How Fiber to the Home (FTTH) Works

Fiber to the Home (FTTH) is the gold standard of broadband access networks. Unlike DOCSIS (cable) or DSL, which repurpose copper infrastructure originally built for television or telephone service, FTTH runs optical fiber from the service provider's central office all the way to the subscriber's premises. This end-to-end optical path eliminates the electromagnetic interference, distance-dependent attenuation, and bandwidth bottlenecks inherent in copper last-mile technologies, delivering symmetrical multi-gigabit speeds over distances that would be impossible with electrical signaling.

To understand FTTH, you need to understand the physics of optical fiber, the network architectures that make fiber economical for residential deployment, and the protocols that let dozens of subscribers share a single strand of glass.

Fiber Optic Fundamentals

An optical fiber is a thin strand of ultra-pure glass (silicon dioxide) that guides light via total internal reflection. The fiber consists of two concentric layers: a core through which light propagates, and a cladding with a slightly lower refractive index that surrounds the core. When light traveling through the core hits the core-cladding boundary at a shallow enough angle (above the critical angle), it reflects back into the core rather than passing through into the cladding. This traps the light inside the core, allowing it to travel kilometers with remarkably low loss.

Modern telecommunications fiber has an attenuation of roughly 0.2 dB/km at the 1550 nm wavelength — meaning a signal can travel 20 km and lose only about 4 dB, compared to coaxial cable which loses 4 dB in a few hundred meters at GHz frequencies. This extraordinary transparency is why fiber dominates long-haul, metro, and increasingly access networks.

Single-Mode vs Multi-Mode Fiber

Optical fibers come in two fundamental types, distinguished by their core diameter and how many spatial modes of light they support:

• Multi-mode fiber (MMF) — Has a large core, typically 50 or 62.5 micrometers in diameter. Multiple spatial modes (paths) of light propagate simultaneously, each traveling at a slightly different speed. This modal dispersion causes the light pulse to spread out over distance, limiting both bandwidth and reach. MMF is used in short-distance applications like data center interconnects (up to ~550 m at 10 Gbps with OM4 fiber) and campus LANs, where cheap VCSEL transceivers can be used.
• Single-mode fiber (SMF) — Has a tiny core, about 8–10 micrometers in diameter. Only one spatial mode propagates, eliminating modal dispersion entirely. The remaining dispersion (chromatic dispersion, where different wavelengths travel at slightly different speeds) is manageable and can be compensated. SMF is used for all telecom applications beyond a few hundred meters, including all FTTH deployments. A single strand of single-mode fiber can carry signals over 100 km without amplification and supports bandwidths measured in terabits per second.

Every FTTH deployment worldwide uses single-mode fiber. The cost difference between SMF and MMF glass is negligible (the raw fiber costs pennies per meter), and SMF provides virtually unlimited bandwidth headroom for decades of technology upgrades without replacing the physical plant.

Wavelengths and the Fiber Spectrum

Telecommunications fiber operates in several wavelength bands within the near-infrared spectrum, chosen because they correspond to low-attenuation windows in silica glass:

• O-band (1260–1360 nm) — "Original" band. Used for upstream transmission in GPON (1310 nm). Has low dispersion, making it suitable for lower-cost transmitters.
• S-band (1460–1530 nm) — "Short" wavelength band. Used by some next-generation PON systems.
• C-band (1530–1565 nm) — "Conventional" band. Lowest attenuation window. Used extensively in long-haul networks with erbium-doped fiber amplifiers (EDFAs). NG-PON2 allocates C-band for certain downstream channels.
• L-band (1565–1625 nm) — "Long" wavelength band. Provides additional capacity alongside C-band.

FTTH systems use wavelength division multiplexing (WDM) to transmit upstream and downstream traffic on different wavelengths over the same single fiber, which is fundamental to PON economics: one fiber serves both directions simultaneously. We will examine the specific wavelength assignments for each PON standard below.

PON Architecture: How FTTH Networks Are Built

The dominant architecture for FTTH is the Passive Optical Network (PON). A PON uses unpowered (passive) optical splitters in the field to distribute a single fiber from the central office to multiple subscribers. "Passive" is the key word: there are no active electronics between the provider's equipment and the subscriber's premises. This eliminates the need for field-powered equipment, dramatically reducing operational costs, failure points, and maintenance requirements.

A PON has three main components:

• OLT (Optical Line Terminal) — Located at the provider's central office or remote cabinet. The OLT is the headend of the PON. It generates the downstream optical signal (broadcast to all ONTs on the PON), receives upstream signals from ONTs, and manages bandwidth allocation. A single OLT chassis can serve thousands of subscribers across many PON ports.
• Passive splitter — An unpowered optical device that splits one input fiber into N output fibers (commonly 1:32 or 1:64). In the downstream direction, it copies the optical signal to all output ports (each getting 1/N of the power). In the upstream direction, it combines signals from all ONTs onto the single feeder fiber. The splitter has no electronics, no power supply, and no moving parts. It sits in a weatherproof enclosure on a pole, in a pedestal, or in a utility vault.
• ONT/ONU (Optical Network Terminal / Optical Network Unit) — Located at the subscriber's premises. The ONT converts between optical signals on the PON and electrical Ethernet or Wi-Fi in the home. It handles encryption, decryption, and timing synchronization with the OLT. The terms ONT and ONU are sometimes used interchangeably; strictly, an ONU serves multiple subscribers (as in an apartment building), while an ONT serves a single premises.

How PON Downstream Works: Broadcast and Select

In the downstream direction (OLT to subscribers), a PON uses a broadcast model. The OLT transmits a continuous stream of data frames on the downstream wavelength. The passive splitter copies this signal to every connected ONT. Each frame is tagged with an identifier (GEM Port ID in GPON, or LLID in EPON) that indicates which subscriber it belongs to. Every ONT receives every downstream frame, but only extracts and processes the frames addressed to it, discarding the rest. To prevent eavesdropping, downstream data is encrypted — GPON uses AES-128 encryption so that even though all ONTs receive all traffic, only the intended ONT can decrypt its data.

How PON Upstream Works: TDMA

The upstream direction (subscribers to OLT) is more complex. All ONTs share the same upstream wavelength on the same fiber. If two ONTs transmitted simultaneously, their signals would collide at the splitter and become unreadable. PON solves this with Time Division Multiple Access (TDMA).

The OLT acts as a central controller, granting each ONT a precise time slot during which it may transmit. This process works as follows:

1. Ranging — When an ONT first connects, the OLT measures the round-trip delay to that ONT. Since each ONT is at a different distance, their signals would arrive at different times. The OLT assigns each ONT an equalization delay so that, from the OLT's perspective, all ONTs appear to be at the same logical distance. This is critical: an ONT 2 km away must delay its transmission relative to one 18 km away, so both signals arrive at exactly the right time.
2. Bandwidth map (BWmap) — Every downstream frame includes a bandwidth allocation map that tells each ONT exactly when (which timeslots) it may transmit during the next upstream frame period. The OLT dynamically adjusts these allocations based on each ONT's queue depth and service-level agreements.
3. Burst-mode transmission — ONTs transmit in short bursts during their assigned timeslots and remain silent at all other times. The OLT's burst-mode receiver must lock onto each ONT's signal extremely quickly (within a few bits), since the optical power level can differ significantly between ONTs at different distances. Guard bands of a few microseconds are inserted between timeslots to account for timing tolerances.

This TDMA upstream architecture is fundamentally different from DOCSIS cable, which uses frequency division in addition to time division, and from DSL, which gives each subscriber a dedicated copper pair and thus avoids sharing entirely.

Wavelength Allocation in PON

Each PON standard assigns specific wavelengths for upstream and downstream transmission on the single shared fiber. This wavelength division multiplexing is what allows bidirectional communication without signal interference:

GPON (Gigabit Passive Optical Network)

GPON, defined by ITU-T G.984, is the most widely deployed PON standard globally. It provides 2.488 Gbps downstream and 1.244 Gbps upstream, shared among up to 64 or 128 ONTs per PON port (though 32 is the most common split ratio in practice). GPON uses 1490 nm for downstream and 1310 nm for upstream. An optional video overlay can use 1550 nm for RF video distribution, though this is increasingly rare as IPTV has replaced analog video.

GPON uses the GEM (GPON Encapsulation Method) protocol to encapsulate Ethernet frames, TDM circuits, and other payloads into 125-microsecond frames. This framing structure is what enables the precise TDMA scheduling that prevents upstream collisions.

XGS-PON (10G Symmetrical PON)

XGS-PON, defined by ITU-T G.9807.1, delivers 10 Gbps symmetrical — 10 Gbps both downstream and upstream. It uses 1577 nm for downstream and 1270 nm for upstream. Crucially, XGS-PON was designed for wavelength coexistence with GPON on the same fiber: the two systems use non-overlapping wavelengths, so operators can deploy XGS-PON incrementally alongside existing GPON without re-splicing any fiber. A wavelength-blocking filter (coexistence element) at the OLT combines and separates the two sets of wavelengths.

This coexistence capability is the key economic advantage of XGS-PON: operators can upgrade individual subscribers from 1 Gbps to 10 Gbps service without touching the outside plant, simply by swapping the ONT and adding an XGS-PON port to the OLT. The passive infrastructure (splitters, fibers) remains unchanged.

10G-EPON

The IEEE 802.3av standard defines 10G-EPON, providing 10 Gbps downstream and either 1 Gbps or 10 Gbps upstream. EPON-based standards are dominant in Asia (particularly Japan, South Korea, and China), while GPON-based standards dominate in Europe and North America. EPON uses native Ethernet framing rather than GEM, with a Multi-Point Control Protocol (MPCP) managing upstream timeslot allocation. The wavelength plan overlaps with GPON's, using 1310 nm upstream and 1490 nm downstream for 1G-EPON.

NG-PON2 (Next-Generation PON 2)

NG-PON2, defined by ITU-T G.989, is the most advanced PON standard currently deployed. It uses Time and Wavelength Division Multiplexing (TWDM): instead of a single downstream and upstream wavelength pair, NG-PON2 defines four to eight wavelength pairs, each carrying 10 Gbps. This provides an aggregate capacity of 40–80 Gbps per PON port. NG-PON2 operates in the L-band (1596–1603 nm) for downstream and C-band (1524–1544 nm) for upstream. ONTs must contain tunable lasers and receivers to select among the available wavelength channels, which increases ONT cost but provides massive capacity.

NG-PON2 also supports point-to-point WDM (PtP WDM) overlay, where individual wavelengths can be dedicated to specific high-bandwidth customers, effectively providing a private wavelength over the shared PON infrastructure.

OLT, ONT, and Splitter Infrastructure in Detail

The OLT

A modern OLT is a chassis-based system installed in the provider's central office (or in a hardened outdoor cabinet for distributed architectures). Each line card in the OLT provides multiple PON ports, where each port serves one PON tree of up to 32 or 64 subscribers. A typical enterprise OLT chassis might have 16 line cards with 16 PON ports each, serving up to 8,192 subscribers from a single chassis (assuming 1:32 split).

The OLT handles:

• Downstream scheduling — Encapsulating Ethernet frames into GEM frames, encrypting them, and transmitting them on the downstream wavelength.
• Upstream bandwidth allocation — Running the Dynamic Bandwidth Allocation (DBA) algorithm that assigns timeslots to each ONT every frame period (125 microseconds). Good DBA implementations can react to traffic bursts within a few hundred microseconds.
• ONT management — Provisioning, activation, software upgrades, and monitoring of all ONTs on the PON via OMCI (ONT Management and Control Interface) or PLOAM (Physical Layer Operations, Administration, and Maintenance) messages.
• Aggregation and uplink — Multiplexing all PON traffic onto 10G/100G uplinks toward the provider's aggregation routers and ultimately the internet backbone, where the traffic enters the BGP routing domain.

Passive Splitters

Optical splitters are manufactured using two technologies:

• Fused Biconical Taper (FBT) — Two or more fibers are fused together and stretched, causing light to couple between them. FBT splitters are cheap for low port counts (1:2, 1:4) but become expensive and uneven in power distribution at higher splits.
• Planar Lightwave Circuit (PLC) — A semiconductor-like manufacturing process etches waveguide patterns into a silica substrate. PLC splitters provide highly uniform splitting ratios, are compact, and scale well to 1:32 or 1:64. They are the standard choice for FTTH.

A 1:32 PLC splitter introduces approximately 17 dB of insertion loss (each split halves the power, and 2^5 = 32, so 5 x 3.4 dB per split = 17 dB). This loss, combined with fiber attenuation over distance, determines the maximum reach of a PON. GPON specifies a maximum optical path loss of 28 dB (Class B+), which supports roughly 20 km reach with a 1:32 split. XGS-PON offers an extended Class N2 budget of 35 dB for longer reach or higher split ratios.

The ONT

The ONT is the subscriber-facing device. Externally, it typically provides one or more Gigabit Ethernet ports (some models include 10G Ethernet), telephone ports (for VoIP/POTS), and sometimes a coaxial RF output for legacy video. Many operators deploy ONTs with built-in Wi-Fi routers (often called "residential gateways" or "home gateways"), combining the optical termination and routing functions in one device.

Internally, the ONT contains a laser transmitter (typically a DFB or EML laser for upstream), an avalanche photodiode (APD) receiver for downstream, a PON MAC chipset that handles TDMA timing and GEM framing, and an Ethernet switch chip. The ONT must synchronize its upstream transmission timing to nanosecond precision, as the TDMA guard bands between timeslots are measured in bits, not bytes.

Deployment Topologies

There are three main physical topologies for deploying FTTH, each with different trade-offs in cost, fiber usage, and flexibility:

Centralized Split

In a centralized split topology, a single splitter is placed close to the OLT (often in the central office or in a nearby fiber distribution hub). From the splitter, individual fibers run all the way to each subscriber. This approach uses more fiber but fewer splitter locations, making it simpler to manage and easier to migrate subscribers between PON ports. It is common in dense urban deployments where the central office is relatively close to all subscribers.

Distributed Split (Cascaded)

Distributed split uses two stages of splitting. A first-stage splitter (e.g., 1:4) is placed at a fiber distribution hub, and second-stage splitters (e.g., 1:8 each) are placed closer to the subscribers, often in neighborhood pedestals or on utility poles. The total split remains the same (4 x 8 = 32), but the distribution fiber between the first and second stages serves groups of subscribers, reducing the total fiber count. This is the most common topology for suburban FTTH deployments where subscribers are spread over larger areas.

Home Run (Point-to-Point) Fiber

Home run fiber runs a dedicated fiber from the central office to each subscriber — no splitters at all. Each subscriber gets the full bandwidth of their own fiber and their own OLT port. This is the most expensive topology per subscriber (it requires N fibers for N subscribers and N OLT ports), but it offers the highest bandwidth potential and the greatest flexibility. Home run is common for business and enterprise FTTH services, and some greenfield residential deployments (particularly in the Nordic countries and the Netherlands) use it as well, betting that the long-term value of a dedicated fiber justifies the higher upfront cost.

Inside the Home: ONT to Router

Fiber enters the home through a wall penetration or conduit and terminates at a fiber termination box (also called a fiber rosette or wall outlet). A short patch cord connects the termination box to the ONT. The fiber connector type is typically SC/APC (angled physical contact) with a green ferrule, which minimizes back-reflections that can degrade PON performance.

The ONT may be a standalone device (an "ONT-only" box with Ethernet ports), or it may be a combination ONT/router/Wi-Fi access point (a "residential gateway"). With standalone ONTs, the subscriber connects their own router to the ONT's Ethernet port. With integrated gateways, the ONT handles routing, NAT, DHCP, and Wi-Fi in addition to optical termination.

Increasingly, operators are deploying ONTs with XGS-PON capability and 10G Ethernet output ports, enabling multi-gigabit service to the home. The subscriber's internal wiring then becomes the bottleneck: Cat5e Ethernet supports 1 Gbps, Cat6 supports up to 5 Gbps (with 2.5GBASE-T and 5GBASE-T), and Cat6a supports 10 Gbps. Wi-Fi 6E and Wi-Fi 7 can deliver multi-gigabit speeds wirelessly, finally matching the capacity of the fiber connection.

Symmetrical vs Asymmetrical Speeds

Traditional GPON is asymmetrical: 2.488 Gbps downstream but only 1.244 Gbps upstream, shared among all subscribers on the PON. This ratio made sense when residential traffic was overwhelmingly download-heavy (web browsing, video streaming). But the growth of video conferencing, cloud backup, live streaming, and content creation has made upstream bandwidth increasingly valuable.

XGS-PON delivers 10 Gbps symmetrical, and this is what most new FTTH deployments now target. Symmetrical service means a subscriber can upload at the same speed as they download, which is transformative for work-from-home productivity, real-time collaboration, and cloud-native workflows.

Even with symmetrical PON capacity, operators may still sell asymmetrical speed tiers (e.g., "2 Gbps down / 1 Gbps up") as a product differentiation and pricing strategy, even though the underlying network is capable of symmetrical delivery. This is a commercial decision, not a technical limitation.

FTTH vs FTTC, FTTB, and FTTN

The "FTTx" family describes how far toward the subscriber the fiber extends before transitioning to copper for the final segment:

• FTTH (Fiber to the Home) — Fiber all the way to the subscriber's premises. No copper in the path. Maximum speed potential with the lowest latency and highest reliability.
• FTTB (Fiber to the Building) — Fiber terminates in the basement or utility room of a multi-dwelling unit (apartment building). The connection from the building's fiber termination to each apartment uses copper — either Ethernet (if the building is wired with Cat5e/Cat6) or VDSL2 over existing telephone wiring. Common in dense urban areas where re-wiring apartments is impractical.
• FTTC (Fiber to the Curb/Cabinet) — Fiber terminates at a street-side cabinet, typically within 300 meters of the subscriber. The last segment uses VDSL2 or G.fast over existing copper telephone lines. FTTC delivers speeds of 50–300 Mbps depending on loop length, significantly better than FTTN but far below true FTTH. BT (Openreach) in the UK deployed FTTC extensively as a faster-to-deploy alternative to full FTTH.
• FTTN (Fiber to the Node) — Fiber terminates at a neighborhood node, which may be 500–1500 meters from subscribers. The long copper run limits speeds to 25–100 Mbps via VDSL2. Australia's NBN initially relied heavily on FTTN, resulting in widespread performance complaints and an eventual pivot toward FTTH and FTTC.

The fundamental trade-off is clear: the closer the fiber gets to the subscriber, the better the performance, but the higher the construction cost. FTTH is the only FTTx variant that completely eliminates copper and its associated limitations.

Comparison to Cable (DOCSIS) and DSL

Understanding FTTH's advantages requires comparing it with the two other major last-mile technologies:

FTTH vs DOCSIS (Cable)

DOCSIS delivers broadband over hybrid fiber-coax (HFC) networks — fiber to a neighborhood node, then coaxial cable to the home. DOCSIS 3.1 supports up to 10 Gbps downstream and 1–2 Gbps upstream in theory, but actual deployments typically offer 1–2 Gbps downstream and 100–200 Mbps upstream. The coaxial cable segment is a shared medium with significant noise challenges, and upstream capacity is constrained by the limited spectrum allocation (5–204 MHz in DOCSIS 3.1, extended to 684 MHz in DOCSIS 4.0).

Key differences from FTTH:

• Upstream capacity — FTTH (XGS-PON) delivers 10 Gbps upstream, versus DOCSIS 3.1's practical 100–200 Mbps. Even DOCSIS 4.0 targets only 6 Gbps upstream.
• Latency — FTTH typically delivers sub-millisecond latency on the access network, while DOCSIS introduces 5–15 ms due to its more complex MAC scheduling.
• Signal integrity — Fiber is immune to electromagnetic interference. Coax is susceptible to ingress noise from loose connectors, damaged cables, and household electronics.
• Distance — Fiber works over 20+ km without signal degradation. Coax performance degrades with each amplifier cascade and every hundred meters of cable.
• Upgrade path — FTTH can be upgraded from GPON to XGS-PON to 25G-PON by changing electronics only. DOCSIS upgrades require replacing the coax plant or squeezing more bandwidth from a fixed spectrum window.

FTTH vs DSL

DSL uses existing telephone copper pairs. VDSL2, the fastest widely deployed variant, reaches 100–300 Mbps over short loops (under 300 m) and drops to 10–30 Mbps over typical 1–2 km loops. G.fast achieves up to 1 Gbps but only over loops shorter than 100 meters. DSL gives each subscriber a dedicated copper pair (no sharing), but the fundamental physics of copper — frequency-dependent attenuation, crosstalk between pairs in the same cable bundle, and electromagnetic susceptibility — impose hard bandwidth limits that cannot be overcome without shortening the copper segment (which is what FTTC and FTTB do).

Fiber has none of these limitations. A single-mode fiber's bandwidth capacity is measured in petabits per second using dense WDM, while a copper pair is limited to a few gigabits at most over short distances. FTTH represents a generational leap, not an incremental improvement.

Real-World Deployment Economics

The economics of FTTH deployment are dominated by civil works — the physical labor of trenching, boring, and stringing fiber cable. The fiber itself is remarkably cheap; the cost breakdown for a typical FTTH deployment looks roughly like this:

• Civil works (trenching, boring, aerial installation) — 60–80% of total cost. In urban areas, trenching through streets and sidewalks costs $30–80 per meter. Boring under roads costs $100–300 per crossing. Aerial installation on existing poles is cheaper ($15–30 per meter) but not available everywhere.
• Fiber cable — 5–10% of total cost. A 288-fiber loose-tube cable costs $2–5 per meter. The fiber itself (a strand of glass) costs fractions of a penny per meter; the cost is in the cable jacket, strength members, and gel filling.
• Splitters and passive components — 2–5% of total cost. A 1:32 PLC splitter costs $30–60. Splice closures, fiber distribution hubs, and patch panels add a few hundred dollars per location.
• OLT equipment — 5–10% of total cost. An XGS-PON OLT port costs roughly $500–1,000 and serves 32–64 subscribers.
• ONT (customer premises equipment) — 5–10% of total cost. A GPON ONT costs $30–80; an XGS-PON ONT costs $80–200. This is a per-subscriber cost that scales linearly.

The total cost per passing (the cost to make fiber available to a premises, whether they subscribe or not) ranges from $800–1,500 in dense urban areas to $2,000–5,000+ in rural areas. The cost per connected subscriber adds the ONT, drop cable installation, and provisioning, typically another $300–600. The enormous variation in cost per passing is driven almost entirely by density: fiber runs past dozens of potential subscribers per trench-meter in a city, but past only one or two per hundred meters in a rural area.

This economic reality explains the global deployment pattern: FTTH coverage is highest in dense countries (South Korea, Japan, Singapore, UAE) and urban areas of larger countries, while rural areas lag behind and often require government subsidies (like the US BEAD program or EU fiber funds) to close the business case.

The Connection to the Internet

From the subscriber's perspective, FTTH provides a fast link to the ISP's network. But what happens after the OLT? The PON traffic is aggregated onto the ISP's metro Ethernet network, routed through their backbone, and then exchanged with other networks via BGP peering at Internet Exchange Points or through transit providers. The global routing fabric that connects all these networks — including the submarine cables that span oceans — is what ultimately determines your end-to-end internet experience. FTTH eliminates the last-mile bottleneck, ensuring that the access network is no longer the weak link in the chain.

You can trace this path from subscriber to the global internet by looking up your ISP's autonomous system. For example, major FTTH providers like Charter (AS20115), Verizon (AS701), or Free/Iliad (AS12322) each operate their own AS and exchange routes with the global internet via BGP.

The Future: 25G-PON, 50G-PON, and Beyond

The PON technology roadmap continues to push bandwidth higher without replacing the outside plant:

25G-PON and 50G-PON

The ITU-T and IEEE are both standardizing the next generation of PON. ITU-T G.9804 (50G-PON) defines 50 Gbps downstream with 25 or 50 Gbps upstream. IEEE is standardizing 25G-EPON and 50G-EPON. These standards will coexist on the same fiber as GPON and XGS-PON via wavelength allocation, maintaining the "upgrade electronics, keep the fiber" model.

The technical challenges at these speeds are significant. At 50 Gbps, the PON must use advanced modulation (PAM-4 or NRZ with equalization), forward error correction (FEC) with higher coding gain, and more sensitive receivers. The burst-mode receivers required for TDMA upstream must lock on even faster — within a few nanoseconds. Despite these challenges, 50G-PON equipment is expected in commercial deployments by 2026–2028.

Coherent Optics to the Home

Looking further ahead, coherent optics — the technology that revolutionized long-haul fiber transmission — may eventually reach the access network. Coherent detection uses a local oscillator laser at the receiver and digital signal processing (DSP) to demodulate the signal, enabling detection of both amplitude and phase across both polarizations of light. This allows spectral efficiencies of 4–8 bits per symbol, compared to 1 bit per symbol with traditional on-off keying (OOK).

Coherent PON (which Nokia and others have demonstrated in labs) could deliver 100 Gbps or more per wavelength, with greater reach and higher split ratios than current PON. The barrier is cost: coherent transceivers currently cost thousands of dollars per unit, while PON ONT optics cost tens of dollars. But silicon photonics integration is driving coherent transceiver costs down rapidly. If coherent ONTs can be manufactured for under $100 — which some industry roadmaps project by the early 2030s — coherent PON could become the dominant FTTH technology, providing virtually unlimited bandwidth on the same single-mode fiber deployed today.

The Fiber Endgame

The physical capacity of a single-mode fiber is staggering. Researchers have demonstrated over 1 petabit per second (1,000 Tbps) on a single standard fiber using dense WDM and multi-band transmission. Today's FTTH networks use a tiny fraction of this capacity — even a 50G-PON uses just two wavelengths. The fiber already in the ground has enough bandwidth headroom to serve subscribers for decades, with each technology generation simply using a slightly larger slice of the fiber's spectrum.

This is the fundamental promise of FTTH: the passive infrastructure (fiber, splitters, conduits) is a one-time investment that will outlast many generations of active electronics. An operator that buries fiber today is deploying infrastructure with a useful life of 30–50 years, capable of supporting technologies that have not yet been invented. No copper-based technology can make that claim.

See Your Connection in Action

Whether your home connects via FTTH, DOCSIS, or DSL, your traffic ultimately enters the BGP routing system once it leaves your ISP's network. You can explore this routing infrastructure with a looking glass tool:

• Look up your IP address to see which network carries your traffic
• Verizon Fios (AS701) — one of the largest FTTH providers in the US
• Free/Iliad (AS12322) — a major European FTTH operator
• Korea Telecom (AS4766) — pioneer of nationwide FTTH in South Korea