Diagnose · the failure atlas

What went wrong in this frame

Tick what you see. The 29 failure modes that could produce it are ranked by how much of each one's signature you ticked, with the look-alikes to rule out. Each mode's page has the causes in order of likelihood and the fix for each, cheapest first.

It ranks what you tick against written signatures. It does not look at your image, and some failures need the night's logs to tell apart, which each page says.

What you see

Tick everything that fits, at full size where it says so. You don’t need to know the cause.

StarsShape, size and what surrounds them, checked at full size.

Every star is stretched the same way, across the whole frameCompare a star in the center with one in each corner. The dashes point the same way and are about the same length everywhere.
Stars are long streaks, many times longer than they are wideAny bright star at full size. The streak is many times longer than the star is wide, with no round core at either end.
Stars are a sharp dot with a streak attachedA bright star at full size: a round, sharp core, then a tail that starts abruptly. Every star has the tail on the same side.
Stars are short arcs that curve around one point, longer toward the edgesThe corners at full size. The arcs lengthen with distance from one point in the frame and all curve the same way around it.
Corner stars are triangles or little comets while the center is sharpAll four corners at full size. The distortion is about the same in each corner, and it grows with distance from the center.
Stars are worse on one side, or in one corner, and fine elsewhereCompare each corner with its opposite at full size. One edge or corner is soft or stretched and its opposite is sharp.
Stars stretch outward from the center, more toward the edges, like seagullsEdge stars at full size. Each points away from the center of the frame with a flared tail. Center stars are fine.
Stars are round but fat, everywhere in the frameCenter and corners alike at full size, in the linear stack. Stars are round, with no direction to them, and all of them fat.
Bright stars have a ring or a glow around themThe brightest star in the frame at full size. A bright ring or a soft disk sits around it, larger than the star itself.
Stars have a dark ring around themStars against nebulosity at full size. The ring is darker than the background beside it.
Stars have spikes your telescope should not makeThe brightest stars. A refractor making spikes at all, or a reflector making extra ones at an odd angle.
Stars grew large and pale as the nebula was stretchedThe stretched image beside the linear stack. Stars that were small in the stack are now wide and washed out.

BackgroundThe empty sky between stars, after a hard screen stretch.

The background is brighter on one side and fades across the frameThe whole frame, zoomed out, with a hard stretch. The ramp runs in one direction and is not centered on the frame.
The corners are darker than the centerThe whole frame, zoomed out, with a hard stretch. The darkening is round, centered, and about the same in each corner.
The whole image has a green tint, worst in the backgroundThe empty sky between stars, after stretching.
Faint diagonal streaks of noise run through the stacked imageThe background of the stack at full size, stretched hard. The streaks are faint, parallel, and run at one angle across the whole frame.
Horizontal or vertical stripesThe background at full size, stretched hard. Regular lines running straight across or straight down, never diagonal.
A glow in one corner or along one edge, the same in every frameThe same corner of several subs, stretched hard. The glow has structure and does not move or change between frames.
The background is pure black and the faint outer nebulosity is goneThe histogram. The left side is cut off at zero instead of showing a hump above it.
Washed out and low in contrast, with a bright gray backgroundThe whole image. The empty sky is a light gray, and the histogram's hump sits well away from the left edge.
The background is smooth and blotchy, like wax, with no grainEmpty sky at full size after noise reduction. No fine grain at all, only soft blotches.
Colored speckle in the backgroundEmpty sky at full size, stretched. Single pixels in different colors where the sky should be an even gray.

The rest of the imageLines, dust, color and edges, in one sub or in the stack.

A single sharp line across one frame, with the stars unaffectedOne sub. The line is thin and straight, and the stars on either side of it are round.
Dust donuts appear doubled after flats are appliedA calibrated frame beside the raw one. A donut that was there once now appears twice, side by side.
The corners come out brighter than the center after calibrationA calibrated frame, stretched hard. The corners are lighter than the middle, the reverse of vignetting.
Colors are garish and unnaturalThe nebula and the stars. Hues are pure and loud, with no muted tones between them.
Odd frames, streaks or seams along the bordersEach edge of the stack, zoomed out. A band along the border looks different from the rest of the image.

The nightWhat changed over the hours, from the logs and the time-sorted subs.

The guide graph shows spikes or excursionsThe guiding software's graph for the hours you imaged.
Subs were fine, then worse from the meridian flip onwardSort the subs by time. The change is a step at the flip, with no slow decline before it.
Subs get softer as the night goes onStar size (HFR or FWHM) per sub, plotted over the session. It climbs steadily.
Halos around bright stars grow through the sessionThe same bright star in an early sub and a late one.
The star count drops suddenly while star size stays the sameStar count and HFR per sub, over the session. The count falls off a cliff and the HFR line does not move.
Only one frame is affected; the ones before and after are fineBlink through the subs on either side of the bad one.
The same in every frame, including the first of the nightBlink through subs from the start and the end of the night. The feature does not change.
It changes from frame to frame as the sky changesCompare subs taken hours apart. The brightness, or its direction, shifts.
Dew or frost on the front lens or the corrector plateShine a red light across the glass at a low angle.

When it appearedWhich step brought it on.

The linear stack looks fine; the problem appears when you stretchOpen the unstretched stack and look at the same spot with a screen stretch only.
It appeared after sharpening, deconvolution, noise reduction or star removalUndo the last step and look again.
It appeared, or got worse, after applying flatsThe same frame calibrated with flats and without them.

Ranked failure modes

Tick what you see and the failure modes that produce it are ranked here, each with the look-alikes to rule out and a link to its page.

All 29 failure modes

Grouped by where they tend to turn up, on a first rig or on a new one: the mechanics of tracking and focus, the quality of the data, and processing.

Acquisition mechanics

usually months 1–3 · 7 modes

Elongated starsEvery star in the frame is stretched into a short oval, all pointing the same way. The center looks as bad as the corners.
Star trailsStars are long streaks, many times longer than they are wide, with no round core at either end.
Stars trail partway through a subIn one frame, every star is a sharp dot with a streak attached. Something moved the rig partway through the exposure.
Worse after the meridian flipGuiding and star shapes are fine until the flip, then get worse at once and stay worse for the rest of the night.
Field rotationStars are short arcs curving around one point in the frame, longer toward the edges.
Bloated, soft starsEvery star is round but fat, across the whole frame, with no stretch in any direction.
Unwanted diffraction spikesBright stars carry spikes where the telescope should not make them: any spike from a refractor, or extra spikes at an odd angle.

Acquisition quality

usually months 3–12 · 13 modes

Triangular corner starsUnverifiedCorner stars are triangles or little comets while the center is sharp, and all four corners look about the same.
One side softOne edge or one corner is consistently worse than the rest of the frame, and the others are fine.
ComaStars stretch outward from the center, more toward the edges, flaring like seagulls. The center is sharp.
VignettingThe corners are darker than the center, evenly, in a ring around the middle of the frame.
Halos around bright starsBright stars have a ring or a soft glow around them, larger than the star itself. Faint stars look normal.
GradientsA smooth ramp in brightness runs across the frame, brighter on one side.
Amp glowA structured glow sits in one corner or along one edge, the same in every frame.
Walking noiseFaint diagonal streaks of noise run through the stacked image, all at one angle.
BandingHorizontal or vertical stripes cross the background.
Lost to cloudThe number of stars in each sub collapses while the stars that remain stay as sharp as before.
Satellite trailsA single sharp line crosses one frame, and the stars around it are unaffected.
Dew on the opticsStars get softer and halos around bright stars grow through the session, sub after sub.
Flats made it worseAfter calibration, dust donuts appear doubled, or the corners come out brighter than the center.

Processing

usually year 1 onward · 9 modes

Green castThe whole image has a green tint, worst in the background.
Clipped blacksThe background is pure black, and the faint outer nebulosity is gone.
Over-stretchedThe image is washed out and low in contrast, with a bright gray background.
Star bloat from stretchingStars grow large and pale as the nebula is stretched.
Dark halosA dark ring surrounds stars, or runs along the edges of bright structure.
Plastic backgroundThe background is smooth and blotchy, like wax, with no grain.
Color noiseUnverifiedColored speckle covers the background: single pixels in different colors where the sky should be even.
Neon colorsColors are oversaturated and unnatural across the image.
Edge artifactsOdd frames, streaks or stacking seams run along the borders of the stacked image.

How this is computed

score(mode) = sum of weight(symptom), over the ticked symptoms the mode lists
weight       = 3  the mode's own signature
               2  its look-alike test, or a strong companion sign
               1  a weak sign it shares with look-alikes
order        = score, high to low; ties in the atlas's own order

Every symptom a failure mode can produce carries a weight, and a mode’s score is the sum of the weights you ticked, shown against the most it could reach. The weights are ours. They were set so that the distinctions the atlas draws come out right, such as fat stars that look fine in the unstretched stack ranking star bloat from stretching ahead of focus. No measured data is behind them yet.

Ties go to the mode the atlas lists first, which runs in the order failures usually appear. The ranking is a starting point for the mode pages, which carry what a single frame cannot tell you.

Source: docs/10-failure-atlas.md § 2 (signatures, causes, look-alikes) and § 4 (each row is a rule for the Diagnostic). 39 symptoms; content and ranking in packages/data/src/diagnose, unit-tested.

Last updated 2026-10-12

01What you see

02Ranked failure modes

03All 29 failure modes