1. Why Glass Beats Copper

Before we touch a single fiber, sit with the basic question: why bother with optical communications at all? Copper has been a perfectly good signal carrier since Samuel Morse. Why throw it out?

The answer comes from comparing five numbers between the two media: bandwidth, loss per kilometer, regenerator spacing, immunity to interference, and weight per gigabit.

1.1 The bandwidth gap

A coaxial cable, the best copper carrier ever deployed at scale, can run a few gigahertz cleanly over short distances. A category-6 Ethernet cable runs 10 Gbps over 100 meters. A high-end DAC (direct-attach copper) cable in a server rack squeaks out 100 Gbps over 3 meters before the eye closes. Push any of those farther and the signal turns to mush.

A single strand of single-mode optical fiber, by contrast, carries thousands of separate wavelength channels in the C-band (1530 to 1565 nm) alone, each running at hundreds of gigabits per second. Modern submarine cables ferry tens of terabits per second through one fiber pair. The MAREA cable that crosses the Atlantic between Virginia Beach and Bilbao runs at 224 Tbps. That is roughly the entire 2010 internet's peak traffic, on one cable.

Where does the bandwidth come from? From the carrier frequency. A signal cannot easily carry more bandwidth than a small fraction of its carrier. Try to amplitude-modulate a 1 GHz radio carrier with a 5 GHz baseband signal and you get nonsense. With copper Ethernet, your carrier might be a few GHz. With optical communications at 1550 nm, your carrier is

$f = c/\lambda = (3 \times 10^8 \text{ m/s})/(1550 \times 10^{-9} \text{ m}) \approx 193\text{ THz}.$

Two hundred terahertz of carrier. Even a tiny fractional bandwidth gives you many tens of terahertz to play in. The fundamental ceiling is enormous, and we are nowhere near it.

Pipe analogy. Copper is a pipe: rough walls, internal friction, bends that cost you flow. Push water through it, friction heats it up, and the friction grows worse the harder you push. Fiber is a perfectly polished tube, almost frictionless, that also happens to allow many independent colored streams of water to flow through at once without mixing. Different colors of light travel through glass without interacting. Different frequencies of current in copper crosstalk like crazy.

1.2 The loss gap

A typical RG-58 coaxial cable loses about 50 dB per 100 m at 1 GHz. Translate to per-km units and you are looking at 500 dB/km. After one kilometer of cable your 1 GHz signal has been attenuated by ten to the fiftieth power: utterly destroyed. Even premium cables for cellular base stations lose 50 to 100 dB/km at GHz frequencies.

A modern single-mode silica fiber at 1550 nm loses about 0.18 dB/km. Three orders of magnitude better than the best copper. After 100 km of fiber you are down 18 dB, a manageable loss that an erbium-doped fiber amplifier can recover in less than a meter of doped fiber.

Why is glass so transparent? Because the losses we discussed for copper (resistive heating in $I^2 R$, dielectric polarization losses, skin-effect surface heating) all scale with frequency. Copper gets worse as you push it harder. Glass, by contrast, is dominated by Rayleigh scattering of light off microscopic density variations, and that scattering scales as $1/\lambda^4$. Move from visible (around 500 nm) to near-IR (1550 nm), and you reduce scattering by a factor of $(1550/500)^4 \approx 92$. The longer the wavelength, the less the photons "see" the molecular density of the glass.

1.3 The regenerator gap

Cables that lose a lot need amplifiers spaced close together. The first transatlantic copper telephone cable (TAT-1, 1956) used 51 vacuum-tube repeaters across the Atlantic. The first transatlantic fiber cable (TAT-8, 1988) used 109 electronic regenerators across roughly the same distance. The current generation of submarine cables uses passive optical amplifiers (no electronics in the deep-sea repeater, just doped fiber and a pump laser) spaced 60 to 90 km apart, and the entire optical signal stays in the fiber from coast to coast. The cost-of-ownership advantage compounds: fewer parts to fail under 10,000 meters of seawater pressure for 25 years.

1.4 The interference gap

Glass does not pick up nearby radio. It is electrically inert. A 50,000-volt motor running next to a fiber cable does nothing measurable to the signal. A 50,000-volt motor next to a copper cable injects spectacular noise spikes through magnetic coupling. This is why fiber gets used inside power substations, near radar transmitters, and on military ships. It is also why fiber is the standard backhaul for any installation that has to survive a lightning strike: the fiber doesn't care, and there's no copper path for the surge to use to fry the equipment on the other end.

1.5 The weight and tap gaps

A pair of single-mode fibers can carry 100 Gbps. A copper bundle that does the same job weighs ten to a hundred times more. This matters in airliners (Boeing's 787 uses fiber for the in-flight entertainment backbone), in ships, and in long-haul ducts where the weight of the cable itself is a logistical problem.

Finally, fiber is harder to tap. Cutting a copper cable lets you splice a parallel listener with relative ease. Cutting fiber drops the light, which is immediately detectable at the receiver. Bend-coupling attacks (Section 11) do exist, but they have a measurable insertion loss signature, and continuous OTDR monitoring picks them up. We will revisit fiber tapping as a hardware-security topic later.

The drawbacks of fiber are real but mostly logistical: connectors are more delicate, splicing requires expensive specialized equipment, and bending past a critical radius will leak light. None of these come close to outweighing the advantages on any link more than a few meters long.

1.6 A short history

Fiber-optic communication is only about as old as broadband. A condensed timeline:

• 1854. John Tyndall demonstrates that light follows a curved water jet from a tank, the first total-internal-reflection light guide.
• 1880. Alexander Graham Bell patents the photophone, a free-space optical voice link.
• 1960. Theodore Maiman demonstrates the first laser. Optical communications becomes plausible.
• 1966. Charles Kao and George Hockham at STC predict that low-loss glass fiber, around 20 dB/km, would be enough for long-haul communication. At the time, the best fibers ran at 1000 dB/km. Kao won the Nobel Prize in 2009 for this insight.
• 1970. Corning's Robert Maurer, Donald Keck, and Peter Schultz produce silica fiber with 20 dB/km loss using titanium-doped silica. The first time fiber crosses the threshold Kao predicted.
• 1976. AT&T Bell Labs runs a 6 Mbps experimental link over 11 km of fiber in Atlanta.
• 1977. First commercial fiber link goes live (Chicago, GTE).
• 1988. TAT-8, the first transatlantic fiber cable, carries 280 Mbps across the Atlantic. The era of submarine fiber begins.
• 1990s. Erbium-doped fiber amplifiers (EDFAs) replace electronic regenerators in long-haul cables. Suddenly you can amplify the optical signal directly without converting to electrical. WDM begins squeezing many wavelengths into one fiber.
• 2010. Coherent transmission systems (digital signal processing on the receive side, recovering both phase and polarization) bring 100 Gbps per channel into commercial deployment.
• 2020s. 800 Gbps per wavelength, 96+ wavelengths per fiber, hundreds of terabits per second per cable on the trans-oceanic routes. PAM-4 and silicon photonics dominate intra-data-center optics.

1.7 The generic fiber-optic system

Every fiber link, from a 10-meter datacenter patch cord to a 10,000 km submarine cable, has the same five-stage block diagram:

Source, optional external modulator, fiber, optional amplifier, photodetector, electrical receiver. The variations are in which boxes are merged, which wavelengths are used, what modulation is used, and how you split or combine wavelengths. Understanding each block is the rest of this chapter.