Same idea, different spacing
Both technologies do the same thing: each service gets its own wavelength, a passive multiplexer combines the wavelengths onto one fiber, and a demultiplexer separates them again at the far end. Fiber itself is cheap; laying it is not. Putting eight or forty services on a fiber pair instead of one is what makes WDM attractive on metro routes.
The difference is the spacing. CWDM places its channels 20 nm apart, across 1270 to 1610 nm. That wide spacing lets lasers drift with temperature and keeps the filters simple, so both the optics and the multiplexers cost less. DWDM packs channels 0.8 nm or 0.4 nm apart in the C-band around 1550 nm. That needs lasers held precisely on their channel and much more complex filter coatings, but it fits many more channels on the fiber and sits exactly in the band where erbium-doped fiber amplifiers work.
When CWDM is enough
CWDM fits routes with modest, predictable capacity: access rings, enterprise links, mobile backhaul and links between a few sites. Its limits are the channel count and the lack of amplification. Plan for those and CWDM goes further than many expect.
For long spans, use the upper channels. Of the 18 CWDM wavelengths, the eight from 1470 to 1610 nm are the ones generally used beyond 40 km, because fiber loss is lower there. Each channel has a passband of about ±6.5 nm, and the optic must stay inside it.
A worked example. The WincoTek 10G SFP+ CWDM 80 km module launches at least −1 dBm and has an APD receiver rated −24 dBm: 23 dB of budget. Keep 3 dB of margin and 20 dB remains for the multiplexer, demultiplexer and fiber. That is enough to run eight 10G channels over about 60 km between two sites with one 8-channel CWDM multiplexer at each end and no amplifier, provided the measured loss of the worst channel stays inside 20 dB. Intermediate sites can drop single channels with an add/drop filter, as long as its loss is in the sum. Check the receiver ceiling as well: on short spans the APD can be overloaded, so attenuate where the received level is too high.
Three multiplexer features are worth asking for. An express (pass-through) port lets you add a second multiplexer with more channels later, without replacing the first. A monitor port taps a small share of the line power so you can spot a fault quickly. A 1310 nm port carries ordinary grey 1310 nm optics alongside the colored channels.
The same passive idea carries 5G fronthaul: colored 25G optics at the radio and the DU share one fiber through passive multiplexers, with six-wavelength and twelve-wavelength plans in use. Whatever the plan, optics and multiplexers must use the same channel grid.
When you need DWDM
DWDM becomes the right choice when one of these is true: more channels than CWDM offers, wavelengths faster than 10G, spans that need amplifiers, or a route that must keep growing for years.
A DWDM optic is ordered by channel, not only by reach: each module must match its multiplexer port. Tunable optics change that logistics. The WincoTek 10G SFP+ tunable DWDM module covers 96 channels on the 50 GHz grid, so one part number can spare a whole channel plan.
At 100G there are two routes. PAM4 DWDM optics such as the WincoTek 100G QSFP28 PAM4 DWDM split 100G across two channels and reach 80 km with a DWDM line system; they tolerate only about ±100 ps/nm of residual dispersion and need an OSNR of 31 dB, so the line needs dispersion compensation and amplification. A pluggable EDFA, such as the WincoTek QSFP28 EDFA with 22 dB gain across 1528 to 1561 nm, can provide that gain from a switch port. Coherent optics take the other route: the 100G CFP2-DCO handles 40,000 ps/nm of chromatic dispersion in its DSP, and 400G QSFP-DD coherent modules put 400G on one wavelength straight from a router port.
For the highest capacity, DWDM transport platforms carry up to 800G per wavelength and 25.6 Tb/s per fiber pair, with 96 channels at 50 GHz, plus protection switching and network management.
Budget and dispersion: the physics behind the choice
Two effects limit a link. Attenuation is lower around 1550 nm than at 1310 nm, which favors the long-wavelength channels for distance. Chromatic dispersion is the opposite: it is much higher at 1550 nm, roughly twenty times the value near 1310 nm on standard fiber, and it spreads pulses as the span grows. At 10G with good optics that is manageable to 80 km. At 100G PAM4 it is the reason for dispersion compensation, and coherent detection is the reason it stops mattering.
A practical way to decide
- Count channels today and in five years. Up to eight long-reach or 18 short-reach channels: CWDM is a candidate. More: DWDM.
- Check the rate per channel. 1G and 10G suit both; 100G and above point to DWDM, PAM4 or coherent.
- Add up the span loss: fiber, multiplexers, add/drop filters, connectors, splices and 3 dB of margin. If it fits the optic budget, you can avoid amplifiers; if not, DWDM with EDFAs.
- Consider operations: monitoring, protection and remote management favor a managed DWDM platform on core routes.
- Start simple where growth is uncertain: a CWDM multiplexer with an express port keeps options open.
Send us the route length, span losses, fiber type and the services to carry, and we will propose a CWDM or DWDM design with the matching optics.
Optical network solutions