Wavelength division multiplexing is the technology you reach for when the fiber plant, not the electronics, has become the limit. Instead of lighting one channel per fiber pair, WDM assigns each channel its own wavelength — its own colour — and combines them onto the same pair. The two disciplines, CWDM and DWDM, are the same idea at two densities, and choosing between them is the first decision in any WDM design. GRT manufactures both, as the CWDM/DWDM systems page outlines; this guide is the engineering version of that choice.
The one difference that explains everything: spacing
CWDM (coarse) spaces channels about 20 nm apart across the spectrum. DWDM (dense) spaces them a fraction of a nanometre apart, on a standardised grid. Wide spacing means the laser at each channel can drift slightly with temperature without colliding with its neighbour — so CWDM lasers need no cooling, and uncooled lasers are dramatically cheaper and simpler. Tight spacing means lasers must be temperature-stabilised and precisely tuned — expensive optics, but the same fiber window fits an order of magnitude more channels. Every practical difference between the two systems is this paragraph wearing different clothes.
Channel count and distance
The wide grid caps CWDM channel counts in the teens — enough to multiply a campus pair handsomely, not enough for a carrier backbone. DWDM counts run to many tens of channels per pair. Distance follows a similar shape: CWDM designs are metro-scale, comfortable across campuses and between nearby sites; DWDM is the long-haul discipline, designed for amplified spans between cities. A useful rule: if your longest span and your channel count both fit comfortably inside CWDM territory, DWDM’s extra capability is capability you are paying to not use.
Economics, honestly stated
CWDM wins on per-channel optics cost and operational simplicity: uncooled pluggable channel optics, passive mux shelves, no amplifiers in typical designs. DWDM wins on capacity per fiber — when the fiber itself is the scarce asset (a leased pair, a trans-city span, a duct you cannot re-enter), cost per channel at high counts favours DWDM despite pricier optics. The crossover is a property of your plant, not of the technologies. This is why WDM is quoted as a designed system: mux, channel optics and patch cords from one hand, against a stated channel plan — the model we use on the CWDM and DWDM pages.
Transparency and growth
Two properties both systems share, and which make WDM strategically attractive. Transparency: each wavelength is an independent channel — Ethernet on one, storage replication on another, a legacy protocol on a third, all ignorant of each other. And staged growth: you install the passive mux once, then light channels as demand arrives, turning a construction project into a purchasing decision. Design the mux for the channel count you will plausibly need at year five; light what you need at year one.
Decision summary
- Choose CWDM when: spans are campus/metro, channels needed are modest, budget per channel matters, simplicity is prized.
- Choose DWDM when: channel counts are high or growing without bound, spans are long, or the fiber pair itself is precious and irreplaceable.
- Choose neither yet when: you still have dark pairs — lighting another pair with plain optics is cheaper than any WDM.
Send us the sites, the distance and the channel count you need today and at year five. The recommendation — including "you do not need WDM" where that is true — comes back with the design.
Can CWDM and DWDM share one fiber?
In engineered designs, a CWDM band can coexist with DWDM channels inside part of the spectrum. It is a design conversation, not an accident to improvise in the field.
What rate can each channel carry?
The mux is rate-transparent; the channel optics decide. 10G per wavelength is a standard design point in both disciplines.
A worked example, in the abstract
Consider a campus with two buildings and one leased fiber pair between them. Today the pair carries a single 10G link, saturated at peak. The CWDM answer: a mux pair at each end and four channel optics — production, backup, storage replication and a spare colour — at uncooled-optics prices, installed in a weekend, with the pair now carrying four independent services. The DWDM answer would also work; it would simply bring carrier-grade channel capacity the campus will not use this decade, at carrier-grade optics prices. Now move the same problem to an operator interconnecting two cities over a single owned pair, with forty services queued: the arithmetic inverts, CWDM runs out of spectrum, and DWDM’s density becomes the cheap option per channel. Same technologies, opposite verdicts — decided entirely by channel count and distance, which is why any WDM recommendation that does not start by asking those two numbers is a sales pitch, not engineering.
Can I start with CWDM and migrate to DWDM on the same fiber?
Yes, with planning: hybrid designs reserve part of the spectrum for DWDM channels inside a CWDM plan. Raise it at design time — retrofitting the reservation is the expensive version.
The passive plant: what actually gets installed
A WDM installation is mostly glass and patience. At each end: a mux/demux unit — a passive optical assembly in a shelf or enclosure — into which the existing fiber pair lands on the line side and the channel optics land on the equipment side, joined by short patch cords that must be labelled with their wavelengths as if the labels were law (in operations, they are). The channel optics themselves are ordinary pluggables wearing specific wavelengths: a 10G CWDM SFP+ is mechanically a module like any other, distinguished by its colour in the plan. Nothing in the plant consumes power except the end equipment you already owned. That passivity is WDM’s operational elegance — the middle of the link cannot crash, hang or need a firmware update — and the source of its one discipline: because nothing in the middle reports anything, the documentation is the monitoring. Keep the channel plan, the cord labels and the loss measurements as one record, and the plant will outlive everyone who installed it.
Carry away the two numbers: channel count now (with honest growth), and the distance between the endpoints. Everything else in WDM — cooling, grids, amplification, price — is commentary on those two numbers. Bring them to the first conversation and the technology will, quite literally, select itself.
One operational coda: whichever discipline you deploy, buy the channel optics and the muxes as one engineered set from one accountable source, and keep the channel plan with the fiber records. Split sourcing — muxes from one catalogue, optics from another, records from memory — is how WDM plants acquire their mysteries.