Integration is not the same as time spent
Four numbers separate the two. The gap between subs, when the shutter is closed and the camera is writing and the mount is settling after a dither. The frames you throw away. The dark hours that end before the target does. And the setup and teardown at each end of every session, which is pure overhead repeated once per night.
Put realistic figures in and the ratio is sobering. Six hours of integration at three minute subs with twelve seconds of overhead, an 85 percent keep rate, four usable dark hours a night and forty minutes of setup works out at 142 subs, 7.6 hours of actual imaging across two nights, plus an hour and twenty minutes of setup — call it 8.9 hours from car door to car door for six hours in the stack. About two thirds of the time outdoors survives, and that is a good case. Lengthen the overhead, drop the keep rate or lose an hour of the window to cloud and it falls fast.
The overhead figure is worth measuring rather than guessing
Twelve seconds sounds trivial next to a three minute sub — it is under seven percent. Put the same twelve seconds next to a thirty second sub and it is 29 percent, and a night of short subs spends nearly a third of the dark hours with the shutter closed. This is the strongest argument against very short subs on a rig with slow downloads, and it is invisible unless you time it.
Timing it is easy: watch the clock across ten frames and divide. What you are counting is everything between the end of one exposure and the start of the next — the download, the write to card or disk, the dither move and however long the guider takes to settle afterwards. Dither settling is usually the largest part and the one most worth attacking, because it can often be reduced by dithering every second or third frame instead of every one.
Keep rate is personal history, not a constant
| What ruins frames | Shows up as |
|---|---|
| Satellites and aircraft | A bright streak, usually removable by the stacking rejection instead of discarding the frame |
| Thin cloud drifting through | Raised background and bloated stars across part of the set |
| Wind gusts | Isolated elongated frames, worse with a long tube and a light mount |
| Guiding excursions | Same as wind, but clustered around whatever caused them |
| Dew on the corrector | A gradual softening across a run, often unnoticed until the frames are inspected |
The number to enter is the one from your own past sessions, and it varies more by site and season than by equipment. Somewhere in the eighties is a common experience for a settled night from a fixed pier; a windy night with a portable setup can drop well below that. The calculator uses it to inflate the frame count, because shooting exactly the number of frames the integration needs and then discarding some leaves you short.
Choosing the sub length
Longer subs mean fewer read noise events, less overhead as a fraction of the night, and fewer files to handle. Shorter subs mean less lost each time something spoils one, less demand on the mount, and less risk of saturating bright stars. The point where the balance tips depends on the read noise of the camera, how bright the sky is where you are, and how well the mount tracks, and none of those are things this page knows.
What the comparison section does show is the mechanical consequence: what a different sub length does to the frame count, the overhead fraction and the total imaging time. Going from three minutes to one minute at twelve seconds of overhead takes the overhead from 6.3 percent to 16.7 percent of the night and triples the number of files. Whether that is worth it is a signal-to-noise question your own test frames answer.
Calibration frames and where the storage actually goes
Darks, flats and bias frames add to the file count without adding integration. They are also the frames most likely to be reused across sessions, which is why the calculator takes a single total rather than trying to schedule them. The bigger storage surprise is not the raws at all: it is the calibrated and registered copies that a stacking run produces, which can be several times the raw volume and which live on the disk for as long as the project is open.
Plan the drive around the working set rather than the archive. The file size calculator is the place to sort out GB against GiB, which is where an apparently adequate drive turns out to be seven percent smaller than the label suggested, and the storage growth calculator is the one for a season rather than a project.
Questions people ask
How many subs do I need for a given integration time?
Divide the target integration by the sub length and round up, then divide again by your expected keep rate. Six hours at three minute subs is 120 frames that have to survive; at an 85 percent keep rate that means shooting 142. Shooting exactly 120 and then discarding the bad ones leaves you short of the target, which is the mistake this arithmetic exists to prevent.
Why does my session deliver so much less integration than the hours suggest?
Four leaks, all of them ordinary. Overhead between subs runs from a few percent on long subs to a third of the night on short ones. Rejected frames take another slice. The usable dark window is shorter than the night, because the target has to be high enough and the sky dark enough. And setup and teardown is repeated every session. Together they commonly halve the ratio between time outdoors and time in the stack.
Are longer subs better?
They reduce read noise contribution and cut overhead as a fraction of the night, which is real. Against that, each spoiled frame costs more integration, bright stars saturate sooner, and the mount has to hold for longer without a correction. Where the balance sits depends on your camera read noise, your sky brightness and your mount, so it is a thing to test rather than to look up. This page shows the mechanical side of the trade — frame counts, overhead fraction and total imaging hours — for two sub lengths side by side.
How much disk space will a project take?
The raw subs are the easy part: frame count times file size, plus the calibration frames. The part that catches people is the working set. Calibrated, registered and normalised copies produced during stacking can run several times the raw volume and stay on the disk while the project is open. Size the drive for the working set, and check GB against GiB while you are at it, because a drive sold as a round number of gigabytes reports about seven percent less.
Should I count calibration frames toward my integration?
No. Integration is signal collected on the target, and darks, flats and bias frames collect none of it — they exist to remove what the sensor and optics add. They do consume session time and disk space, which is why the calculator counts them for storage and leaves them out of the integration total. Many people shoot them once and reuse them across sessions, which changes the storage picture but never the integration one.