Wavelength-Division Multiplexing and Optical Amplification

Introduction and Context

The capacity of a single optical fibre can exceed several terabits per second, yet a conventional intensity-modulated system using one optical carrier occupies only a small fraction of the available fibre spectrum. Wavelength-division multiplexing (WDM) addresses this limitation by transmitting multiple independent optical channels at different wavelengths through the same fibre. Each wavelength functions as a separate logical communication channel.

As WDM systems evolved from a few widely separated wavelengths to hundreds of closely spaced channels, optical amplifiers became essential. In-line optical-electrical-optical regeneration is expensive and limits transparency. Optical amplifiers provide direct optical gain, allowing many channels to be amplified simultaneously without converting them to the electrical domain. However, amplifier noise, gain tilt, saturation, nonlinear effects, and channel crosstalk impose important design constraints.

This chapter develops the principles needed to design coarse WDM (CWDM) and dense WDM (DWDM) systems, establish wavelength plans, evaluate crosstalk and power budgets, and compare erbium-doped fibre amplifiers (EDFAs), Raman amplifiers, and semiconductor optical amplifiers (SOAs).

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain WDM architecture, wavelength grids, and multiplexing/demultiplexing methods.

2. Distinguish CWDM, DWDM, and related frequency-grid standards.

3. Design a preliminary wavelength plan subject to passband, spacing, loss, and dispersion constraints.

4. Calculate optical power budgets, amplifier gain, amplified spontaneous emission (ASE), and noise figure.

5. Explain channel crosstalk and gain equalization requirements.

6. Compare EDFAs, Raman amplifiers, and SOAs for practical network applications.

7. Evaluate a multi-span link and identify suitable engineering and operational practices.

1. WDM Operating Principle

In WDM, optical carriers with centre wavelengths \(\lambda_1,\lambda_2,\ldots,\lambda_N\) are independently modulated and combined by an optical multiplexer. The combined signal propagates through a fibre and is separated at the receiver by a demultiplexer. The wavelength separation is commonly expressed in frequency because frequency grids are more directly related to optical filtering:

\[

f=\frac{c}{\lambda}

\]

where \(f\) is optical frequency, \(c\) is the speed of light in vacuum, and \(\lambda\) is wavelength. For a small wavelength separation,

\[

\Delta f \approx \frac{c}{\lambda^2}\Delta\lambda.

\]

This approximation assumes that \(\Delta\lambda\ll\lambda\). At 1550 nm, 1 nm corresponds approximately to 125 GHz. Thus, a 20-nm channel separation corresponds to about 2.5 THz.

A WDM terminal normally contains transmitters, wavelength-selective multiplexers, optical amplifiers where required, fibre spans, demultiplexers, receivers, monitoring components, and network-control functions. Modern coherent systems may use polarization multiplexing and high-order modulation on each wavelength, but the wavelength-division principle remains the same.

Important terminology

A channel is an assigned optical carrier and its modulated information. Channel spacing is the frequency or wavelength interval between adjacent nominal carriers. The free spectral range is the periodic spacing of a periodic filter response. Insertion loss is the power loss introduced by a passive component. Isolation is the suppression of an undesired wavelength at a selected port. Crosstalk is unwanted coupling of one channel into another. Gain tilt is the variation of amplifier gain across the WDM band.

2. CWDM and DWDM

2.1 Coarse WDM

CWDM uses relatively wide channel spacing, commonly 20 nm, and is associated with a wavelength range extending approximately from 1270 to 1610 nm, depending on the implementation and transceiver specification. The wide spacing permits relatively inexpensive thin-film filters, arrayed components, and uncooled laser transmitters. CWDM is attractive for metropolitan networks, access aggregation, enterprise interconnection, and moderate-capacity links.

CWDM has lower spectral efficiency and generally fewer channels than DWDM. The broad spacing also reduces sensitivity to laser wavelength drift and filter-shape tolerances. However, optical amplification across the entire CWDM wavelength range is more difficult than amplification in the conventional C-band, and many CWDM systems therefore use shorter spans or amplified subsets of the band.

2.2 Dense WDM

DWDM uses closely spaced channels, often based on standardized frequency grids such as 100 GHz, 50 GHz, or 25 GHz. The C-band near 1530–1565 nm is particularly important because silica fibre loss is low and erbium-doped fibre amplifiers operate effectively there. The L-band, approximately 1565–1625 nm, can provide additional capacity but requires suitable amplifiers, filters, and transceivers.

DWDM supports high spectral efficiency and long-haul operation. It requires frequency-stable lasers, narrowband filters, accurate wavelength management, dispersion and nonlinear-effect control, and gain equalization. A DWDM system may carry tens or hundreds of channels, each at 10, 25, 100, 400, or 800 Gb/s depending on modulation, symbol rate, coding, and optical signal-to-noise ratio (OSNR).

2.3 Frequency grids

A frequency-grid plan specifies nominal channel frequencies relative to a reference. If adjacent channels have spacing \(\Delta f\), a simple grid is

\[

f_k=f_0+k\Delta f,

\]

where \(f_0\) is a reference frequency and \(k\) is an integer channel index. The exact reference and allowed grids should follow the applicable ITU-T recommendation and equipment specification.

Frequency spacing must accommodate the occupied signal bandwidth, filter roll-off, laser frequency error, temperature drift, nonlinear spectral broadening, and guard margin. A 50-GHz grid is not automatically suitable for every 50-GHz-channel signal: the transmitter spectrum and filter cascade must also fit within the slot.

3. WDM Components and Architectures

3.1 Multiplexers and demultiplexers

Common technologies include thin-film interference filters, diffraction gratings, arrayed-waveguide gratings (AWGs), planar lightwave circuits, and reconfigurable optical add-drop multiplexers (ROADMs). AWGs provide compact, repeatable multi-channel filtering but are sensitive to temperature and wavelength alignment. Thin-film filters can provide good isolation and modular add/drop operation. ROADMs enable remotely controlled wavelength routing and improve network flexibility.

A practical multiplexer must be assessed using insertion loss, channel passband, adjacent-channel isolation, non-adjacent isolation, polarization-dependent loss, polarization-mode dispersion, return loss, and temperature stability.

3.2 Add-drop architectures

An optical add-drop multiplexer removes selected wavelengths from a through fibre and inserts new channels. Fixed add-drop multiplexers are simple but inflexible. Reconfigurable devices use wavelength-selective switches and are controlled by a network management system. In a mesh network, a wavelength may pass through several ROADMs; therefore, filter narrowing and cumulative passband distortion must be included in the design.

3.3 Transmitters and receivers

CWDM transmitters may use directly modulated, uncooled distributed-feedback lasers. DWDM transmitters generally require tighter wavelength control and may use cooled DFB lasers, tunable lasers, external modulators, or coherent transmitters. Receivers can be direct-detection or coherent. Coherent receivers provide access to amplitude, phase, and polarization information and support advanced digital signal processing, but they require greater system complexity.

4. Wavelength-Plan Design

A wavelength plan should be developed before selecting components. The process includes:

1. Define the fibre type, operating band, span lengths, data rates, modulation formats, and required reach.

2. Select a standardized frequency grid and reserve channels for protection, expansion, and monitoring.

3. Check transmitter tuning range and wavelength accuracy.

4. Verify that channel bandwidth fits the optical slot after filter cascading.

5. Check fibre attenuation, chromatic dispersion, polarization effects, and nonlinear penalties.

6. Establish launch-power limits and amplifier operating points.

7. Specify channel isolation and acceptable crosstalk.

8. Define power equalization and monitoring procedures.

The usable optical band is not simply the nominal filter band. It is the intersection of the transmitter range, multiplexer and demultiplexer passbands, fibre and amplifier operating windows, and receiver sensitivity requirements.

Worked Example 1: Comparing wavelength and frequency spacing

Consider adjacent CWDM wavelengths at 1550 nm and 1570 nm. Using \(c=3.0\times10^8\) m/s and the small-separation approximation,

\[

\Delta f\approx\frac{3.0\times10^8}{(1.55\times10^{-6})^2}(20\times10^{-9}).

\]

Since \((1.55\times10^{-6})^2=2.4025\times10^{-12}\) m²,

\[

\Delta f\approx 2.50\times10^{12}\text{ Hz}=2.50\text{ THz}.

\]

The result illustrates why a 20-nm CWDM spacing is much wider than a 100-GHz DWDM spacing. The approximation is sufficiently accurate for this engineering comparison; exact frequency calculation should be used when defining a standards-compliant grid.

5. Crosstalk and System Impairments

Crosstalk occurs when energy from one channel reaches another channel’s receiver. It may be caused by imperfect multiplexer isolation, filter leakage, reflections, connector contamination, four-wave mixing, Raman interaction, or transmitter spectral broadening.

If a desired channel has power \(P_s\) and an interfering contribution has power \(P_i\), a simple crosstalk ratio is

\[

X=10\log_{10}\left(\frac{P_i}{P_s}\right)\text{ dB}.

\]

Because \(P_i<P_s\), \(X\) is negative. Some specifications instead report isolation as \(I=10\log_{10}(P_s/P_i)\), a positive number. These conventions must not be confused.

For several independent interferers, powers—not dB values—must be added:

\[

P_{i,\mathrm{total}}=\sum_j P_{i,j}.

\]

The resulting crosstalk can reduce signal-to-noise ratio, increase bit errors, and, in coherent systems, produce phase and constellation distortions. Adjacent-channel crosstalk is usually more critical than non-adjacent crosstalk.

Other major impairments include chromatic dispersion, polarization-mode dispersion, fibre nonlinearities, ASE noise, filter narrowing, connector loss, and stimulated Brillouin or Raman scattering. WDM design must treat these effects jointly. Increasing channel launch power improves received signal power only up to the point at which nonlinear penalties and amplifier saturation become dominant.

6. Optical Power Budget and OSNR

For a passive link, the received power in dBm can be estimated as

\[

P_R=P_T-L_F-L_C-L_M,

\]

where \(P_T\) is transmitter power, \(L_F\) is fibre attenuation, \(L_C\) is total connector and splice loss, and \(L_M\) is an engineering margin. If fibre attenuation is \(\alpha\) dB/km over length \(L\), then fibre loss is \(\alpha L\).

For an amplified span, the amplifier gain approximately compensates the preceding span loss, but the amplifier also adds ASE noise. The relevant quality metric for multi-channel systems is often OSNR rather than merely received power. OSNR is the ratio of signal power to ASE noise power measured in a specified reference bandwidth.

For an optical amplifier with noise figure \(F

\), input signal-to-noise ratio \(\mathrm{SNR}_{in}\), and gain \(G\), the idealized relation is

\[

\mathrm{SNR}_{out}=\frac{\mathrm{SNR}_{in}}{F},

\]

provided the noise figure is expressed as a linear ratio and the amplifier operates in its specified regime. In decibels,

\[

NF=10\log_{10}F.

\]

For a high-gain phase-insensitive amplifier, a common approximation for ASE spectral density per polarization is

\[

S_{ASE}\approx n_{sp}h\nu(G-1),

\]

where \(n_{sp}\) is the spontaneous-emission factor, \(h\) is Planck’s constant, \(\nu\) is optical frequency, and \(G\) is linear gain. Exact OSNR calculations must account for polarization conventions, optical reference bandwidth, gain, input noise, and amplifier architecture.

Worked Example 2: Passive power budget

A transmitter launches +2 dBm into 40 km of fibre with attenuation 0.25 dB/km. There are two connectors with 0.5 dB loss each, and the engineering margin is 3 dB. Fibre loss is

\[

40(0.25)=10\text{ dB}.

\]

Total loss is \(10+1+3=14\) dB. Therefore,

\[

P_R=2-14=-12\text{ dBm}.

\]

If the receiver sensitivity is −18 dBm, the margin remaining beyond the specified engineering margin is 6 dB. This calculation assumes linear additive loss in dB and does not include dispersion, nonlinear penalties, or crosstalk.

7. Optical Amplifiers

An optical amplifier increases optical power without first converting the signal to an electrical form. Important specifications include gain, saturated output power, noise figure, gain bandwidth, gain flatness, input dynamic range, polarization dependence, transient response, and reliability.

7.1 Erbium-doped fibre amplifier

An EDFA uses erbium ions doped into a short fibre. Pump light, commonly near 980 or 1480 nm, excites the erbium ions. Stimulated emission near 1550 nm amplifies signals in the C-band; modified designs operate in the L-band.

EDFAs offer high gain, high output power, low noise figure, good coupling to standard fibre, and simultaneous amplification of many WDM channels. They are therefore dominant in long-haul C-band systems. Limitations include gain variation with wavelength and loading, transient gain changes when channels are added or removed, ASE accumulation, pump ageing, and difficulty covering very broad bands with uniform gain.

7.2 Raman amplifiers

A Raman amplifier uses stimulated Raman scattering in the transmission fibre or a dedicated fibre. Pump photons at a shorter wavelength transfer energy to signal photons at a longer wavelength. Distributed Raman amplification reduces the effective noise contribution by providing gain along the span rather than only at its end.

Raman systems can provide low effective noise figure, flexible gain wavelengths, and improved span reach. They require high pump power, careful pump-safety control, wavelength-dependent design, and management of pump-to-signal and signal-to-signal interactions. Counter-propagating pumping is common because it reduces pump intensity at the transmitter and can improve noise performance; co-propagating arrangements may have different transient and implementation characteristics.

7.3 Semiconductor optical amplifiers

An SOA uses stimulated emission in a semiconductor waveguide. It is compact, electrically pumped, and potentially integrable with other photonic devices. SOAs are useful in optical switching, access networks, burst-mode applications, and integrated transmit/receive modules.

Compared with EDFAs, SOAs commonly exhibit higher noise, stronger gain saturation, polarization sensitivity, pattern effects, and nonlinear distortion. Their fast carrier dynamics can be useful for switching but problematic for transparent multi-channel amplification.

7.4 Comparison

EDFAs are preferred for high-power, low-noise, multi-channel C-band line amplification. Raman amplifiers are attractive when very long reach, low effective noise figure, or extended spectral coverage is required. SOAs are preferred where compactness, integration, fast switching, or moderate gain is more important than the lowest noise and highest output power. No amplifier is universally superior; the decision depends on span length, channel count, band, launch power, cost, control complexity, and maintenance capability.

8. Gain Saturation and Gain Equalization

An amplifier has finite pump power and finite inversion. As total input power increases, gain decreases and output power approaches saturation. In a WDM system, a strong channel may consume a disproportionate amount of available gain, causing weaker channels to receive less gain. Channel loading therefore affects gain, noise, and transient response.

Gain equalization seeks to make the net channel gain approximately uniform. Techniques include:

• gain-flattening filters placed within or after EDFAs;
• variable optical attenuators;
• dynamic gain equalizers;
• per-channel power control;
• amplifier pump-current control;
• Raman-pump spectral shaping; and
• closed-loop monitoring using optical-channel monitors.

If an amplifier has gain \(G_k\) for channel \(k\), a simple equalization target is to select attenuation \(A_k\) so that

\[

G_k-A_k\approx G_0,

\]

where \(G_0\) is the desired net gain. This relationship is a design approximation in dB and assumes that the equalizer attenuation does not introduce unacceptable noise or power penalties. In cascaded systems, equalization must consider the accumulated spectral shape, not merely the response of one amplifier.

A practical operating target is often a controlled power per channel at each amplifier input and output. Excessive equalization loss wastes power margin, while insufficient equalization causes channel-to-channel OSNR variation and may produce early saturation.

9. Illustrative Case: DWDM Metropolitan-to-Long-Haul Upgrade

Background

This illustrative case concerns a 16-channel metropolitan system operating over eight fibre spans. Each channel carries 100 Gb/s in the C-band. The operator wishes to expand to 40 channels without replacing the installed fibre.

Problem

The original terminal uses fixed 100-GHz filters and amplifiers designed for 16 channels. Expansion introduces three risks: insufficient passband capacity, gain tilt under the new loading, and inadequate OSNR after multiple amplifier stages.

Analysis

The engineering team first verifies the fibre attenuation and chromatic-dispersion map. It then measures the existing multiplexer passband, amplifier gain spectrum, output saturation, and per-channel OSNR. The new channel plan is placed on a standardized 100-GHz grid with reserved channels at both band edges. Measurements show that the existing amplifiers provide approximately 5 dB more gain at the short-wavelength side than at the long-wavelength side under the proposed loading.

Approach and solution

The operator installs gain-flattening filters and dynamic gain equalizers, upgrades the optical-channel monitor, and defines a launch-power target per channel rather than a single total-power target. Channels are introduced in stages. During commissioning, each channel is measured at the transmitter, after each span, and at the receiver. Automatic power control is configured to prevent a newly added channel from causing a transient gain increase in existing channels.

The network control system also reserves spare channels and records wavelength, power, alarm, and optical performance data. Before commercial operation, the design is tested under worst-case channel loading, amplifier failure, fibre repair, and channel add/drop conditions.

Outcome

The upgrade succeeds because the limiting issue was not the nominal fibre capacity alone; it was the interaction of amplifier saturation, filter response, OSNR, and operational transients. Equalization produces a more uniform channel power profile, while staged turn-up protects active services.

Learning points

1. Capacity expansion requires characterization of the complete optical path.

2. Total output power is not a sufficient indicator of channel quality.

3. Gain equalization must be designed for cascaded amplifiers and changing channel populations.

4. Operational procedures are part of optical-system engineering.

10. Design, Safety, and Governance Considerations

Best practice requires calibrated optical-power meters, wavelength meters, OSNR analyzers, and documented connector inspection and cleaning. Laser safety procedures must follow the applicable classification and national regulations. WDM systems may contain invisible, high-power radiation; open connectors should never be viewed directly.

Network operators should maintain accurate wavelength inventories, access controls, alarm logs, and configuration backups. Automated control must be designed with safe power limits and rollback mechanisms. AI-assisted monitoring can identify abnormal gain tilt, predict pump degradation, or detect gradual connector loss, but decisions affecting live channels require validated thresholds, explainable alarms, and human oversight. Training data should be protected because performance records can reveal network topology and operational vulnerabilities.

11. Contemporary Developments and Research Directions

Current development areas include flexible-grid elastic optical networks, sliceable transponders, C+L-band systems, probabilistic constellation shaping, space-division multiplexing, integrated silicon photonics, and open programmable line systems. Flexible-grid networks allocate spectrum according to channel bandwidth rather than a fixed slot size. This improves spectral efficiency but increases planning and filter-management complexity.

Distributed Raman amplification is being combined with EDFAs to extend reach and improve OSNR. Photonic integration is reducing package size and power consumption, while coherent digital signal processing compensates selected linear impairments and monitors channel quality. Nevertheless, digital compensation cannot eliminate all optical nonlinearities, amplifier noise, or inadequate physical-layer margins.

Chapter Summary

WDM increases fibre capacity by transmitting independent information channels on distinct optical wavelengths. CWDM uses wide spacing and economical components, whereas DWDM uses closely spaced standardized frequency grids and supports high capacity over amplified links. A sound wavelength plan must account for transmitter stability, filter passbands, fibre impairments, crosstalk, amplifier bandwidth, channel loading, and future expansion.

Optical power budgets use additive dB losses, but WDM performance is often governed by OSNR, gain saturation, nonlinear effects, and filter accumulation. EDFAs provide high-power, low-noise C-band amplification; Raman amplifiers offer flexible and potentially low-noise distributed gain; SOAs provide compact and fast integrated amplification but generally with greater noise and nonlinear penalties. Gain equalization is essential in cascaded WDM systems because amplifier gain varies with wavelength and channel loading. Safe commissioning, calibrated measurements, automated but supervised control, and documented wavelength governance are necessary for reliable operation.

Key Terms

WDM: Multiplexing independent optical channels on different wavelengths.

CWDM: WDM using relatively wide wavelength spacing.

DWDM: WDM using closely spaced standardized frequency channels.

Channel spacing: Frequency or wavelength separation between adjacent channels.

AWG: Arrayed-waveguide grating used for wavelength-selective multiplexing or demultiplexing.

ROADM: Reconfigurable optical add-drop multiplexer.

Insertion loss: Power loss introduced by an optical component.

Isolation: Suppression of an undesired channel at a selected port.

Crosstalk: Unwanted transfer of energy from one optical channel to another.

EDFA: Erbium-doped fibre amplifier.

Raman amplifier: Amplifier based on stimulated Raman scattering.

SOA: Semiconductor optical amplifier.

ASE: Amplified spontaneous emission noise.

Noise figure: Degradation of signal-to-noise ratio caused by an amplifier.

Gain saturation: Reduction of amplifier gain as input or total optical power increases.

Gain tilt: Variation of amplifier gain across the wavelength band.

Gain equalization: Adjustment of channel gains toward a common target.

OSNR: Optical signal-to-noise ratio measured over a specified reference bandwidth.

Flexible grid: WDM arrangement allowing variable spectrum slots for channels.

Review Questions

Short Answer Questions

1. Why is frequency spacing generally preferred to wavelength spacing for DWDM grid specification?

2. State two advantages and two limitations of CWDM.

3. What is the function of a gain-flattening filter?

4. Define OSNR and explain why it is important in amplified WDM links.

5. Name the principal physical mechanism used by an EDFA.

Long Answer Questions

1. Explain the architecture of a DWDM link, including transmitters, multiplexers, amplifiers, ROADMs, demultiplexers, and receivers.

2. Compare EDFAs, Raman amplifiers, and SOAs in terms of operating principle, gain, noise, bandwidth, saturation, and applications.

3. Develop a complete wavelength-plan design procedure for a multi-span DWDM network.

4. Explain the sources and consequences of crosstalk in WDM systems and describe mitigation methods.

5. Discuss gain saturation, gain tilt, ASE accumulation, and gain equalization in cascaded optical amplifiers.

Analytical / Application Questions

1. A 32-channel system uses 50-GHz spacing. Explain the measurements and calculations required to determine whether each channel fits within its slot after transmitter, multiplexer, and ROADM filtering.

2. A link has four amplifier spans with unequal gain tilt. Propose a monitoring and equalization strategy that remains stable when channels are added or removed.

3. An operator must choose between an EDFA and distributed Raman amplification for a long-reach link. Evaluate the choice using reach, OSNR, pump power, safety, cost, maintenance, and channel-loading criteria.

Multiple Choice Questions with Answers

Question 1

Which feature most directly distinguishes DWDM from CWDM?

A. DWDM uses much narrower channel spacing

B. DWDM requires electrical regeneration after every span

C. CWDM always operates only in the O-band

D. CWDM cannot carry digital signals

Correct Answer: A. DWDM uses much narrower channel spacing

Question 2

At approximately 1550 nm, a 1 nm wavelength separation corresponds to approximately:

A. 1.25 GHz

B. 12.5 GHz

C. 125 GHz

D. 1.25 THz

Correct Answer: C. 125 GHz

Question 3

Which amplifier is most strongly associated with high-power C-band multi-channel line amplification?

A. EDFA

B. SOA only

C. Photodiode amplifier

D. Electro-absorption modulator

Correct Answer: A. EDFA

Question 4

If a desired channel power is 1 mW and an interfering channel contributes 0.001 mW, the crosstalk ratio is:

A. −3 dB

B. −10 dB

C. −30 dB

D. −60 dB

Correct Answer: C. −30 dB

Question 5

Which quantity must be added linearly rather than directly in decibels when combining independent interferers?

A. Optical powers

B. Fibre lengths

C. Wavelength indices

D. Frequency-grid labels

Correct Answer: A. Optical powers

Question 6

An EDFA primarily obtains optical gain through:

A. Stimulated emission from erbium ions

B. Thermal expansion of fibre

C. Photovoltaic conversion

D. Four-wave mixing only

Correct Answer: A. Stimulated emission from erbium ions

Question 7

The main purpose of gain equalization in a WDM amplifier is to:

A. Make channel gains more uniform across the band

B. Convert optical signals into electrical signals

C. Increase connector loss

D. Eliminate all chromatic dispersion

Correct Answer: A. Make channel gains more uniform across the band

Question 8

A 40 km fibre has attenuation of 0.25 dB/km. Its fibre attenuation loss is:

A. 0.625 dB

B. 10 dB

C. 16 dB

D. 100 dB

Correct Answer: B. 10 dB

Question 9

Which amplifier commonly provides distributed gain along the transmission fibre?

A. Raman amplifier

B. EDFA only

C. SOA only

D. PIN receiver

Correct Answer: A. Raman amplifier

Question 10

For an amplifier noise figure expressed as a linear ratio F, the output SNR under the idealized noise-figure definition is:

A. F times the input SNR

B. The input SNR divided by F

C. The input SNR plus F in dB without conversion

D. Independent of F

Correct Answer: B. The input SNR divided by F