Coverage and revisit
What you will accomplish
A coverage number you can defend, arrived at by refining the analysis until it stops changing rather than by accepting the first result.
Coverage is the easiest quantity in this software to get confidently wrong, and the two ways it goes wrong are both settings, not physics.
Prerequisites
- A constellation, from generation, synthesis, or import.
Steps
Take a first look, coarsely
orbitforge coverage \
--constellation ph1.json \
--duration-hours 6 --step-seconds 120 \
--grid-deg 20 --min-elevation-deg 10 mean coverage 62.9% (area-weighted), min 0.0%, fully covered 1.1%, any 60.0%
max fold 3, max coverage gap 2160 sRefine the grid and watch what moves
for g in 10 5; do
orbitforge coverage --constellation ph1.json \
--duration-hours 6 --step-seconds 120 \
--grid-deg $g --min-elevation-deg 10
done--- grid 10 deg ---
mean coverage 70.1% (area-weighted), min 0.0%, fully covered 10.8%, any 68.4%
max fold 3, max coverage gap 5760 s
--- grid 5 deg ---
mean coverage 71.2% (area-weighted), min 0.0%, fully covered 11.0%, any 73.0%
max fold 3, max coverage gap 5760 sReading the convergence
| Grid | Mean coverage | Fully covered | Max gap |
|---|---|---|---|
| 20 deg | 62.9 percent | 1.1 percent | 2160 s |
| 10 deg | 70.1 percent | 10.8 percent | 5760 s |
| 5 deg | 71.2 percent | 11.0 percent | 5760 s |
The 20-degree run under-reported the worst gap by a factor of 2.7, and it did so in the safe-looking direction: 2160 seconds against a true 5760.
A coarse grid does not sample the points where the gaps are. Quoting 36 minutes when the answer is 96 minutes is the kind of error that survives a design review and fails in service.
Between 10 and 5 degrees the max gap does not move at all and the mean shifts by 1.1 points. That is what convergence looks like: refine until the numbers you care about stop changing, then quote them.
fully covered moved from 1.1 to 10.8 percent between 20 and 10 degrees, which
is a tenfold change. Any statistic that swings that hard under refinement was not
converged, and reporting the first one would have been reporting the grid rather
than the constellation.
The two settings that decide your answer
| Setting | Failure when too coarse | How to check |
|---|---|---|
--grid-deg | Holes fall between sample points; gaps under-reported | Halve it; re-run; see whether the max gap grows |
--step-seconds | Gaps shorter than the step are invisible | Halve it; re-run; see whether the max gap grows |
Both fail in the same direction, and it is the flattering one. That is what makes convergence a habit worth having rather than an optional refinement.
The grid is two-dimensional, so halving the spacing quadruples the point count and roughly quadruples the run time. Explore coarse, converge before quoting.
Always state the elevation mask
The same constellation, with and without a realistic mask:
| Setting | Mean coverage | Fully covered |
|---|---|---|
| 10-degree mask | 70.1 percent | 10.8 percent |
| 25-degree mask | 29.6 percent | 0.0 percent |
A satellite at 5 degrees elevation is geometrically visible and practically useless: longest slant range, most atmosphere. Coverage quoted without a mask is a number no real ground segment will achieve.
State the mask every time you quote a coverage figure. The two rows above are the same design.
Reading the statistics honestly
| Statistic | What it hides |
|---|---|
| Mean coverage | Structural holes. A healthy mean coexists with min 0.0% |
min | Nothing. It is the most honest number in the output |
fully covered | Whether anything is continuously served. Zero here rules out uninterrupted service |
| Max gap | Usually the binding requirement. Compare against what the mission tolerates |
| Max fold | Redundancy. Fold of 1 means a single failure creates an outage |
min 0.0% appears in every run above because at 53 degrees inclination the
design never reaches the poles. That is structural, not statistical, and no
number of satellites fixes it.
A convergence checklist
- Run coarse to explore.
- Halve the grid; re-run. Did the max gap grow?
- Halve the step; re-run. Did the max gap grow?
- Repeat until the numbers you intend to quote stop moving.
- State the elevation mask, the window, the grid, and the step alongside the result.
That last item is what makes the number checkable by someone else, which is the actual standard for a design review.
Next steps
- Closing a link, because coverage says nothing about whether a usable signal arrives.
coveragereference for sensor footprints and the full flag set.
main (pre-release)