Optical Power Budget: How to Calculate a Fiber Link

A 10 km transceiver fails on a 7 km path. The datasheet said 10 km, the map said 7 km, so why? Because "10 km" is a marketing number derived from a power budget, and your real path includes patch panels, splices and bends the map never mentions. The optical loss budget is the arithmetic that closes this gap — calculated on paper, before cable is pulled, and re-checked against measured receive power afterwards.

Why Datasheet Reach Is Not Distance

Every transceiver class carries a name that looks like a distance — 10GBASE-SR, 1000BASE-LX, 10GBASE-ER, 40GBASE-FR — and every one of those names is really a power budget. The standard that defines the class (an IEEE 802.3 clause, or an MSA such as 100G CWDM4) specifies a minimum launch power and a minimum receiver sensitivity, subtracts them, and divides by an assumed per-kilometre attenuation to arrive at a "reach". The assumption is the weak part: it usually includes a fixed number of connectors and splices, almost never includes a splitter or a WDM mux, and says nothing at all about your patch panels.

That is how the same 10 km optic is comfortable at 6 km and fails at 7 km. The reach figure is a design aid, not a guarantee. The only defensible method is to compute the budget of the real path and compare it with the optic you intend to buy, before the cable is ordered.

Note the pattern: throughput rises, budget falls. Four parallel 25G lanes must share a link that a single 10G lane handled comfortably, and forward error correction exists precisely because the optical margin got thinner. The faster the link, the more the arithmetic matters.

A second trap is the difference between minimum and typical values. A datasheet quotes transmit power as a range (roughly -8.2 to +0.5 dBm for 10GBASE-LR) and sensitivity as a worst case. A budget built from typical values can be 3 dB optimistic, and 3 dB is the entire engineering margin.

The Two Numbers That Define Everything

Power budget (dB) = Tx power (min) - Rx sensitivity (min)
Total link loss (dB) = fiber + connectors + splices + splitters + WDM modules
Link is valid when:  Power budget >= Total link loss + margin

Use minimum transmit power and minimum sensitivity, not typical values. Optical components vary, and the worst-case pair is what determines whether a link stays up in a hot cabinet in August. A 3 dB engineering margin is the usual floor; 5 dB is comfortable for long-lived external plant.

Typical Loss Values per Element

Wavelength, Attenuation and Where dB/km Comes From

Single-mode attenuation is not constant across the spectrum. The 1310 nm window (O-band) sits near a zero-dispersion point but has higher attenuation; the 1550 nm window (C-band) has the lowest loss, which is why long-haul and DWDM systems live there. Between roughly 1380 and 1440 nm the hydroxyl (OH-) peak from water absorbed in the glass raises loss sharply — the "water peak". Low-water-peak fibre (G.652.D) suppresses it and opens the 1383 nm region to CWDM channels, but fibre already sitting in your ducts may be twenty-year-old G.652.B where that window is unusable.

Use the manufacturer's measured figure for installed plant, never the catalogue number. An OTDR trace gives per-kilometre attenuation for the actual cable, including the section that was spliced badly a decade ago and the one with a tight bend behind a cabinet door. If no trace exists, budget 0.4 dB/km at 1310 nm and 0.3 dB/km at 1550 nm as a pessimistic placeholder until you measure.

Insertion Loss and Return Loss Are Different Problems

The budget arithmetic only covers insertion loss — how much power a component removes from the forward path. Return loss is how much power reflects back toward the transmitter, and it never appears in the sum. It shows up instead as a link that calculates perfectly clean but produces CRC errors, intermittent flaps, or in the worst case a damaged laser in a DWDM amplifier chain.

• Mated connector pairs at 0.3 dB insertion loss should still have return loss better than 45 dB (UPC) or 60 dB (APC). A dirty endface or scratched ferrule ruins return loss long before insertion loss moves measurably.
• APC/UPC mismatch is a classic field error: an angled APC connector (green) mated to a flat UPC surface (blue) gives roughly 14 dB return loss and several dB of insertion loss. Always check colour codes at the patch panel.
• Macrobends from cable pulled too tight around a corner or left unsecured in a tray leak power and reflect it. They are invisible on a budget sheet and unmissable on an OTDR trace.
• Dirty connectors are the most common cause of an unexplained 2 dB loss where the sheet predicted 0.3 dB. Clean and re-measure before replacing anything.

Worked Example 1: 500 m OM4 Link in a Plant

Fiber:    0.5 km x 3.0 dB/km      = 1.50 dB
Connectors: 4 mated pairs x 0.3 dB = 1.20 dB   (2 patch panels + 2 equipment ports)
Splices:    1 fusion x 0.1 dB      = 0.10 dB
Total link loss                   = 2.80 dB
Budget of a typical 10GBASE-SR pair (approx 7.5 dB) - 2.80 dB => >4 dB margin. Valid.

Worked Example 2: 60 km DWDM Metro Span

Optics: 10G SFP+ ZR, power budget 23 dB, 1550 nm
Fiber:     60 km x 0.25 dB/km     = 15.00 dB
Splices:   4 x 0.1 dB             =  0.40 dB
Connectors: 6 pairs x 0.3 dB      =  1.80 dB
Mux + demux pair (16 ch, 100 GHz) =  7.50 dB
Total link loss                   = 24.70 dB
Budget 23 dB < 24.70 dB  =>  LINK WILL NOT COME UP without amplification.

Fixes: place an EDFA (or two) on the span, move to a coherent 100G optic with a
larger budget, or remove patch-panel pairs from the path.

This is exactly the failure the datasheet "10 km / 40 km / 80 km" labels hide: the rated distance assumes a clean path with minimal connectors and no mux/demux in line.

GPON Example: When Splitter Loss Dominates

10 km, 1:32 split, 4 splices, 2 connector pairs
Fiber:     10 x 0.35      = 3.50 dB
Spices:    4 x 0.1        = 0.40 dB
Connectors: 2 x 0.3       = 0.60 dB  (some designs use 0.5 dB/pair)
Splitter:  1:32           = 17.50 dB
Total                     = 22.00 dB
Power budget 28 dB => 6 dB margin  (comfortable)
Power budget 23 dB => 1 dB margin  (tight: any added loss breaks it)

A Repeatable Design Worksheet

Budgets go wrong when they are done in a spreadsheet once and never revisited. Keep the calculation as a small table with one row per element and one column per value, and store it with the link documentation.

Element              Qty     Unit loss     Subtotal
Fiber, 1310 nm       7.0 km  0.35 dB/km     2.45 dB
Fusion splices          6    0.10 dB        0.60 dB
Connector pairs         4    0.30 dB        1.20 dB
WDM mux/demux pair      1    7.50 dB        7.50 dB
-----------------------------------------------
Total link loss                              11.75 dB

Optic budget (10GBASE-ER)                    19.00 dB
Engineering margin       19.00 - 11.75   =   7.25 dB   PASS
Minimum acceptable margin (3 dB)                        PASS

After any change, recompute. Adding one patch-panel pair plus
a mux/demux pair consumes 7.8 dB of the 7.25 dB margin.

Two habits make the worksheet trustworthy. First, count every mated pair, including the ones inside patch panels and at equipment ports — designers routinely forget that a patch panel is two connectors, not one. Second, record the wavelength each element was budgeted at; a fibre figure measured at 1550 nm is not valid for a 1310 nm optic.

Designing the Margin, Not Just the Link

The 3 dB floor is not arbitrary: it absorbs connector cleaning variance, temperature-driven laser drift, transmitter ageing, and the difference between the datasheet and the shipped part. For external plant with splices in the ground, a 5 dB target is more appropriate, because every repair splice added over the life of the cable costs 0.1-0.3 dB and access to the plant is expensive.

Plan for growth as well. If a link has 2 dB of margin today and you intend to add a CWDM module next year, budget the module now — a 4-channel CWDM mux/demux pair costs about 3 dB, which already exceeds that margin. The cheapest time to place an amplifier or choose a higher-budget optic is during the initial design, not after the link has been commissioned and documented.

Verify Against Reality

1. Read the measured receive power on both ends (show interfaces … transceiver / show interfaces optics) and compare to the RX sensitivity in the datasheet — not to the "typical" value.
2. Use an OTDR to locate unexpected splices, macrobends or dirty connectors; a connector that measures 2 dB instead of 0.3 dB is usually dirt or a damaged ferrule, not cable loss.
3. Match polish types: an APC/UPC mismatch creates high reflectance, which shows up as intermittent CRC/input errors rather than a dead link.
4. Recheck the budget after any topology change — adding one patch panel pair plus a WDM module can consume the whole margin.
5. Record the calculated budget and the measured power in documentation; the next engineer comparing the two will immediately see degradation.

Frequently Missed Loss Elements

When a link comes up with less margin than designed, walk this list before blaming the optic. In practice the discrepancy is almost always a connector count error, a dirty endface, or an undocumented attenuator installed during an earlier troubleshooting session.

Finally, keep the design target and the measured result side by side in the link record: budgeted 11.75 dB, measured 12.4 dB on commissioning. The next engineer comparing those two numbers six months later has an instant baseline for "is this link degrading" — which is the whole point of doing the arithmetic in the first place.

相关阅读：800G ZR / ZR+ 相干可插拔光模块指南、H3C 光模块与链路维护指南 以及 相干光模块：Pre-FEC BER、OSNR 与 DSP 检查。

原文链接：Fiber Optic Link Loss Budget: How to Calculate Before You Build