DWDM Explained: Dense Wavelength Division Multiplexing

DWDM (Dense Wavelength Division Multiplexing) is the technology that carries virtually all intercontinental traffic: it combines dozens of optical channels, each on a different wavelength, onto a single fiber pair, multiplying capacity without laying new cable. This tutorial explains how a DWDM system works - transmitters, multiplexers, EDFAs and demultiplexers - and why the C-band, channel spacing and fiber type all matter.

What Is DWDM?

DWDM is an optical multiplexing technique that combines multiple discrete transport channels, each using a different wavelength, and transmits them over a single optical fiber. If you multiplex 32 STM-1 signals, the fiber capacity jumps from 2.5 Gbps to 80 Gbps. At the receiving end, a demultiplexer separates the combined wavelengths back into individual channels.

It helps to picture a fiber as a highway that is already built. DWDM does not widen the road, it adds lanes. Each lane carries an independent signal at its own wavelength, so a fiber pair that once carried a single 10G link can carry 40, 80 or 96 channels of 10G, 100G or 400G with no re-trenching, no new duct space and no new right-of-way negotiations.

Three physical properties make this practical. First, light near 1550 nm travels with very low loss - about 0.2 dB/km in modern single-mode fiber - so signals survive tens of kilometres before they need help. Second, an erbium-doped fiber amplifier can boost every channel simultaneously, because amplification happens in the optical domain instead of per-channel in electronics. Third, the ITU-T wavelength grid gives every vendor a common set of anchors, so a transmitter from one supplier can be added to a system built by another.

Why DWDM Matters: The Capacity Economics

Laying new fiber is the single most expensive item in a transport network. Material cost is minor compared with civil works: permits, trenching, duct, splicing and re-instatement can consume 70-80 percent of a metro or long-haul build budget, and the schedule is measured in months or years. DWDM attacks that cost curve directly.

• Pay-per-lane growth: you light up additional wavelengths only when demand arrives, so capital spending matches revenue.
• Protocol transparency: a wavelength carries whatever you put on it - 10G/100G Ethernet, OTN, Fibre Channel, SDH, CPRI fronthaul - because the optical layer does not inspect the payload.
• Reach without regeneration: optical amplification plus dispersion management pushes reach to hundreds or thousands of kilometres before an OEO regeneration is needed.
• Stranded-capacity recovery: dark fiber already leased from a carrier is monetised many times over.

The classic entry point is a simple point-to-point system: two sites, one fiber pair, a passive MUX at each end and a pair of amplifiers. Everything else - ring, mesh, ROADM - is an evolution of that building block.

DWDM System Components

• Transmitters - lasers, one per channel, each modulated at a specific wavelength.
• Multiplexer (MUX) - passive device that merges the wavelengths into one fiber.
• Demultiplexer (DEMUX) - separates the combined light back into individual wavelengths.
• Optical isolator - minimizes back reflection.
• EDFA - erbium-doped fiber amplifier that boosts all channels at once.
• OADM - optical add-drop multiplexer that adds or drops specific wavelengths at intermediate nodes.

In a modern system these functions are packaged as transponders (one client signal in, one wavelength out), muxponders (several lower-rate clients groomed into one high-rate wavelength), ROADMs (reconfigurable add/drop with wavelength-selective switching, so a node can drop 100G to one direction and pass the rest through), and optical supervisory channel equipment that carries management traffic on its own out-of-band wavelength.

A useful mental model is the layered stack defined by ITU-T G.872: the optical channel layer (OCh, the end-to-end wavelength), the optical multiplex section (OMS, between amplifiers) and the optical transmission section (OTS, between two adjacent passive/active elements). Most troubleshooting is an exercise in figuring out which of those three layers has failed.

The ITU-T Wavelength Grid

DWDM channels are not placed arbitrarily. ITU-T G.694.1 defines an absolute frequency grid anchored at 193.1 THz (approximately 1552.52 nm) with channel spacing of 12.5, 25, 50, 100 GHz and wider multiples. Working in frequency rather than wavelength is deliberate: frequency spacing is what determines crosstalk between channels, and the grid stays valid as the speed of the light through the fiber changes with temperature.

Practical consequences of the grid are worth remembering:

• A 100 GHz grid channel count in C-band is roughly 45 usable channels; 50 GHz doubles that to about 90-96.
• Narrower spacing means tighter laser tolerances, better filters, and more sensitivity to non-linear effects.
• Flexible-grid (flexgrid) systems abandon fixed spacing entirely and allocate arbitrary width slots - the foundation of 400G and 800G transport where a channel may occupy 75 GHz or 100 GHz by itself.

For a detailed walk-through of the trade-off, see our DWDM 50GHz vs 100GHz channel spacing on the ITU grid article.

Optical Fiber Considerations

DWDM systems primarily operate in the C-band (1530-1565 nm) where attenuation is lowest and EDFAs work. Conventional fibers have higher dispersion at these wavelengths, so dispersion-compensating fibers or ITU G.655 fiber (lower dispersion, supports high-speed long-haul transmission, suppresses four-wave mixing) are preferred for backbone networks; ITU G.652 fiber is common in access networks.

The main fiber families you will meet in a field survey are:

Water-peak legacy fiber (the old "1400 nm peak" grade) is largely irrelevant in C-band designs, but its higher loss at 1383 nm matters if you are considering a CWDM overlay.

EDFA Amplification

In a DWDM system, a post-amplifier boosts the signal at the transmitter, a pre-amplifier strengthens it at the receiver, and in-line EDFAs compensate for span loss (typically every 50-100 km). EDFAs amplify the 1530-1570 nm window using pump lasers at 980 nm or 1480 nm, achieving around 30 dB of gain - for all wavelengths simultaneously, which is what makes multi-thousand-kilometer DWDM links economical.

Four amplifier characteristics drive most design decisions:

• Gain flatness: practical EDFAs have a gain ripple across the band. Without gain-flattening filters, channels at the edges of the band are amplified less than those in the middle, and after eight or ten spans the imbalance becomes unworkable. This is why pre-emphasis (deliberately launching edge channels at slightly higher power) is common.
• Noise figure: typically 5-6 dB for a good C-band EDFA. Noise accumulates span by span, so total OSNR degrades roughly as 10*log N plus the amplifier noise contribution. A link budget built only on optical power and ignoring OSNR will pass on day one and fail after an amplifier is added.
• Saturation and AGC: an EDFA has a finite total output power shared across all channels. Adding channels while running in constant-gain mode drops per-channel power; constant-output-power mode keeps the total but changes each channel's level. Understand which mode your amplifiers are in before you compare power measurements.
• Transient response: if a channel fails or is added abruptly, surviving channels can experience a power excursion until the amplifier's control loop settles. Fast transient suppression protects the rest of the traffic.

Beyond EDFA, Raman amplification injects a pump wavelength in the opposite direction to the signal and distributes gain along the fiber itself. Raman is often used to extend span length or to improve OSNR at the start of a long link, sometimes in combination with an EDFA (hybrid amplification).

Channel Spacing: The "Dense" in DWDM

Channel spacing determines how many wavelengths fit in the C-band:

Commercial systems commonly support 40, 80 or 160 channels over one fiber pair with 50/100 GHz grids, and with EDFAs the system can span thousands of kilometers.

The "dense" label contrasts with CWDM's 20 nm spacing. Narrower spacing multiplies capacity but demands DFB or tunable ITU-grid lasers with wavelength stability of a few GHz, tighter demultiplexer filters, and more care about four-wave mixing - the non-linear effect that creates spurious mixing products exactly on neighbouring channels, which is why DWDM avoids operating near the zero-dispersion wavelength.

DWDM vs CWDM

CWDM uses wide 20 nm spacing, cheap optics and is fine for metro distances; DWDM packs channels tightly with stable laser sources, costs more per port, and is the backbone technology for long-haul and submarine links.

The practical selection question is rarely about raw channel count. CWDM's wide spacing means uncooled lasers and passive, temperature-insensitive muxes, so it wins on cost for a 10 km campus or metro ring with a handful of services. DWDM's narrow spacing means cooled, wavelength-locked lasers and amplification, so it wins whenever you must overcome span loss with amplifiers or you need more than eighteen-ish channels. Our full CWDM vs DWDM comparison works through the numbers.

Linear and Non-Linear Impairments

Every DWDM link fights four families of degradation:

• Attenuation - solved by amplification, but amplifier spacing is limited by OSNR, not by power.
• Chromatic dispersion - different wavelengths travel at different speeds. At 10G it can be tolerated over hundreds of kilometres; at 100G and beyond, every 80 km span needs compensation (DCF modules, dispersion-shifted fiber, or electronic compensation in coherent DSP).
• Polarisation mode dispersion (PMD) - caused by fiber asymmetry and mechanical stress. Older fiber can have PMD values that cap the usable bit rate regardless of how much power you apply. Measuring PMD before an upgrade is standard practice.
• Non-linear effects - four-wave mixing, self-phase modulation, cross-phase modulation and stimulated Brillouin scattering. They scale with launch power, which is why "turn the power up" is almost never a valid DWDM fix: it buys a little OSNR and pays for it with distortion.

Coherent detection with powerful DSP has moved much of this problem into the digital domain, which is why coherent pluggables such as 400G ZR and ZR+ have changed DWDM design habits. See 800G ZR/ZR+ coherent pluggables and pre-FEC BER and OSNR checks.

Network Topologies: Point-to-Point, Ring and Mesh

Three topologies cover almost every deployment:

A fixed OADM is cheap and uses no power, but the wavelengths it drops are welded in place; adding a new service means truck-rolling equipment. A ROADM with wavelength-selective switches lets you re-plan connectivity from a network management system, which is the direction every large operator has taken because it turns a field job into a software job.

Protection and Restoration

DWDM links are usually protected optically, below the client protocol:

• 1+1 optical protection - the same signal is bridged onto two fibers; the receiver selects the better one. Switching is typically under 50 ms and is transparent to Ethernet, OTN and SDH alike.
• OLP (optical line protection) - a dedicated protection box with an optical switch and power monitoring on a working and protection path. Common on leased-fiber routes where the operator does not own both strands.
• OMS protection / optical ring protection - wavelengths travel in one direction around the ring; on a fiber cut, the node loops traffic back on the opposite side.
• Client-layer protection - MPLS-TP, OTN SNCP or ERPS still exist above the optical layer, but relying on them means the optical layer has no independent means of fast recovery.

Whichever scheme you use, verify that the working and protection paths take genuinely diverse routes. Two fibers in the same duct are one cut away from a simultaneous failure, and "protected" links that share a conduit are the most common false comfort in transport engineering.

Commissioning and Day-to-Day Care

A DWDM system is not finished when the fiber is spliced. The commissioning sequence that consistently works is:

1. Measure and record the fiber plant: end-to-end loss, per-span loss, OTDR traces for both directions, and return loss at every connector.
2. Verify amplifier input/output power and confirm the gain setting matches the span design.
3. Set per-channel launch power from the design sheet, then confirm the received power and OSNR on every channel - not just the first and last.
4. Confirm the supervisory channel and management connectivity from every node.
5. Baseline everything: a spreadsheet of power and OSNR per channel at commissioning is the single most valuable troubleshooting document you will own.

Once live, the routine that prevents most outages is a periodic power and OSNR audit. A slow drift of one channel's power signals a degrading laser; a uniform drop across all channels points at a connector or splice; a rise in OSNR of one channel suggests a neighbour was added without re-balancing. Our DWDM commissioning checklist and optical power budget calculation guide cover both the build and the steady state.

Common DWDM Troubleshooting Patterns

Key Takeaways

• DWDM adds wavelength lanes to fiber you already own; the business case is civil-works avoidance, not optics cost.
• The ITU-T grid, the C-band and EDFA amplification are the three pillars that make dense spacing work.
• Channel spacing (50 GHz vs 100 GHz) is the main lever on capacity, and it directly affects laser cost and non-linear tolerance.
• Optical-layer protection at 50 ms is faster than most client-layer schemes and should be designed in from the start.
• Commissioning baselines - power and OSNR per channel - are what make later faults diagnosable in minutes instead of days.

Related articles: DWDM system components: MUX, EDFA and DEMUX explained, FC vs iSCSI SAN comparison, Data center ToR switch selection, and NetApp ONTAP SnapMirror guide.

原文链接：https://rfwireless-world.com/tutorials/dwdm-tutorial-basics