Dark frames remove thermal signal and amplifier glow from your light frames, but each dark is itself noisy, and that noise gets injected during subtraction. Averaging many darks into a master dark smooths it out. This calculator shows the diminishing-returns curve so you take enough darks without wasting a clear night.
How it works
A master dark made by averaging N darks has noise reduced by the square root of N. The residual noise it adds to each calibrated light, relative to one raw dark, is:
residual(N) = 1 / √NThe marginal benefit of one more frame is the difference between 1/√N and
1/√(N+1), which shrinks rapidly. The tool reports both the residual penalty at
your chosen count and the gain from adding a single extra frame, so you can see
when the curve has flattened.
Example and tips
With 1 dark the residual is 1.0 — the master adds as much noise as a single dark. With 16 darks it drops to 0.25, a 75 percent reduction. By 50 darks it is about 0.14, and pushing to 100 only reaches 0.10. Most astrophotographers find 20 to 50 darks gives a clean master with little benefit beyond. Always match exposure, gain, and temperature to your lights, and reuse a temperature-matched master dark library so you are not re-shooting darks every session.
The diminishing-returns curve in detail
The noise reduction is rapid at first and then nearly flat:
| Dark frames | Residual noise (fraction) | Gain from one more |
|---|---|---|
| 1 | 1.000 | – |
| 4 | 0.500 | large |
| 9 | 0.333 | moderate |
| 16 | 0.250 | smaller |
| 25 | 0.200 | smaller still |
| 50 | 0.141 | tiny |
| 100 | 0.100 | negligible |
The jump from 1 to 4 darks halves the residual noise — four frames bought you as much as going from 50 to 200 frames would later. This is the core insight: the first ten or so darks provide most of the achievable benefit, and every dark beyond that point returns diminishing gain.
When to stop: the practical sweet spot
Most astrophotographers settle between 20 and 50 darks as a calibration library:
- 20 darks gives a residual of about 0.22 — already a good calibration.
- 30 darks reaches about 0.18.
- 50 darks is around 0.14, and is a common upper-limit target for sensors with significant thermal current.
Going beyond 50 is most worthwhile for very long total exposures (many hours of integrated light) where even a small reduction in calibration noise matters, or for sensors that run warm and generate substantial fixed-pattern noise.
Temperature matching is more important than count
The single biggest source of dark-calibration error is not count — it is temperature mismatch. Dark current in a CMOS or CCD sensor approximately doubles for every 6–8°C rise in temperature. A master dark taken at 10°C will not correctly subtract the thermal signal from a light frame taken at 20°C, no matter how many darks you averaged. If your camera has no cooling, try to take darks immediately after your lights while the sensor is still at operating temperature, or build a library indexed by ambient temperature.