// Optical Networking

DWDM Basics
Dense Wavelength Division Multiplexing

A complete beginner to intermediate guide to optical networking and DWDM. Covers the physics, components, design principles, and real-world applications. Interview questions included.

Jump to What is DWDM The physics C band spectrum Components Network design DWDM vs grey wave Interview questions
Foundation

What is DWDM?

DWDM stands for Dense Wavelength Division Multiplexing. The name sounds complex but the concept is simple: instead of sending one stream of data through a fiber optic cable, DWDM sends multiple streams simultaneously by using different colours of light for each one.

Think about a glass prism. White light goes in one side, and a rainbow of colours comes out the other. DWDM works on exactly the same principle. Each colour of light carries its own independent data stream, so one pair of fiber can carry the traffic that would otherwise need dozens of separate cables.

🌈
The prism analogy

White light entering a prism contains many wavelengths mixed together. The prism separates them into individual colours. DWDM does exactly this: a filter combines multiple laser wavelengths onto one fiber, and a filter at the other end splits them back out. Each wavelength is completely independent and invisible to the others.

Why use DWDM?

Fiber relief

When you have too few physical fiber strands but need far more bandwidth, DWDM lets you multiply capacity without digging new cable. One fiber pair can carry 48 or 96 independent wavelengths.

Traffic isolation

Each wavelength is logically separate. You can give one customer their own dedicated wavelength on shared physical fiber with complete isolation from all other traffic.

Long distance efficiency

DWDM with coherent optics can transmit hundreds of kilometers without regeneration. It is far more cost effective than running separate dark fiber for each circuit.

How it looks in practice

Site A multiplexed MUX Ch 1 • 100G Ch 2 • 100G Ch 3 • 100G Ch 4 • 100G Ch 5 • 100G Ch 6 • 100G ...up to 96 channels Site B (DEMUX)
The physics

Light as a carrier of data

You do not need a physics degree to understand DWDM but knowing the basics helps you design better networks and troubleshoot problems faster. These are the four key physical properties you will deal with every day.

Optical power

How bright the laser is, measured in dBm. Too bright and you blind the receiver. Too dim and you lose the signal. The target receive power is engineered to sit precisely in the receiver's sensitivity window.

Attenuation

Signal loss over distance. Standard SMF-28 fiber loses about 0.25 dB per kilometer. Over 80km that is 20 dB of loss. Amplifiers are placed along the span to compensate and keep signal strength within range.

Chromatic dispersion

Different wavelengths travel at slightly different speeds through glass. Over long distances they spread apart, causing the signal to blur. This is compensated in the engineering phase using dispersion compensation fiber or coherent DSP.

Signal to noise ratio (OSNR)

Think of tuning a radio late at night. When you dial in to a station you are improving the ratio of the signal you want versus the background noise. OSNR in optical networks works the same way. Amplifiers add a small amount of noise called ASE (amplified spontaneous emission) with every amplification. Too many amplifiers in series and your noise floor rises until the receiver cannot distinguish the signal from the noise. Good network engineering keeps OSNR above the threshold required by your transponders, typically 16 to 20 dB for 100G and higher for 400G.

Rule of thumb: Every 10 dB of loss halves the signal power. At 20 dB loss you have lost 99% of the original power. Amplifiers restore this. The design goal is to ensure the signal arrives at the receiver within its specified sensitivity window every time, across every wavelength, across temperature changes and fiber aging.
Spectrum

The C band and channel plan

The entire electromagnetic spectrum spans from gamma rays to radio waves. Visible light is a tiny slice of that. DWDM operates in the infrared part of the spectrum, in a region the industry has defined as the C band, centred around 1550 nanometers. This is standardised globally so a channel number means the same thing to every vendor.

C band
UV (shorter wavelength) Visible light Infrared (longer wavelength)
🥧
The pie analogy

The C band is a fixed pie. You cannot make it bigger. All you can do is decide how to slice it. Bigger slices (100 GHz spacing) means fewer channels but each channel can carry more data. Smaller slices (50 GHz) gives you more channels but each one carries less. The industry most commonly uses 48 channels at 100 GHz or 96 channels at 50 GHz.

Channel spacing comparison

Channel spacing Number of channels Max per-channel rate Total capacity Use case
100 GHz 48 400G 19.2 Tbps Long haul, high capacity backbone
50 GHz 96 200G 19.2 Tbps Metro, more wavelengths needed
Flex grid Variable Up to 1.2T Varies Hyperscale, adaptive networks
Key point for interviews: The C band is defined from 1530nm to 1565nm. Channel 33 on the ITU grid is the reference channel at 193.1 THz. Channels are numbered and spaced precisely so that all vendors interoperate on the same grid. This standardisation is what makes multi-vendor DWDM networks possible.
Network components

The building blocks of a DWDM network

Every DWDM network is built from the same set of fundamental components. Understanding what each one does and why it exists is essential for both design and troubleshooting.

💡
Transponder
Grey to DWDM
🔀
MUX filter
Combines wavelengths
📡
Fiber span
SMF-28
🔊
EDFA amplifier
Restores power
🔄
ROADM
Add/drop/pass
🔀
DEMUX filter
Splits wavelengths
📥
Receiver
Client side

Transponder and pluggable optic (SFP, QSFP)

The laser source. Takes a grey wave signal from your router or switch and converts it to a precisely tuned DWDM wavelength. Modern pluggable coherent optics like QSFP-DD can do this in a single module. Tunable to any ITU grid channel, so you can change which wavelength a port uses from software without touching hardware.

MUX/DEMUX filter (FOADM)

The glass prism. A completely passive device with no power, no CPU, no moving parts. Just mirrors and thin-film filters that combine (MUX) multiple wavelengths onto one fiber at the transmit end and split them back apart (DEMUX) at the receive end. Also called a Fixed Optical Add/Drop Multiplexer or a Mux/Demux shelf.

EDFA amplifier (Erbium Doped Fiber Amplifier)

Placed every 60 to 80 kilometers on a long haul span to compensate for fiber attenuation. An EDFA amplifies all wavelengths simultaneously in the optical domain without converting to electrical signals. It uses a short section of erbium doped fiber pumped with a 980nm laser. The key difference from a regenerator is that no signal processing happens, it is purely optical amplification.

ROADM (Reconfigurable Optical Add/Drop Multiplexer)

The intelligent switching layer. A ROADM detects incoming wavelengths and decides whether to drop them locally (add/drop) or pass them through to the next node. Unlike a passive FOADM, a ROADM is software-configurable. You can change which wavelengths drop at which site without touching fiber. It also handles all optical power balancing, gain, and tilt adjustments automatically. Most modern networks are built with ROADMs rather than passive filters.

Network design

How a DWDM network is designed

Designing a DWDM network is an engineering process. All the physics has to be calculated before a single piece of hardware is ordered to guarantee the network will work. Here is how it is done step by step.

Design inputs

01
Network topologyWhat sites exist, how they are connected, ring or mesh or point to point
02
Fiber characteristicsOTDR traces, fiber type, span lengths, loss measurements
03
Traffic requirementsHow much bandwidth, where it needs to go, which sites are hubs vs pass-through
04
Protection requirementsWorking and protect paths, ring protection, restoration time objectives

Design outputs

A
Amplifier placementExactly where EDFAs go, their gain settings, target output power
B
Node type selectionROADM vs passive filter at each site based on traffic pattern
C
OSNR simulationMathematical proof that signal quality meets threshold across all channels end to end
D
Bill of materialsEvery chassis, card, amplifier, filter, and optic required for the build
Important: Once a DWDM network is correctly engineered and installed, adding new channels is plug and play. You plug in a new card and optic on each end and the ROADM automatically handles all the optical power balancing. No re-engineering is needed unless you change the physical topology.
Comparison

DWDM vs grey wave optics

Grey wave optics are the standard 1310nm and 1550nm transceivers most engineers are familiar with. Understanding the difference between grey wave and DWDM coherent optics is a common interview topic.

Grey wave
Operates at broad wavelengths like 1310nm or 1550nm. Not tunable. Cannot be multiplexed with other wavelengths because it occupies the full available spectrum. Works for short point to point runs. Much cheaper than DWDM optics. Think of it as a spotlight.
DWDM coherent
Tunable to a precise ITU grid channel, down to 1/100th of a nanometer. Multiple DWDM wavelengths can share one fiber without interfering. Supports longer distances, higher speeds, and amplification. Significantly more expensive but cost per bit is lower at scale.
Coherent DSP
Modern coherent optics include a digital signal processor that compensates for chromatic dispersion, polarization mode dispersion, and nonlinear effects in software. This is what enables 100G, 400G, and 1.2T over thousands of kilometers without hardware dispersion compensation.
ZR and OpenZR+
Newer pluggable coherent standards that bring DWDM capability into standard QSFP-DD form factors. Allow routers to directly connect to DWDM networks without separate transponder hardware. Supported by open standards so multiple vendors interoperate.
When to use each: Grey wave for short runs under 10km where cost matters and you have plenty of fiber. DWDM coherent when you need long distances, multiple services on one fiber pair, or scalability beyond what a single wavelength can carry. The crossover point is typically when you need more than 2 to 4 services on the same fiber route or distances exceed 40km.
Interview prep

Interview questions

These are the most commonly asked DWDM and optical networking questions in network engineering interviews at ISPs, hyperscalers, and telecoms. Click each question to reveal the answer.

Basic What does DWDM stand for and what problem does it solve?
Dense Wavelength Division Multiplexing. It solves the problem of fiber exhaust: when you have too few physical fiber strands but need to carry more traffic. By putting multiple wavelengths (colours of light) on a single fiber pair, each carrying an independent data stream, DWDM multiplies the capacity of existing fiber infrastructure without digging new cable. A single fiber pair can carry up to 96 wavelengths, each at 100G or 400G, giving over 38 Tbps of total capacity on two strands of glass.
Basic What is the C band and why does DWDM use it?
The C band (Conventional band) spans approximately 1530nm to 1565nm in the infrared spectrum. DWDM uses it for three reasons: first, standard SMF-28 fiber has its lowest attenuation in this region at around 0.2 dB/km. Second, EDFA erbium doped fiber amplifiers operate natively in this wavelength range, enabling efficient optical amplification. Third, the ITU has standardised channel grids within this band, ensuring multi-vendor interoperability. All DWDM vendors use the same channel numbering, so channel 33 at 193.1 THz means the same thing to every manufacturer.
Basic What is the difference between a MUX filter and a ROADM?
A MUX/DEMUX filter (also called FOADM or Mux/Demux shelf) is a completely passive device. It uses thin-film filters and mirrors to combine or split wavelengths. It has no power, no CPU, and no intelligence. Every wavelength that enters drops at that site, there is no pass-through capability. A ROADM (Reconfigurable Optical Add/Drop Multiplexer) is an active, intelligent device. It detects incoming wavelengths and decides whether to drop them locally or pass them through to the next site. It is software-configurable, so you can change routing of wavelengths without touching fiber. ROADMs also handle optical power balancing automatically. Use a passive filter when all traffic must drop at every site. Use a ROADM when some traffic needs to pass through sites without dropping.
Intermediate What is chromatic dispersion and how is it managed?
Chromatic dispersion occurs because different wavelengths of light travel at slightly different speeds through glass fiber. Over long distances, the faster wavelengths arrive ahead of the slower ones, causing the optical pulse to spread out in time. This pulse broadening causes adjacent bits to overlap, degrading the signal quality and increasing the bit error rate. Historically this was corrected with Dispersion Compensating Fiber (DCF) inserted into the span. Modern coherent transponders with integrated DSP (Digital Signal Processing) correct for chromatic dispersion electronically in the receive path, eliminating the need for DCF entirely. This is one of the key advantages of coherent technology: the DSP can compensate for the accumulated dispersion of thousands of kilometers entirely in software.
Intermediate What is OSNR and why does it matter for 400G?
OSNR stands for Optical Signal to Noise Ratio. It is the ratio of signal power to noise power measured in dB across the channel bandwidth. Every EDFA amplifier adds a small amount of noise called Amplified Spontaneous Emission (ASE). On a long haul route with many amplifiers in cascade, this noise accumulates. If OSNR falls below the required threshold the receiver cannot reliably distinguish ones from zeros and the bit error rate rises. For 100G DP-QPSK the required OSNR is typically around 16 dB. For 400G DP-16QAM the required OSNR is around 24 to 26 dB because 16QAM has more tightly packed constellation points that are more sensitive to noise. This is why 400G does not travel as far as 100G on the same amplifier chain without additional engineering, and why some long haul 400G deployments use probabilistic constellation shaping to reduce the OSNR requirement.
Intermediate What is the difference between 100 GHz and 50 GHz channel spacing?
Channel spacing refers to how wide each slice of the C band spectrum is. At 100 GHz spacing you fit 48 channels in the C band, each wide enough to support up to 400G per channel using modern modulation formats. At 50 GHz spacing you fit 96 channels but each is narrower, limiting per-channel rates to 200G or lower with current technology. The total capacity works out similarly because more channels at lower rates roughly equals fewer channels at higher rates. The choice depends on your needs: if you need more than 48 wavelengths on a route, 50 GHz spacing gives you more channels. If you need maximum per-channel bandwidth, 100 GHz gives you room for 400G and future 800G or 1.2T optics. Flex grid systems allow variable channel widths, allocating wider spectrum slices to high-rate channels and narrower slices where lower rates suffice.
Advanced How does coherent detection differ from direct detection and what advantages does it provide?
Direct detection simply measures the intensity of the incoming light. It only recovers the amplitude of the signal, discarding phase and polarization information. Coherent detection mixes the incoming signal with a local oscillator laser and detects the resulting interference. This recovers both amplitude and phase information across both polarizations. The advantages are significant: first, coherent receivers can use advanced modulation formats like QPSK, 8QAM, 16QAM, and 64QAM that encode multiple bits per symbol using phase and amplitude together, dramatically increasing spectral efficiency. Second, all impairments including chromatic dispersion, polarization mode dispersion, and nonlinear effects can be compensated digitally in the DSP rather than optically. Third, coherent receivers have much better sensitivity, enabling longer reach before regeneration. Fourth, coherent systems can use both polarizations of light independently, effectively doubling capacity on the same wavelength. This is called Dual Polarization or DP, which is why you see terms like DP-QPSK or DP-16QAM.
Advanced A BGP session is dropping intermittently on a DWDM wavelength. How do you troubleshoot?
Start at the optical layer and work up. First check the optical monitoring on the DWDM system: look at receive power on the affected wavelength and compare to the expected power from the design. A receive power that is marginal, near the receiver sensitivity threshold, or fluctuating will cause intermittent bit errors that manifest as BGP hold timer expiry at the IP layer. Next check OSNR on that channel. Look at the pre-FEC BER (bit error rate before Forward Error Correction) on the transponder. If pre-FEC BER is high but post-FEC is clean, you are marginal and a small degradation will push you over the edge. Look at optical power history and check for gradual degradation, which indicates a dirty connector, a failing EDFA, or fiber degradation such as a micro-bend. On the fiber plant, run an OTDR trace on the affected span. Check connector cleanliness at all patch panels in the path. If the issue is temperature-dependent, a loose connector is a common cause. At the ROADM look at gain tilt, which means some wavelengths are louder than others, compressing the amplifiers and degrading OSNR on the affected channel. Once optical health is confirmed, check the IP layer for MTU mismatches, which cause large BGP UPDATE packets to be fragmented or dropped silently.
Try the network tools

Paste a real network log into the anomaly detector or the log redactor and see them in action.

Anomaly Detector Log Redactor