Link Budget
One column of additions in decibels that decides whether a radio link works. Transmit power plus gains minus losses gives the level at the receiver; compare it to what the receiver needs and the difference is the margin.
A link budget is the arithmetic that decides whether a radio link works. It sums, in decibels, everything that adds to or subtracts from the signal between the transmitter and the receiver, compares the result to the weakest signal the receiver can actually use, and produces one number: the margin. Most failed captures are not bad luck, they are a negative margin nobody computed.
What it is
Because decibels turn multiplication into addition, the whole radio path is a single column of sums:
EIRP = TX power + TX antenna gain - TX feed loss
RX level = EIRP - path loss + RX antenna gain - RX feed loss
sensitivity = -174 dBm/Hz + 10 log10(bandwidth) + noise figure + required SNR
margin = RX level - sensitivity
EIRP (Effective Isotropic Radiated Power) is what the transmitter effectively radiates in its best direction, and it is also the quantity regulators cap. Path loss is almost always the dominant term.
Why it matters
Every practical question in wireless work is a link budget in disguise. How far will this 868 MHz sensor reach? Why does the same dongle hear an aircraft 300 km away but not a doorbell across the garden? Is my capture failing because of the antenna, the coax, or the noise floor? Will an LNA help? The budget answers all of them with the same five lines, and it tells you which term to spend money on, which is usually not the one people buy first.
Free-space path loss
The dominant loss is simply the signal spreading over a growing sphere:
FSPL (dB) = 20 log10(d in km) + 20 log10(f in MHz) + 32.44
Two consequences are worth memorising: doubling the distance costs 6 dB, and doubling the frequency costs 6 dB for the same antenna gains. FSPL is the best case, a vacuum with nothing in the way. Real links add obstruction, ground reflection, foliage, and fading on top, which is why a design keeps a fade margin rather than aiming for exactly 0 dB.
Noise, sensitivity and margin
A receiver is not limited by how small a signal is but by how far it sits above the noise. Thermal noise is -174 dBm in every hertz at room temperature, so widening the bandwidth raises the floor by 10 log10(B): doubling the bandwidth costs 3 dB. The receiver adds its own noise, and the noise figure is exactly how much. Add the SNR the demodulator needs and you have sensitivity. Halving the bandwidth to match the signal buys 3 dB for free, and an LNA at the antenna beats the same LNA at the radio because the first stage sets the noise figure of the whole chain.
Pitfalls
- Mixing units. Every term must be dB, and the two endpoints dBm. A gain in "times" or a distance in metres inside a formula expecting kilometres is the classic silent error.
- Forgetting the feed. Thin coax at high frequency is expensive: RG-174 costs roughly 0.8 dB per metre at 868 MHz and about 1.55 dB per metre at 2.4 GHz. Three metres of the wrong cable can swallow the antenna you just bought.
- Quoting a sensitivity without a bandwidth. A "-110 dBm" receiver figure is meaningless until you know the bandwidth and the SNR it assumes.
- Believing FSPL indoors. Add 10 to 20 dB for a cluttered outdoor link and 20 to 30 dB for anything indoors or mobile, otherwise the budget is optimistic fiction.