Part 1 of 3 - Foundations and History

One fiber, one set of passive components, eighteen "colors." CWDM solves an expensive engineering problem in the simplest possible way - it scales fiber capacity by an order of magnitude without needing a refrigerator for every wavelength. This part covers what CWDM is, why channel spacing matters, and how the technology came to exist.

1. Introduction: Why We Need "Coarse" WDM

In the cost structure of optical communications, one rule has held for decades: cost per channel is inversely proportional to channel spacing.

Wavelength Division Multiplexing (WDM) multiplexes optical signals at different wavelengths onto a single fiber - in essence, trading color for capacity. But the closer adjacent channels sit, the more precisely each laser wavelength must be manufactured and controlled. That demands thermoelectric coolers (TECs), wavelength locking, and tighter filters. On long-haul routes, all of this is worth it: long-haul fiber is expensive and capacity is scarce.

The problem lies in metro and access networks: short distances, comparatively abundant fiber, a messy mix of protocols, and a very small budget. Applying a long-haul DWDM (Dense WDM) architecture to the metro market is like delivering takeout in a heavy-duty truck - technically feasible, economically absurd.

CWDM (Coarse Wavelength Division Multiplexing) was created for exactly this scenario. Its core idea: relax the channel spacing from 0.8 nm to 20 nm, trading away wavelength precision in exchange for eliminating the cooling system, thereby cutting cost per channel to a fraction of DWDM.

In one sentence, the division of labor is:

2. The History and Origins of CWDM

2.1 The Original Driver of WDM: Fiber Exhaust

In the early-to-mid 1990s, WDM first appeared in the long-haul market in the form of DWDM. Its primary driver was not "bandwidth demand" but fiber exhaust - the cable already buried between cities had run out of strands, and the cost of trenching and laying new cable was unacceptable. If you cannot dig, you have to work with the fibers you already have.

This point is essential: WDM was never a technology demonstration - it was a cost-avoidance technology. Only with that in mind does CWDM's later positioning make sense.

2.2 Before the Word "CWDM" Existed: The 1980s 850 nm Multimode Prototype

The term "CWDM" itself did not enter mainstream industry vocabulary until around 1996. "Coarse" was coined as a contrast to "dense" in describing WDM - and "dense" had to exist first.

Yet the technical practice goes back much further. As early as the early 1980s, multi-wavelength transmission already existed in 850 nm window LANs over multimode fiber with 25 nm spacing. Typical applications included multi-channel video distribution and bidirectional, latency-sensitive telemetry and control carried over a single optical fiber. A four-channel bidirectional WDM data link was documented as early as 1985.

Through the mid-to-late 1990s this market remained dominated by 850 nm multimode LAN applications, helped by two new component classes that reduced cost and increased packaging density: Vertical Cavity Surface Emitting Lasers (VCSELs) and Thin-Film Filters (TFFs).

2.3 Late 1990s: A Very Different Set of Metro Requirements

In the late 1990s, WDM products began moving from long-haul into metro and regional networks. But the metro requirement profile is completely different:

Short distances;

More fiber available;

Beyond SONET/SDH, a large set of heterogeneous protocols - Gigabit Ethernet, Fibre Channel - must be supported;

Individual bandwidth demands are often smaller;

Willingness and ability to pay are far lower.

The industry explored two directions: wavelength banding / hierarchical WDM, and more coarsely spaced wavelengths.

An intermediate product emerged, called Metro DWDM: shorter distances allowed expensive dispersion compensators to be removed, but expensive EDFAs were still required. In the end, long-haul DWDM vendors never provided genuinely economic packaging for the metro space, and Metro DWDM inherited too many undesirable characteristics from products engineered for the inter-exchange market.

CWDM then surfaced as the obvious alternative. But the CWDM products developed for short-distance LAN applications had to be re-engineered to provide a wavelength range suited to metro transmission distances.

2.4 A Frequently Overlooked Branch: IEEE and "WWDM"

In the late 1990s, CWDM became a subject of interest within the IEEE 802.3 High Speed Study Group (10 GbE). The goal was to solve dispersion and loss problems for 10 Gigabit Ethernet over the installed base of multimode fiber in building and campus environments: four wavelengths in the 850 nm or 1310 nm windows, extending the life of the existing cabling infrastructure.

To differentiate the two LAN windows, the study group referred to the 850 nm set as CWDM and the 1310 nm set as WWDM (Wide WDM).

The output of this branch was 10GBASE-LX4: the 1310 nm window, 4 lanes with 24.5 nm spacing, each lane at 3.125 Gbit/s, totaling 10 Gbit/s. Nominal wavelengths are 1275.7 / 1300.2 / 1324.7 / 1349.2 nm, specified to 10 km (some commercial WWDM devices reach 20 km).

Note the detail: 10GBASE-LX4 and the ITU CWDM grid are almost identical - the only difference being 24.5 nm rather than 20 nm spacing. There was even discussion of merging the two standards, which would have reduced the CWDM channel count from 18 to 17. They were never merged, but this history shows that CWDM was, from the very beginning, an "good enough" engineering compromise.

In parallel, in January 2002 the Optical Internetworking Forum (OIF) began specifying a Very Short Reach (VSR) Level 5 interface for intra-office and client interconnects at OC-768 (40 Gbit/s). One of the proposed options was 4 x 10 Gbit/s CWDM over single-mode fiber in the O-band, directly adopting the 10GBASE-LX4 wavelength plan.

2.5 2002–2003: Standardization

Standardization was the dividing line between "vendor-proprietary schemes" and "volume industry." Two Recommendations define today's landscape:

ITU-T G.694.2 - defines the CWDM wavelength grid. First published in June 2002 and revised in December 2003, it specifies 18 channels spaced at 20 nm across the 1271–1611 nm range (frequently simplified to 1270–1610 nm).

ITU-T G.695 - defines CWDM optical interfaces (application codes and optical interface parameters). It was approved in November 2003. Contemporary press coverage was blunt: in a cost-conscious telecom market, CWDM was seen as a cheaper and simpler alternative to DWDM, because the wide channel spacing allows the use of uncooled lasers.

2.6 After 2000: Two Curves Diverge

After standardization, DWDM and CWDM followed completely different evolutionary paths:

DWDM kept tightening spacing: 200 GHz → 100 GHz → 50 GHz → 25 GHz, channel counts pushing past 96, combined with EDFAs and coherent modulation, heading all the way to long-haul and ultra-long-haul.

CWDM held firm at 20 nm: channel count capped at 18, no amplification, no wavelength locking, cost squeezed to the extreme.

This was not stagnation, but a deliberate choice. Once that 20 nm figure narrows, the entire cost model collapses (see Section 3.2). CWDM's value lies precisely in what it refuses to improve.

2.7 2010s–2020s: CWDM's Second Explosion

CWDM's real volume explosion happened more than a decade after its birth, driven by two scenarios that were entirely outside its original designers' field of view:

1. Inside the data center and short-reach DCI. In the 100G era, the CWDM4 MSA packed four CWDM wavelengths (1271 / 1291 / 1311 / 1331 nm, 25G each) into a single QSFP28 module to deliver 100G over 2 km / 10 km. This became one of the primary 100G data center optical interconnect solutions.

2. 5G fronthaul. Once the 25G eCPRI fronthaul interface became mainstream, CWDM - passive, low-cost, and needing no power - became the standard answer for fronthaul fiber capacity expansion (see Section 5.2).

With that, CWDM completed an identity shift: from "a cheap metro substitute" to foundational infrastructure for fronthaul and data centers.