The Problem With Last-Mile Fiber
Running fiber to every home in a neighborhood is expensive, and not because of the fiber itself. Fiber is cheap. The cost is in the civil work — trenching, conduit, restoration. Every fiber strand you can eliminate from a trench run directly reduces the capital cost of a deployment.
The standard GPON architecture runs one fiber per home from an OLT (Optical Line Terminal) at the headend to an ONT (Optical Network Terminal) at the premises, using passive optical splitters to serve 32 or 64 homes per PON port. This works well for greenfield deployments with fresh conduit. It works poorly when you're threading new fiber through shared ducts with limited fill capacity, or trying to serve a neighborhood from a small equipment cabinet rather than a full central office.
CWDM (Coarse Wavelength Division Multiplexing) offers a different tradeoff.
How CWDM Works
CWDM uses a single fiber strand to carry multiple independent optical channels, each on a different wavelength. The ITU standard defines 18 channels from 1270nm to 1610nm at 20nm spacing. Each channel is independent — a separate logical fiber. A CWDM mux/demux is a passive device, no power required, that splits or combines these wavelengths.
A typical 8-channel CWDM deployment gives you 8 logical fibers over 1 physical fiber pair. At 10Gbps per channel, that's 80Gbps aggregate on a single strand.
For neighborhood deployment, this means:
- One conduit run from headend to neighborhood distribution point
- One fiber pair carries 8 independent subscriber circuits (or 8 GPON PON legs)
- Each PON leg serves up to 32 homes via passive splitters
- 256 homes served by a single fiber pair to the distribution point
Compare this to running 8 individual fiber pairs — 8× the conduit fill, 8× the splice count, 8× the failure surface.
Physical Design
Headend OLT
│
├─ PON port 1 ──→ λ1271 ─┐
├─ PON port 2 ──→ λ1291 ─┤
├─ PON port 3 ──→ λ1311 ─┤ CWDM Mux ──────────────── single fiber ──────────────── CWDM Demux
├─ PON port 4 ──→ λ1331 ─┤ (headend) (neighborhood)
├─ PON port 5 ──→ λ1351 ─┤ │ │ │ │
├─ PON port 6 ──→ λ1371 ─┤ Splitters
├─ PON port 7 ──→ λ1391 ─┤ (1:32 passive)
└─ PON port 8 ──→ λ1411 ─┘ │
Homes (ONTs)The neighborhood distribution point contains only passive optics: the CWDM demux, passive splitters, and fiber splice trays. No power. No HVAC. No maintenance contracts. It fits in a standard fiber distribution cabinet.
Budget Analysis (128-home block)
| Approach | Fiber runs | Civil cost (est.) | Active electronics | Total CapEx est. |
|----------|-----------|------------------|--------------------|-----------------|
| 4× dedicated fiber pairs | 4 pairs × 500ft | $18,000 | 4× OLT line cards | High |
| 1× fiber pair + CWDM | 1 pair × 500ft | $4,500 | 4× OLT line cards | Low |
| Delta | | −$13,500 | Same | ~40% lower |
The CWDM mux/demux hardware costs roughly $800–1,200 per 8-channel passive unit. The civil savings on a neighborhood-scale deployment typically exceed the hardware cost by 10× or more.
Failure Modes and Mitigation
Single-fiber failure affects all 8 channels. Mitigate with a protection fiber pair (CWDM mux on protection path, automatic failover at OLT). This is the standard carrier practice for CWDM rings.
Wavelength contention. CWDM demux isolation is typically 30–40dB between adjacent channels. SFP+ transceivers must be wavelength-specific (e.g., CWDM-SFP10G-1271). Using standard 1310nm SFPs on a CWDM mux will leak signal across channels.
Insertion loss budget. Each passive CWDM element adds ~1–2dB insertion loss per channel. A 32-way passive splitter adds ~15dB. Total span loss on a GPON link must stay under ~28dB (class B+) or ~32dB (class C+). Design the link budget before ordering hardware.
When This Is Worth Doing
CWDM muxing makes economic sense when:
1. You have limited conduit capacity (can't pull more fiber)
2. Civil restoration costs dominate your CapEx
3. You need to serve a neighborhood from a remote cabinet without utility power at the distribution point
4. You're upgrading an existing copper plant where conduit is already occupied
It does *not* make sense when you're laying fresh conduit with spare capacity, or when the distribution point already has power and climate control (in which case an active CWDM transponder rack with protection switching is more flexible).
Conclusion
CWDM passive muxing is an underused tool for last-mile fiber economics. The hardware is commodity, the design is straightforward, and the civil cost reduction on a per-home basis is significant. For an ISP deploying fiber in established neighborhoods where conduit space is constrained, routing 8 PON legs over a single fiber pair is the right call.
The plan outlined here was developed during ISP fiber deployment work in the early 2000s and remains applicable to current GPON and XGS-PON deployments.