CWDM vs DWDM

Both schemes carry independent channels on separate wavelengths through one fiber. The grid spacing determines everything downstream: laser cost, channel count, amplification, and reach.

CWDM (ITU-T G.694.2) defines 18 channels on a 20 nm grid from 1271 to 1611 nm. Twenty nanometers is generous by design: it swallows the ~0.1 nm/°C drift of an uncooled DFB across its whole temperature range, so CWDM transmitters need no TEC and no wavelength locker, just a binned laser. The result is the cheapest form of fiber-pair multiplication, standard in metro access, mobile fronthaul, and campus links. The constraints follow from the same breadth: channels span the water-peak and high-loss regions (older G.652 fiber loses several of the 18), and, decisively, most of the grid lies outside the erbium window, so CWDM links are unamplified, budget-limited to tens of kilometers, typically ≤8 or ≤16 channels in practice.

DWDM (G.694.1) packs channels at 100 GHz (~0.8 nm), 50 GHz, or on modern flexible grids down to 12.5 GHz granularity, almost always inside the C-band (~1530–1565 nm; L-band for expansion). Forty to ninety-six channels ride one fiber through shared EDFAs: one amplifier per span for every channel simultaneously, which is the economic engine of long-haul transport. The price of density: temperature-stabilized, wavelength-locked transmitters (or tunables), tighter filters (AWGs, WSS), and management of the nonlinear interactions that close channel spacing invites (XPM, FWM).

Selection heuristic: no amplification needed, short reach, cost-driven → CWDM. Amplified reach, high aggregate capacity, coherent formats, ROADM-based networks → DWDM. Hybrid deployments layer a few DWDM channels inside one CWDM window (the 1531/1551 nm slots) when an access network outgrows its grid.