Fiber loss per km, and how much loss is acceptable
How much light fiber loses per kilometer, the TIA-568 maximums you design to, and how to turn them into the number a link should test under.
Attenuation by fiber type
| Fiber | Wavelength | TIA-568 max | Typical new cable |
|---|---|---|---|
| Single-mode, outside plant | 1310 nm | 0.50 | 0.30–0.35 |
| Single-mode, outside plant | 1550 nm | 0.50 | 0.20–0.25 |
| Single-mode, inside plant | 1310 / 1550 nm | 1.0 | — |
| Multimode OM3, OM4, OM5 | 850 nm | 3.0 | — |
| Multimode OM1, OM2 | 850 nm | 3.5 | — |
| Multimode, all grades | 1300 nm | 1.5 | — |
The TIA-568.3 maximums are what a cable must meet; they are the right numbers for a spec or a worst-case design. The typical column is what modern outside plant single-mode cable tends to measure, useful for predicting a test result; for the other types, take the typical figure from the cable datasheet. For reference, Corning lists its SMF-28 Ultra fiber at 0.32 dB/km maximum at 1310 nm and 0.18 at 1550 nm before cabling. Your cable's datasheet beats both columns.
To convert, multiply dB/km by 1.609 for dB per mile, or by 0.3048 for dB per 1,000 ft.
Fiber loss by length
| Length | 1310 typ. 0.35 | 1550 typ. 0.25 | TIA max 0.50 |
|---|---|---|---|
| 1 km (0.6 mi) | 0.35 | 0.25 | 0.50 |
| 2 km (1.2 mi) | 0.70 | 0.50 | 1.00 |
| 5 km (3.1 mi) | 1.75 | 1.25 | 2.50 |
| 10 km (6.2 mi) | 3.50 | 2.50 | 5.00 |
| 20 km (12.4 mi) | 7.00 | 5.00 | 10.00 |
| 40 km (24.9 mi) | 14.00 | 10.00 | 20.00 |
How much dB loss is acceptable?
There is no single number, because acceptable loss depends on the link. Work it out from its parts, the same way the loss budget calculator does:
Maximum acceptable loss = length × fiber max + connector pairs × 0.75 + splices × 0.3
Worked example
A 5 km single-mode outside plant link with a patch panel at each end (two mated pairs) and four splices. Under TIA-568 maximums: 5 × 0.5 + 2 × 0.75 + 4 × 0.3 = 2.5 + 1.5 + 1.2 = 5.2 dB. That is the pass/fail line for an acceptance test.
With typical values (0.35 dB/km, 0.3 per connector pair, 0.05 per splice) the same link should read about 1.75 + 0.6 + 0.2 = 2.6 dB at 1310 nm. If it tests at 4.8 dB it passes, but something is costing twice what it should, and a dirty connector or a bend is worth looking for before you leave.
The other limit is the equipment: the link must also fit inside the optics' power budget with some margin. On PON networks the splitter dominates; see the splitter loss calculator.
Why loss changes with wavelength
Single-mode fiber loses less at 1550 nm than at 1310 nm, because scattering in the glass falls with longer wavelengths. But longer wavelengths leak out of bends more easily, so a link that tests well at 1310 nm and poorly at 1550 nm (or 1625 nm) usually has a macrobend: a tight loop in a tray, a crushed spot, or a cable pulled around a sharp corner.