Star trackers — arcseconds from a camera and a catalogue
A star tracker is the only sensor that gives absolute, drift-free attitude anywhere in the solar system. It photographs a patch of sky, centroids the stars, matches the pattern against an onboard catalogue, and solves for orientation. What separates a $10 k CubeSat unit from a $1 M flagship instrument is accuracy, robustness to stray light, and how quickly it recovers when lost.
Symbols used on this page are defined in the Nomenclature → attitude & rotations.
The specs that matter
| Spec | Meaning | Why it bites |
|---|---|---|
| Cross-boresight accuracy | error about the two axes perpendicular to the boresight | the headline number; typically 1–10″ |
| Roll (about-boresight) accuracy | error about the viewing direction | 5–10× worse than cross-boresight — the reason to fly two heads |
| NEA (noise-equivalent angle) | random, per-measurement noise | averages down with $1/\sqrt N$ frames — the benign part |
| Low-frequency / thermal error | slowly-varying bias | does not average down; usually the real budget driver |
| Update rate | solutions per second | sets how far the gyro must coast between fixes |
| Exclusion angles | keep-out cones for Sun / Earth / Moon | drives head placement and mission geometry |
| Max slew rate | rate at which it still tracks | if exceeded, stars streak and you lose lock during a manoeuvre |
| Lost-in-space time | full-sky acquisition with no prior | the recovery path after a safe-mode event |
Why geometry beats averaging
Star-direction measurements only constrain rotations perpendicular to the measured vector. Adding more stars in the same small patch averages the noise but does almost nothing for the rotation about that patch. This is exactly the conditioning story behind Wahba's problem — and the same effect as GDOP in GNSS:
Representative units
| Unit | Class | Cross-boresight | Notes |
|---|---|---|---|
| Blue Canyon Standard NST | CubeSat / smallsat | ~6–8″ | very widely flown, compact, low power |
| Sodern Auriga / Hydra | smallsat → flagship | ~1–5″ | Hydra uses multiple heads into one processor |
| Ball CT-2020 / CT-6xx | large spacecraft | ~1–3″ | long heritage on science missions |
| Terma HE-5AS | medium/large | ~2″ | robust radiation performance |
| Berlin Space Tech ST200/400 | CubeSat | ~10–30″ | very small and inexpensive |
| Jena-Optronik ASTRO APS | flagship / GEO | ~1″ | APS detector, high radiation tolerance |
Integration notes
- Baffle design dominates real performance. The stray-light rejection of the baffle, not the detector, usually decides whether you keep lock near the Sun-exclusion limit.
- Thermo-elastic alignment between the tracker and the payload boresight is often the largest term in the whole pointing budget — see the ADCS jitter budget, where it beats every electronic error.
- Time-tagging. The solution is valid at the middle of the exposure, not at packet arrival; getting this wrong injects a rate-dependent bias straight into the filter.
- Pair with the gyro, don't choose between them. The tracker bounds drift; the gyro bridges the gaps and rejects tracker outliers. That is exactly the MEKF.
Related: star-tracker theory · attitude & the MEKF · back to Hardware.