Link budget for a 5 km point-to-point 2.4 GHz link: how do I calculate path loss and fade margin?

Start with free-space path loss: FSPL = 32.44 + 20 × log10(d in km) + 20 × log10(f in MHz). For 5 km and 2400 MHz that is 32.44 + 13.98 + 67.60 = 114.0 dB.

Then add everything in order: 15 dBm - 1 dB + 19 dBi = 33 dBm EIRP, minus 114 dB path loss, plus 19 dBi receive antenna, minus 1 dB cable = -63 dBm at the receiver. Against -80 dBm sensitivity the fade margin is 17 dB.

Line of sight in the optical sense is not enough. The first Fresnel zone should be mostly clear, or the path loss will be higher than the free-space figure. Its radius at the midpoint is sqrt(λ × d / 4) for a path of length d: with λ = 0.125 m and d = 5000 m that is sqrt(156) = 12.5 m. The usual target is at least 60 percent clearance, 7.5 m, above roofs, trees and the ground along the whole path.

Earth curvature is small at this distance, about 0.4 m of bulge at mid-path with standard refraction, but it counts in the same clearance sum.

On the margin: 17 dB is workable for a short fixed link with a clear Fresnel zone, and 10 dB would be the point where I would start to worry. Margin has to cover antenna misalignment, connector ageing, wet foliage, interference raising the noise floor, and multipath fading. Rain attenuation is negligible at 2.4 GHz; it becomes a design factor above roughly 10 GHz.

Remember that sensitivity depends on the data rate. The same radio may need -70 dBm for its fastest modulation, which leaves only 7 dB, so expect the link to drop to a lower rate during fades.

Check the EIRP against the local regulations before buying antennas. Licence-exempt bands have a limit on radiated power, and in many countries it is expressed as EIRP, so a higher-gain antenna means the transmitter power must be turned down by the same amount. The 33 dBm in this budget is not permitted everywhere. The limits differ by country and by band, so look up the current rules of your regulator.

High-gain antennas are still worth having under an EIRP cap, because the gain at the receive end is not limited and the narrow beam rejects interference.

A sanity check on the noise side, since the 2.4 GHz band is busy. Thermal noise is -174 dBm/Hz at room temperature; in a 20 MHz channel that is -174 + 73 = -101 dBm, and with a 6 dB receiver noise figure the floor is about -95 dBm. A -63 dBm signal would be 32 dB above that in a quiet band.

In practice other networks raise the floor, and it is the signal-to-interference ratio that sets the data rate, so do a scan at both ends on the intended channel before committing to it.