How optical fibre is made: preforms, drawing, coating, proof test
A telecom fibre is a glass thread 125 µm across, drawn from a preform the size of a baseball bat at tens of metres per second, coated within a fraction of a second and stressed to 1 % strain along its whole length before it is sold. The manufacturing route explains most of the numbers an engineer meets later: why mode field diameter is specified to ±0.4 µm, why splices between vendors lose a little more, why installation strain is capped at a fraction of the proof test, why modern fibre is hydrogen-safe and why 200 µm fibres need different splicer settings. This page follows the process from raw chemicals to the spool and ends with what it means in the field.
Materials
| Material | Role |
|---|
| SiCl₄ (silicon tetrachloride) | ultra-pure silica source; oxidised or hydrolysed in a flame or furnace to SiO₂ soot |
| GeCl₄ → GeO₂ | raises the refractive index — the core dopant of almost all telecom fibres; slightly increases Rayleigh scattering |
| Fluorine (SiF₄, C₂F₆, SF₆) | lowers the index — depressed claddings, the trench of bend-insensitive G.657 fibre, the cladding of pure-silica-core fibres |
| POCl₃ → P₂O₅ | lowers viscosity, used in MCVD claddings; makes glass hydrogen-sensitive — a reason old fibres age badly |
| Al₂O₃, Er, Yb, Tm | active fibres for amplifiers and lasers, added by solution doping of the porous soot |
| Cl₂ and He | drying and consolidation: chlorine removes OH to below 1 ppb (low water peak), helium purges the porous body |
| UV-curable acrylates | dual coating: a soft primary (modulus ~1 MPa) cushions the glass, a hard secondary (~1 GPa) protects it |
| Colour inks, ribbon matrix | identification (12 colours) and ribbon assembly |
| Process | How it works | Who uses it | Strengths |
|---|
| MCVD (modified chemical vapour deposition, Bell Labs 1974) | vapours flow inside a rotating silica tube heated by an external torch; soot deposits layer by layer (cladding, then core); the tube is collapsed into a solid rod | most manufacturers, all specialty fibre makers | flexible: erbium-doped, polarization-maintaining, dispersion-compensating fibres; moderate preform size |
| PCVD (plasma-activated CVD) | like MCVD, but a microwave plasma replaces the torch; thousands of very thin layers | graded-index multimode makers | precise index profiles — OM3/OM4/OM5 |
| OVD (outside vapour deposition) | soot from a flame is deposited on a rotating target rod, core layers first; the rod is removed, the porous body is dried in Cl₂ and sintered to clear glass | large-volume single-mode and multimode production | very low OH, large preforms, core and cladding in one process |
| VAD (vapour-phase axial deposition) | soot grows on the end of a rotating rod that is pulled upward; continuous process | large-volume single-mode production | very long preforms, low OH, high throughput |
| Overcladding (RIC — rod in cylinder; soot overclad; sleeving) | the core rod, which carries the whole index profile, is surrounded by cheap synthetic silica — a cylinder fused on during draw, or extra soot | everyone | the expensive doped glass is only a few percent of the volume |
A finished preform is 100–200 mm in diameter and 1–3 m long. A 150 mm × 2 m preform draws into roughly 2 500–3 000 km of 125 µm fibre; the largest modern preforms yield more than 5 000 km.
Drawing
| Step | What happens | Numbers |
|---|
| Furnace | the preform tip softens in a graphite or zirconia induction furnace | 1 900–2 100 °C |
| Neck-down and draw | a thread is pulled from the molten tip; speed sets the diameter | 20–40 m/s (1 200–2 500 m/min), modern towers faster |
| Diameter control | a laser gauge measures the bare fibre and feeds back to the capstan speed | 125 ± 0.7 µm specified, ±0.1 µm achieved |
| Cooling | helium-filled tube cools the glass enough to be coated | a few hundred °C in less than a metre |
| Primary coating | soft acrylate applied through a pressurised die, cured by UV lamps | to ~190 µm |
| Secondary coating | hard acrylate, cured | 245 ± 10 µm standard; 200 ± 10 µm for high-density cables |
| Spinning | the fibre is twisted alternately clockwise and counter-clockwise during draw, mixing the birefringence axes | PMD coefficient < 0.1 ps/√km instead of 0.5–2 ps/√km on 1980s fibre |
| Capstan and winding | tension-controlled winding on spools | 25–50 km per spool; towers are 20–40 m tall |
Colouring adds an ink layer (or a coloured secondary); ribbons bond 4–12 coloured fibres in a matrix, rollable ribbons bond them intermittently so they can be rolled into a tube.
Proof test and quality assurance
| Test | Method (IEC 60793-1-xx) | Specification for G.652.D |
|---|
| Proof test | every metre runs through a tension of 0.69 GPa (100 kpsi, ≈ 1 % strain) for ≥ 1 s (part 1-30); submarine and special fibres 1.38 GPa | any flaw that would break below that stress breaks here, not in the ground |
| Geometry | cladding diameter, non-circularity, core–cladding concentricity (part 1-20) | 125 ± 0.7 µm; ≤ 1 %; ≤ 0.6 µm — connector and splice alignment depend on it |
| Attenuation | cutback (part 1-40) at 1310/1550/1625 nm plus a spectral scan; OTDR on every spool for uniformity and point defects | ≤ 0.35 / 0.21 / 0.23 dB/km typical; point defects ≤ 0.05 dB |
| Mode field diameter | far-field scan (part 1-45) | 9.2 ± 0.4 µm at 1310 nm |
| Cutoff wavelength | part 1-44 | cabled cutoff ≤ 1 260 nm |
| Chromatic dispersion | phase-shift (part 1-42) | λ₀ = 1 300–1 324 nm, slope ≤ 0.092 ps/(nm²·km) |
| PMD | part 1-48 | link design value ≤ 0.20 ps/√km |
| Macrobend loss | part 1-47 | ≤ 0.1 dB for 100 turns at 30 mm radius (1625 nm) |
| Coating | strip force 1.3–8.9 N (part 1-32), coating geometry, curl | consistent stripping in the field |
| Hydrogen ageing | IEC 60793-2-50 | ≤ 0.01 dB/km increase after exposure — modern fibre is hydrogen-safe |
Every spool ships with a data sheet of measured values. Cable makers keep the mapping spool → cable → fibre number; ask for those fibre data sheets during acceptance — they are the first baseline of your plant (Documentation & labelling).
Special fibres and what is different in making them
| Fibre | Manufacturing difference | Use |
|---|
| Bend-insensitive G.657 | a fluorine-doped trench ring deposited around the core (MCVD/PCVD or OVD) | FTTH, high-density panels, 200 µm cables |
| Graded-index multimode OM3–OM5 | thousands of layers with a parabolic (α ≈ 2) index profile; differential mode delay measured on every preform | short-reach VCSEL links |
| Erbium-doped (EDF) | porous core soot soaked in Er/Al solution before sintering; short lengths (metres) per amplifier | EDFAs |
| Polarization-maintaining (PANDA, bow-tie) | holes drilled in the preform and boron-doped stress rods inserted, or shaped stress regions | coherent transceivers internally, sensors, pump combiners |
| Dispersion-compensating (DCF) | small core with high index step giving large negative dispersion | legacy 10G/40G compensation |
| Pure-silica-core G.654 | undoped core, fluorine-doped cladding; large effective area variants (G.654.E) | submarine and long-haul coherent: 0.15–0.17 dB/km |
| Hollow-core (NANF and relatives) | stack-and-draw of silica capillaries forming an air core; light travels 1.46 × faster and almost without non-linearity | latency-critical links, research; losses below 0.2 dB/km reported in the 2020s |
| Multicore | 4–7 cores in a 125–200 µm cladding, drawn from a stacked preform | space-division multiplexing, submarine systems since the 2020s |
| Radiation-hard | fluorine-doped core and cladding, no germanium | nuclear, space, accelerators |
| Carbon or metal coated | hermetic carbon layer applied in the draw tower before the acrylate | oil and gas, hydrogen-rich and high-temperature environments |
Why a network engineer cares
| Manufacturing fact | Field consequence |
|---|
| MFD specified to ±0.4 µm | splices between fibres from different vendors or eras can lose 0.02–0.05 dB more; gainers and losers on the OTDR (Splicing & termination) |
| Proof test at 1 % strain | long-term installed strain is limited to 20 % of proof (0.2 %), installation to 60 % (Reliability & ageing) |
| Concentricity ≤ 0.6 µm | connector loss is dominated by ferrule and polish, not by the glass |
| Spinning since the late 1990s | 100G coherent runs on any fibre made after about 2000; pre-1995 fibre needs a PMD test (Fibre characterization) |
| Low-OH drying | the 1383 nm water peak is gone: E-band CWDM and 1625/1650 nm monitoring are usable (Attenuation & windows) |
| 200 µm coating | smaller cables, but holders, stripper settings and cleaver clamps must match; mixed 200/250 µm splicing needs the right holders |
| Cabled attenuation > fibre attenuation | design with the cabled value from the cable data sheet, not the fibre value (Cable construction) |
| Ribbons | mass fusion splicers, 12 fibres per splice; identification by ribbon position |
In CodingBox
A transceiver knows nothing about how the fibre was made, but its rated lengths for SMF, OM3 and OM4 assume fibre made to these tolerances — G.652.D or the stated OM grade. CodingBox shows those length fields; on unusual fibre (G.654, G.653, hollow-core) the rated distances do not apply and the link budget has to be built from the fibre's own data sheet (Check transceiver, Fibre types).