How long: per sub and per target

A thin line beside a passage says how far its claim is proven: measured, implemented, illustrative, inferred. Colored numbers carry the same mark. A passage with no line is stated. How claims are marked.

A sub is one single exposure; the Seestar takes many and stacks them into the picture you see. So there are two lengths to choose: how long each sub is, and how long you stay on the target in total. This Note takes them in that order.

Read noise against sky noise

Every time the camera reads its sensor, it adds a little random noise of its own, the same amount whether the sub was 1 second or 60. That is read noise. Meanwhile the sky glow, and the heat of the sensor, pour in a steady trickle of electrons, and their noise grows with the square root of the time you expose.

For a faint target sitting on the sky background, the signal to noise of a stack (how far the target stands above the random scatter around it) works out like this:

Hold the total time T fixed and only the last term depends on how you slice it: more, shorter subs mean more reads, and each read adds R². How much that costs depends on how big R² is next to the sky's B·t in one sub. A common rule of thumb is to stop lengthening subs once the sky's share is about ten times the read noise's; past that point read noise costs you under 5 % of the signal to noise.

With read noise this low, the rule of thumb is met at about 8 s on a sky giving 1 e⁻ per second per pixel, and at about 40 s on a very dark sky or behind a narrow filter giving 0.2 (illustrative). A light-pollution filter passes only narrow color bands, so it cuts the sky's trickle and pushes the useful sub length up; that is one honest reason to pick longer subs.

My sensor's numbers

These come from How calibration works, measured on the telephoto camera of one Seestar S30 Pro (firmware 8.46, gain 200) from bias and dark frames I shot on September 29, 2026. They describe my unit, not every Seestar.

• Read noise: 17.2 ADU, or 0.89 e⁻, on every read (measured). ADU are the camera's raw counting units; e⁻ are electrons.
• Bias pedestal: about 1,112 ADU on every frame before any light arrives (measured).
• Straight-line response: on the S30, S30 Pro and S50, every color stays within about 5 % of a straight line up to 90 % of its clip level, then clips, with no gentle roll-off (measured).
• Green stops lowest. In the raw frames, green stops counting between 49,808 and 54,288 on the 6 units with normal color handling, while red and blue go on to 65,535 (measured). That comes from the bright-star colors Experiment (8 units of 4 models, mostly with the IR-cut filter). It is exploratory: every night in it was used to work the method out, so nothing has yet been checked on fresh data. Why bright stars turn green or purple explains what it does to your pictures.

If you want your own sky's rate in electrons: the two read-noise figures above imply about 19 ADU per electron at gain 200 on my unit (17.2 ÷ 0.89) (inferred). Take the background level of a green pixel in a raw sub, subtract about 1,112, divide by 19 and then by the sub length. I haven't checked that recipe against a separately measured sky, and green is the color to use because the camera scales red and blue up before writing the file (inferred).

What doubling the sub does to bright stars

A star's peak grows in step with the sub length, so every doubling doubles it. Here is a made-up star whose brightest pixel collects 12,000 ADU above the pedestal in a 10 s sub (illustrative):

The 40,000 line is the Experiment's own advice for brightness measurements (photometry), 0.80 of the lowest green ceiling it measured; on three of the four models the response stayed straight up to there, and on the one S50 Pro it asks you to check with fainter stars too, peaking under about 20,000 (measured). In magnitudes (the astronomer's brightness scale, where a bigger number is fainter), each doubling of the sub moves the brightest star you can record without clipping about 0.75 magnitude fainter (2.5 × log₁₀ 2), and going from 10 to 30 s moves it about 1.2 magnitudes. For a field with a few very bright stars, or for measuring variable stars, shorter subs keep more stars usable.

Trailing and field rotation

In alt-az mode the Seestar tracks by turning in altitude and azimuth, and the field slowly rotates around its center as the night goes on. It turns fastest for targets high in the sky. Within one sub, stars away from the center smear into short arcs, and the smear grows with the sub length.

Because the arc grows in step with the exposure, a 20 s sub smears a corner star about twice as far as a 10 s sub of the same target, and a 30 s sub three times as far (illustrative). EQ (equatorial) mode, with the Seestar tilted on a wedge so that it turns around one axis lined up with the Earth's, removes the rotation; Alt-az or EQ mode covers why the field turns and when the wedge is worth it.

Rejected subs

A smart telescope leaves out of the stack the subs it can't use, for example when the stars trailed or the mount shook, and a longer sub gives something more time to go wrong. These numbers are third-party, from owners posting on the AstroBin forum, not from my frames:

• One S50 owner, saving every sub in alt-az mode without live stacking, reports about 95 % usable at 10 s, 75 to 80 % at 20 s and about 60 % at 30 s.
• Another owner, in EQ mode, lost about 40 % of one session's 30 s subs to smeared and streaked stars; after leveling the wedge more carefully, later 30 s sessions lost few.
• Several owners say they stay at 10 s in alt-az, and some use 20 s in EQ mode.

I haven't measured rejection rate against sub length on my own frames yet, so I can't give you a number of mine.

Warm nights

The sensor's heat adds electrons the way the sky does, a steady trickle per second, so its noise adds to the sky's rather than to the read noise. In my darks the dark current was 0.33 to 0.62 e⁻ per second at 29.3 to 33.2 °C, and it doubles every 4.3 °C or so (measured). As the calibration page says, the 5.8 °C doubling in its model and the 4.3 °C from my darks don't agree yet, so treat the doubling as a rough guide (measured).

The camera's black-level clamp takes the average dark current back off, so you don't see it as a brighter background, only as extra noise and warm pixels (measured). On a warm night that extra trickle makes read noise matter even less, which tilts nothing toward long subs: it simply raises the floor every sub sits on (inferred).

Per target: how much total time

Once each sub is long enough that read noise hardly matters, the signal to noise of the stack grows with the square root of the total exposure. Four times the time gives twice the signal to noise.

Exposure isn't time on target. Time goes between subs too: saving each frame, shifting the pointing a little every few subs (dithering) and waiting for it to settle, refocusing, and working out again where the telescope points. Owners on the AstroBin forum report capture time well below time on target: one S50 Pro owner timed about 3 s between live-stacked subs and a 16 to 17 s pause about every five subs, and estimates that works out to about 62 % of the time spent exposing with 10 s subs, 83 % with 30 s and 91 % with 60 s, before any subs are rejected; another got about 1.5 hours of exposure out of a 3-hour session. These are third-party estimates, not my measurements. The same thread makes a useful point: a high share of subs kept says nothing about how much of the night was spent exposing.

So overhead favors longer subs, as a dark sky or a narrow filter does, and it pulls against rejections, which favor shorter ones. At 62 %, four hours on target give about 2.5 hours of exposure, or about 1.6 times the signal to noise of one hour of exposure rather than 2 (illustrative).

Star positions sharpen over a night

On my M 31 night, the position of a moderately bright star (Gaia G 11 to 12) averaged over the night was off from the catalog of the Gaia space telescope by 48 thousandths of an arcsecond, typical of 171 stars, while a single 10-second frame was about 8 times worse, 0.11 pixel (measured). That is from the star-positions Experiment, scored on 88 frames held out from building the method. Two qualifiers travel with it. The 48 comes from the nights the method was built on, so the Experiment doesn't count it as confirmed; locked in advance and run once on another owner's S30 Pro in EQ mode, the same measure came out 72 and 86 thousandths of an arcsecond on two nights (measured). That telescope differs from mine in unit, mount mode and sharper focus at once, so the Experiment can't yet say which matters most. And on alt-az nights where the field barely rotates, a star sits on the same spot of the sensor all night, so part of its error repeats in every frame and more frames don't average it away; that comes from a follow-up look the Experiment counts as exploratory, and its advice is to dither the pointing in both directions (measured).

So which sub length?

This is guidance from the reasoning above, not a measurement (illustrative).

Whatever you pick, more total time on the target does more than a longer sub. SkyCruncher's night planner suggests 10 s frames for deep-sky targets and shows how long each target is up, so you can see how many hours a night can give it (implemented).

Where these numbers come from

• How calibration works (updated October 9, 2026): read noise, bias pedestal, the straight-line response, dark current and its doubling, the clamp. One S30 Pro, plus clip levels on an S30, an S30 Pro and an S50.
• Why a Seestar's bright stars turn green or purple (Experiment, October 9, 2026, exploratory): green's ceiling and the 40,000 line.
• How precisely can a 30 mm smart telescope place a star? (Experiment, Confirmed on other owners' telescopes; my nights of September 18, 2026 not confirmed): the night-average and single-frame errors and the dithering advice.
• Third-party, stated, not ours: two AstroBin forum threads, on Seestar S50 exposure times and on S50 Pro capture time against wall time.

The lines beside passages mark how far each claim is proven; see how claims are marked. More Notes and guides are on the Learn page.