Sensors measure the satellite's orientation against the Sun, Earth and stars, and wheels, magnets or thrusters turn it to face the right way.
A satellite's attitude is the direction it faces. A camera must point at the ground, an antenna at its coverage area, and solar panels at the Sun. Getting this right is the job of the attitude control system.
It works in a loop: measure where you point, compare with where you should point, and push the satellite to close the gap. Then measure again.
NASA's small-spacecraft review splits the job in two. Determination uses sensors: sun sensors, magnetometers, gyroscopes and star trackers. Control uses actuators: reaction wheels, magnetic torquers and thrusters.
ISRO's GSAT-19 is a typical example. ISRO says Sun, Earth and star sensors plus gyroscopes give it an orientation reference, and its Attitude and Orbit Control System turns it with momentum wheels, magnetic torquers and thrusters.
No single sensor is enough. A sun sensor gives the Sun's direction to about 0.1°, NASA says, but one direction cannot fix a full three-axis orientation, so a second source such as a magnetometer is needed. Star trackers are the most precise sensors of all.
From geostationary height, about 36,000 km up, a pointing error of just 1° moves an antenna beam more than 600 km on the ground (see the calculation below).
Low Earth orbit imaging satellites face the same problem at a smaller scale. A tiny wobble while the camera is open smears the picture. For a camera 500 km up, a star tracker error of 8 arcseconds shifts the view by only about 19 m; a sun sensor alone would leave it off by close to a kilometre.
Without pointing, a satellite loses more than its aim. Solar panels turn away from the Sun and power fades. ISRO concluded that its Mars Orbiter Mission, having likely run out of propellant, could not hold the attitude needed for steady power, and it declared the spacecraft at end of life in 2022.
Note how high the NavIC satellites sit — at that distance, each must hold its antenna pointed at India to a fraction of a degree. Open the NavIC tracker
Pointing error turns into a ground error by simple trigonometry. For a satellite at distance d looking straight down, an angular error α shifts the aim point by
x = d × tan α
Geostationary satellite, d ≈ 35,786 km (the geostationary altitude):
Low-orbit camera at d = 500 km: α = 1° gives x = 500 × 0.01746 ≈ 8.7 km.
Low-orbit imager at d = 500 km with a star tracker that knows its pointing to 8 arcseconds (NASA's table value). One arcsecond is 1/206,265 radian, so
x = 500,000 m × 8 ÷ 206,265 ≈ 19 m.
That is why imaging satellites use star trackers: sun sensors (about 0.1°, NASA) would give 500,000 × 0.001745 ≈ 870 m.
No single sensor gives full three-axis attitude. A sun sensor gives one direction; a magnetometer gives another. NASA notes that combining two direction vectors allows three-axis solutions such as the TRIAD and QUEST algorithms. Magnetometers need care: NASA warns that nearby reaction wheels, torquers and power lines distort their readings, so some missions mount them on a boom. Gyros fill the gaps between slower star-tracker updates, but their bias drifts and must be estimated onboard.
Its orientation in space — which direction its antennas, cameras and solar panels are facing.
Sensors: star trackers and sun sensors (also gyroscopes, magnetometers, Earth sensors). Actuators: reaction or momentum wheels and magnetic torquers (also thrusters).
About 36,000 km × tan 0.1° ≈ 62 km.