Fiber Basics
The physical layer under everything else. Get this right and the optics behave; get it wrong and no amount of configuration will save the link.
What "dark fiber" actually is
Dark fiber is installed optical fiber with no light on it — no transceivers, no service, nothing lit. A carrier or landlord leases you the raw glass between two points (an A site and a Z site) and you are responsible for lighting it.
Think of it exactly like leasing an empty conduit: the value is the path, but it does nothing until you put equipment on both ends. Your optical gear (Ekinops) is what makes it "lit."
You are buying a path, not a service. Everything in this course after "fiber basics" is about what you attach to each end of that path.
Strands and pairs
A fiber cable contains many individual glass strands. A single strand carries light in one direction at a time in typical designs, so most services use a pair: one strand to transmit (Tx) and one to receive (Rx). Each strand has an ID (from the cable/splice records) that you must track end to end.
Assuming "a fiber" means a pair. In records and hand-offs, "fiber 12" is usually one strand. Always confirm whether you have the two strands that form a working pair, and which strand ID is Tx vs Rx at each site.
Single-mode vs multimode
| Property | Single-mode (SMF) | Multimode (MMF) |
|---|---|---|
| Core size | Small (~9 µm) | Large (~50/62.5 µm) |
| Distance | Long — km to hundreds of km | Short — within a building/campus |
| Light source | Laser | LED / VCSEL |
| Typical use here | Dark fiber transport, WDM, DWDM | In-building patching, short runs |
| Jacket color (common) | Yellow | Aqua / orange / other |
Dark-fiber optical transport is single-mode. If you see multimode on a transport path, something is wrong — you cannot run long-haul WDM over MMF.
Why the core size decides everything
Single-mode fiber has a core of only about 9 µm — narrow enough that light can travel in essentially one path (one mode) down the glass. Multimode's much larger core (50 µm for OM2–OM5, 62.5 µm for OM1) lets light take many paths at once. Those paths arrive at slightly different times (modal dispersion), which smears the pulse and kills reach and bit-rate over distance.
WDM depends on precise, narrow-linewidth laser wavelengths staying clean over tens or hundreds of km. Only single-mode's single-path geometry keeps a pulse coherent that far — which is exactly why every WDM/long-haul design is single-mode. MMF is a short-reach, in-building technology.
| Grade | Type | Core | Notes for transport work |
|---|---|---|---|
| OS1 | SMF | ~9 µm | Indoor/tight-buffered single-mode. Slightly higher attenuation spec. |
| OS2 | SMF | ~9 µm | Loose-tube outside-plant SMF, low water-peak. The dark-fiber standard for long spans. |
| OM1 | MMF | 62.5 µm | Legacy. Orange jacket. Short in-building only. |
| OM2 | MMF | 50 µm | Legacy. Orange jacket. |
| OM3 / OM4 | MMF | 50 µm | Laser-optimized (VCSEL, 850 nm). Aqua jacket. Data-center reach only. |
| OM5 | MMF | 50 µm | Wideband MMF for short-reach SWDM. Lime jacket. Still not transport. |
Wavelengths and optical bands
Single-mode fiber has two classic low-loss operating windows — 1310 nm and 1550 nm. WDM systems live mostly in the C-band around 1550 nm because that's where fiber loss is lowest and where erbium amplifiers (EDFAs) work. The named bands:
| Band | Range (nm) | Where you meet it |
|---|---|---|
| O (Original) | 1260–1360 | 1310 nm grey optics, short-reach client side, CWDM lower channels. |
| E (Extended) | 1360–1460 | The old water-peak region (~1383 nm) — high loss in legacy fiber. |
| S (Short) | 1460–1530 | Some CWDM channels. |
| C (Conventional) | 1530–1565 | The DWDM workhorse. Lowest loss, EDFA-amplifiable. |
| L (Long) | 1565–1625 | Band extension for extra DWDM channels on high-capacity systems. |
Assuming any wavelength runs fine on old glass. Legacy (non-low-water-peak) fiber has a big attenuation spike near 1383 nm in the E-band. It rarely matters for C-band DWDM, but it can wreck a CWDM plan that tries to use channels in that region. Confirm the fiber's water-peak spec before planning CWDM.
Connectors: LC/SC and UPC/APC
Two independent choices describe a connector: its form factor (the physical latch/ferrule) and its polish (the endface geometry).
Form factor
- LC — small, push-pull latch. The common modern choice on transponders and patch panels.
- SC — larger, square, push-pull. Common on older gear and some outside-plant panels.
Polish
- UPC (Ultra Physical Contact) — flat/domed endface, usually blue housing.
- APC (Angled Physical Contact) — 8° angled endface, usually green housing. Much better return loss.
Mating APC to UPC. The angled and flat endfaces don't seat properly — you get high loss and high reflection, and you can damage the ferrules. Green mates green, blue mates blue. Match the polish before you plug anything in.
Return loss: why APC exists
The difference between UPC and APC is the angle of the polished endface, and it shows up directly in return loss (reflected light). An angled face sends the reflection off into the cladding instead of straight back up the fiber.
| Polish | Endface | Typical return loss | Housing | Use |
|---|---|---|---|---|
| UPC | Domed, flat (0°) | ≥ 50 dB | Blue | Data, most client/grey optics, general patching. |
| APC | Angled 8° | ≥ 60 dB | Green | Transport, DWDM, high-rate coherent, analog/RF-over-fiber. |
Remember return loss is a positive number where bigger is better: 60 dB of return loss means only one-millionth of the light reflects back, versus one-hundred-thousandth at 50 dB. High reflections feed back into the laser and destabilize it — a real problem on narrow-linewidth DWDM sources — which is why transport work standardizes on APC.
MPO / MTP — many fibers, one connector
Beyond single-strand LC/SC, high-count links use MPO (a.k.a. MTP, a branded MPO): one rectangular ferrule carrying 8, 12, 16, or 24 fibers in a row. You'll see MPO on 40G/100G/400G parallel optics and on structured-cabling trunks that break out to LC at a cassette.
- Has a key (up/down) and pin/no-pin (male/female) genders — mismatches don't mate.
- Comes in polarity methods A, B, and C (see below) that route the fiber positions differently.
- One dirty MPO endface = up to 24 links down at once. Inspection matters even more here.
Patch panels, jumpers, meet-me rooms, demarcation
- Patch panel: where outside-plant fiber terminates onto connectors you can reach. Every panel port is a potential loss point.
- Jumper (patch cord): a short fiber that connects a panel port to your equipment. Wrong length, wrong polish, or a dirty jumper is a frequent turn-up failure.
- Meet-me room (MMR): a shared space (often in a carrier hotel) where different providers cross-connect. Your "dark fiber" often hands off here.
- Demarcation point (demarc): the exact port/panel where the carrier's responsibility ends and yours begins. Know it precisely — it defines who fixes what.
Tx/Rx polarity
The transmit fiber at one end must land on the receive at the other. Cross them and each transceiver is shouting into another transmitter and hearing nothing — the link stays down even though light is present on the glass.
Polarity is a connectivity problem, not a power problem. A power meter can read good light while the strands are swapped, because it doesn't care which direction the signal is supposed to go.
Duplex-LC rollover (A-to-B)
A duplex-LC jumper holds two fibers side by side, labeled A and B. Correct polarity is achieved by a rollover: position A at one end lands on position B at the other, so every Tx meets the far-end Rx. A straight-through (A-to-A) duplex cable would cross the pair — Tx into Tx.
MPO polarity: methods A / B / C
Higher-count MPO trunks solve the same Tx→Rx problem with three defined wiring conventions. You don't have to memorize the wiring, but you must not mix methods within one channel — an A-cassette on a B-trunk crosses fibers.
- Method A (straight): key-up to key-down; polarity fixed by flipping one duplex end at the breakout.
- Method B (flipped): key-up to key-up; the trunk itself reverses fiber positions.
- Method C (pairs-flipped): adjacent fibers swapped in pairs inside the trunk.
Fiber pair vs single-fiber (BiDi)
There are two ways to get both directions of a link across dark fiber. Know which one a service uses before you order strands or optics.
Fiber pair (two strands)
The classic design: one strand for Tx, one for Rx, usually the same wavelength in each direction. Simple, symmetric, easy to test — but it consumes two strands per link.
Single-fiber BiDi (one strand)
Both directions ride one strand using two different wavelengths (e.g. 1310 nm one way, 1550 nm the other). Doubles strand efficiency — but the two ends are not identical: they're a matched Tx/Rx-wavelength pair (often sold as "U" and "D" / upstream and downstream).
Ordering two identical BiDi optics. BiDi ends are complementary: one transmits 1310/receives 1550, the other transmits 1550/receives 1310. Two of the same "half" will both talk on the same color and hear nothing. Always pair the matching upstream/downstream halves.
Optical power in dBm
Optical power is measured in dBm — decibels relative to 1 milliwatt. It's a logarithmic scale, so:
- 0 dBm = 1 mW. Positive dBm = more than 1 mW. Negative dBm = less than 1 mW (typical for received signals, e.g. −18 dBm).
- Every −3 dB = roughly half the power. Every −10 dB = one tenth.
- Loss is measured in dB (a ratio/difference). Absolute power is in dBm.
Because it's logarithmic, you can just add and subtract dB values along a path — which is exactly what a loss budget does.
dB is a ratio (a difference, no reference) — use it for loss and gain. dBm is an absolute power referenced to 1 mW — use it for what a meter actually reads. Subtract two dBm readings and you get dB: −6 dBm − (−3 dBm) = −3 dB of loss.
| Power (dBm) | Linear power | Anchor to remember |
|---|---|---|
| +10 | 10 mW | +10 dB = ×10 |
| +3 | ≈ 2 mW | +3 dB = ×2 (double) |
| 0 | 1 mW | the reference point |
| −3 | ≈ 0.5 mW | −3 dB = half |
| −10 | 0.1 mW | −10 dB = ÷10 |
| −30 | 1 µW | typical low Rx region |
Two shortcuts cover almost all field mental math: 3 dB ≈ ×/÷2 and 10 dB ≈ ×/÷10. Chain them: −13 dB is roughly one-twentieth of the power (÷10 then ÷2).
Insertion loss, return loss, reflectance
| Term | What it measures | You want it to be… |
|---|---|---|
| Insertion loss | Power lost passing through a component or span (connector, splice, fiber) | Low (small dB) |
| Return loss | How much light is reflected back toward the source, expressed as a positive dB number | High (a big return-loss number = little reflection) |
| Reflectance | The reflection at a single point, expressed as a negative dB number | Very negative (e.g. −55 dB is better than −35 dB) |
Insertion loss is about the light that gets through; return loss / reflectance is about the light that bounces back. High reflections can destabilize lasers and corrupt high-rate signals — which is why APC connectors matter on transport links.
Putting numbers on it
- Insertion loss example: a mated connector pair spec'd at 0.3 dB passes 93% of the light (−0.3 dB ≈ ×0.93). A dirty one measuring 1.5 dB passes only about 71% — you've silently lost a fifth of your budget at one port.
- Return loss example: a clean APC connector at 60 dB RL reflects one part in a million (10⁻⁶) of the light. A degraded connector at 30 dB RL reflects one part in a thousand (10⁻³) — a thousand times more reflection back at the laser.
- Reflectance example: a single event at −55 dB reflectance is better than −35 dB (more negative = less reflected). An open/unmated APC connector still reflects far less than an open UPC or a flat glass-to-air break.
Insertion loss and return loss are related but independent: a connector can pass light well (low IL) yet reflect badly (poor RL), for example a clean-but-flat-polished UPC on a high-rate coherent link. On transport you care about both.
Dirty connectors: the #1 avoidable outage
A single speck of dust on a fiber endface sits right in the light path. On single-mode fiber the core is ~9 µm — a common airborne particle is ~1 µm, so a handful of specks can cover a meaningful fraction of the core. A dirty connector causes high insertion loss, high reflectance, intermittent errors, and it transfers contamination to the mating connector, spreading the problem to good hardware.
Inspect with a scope, grade against IEC 61300-3-35
"Looks clean to the eye" is meaningless at 9 µm. Use a fiber inspection scope (video probe) that images the endface and grades it against the IEC 61300-3-35 standard. That standard divides the endface into concentric zones and sets pass/fail limits for scratches and particles in each:
- Core zone (center, where the light is): the strictest — essentially zero tolerance for any defect on single-mode.
- Cladding zone: a few small defects allowed.
- Adhesive & contact/ferrule zones: progressively looser limits farther from the core.
Connecting a dirty, powered endface. Under the concentrated optical power of an amplified DWDM system, a particle in the core zone can absorb energy, heat up, and permanently burn a pit into the endface — turning a wipe-it-off problem into a replace-the-connector problem, and dropping every channel on that fiber. Never mate under power without inspecting first.
Plugging in without inspecting. The field rule is inspect → clean → inspect → connect — every connector, every time. Use dry cleaning tools (cassette cleaner, click-cleaner) first, then wet-then-dry for stubborn film. Never blow on a connector, wipe it on your shirt, or touch the ferrule with a finger. And always cap unused connectors.
1. You measure −6 dBm at a receiver and expected −3 dBm. Roughly how much power did you lose?
2. A green (APC) jumper won't seat cleanly into a blue (UPC) bulkhead. What's happening?
3. Power looks good on both strands but the link is down. Best first guess?