How to photograph star trails, and what the tutorials disagree about

Put the camera on a tripod, shoot short exposures back to back with the smallest gap the camera allows, and stack them afterwards; the sky turns 15.041° an hour, so a two-hour sequence draws a 30° arc whatever else you change. Everything after that first sentence is detail, and this page gives the detail twice over: what twenty-four published guides recommend, and what one real night measured on our own files actually did.

1. What a star trail is, in one minute

The stars are not moving. You are, with the ground, once every sidereal day — 23 hours 56 minutes and 4 seconds. That works out at 15.041 degrees per hour, and it is the only number in star-trail photography that nothing you own can change. Point a fixed camera at the sky, leave the shutter open, and every star draws an arc that long.

So arc length is a question about elapsed time, not about exposure, aperture or ISO. This is the single most useful thing to internalise, because it decides how long you are standing in a field.

The frame counts assume no gap between exposures, which is never quite true. Our own example night below ran at 81 % duty cycle and still produced continuous arcs; see the duty-cycle number.

2. Before you go: where, and when

How dark does it actually need to be?

Less dark than for the Milky Way, and this is good news. A star trail is built out of the brightest stars in the sky, and those survive a raised background that would erase faint nebulosity entirely. A stack that equalises each frame's sky before combining also fights the glow directly. You will get more stars and better colour from a darker site, but a Bortle 4 rural field is a perfectly good place to learn.

The Moon is a setting, not an obstacle

A bright Moon washes out faint stars, and it also lights your foreground for free. Both are true, and which one matters depends on the picture you want. A new Moon gives the most stars and a black foreground you will have to light yourself. A quarter Moon low in the sky gives fewer stars and a landscape that looks like a landscape. What actually counts is the Moon's altitude during your sequence, not the phase printed on a calendar: a full Moon below the horizon is no Moon at all.

Season, and the trap at high latitude

Winter nights are long, cold and often clear, and cold air holds less moisture. Summer nights are short, and above about 49° of latitude midsummer contains no astronomical darkness at all — that threshold runs through Paris and Vancouver, so it catches far more people than they expect. The window widens quickly with latitude: around 52° it lasts roughly two months, and at 60° about four. Anyone in Britain, Scandinavia, Canada or southern Patagonia should check for real darkness before planning a summer night, because the sky can simply refuse.

The free half of the satellite problem

Satellites shine by reflected sunlight, so they are only visible while you are in darkness and they, several hundred kilometres up, are not. That geometry holds for roughly the first and last couple of hours of the night, and around local midnight most low-orbit satellites are inside the Earth's shadow and go dark. A survey of Zwicky Transient Facility images found that 64 per cent of Starlink trails were recorded during twilight, and about 70 per cent of them below 40° of elevation (Mróz et al., 2022).

The practical version: start an hour later than you were going to, and point away from the horizon. That throws away most of your satellite passes and costs nothing. It is the least-repeated useful advice in the whole subject.

3. What to carry

4. The settings

Three recipes, by how many of these you have done. They differ less than you would expect, because the sky does not care how experienced you are.

Why ISO is an aesthetic decision

ISO 800 is the most-quoted starting point in the published tutorials, and the reason it is a starting point rather than an answer is that ISO trades star count against star colour. Push it and more faint stars clear the noise floor, so the sky fills up — and the bright stars saturate toward white, so Betelgeuse stops looking orange. Pull it back and you get fewer stars with more colour in them. Several tutorials state the trade in opposite directions and both are right; they simply want different pictures.

Trail brightness is decided per frame, not by the night

This one surprises people, and it follows directly from how the stacking works. A star-trail stack keeps the brightest value each pixel ever had, not the sum of them. So adding frames makes your trails longer and does nothing at all to make them brighter: the brightness of any point on an arc is whatever the one frame that drew it recorded there.

Two consequences worth carrying outside with you. Aperture and ISO are your only brightness controls, and they act on every frame equally, so a sequence that is one stop under is one stop under for four hours. And a faint star that never clears the noise in a single 30-second frame will not appear in the stack however many frames you take — which is the exact opposite of deep-sky work, where the whole point of adding frames is to lift things that no single frame could show. If you want more stars in the picture, you change ISO or aperture; if you want longer arcs, you stay out later.

This is also why the sky background does not average away. A maximum samples the top of the noise at every pixel instead of the middle of it, so the raw background gets noisier as frames accumulate. Fighting that is a job for the stacker, not for the camera — per-frame sky equalisation is what does it.

The number nobody publishes: duty cycle

Every guide tells you to keep the gap between frames short. None of them tells you what "short" means, because the answer depends on the frame length. The useful figure is shutter-open time divided by the interval.

Above 90 % you can stop thinking about it. Below it, duty cycle stops being the whole answer and the question becomes how far the sky moved during the gap in pixels, which depends on your focal length and where in the frame you are looking — the example night below sat at 81 % and still drew continuous arcs, because three seconds at 24 mm is less than a pixel. The arithmetic that turns a gap in seconds into a gap in pixels is a page of its own. A good stacker can close small gaps afterwards, but it can only interpolate across what you did not record.

What to switch off, and why

Focusing, which is where nights are lost

Switch to manual focus, point at the brightest star you can find, magnify it as far as live view goes, and turn the ring until the dot is as small as it will get. Then nudge past and come back, so you know you found the minimum rather than a plateau. Two checks worth making: confirm the dot is a star and not a hot pixel by watching whether it changes size as you turn the ring — a hot pixel will not — and do not trust the engraved infinity mark, because most modern lenses focus past infinity by design and the mark is not a stop. Once it is right, tape the ring.

If your foreground is close enough to matter, you have a genuine conflict, and the honest answers are either to focus at the hyperfocal distance and accept slightly softer stars, or to shoot the foreground as a separate frame at a different focus and blend it. There is no setting that resolves it.

The 500 rule tells you the longest exposure before stars trail — and for this subject you want the opposite of what it protects →

5. Where to point the camera

This is the compositional decision, and it is entirely determined by geometry. The sky turns about one axis; where that axis sits relative to your frame decides the shape of every arc in it.

Finding the pole, in both hemispheres

In the north, extend the line from Merak through Dubhe — the two stars at the end of the Plough's bowl — about five times, and you land on Polaris. Polaris sits about 0.7° from the true pole, which is invisible at any wide-angle focal length and only matters if you are polar-aligning a tracker.

In the south there is no useful pole star. Sigma Octantis is within about a degree of the pole but shines at magnitude 5.5, which is barely naked-eye at a dark site and hopeless anywhere else. Southern shooters use geometry instead: extend the long axis of the Southern Cross, from Gacrux through Acrux, by about four and a half times its own length — roughly 27° — and cross-check it against the perpendicular bisector of the line joining Alpha and Beta Centauri. Where the two lines meet is the pole. The practical consequence is that the hub of a southern vortex is genuinely empty; there is no bright anchor star sitting in the middle of it.

The ground under the sky

Every tutorial in the survey that discusses composition says the same thing, and they are right: arcs alone are not a photograph. A tree line, a ruin, a lone building, a ridge — something with a recognisable edge gives the circles a scale and a reason. Roughly seventy per cent sky is one stated ratio; the real rule is that the foreground should be worth looking at on its own.

If you want to light it, do it on the first frame and the last frame only, at low power, from off to one side. A lighten stack keeps the brightest value at every pixel, so lighting one frame lights the whole picture — and lighting all of them makes the foreground look like a film set.

6. One real night, measured

Everything above is what people recommend. This is what one sequence actually was, read out of the RAW files themselves rather than remembered. Every one of the 380 frames was shot at 15 seconds, f/1.6 and ISO 5000, and the sequence ran from 00:33:32 to 02:30:02 on 13 August 2026.

Two things about that table are worth arguing with, and both are deliberate.

15 seconds and ISO 5000 sit outside the tutorial consensus, which is 30 seconds at ISO 800. Short bright frames are a legitimate trade: you lose star colour and gain resilience, because a car headlight or a gust ruins fifteen seconds rather than a minute. At f/1.6 under a light-polluted sky the background is already the limiting factor well before 15 seconds, so a longer exposure would mostly have collected more town glow. It is not the recipe to start with; it is a defensible answer to that particular field.

An 81 % duty cycle is lower than it should be. Three seconds is a fifth of a 15-second frame and a twenty-fifth of a 75-second one, so short frames pay five times over for the same pause. The arcs came out continuous anyway, because a 3-second gap at this focal length is under a pixel of sky movement near the pole and about 0.9 pixels at the equator — which is the whole point of doing the arithmetic instead of guessing.

7. The lines that are not stars

This is the part of star-trail photography that has genuinely changed in the last five years, and no tutorial written before about 2021 accounts for it.

Jonathan McDowell's satellite catalogue counted 12,881 Starlink satellites launched and 11,093 still in orbit on 31 August 2026 (Planet4589), and Starlink is one constellation among several. CelesTrak put the total tracked population at about 35,000 objects on orbit in early September 2026. Published forward plans across all announced programmes run to well over a million objects. These numbers go stale quickly and are worth checking rather than quoting.

The measured consequence is not speculative. The SPHEREx mission analysed about 6,000 exposures collected in 2025 and found that 73 per cent of them already carried a satellite trail, averaging 2.18 trails per exposure (Borlaff et al., 2026). That measurement matched the prediction made in advance of it, which is the part worth worrying about: the models forecasting a worse sky are working correctly. A study of two decades of Hubble images found 2.7 per cent of its individual exposures crossed, at a typical exposure length of eleven minutes (Kruk et al., 2023), and the Rubin Observatory expects around a tenth of all its images to carry a trail.

Telling one from the other in your own frames

There is a fourth thing you will see and it is not in that table, because what identifies it is timing rather than shape. A tumbling satellite can catch the sun on a flat surface and flash: a bright dot or short dash that exists in exactly one frame. The test that separates a glint from a hot pixel is to look at the frames either side — a hot pixel sits at the same place on the sensor every time, and a glint happens once, at a position in the sky.

The beading has an exact cause. Aviation regulations require an anti-collision light with an effective flash frequency of not less than 40 and not more than 100 cycles per minute (14 CFR 25.1401), which is about once a second. At one flash per second, a 30-second frame records roughly thirty separate bright beads along the aircraft's path. In a lighten stack, where the brightest value at each pixel wins, every one of those beads is welded permanently into the finished picture.

Which is worse depends on where you stand. On 23 July 2026 Flightradar24 reported tracking 153,359 commercial flights in a single day, a record, and 287,364 flights of every kind (Flightradar24). Aircraft fly all night and they fly where people live, which is where most readers of this page will be standing. Satellites are concentrated at dusk and dawn but they are everywhere, including the dark site you drove three hours to reach, and their numbers are growing much faster than air traffic. Aircraft are the bigger nuisance today, near people. Satellites are the bigger problem on trend.

What the usual fixes cost

Sigma clipping — discarding the outlier bright samples at each pixel — does remove satellite trails, and it removes your meteors with them, because a meteor is statistically identical to the thing you are rejecting. Gap filling in a trail stacker bridges missing time between frames; it was never designed to delete an intruder and will not. Content-aware fill invents arcs that do not follow the concentric geometry of the real ones, which is visible the moment anyone looks closely.

The approach that keeps the sky is to find each straight track by measuring it, then repair its footprint from a neighbouring frame in which the same patch of sky was not crossed. What replaces the line is not painted or interpolated — it is what the camera itself recorded one frame earlier or later. How that measurement works, and where it declines to act is its own page.

Here is what that came to on the night at the top of this page. The stacker's own straight-line search was run over the 380 frames at the sensitivity it ships on, with nothing overridden: 309 of the 380 frames carried at least one straight track, 739 tracks were found in total, and 488 of them were removed automatically. Not one was refused because the repair could not reach it. The other 251 were flagged for review rather than deleted, which is the tool declining to act on its own evidence: the more frames a track appears in, the more likely it is something you meant to keep.

8. What happens to the folder afterwards

Stacking star trails is conceptually simple: for every pixel, keep the brightest value any frame recorded there. A star that moved across the frame is bright somewhere new in each frame, so the maximum draws its whole path. That operator is called lighten, or maximum, and it is why star trails are the easiest kind of stacking to do badly and hard to do well.

Two things it does that nobody mentions. It keeps every bright intruder, perfectly, for the same reason it keeps the stars. And because it samples the upper tail of the noise at each pixel rather than averaging it away, a plain maximum makes the sky background noisier as you add frames, not cleaner — the opposite of what stacking does everywhere else in astrophotography. Equalising each frame's sky level before taking the maximum is what stops a brightening cloud or a rising Moon from winning the whole picture. Why a plain lighten stack loses your faintest trails has the measurements.

Stack your own night in the browser The four steps, start to finish →

9. What was not done to these pictures

Being precise about this is the point of the section, because "straight out of camera" is claimed constantly and almost never true.

The result is a magenta-leaning sky, visible noise in the shadows, and hay bales lit an unglamorous red by somebody's torch. All of that is real, and the cast in particular is a decision rather than an accident: a camera's own colour primaries are not sRGB, and converting between them takes a second transform that would sit underneath any judgement you later made about star colour. A finished print would neutralise the cast in about four minutes in any raw editor, and you should — but then it would be a picture about editing, and the numbers on this page would be describing something else.

10. The tutorials, and where they disagree

Twenty-four published guides to star trails were read for this page: photography tutorials, magazine features, two manufacturer pages and the documentation of two stacking programs. Fourteen of them give concrete numbers, and those are below, so you can see the spread rather than take one recipe on trust. Every figure in the table was read from the source's own page, and the two positions this page calls wrong were each re-checked against the original before it said so.

The six real disagreements

Three claims that recur and are simply wrong: that long-exposure noise reduction only affects JPEGs (it is dark-frame subtraction applied to the raw data, and it doubles frame time regardless of format); that you can set focus by the engraved infinity mark (most modern lenses focus past it); and that focus is best judged by minimising chromatic aberration (colour fringing does not vanish at best focus).

If you only remember five things

1. Time makes the arc. An hour is 15°. Nothing else you set changes that.
2. Turn long-exposure noise reduction off. It is the usual reason trails come out dotted.
3. Fix everything and touch nothing. Manual exposure, manual focus, manual white balance, and hands off the camera until it is finished.
4. Start an hour later. You lose an hour of arc and most of your satellite passes.
5. Put something in front of the sky. Circles on their own are a diagram.