CWDM vs DWDM: which fits your metro network
Guide · Metro optical transport · Last updated: 2026-11-19
CWDM suits metro links that need a handful of wavelengths over short-to-medium distances at the lowest entry cost, while DWDM suits high-capacity metro routes where fiber pairs are scarce or amplification is required. Most buyers match CWDM to access and aggregation rings and DWDM to core metro routes; YT CIG supplies both technology families as a multi-vendor supplier.
CWDM and DWDM: the 20 nm gap that changes the equipment
CWDM places a small number of channels widely apart on the fiber so that cheap, uncooled optics can be used, while DWDM packs many channels tightly into the 1550 nm window so that far more traffic fits on the same fiber pair at the price of precision optics and amplification. Both technologies multiplex independent wavelengths onto one fiber pair; everything that follows (channel count, capacity, reach, cost, operations) flows from that single difference in spacing.
The coarse spacing is what makes CWDM inexpensive. With 20 nm between channels, a laser’s center wavelength may drift with temperature without spilling into a neighbor’s passband, so CWDM transceivers use uncooled DFB lasers with no thermoelectric cooler, and the multiplexers are passive thin-film-filter devices that need no power. DWDM works the other way: sub-nanometer spacing demands lasers locked to an absolute frequency grid, temperature-stabilized and wavelength-accurate, plus narrow-passband filters, optical amplification and usually a managed line system.
In a metro context the two technologies sit at different layers. CWDM typically terminates access and aggregation rings, cell-site backhaul and campus links, where a few services ride passive multiplexers straight off switch or router ports. DWDM belongs in the metro core and regional layer, where fiber pairs are scarce, channel counts are high and spans may need amplification.
How do ITU-T G.694.2 and G.694.1 define the wavelength grids?
ITU-T G.694.2 defines the CWDM grid as 18 nominal center wavelengths from 1271 nm to 1611 nm in fixed 20 nm steps, while ITU-T G.694.1 defines the DWDM grid as a frequency lattice anchored at 193.1 THz (1552.52 nm) with fixed spacings of 12.5, 25, 50 and 100 GHz (ITU-T G.694.2; ITU-T G.694.1). The CWDM grid is wavelength-based; the DWDM grid is frequency-based.
Two practical consequences follow from those definitions. First, the DWDM grid has included a flexible-grid option for high-bit-rate super-channels since the 02/2012 edition, n × 6.25 GHz central frequencies with 12.5 GHz slot granularity, the 193.1 THz anchor unchanged (ITU-T G.694.1), so modern DWDM line systems can carve variable-width slots; the fixed 20 nm CWDM grid has no equivalent. Second, channel placement interacts with fiber type. The CWDM channels at 1391, 1411 and 1431 nm sit just above the 1383 nm water peak, with 1351 and 1371 nm below it: on legacy G.652.A/B fiber that peak attenuates the 1391–1431 nm channels heavily, whereas low-water-peak G.652.D fiber keeps attenuation at 1383 nm at or below the 1310 nm figure, effectively opening the full 18-channel grid (ITU-T G.652). The fiber class decides whether the 1391, 1411 and 1431 nm channels can be used at all, so the survey comes before the channel plan.
| Attribute | CWDM grid (ITU-T G.694.2) | DWDM grid (ITU-T G.694.1) |
|---|---|---|
| Definition basis | Nominal center wavelengths | Absolute frequency grid anchored at 193.1 THz (1552.52 nm) |
| Spacing | 20 nm (≈ 2.5 THz at 1550 nm) | 12.5 / 25 / 50 / 100 GHz; 100 GHz ≈ 0.8 nm |
| Channel span | 18 channels, 1271–1611 nm (O, E, S, C and L bands) | Fixed and flexible grids; deployed mainly in the C-band, 1530–1565 nm |
| Practical passive mux size | 4, 8, 16 or 18 ports | 40 ports at 100 GHz or 80 ports at 50 GHz per C-band |
| Laser stabilization | Uncooled DFB lasers | TEC-cooled DFB lasers locked to the grid |
| Grid evolution | Fixed grid, unchanged since 2003 | Flexible grid introduced in the 02/2012 edition (n × 6.25 GHz central frequencies, 12.5 GHz slot granularity, 193.1 THz anchor unchanged); carried forward in edition 3.0 (2020) |
Channel counts, lane rates and capacity per fiber pair
A passive CWDM plant typically carries 8–18 wavelengths at 1G–25G per lane (tens to low hundreds of gigabits per fiber pair) while a DWDM C-band plant carries 40–80 wavelengths at 10G–400G each, from hundreds of gigabits to multiple terabits per pair.
On the CWDM side, the eight channels from 1471 to 1611 nm are the workhorses: Cisco’s CWDM-SFP10G family, for example, is specified to carry up to eight 10 Gigabit Ethernet channels over single-mode fiber on those wavelengths (Cisco CWDM-SFP10G datasheet). On G.652.D fiber the O- and E-band channels (1271–1451 nm) become usable as well, which is what takes a full 18-channel mux into reach. Per-lane rates cover 1G SFP, 10G SFP+ and 25G NRZ, the last being the basis of CWDM4, where four 25G lanes at 1271, 1291, 1311 and 1331 nm form a 100G link over up to 2 km of duplex single-mode fiber (CWDM4 MSA Technical Specification v1.0). YT CIG’s equipment catalog reflects exactly this spread, from QSFP-100G-CWDM4 modules to 10G CWDM SFP+ parts such as the FTLX1471D3BCL-EX.
On the DWDM side, a passive 40-port C-band mux carrying 10G per wavelength yields 400G per fiber pair, a common metro aggregation starting point. An 80-port mux at 50 GHz doubles the count, and coherent per-wavelength rates raise it further: the OIF 400ZR agreement defines a 400G coherent DWDM interface in pluggable QSFP-DD/OSFP form factors for exactly these metro and data-center-interconnect spans (OIF 400ZR Implementation Agreement, 2020), which places 40 × 400G = 16 Tbps within grid arithmetic for an equipped C-band. In practice, YT CIG stocks 10G DWDM XFPs such as the Huawei DWDM-XFP10G-1550-12 alongside transport platforms like the Huawei OSN1800V and the Alcatel 1830PSS-16II that host 100G and 400G line cards.
Which reach limits apply on metro fiber plants?
CWDM is essentially an unamplified technology whose spans are bounded in the tens of kilometers, while DWDM extends further through C-band amplification and coherent interfaces, the 400ZR agreement targets a minimum 80 km reach and permits amplified links up to 120 km (OIF 400ZR Implementation Agreement, 2020).
The CWDM ceiling has three parts. Loss: passive muxes and uncooled lasers leave modest link budgets, and reach is whatever the optic’s datasheet says, Cisco’s 10G CWDM receiver specification, for instance, is referenced to ITU-T G.652 maximum dispersion over 40-km single-mode spans (Cisco CWDM-SFP10G datasheet), and 2 km is the CWDM4 MSA limit for 100G CWDM4 links (CWDM4 MSA Technical Specification v1.0). Amplification: the erbium-doped fiber amplifier operates in the C-band, roughly 1530–1565 nm, so it cannot boost the 1270–1451 nm CWDM channels at all, a passive CWDM plant has no practical mid-span amplifier option across most of its grid. Fiber type: the E-band channels (1391/1411/1431 nm) require low-water-peak G.652.D fiber, as covered above (ITU-T G.652).
DWDM’s reach advantages compound. Its channels sit inside the C-band where EDFAs are standard equipment, so spans extend stage by stage with powered line amplifiers. Single-wavelength 10G ZR optics already reach 80 km on G.652 fiber (Cisco SFP+ transceiver installation notes); coherent 100G/400G interfaces manage chromatic dispersion in the receiver DSP instead of with passive compensation modules, which is what makes 400ZR’s 80–120 km targets realistic on metro routes (OIF 400ZR Implementation Agreement, 2020). A 60–120 km crossing that has no new fiber budget lands on a DWDM build, passive plus amplified; the same route at 40 km does not.
How does the cost profile differ between CWDM and DWDM?
CWDM is the cheapest way to put several services on one fiber pair when channel counts are small and spans are short; DWDM costs more per wavelength but can end up cheaper per transported bit once channel counts, reach or fiber-scarcity costs rise. The CWDM bill of materials is structurally light: uncooled DFB lasers avoid thermoelectric coolers and their control loops, cutting module cost, size and power; the wide 20 nm tolerances keep filters inexpensive; multiplexers are passive thin-film devices; and no line system, amplifier or management layer is required, CWDM SFP+ optics plug directly into ordinary switch ports. The DWDM bill of materials carries the precision burden: TEC-cooled, grid-locked lasers; narrow-passband filters or athermal arrayed-waveguide gratings; EDFAs and their supervision; and, beyond the passive layer, transport platforms with OTN switching and protection. Coherent 400G-class modules cost a multiple of CWDM4-class modules, and an equipped DWDM line system is a powered, monitored asset where a CWDM plant is a passive strand of glass with connectors.
The crossover is per-bit and per-route. For up to eight channels within tens of kilometers, CWDM wins end to end because almost nothing besides the optics and a passive mux is bought. When the alternative is trenching new fiber, leasing dark fiber or standing up one parallel link per service, DWDM’s density repays its higher unit cost, eighty wavelengths on one pair against eighty fiber pairs changes the economics of the whole route. Operating costs follow the same line: passive CWDM rings need almost no powered infrastructure, while DWDM line systems demand power, cooling, monitoring and staff trained on the platform.
The key parameters side by side
Per-product reach is datasheet-governed, and the fiber class (G.652 vs G.652.D) has to be settled before any CWDM channel plan is locked (ITU-T G.652).
| Parameter | CWDM (ITU-T G.694.2) | DWDM (ITU-T G.694.1) |
|---|---|---|
| Grid | 18 channels, 1271–1611 nm, 20 nm steps | Frequency grid anchored at 193.1 THz; 12.5/25/50/100 GHz |
| Channel spacing | 20 nm (≈ 2.5 THz) | 100 GHz (≈ 0.8 nm) typical; 50 GHz for 80-channel builds |
| Typical passive mux | 4/8/16/18 ports | 40 ports at 100 GHz or 80 ports at 50 GHz (C-band) |
| Per-lane rates (typical) | 1G, 10G, 25G (CWDM4 MSA: 4 × 25G = 100G over 2 km) | 10G, 100G, 400G coherent (OIF 400ZR) |
| Capacity per fiber pair (passive, typical) | Up to 8 × 10G = 80G (1471–1611 nm channels; Cisco CWDM-SFP10G datasheet) | 40 × 10G = 400G; 40 × 400G = 16T (grid arithmetic) |
| Reach without amplifiers | 40-km class for 10G CWDM; 2 km for CWDM4-class 100G | 80 km minimum target under 400ZR; ≤ 120 km amplified (OIF 400ZR) |
| Amplification | Not practical below 1530 nm (EDFA covers the C-band only) | EDFA standard in the C-band; spans extend stage by stage |
| Laser | Uncooled DFB | TEC-cooled DFB |
| Fiber requirement | G.652.D needed for E-band channels (1391/1411/1431 nm) | C-band channels work on standard G.652 |
| Cost per wavelength | Lowest | Higher; falls per bit as channel count rises |
| Typical fit | Access and aggregation rings, campus and cell-site links | Core metro and regional routes, fiber-lean builds, high capacity |
CWDM’s capacity ceiling is set by its grid: eight S/C/L-band channels at 10G is the practical passive baseline, and 100G arrives only through CWDM4-class optics at 2 km. Every CWDM row is unamplified by design, while every DWDM row assumes amplification is available in the C-band. The cost rows describe differing cost structures, and the crossover between them depends on channel count and fiber availability.
Which metro scenarios favor CWDM?
Four recurring scenarios fit a route with up to eight active channels, spans inside tens of kilometers, spare fiber for future needs and no amplification plan. First, cell-site and access aggregation: 10G CWDM SFP+ optics on the 1471–1611 nm channels carry mobile backhaul and fixed-access rings straight between switch ports, with Cisco’s own datasheet framing the eight-channel, 40-km-class envelope (Cisco CWDM-SFP10G datasheet). Second, campus and enterprise multi-service links: data, storage and voice services each take a wavelength on a passive mux, which avoids building parallel fiber. Third, short 100G interconnects: CWDM4-class QSFP28 modules deliver 4 × 25G over up to 2 km of duplex single-mode fiber (CWDM4 MSA Technical Specification v1.0), the standard budget option inside data halls and campus rings. Fourth, low-opex builds: a passive CWDM plant has no powered line equipment, which suits unstaffed sites and tight maintenance budgets.
The equipment catalog mirrors these scenarios across vendors: QSFP-100G-CWDM4 modules and CWDM4-class parts such as the ZXS-Q8CWDM4Z-00, plus Finisar 10G CWDM SFP+ optics like the FTLX1471D3BCL-EX, are held in YT CIG’s equipment stock. The sibling guide on 100G QSFP28 transceivers covers the CWDM4-versus-LR4 choice in depth, and the 10G SFP+ guide explains how to read the part-number scheme behind these optics.
Which metro scenarios call for DWDM or a hybrid build?
DWDM enters when channel counts pass what the CWDM grid can carry, spans pass unamplified reach, fiber pairs are scarce enough that new construction dominates the budget, or the design needs a managed transport platform with protection. Fiber exhaustion is the first case: a metro core route carrying 40 wavelengths on one pair removes the need for 40 parallel fibers, at 100G per wavelength that is 4 Tbps on a single strand pair, and 400ZR-class coherent interfaces push per-wavelength rates to 400G for 80–120 km metro and data-center-interconnect spans (OIF 400ZR Implementation Agreement, 2020). Long aggregation legs: spans beyond CWDM’s 40-km-class envelope work with an amplified line system, since the C-band is exactly where EDFAs operate. Managed platforms: OTN-class transport such as the Huawei OSN1800V or the Alcatel 1830PSS-16II hosts DWDM line cards, protection switching and centralized management, the stock pages for both platforms and for 10G DWDM XFPs like the Huawei DWDM-XFP10G-1550-12 show what that layer looks like in practice, and the guide on 100G DWDM transceivers details the C-band channel plans behind them.
Most metro upgrades land on hybrid builds. The standard pattern overlays a DWDM C-band plant (1530–1565 nm) with CWDM’s O- and E-band channels (1271–1451 nm) on the same fiber pair: the two grids do not collide, and the CWDM wavelengths keep serving low-rate access services while the C-band carries the dense core traffic. One collision must be planned around: CWDM’s 1531 and 1551 nm channels sit inside the C-band and cannot be used simultaneously with C-band DWDM on the same fiber (ITU-T G.694.2; ITU-T G.694.1).
Which one fits your metro network?
Work through three checks. Count the services needed on day one and over a three-year horizon, that is the wavelength count, and it separates an eight-channel problem from an eighty-channel one.
Measure the spans and check the fiber class: if any E-band CWDM channel is in the plan, the plant must be G.652.D (ITU-T G.652). If a metro route needs more than eight wavelengths and the alternative is trenching, the trench cost dominates the comparison, whatever the optics cost.
Up to about eight wavelengths, tens of kilometers, fiber available, no amplification: CWDM passive muxes with CWDM and CWDM4 optics are the fit. Fiber pairs scarce, more than eight wavelengths, spans past 40 km, or coherent 100G/400G services: DWDM, whether as a passive 40-port mux or a platform-based line system. Most real metro upgrades are hybrids, with DWDM C-band in the core, CWDM O/E-band channels toward the edge, planned around the 1531/1551 nm collision noted above.
YT CIG stocks both technology families across vendors, CWDM4-class QSFP28 and 10G CWDM SFP+ optics on one side, 10G DWDM XFPs and OTN transport platforms on the other; the transmission category lists the transport layer and the multi-vendor optics case study shows how mixed-vendor optics are sourced in practice.
When requesting pricing, state the wavelength plan, span lengths and fiber type so the optics’ reach ratings can be matched to the actual spans.
What do buyers still ask about CWDM and DWDM?
Where exactly does a CWDM plan collide with a DWDM C-band plan?
At two channels: 1531 and 1551 nm are the CWDM wavelengths that sit inside the C-band, and they cannot run alongside C-band DWDM on the same fiber (ITU-T G.694.2; ITU-T G.694.1). Everything else separates cleanly: the DWDM C-band (1530–1565 nm) and the CWDM O- and E-band channels (1271–1451 nm) occupy disjoint spectrum, so a passive combiner can put both plants on one fiber pair. That is the standard hybrid metro build.
A CWDM route needs 50 km. Can it stay CWDM?
Not on the 10G datasheet envelope. Cisco’s CWDM-SFP10G receiver specification is referenced to ITU-T G.652 maximum dispersion over 40-km single-mode spans (Cisco CWDM-SFP10G datasheet), and EDFAs cover the C-band only, so a passive CWDM plant has no mid-span amplifier option on the 1270–1451 nm channels. A 50 km route therefore becomes a DWDM C-band span. At 100G the question is narrower still: the CWDM4 MSA caps reach at 2 km over duplex single-mode fiber (CWDM4 MSA Technical Specification v1.0).
Is CWDM4 the same as 100GBASE-LR4?
No. Both are 4 × 25G duplex-LC single-mode modules, but CWDM4 follows the MSA: four lanes at 1271, 1291, 1311 and 1331 nm with 2 km reach (CWDM4 MSA Technical Specification v1.0), while 100GBASE-LR4 follows IEEE 802.3bm: LAN-WDM lanes at 1295.56, 1300.05, 1304.58 and 1309.14 nm with 10 km reach (IEEE 802.3bm-2015). They are not interchangeable at the optical layer.
Why does the DWDM grid count in frequency?
Because every spacing is defined as an offset from the 193.1 THz anchor (1552.52 nm): 12.5, 25, 50 or 100 GHz in the fixed grid, with 50 GHz used for 80-channel C-band builds (ITU-T G.694.1). 100 GHz is about 0.8 nm at 1550 nm, and the wavelength equivalent drifts slightly across the band, which is why channel plans are written in frequency. The recommendation has also carried a flexible grid for super-channels since its 02/2012 edition (ITU-T G.694.1).
Do CWDM transceivers need thermoelectric cooling?
No, that is a direct consequence of the 20 nm grid: uncooled DFB lasers can drift across several nanometers over temperature without leaving their passband, which is why CWDM optics run cooler, cost less and omit the TEC entirely (ITU-T G.694.2). DWDM lasers, locked to sub-nanometer spacing, do need thermoelectric cooling to stay on grid.
Can a CWDM metro ring be upgraded to DWDM later?
Yes, by three routes. Overlay a DWDM C-band plant beside the existing CWDM O/E-band channels on the same fiber; replace the passive CWDM mux with a DWDM mux when the CWDM channels reach end of life; or keep CWDM4-class optics for short spans and add a DWDM transport platform in the core, the platform layer YT CIG stocks, such as the Huawei OSN1800V and Alcatel 1830PSS-16II, is designed for that retrofit. Plan wavelengths at design time so the 1531/1551 nm collision never materializes (ITU-T G.694.2; ITU-T G.694.1).
Need a specific part?
Send the part number and quantity — we confirm condition, availability and price before you commit.