Networks
DWDM vs CWDM: which fits your link?
· 7 min read
CWDM and DWDM both carry multiple channels over one fibre pair by giving each channel its own wavelength. CWDM spaces channels 20nm apart, supports up to 18 channels, cannot be optically amplified and suits shorter, lower-count links at lower cost. DWDM spaces channels 0.8nm or closer, supports 40, 80 or more channels, works with optical amplification and suits long-haul and high-capacity routes. If your link is under about 80km and needs fewer than 8 channels, CWDM is usually the economic answer.
Leasing more dark fibre is expensive and often simply unavailable on the route you want. Wavelength division multiplexing is the standard answer: put several independent channels onto one fibre pair by giving each its own colour of light. The question is which flavour.
Channel spacing is the underlying difference
CWDM (coarse wavelength division multiplexing) spaces channels 20nm apart across the 1270nm to 1610nm range, giving up to 18 channels. The wide spacing is what makes it cheap: the lasers do not need temperature stabilisation, because a drift of a nanometre or two does not push a channel into its neighbour.
DWDM (dense wavelength division multiplexing) packs channels into the narrow C band around 1550nm at 0.8nm spacing or tighter, supporting 40, 80 or 96 channels and more on modern systems. That density demands temperature-controlled, wavelength-locked lasers, which is the main reason the optics cost more.
Amplification is the practical dividing line
This is the difference that decides most real designs. Erbium-doped fibre amplifiers (EDFAs) work in the C band, around 1530nm to 1565nm. DWDM channels sit inside that window, so an entire DWDM system can be amplified with a single device.
CWDM channels are spread from 1270nm to 1610nm, mostly outside the EDFA window. There is no practical way to amplify a CWDM system as a whole. When a CWDM link runs out of optical budget, the options are regeneration at cost and complexity, or a redesign onto DWDM. This is why CWDM tends to stop being viable somewhere around 80km, while DWDM reaches hundreds or thousands of kilometres with amplification.
Choosing between them
CWDM is usually right when:
- The link is under roughly 80km, typically metro or campus
- You need 8 channels or fewer today and modest growth later
- Capital cost matters more than headroom
- There is no realistic path to needing amplification
DWDM is usually right when:
- The route is long-haul, or metro with amplification needed
- You need more than 8 channels, or expect to within the equipment's life
- You are carrying 100G or 400G channels, where coherent optics are DWDM anyway
- Fibre is scarce or expensive enough that maximising each pair pays for the optics
- You want the option of adding capacity later without touching the passive plant
What to measure before you design
Whichever you pick, the design stands on the optical budget, and the optical budget must be built from measurements rather than the provider's route estimate.
- OTDR the fibre to get real length, real loss and the location of every splice and connector
- Measure end-to-end insertion loss rather than calculating it from an assumed loss per kilometre
- Account for the mux and demux insertion loss, which is significant and often forgotten
- Include connector and splice losses at both ends, plus a margin for future repairs
- Leave headroom for ageing and for the splices that a future cable cut will add
That last point catches people out. A route that is repaired after a fibre cut comes back with additional splices and slightly more loss than it had originally. A design with no margin works perfectly until the first time a digger finds the duct.
Where the cost actually sits
The passive multiplexers are a modest part of the bill. Most of the cost is in the optics, one per channel per end, and the difference between CWDM and DWDM optics is substantial. That is why channel count drives the economics more than distance does at the shorter end: a two-channel link is cheap either way, and a forty-channel link is only sensible on DWDM.
It is worth designing the passive plant for where you expect to be in five years even if you only populate a few channels now. Adding optics to an existing mux is straightforward. Replacing the mux because you outgrew it means going back to the fibre.