Satellite Link Budget: Rain, Absorption & Availability

Why Satellite Links Above 10 GHz Are Different

A terrestrial microwave hop between two towers is almost boring once the path loss is known: add a few dB of rain margin and you are done. A satellite downlink above 10 GHz is not. The signal crosses the whole troposphere at a slant, rain attenuation grows steeply with frequency, and the availability target — say 99.9 % of the year, which still allows 8.76 hours of outage — is set by the worst few hours of weather at one particular place.

That last point matters more than any other. Rain, cloud and water vapour are properties of the site, not of the latitude band it sits in. This walkthrough designs a Ka-band direct-to-home downlink for a 60 cm dish in London and shows every number the Satellite Link Budget Analyzer uses along the way.

What the Tool Computes

The analyzer reads the site's climate from the ITU-R digital maps and runs the ITU-R prediction methods on it:

The implementation is a port of the open-source ITU-Rpy library and agrees with it within 0.001 % at twelve test sites. Tropospheric scintillation is not included — P.618 needs the receive antenna's diameter and efficiency for it — and the result says so.

The Example: Ka-band Direct-to-Home Downlink

Three of those numbers are derived, and it is worth doing them once by hand. Slant range. A GEO satellite 35,786 km above the equator is further away than its altitude from anywhere but the sub-satellite point. With Earth radius $R_E$ = 6378 km and orbit radius $r$ = 42,164 km: G/T. A 0.60 m dish at 65 % efficiency has a gain of $10\log_{10}\left(0.65\,(\pi D f/c)^2\right)$ = 40.1 dBi at 20 GHz. With a 150 K system temperature (21.8 dBK), G/T = 18.4 dB/K. Required Eb/N0. ETSI EN 302 307-1 gives an ideal Es/N0 of 7.91 dB for 8PSK 3/4. Each symbol carries 3 × 0.75 = 2.25 information bits, so $E_b/N_0 = 7.91 - 10\log_{10}(2.25) = 4.4$ dB.

The Site's Climate

Before any attenuation is computed, the tool reads the climate at 51.5° N, 0.1° W and lists every value it used:

The same analysis at Singapore reads $R_{0.01}$ = 100 mm/h. At Madrid it reads 25.1 mm/h — almost London's figure, but with a rain height, cloud climate and elevation angle different enough to change the answer, as we will see.

The Budget at 99.9 %

The link does not close at 99.9 %. Rain is the largest atmospheric term, but not by as much as the folklore suggests: gas and cloud add another 1.94 dB between them.

Why are gas and cloud taken at 1 % when the target is 0.1 %? Because P.618-13 section 2.5 says so. Gas and cloud do not grow the way rain does in the rarest fades, so below 1 % exceedance the method holds them at their 1 % values and combines them with the rain attenuation at the target probability:

with $A_S$ = 0 here because scintillation is not modelled.

The Availability Curve

The tool evaluates the whole budget at every availability, not just the target:

This 60 cm terminal closes to about 99.66 % of the year. Going from 99.9 % to 99.99 % costs another 8.1 dB of rain attenuation — each extra nine is far more expensive than the last.

Monte Carlo: Uncertainty Around the Nominal Budget

With a paid account the tool runs 100,000 trials around the nominal budget (10,000 on the free tier). It varies EIRP and G/T (σ = 0.3 dB each), adds a pointing loss (exponential, mean 0.2 dB), scales the rain attenuation by a log-normal rain-rate spread (σ = 0.5) and adds a fading term (|normal|, σ = 0.4 dB). This is an uncertainty model the tool applies around the ITU-R result — it is not itself an ITU-R method.

For the 60 cm design at 99.9 %:

The median sits below the nominal −2.09 dB because pointing and fading only ever subtract. The p5 tail is long because the rain-rate spread multiplies an attenuation that is already 4.2 dB: a rainier-than-predicted year hurts far more than a drier one helps.

Closing the Link

Either a 90 cm dish or dropping to QPSK 3/4 closes the link nominally. Neither makes the p5 margin positive — if you need 95 % of terminals to meet 99.9 % in a wetter-than-average year, you need more still. Adaptive coding and modulation, which DVB-S2 supports, exists for exactly this: run 8PSK in clear sky and fall back to QPSK in a fade.

And 99.99 %? Even the 90 cm dish is 6.68 dB short there, with 12.29 dB of rain.

The Site Matters More Than the Latitude

Run the same 60 cm, 8PSK 3/4 terminal elsewhere (each site at its own GEO elevation):

Madrid and London have almost the same $R_{0.01}$, yet Madrid's higher elevation and drier cloud climate put it 2.15 dB ahead. Singapore, with a satellite nearly overhead, is still 19 dB short: tropical Ka-band links need a different design altogether.

Terrestrial Mode

Switch the link type to terrestrial and the tool stops treating the path as a slant through the rain layer. Rain follows the ITU-R P.530-17 path-reduction method over the path length, and gas is the ground-level specific attenuation times that length. For a 23 GHz, 15 km hop at the same London site:

Gaseous absorption adds 5.1 dB over the 15 km: 23 GHz sits right next to the 22.2 GHz water-vapour line. The terrestrial result also states what it does not include — multipath fading, which P.530 predicts from geoclimatic and terrain data the tool does not have.

What This Means Operationally

For a broadcast service, 99.9 % (8.76 h/yr of outage) is typical and 99.7 % is often accepted at Ka-band. Aviation or financial links need 99.99 % (52.6 min/yr) or better, and in London that means another 8.1 dB of atmospheric loss at 20 GHz beyond the 99.9 % design.

Run the scenario for your own site before you size a dish. A latitude and a rain "zone" are not enough — the climate values the tool lists are the ones that set your outage hours.

Updated 2026-09-19. An earlier version of this article used rain-zone climate, an older P.838 coefficient table and a clear-sky budget that did not match the tool. Every number above now comes from the corrected analyzer. Satellite Link Budget Analyzer