Light-loss budgets

THE SHORT VERSION

A transmitter puts a certain amount of light into the fiber. A receiver needs a certain minimum to read the signal. Everything in between — glass, connectors, splices, splitters — takes a cut. The budget is the difference between what you start with and what you need; the link works if everything's cuts added up fit inside it, with some left over for a bad day.

DB FROM ZERO

Light power spans a huge range, so the trade counts it on a logarithmic scale. Two units do all the work, and mixing them up is the classic mistake:

• dB is a ratio — how much bigger or smaller one power is than another: dB = 10 × log₁₀(P₂ ÷ P₁). Losses and gains are in dB.
• dBm is an absolute power, measured against 1 milliwatt: 0 dBm = 1 mW. Transmitter output and receiver readings are in dBm.

Landmarks worth memorizing: every −3 dB halves the power, every −10 dB divides it by ten. So −10 dBm = 0.1 mW, −20 dBm = 0.01 mW, and −3 dBm is about 0.5 mW.

Why losses simply add: a fiber passes some fraction of light, a connector passes a fraction of that, and so on — real power multiplies through the fractions. Logarithms turn multiplication into addition, so in dB you just add the losses up. The unit arithmetic follows: dBm − dB = dBm (a power minus a loss is a power), and dBm − dBm = dB (the gap between two powers is a ratio).

THE BUDGET EQUATION

available budget = transmitter power (dBm) − receiver sensitivity (dBm)

link loss = (fiber km × dB/km)
          + (connector pairs × dB each)
          + (splices × dB each)
          + splitter loss

margin = available budget − link loss

Use the transmitter's minimum rated output and the receiver's sensitivity from the datasheets — worst case on both ends, so the answer holds for any unit off the shelf. The margin is your allowance for aging, repairs, temperature, and the splice someone adds in ten years. How much to keep is a design choice; about 3 dB is a common rule of thumb, not a standard.

TYPICAL REFERENCE VALUES

Splitters have a floor set by plain arithmetic: split the light N ways and each leg gets 1/N of it, a loss of 10 × log₁₀(N) dB. Real splitters lose a bit more (excess loss, plus unevenness between legs), so always budget the datasheet value, never the ideal:

The ~2 dB gap between the 1×32 ideal (15.1 dB) and the typical datasheet figure (≈ 17 dB) is real-world excess loss — exactly the kind of detail a budget built on ideal math gets wrong.

WORKED EXAMPLE — A PON LINK, END TO END

A passive optical network: one OLT port, one 1×32 splitter in the field, and a customer's ONT. The optics values are illustrative — in the range of common GPON optics, but use your own datasheets in real life:

• Fiber: 12 km, budgeted at 0.35 dB/km (the 1310 nm planning value used for every wavelength — a conservative simplification)
• Fusion splices: 6, at 0.1 dB each (typical)
• Mated connector pairs: 4, at the 0.75 dB TIA allowance each (worst case)
• OLT transmitter, minimum output: +1.5 dBm (illustrative)
• ONT receiver sensitivity: −27.0 dBm (illustrative)
• Splitter: 1×32, at 17.0 dB (typical datasheet figure)

available budget = +1.5 − (−27.0)          = 28.5 dB

fiber            = 12 km × 0.35 dB/km      =  4.20 dB
connector pairs  = 4 × 0.75 dB             =  3.00 dB
fusion splices   = 6 × 0.1 dB              =  0.60 dB
splitter (1×32)                            = 17.00 dB
                                             --------
link loss                                  = 24.80 dB

margin           = 28.5 − 24.80            =  3.70 dB
expected at ONT  = +1.5 dBm − 24.80 dB     = −23.30 dBm

The link closes with 3.7 dB to spare, and a meter at the ONT should read near −23.3 dBm on a worst-case transmitter. Look at where the budget went: the splitter alone is more than two-thirds of the loss. And the connectors, budgeted at their allowance, cost five times what the splices do — which is why clean, good connections matter so much.

One more check worth knowing: receivers also have a maximum (overload) input. A very short link with no splitter can deliver too much light, and the cure is an attenuator, not more fiber.

WHAT EATS BUDGETS IN THE REAL WORLD

• Dirty endfaces. The core of a singlemode fiber is only about 8 to 9 microns across; a speck of dust you can't see covers a lot of it. One contaminated connector can cost more than every splice in the example combined, and mating a dirty connector can transfer the dirt and scratch both faces. Inspect, clean, inspect again — every time, both sides.
• Macrobends. A fiber pulled too tight around a corner, a clip, or a slack loop leaks light. Bends hurt longer wavelengths more, so a link can pass at one wavelength and fail at another — the OTDR note shows how to catch it.
• One bad splice. Budgeted at 0.1 dB, a poor one can measure many times that. A budget with 3.7 dB of margin can absorb one surprise, not a handful.

SOURCES AND LIMITS

The connector allowance is from the TIA-568 family. Splice, fiber, and splitter figures are typical values; the optics in the example are illustrative. Datasheets for the actual parts outrank every number here. To practice this with real numbers, Light Show is built around exactly this math.

← All field notes