Satellite Link Budget: Rain, Absorption & Availability

Why Satellite Links Are Different

Here's something most people don't realize until they've actually deployed a satellite system: a terrestrial microwave link between two fixed towers is almost boring once you've calculated the path loss. Add a few dB of rain margin, maybe pad it a bit if you're cautious, and you're done. Satellite links above 10 GHz? Completely different beast.

Rain attenuation at 20 GHz can slam you with more than 20 dB of loss during a tropical downpour — not a sprinkle, but a real storm. Meanwhile, gaseous absorption from oxygen and water vapor is sitting there quietly adding 0.5–3 dB depending on your elevation angle. Cloud liquid water throws in another 1–2 dB when you're looking up at high angles. And through all of this, your system needs to hit a specified availability target — say 99.9% of the year, which sounds great until you realize that's still 8.76 hours of outage annually that you have to account for.

ITU-R has published a suite of propagation models that let you translate rain rate statistics into attenuation exceedance probabilities with actual physical meaning. The Satellite Link Budget Analyzer implements these directly: P.618-13 for rain and scintillation, P.676-13 for gaseous absorption, and P.840-8 for cloud attenuation. No external library dependencies, no black-box calculations. The tool couples these models with a Monte Carlo simulation that varies rain rate, pointing loss, EIRP, and G/T to generate annual availability curves that actually reflect the messy reality of deployed systems.

The Example: Ka-band Direct Broadcast Downlink

Let's work through a real scenario that'll make the differences clear. We're looking at a Ka-band direct broadcast satellite downlink — specifically the 19.7–20.2 GHz band — feeding a 60 cm consumer dish somewhere in a temperate maritime climate. That puts us in ITU-R rain zone K, where the rain rate exceeded 0.01% of the year (R₀.₀₁) is 30 mm/hr. Not the worst zone by any means, but not benign either.

Here are the link parameters we're working with:

That Eb/N0 requirement comes from DVB-S2 8PSK 3/4 coding — a reasonably aggressive modulation scheme that squeezes decent spectral efficiency out of the available bandwidth. You could back off to QPSK 1/2 for more robustness, but you'd sacrifice throughput.

Computing the Nominal Clear-Sky Budget

Start with the basics. Free-space path loss at 20 GHz over a GEO distance is substantial:

That's a big number, but it's deterministic. The geometry doesn't change (much — we'll get to that).

The tool computes received carrier-to-noise density ratio from first principles:

where k = −228.6 dBW/K/Hz is Boltzmann's constant. Now here's where the atmospheric physics comes in. At 35° elevation angle, the P.676 gaseous absorption model gives approximately 0.8 dB of combined oxygen and water vapor absorption. This number varies significantly with surface humidity and temperature — the tool uses the ITU-R standard reference atmosphere, but in a real deployment you'd want to check local conditions. The P.840 cloud model adds another 0.3 dB assuming a liquid water path of 10 g/m², which is typical for non-raining clouds at mid-latitudes.

Plug in the numbers: clear-sky C/N0 = 52 − 209.5 − 0.8 − 0.3 + 12.8 + 228.6 = 82.8 dBHz.

With a 45 Msps symbol rate, the noise bandwidth is 10·log₁₀(45×10⁶) = 76.5 dBHz. So our Eb/N0 = 82.8 − 76.5 = 6.3 dB. Wait — we need 7.2 dB for the DVB-S2 decoder to lock reliably. We're short by 0.9 dB even in clear sky.

Actually, let me recalculate that more carefully with the exact G/T. With a 0.60 m dish at 65% efficiency, the receive gain is about 10·log₁₀(η·(π·D·f/c)²) = 10·log₁₀(0.65·(π·0.6·20×10⁹/3×10⁸)²) ≈ 41.6 dBi. System temperature is 150 K (21.8 dBK), so G/T = 41.6 − 21.8 = 19.8 dB/K. Let me correct that table value — the 12.8 dB/K was too pessimistic.

With the corrected G/T: C/N0 = 52 − 209.5 − 0.8 − 0.3 + 19.8 + 228.6 = 89.8 dBHz. Eb/N0 = 89.8 − 76.5 = 13.3 dB. Now we have 13.3 − 7.2 = 6.1 dB of clear-sky margin. Much better. But hold on — all of that margin is going to get eaten by rain.

ITU-R P.618 Rain Attenuation

The P.618-13 rain attenuation model is where things get interesting. It computes the attenuation exceeded for p% of the year based on your rain zone and geometry. The calculation sequence:

1. Specific rain attenuation: γ_R = k × R₀.₀₁^α. At 20 GHz horizontal polarization, the P.838 coefficients are k ≈ 0.0751 and α ≈ 1.099. With R₀.₀₁ = 30 mm/hr, γ_R = 0.0751 × 30^1.099 ≈ 2.85 dB/km.

1. Effective slant path through rain: L_S = (h_R − h_S)/sin(θ), where h_R is the rain height (about 3.5 km at mid-latitudes from the 0°C isotherm data), h_S is station height (assume sea level), and θ = 35° elevation. So L_S = 3.5/sin(35°) ≈ 6.1 km.

1. Horizontal reduction factor: The actual path through rain is shorter than the geometric slant path because rain cells have finite horizontal extent. P.618 applies a reduction factor r₀.₀₁ that depends on latitude and frequency. At 45°N and 20 GHz, r₀.₀₁ ≈ 0.36.

1. Attenuation exceeded 0.01% of the year: A₀.₀₁ = γ_R × L_S × r₀.₀₁ = 2.85 × 6.1 × 0.36 ≈ 6.3 dB. This is the fade depth exceeded about 52 minutes per year.

1. Scale to other percentages: P.618 Equation 6 provides a power-law scaling. For 0.1% of the year (99.9% availability), the attenuation is roughly A₀.₁ ≈ A₀.₀₁ × 0.12 ≈ 0.76 dB. Wait, that doesn't match what I said earlier. Let me recalculate using the exact P.618 formula, which is more complex than a simple power law and includes latitude and frequency dependence.

Actually, the P.618 model uses: A_p = A₀.₀₁ × (p/0.01)^[−(0.655 + 0.033·ln(p) − 0.045·ln(A₀.₀₁) − β·(1−p)·sin(θ))] where β is a function of latitude. For p = 0.1% and our geometry, this gives A₀.₁ ≈ 3.2 dB.

So with 6.1 dB of clear-sky margin and 3.2 dB of rain attenuation at 99.9% availability, we have 6.1 − 3.2 = 2.9 dB of residual margin. That's cutting it close, but it technically closes the link.

Monte Carlo: Availability Curves With Uncertainty

Here's the problem with the nominal calculation: it assumes everything sits exactly at its design center value. In the real world, satellite EIRP drifts ±1 dB over the spacecraft lifetime — you're at beam center when the satellite is fresh, but as it ages and the transponder degrades, you might be down 0.8 dB. Pointing loss varies ±0.5 dB due to wind loading on the dish, thermal expansion of the mount, even the weight of ice accumulation in winter. And the ITU-R rain zone boundaries? Those are statistical fits to sparse rain gauge data. Your actual location might be 20% wetter than the zone average.

Run a Monte Carlo simulation with 100,000 trials, varying EIRP (±1 dB uniform), pointing loss (0 to 1 dB), G/T (±0.5 dB), and rain rate (±20% log-normal). The availability curve output shows median, 10th percentile, and 90th percentile annual availability as a function of added fade margin:

Look at that 10th percentile column. To guarantee 99.9% availability (your spec) at the 10th percentile of system performance — meaning 90% of your deployed terminals meet spec — you need 3 dB of additional fade margin beyond the nominal calculation.

How do you get 3 dB? Upsize the dish from 60 cm to about 75 cm (that's a 3 dB gain increase from the larger aperture). Or run the satellite transponder at higher power, if you have the DC power budget and thermal margin. Or switch to a more robust modulation — QPSK 1/2 instead of 8PSK 3/4 — but you'll cut your data rate nearly in half.

Most operators underestimate this. They design to the median case and then field angry calls when 10% of their customer base experiences dropouts during storms. The Monte Carlo tells you what margin you actually need to sleep at night.

Terrestrial vs. Satellite Mode

Switch the link type to "terrestrial" in the tool and you're modeling a fixed point-to-point microwave link using the same ITU-R rain model, but now it's a single-layer rain cell rather than a slant path through the atmosphere. The P.838 specific attenuation coefficients are identical; the difference is that the path length through rain is just your link distance rather than being computed from orbital geometry and rain height.

This mode is useful when you're comparing a satellite path to an alternative terrestrial backhaul route. Say you're trying to decide between a Ka-band satellite hop and a 23 GHz terrestrial link across 15 km. Same rain zone, same frequency band (roughly), but very different path geometries. The terrestrial link might actually have worse rain fade because the entire 15 km path can be immersed in rain, whereas the satellite slant path only intersects 6 km of rain height at 35° elevation.

What the Numbers Mean Operationally

For a commercial broadcast operator, 99.9% annual availability translates to 8.76 hours of outage per year. That's acceptable for entertainment services — nobody's going to sue you because they missed half a football game during a thunderstorm.

For aviation safety communications or financial trading links, you need 99.99% (52 minutes per year) or even 99.999% (5.2 minutes per year). Each additional "nine" costs you roughly 3–4 dB of margin, which translates directly into satellite power, antenna size, or both. A 99.999% link might need a 2-meter dish where a 99.9% link could get by with 60 cm.

The Monte Carlo output gives you the margin required not just for a single nominal system sitting in perfect conditions, but across your entire fleet of deployed terminals and over the 15-year orbital life of the satellite. This is the difference between a paper link budget that looks great in PowerPoint and a deployment confidence interval that actually predicts field performance. Most engineers skip this step and regret it later when they're buying larger dishes to retrofit angry customers.

Satellite Link Budget Analyzer