How spy satellites work
Published 19 October 2026 · All guides
In short: a spy satellite's detail is set by the width of its mirror and how far away it is. Its timing is set by its orbit, so it sees any one place only now and then. A camera needs daylight and clear skies, and works best with a low sun that casts shadows. Radar and signals collection need neither, which is why real programmes fly all three.
The mirror sets the limit
No camera can resolve detail finer than its optics allow. Light spreads as it passes through an aperture, so two points closer together than a certain angle blur into one. That angle is roughly 1.22 × the wavelength ÷ the mirror's diameter. Multiply it by the distance to the ground and you get the smallest thing the satellite can separate.
Visible light has a wavelength of about half a micrometre. Working that through:
- A 1.1 m mirror at 617 km resolves about 34 cm. That is roughly WorldView-3, a commercial satellite launched in August 2014, which sells imagery at 31 cm. The arithmetic lands within a few centimetres of what it actually delivers.
- A 2.4 m mirror at 250 km resolves about 6 cm. A 2.4 m mirror is the size of Hubble's, and in 2012 the US National Reconnaissance Office (NRO) gave NASA two surplus telescopes of exactly that size. One of them became the Nancy Grace Roman Space Telescope.
- The same 2.4 m mirror from geostationary orbit, 35,786 km up, resolves about 9 m. That is why nobody does fine imaging from there.
The air adds its own blur, so real systems do worse than the formula. In August 2019 a US President posted a classified image of a damaged launch pad at Iran's Semnan site. Amateur satellite trackers matched the viewing angle to a pass by USA 224, widely believed to be a KH-11 reconnaissance satellite, and estimated the image at about 10 cm per pixel.
Closer is sharper, but faster
Because detail falls with distance, imaging satellites fly low, typically a few hundred kilometres up. The price is speed: at that height a satellite circles the Earth in about an hour and a half, so it is over any one spot for a few minutes at most. It also feels the top of the atmosphere, which slowly drags it down and has to be countered with propellant.
Sunlight and shadow
An ordinary camera needs the ground to be lit. Less obviously, noon is not the best time. With the sun overhead there are no shadows, and shadows are what make a building, a vehicle or a missile readable from straight above. An analyst who knows the sun's angle can work out a structure's height from the length of its shadow.
That is why so many imaging satellites use a sun-synchronous orbit, which crosses each latitude at the same local time every day. Europe's Sentinel-2 satellites, for example, pass over at about 10:30 in the morning. Every picture of a place then has the same lighting, so changes stand out.
Cloud and night
There are two ways round the dark and the weather, and both cost detail or money.
- Thermal infrared sees heat, so it works at night: a running engine, a warm building, a recently used runway. But its wavelength is around twenty times that of visible light, so the same mirror resolves roughly twenty times coarser.
- Synthetic aperture radar brings its own illumination, a radio pulse, and sees through cloud and darkness alike. Commercial operators such as ICEYE and Capella Space now sell radar images to anyone.
How often it comes back
A single low satellite passes over a given place only occasionally, and the times shift from day to day because the Earth turns under the orbit (see orbital mechanics without the maths). Orbits are also predictable: anyone with tracking data can work out when a satellite will be overhead and time what they do around it.
The answer is numbers. Planet Labs flies a constellation of small satellites that images nearly all of the Earth's land every day, at a few metres per pixel. A few large satellites give detail; many small ones give frequency. Real programmes want both.
Listening instead of looking
Signals intelligence (SIGINT) satellites do not take pictures at all. They pick up radars, radio links and telemetry. The first US reconnaissance satellite to work was one of these: GRAB, launched on 22 June 1960 by the Naval Research Laboratory under the cover of a solar science experiment, mapped Soviet air-defence radars. Its true mission was declassified in 1998.
A transmitter is as loud at 3 a.m. as at noon and as loud under cloud as in clear sky, so signals collection is the one kind that ignores the weather and the time of day. Some SIGINT satellites sit in geostationary orbit, where they can listen to a whole hemisphere at once. When one was launched in November 2010, the NRO's director called it the largest satellite in the world.
Where it started: film in a bucket
The first imaging spy satellites, the US Corona programme, shot photographic film and dropped it back to Earth in a capsule that an aircraft caught in mid-air under its parachute. The first successful recovery came from Discoverer 14 in August 1960, and that single mission photographed more of the Soviet Union than every U-2 flight before it. The later cameras resolved about 1.8 m. Corona flew until 1972 and its imagery was declassified in February 1995.
Where Celestrike comes in
Celestrike runs on exactly these trade-offs. A camera's detail comes from its aperture and its distance; an imaging satellite collects best with the sun low, slows down at night and under storms, and only sees a site when its real orbit brings it over. A signals satellite listens through all of it. Celestrike is a free spy-satellite strategy game for iPhone and Android, played on real orbits — on the App Store and Google Play.