Electronics Guide

Wavelength and Frequency

Optical communications specify channels either by wavelength in nanometers or by frequency in terahertz, the two being related by the speed of light. ITU-T Recommendation G.694.1 defines the standardized dense WDM frequency grids, anchored to a reference frequency of 193.1 THz, which corresponds to a wavelength of 1552.52 nm. Channel spacings of 100 GHz, 50 GHz, 25 GHz, or 12.5 GHz define the available slots in the optical spectrum.

The C-band, spanning roughly 1530 to 1565 nm, is the most commonly used spectral region because it combines the loss minimum of silica fiber with the gain window of erbium-doped fiber amplifiers. Its roughly 4.4 THz of usable bandwidth holds about 88 channels on a 50 GHz grid, and modern extended C-band amplifiers stretch that to approximately 6 THz. The L-band from 1565 to 1625 nm provides comparable additional capacity with extended-band EDFAs. Newer systems also reach into the S-band (1460 to 1530 nm), where thulium-doped fiber amplifiers or Raman gain replace erbium.

Channel Capacity Limits

The maximum data rate per channel depends on the spectral efficiency achievable with the modulation format and on the available signal-to-noise ratio. Simple on-off keying achieves about 1 bit per second per hertz of bandwidth, while advanced coherent formats with higher-order QAM can exceed 6 bits per second per hertz. For example, a polarization-multiplexed 16-QAM signal carrying 400 Gbps in a 75 GHz slot reaches roughly 5.3 bits per second per hertz. The Shannon-Hartley limit establishes the theoretical maximum based on channel bandwidth and SNR.

Every format trades spectral efficiency against reach. Polarization-multiplexed QPSK carries 4 bits per symbol and tolerates low OSNR, so it serves transoceanic spans; PM-16QAM doubles that to 8 bits per symbol but demands roughly 7 dB more OSNR, confining it to regional and metro distances; PM-64QAM reaches 12 bits per symbol over short, high-quality links. Modern transponders blur these steps with probabilistic constellation shaping, which adjusts the statistical distribution of constellation points to tune capacity in fine increments and to recover part of the theoretical shaping gain.

Quoted line rates also include overhead. Forward error correction adds roughly 15 to 30 percent redundancy, so a transponder delivering 400 Gbps of client traffic transmits a noticeably higher symbol payload on the fiber. Soft-decision FEC provides coding gains that translate directly into span budget, and the choice of FEC is often what separates two otherwise identical products in reach.

Optical Bandwidth and Grid Flexibility

Traditional fixed-grid systems allocate channels on the standardized spacing regardless of actual signal bandwidth, which wastes spectrum when a signal is narrower than its slot and forbids the signal entirely when it is wider. Flexible-grid (flexi-grid) systems, introduced in the 2012 revision of ITU-T G.694.1, define frequency slots as integer multiples of a 12.5 GHz slot-width granularity, with nominal central frequencies placed on a 6.25 GHz raster. The finer center-frequency raster exists so that slots an odd number of 12.5 GHz increments wide can be packed adjacently without leaving gaps.

The practical effect is that each signal consumes only the spectrum it needs. A 400 Gbps PM-16QAM carrier fits comfortably in 75 GHz, while a 1.6 Tbps single-carrier signal at roughly 200 GBaud requires on the order of 200 GHz. On a rigid 50 GHz grid the latter would be unusable. Super-channels, which combine several carriers into one managed entity provisioned and switched as a unit, extend the same idea to capacities beyond what a single carrier supports.

DWDM Characteristics

Dense wavelength division multiplexing (DWDM) uses closely spaced channels, typically at 100 GHz (0.8 nm) or 50 GHz (0.4 nm) intervals. The term "dense" distinguishes these systems from coarse WDM with wider spacing. DWDM is the technology of choice for long-haul and submarine telecommunications links, where maximizing fiber capacity justifies the investment in precision optical components.

Wavelength Stability Requirements

DWDM systems require transmitter wavelengths stable within a small fraction of the channel spacing to prevent interference with adjacent channels. A distributed feedback laser drifts roughly 0.1 nm per degree Celsius, which would sweep across several 50 GHz channels over a normal equipment temperature range, so every DWDM source sits on a thermoelectric cooler under closed-loop control. A wavelength locker, typically an etalon-based discriminator that compares transmitted and reflected power, holds the carrier to within a few picometers of the assigned grid frequency. Adherence to the ITU-T grid is what allows transponders, filters, and line systems from different vendors to interoperate.

DWDM Components

Key DWDM components include distributed feedback lasers or external cavity lasers with precise wavelength control, thin-film or arrayed waveguide grating multiplexers and demultiplexers, optical amplifiers with flat gain across the operating band, and dispersion compensation modules. High-performance optical filters with steep roll-off enable tight channel spacing.

Tunable lasers have displaced fixed-wavelength sources in most modern equipment. A single tunable transponder covering the full C-band replaces the dozens of wavelength-specific spares an operator would otherwise stock, and it allows a wavelength to be reassigned by software when the network is reconfigured. External cavity lasers and sampled-grating distributed Bragg reflector designs provide the necessary tuning range with the narrow linewidth that coherent detection demands, since phase noise in the transmit or local-oscillator laser directly limits how high an order of QAM the receiver can resolve.

Amplification in DWDM Systems

Erbium-doped fiber amplifiers provide simultaneous amplification of all C-band DWDM channels, eliminating the need for per-channel regeneration. The intrinsic erbium gain spectrum is far from flat, so gain-flattening filters trim the peak near 1532 nm and equalize the response across the band; without them, channel powers would diverge by many decibels after a handful of cascaded amplifiers. Distributed Raman amplification, which pumps the transmission fiber itself, adds gain where it improves noise performance most and extends coverage to wavelengths erbium cannot reach.

Amplifiers must also handle changing channel loading gracefully. When a fiber cut or a wavelength deletion removes many channels at once, the surviving channels would otherwise absorb the freed inversion energy and surge in power. Automatic gain control and transient suppression circuits react within microseconds to hold per-channel power constant, protecting downstream receivers from overload.

CWDM Characteristics

Coarse wavelength division multiplexing (CWDM) uses a channel spacing of 20 nm, enabling the use of less expensive uncooled laser sources with wider wavelength tolerance. ITU-T Recommendation G.694.2 defines an 18-channel grid spanning 1271 to 1611 nm, though practical deployments often use a subset of 8 to 16 channels. The wider spacing relaxes filter requirements and reduces overall system cost.

The 20 nm spacing is not arbitrary. It accommodates the combined wavelength uncertainty of an uncooled laser across its full operating temperature range, plus manufacturing tolerance and aging drift, with margin left for the filter passband. Because the grid is specified in wavelength rather than frequency, CWDM channels are conventionally named by their nominal center wavelength, such as the widely deployed eight-channel set from 1471 to 1611 nm that sits above the water absorption region.

CWDM Applications

CWDM is well-suited for enterprise networks, metropolitan area networks, and access network applications where moderate capacity increases are needed without the cost of DWDM. Typical applications include campus backbone links, storage area network extensions, and cable television systems. CWDM can also provide overlay capacity on existing single-channel fiber links.

Mobile fronthaul is a prominent modern use. Where an operator has one fiber pair, or even one fiber, running to a cell site, passive CWDM multiplexers let several radio units share it without active equipment in the field. Related medium-density grids with narrower spacing serve the same role when more channels are needed than the coarse grid provides. Because the multiplexer is entirely passive, it consumes no power, needs no management interface, and survives the temperature extremes of an outdoor cabinet.

Cost Advantages

The primary advantage of CWDM is reduced component cost. Uncooled lasers eliminate the thermoelectric cooler and its control circuitry. Wider filter passbands use simpler thin-film or polymer technologies. These savings are significant in cost-sensitive applications, though CWDM capacity is inherently limited compared with DWDM.

Distance Limitations

CWDM systems face attenuation penalties on the channels affected by the water absorption peak near 1383 nm. Low-water-peak fiber, specified in ITU-T G.652.C and G.652.D, eliminates this absorption, enabling all CWDM channels to achieve similar reach. Because no single optical amplifier spans all CWDM wavelengths, these systems are generally limited to the reach of unamplified transmission.

Thin-Film Filters

Thin-film interference filters use stacks of dielectric layers with precisely controlled thicknesses to create wavelength-selective transmission and reflection. Each filter passes one channel and reflects the rest, so channels are added or dropped one at a time by cascading filters in a serial chain. This construction is simple and inexpensive, and it excels at low channel counts, but insertion loss accumulates along the chain: the last channel in a long cascade traverses every preceding filter. Thin-film devices therefore dominate CWDM multiplexers and small DWDM add-drop modules, while high-channel-count DWDM multiplexing turns to other technologies.

Because the passband of a tilted thin-film filter shifts with angle of incidence and with temperature, mechanical stability and athermal packaging matter as much as the coating design itself. Adjacent-channel isolation of 25 dB or more is a normal requirement so that leakage from neighboring wavelengths does not degrade the recovered signal.

Arrayed Waveguide Gratings

Arrayed waveguide gratings (AWGs) provide integrated multiplexing and demultiplexing of many channels in a single planar lightwave circuit. Light entering a slab coupler spreads across an array of waveguides whose lengths differ by a constant increment; the resulting wavelength-dependent phase tilt refocuses each wavelength onto a different output port in a second slab coupler. Because all channels are separated in one device rather than one filter at a time, insertion loss is essentially uniform across ports and does not grow with channel count. This property makes AWGs the technology of choice for 40-, 80-, and 96-channel DWDM multiplexers.

The chief engineering challenge is thermal stability, since the refractive index of silica waveguides varies with temperature and shifts the whole channel comb. Devices are either actively temperature-controlled with a heater and thermistor or built as athermal AWGs, in which a compensating material with an opposite thermo-optic coefficient is inserted into the light path. The same planar platform supports interleavers, which split an incoming comb into odd and even channels so that two multiplexers with relaxed spacing can serve a grid twice as dense.

Fiber Bragg Gratings

Fiber Bragg gratings (FBGs) are periodic refractive index variations written into optical fiber that reflect specific wavelengths while transmitting others. FBG-based multiplexers cascade multiple gratings, often paired with optical circulators, to separate channels. The technology also provides dispersion compensation and wavelength reference functions in WDM systems.

Echelle Gratings and Other Technologies

Echelle diffraction gratings, used at high diffraction orders, achieve strong dispersion in a compact free-space or planar geometry and serve as an alternative to the AWG where footprint matters. Microring resonators take the opposite approach: a ring coupled to a bus waveguide extracts only the wavelengths matching its resonance, and a cascade of rings with slightly different radii demultiplexes a comb in an area far smaller than any grating device.

The trade-off is sensitivity. A ring resonance shifts by roughly 0.1 nm per degree Celsius in silicon, and nanometer-scale fabrication variation alters the radius enough to move a channel off grid, so practical microring multiplexers require integrated heaters with per-ring feedback control. The resulting power and calibration overhead is the principal obstacle to their wider use, and reducing it is an active area of silicon photonics research.

Wavelength-Selective Switches

Wavelength-selective switches (WSS) route individual wavelengths from an input port to any of multiple output ports, enabling flexible optical networks. A diffraction grating disperses the incoming channels across a steering element, which deflects each wavelength toward its chosen output fiber, and the light is then recombined. Liquid-crystal-on-silicon (LCoS) implementations have largely displaced MEMS micromirrors, because addressing a two-dimensional pixel array lets the device define arbitrary passband edges in software rather than at the granularity of a fixed mirror. This is precisely what flexible-grid operation requires: the same hardware can carve a 37.5 GHz slot or a 200 GHz slot on command.

Beyond routing, a WSS provides per-channel attenuation. Setting an independent loss for every wavelength is how operators equalize channel powers across a link and compensate for the wavelength-dependent gain of the amplifier chain. Blocking a channel is simply the limiting case of attenuating it, which lets the same device enforce wavelength isolation between network segments.

ROADM Architecture

Reconfigurable optical add-drop multiplexers (ROADMs) use wavelength-selective switches to dynamically add, drop, or pass through wavelength channels at network nodes. A node is characterized by its degree, the number of fiber directions it terminates; a two-degree node sits in a line, while a four- or eight-degree node anchors a mesh. Each degree carries its own WSS pair, and a local add-drop structure connects transponders to any of them.

The three qualifiers describing add-drop flexibility have precise meanings. Colorless means any transponder port accepts any wavelength, removing the fixed wavelength-to-port binding of early designs. Directionless means a transponder can be steered toward any degree of the node without a technician moving a fiber. Contentionless means the same wavelength can be added or dropped more than once at a node, toward different directions, which earlier designs blocked because two identical wavelengths would collide in a shared multiplexer. Together, colorless, directionless, and contentionless (CDC) architectures allow a wavelength to be provisioned or rerouted entirely from a management system, which is the foundation of software-defined optical networking. The flexibility costs money and optical loss, so many operators deploy CDC at core nodes and simpler fixed add-drop structures at the edge.

Optical Cross-Connects

Large-scale optical cross-connects switch wavelengths between multiple fiber pairs, enabling mesh network topologies without electrical conversion. These systems combine wavelength-selective switching with fiber switching to create flexible, high-capacity network nodes. All-optical switching avoids the power consumption and latency of electrical regeneration.

Network Management and Control

Software-defined networking principles applied to optical networks enable centralized control of wavelength routing across the network. Optical channel monitors track power levels and wavelengths throughout the system. Automatic power control and wavelength drift compensation maintain performance as conditions change.

The control layer depends on an accurate optical model. Before a controller may turn up a lightpath, it must predict whether the route will meet its OSNR and nonlinear penalty budget, which requires knowledge of span losses, amplifier gain profiles, and the interaction between the new channel and its neighbors. Planning tools traditionally answered this question offline with conservative margins. Modern systems increasingly maintain a continuously calibrated model fed by live telemetry, narrowing those margins and allowing capacity that would otherwise have been held in reserve to be sold. Streaming telemetry and standardized data models are what make such closed-loop operation possible across equipment from multiple suppliers.

Optical Signal-to-Noise Ratio Budgeting

OSNR is the master budget of a WDM link. Each amplifier adds spontaneous emission noise, and because the noise accumulates while the signal is merely restored to its original level, OSNR degrades monotonically along the chain. For an amplified link of identical spans, the received OSNR in a 0.1 nm reference bandwidth follows a compact expression: 58 dB plus the per-channel launch power in dBm, minus the span loss in dB, minus the amplifier noise figure in dB, minus ten times the base-ten logarithm of the number of spans.

The formula makes the design levers explicit. Doubling the span count costs 3 dB. A 100 km span of standard fiber at roughly 0.2 dB/km imposes about 20 dB of loss, so shortening spans or reducing splice and connector loss buys OSNR directly. A good EDFA contributes a noise figure near 5 dB, and distributed Raman amplification can produce an effective noise figure below that of any discrete amplifier, which is why long-haul systems often combine the two. Raising launch power appears to help without limit, but nonlinear impairments impose a ceiling, and the optimum sits a few decibels below the point where nonlinear penalty overtakes noise improvement.

Chromatic Dispersion

Chromatic dispersion causes different wavelengths to travel at different speeds, spreading optical pulses over distance. WDM systems must manage dispersion across all channels, either through dispersion-shifted fiber, dispersion compensating fiber, or electronic compensation in coherent receivers. The optimal dispersion map may differ across channels in a wide-bandwidth system because the dispersion of standard single-mode fiber varies with wavelength.

Fiber Nonlinearities

High optical power densities in WDM systems activate fiber nonlinear effects, including self-phase modulation, cross-phase modulation, and four-wave mixing. These effects create interchannel crosstalk and signal distortion. System design balances launch power to maximize optical signal-to-noise ratio (OSNR) while limiting nonlinear impairments. Unequal channel spacing can mitigate four-wave mixing, which is most severe where local dispersion is near zero.

The dispersion dependence of four-wave mixing produced one of the field's instructive reversals. Dispersion-shifted fiber, engineered to place the dispersion zero at 1550 nm, was excellent for single-channel systems but proved poorly suited to dense WDM, because near-zero local dispersion keeps neighboring channels phase-matched and lets mixing products build coherently. Non-zero dispersion-shifted fiber was standardized to restore a small, deliberate amount of dispersion that destroys the phase matching while remaining easy to compensate.

Stimulated Raman scattering becomes a first-order concern in wideband systems. It transfers power from shorter to longer wavelengths across the transmitted comb, tilting the spectrum so that blue channels are depleted and red channels are amplified. Across a combined C+L band spanning roughly 10 THz, this inter-band tilt can amount to several decibels per span and must be pre-compensated by shaping the launch spectrum and by adjusting amplifier gain profiles.

Polarization Effects

Polarization mode dispersion and polarization-dependent loss vary with wavelength and can differ significantly across a WDM band. Coherent receivers track and compensate polarization effects digitally, but system design must ensure adequate OSNR margin for worst-case polarization conditions on all channels. Because the state of polarization drifts with temperature and mechanical disturbance, and can rotate rapidly when a cable is struck or a lightning surge passes nearby, the receiver's tracking loop must be fast enough to follow transients as well as slow drift.

Filter Cascading and Passband Narrowing

A wavelength crossing a mesh network passes through the filters of every ROADM along its path. Each passband is slightly narrower than the nominal slot, and the intersection of many such passbands is narrower still, so the effective channel width shrinks with each node traversed. The resulting spectral clipping distorts the signal and costs OSNR margin, and it sets a practical limit on how many nodes a wavelength may cross before regeneration becomes necessary.

Design responses include specifying filters with steep, well-controlled edges, allocating a slightly wider flexible-grid slot than the signal strictly requires, and modeling the cumulative penalty during network planning so that lightpaths exceeding the budget are routed differently or assigned a lower-order modulation format. Transmitter-side pulse shaping that keeps the signal spectrum compact also reduces sensitivity to the effect.

Channel Monitoring and Management

Optical channel monitors measure power and wavelength for each channel at key network points. Optical spectrum analyzers provide detailed characterization during installation and troubleshooting. Channel power equalization maintains uniform performance across all wavelengths despite varying transmitter powers and wavelength-dependent losses.

Point-to-Point Systems

The simplest WDM architecture multiplexes all channels at one end, transmits through amplified fiber spans, and demultiplexes at the far end. This topology is common for long-haul terrestrial and submarine links, where traffic naturally aggregates between major population centers.

Submarine systems are the extreme case and illustrate a different optimization. Repeaters on the seabed draw power from a conductor in the cable fed from shore, so total electrical power, not spectrum, is the binding constraint. That inverts the usual design logic: rather than maximizing capacity per fiber pair, modern transoceanic cables carry more fiber pairs operated at lower power each, because amplifier efficiency improves when a pair runs below its capacity-optimal launch power. Repeater spacing of roughly 50 to 100 km, a design life measured in decades, and the impossibility of convenient repair also push these systems toward conservative margins and highly proven components.

Ring Networks

Metropolitan networks often use ring topologies with add-drop nodes accessing wavelength channels around the ring. Protection switching can restore service within tens of milliseconds, traditionally targeting the 50 millisecond bound inherited from SONET/SDH, by routing traffic the opposite direction around the ring. Wavelength reuse on different ring segments increases effective capacity.

Mesh Networks

Mesh topologies with multiple interconnected nodes provide resilience and routing flexibility. Wavelength routing through optical cross-connects creates end-to-end lightpaths without intermediate electrical conversion. Network planning algorithms solve the routing and wavelength assignment (RWA) problem to optimize capacity across the mesh.

Passive Optical Networks

WDM-PON uses different wavelengths to serve different subscribers from a shared optical line terminal, providing dedicated bandwidth to each user. This architecture offers higher capacity than time-division PON while maintaining a passive outside plant. WDM-PON is attractive for business services and 5G mobile fronthaul.

Optical Transport Network Framing

Client signals rarely travel over a wavelength in their native form. The Optical Transport Network hierarchy defined in ITU-T Recommendation G.709 wraps each client in a digital frame that carries forward error correction, performance monitoring counters, and management overhead, then maps that frame onto an optical channel. The wrapper is what makes a wavelength manageable: an operator can measure bit error rate, localize a fault to a specific span, and verify a service-level guarantee without interpreting the client protocol at all. It also allows a single wavelength to carry a mixture of Ethernet, storage, and legacy time-division traffic multiplexed into one container.

Coherent Pluggable Transceivers

Coherent optics have shrunk from line cards to pluggable modules in standard router and switch cages, which changes who owns the transponder. The OIF 400ZR Implementation Agreement, published in 2020, specifies an interoperable 400 Gigabit Ethernet interface using polarization-multiplexed 16-QAM at 59.84 GBaud with concatenated FEC, targeting amplified point-to-point DWDM links of 120 km or less and unamplified links to about 80 km. It deliberately trades reach for low power and small size, suiting the data center interconnect application it was designed for.

The OpenZR+ multi-source agreement extends the same form factor with a stronger open FEC and additional line rates, reaching metro and regional distances and adding OTN framing that 400ZR omits. An 800ZR agreement carries the approach to the next rate. Where these pluggables apply, they let an operator run a wavelength directly from a router port into the line system, eliminating the short-reach optics and the separate transponder chassis that a traditional architecture requires.

Open Line Systems and Alien Wavelengths

Traditionally a vendor supplied the transponders and the amplifier and ROADM line system together, guaranteeing performance for the combination. Disaggregation separates the two. An open line system provides amplification, wavelength routing, and spectrum management, while transponders from any supplier occupy the spectrum; a wavelength generated by third-party equipment is called an alien wavelength.

The benefit is that transponder technology, which advances quickly, can be refreshed without replacing the line system, which advances slowly and is expensive to touch. The cost is that responsibility for the optical budget shifts to the operator. Open APIs, standardized performance monitoring, and careful modeling of amplifier gain, filter penalties, and nonlinear interaction between foreign and native channels are what make the arrangement workable in practice.

Coherent WDM Systems

Coherent detection with digital signal processing has transformed WDM systems by enabling higher-order modulation formats, electronic compensation of linear impairments, and polarization multiplexing that doubles spectral efficiency. Because the receiver recovers the full optical field rather than only intensity, chromatic dispersion and polarization mode dispersion become linear distortions that a digital filter can invert. This eliminated the inline dispersion compensating fiber that earlier systems required, simplifying amplifier huts and removing the loss and nonlinearity those modules introduced.

Commercial coherent transponders now span 400 Gbps to 1.6 Tbps per wavelength, using formats from QPSK to 64-QAM combined with probabilistic constellation shaping. Capacity at the top of that range comes from symbol rate as much as from constellation order: single-carrier 1.6 Tbps operation runs near 200 GBaud and consequently occupies roughly 200 GHz of spectrum, so only about two dozen such channels fit the C-band. High-rate wavelengths therefore depend on flexible-grid line systems, and the choice between many modest channels and few very wide ones becomes a network planning decision rather than a fixed property of the equipment.

Super-Channels and Nyquist WDM

Super-channels combine multiple closely spaced carriers into a single manageable entity, simplifying high-capacity transmission. Nyquist WDM uses pulse shaping to minimize spectral guard bands between channels, approaching the theoretical limit of spectral efficiency. These techniques maximize capacity within the available amplifier bandwidth.

Ultra-Wideband WDM

Extending WDM beyond the C-band and L-band exploits more of the low-loss window of installed silica fiber, which is attractive because it multiplies capacity without trenching new cable. Combined C+L systems are now commercially deployed and place roughly 12 THz of spectrum on a fiber pair, about double a C-band-only system. The S-band is the next step: erbium provides no gain there, so amplification relies on thulium-doped fiber amplifiers, semiconductor optical amplifiers, or multi-pump Raman gain. Laboratory and field trials have demonstrated transmission across combined S-, C-, and L-band spectrum exceeding 15 THz, with reported throughput above 200 Tbps on field-deployed fiber.

Practical multi-band systems face problems that a single-band design never encounters. Stimulated Raman scattering couples the bands together, so power added in one region depletes another. Chromatic dispersion differs substantially between the short and long ends of the spectrum, requiring per-band equalization. Fiber loss rises outside the 1550 nm minimum, shortening reach for the outer bands, and every band needs its own amplifier, multiplexer, and transponder family. Whether multi-band transmission or space-division multiplexing offers the better cost per bit remains an open question that depends on the price of the components each approach requires.