Aerial free-space optical (FSO) links — ground-to-HAPS, ground-to-UAV, and platform-to-platform — promise fibre-class data rates without spectrum licensing. But an aerial optical link lives or dies on three impairments that a spreadsheet cannot capture together: atmospheric turbulence, weather-driven attenuation, and pointing error between two platforms that are never perfectly still. Predicting link performance means combining all three, statistically, over a realistic geometry. This article walks through how that analysis is actually done, and how SCOPT's FSO Aerial Link Analyzer models each stage.
Why aerial FSO is a harder problem than terrestrial FSO
A rooftop-to-rooftop FSO link sees a fixed geometry and a roughly horizontal, uniform atmosphere. An aerial link breaks both assumptions:
- The path is slanted and layered. A ground station shooting up to a High-Altitude Platform Station (HAPS) at 18–20 km crosses a rapidly thinning atmosphere. The turbulence strength Cn2 and the air density both change by orders of magnitude along the path, so the channel must be integrated layer by layer, not treated as uniform.
- The platforms move. A hovering UAV drifts; a HAPS oscillates in the stratospheric wind. That converts into a time-varying misalignment between transmit beam and receive aperture — pointing error — which is often the dominant fade mechanism at high SNR.
- Weather sits in discrete layers. A slant path may pass through fog near the ground, a clear-air remainder, then a cloud or aerosol layer. Attenuation is not a single number; it accumulates in decibels across each layer.
The channel model, one stage at a time
1. Turbulence — from Rytov variance to fading statistics
The starting point is the Rytov variance, which quantifies the accumulated turbulence-induced scintillation along the slant path as a function of wavelength, zenith angle, and the altitude profile of Cn2 (for example a Hufnagel–Valley profile). The Rytov variance then sets the parameters of a statistical fading distribution for the received irradiance. Common choices are:
- Gamma–Gamma — the workhorse for moderate-to-strong turbulence, with its α and β parameters derived from the large- and small-scale scintillation.
- Málaga (ℳ) — a more general distribution that contains Gamma–Gamma and several others as special cases, useful when a bright deterministic component is present.
- Weibull and log-normal — for weak turbulence or specific measured channels.
Choosing the wrong distribution, or the wrong turbulence regime, is one of the most common ways an aerial-link budget ends up optimistic.
2. Attenuation — layered Beer–Lambert
Molecular and aerosol absorption/scattering are handled with a Beer–Lambert law, but the key for aerial links is that it is applied per layer and summed in dB. The extinction coefficient in each layer comes from a visibility model — typically Kim or Kruse — with the wavelength exponent q(V) that governs how strongly attenuation depends on wavelength in fog and haze. Distinct presets for fog, water cloud, cirrus and even volcanic-aerosol layers let you assemble a realistic column rather than assuming clear air.
3. Pointing error — the platform-stability penalty
When the transmit beam of half-width wb and the receive aperture of radius ra are misaligned by a random radial displacement, the fraction of collected power drops. That displacement is described by a statistical distribution — Rayleigh for symmetric jitter, or Hoyt and Beckmann when the horizontal and vertical jitter differ or a boresight offset exists. The resulting model is parameterised by the fraction of power collected at perfect alignment (A0) and an equivalent beamwidth (ηs). For a stable link this penalty is small; for a drifting UAV it can dominate the fade budget.
The four metrics that decide the link
With the composite channel in hand — turbulence × attenuation × pointing — the design questions reduce to four performance metrics, each computed by Monte Carlo over the channel statistics:
- Average bit-error rate (BER) across common modulation schemes (OOK, BPSK, M-PSK, M-QAM), for both intensity-modulation/direct-detection and coherent receivers.
- Ergodic capacity — the Shannon channel capacity averaged over the fading, i.e. the throughput ceiling the link can sustain.
- Outage probability — the probability that the instantaneous SNR falls below a threshold, which is the metric that maps directly to an availability specification.
- Link budget — received power, geometric and atmospheric path loss, and the remaining margin.
Why Monte Carlo, and why validation matters
Closed-form expressions exist for some of these metrics under simplifying assumptions, but a realistic aerial link — layered attenuation, a specific pointing distribution, a chosen modulation — usually has no clean analytical BER or capacity. A Monte Carlo approach samples the composite channel directly, which is why the FSO Aerial Link Analyzer is built around it. The trade-off is convergence: too few samples and the outage estimate is noisy; enough samples and it settles to a stable value.
Because a simulation is only as trustworthy as its validation, the analyzer's models are cross-checked two ways. First, every analytically tractable sub-model — Rytov variance versus HAP altitude, the Kim/Kruse attenuation identities, the pointing-error parameters A0 and ηs — is compared against its closed-form expression, agreeing to machine precision. Second, full end-to-end outage and BER curves are reproduced against results published in the peer-reviewed literature on HAPS and hovering-UAV FSO systems. That is the difference between a plausible plot and an engineering result you can put in a design review.
Where this fits in a project
The practical value of pulling turbulence, attenuation and pointing into one tool is that you can answer real design questions before committing to hardware: Will a 1550 nm ground-to-HAPS link close at 20 km under moderate turbulence and thin cirrus? How much does the outage probability improve if aperture diameter is doubled? Which modulation gives the best capacity at the available margin? Is the pointing budget realistic for this platform's stability? Getting those answers on a laptop, before procurement, is exactly where link-analysis software earns its place.
Frequently asked questions
What turbulence models does aerial FSO analysis need?
At minimum, a Rytov-variance calculation over a slant path with an altitude-dependent Cn2 profile, feeding a fading distribution appropriate to the regime — Gamma–Gamma or Málaga for moderate-to-strong turbulence, log-normal or Weibull for weak turbulence.
Why is pointing error so important for HAPS and UAV links?
Because the platforms move. Unlike a fixed terrestrial link, an aerial platform oscillates or drifts, so the beam–aperture misalignment is a time-varying random process. At high SNR, where turbulence fades are less limiting, residual pointing error frequently becomes the dominant cause of outage.
How is attenuation handled on a slant path?
As a layered Beer–Lambert column: each atmospheric layer (fog, clear air, cloud, aerosol) contributes an extinction in dB from a visibility model such as Kim or Kruse, and the layers are summed. This is more realistic than applying a single horizontal-path attenuation figure to a vertical link.
Can these links be analysed without measured field data?
Yes — that is the point of a physics-based Monte Carlo tool. Given a geometry, wavelength, turbulence profile, weather column, pointing statistics and modulation, it predicts BER, capacity, outage and margin. Field measurements then refine the inputs rather than being a prerequisite to start designing.
SCOPT's FSO Aerial Link Analyzer implements the workflow described here — Gamma–Gamma / Málaga / Weibull turbulence, Kim/Kruse layered attenuation, Rayleigh/Hoyt/Beckmann pointing error, and Monte Carlo BER, ergodic capacity, outage probability and link budget — validated against analytical formulas and published results.