitter strictly enforces linearity in each compo-

nent associated with the CO-OFDM transmitter. Second, in the single-carrier systems, the information

is coded in the time domain, whereas in CO-OFDM, the information is encoded in the frequency

domain. Given these two basic differences, CO-OFDM provides the following advantages:

1. Ease of signal processing. In CO-OFDM-based multicarrier systems, with the use of pilot sym-

bols and pilot subcarriers, channel estimation and phase estimation can be relatively simple.

However, in single-carrier coherent systems, channel estimation has to rely on blind equalization,

such as the constant modulus algorithm (CMA) or decision feedback, all of which are prone to

error propagation. The phase estimation usually adopts the Viterbi algorithm, which is mostly

effective for the pure phase modulation and less effective for other constellation modulation.

Furthermore, differential phase coding needs to be employed to resolve the intrinsic phase ambi-

guity for the Mth-power law algorithm, resulting in approximately a factor of 2 for the bit error

rate (BER) increase.6

2. Higher-order modulation. With increasing spectral efficiency in transmission, optical transmis-

sion systems should be designed to support the higher-order modulation format. For CO-OFDM

systems, the scale-to-higher-order modulation format can be simply achieved via the software to

reconfigure DSP and DAC. In contrast, the higher-order, single-carrier optical system requires

more complicated optical modulator configurations, which inevitably increase system complexity

and cost.

36113.3 OPM using coherent detection

3. Tight bounding of spectral components. The OFDM spectral shape is tightly bounded and is more

tolerant to the filter narrowing effect. Even if the edge subcarriers are attenuated by the narrow-

ing filtering, certain bit- and power-loading schemes in the similar spirit to water-filling algo-

rithm20 can be employed to mitigate the effect. In contrast, the filtering narrowing effect not

only causes pulse distortion, but also makes it susceptible to timing jitter in single-carrier

systems.

13.3 OPM USING COHERENT DETECTIONVarious monitoring techniques using coherent detection for both single-carrier and multicarrier sys-

tems have been proposed in the literature to monitor one or multiple parameters. Generally, OPMs

with coherent detection can be grouped into two categories: OPM without receiver electrical equali-

zation, and OPM with receiver electrical equalization. OPM without receiver electrical equalization

usually relies on external devices such as an optical spectrum analyzer (OSA),21 RF devices,22

frequency-selective polarimeter,23 and so on. Recently, many OPMs have been proposed based

on the receiver electrical equalization technique,24 which takes advantage of powerful and cost-

effective silicon signal-processing capabilities. They do not require expensive external devices to

evaluate optical properties or to tap the optical signal, which eventually reduces the effective

received optical power. In addition, DSP-based OPM techniques are adaptable to varying data rates

and modulation formats, and are capable of realizing jointly monitoring OSNR, BER, Q-factor, CD,

and PMD.

13.3.1 OPM without receiver electrical equalization13.3.1.1 OSNR monitoringWith the development of laser technology, tunable lasers with wide wavelength coverage, continuous-

wavelength tunability, and narrow spectral linewidth are commercially available. Coherent receivers

incorporated with such a tunable local laser can be used as a high-spectral-resolution OSA.

OSAs with coherent detection can be used for precise identification of optical power, OSNR,21 data

rate, and modulation format of the signal in the optical domain, which cannot be achieved by

conventional grating-based OSAs.

Baney et al.25 demonstrated a coherent optical spectrum analysis method based on a swept-tuned

optical LO and a coherent receiver that provides fine resolution and high dynamic range. Figure 13.7

is a simplified block diagram of the proposed coherent optical spectrum analyzer (COSA). The

incoming signal is combined with a tunable LO via an optical coupler, and a balanced receiver is

used to perform OE conversion. The LO frequency is swept across the measurement wavelength

range to display the optical spectrum. Figure 13.8 shows the DFB-LD spectral measurement by a

COSA and a grating-based OSA that had been set to the narrowest resolution of �80 pm. The LO

emitted 2-mW power and scanned at a nominal rate of 62 GHz/s.25 It was reported that COSA

was able to clearly resolve the DFB lineshape, including the relaxation sidebands and the central

laser peak. Furthermore, the dynamic range achieved by COSA was as high as 70 dB.

Tian et al.21 proposed an in-band polarization-assisted OSNR and spectrum monitoring technique

based on the swept coherent detection. As shown in Figure 13.9, the swept coherent detector consists

Wavelength (nm)

1560.90 1561.00 1561.15

30 GHz–80

–60

–40

–20

0

COSA OSA

Opt

ical

pow

er (

dBm

)

FIGURE 13.8

Comparison of measurement of DFB-LD linewidth by COSA and by diffraction grating-based OSA.25 Copyright ©

2002 IEEE.

LO

Δout

TZ

TZCoupler

n(t) = no + g t

ELO(t)

Es(t) EA(t)

EB(t)

In

PD

PD

FIGURE 13.7

Simplified block diagram of coherent optical spectrum analyzer. LO, local oscillator; PD, photodetector; TZ,

transimpedance amplifier.25 Copyright © 2002 IEEE.

Incoming signal

RF powermeter 50:50

pS

pLO

PC2

PC3

LO WavelengthtunableSwept coherent detector

FIGURE 13.9

Schematic for in-band OSNR and spectrum monitoring based on swept coherent detection. FBC, fiber Bragg

grating; PC, polarization controller; VOA, variable optical attenuator.21 Copyright © 2006 IEEE.



of a tunable LO laser, two polarization controllers to adjust the polarization states of the signal and

LO, a balanced detector, and a RF power meter. By adjusting the polarization of the signal and LO

laser, the proposed method was capable of measuring the signal and the in-band ASE noise spectra

separately; the OSNR then can be obtained by integrating the measured spectra.

The average RF power at the output of the balanced detector is

hPEðtÞi / jALOjXk

jAS;kjP!

LO � P!S

!2

þ A2LO

Xk

AiP!LO � P!ASE;i

!2* +

/ jALOj2Ps P!LO � P!S

� �2þ 1

2jALOPASE;

(13.17)

where ALO is the LO signal amplitude, and PASE and PS are, respectively, the ASE noise power and

signal power. P!S and P

!LO are the signal and LO polarization states. Ai and P

!ASE are the amplitude

and polarization states of the ASE at frequency fi. h�i denotes the time average. If P!S⊥P

!LO, then

P!

SP!LO ¼ 0, the minimal RF power is detected, and the output is a direct indicator of the in-band

ASE power at a specific frequency. If P!

SjjP!

LO, then P!SP!LO ¼ 1, the RF power meter gives the max-

imum power, which contains the signal power and the half of the in-band ASE power. The monitored

optical spectrum can be obtained by scanning the wavelength of the LO over the optical signal, as

shown Figure 13.10. Then the OSNR can be obtained by integrating over the measured spectra for

signal and noise power. However, the accuracy of the polarization-assisted OSNR monitoring

scheme could be affected by PMD and polarization scattering induced by interchannel cross phase

modulation (XPM) in WDM systems26,27; therefore, additional measurement must be used to reduce

errors caused by polarization scattering.

13.3.1.2 CD/PMD monitoringHui et al.22 proposed a combined CD and PMD monitoring technique based on RF signal processing

methods that depends on the signal processing on the RF signal after heterodyne detection using exter-

nal devices such as RF amplifiers, RF filters, RF mixers, and so on. The principle and experiment setup

fLOtunable

Max power, pS // pLO

Min power, pS ^ pLO In-bandASE

Signalspectrum

fi

Optical frequency

f2B

FIGURE 13.10

Operating principle of in-band, high-resolution swept coherent detection scheme.21 Copyright © 2006 IEEE.

Tunablelaser

Polarizationcontroller/scrambler

PD

3-dBcoupler

OSA ESA

EDFASMF

PMDemulator

10-Gbit/sRZ

BPF

BPF

25 GHz

BPF

10 GHz

BPF

10 GHz

BPF

5 GHz

15 GHz

RF amp

Oscilloscope

GPIB

Power meter

ADC

CD and PMDcalculation

Mixer

FIGURE 13.11

Block diagram of CD and PMD monitoring using coherent detection. ADC, analog-to-digital converter; BPF,

bandpass filter; ESA, electrical spectrum analyzer; OSA, optical spectrum analyzer.22 Copyright © 2005 IEEE.


are shown in Figure 13.11. Since RF technology is more mature than the optical alternative, extending

signal processing from the optical domain into the RF domain greatly enhances the system’s function-

ality and flexibility. CD monitoring is based on the fact that the optical spectrum with RZ or NRZ dig-

ital modulation typically has two redundant clock-frequency components. Due to the CD, these two

clock components propagate at different speeds. Since the coherent heterodyne detection linearly

shifts the optical spectrum into the RF domain, the relative phase-delay information of the optical sig-

nal is preserved. The fiber CD can be evaluated from the relative time delay between two recovered RF

clocks. In the RF domain, the carrier and the two clock frequencies can be selected separately by three

bandpass filters (BPFs). As shown in Figure 13.11, because of the heterodyne detection, the frequency

of two RF clocks are both shifted with a frequency equal to the IF carrier frequency. Thus, the carrier

component is further split into two components, each used to mix with the upper and lower sidebands

independently to generate two baseband clocks. The CD can be evaluated from the relative time delay

Dt between these two recovered clocks by

Dt ¼ Dl2Rb=c; (13.18)

where D is the fiber accumulated CD, Rb is the data rate, l is the signal wavelength, and C is the

speed of light.

The basic idea of PMD monitoring is to measure the differential polarization walk-off between

any two different frequency components within the optical spectrum. When only considering the

first-order DGD (Dt), a relatively angular walk-off between two frequencies with Df distance can

be used to represent such polarization walk-off:

D’ ¼ p � Df � Dt: (13.19)

Since the RF signal is a linear representation of the optical signal, to perform such measurement, the

power of two RF frequencies, f1 and f2, is measured. In Reference 22, the down-converted optical


carrier ( f1) and lower sideband ( f2) are measured. The RF power of two selected frequencies can be

expressed as22

P1 ¼ �1PloPsigðf1Þ cos2ð’ÞP2 ¼ �2PloPsigðf2Þ cos2ð’þ D’Þ ; (13.20)

where �1/2 is the combined effect of two detectors’ responsivities and the relative amplitude of RF

frequency f1/2. Plo/sig is the optical power of LO/signal. ’ represents the polarization mismatch

between one of the selected frequencies and the LO. D’ is the SOP angle between frequencies

f1 and f2. Normalized powers of P1 and P2 are measured to eliminate the uncertainty of the power

spectral densities of two frequencies, detector responsivity and RF amplifier gain. The first-order

DGD can be obtained through the normalized power cos2’(t) and cos2(’(t) þ D’). However, thesemethods are not bit rate or modulation-format transparent because several RF BPFs with particular

center frequencies are needed in the implementation of these approaches.

Roudas et al.23 proposed a PMD monitoring technique based on a frequency-selective polarimeter

using coherent heterodyne detection. Due to the inherent high-frequency resolution and power sen-

sitivity of coherent detection, the frequency-selective polarimeter with coherent detection offers

superior accuracy compared to its direct detection counterparts. The proposed method is capable

of measuring variation in the Stokes parameters as a function of frequency. Figure 13.12 shows

the block diagram of such a coherent detection–based, frequency-selective polarimeter, which con-

sists of a 3-dB coupler, balanced receiver, polarization transformer, and electronic preamplifier.

A BPF with a center frequency of fc is used to cut a very thin slice of the modulated signal.

A square-law detector and a low-pass filter are used to measure the power of this spectral slice,

which can be used to estimate the SOP of the received signal. When the LO and the received signal

are both planar monochromic waves, the photocurrent at the output of the low-pass filter is

iLPF ¼ R2PsPloð1þ eseloÞ; (13.21)

Square-lawdetector

LO

EIo

ErEr1

Er2

3-dBcoupler

i2

i1

itot

PA

BPF (.)2 LPF

PH

PHPT

FIGURE 13.12

Schematic of coherent frequency–selective polarimeter. BPF, bandpass filter; LPF, low-pass filter; PA, power

amplifier; PH, photodetector; PT, polarization transformer.23 Copyright © 2004 IEEE.


where R is the responsivity of the photodiode, and ps and plo are the average power of the received

signal and LO, respectively. es; elo are the normalized Stokes vectors corresponding to the SOP of the

received signal and LO, respectively. The Stokes components [Sx, Sy, Sz] of the signal around the fre-

quency fc can be estimated with the following equation:

S 1ð Þx S

1ð Þy S 1ð Þ

z

S 2ð Þx S

2ð Þy S 2ð Þ

z

S 3ð Þx S

3ð Þy S 3ð Þ

z

2664

3775

Sx

Sy

Sz

264

375 ¼ 1

R2PsPlo

iLPFj 1ð ÞiLPFj 2ð ÞiLPFj 3ð Þ

264

375� 1: (13.22)

R2PsPlo can be estimated from two measurements of the photocurrent corresponding to two antiparallel

(in Stokes space) LO-SOPs. iLPF|(k), k¼ 1, 2, 3 represent three different measurements of the photocurrent

corresponding to three noncoplanar LO-SOPswith knownStokes components ½Skx; Sky; Skz ; k ¼ 1; 2; 3. Forexample, LO-SOPs can be set to 0�, linear 45�, and right- or left-circular polarization. Estimation of the

Stokes parameter variance as a function of frequency can be performed by tuning the LO frequency at

closed-space intervals and repeating Equations (13.21) and (13.22) for each frequency.

13.3.2 OPM with receiver electrical equalizationFor single-carrier coherent optical systems, Hauske28 demonstrated the CD, DGD, and OSNR moni-

toring techniques by analyzing finite-impulse response (FIR) filter coefficients. This technique uses a

polarization-diversified coherent receiver, as shown in Figure 13.13, to linearly map the optical sig-

nal into electrical domains. A bank of FIR filters is applied to the digital signal after analog-to-digital

conversion. The filtering is induced by the blind adaptive algorithm to minimize ISI, and thus BER is

minimized. The equalizer filter consists of four complex-valued FIR filters arranged in a butterfly

structure. The filter’s transfer function is described with a single Jones matrix,

H�1ð f Þ ¼ H�1XXð f Þ H�1

YX ð f ÞH�1

XY ð f Þ H�1YY ð f Þ

" #; (13.23)

where the matrix elements are the transfer functions of complex-valued filters. The filter’s transfer

function can be assumed as the inverse of the fiber link once the tap algorithm for blind adaptation

is converged. Then channel parameters such as CD and DGD can be obtained from filter coeffi-

cients. Experiment results based on the 111-Gb/s PolMUX-RZ-DQPSK data showed that the OSNR

Y-pol Y�-pol

X-pol X�-pol

Y-pol

X-pol

Tran

smitt

er

Fib

er c

hann

el F

Coh

eren

t rec

eptio

nA

DC

Clo

ck r

ecov

ery HXX

HXY

HYY

Equ

aliz

er H

Car

rier

reco

very

Dec

ider

+

+

HYX

FIGURE 13.13

Coherent transmission system with butterfly-structured equalizer filter.28 Copyright © 2008 IEEE.


could be estimated with a precision of 1 dB. At zero DGD, the highest standard deviation of CD

monitoring is 63 ps/nm.

Real-time in-service CD and PMD monitoring based on the coefficient extraction of an FIR filter

were demonstrated.29 Figure 13.14 shows the experimental setup. The system under test consists of

an 80-channel, dual-polarization QPSK system and 800-km fiber with distributed high PMD. Nortel

eDC40G circuit packs are used as the dual-polarization QPSK transceiver. The system uses the OC-

192 signal as the payload. Compared with independent measurements of the PMD, the DGD monitor

exhibited tolerance of 12 ps (95% confidence level) over a range of 10–123 ps. Experimental results

showed that CD monitoring is independent of instantaneous DGD. The average CD monitor reading

was within 32 ps/nm of the independently measured CD.

For multicarrier coherent optical systems, with the emergence of CO-OFDM systems, the combi-

nation of coherent detection and multicarrier transmission brings about an advanced OPM concept,

the so-called OCE.30,31 As proposed by Shieh et al.,30 OCE is equivalent to channel estimation in

wireless OFDM communication systems. Because all optical parameters, including OSNR, CD,

Polarimeter

Tunabletransceiver

#2

OSA

WS

S

Noiseloading

Span 8100-km TW-RS

x3

IV or V or VI I or II or IIIDGE

100-km TWRS100-km TWRS100-km TWRS

30-km SSMF

x3

BoosterEDFA

Polarizationcontroller 1

Tunabletransceiver

#1

Tunablelaser

PBC

Delay

QPSK

QPSK

MUX

MUX

11.5Gb/s231–1 PRBS

D2

D1 D3

D4

80 loading channels13

79

80

24

WS

S

……

FIGURE 13.14

Block diagram of real-time monitoring transmission experiment. Tunable laser and polarimeter were used to

measure PMD of channel independently of monitor.29 Copyright © 2008 IEEE.

TX procDAC

TX procDAC

X2

X1

s2(t)

s1(t)

I/Q mod

I/Q mod

n times

2�4 90−hybrid

2�4 90�hybrid

r2(t)

r1(t)

Y1

Y2ADC,RX proc

ADC,RX proc

FIGURE 13.15

Schematic of investigated OFDM system. OFDM parameters: data rate 100 Gb/s, 256 subcarriers, 16-QAM,

12.5% CP. Twelve spans: 80 km-SSMF per span, D¼ 17 ps/nm/km, mean PMD¼ 10 ps.32 Copyright© 2009 IEEE.


and PMD, are embedded in the optical channel response, most of them can be extracted and accu-

rately monitored at the same time after OCE. Therefore, in CO-OFDM systems, the principle of

OCE is to estimate the optical channel response by using training symbols. Once channel response

is known, OPM is the extraction of various optical parameters from the estimated optical channel

response. More importantly, performance monitoring by OCE is basically free because it is embed-

ded as a part of intrinsic receiver signal processing. Such a monitoring device could be also placed

anywhere in the network without concern about the large residual CD of the monitored signal.

A similar concept of channel estimation based on optical OFDM with coherent detection is pro-

posed by Mayrock et al.32 A simulated optical OFDM system with polarization-diverse coherent

receiver and simulation parameters are shown, respectively, in Figure 13.15. Two orthogonally

polarized, optical OFDM signals are generated independently. At the receiver, with the DSP, the

optical channel matrix is obtained through the pilot symbols. Then the accumulated CD and the dif-

ferential group delay (DGD) on the subcarrier basis can be extracted from the channel matrix. To

minimize the ASE noise impact, Savitzky-Golay filters are used to improve monitoring accuracy.

Figure 13.16(a) depicts estimated DGDs that were obtained at an OSNR of 20 dB with polynomial

degree p ¼ 1. W is the number of data samples processed by the filter. Figure 13.16(b) summarizes

the estimated inverse SNR versus optical input power. For low optical powers, N/S is dominated by

ASE noise, and thus an estimate for the OSNR can be deduced directly.31 At higher signal power

levels, self-phase modulation is the dominant additive noise. Thus, nonlinear signal degradation

can be identified with some additional effort by intentionally inserted signal power variations.33

A detailed discussion of the signal processing of OPM with OCE in CO-OFDM systems is the

focus of Section 13.4.

13.4 OPM IN CO-OFDM SYSTEMSAs shown in Section 13.2.3, the principle of OFDM is to use a large number of low-speed orthogonal

subcarriers to transmit a high-speed data stream. Therefore, each subcarrier only occupies a narrow

frequency band, and the channel response for each subcarrier is approximately flat, even though the

(a) (b)

0 50 100 150 200 250

Carrier Index

120

140

160

180

200

220

240

260Δt

(d)

(ps)

ReferenceW = 5W = 11

W = 17

Branch 1

Branch 2

–10

–12

–14

–16

–18

–20

–22–16 –14 –12 –10 –8 –6 –4 –2 0 2

Popt (dBm)

10 lo

g|N

/S|

FIGURE 13.16

(a) Estimated DGDs at OSNR ¼ 20 dB using Savitzky-Golay filtering. (b) Estimated inverse SNR versus optical

input power.32 Copyright © 2009 IEEE.

36913.4 OPM in CO-OFDM systems

global channel response is not. The global channel response is simply the combination of channel

responses of all the subcarriers. Furthermore, by using coherent detection and polarization-diversity

detection, the optical field can be linearly down-converted to the electrical domain. Consequently,

the channel information can be obtained through receiver signal processing. It should be noted that

OPM by OCE in a CO-OFDM system can achieve a very fast response, such as less than a micro-

second if tens of OFDM symbols are used as the preamble for the OCE. This monitoring speed could

be sufficient to accommodate the CD and OSNR change from the environment disturbance. OCE can

also be conducted in single-carrier systems but with increased computation complexity.11 This section

mainly discusses OPM in CO-OFDM systems.

13.4.1 Optical channel modelIt is well-known that single-mode fiber supports two polarization modes. Thus, instead of being

represented as a single element, the CO-OFDM signal model requires the mathematical description

of the polarization effects as well as the fiber CD. Therefore, the optical channel model for a CO-

OFDM signal can be treated as a two-input-two-output (TITO), multiple-input multiple-output

(MIMO) OFDM model, which is intrinsically represented by a two-element Jones vector familiar

to the optical communication community. Figure 13.17 shows a complete TITO-MIMO CO-OFDM

system that consists of two CO-OFDM transmitters (one for each polarization), an optical link, and

two CO-OFDM receivers. The dashed line on the devices indicates variations of the MIMO architec-

ture with the option to remove the device from the configuration. The other MIMO architecture

includes single-input-single-output (SISO), single-input-two-output (SITO), and two-input-single-

output (TISO). As discussed in Section 13.2.3, the CO-OFDM transmitter comprises an RF-OFDM

EDFA

Fiber

Optical OFDM transmitter I

Optical OFDM transmitter II

PBCFiber

EDFA

PBS

Optical OFDM receiver I

Optical OFDM receiver II

Optical link with PMD/PDL

Span 1 Span m

FIGURE 13.17

Conceptual diagram of TITO coherent optical MIMO-OFDM model.


transmitter and RTO up-converter, whereas the CO-OFDM receiver comprises an OTR down-

converter and RF-OFDM receiver. Each fiber span includes the effects of CD and PMD/PDL. The

optical noise is added from the optical amplifiers (OAs) at the end of each span. The fiber nonlinear

effect is not considered in this channel model, but we will investigate the influence of fiber nonlinea-

rities on OPM through OCE.

Similar to the single-polarization OFDM signal model represented by Equations (13.7)–(13.9),

the transmitted OFDM time-domain signal s(t) of the MIMO-OFDM model is described using the

Jones vector given by

sðtÞ ¼X1i¼�1

XNsc

k¼1

cikPðt� iTsÞej2pfkðt�iTsÞ; (13.24)

T x y T
sðtÞ ¼ ½ sx sy ; cik ¼ ½ cik cik ; (13.25)
and

PðtÞ ¼ 1 ð0 < t < TsÞ0 ðt � 0; t > TsÞ :

�(13.26)

We use i and k as the indices for the OFDM symbol and OFDM subcarrier, respectively. sx and sy arethe two polarization components for s(t), and cik is the transmitted OFDM information symbol in the

form of Jones vector for the kth subcarrier in the ith OFDM symbol. The Jones vector cik is employed

to describe the generic OFDM information symbol regardless of any polarization configuration

for the OFDM transmitter. The received information symbol after the proper DFT window and

frequency offset synchronization is given by

c0ik ¼ ejfiejFDðfkÞTkcik þ nik; (13.27)

N

Tk ¼Yl¼1

exp ð�1

2jbl fk � 1

2alÞs; (13.28)

and

FDð fkÞ ¼ f0 þ 2pt0 fk þ pcDt f2k =f

2LD1; (13.29)

37113.4 OPM in CO-OFDM systems

where c0ik ¼ ½ c0xik c

0yikT is the received information symbol in the form of the Jones vector for the

kth subcarrier in the ith OFDM symbol, nik ¼ ½ cxik cyik T is the noise including two polarization

components, Tk is the Jones matrix for the fiber link, N is the number of PMD/PDL cascading

elements represented by their birefringence vector bl and PDL vector al,34 s is the Pauli matrix

vector,34 fk is the frequency of kth subcarrier, FD( fk) is the phase owing to the fiber-accumulated

CD Di, and fi is the OFDM symbol phase noise owing to the phase noises from the lasers and

RF-LO at both the transmitter and receiver.2 fi is usually dominated by laser phase noise.

In the channel model of Equation (13.27), CD and DGD are independent of the OFDM symbol

index because they are treated as slowly varying compared within the time duration of OFDM sym-

bols. The laser phase-noise term becomes CPE after FFT, common to one OFDM symbol. Although

the transmission model of Equation (13.27) includes the frequency responses of transmitter and

receiver components, their effect can be considered as stationary and calibrated out in the initial

stage, and will not be discussed.

13.4.2 Principle of OPM through optical channel estimationIn order to perform the channel estimation, the phase noise fi for each OFDM symbol has to be

obtained through pilot subcarriers introduced in Section 13.2.3. Removing the phase noise fi from

Equation (13.27), we obtain

cp0ik ¼ Hð fkÞcik þ npik; (13.30)

jFDðfkÞ
Hð fkÞ ¼ e Tk; (13.31)
and

Hð fkÞ ¼Hxxð fkÞ Hxyð fkÞHyxð fkÞ Hyyð fkÞ� �

; (13.32)

where C p0ik ¼ C 0

ike�jfi , and npik ¼ nike

�jfi are, respectively, the received symbol and noise after

phase noise compensation. H( fk) is a 2 � 2 channel response matrix that includes CD and DGD/

PDL. The received signal after channel compensation and phase noise compensation is

cRik ¼ H�1 fkð Þc p0ik: (13.33)

The goal of OCE is to estimate the four elements of H( fk) through signal processing. Once H( fk) isknown, OPM is the extraction of various optical parameters from the estimated optical channel

response. One possible method to estimate H( fk) for the TITO-MIMO-OFDM system is by using

training symbols in the preamble using alternate polarization launch35—that is, successive transmis-

sion ofc10

� �and

0

c2

� �. Thus, the four elements of H( fk) can be expressed as

Hxx ¼< c01=c1 >Hyx ¼< c02=c1 >Hxy ¼< c01=c2 >Hyy ¼< c02=c2 >;

(13.34)

where c01 and c02 are the received training symbols, and h�i denotes the average overall training

symbols.


13.4.2.1 CD monitoringFrom Equation (13.27), it is apparent that phase change of the channel response is mainly induced by

the CD. Once the channel transfer function H( fk) is known, the subcarrier phase is given by

Fð fkÞ ¼ argðHð fkÞÞ; (13.35)

where arg(�) stands for the phase for a complex signal. According to Equation (13.29), the accumu-

lated chromatic dispersion Dt can be estimated by a simple second curve fitting of F( fk) as a func-

tion of the subcarrier frequency. Note that CD monitoring is only based on the phase curves of the

channel response, and therefore immune to the DGD-induced amplitude change.

13.4.2.2 DGD monitoringIt can be shown that the amplitude response of Uxx( fk) and Uvx( fk) can be expressed as28

jUxxð fkÞj2 ¼ aþ b cosð2pfktÞ aþ b � 1

jUyxð fkÞj2 ¼ 1� a� b cosð2pfktÞ a� b � 1;

((13.36)

where a and b are constants. |Uxx( fk)|2 and |Uyx( fk)|

2 essentially represent the multipath interference

due to DGD t. By analyzing Equation (13.36), such interference/fading is periodic, which follows a

cosine function with a period determined by t. Thus, the amplitude fading of the OFDM subcarrier is

caused by DGD, which can be estimated by the inverse of the fading period.

13.4.2.3 System Q-factor monitoringAnother important parameter to monitor is the system Q-factor. A live system could run error-free

even without FEC for an extended period, making it hard to detect the system margin by measuring

BER directly. From Equation (13.27), we can see that each subcarrier channel is essentially a linear

channel with additive white Gaussian noise. Subsequently, the BER of a QPSK system is given by36

BER ¼ 0:5 � er f c ESNR=ffiffiffi2

p� �(13.37)

and

ESNR ¼ hhc0kii2i =d2kik; (13.38)

where ESNR is the (electrical) SNR ratio per bit, hik stands for the averaging over the subcarriers or

the index k, hciikii is the expectation value of the received symbol for subcarrier k, and

dk ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffihjc0ikj2ii � hciiki2i

qis the standard deviation of the received symbol for subcarrier k. Equation

(13.38) shows that ESNR can be obtained by first constructing the constellation of the received sym-

bol and then performing the computation of ESNR for each constellation point. We further convert

the BER in Equation (13.37) into the Q-value,37 which is commonly used in the optical community.

From Equation (13.37), the system Q is thus given by

Q ¼ 10 log10ð2ESNRÞ: (13.39)

From Equations (13.38) and (13.39), the system Q can be effectively monitored by computing the

subcarrier symbol spread in the constellation diagram. Since the CD-induced delay spread and the

ISI can be completely removed in a CO-OFDM system, the electrical noise characterized by d is

predominately from the accumulation of ASE noise from optical amplifiers, and it can be shown that

37313.5 Progress in OPM for CO-OFDM systems

1

ESNR¼ A

1

OSNRþ B; (13.40)

where A is a proportional constant between ESNR and OSNR, and B is attributed to the background

noise not accounted for by ASE noise, which is mainly from the phase noise of the transmit/receive

lasers. From Equations (13.37)–(13.40), we can see that by acquiring hc0iki and dk through receiver

signal processing, the ESNR for the OFDM signal can be computed, and subsequently both the sys-

tem Q and the OSNR can be monitored. The coefficients A and B in Equation (13.40) can be

obtained empirically with a calibration procedure by measuring ESNRs against a series of known

OSNRs and performing a linear fit between 1/ESNR and 1/OSNR.It is quite instructive to explicitly write out the ideal coherent detection performance for a QPSK-

modulated CO-OFDM system where the linewidths of the transmit/receive lasers are assumed to be

zero. From Equation (13.37), the corresponding BER, Q, and ESNR in this ideal condition can be

given by

BER ¼ 0:5 � er f c OSNRB0

R

� ; (13.41)

�
Q ¼ 10 log10 2OSNR
B0

R; (13.42)

and

ESNR ¼ 2OSNRB0

R; (13.43)

where B0 is the optical ASE noise bandwidth used for OSNR measurement (�12.5 GHz for 0.1-nm

bandwidth), R � Nsc Df is the total system symbol transmission rate, and Nsc and Df are the number

of subcarriers and channel spacing of the subcarriers, respectively.

13.5 PROGRESS IN OPM FOR CO-OFDM SYSTEMSIn this section we report the recent progress in OPM for CO-OFDM systems in terms of simulation

and experimental demonstration. We show that in polarization, multiplexed CO-OFDM systems,

critical optical system parameters including fiber CD, PMD, Q-value, and OSNR can be accurately

monitored without resorting to separate monitoring devices.

13.5.1 Simulation model and resultsAMonte Carlo simulation is carried out to demonstrate the CD, Q, and OSNR monitoring for a SISO-

MIMO-CO-OFDM system. The OFDM parameters are a symbol period of 25.6 ns, a guard time of

3.2 ns, and 256 subcarriers. BPSK encoding is used for each subcarrier, resulting in a total bit rate

of 10 Gb/s. The linewidth of the transmitter and receiver lasers are assumed to be 100 kHz each, which

is close to the value achieved with commercially available semiconductor lasers.38,39 The linked ASE

noise from the optical amplifiers is assumed to be additive white Gaussian noise, and the phase noise

of the lasers is modeled as white frequency noise characterized by its linewidth. The CD is assumed

0

20000

40000

60000

0 20000 40000 60000

Link CD (ps/nm)

Mon

itore

d C

D (

ps/n

m)

–100

–50

0

50

100

Mon

itore

d C

D e

rror

(ps/

nm)

Monitored CD (ps/nm)Monitored CD error (ps/nm)

FIGURE 13.18

Performance of CD monitoring through channel estimation.


to be constant within the OFDM spectrum. Eight-block OFDM symbols, each consisting of 100

OFDM symbols, are used for extracting various parameters, including CD, system Q, and OSNR.

In the following text, we use “calculate” to mean the BER results obtained by Monte Carlo simulation,

and “monitor” to mean the interpolation results obtained by Equations (13.41)–(13.43).

Figure 13.18 shows the monitored CD from the receiver signal processing. The input OSNR is set

at 3.8 dB, which gives a BER of 10�3 for a CD below 34,000 ps/nm. We can see that CD up to

50,000 ps/nm can be monitored with an accuracy of 50 ps/nm. The simultaneous large dynamic

range and good accuracy of CD monitoring are the unique features of the OFDM modulation format,

namely, a large number of subcarriers spread across a wide spectrum of 10 GHz, resulting in good

accuracy, and narrow subcarrier channel spacing of 44.6 MHz, resulting in wide dynamic range. This

wide dynamic range is an improvement of more than one order of magnitude over a prior report

using single or a few auxiliary subcarriers.40

Figure 13.19 shows the monitored system Q and OSNR though OCE. The Q is calculated from

7 to 12 dB by Monte Carlo simulation—that is, direct BER simulation with signal duration of

5

10

15

20

25

0 5 10 15 20 25 30

OSNR (dB)

Sys

tem

Q (

dB)

–14

–10

–6

–2

2

Mon

itore

d O

SN

Rer

ror

(dB

)Calculated Q

Monitored Q

OSNR errorQ Margin

FIGURE 13.19

Monitored system Q and OSNR as function of input OSNR.


20.5 ms, represented by the solid squares in Figure 13.19. This demonstrates good agreement with the

monitored Q by Equation (13.39). Beyond that, we rely on Equation (13.39) for system Q estimation.

To appreciate the advantage of this approach, for instance, at an input OSNR of 20 dB, the system Q

for this OSNR is monitored to be 21.3 dB, which gives a Q-margin of 11.5 dB over a BER of 10�3.

Such a method of Q-margin prediction at high OSNRs is similar to that in direct detected systems.35

Thus, the margin monitoring is achieved nonintrusively. Note that this level of system margin cannot

be measured directly. Additionally, the OSNR is monitored by computing ESNR and estimating

OSNR using Equation (13.40). The curve with solid triangles in Figure 13.19 shows that the OSNR

can be monitored with errors within 0.5 dB for an input OSNR dynamic range of 1–20 dB. The max-

imum OSNR that can be monitored is limited by laser phase noise.

13.5.2 Optical performance monitoring in CO-OFDM systems with 4-QAMIn this section we focus on the experimental demonstration of OSNR, Q-factor, fiber CD, and DGD

monitoring through OCE in CO-OFDM systems.31

13.5.2.1 Experiment setupThe experimental setup in this work comprises a generic SITO-MIMO CO-OFDM transmission sys-

tem, as shown in Figure 13.20. The transmitter and receiver laser sources in this work both have a

specified linewidth of less than 100 kHz. The optical carrier wavelength is about 1555 nm. The data

rate is 10.7 Gb/s. The OFDM parameters used in the experiment are listed in Table 13.1. In the

experiment, 11 pilot subcarriers are used for phase estimation. The spacing of OFDM subcarriers

TDS PBS

Balancedreceiver I

Opticalhybrid

LD1

DMZ

I Q

AWG

OBF

DGD

Optical attenuator

Opticalhybrid

LD2

Balancedreceiver II

Laser

EDFATransmitter

Polarization diversity receiver

FIGURE 13.20

Experimental setup for optical performance monitoring with SITO-MIMO CO-OFDM system. (There is a

polarization controller before DGD emulator.) AWG, arbitrary waveform generator; DMZ, dual MZ modulator;

EDFA, erbium-doped fiber amplifier; TDS, time-domain sampling scope.

Table 13.1 OFDM Parameters for 4-QAM Transmission

FFTLength

SubcarrierNumber

SymbolPeriod

GuardInterval

Pilot SubcarrierNumber

Effective DataRate

128 88 12.8 ns 1.6 ns 11 10.7 Gb/s


is 1/(12.8 � 10�9) ¼ 78.125 MHz. The RF-OFDM sequence is generated by an arbitrary waveform

generator at 10 GSa/s to emulate the DAC, and the electrical-to-optical (EO) direct up-conversion

is realized by a dual MZ modulator configured as an optical I/Q modulator. A home-built PMD emu-

lator follows the transmitter to generate different DGD. A recirculating loop, including 100.8-km

standard SMF fiber and an EDFA to compensate the loss, is used to emulate the long-haul transmis-

sion. As the PMD emulator is outside the recirculating loop, only the first-order PMD is emulated.

The optical signal coupled out from the recirculating loop transmits through an optical attenuator

and another EDFA to evaluate the system performance through ASE noise loading. The output signal

after fiber transmission is detected using a polarization-diversity coherent receiver with intermediate-

frequency (IF) down-conversion detection. After the optical-to-electrical (OE) down-conversion,

the RF-OFDM signal is sampled by a real-time scope (TDS) at 20 GSa/s to emulate the ADC, and the

resultant digital sequences are uploaded to a computer for DSP.

The OCE is conducted by sending preambles in the OFDM frame. In this experiment, 40 training

symbols (equivalent to an OPM response time in the order of microseconds) are used in the preamble

to keep a balance between OCE accuracy and OPM response speed.

13.5.2.2 OSNR and Q-factor monitoringIn this section we monitor the OSNR and Q-factor by loading different levels of ASE noise in a

back-to-back transmission. The OSNR measured by an OSA is used as the reference. Both OSNR

and Q-factor monitoring are derived from the electrical SNR of the received RF-OFDM signal

according to Equations (13.42) and (13.43). SNR (per bit) is calculated from the noise spreading

of received signal constellations based on Equation (13.38).

The best SNR in our system is about 17 dB, which is used to determine and subtract the back-

ground noise (mainly from the RF components). The OSNR monitoring result is shown in Fig-

ure 13.21(a). Due to background noise, it is difficult to monitor the high OSNR in the system.

However, the monitored OSNR error is within 0.5 dB over a range from 6 to 18 dB, which covers

the main OSNR dynamic range of interest.

Figure 13.21(b) shows themonitoredQ-factor result. TheQ-factor calculated from the BER, the cal-

culated Q, is included for comparison. It can be seen that the monitored Q agrees with the calculated Q

within 0.6 dB, which implies that our system is dominated byGaussian noise.When the BER is low, it is

difficult to obtain the meaningful calculated Q, which signifies the importance of Q-factor monitoring.

The calculated Q is limited to 12.7 dB due to the maximum number of OFDM symbols processed.

13.5.2.3 CD and DGD monitoringThe CO-OFDM channel response is mainly determined by CD and PMD in the optical link. In this

section we first estimate the channel response and then extract the parameters for CD and DGD in

the long-haul transmission.

5

10

15

20

5 10 15 20

OSNR by OSA (dB)

Q-f

acto

r (d

B)

–2

–1

0

1

2

Q-f

acto

r er

ror

(dB

)

Monitored QCalculated QQ-factor error

(b)

4

8

12

16

20

5 10 15 20OSNR by OSA (dB)

Mon

itore

d O

SN

R (

dB)

–2

–1

0

1

2

Mon

itorin

g er

ror

(dB

)

(a)

FIGURE 13.21

(a) OSNR monitoring result. (b) Q-factor monitoring result. Both are measured in back-to-back transmission.


Figure 13.22 shows an example of channel response of the 1008-km and 900-ps DGD transmis-

sion. Since we use a SITO-MIMO architecture, we can estimate the channel responses for both polar-

ization components corresponding to Hxx( fk) and Hyx( fk) in Equation (13.32). Figure 13.22 clearly

shows that due to the CD the phase response is parabolic and the magnitude has an apparent

DGD-induced, frequency-selective fading. As discussed in Section 13.4.2, CD is estimated by sec-

ond curve fitting of the phase response, and the DGD is estimated by the inverse of the period of

the fading as shown in Equation (13.36); for example, the period in Figure 13.22(a) is 1.1 GHz,

which corresponds to a DGD of 909 ps.

Figure 13.23(a) shows the CD monitoring result versus transmission distance. Without DGD, the

monitored CD increases linearly with the transmission distance. The CD parameter is around 16.4 ps/

nm/km, which corresponds to the dispersion of the SMF used in the experiment. The CD monitoring

result with fixed 900-ps DGD is also shown in Figure 13.23(a). Compared with the non-DGD moni-

toring result, the DGD causes about 5% of the monitoring variation. The main reason for such mon-

itoring variation is the phase ripples, as illustrated in Figure 13.22. The launch power into the fiber is

about 0 dBm. It has been shown that the fiber nonlinearity at this power level is very strong.41

0 10 20 30 40 50 60 70 80 90–20

–15

–10

–5

Frequency (�78.125 MHz)

Mag

nitu

de

(dB

, rel

ativ

e)

Pha

se (r

ad)

–25

–15

–10

–5

5

0

–20

0 10 20 30 40 50 60 70 80 90–20

–15

–10

–5

Frequency (�78.125 MHz)

Mag

nitu

de

(dB

, rel

ativ

e)

Pha

se (r

ad)

–25

–15

–10

–5

0

5

–20

(a)

(b)

FIGURE 13.22

Estimated channel responses for (a) x and (b) y polarization components X-axes are the frequencies

normalized to OFDM subcarrier spacing.


Therefore, Figure 13.23(a) also in essence indicates that the CD monitoring is robust against fiber

nonlinearity.

Figure 13.23(b) shows the CD monitoring error versus varying DGD at two different OSNR con-

ditions after 1008-km transmission with about �7-dBm launch power. The high OSNR (13–14 dB)

condition has a BER smaller than 10�5, whereas the low OSNR (6.5–7.5 dB) has a BER greater than

10�3. The CD monitoring errors of both OSNR conditions are within 5%, which shows that the CD

monitoring can be immune to DGD and ASE noise.

13.5.3 OPM in CO-OFDM systems with 16-QAM modulationThe OPM in CO-OFDM systems with 16-QAM is conducted in this section. The experiment setup is

similar to the one used for 4-QAM, excepting that the PMD effect is excluded. The data rate is

10 Gb/s and the OFDM parameters are shown in Table 13.2.

0

4000

8000

12000

16000

20000

0 200 400 600 800 1000Transmission distance (km)

Chr

omat

ic d

ispe

rsio

n

–10%

–5%

0%

5%

10%

DG

D-in

duce

d C

Dva

riatio

n (d

B)

Without DGDWith DGDDGD-induced CD variation

(a)

–4%

–2%

0%

2%

4%

0 400 800 1200DGD (ps)

Rel

ativ

e er

ror

High OSNRLow OSNR

(b)

FIGURE 13.23

(a) CD monitoring versus transmission distance with and without DGD. (b) CD monitoring error versus DGD

after 1008-km transmission.

Table 13.2 OFDM Parameters for 16-QAM Transmission

FFT LengthNumber ofSubcarrier

Symbol Period Guard IntervalNumber of PilotSubcarrier

Data Rate

128 44 13.6 ns 0.8 ns 8 10 Gb/s

37913.6 OPM experiment results

13.6 OPM EXPERIMENT RESULTSFigure 13.24 shows the monitoring results of OSNR and Q-factor in the back-to-back transmission.

When the OSNR is below 18 dB, the OSNR monitoring error is below 0.5 dB, as shown in

Figure 13.24(a). The bigger monitoring error beyond 18 dB is due to the finite SNR of the RF com-

ponents in the transmitter and receiver.

Figure 13.25 shows the CD monitoring results. The monitored CD linearly increases with the trans-

mission distance and the CD coefficient of the transmission fiber is calculated as 16.22 ps/nm/km.

8

10

12

14

16

18

20

22

8 10 12 14 16 18 20OSNR by OSA (dB)

Mon

itore

d O

SN

R (

dB)

–2

–1

0

1

2

Mon

itorin

g er

ror

(dB

)

OSNR-monOSNR-mon offset

6

8

10

12

14

16

8 10 12 14 16 18 20

OSNR by OSA (dB)

Q-f

acto

r (d

B)

–2

–1

0

1

2

Q-f

acto

r er

ror

(dB

)

Monitored QCalculated QQ-factor error

4

(b)

(a)

FIGURE 13.24

Monitoring results in CO-OFDM system with 16-QAM for (a) OSNR and (b) Q-factor.

0

4000

8000

12000

16000

20000

0 200 400 600 800 1000 1200Transmission distance (km)

Chr

omat

ic d

ispe

rsio

n(p

s/nm

)

FIGURE 13.25

CD monitoring result in CO-OFDM system with 16-QAM.


381References

13.7 SUMMARYWe have reviewed the principle and progress of the OPM techniques in coherent optical systems

with a focus on MCM. We have presented the theoretical background as well as experimental dem-

onstration of OPM in CO-OFDM systems. In particular, we have taken advantage of OCE using

advanced DSP to realize a fast and joint estimation of channel parameters as OSNR, Q-factor,

CD, and PMD.

REFERENCES1. Giles RC, Reichman KC. Optical self-homodyne DPSK transmission at 1-Gbit/s and 2-Gbit/s over 86 km of

fiber. IEE Electron Lett 1987;23(22):1180–1.2. Kazovsky LG, Benedetto S, Willner AE. Optical fiber communication systems. Norwood, CT: Artech

House; 1996.

3. Okoshi T, Kikuchi K. Coherent optical fiber communications. Berlin: Springer; 1988.4. Noe R, Sandel D, Yoshida-Dierolf M, Hinz S, Mirvoda V, Schopflin A, et al. Polarization mode dispersion

compensation at 10, 20, and 40 Gb/s with various optical equalizers. IEEE/OSA J Lightwave Technol1999;17(9):1602–16.

5. Savory SJ, Gavioli G, Killey RI, Bayvel P. Electronic compensation of chromatic dispersion using a digital

coherent receiver. OSA Opt Express 2007;15(5):2120–6.6. Ly-Gagnon DS, Tsukarnoto S, Katoh K, Kikuchi K. Coherent detection of optical quadrature please-shift

keying signals with carrier phase estimation. IEEE/OSA J Lightwave Technol. 2006;24(1):12–21.7. Charlet G, Renaudier J, Salsi M, Mardoyan H, Tran P, Bigo S. Efficient mitigation of fiber impairments in

an ultra-long haul transmission of 40 Gbit/s polarizationmultiplexed data, by digital processing in a coherent

receiver. In: Technical digest of optical fiber communication conference and exposition and the nationalfiber optic engineers conference (OFC/NFOEC), paper PDP17. Anaheim, CA; 2007.

8. Shieh W, Yi X, Ma Y, Tang Y. Theoretical and experimental study on PMD-supported transmission using

polarization diversity in coherent optical OFDM systems. OSA Opt Express 2007;15(16):9936–47.9. Jansen SL, Morita I, Tanaka H. 16 � 52.5-Gb/s, 50-GHz spaced, POLMUX-COOFDM transmission over

4,160 km of SSMF enabled by MIMO processing KDDI R&D laboratories. In: Proc. of European confer-ence on optical communications (ECOC), paper PD1.3. Berlin; 2007.

10. Taylor MG. Coherent detection method using DSP for demodulation of signal and subsequent equalization

of propagation impairments. IEEE Photon Technol Lett 2004;16(2):674–6.11. Ip E, Kahn JM. Digital equalization of chromatic dispersion and polarization mode dispersion. IEEE/OSA

J Lightwave Technol 2007;25(8):2033–43.12. Kikuchi K. Phase-diversity homodyne receiver for coherent optical communications. In: Technical digest of

optical amplifiers and their applications/coherent optical technologies and applications, paper CThB3.

Whistler, BC, Canada; 2006.

13. Seb J Savory. Digital filters for coherent optical receivers. OSA Opt Express 2008;16(2):804–17.14. Shieh W, Athaudage C. Coherent optical orthogonal frequency division multiplexing. IEE Electron Lett

2006;42(10):587–9.15. Shieh W, Yi X, Tang Y. Transmission experiment of multi-gigabit coherent optical OFDM systems over

1000 km SSMF fiber. IEE Electron Lett 2007;43(3):183–5.16. Jansen SL, Morita I, Takeda N, Tanaka H. 20-Gb/s OFDM transmission over 4,160-km SSMF enabled by RF-

pilot tone phase noise compensation. In: Technical digest of optical fiber communication conference and expo-sition and the national fiber optic engineers conference (OFC/NFOEC), paper PDP15. Anaheim, CA; 2007.


17. Roberts K. Electronic dispersion compensation beyond 10 Gb/s. In: Technical digest of IEEE LEOS summertopical meeting, paper MA2.3. Portland, OR; 2007.

18. Sitch J. Implementation aspects of high-speed DSP for transmitter and receiver signal processing. In: Tech-nical digest of IEEE LEOS summer topical meeting, paper MA4.3. Portland, OR; 2007.

19. Yang Q, Chen S, Ma Y, Shieh W. Real-time reception of multi-gigabit coherent optical OFDM signals. OSAOpt Express 2009;17(10):7985–92.

20. Proakis J, Salehi M. Digital communications. 5th ed. Singapore: McGraw-Hill International Edition; 2008.

21. Tian X, Su Y, Hu W, Leng L, Hu P, He H, et al. Precise in-band OSNR and spectrum monitoring using

high-resolution swept coherent detection. IEEE Photon Technol Lett 2006;18(1):145–7.22. Fu B, Hui R. Fiber chromatic dispersion and polarization-mode dispersion monitoring using coherent detec-

tion. IEEE Photon Technol Lett 2005;17(7):1561–3.23. Roudas I, Piech G, Mlejnek M, Mauro Y, Chowdhury D, Vasilyev M. Coherent frequency-selective polar-

imeter for polarization-mode dispersion monitoring. IEEE/OSA J Lightwave Technol 2004;22(4):953–67.24. Buchali F. Electronic dispersion compensation for enhanced optical transmission. In: Technical digest of

optical fiber communication conference and exposition and the national fiber optic engineers conference(OFC/NFOEC), paper OWR5, Anaheim, CA; 2006.

25. Baney DM, Szafraniec B, Motamedi A. Coherent optical spectrum analyzer. IEEE Photon Technol Lett2002;14(3):355–7.

26. Lee JH, lung DK, Kim CH, Chung YC. OSNR monitoring technique using polarization-nulling method.

IEEE Photon Technol Lett 2001;13(1):88–90.27. Xie C, Mtiller L, Kilper DC, Mollenauer LF. Impact of cross-phase modulation induced polarization scat-

tering on optical PMD compensation in WDM systems. OSA Opt Lett 2003;28(23):2303–5.28. Hauske FN, Geyer J, Kuschnerov M, Piyawanno K, Duthel T, Fludger CRS, et al. Optical performance

monitoring from FIR filter coefficients in coherent receivers. In: Technical digest of optical fiber communi-cation conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paperOThW2. San Diego, CA; 2008.

29. Woodward SL, Nelson LE, Feuer MD, Zhou X, Magill PD, Foo S, et al. Characterization of real-time PMD

and chromatic dispersion monitoring in a high-PMD 46-Gb/s transmission system. IEEE Photon TechnolLett 2008;20(24):2048–50.

30. Shieh W, Tucker R, Chen W, Yi X, Pendock G. Optical performance monitoring in coherent optical OFDM

systems. OSA Opt Express 2007;15(2):350–6.31. Yi X, Shieh W, Ma Y, Tang Y, Pendock GJ. Experimental demonstration of optical performance monitor-

ing in coherent optical OFDM systems. In: Technical digest of optical fiber communication conference andexposition and the national fiber optic engineers conference (OFC/NFOEC), paper OThW3. San Diego,

CA; 2008.

32. Mayrock M, Haunstein H. Performance monitoring in optical OFDM systems. In: Proc. conference on opti-cal fiber communication (OFC), paper OWM3. San Diego, CA 2009.

33. Mayrock M, Haunstein H. Optical monitoring for non-linearity identification in CO-OFDM transmission

systems. In: Proc. conference on optical fiber communication (OFC), paper JThA58. San Diego, CA 2008.

34. Gisin N, Huttner B. Combined effects of polarization mode dispersion and polarization dependent losses in

optical fibers. Opt Commun 1997;142(1–3):119–25.35. Shieh W, Yi X, Ma Y, Tang Y. Theoretical and experimental study on PMD-supported transmission using

polarization diversity in coherent optical OFDM systems. Opt. Express 2007;15:9936–47.36. Proakis J. Digital communications. 3rd ed. New York: WCB/McGraw-Hill; 1995 [chapter 5].

37. Bergano NS, Kerfoot FW, Davidsion CR. Margin measurements in optical amplifier system. IEEE PhotonTechnol Lett 1993;5(3):304–6.

38. Berger JD, Anthon D. Tunable MEMS devices for optical networks. Opt Photon News 2003;14:43–9.

383References

39. Ip E, Kahn J, Anthon D, Hutchins J. Linewidth measurements of MEMS-based tunable lasers for phase-

locking applications. IEEE Photon Technol Lett 2005;17(10):2029–31.40. Liu A, Pendock GJ, Tucker RS. Improved chromatic dispersion monitoring using single RF monitoring

tone. OSA Opt Express 2006;14(11):4611–6.41. Ma Y, Shieh W, Yi X. Characterization of nonlinearity performance for coherent optical OFDM signals

under influence of PMD. IEE Electron Lett 2007;43(17):943–5.

Documents

Chapter 13 - Optical performance monitoring for coherent ...sb.uta.cl/ebooks/Optical Performance Monitoring/3-s2.0...Configuration of coherent receiver with balanced detector. 13.2