Exposure calculator
The shortest sub where read noise stops mattering. Past this point longer subs add almost nothing, so you can keep them short and throw away fewer frames to wind, planes and guiding.
Read the guide: How long should your subs be? Sub-exposure length explained
Shortest useful sub
--
Noise added by read noise, by sub length
Total integration
| Total time here | Subs | Same SNR at a dark site |
|---|
How to use the exposure calculator
- Pick your camera and the gain you shoot at. The read noise fills in from the manufacturer's figures; enter your own if you have measured your camera.
- Pick your telescope and any reducer. Aperture, focal length and central obstruction set how much sky light each pixel collects.
- Pick your filter, then your sky as a Bortle class or, better, an SQM meter reading.
- Use the result as your shortest sub. Anything longer is fine until stars clip or guiding errors show.
Questions
How long should my sub-exposures be?
Long enough that noise from the sky background swamps your camera's read noise. Past that point, longer subs add almost nothing to the final signal to noise ratio, so shorter subs cost you nothing and you lose less to planes, wind and guiding.
Why are my subs so short under a bright sky?
A bright sky fills each pixel with background light quickly, so read noise stops mattering within seconds. That is also why you need more total integration under light pollution: this sky is shown as a multiple of a dark site.
Does gain or ISO change the answer?
Yes, through read noise. Higher gain usually means lower read noise, so the sky swamps it sooner and the shortest useful sub gets shorter. Many modern CMOS cameras drop sharply in read noise when their high conversion gain mode switches on.
How long should narrowband subs be?
Much longer than broadband. A 3 nm filter passes about 1% of the sky light of a 300 nm luminance filter, so the sky takes about 100 times longer to swamp read noise. Several minutes to 10 minutes is typical.
How accurate is the result?
It is a physical estimate, not a measurement. The method panel lists the assumptions. For a precise figure, measure the background of a test sub in electrons and compare it with the target shown here.
Method
Each sub carries two kinds of noise that do not depend on the target: read noise, fixed per frame, and shot noise from the sky, which grows with exposure. Once the sky signal in a sub is large compared with the read noise squared, read noise stops mattering.
Sky signal per pixel
flux = 10080 x 10^(-0.4 x SQM) photons / s / cm² / nm / arcsec² rate = flux x bandwidth x area x scale² x QE x 0.81 e- / s / pixel area = pi x (aperture / 2)² x (1 - obstruction²) cm² scale = 206.265 x pixel size / focal length arcsec / pixel
10080 photons per second per cm² per nm is a magnitude 0 source at 550 nm. The 0.81 covers optics and filter transmission (0.9 each).
Sub length
Total noise in a sub is sqrt(sky + RN²). To keep read noise from adding more than a fraction p to it:
sky electrons >= RN² / ((1 + p)² - 1) p = 5% gives 9.76 x RN² sub = sky electrons / rate
Integration
When sky noise dominates, the time needed for a given signal to noise ratio on faint detail scales with sky brightness. The dark-site column compares against SQM 21.7:
penalty = 10^(0.4 x (21.7 - SQM))
Assumptions
- The sky spectrum is treated as flat across the filter band. Real light pollution has strong emission lines, so narrowband results can be off by a factor of two in either direction. Treat them as a starting point.
- On a color sensor, broadband light (over 30 nm) is split three ways by the Bayer matrix. Narrow lines land mostly in one channel at full width.
- Peak QE is used across the band, which overstates the signal a little and makes subs slightly short.
- Dark current is ignored, which is fair for cooled cameras at -10 °C or colder.
- Bortle classes map to typical zenith SQM values: 1: 21.9, 2: 21.6, 3: 21.4, 4: 20.8, 5: 20.1, 6: 19.3, 7: 18.6, 8: 18.1, 9: 17.5. A meter reading from your site beats any class.
- Camera presets come from each manufacturer's manual or spec page (ZWO, QHY and Player One). Where a read noise value is only shown in a chart, it is marked "about". Read noise varies a little between units, so your own camera may measure slightly differently.