Rain Fade Time Series
Synthesise a second by second rain fade series for a microwave link with ITU-R P.1853-2, and see how often and for how long fades exceed the margin.
Free to use. Sign up free to save, download and print.
Worked example: 23 GHz over 8 km with a 30 dB margin
Fades of 10 s or longer: 12 a year
Loading the Rain Fade Time Series…
How it works
Rain fade time series: the ITU-R P.1853 method
ITU-R P.530-19 gives the rain fade exceeded for a percentage of an average year, but not how that time is split into fades. ITU-R P.1853-2 Annex 3 makes a series, one sample a second, whose statistics follow P.530-19. It fits a log-normal distribution to the P.530-19 rain fade from 0.01 to 1 per cent of the time, below the probability of rain P₀, then passes Gaussian noise through two low-pass filters, so fades rise and fall at realistic rates, and maps it onto that distribution. For the rest of the year the series has no rain fade.
With a fade margin, P.530-19 gives two figures to compare with the series: the time the margin is exceeded, found by solving its rain fade equation for the margin, and the outage intensity, the number of fades of 10 seconds or longer a year, from its equation (78). The series counts its own.
ln Ap = σR Q⁻¹(p / P₀) + mR
αR = Q⁻¹(P₀ / 100)
Xᵢ(k) = ρᵢ Xᵢ(k − 1) + √(1 − ρᵢ²) n(k), with ρᵢ = e^(−βᵢ Ts)
G = 0.3746 X₁ + 0.7738 X₂
A = exp(σR Q⁻¹[(100 / P₀) Q(G)] + mR) when G > αR, otherwise 0
N₁₀ₛ = 1 + 1313 p^0.945
- Ap, A
- rain fade exceeded for p per cent of the time, and the fade in the series each second, in dB
- p
- time percentage, of an average year
- P₀
- probability of rain, in per cent of the year, from ITU-R P.837-8 or local data
- mR, σR
- mean and standard deviation of ln A in the fitted log-normal distribution
- Q, Q⁻¹
- Gaussian tail probability, as in ITU-R P.1057, and its inverse
- αR
- threshold on G; below it the series has no rain fade
- n, X₁, X₂, G
- unit Gaussian noise, one sample a second with k counting the seconds, the two filtered series and their weighted sum
- β₁, β₂, Ts
- filter constants, 9.0186 × 10⁻⁴ and 5.0990 × 10⁻⁵ per second, and the 1 s sampling period
- N₁₀ₛ
- fades of 10 s or longer a year at a margin exceeded for p per cent of the time, ITU-R P.530-19 equation (78)
Standards
- ITU-R P.1853-2 Annex 3 time series synthesis of rain attenuation on terrestrial paths
- ITU-R P.530-19 §2.4.1 rain attenuation and §2.4.5 event statistics
- ITU-R P.837-8 rain rate and probability of rain
Worked example
Worked example: 23 GHz over 8 km with a 30 dB margin
A 23 GHz link over 8 km on vertical polarisation, with R₀.₀₁ entered as 45 mm/h, the probability of rain P₀ entered as 4 per cent and a 30 dB fade margin.
| Step | Working | Result |
|---|---|---|
| Log-normal fit | ln Ap against Q⁻¹(p / 4), nine points from 0.01 to 1 per cent | mR = 0.156, σR = 1.089 |
| Rain threshold | αR = Q⁻¹(4 / 100) | 1.751 |
| Time above the margin | P.530-19 Ap = 30 dB, solved for p | 0.00617 per cent |
| Minutes a year | 0.00617 / 100 × 525,960 | 32.5 min |
| Fades of 10 s or longer | 1 + 1313 × 0.00617^0.945 | 12 a year |
The 12 fades a year is the ITU-R P.530-19 outage intensity, its estimate of unavailability events at this margin, and the fit stays within 0.69 dB of P.530-19 from 0.01 to 1 per cent. A synthesised year is one random example: fades this deep come from a few long events, so the series' own count and longest fade change with the seed, and more years steady them. R₀.₀₁ and P₀ are entered here; for a site without long-term records, the tool reads both from the ITU-R P.837-8 maps.
How long does rain fade last?
ITU-R P.530-19 says there is little information yet on the overall distribution of fade duration, though rain fades last much longer than multipath fades. It gives the number of fades of 10 seconds or longer a year as N₁₀ₛ = 1 + 1313 p^0.945, for a margin exceeded p per cent of the year, from measurements at 12.3 to 83 GHz on paths of 1.2 to 43 km. The series adds the rest: a count of every crossing of the margin, the longest events and the share of outage in events of 10 seconds or longer.
What counts as a fade event?
ITU-R P.530-19 defines an attenuation event as a fade lasting 10 seconds or longer, the events that count as unavailability on a digital link. The tool also counts shorter crossings of the margin and shows the share of outage in the longer events. Its count need not match equation (78): the two are separate ITU-R models, and neither gives an accuracy for the count.
Is this the rain that actually fell at my site?
No. The series is synthetic. It reproduces the long-term rain fade statistics for the site, so it is a realistic example of how fades come and go, but its dates and single events are not past weather.
Why does the deepest part of the curve differ from P.530-19?
A synthetic series is one random example. The time spent in shallow fades settles quickly, but the deepest fades come from a few long events and vary from run to run. Synthesise more years or try other seeds to see the spread.
What frequencies and path lengths does it cover?
ITU-R P.1853-2 gives its terrestrial method for 4 to 40 GHz and paths of 2 to 60 km, so the tool works in that range only. For a single availability figure at other frequencies, use the Rain Fade Calculator.
Does any data leave my browser?
The synthesis runs in your browser. The path midpoint is sent to the noIM₃ server to read the ITU-R P.837-8 maps. Nothing else you enter is sent.
Sign up free to keep your work.
Free to use. Saving, downloading and printing need a free account.