Calibration mystifies beginners because it's taught as ritual — 'take 30 of these with the cap on' — instead of as what it is: photographing your equipment's defects separately, so software can remove them from your photographs of the sky. Each frame type isolates one defect.
DARKS photograph what the sensor does in total darkness: thermal signal and hot pixels accumulating over time. Cap the scope, match your lights' exposure length, gain and sensor temperature exactly, and shoot 20–30. Subtracting the master dark removes the heat signature. Set-point cooling is what makes dark libraries reusable — shoot once at -10 °C, use for months.
FLATS photograph your optical train's fingerprint: vignetting (corners darker than center) and dust motes (those out-of-focus donuts). Illuminate the objective evenly, expose to put the histogram peak near half scale, shoot 20–30 — same focus, same rotation, same everything as your lights. Division by the master flat evens the field. Flats are the calibration frame that most improves a suburban image, because gradients and dust are what stretching amplifies first.
BIAS frames photograph the read-out pattern itself: the shortest possible exposure with the cap on. Modern CMOS practice often replaces bias with DARK-FLATS — darks matched to your flat exposure — which calibrate the flats more reliably on sensors with unstable bias behavior. Use one or the other consistently; your stacking software's documentation tells you which it expects.
The matching rules and shelf lives
Darks match lights on exposure, gain, offset and temperature. Flats match lights on optical configuration — focus position, camera rotation, everything in the light path — which is why you shoot them before touching anything after a session. Bias/dark-flats match the flats.
Shelf life: bias is nearly immortal; darks age over months as the sensor's hot-pixel population evolves (refresh ~2× a year, or after firmware/driver changes); flats die the instant the optical train changes — a refocus large enough to move dust shadows, a camera rotation, or any disassembly. When in doubt, flats are cheap: reshoot.
- Uncooled-camera darks from a different ambient temperature than the lights. Without set-point cooling, match conditions per session or rely on dithering + outlier rejection instead.
- Shooting flats through a different filter than the lights. Dust sits on filter surfaces too — flats are per-filter.
- One master dark for every exposure length 'close enough'. Scaling darks works only when your software does it deliberately (and bias is handled correctly) — not by accident.