Constellation size is the number that travels furthest in a press release and explains the least about performance. Two radar constellations of identical size can differ widely in how often either one returns to a given site.

Where the headline number comes from

The arithmetic behind most revisit claims is straightforward, which is precisely the problem. Orbital period divided by satellite count, adjusted for swath, yields a global average that is arithmetically true and operationally useless.

That average describes the whole planet over a long interval. It says nothing about a 30-kilometre site at 52 degrees north on a Tuesday, and the site is what the buyer of an image actually cares about.

What sets the interval between passes

Four parameters govern revisit, and satellite count is only one of them. A market survey of SAR imagery providers, such as how eight SAR operators compare on observationdata.com, publishes a revisit figure for each operator, and those figures become comparable only once the geometry behind them is known.

Constellation geometry is conventionally written in Walker notation as i: T/P/F, where i is inclination, T the total number of satellites, P the number of equally spaced orbital planes, and F the phasing factor. The two standard families divide along the spread of their ascending nodes: a Walker-Star arrangement distributes the right ascension of the ascending node across 180 degrees and suits near-polar missions, whilst a Walker-Delta spreads it across the full 360 and suits inclined ones. Analytical treatments of revisit time work directly from that geometry, and the design choice moves the resulting figure more than adding hardware does.

Only the third of those four is routinely published, and the first, which carries the most weight, is almost never stated outright.

Two satellites, one plane, almost nothing gained

The clearest illustration needs no simulation. Place two identical radar satellites in the same orbital plane, separated in true anomaly, and each passes over a mid-latitude site at nearly the same local solar time, one behind the other. The second satellite adds a few minutes of latency improvement and almost no new access.

Move the second satellite into a plane whose ascending node is offset by 90 degrees, and the same pair delivers something close to twice the daily opportunities over that site. The hardware bill is identical. Only the node spacing changed, and the revisit roughly halves as a result.

The same two satellites in one plane and in two planes offset by 90 degrees of right ascension. Source: own diagram, following standard Walker constellation geometry.

This is well understood in the formation flying and distributed systems community, and the topics of interest for this conference placed constellation geometry alongside relative navigation for exactly that reason. It travels poorly into commercial datasheets, where one revisit figure stands in for the entire design.

Advertised count against operating count

A second gap sits underneath the first. Providers publish launched totals, planned totals, and operating totals more or less interchangeably, and the three diverge substantially.

Planet states that it has deployed 462 satellites since 2013 and currently operates approximately 200 in orbit. Capella Space offers a sharper case: the public catalogue of Capella payloads lists 19 launched objects, of which 9 remained in orbit in August 2026, against an announced constellation of 36. None of those figures is dishonest in isolation. They simply answer three different questions, and only the operating count bears on revisit.

The gap is structural rather than exceptional. Early smallsat generations are retired as newer ones fly, and low orbits decay, so a fleet growing on paper can hold steady or shrink in operation.

Why the radar case is cleaner than the optical one

For optical constellations, this argument is usually swamped by weather. Roughly 67 per cent of the Earth's surface carries cloud at any moment, so an optical revisit figure is a statement about opportunities rather than images, and the conversion between the two is a local climatology problem.

Radar removes that term. A SAR pass is a usable pass in nearly all conditions, with heavy precipitation at X-band the main exception. What remains after weather drops out is orbital geometry alone, which is why radar constellations are the honest test case for the whole question. The dawn-dusk sun-synchronous orbits favoured by several radar operators tighten the picture further, since power and thermal conditions stay stable across the orbit.

Reading a published revisit figure

A revisit number becomes interpretable as soon as four things are known about it. Datasheets list mode and swath readily enough; what they rarely state is which of them the figure assumes.

• Whether the figure is a global mean or an estimate for a specific latitude, because the two are rarely the same number.
• Whether it counts access opportunities or confirmed collections after tasking, since queue competition removes a share of the former.
• Which imaging mode and swath the figure assumes, because a narrow spotlight mode collapses the access corridor.
• How many satellites were operating on the date the figure was calculated, rather than launched or planned.

Asked in that order, the four questions convert a marketing figure into an engineering one. They also tend to reveal which operators have thought carefully about the answer and which are quoting an average computed once and reused since.

Satellite count will keep leading the announcements, because it is a single rising number and node spread is not. Anyone specifying a radar tasking contract is better served by asking how many planes the constellation occupies, and at what inclination, before asking how many satellites fly in it.