Electronics Guide

Frame Structure and Rates

The fundamental SONET building block is the Synchronous Transport Signal level 1 (STS-1), which carries a frame of 810 bytes transmitted 8,000 times per second for a line rate of 51.84 megabits per second. The frame is conventionally drawn as 9 rows by 90 columns, with the first 3 columns devoted to transport overhead and the remainder forming the synchronous payload envelope. The equivalent optical signal is the Optical Carrier level 1 (OC-1). The SDH counterpart, Synchronous Transport Module level 1 (STM-1), is three times larger at 155.52 megabits per second and aligns with OC-3.

Higher rates are formed by byte-interleaving the basic signals. Common levels include OC-3/STM-1 at 155.52 megabits per second, OC-12/STM-4 at 622.08 megabits per second, OC-48/STM-16 at 2.488 gigabits per second, OC-192/STM-64 at 9.953 gigabits per second, and OC-768/STM-256 at 39.813 gigabits per second. Each step multiplies the rate by four.

Contiguous concatenation joins payload capacity into a single clear channel rather than independent tributaries, which is convenient for packet traffic that does not subdivide neatly. SONET marks a concatenated signal with a "c" suffix on the transport signal, as in STS-3c or STS-48c. SDH instead names the concatenated container, using the form VC-4-Xc with X equal to 4, 16, 64, or 256; a VC-4-4c fills an STM-4 with one contiguous payload of 599.04 megabits per second. In both families the concatenated payload carries a single set of path overhead, so the network treats it as one indivisible circuit end to end.

Layered Overhead and Pointers

SONET and SDH define a layered overhead model that mirrors the network elements a signal traverses. Section (regenerator section in SDH) overhead is processed at every repeater and carries framing bytes and a section-level error check. Line (multiplex section in SDH) overhead is processed where signals are multiplexed and carries pointers and automatic protection switching commands. Path overhead travels end to end with the payload and includes a path trace identifier and a bit-interleaved parity check that lets the receiver confirm the signal arrived intact from its true source.

The pointer mechanism is a defining feature. Rather than forcing every tributary to start at a fixed byte position, SONET and SDH let the payload "float" within the frame and record its starting offset in the line overhead. Small frequency differences between nodes are absorbed by incrementing or decrementing the pointer, an operation called pointer justification. This avoids the slip buffers and bit stuffing that plagued earlier asynchronous systems.

Virtual Tributaries and Concatenation

To carry sub-rate services, the payload envelope is subdivided into virtual tributaries (SONET) or virtual containers (SDH). A VT1.5 transports a 1.544 megabit-per-second DS1, while a VC-12 transports a 2.048 megabit-per-second E1, matching the two regional digital telephony hierarchies. These structures are why a carrier could deliver a leased T1 or E1 circuit anywhere on its footprint and monitor it end to end without unwinding the high-rate signal at every node.

Virtual concatenation, added later, relaxes the rigid step sizes of contiguous concatenation. Rather than reserving a full VC-4-4c for a 200 megabit-per-second service, the endpoints group an arbitrary number of independent containers, route them separately if necessary, and reassemble them at the far end, wasting far less capacity. The Link Capacity Adjustment Scheme extends this by adding or removing members without interrupting traffic, and the Generic Framing Procedure adapts Ethernet, Fibre Channel, and other packet payloads into the containers. Together these mechanisms let carriers sell Ethernet services over installed SONET and SDH plant, which extended the life of that infrastructure considerably.

Legacy Status

SONET and SDH remain in service, particularly for legacy private-line and utility circuits, but they are no longer where new capacity is built. Their fixed hierarchy, their heavy bias toward 64 kilobit-per-second voice channels, and their lack of a standard forward error correction field all fit poorly with traffic that is now almost entirely packet data at 10 gigabits per second and above. Carriers have progressively migrated these services onto OTN and Carrier Ethernet, often emulating the legacy circuit over the newer transport so that customer equipment need not change.

The Optical Layer Model

ITU-T Recommendation G.872 defines the architecture that G.709 framing implements. Three optical layers stack beneath the digital containers. The optical transmission section spans a single fiber segment between amplifiers or regenerators. The optical multiplex section spans the aggregate multi-wavelength signal between the points where wavelengths are combined and separated. The optical channel spans a single wavelength end to end, from the transmitting transponder to the receiving one, possibly crossing several multiplex sections and many transmission sections along the way.

This layering matters operationally because each layer has its own failure modes and its own maintenance signals. A fiber cut affects a transmission section, an amplifier failure affects a multiplex section, and a transponder failure affects only one optical channel. Separating the layers lets management systems suppress the alarm storm a single fault would otherwise produce and identify the true root cause.

The ODU and OTU Container Hierarchy

G.709 defines a layered set of containers. The Optical Channel Payload Unit (OPU) holds the adapted client signal and a small amount of justification overhead that accommodates rate differences. The Optical Channel Data Unit (ODU) adds path-level overhead for tandem connection monitoring and maintenance signals. The Optical Channel Transport Unit (OTU) adds section-level overhead, frame alignment, and the forward error correction field, and it represents the signal actually launched onto a wavelength.

These containers come in rates designated by a numeric suffix, and each OTU runs slightly faster than the ODU it carries because of the added overhead and error-correction field. ODU1 runs at 2.499 gigabits per second and its OTU1 at 2.666, sized to transport an OC-48/STM-16 client. ODU2 runs at 10.037 gigabits per second and OTU2 at 10.709. ODU3 runs at 40.319 gigabits per second and OTU3 at 43.018. ODU4 runs at 104.794 gigabits per second and OTU4 at 111.810, sized to carry a 100 Gigabit Ethernet client. A low-order container, ODU0, was added at 1.244 gigabits per second to transport a single Gigabit Ethernet signal efficiently, and a flexible container, ODUflex, sizes itself to the client: for constant-bit-rate clients it matches the client rate directly, and for packet clients it is provisioned as a whole number of tributary slots.

Lower-rate ODUs are multiplexed into higher-rate ODUs through that tributary-slot structure, so a single OTU4 wavelength can groom a mixture of ODU0, ODU2, and ODUflex signals. The result is a digital cross-connect fabric operating at the transport layer: a carrier can hand off a 1, 10, or 100 gigabit-per-second service anywhere on the network without dedicating a whole wavelength to it, and can monitor that service independently of the wavelength that happens to carry it.

Frame Format and Overhead

The OTU frame is organized as 4 rows by 4,080 columns. The first 16 columns carry frame alignment and overhead, columns 17 through 3,824 carry the OPU and its client payload, and the final 256 columns (columns 3,825 through 4,080) carry the forward error correction. Unlike SONET and SDH, whose frame repetition rate is fixed at 8,000 hertz, the OTN frame period shortens as the line rate increases, so the byte structure stays constant while the clock scales. The overhead provides six levels of tandem connection monitoring, allowing each carrier in a multi-operator path to monitor only its own segment and to localize faults precisely.

Tandem connection monitoring is one of the practical advantages OTN holds over SONET and SDH. A circuit that crosses three carriers can be instrumented so that each carrier sees error counts for its own segment alone. When the end-to-end path degrades, the operators do not have to argue about whose network is at fault; the nested monitoring levels answer the question directly.

Beyond 100G: OTUCn, ODUCn, and FlexO

The OTUk hierarchy stopped at 100 gigabits per second, and defining a new fixed container for every future client rate proved impractical. Later revisions of G.709 therefore introduced a modular structure in which the suffix is the Roman numeral C, denoting 100, followed by a multiplier: an ODUCn is built from n instances of a roughly 100 gigabit-per-second slice, so an ODUC2 carries a 200 Gigabit Ethernet client and an ODUC4 a 400 Gigabit Ethernet client. The payload area, the OPUCn, is divided into tributary slots of nominally 5 gigabits per second, giving 20n slots. The coarser slot granularity suits modern clients such as 25 Gigabit Ethernet, 100 Gigabit Ethernet, and 400 Gigabit Ethernet far better than the 1.25 gigabit-per-second slots used at lower rates.

Because an ODUCn is defined independently of any particular optical interface, a separate mechanism carries it over real hardware. Flexible OTN (FlexO), specified in the G.709.1 series, bonds a group of 100, 200, or 400 gigabit-per-second physical interfaces into a single logical pipe of the required size. The arrangement mirrors Flexible Ethernet on the client side and decouples the digital container rate from the number and speed of the optical lanes available, so an operator can carry a 400 gigabit-per-second service over whatever combination of interfaces the installed transponders provide.

Dense and Coarse WDM

Dense WDM (DWDM) packs channels tightly within the low-loss fiber windows, predominantly the C-band near 1,550 nanometers and the adjacent L-band. ITU-T Recommendation G.694.1 defines a frequency grid anchored at 193.1 terahertz with spacings of 100, 50, 25, or 12.5 gigahertz. A conventional C-band system on the 50-gigahertz grid provides roughly 80 to 96 channels; extending into the L-band with a second amplifier chain roughly doubles that count. The narrow spacing demands temperature-stabilized lasers and precise filters, but it maximizes the number of carriers per fiber.

Coarse WDM (CWDM), defined in G.694.2, instead spaces channels 20 nanometers apart on a grid running from 1,271 to 1,611 nanometers, for 18 channels in total. The relaxed spacing tolerates the wavelength drift of an uncooled laser over temperature, so CWDM transceivers need no thermoelectric cooler and use inexpensive thin-film filters. That economy makes CWDM attractive for shorter metropolitan, access, and mobile fronthaul links where channel count matters less than cost. Its drawback is that most CWDM wavelengths fall outside the erbium amplifier's gain window, so CWDM spans are limited to what the transmitter power and receiver sensitivity can reach unaided.

Optical Amplification and the Flexible Grid

Long-haul DWDM depends on optical amplifiers that boost all channels at once without converting back to electrical form. The erbium-doped fiber amplifier (EDFA), which provides gain across the C-band, made transoceanic and continental DWDM economical by eliminating per-channel regenerators. Raman amplification, which uses the fiber itself as a distributed gain medium pumped by high-power lasers, extends reach further and improves the signal-to-noise ratio. More recent systems adopt a flexible grid, added to G.694.1 in 2012, that allocates spectrum in slot widths that are multiples of 12.5 gigahertz, positioned on a 6.25-gigahertz central-frequency granularity, so that superchannels using advanced modulation can occupy exactly the bandwidth they need rather than a fixed channel slot.

Amplification is not free. Every amplifier adds spontaneous-emission noise, so the optical signal-to-noise ratio degrades with each span, and the accumulated noise ultimately sets the reach of a channel. Raising launch power to compensate eventually backfires, because the fiber's nonlinear refractive index converts excess power into distortion. Line engineering therefore searches for the launch power that balances noise against nonlinearity, and low-loss, large-effective-area fiber types such as those in the ITU-T G.652 and G.654 families widen that operating window on long routes.

Coherent Transmission and Pluggable Optics

Modern coherent transceivers transmit far more than one bit per symbol by modulating both amplitude and phase across two polarizations, using formats such as polarization-multiplexed quadrature phase-shift keying and higher-order quadrature amplitude modulation. A coherent receiver mixes the incoming signal with a local-oscillator laser and digitizes the result, which preserves the full optical field and lets digital signal processing undo chromatic dispersion and polarization-mode dispersion electronically. Optical dispersion-compensating modules, once required on every span, disappeared from the line system as a result.

The same digital signal processing makes capacity tunable. A given transceiver can trade modulation order, symbol rate, and error-correction overhead against reach, so an operator provisions the highest rate a particular route will support rather than accepting a fixed channel size. Commercial coherent interfaces now span 100 to 800 gigabits per second per wavelength, with higher rates reserved for shorter, higher-quality routes.

Coherent optics have also shrunk from line cards into pluggable modules. The Optical Internetworking Forum's 400ZR specification defines a 400 Gigabit Ethernet interface in a QSFP-DD module, targeting amplified point-to-point DWDM links of roughly 80 to 120 kilometers for data-center interconnect. The industry OpenZR+ specification extends the same form factor with the stronger open forward error correction from OpenROADM, selectable 100, 200, 300, and 400 gigabit-per-second line rates, and greater dispersion tolerance, reaching well beyond 400ZR. Because these modules plug directly into router and switch ports, they let operators build an "IP over DWDM" network that eliminates a separate transponder shelf, at the cost of managing the optical layer from equipment that was not traditionally optical.

Switching Technology

The core of a modern ROADM is the wavelength-selective switch, commonly built with liquid-crystal-on-silicon or microelectromechanical mirror arrays. The switch disperses the incoming spectrum with a grating, steers each wavelength independently to a chosen output port by tilting the corresponding pixels or micromirrors, and recombines the result. Because the steering element is addressed pixel by pixel, a wavelength-selective switch also supports the flexible grid, passing spectral slices of arbitrary width rather than fixed channels.

A ROADM node is characterized by its degree, meaning the number of fiber directions it terminates. A two-degree node sits inline on a route; a four- or eight-degree node is a junction where several routes meet, and it requires a wavelength-selective switch per direction plus an interconnect structure among them. Degree count drives cost, footprint, and the optical loss the amplifiers must make up.

ROADM add-drop architectures are described by three further properties. A colorless add-drop port can receive any wavelength rather than being hardwired to one. A directionless port can route an added or dropped wavelength toward any fiber direction at a multi-degree node. A contentionless design allows the same wavelength to be added or dropped on multiple directions simultaneously without internal blocking. A ROADM combining all three, abbreviated CDC (and CDC-F when it also handles the flexible grid), gives operators the greatest freedom to assign and reroute wavelengths dynamically, which is what allows a control plane to restore a failed circuit onto a different route in the optical domain rather than through an electrical regeneration hop.

Linear and Ring Protection

Linear automatic protection switching reserves a backup path for a working path between two points. In 1+1 protection, the signal is transmitted simultaneously on both paths and the receiver selects the better one, giving the fastest recovery. In 1:1 or 1:N protection, the backup path is shared and is activated only when a fault is detected, conserving capacity at the cost of a signaling exchange before the switch.

Ring topologies dominated the SONET and SDH era and appear in two forms. The unidirectional path-switched ring, known in SDH as subnetwork connection protection, bridges each circuit onto both directions of the ring and lets the receiving node choose the healthier copy; it is simple and fast but consumes half the ring capacity. The bidirectional line-switched ring, known in SDH as the multiplex section shared protection ring, instead reserves protection capacity that all spans share and loops traffic back around the surviving fibers when a span fails; it carries more working traffic but requires signaling among the nodes. Both were engineered to the classic target of restoring service within 50 milliseconds, short enough that voice calls and legacy circuits survive the event unbroken. OTN inherits the same discipline through the linear and ring protection schemes of ITU-T Recommendation G.873.

Mesh Restoration and Control Planes

Mesh networks restore traffic by computing an alternate route across the topology after a failure, which uses capacity more efficiently than dedicated ring protection but generally restores more slowly. Distributed control planes such as Generalized Multiprotocol Label Switching (GMPLS), and increasingly centralized software-defined networking controllers, automate path computation, signal the new route to each node, and reconfigure ROADMs and cross-connects. These control planes also enable services such as bandwidth-on-demand and rapid provisioning of new wavelength circuits.