Skip to content

Astro test frame shot list

Jump to bottom
gregorydearth edited this page Sep 14, 2026 · 2 revisions

Astro test frame shot list — Hyper Browser RAW inspector

Shoot everything in RAW (uncompressed if the body offers it). Turn off any in-camera lens/shading compensation and long-exposure noise reduction. Manual lenses report nothing in EXIF, so put the lens and aperture in the filename, e.g. wide_tak55_f1.8_a7r3.ARW or wide_23mm_f2_a6000.ARW.

One frame each is enough. Three of the flats is better if it's no trouble — they average.

Bodies: A7R = a7R IIIa (full frame), A6000 = a6000 (APS-C).

A print-friendly copy for use in the field: astro-shot-list.pdf.

Priority 1 — the core vignetting pair (most valuable)

# Body Lens Aperture Shot
1 A7R Takumar 55mm f/1.8 wide open Wide field night sky, plain sky, no foreground
2 A7R Takumar 55mm f/1.8 wide open Flat (see below), same lens and aperture

These two together let me check a fit made from frame 1 against the truth in frame 2. If they match within a few percent, the algorithm is validated and no flat is ever needed again.

Priority 2 — severe falloff (the case the feature is really for)

# Body Lens Aperture Shot
3 A6000 23mm APS-C lens wide open Wide field night sky
4 A6000 23mm APS-C lens same aperture Flat

APS-C-native lenses have barely enough image circle for the sensor, so these should vignette hard. This pair is worth more to the vignetting work than anything else on the list.

Priority 3 — sensor comparison (same glass, two sensors)

# Body Lens Aperture Shot
5 A6000 Takumar 55mm f/1.8 Wide field night sky
6 A6000 Takumar 55mm f/1.8 Flat

Same 55mm as #1 and #2, so the only variable is the sensor: photosite size, black level, green split, read noise. The APS-C body only sees the centre of a 35mm image circle, so expect mild falloff — that is the point. It is the do-no-harm case, where a fit could invent a correction that isn't there.

Priority 4 — the two failure cases

# Body Lens Aperture Shot
7 A7R Takumar 55mm f/1.8 Landscape astro: sky with a treeline or horizon in frame
8 A6000 Takumar 135mm f/3.5 or wide open Lunar, moon centred

#7 breaks the assumption that every annulus sees the same sky, and tells me how hard to make the fit resist it. #8 checks the automatic guard on the other body — the a7R lunar frame already supplied (_DSC2596) covers the full-frame side, so this one is optional.

Optional extra, if it's easy

# Body Lens Aperture Shot
9 A7R Takumar 55mm f/5.6 Wide field night sky, plus a flat at f/5.6

Same lens and body at a second aperture gives a second falloff magnitude to check the fit tracks the optics rather than fitting noise. Nice to have, not needed.

How to shoot a flat

The flat measures how much of an even light actually reaches each photosite, so the only thing that matters is that the light is genuinely even.

  1. Same lens, same focal length, same aperture as the light frame it pairs with. A flat is only valid for the aperture it was shot at.
  2. Target, best first:
    • Twilight sky, 20–30 minutes after sunset, pointed straight up, away from the sunset glow.
    • A white T-shirt or two stretched over the lens hood, pointed at that sky.
    • An evenly lit white wall, camera square to it, shot slightly defocused.
  3. Exposure: aim for the histogram centred around the middle, roughly 50–60%. Do not clip. Underexposed is far better than clipped.
  4. ISO: as low as the body goes (100), since there's no reason for noise here.
  5. Keep the camera in the same orientation as the light frames, and no filters on.

What I do with them

Frames 1–6 validate the synthetic flat-field fit against real falloff, across two sensors and two image-circle situations. Frames 7–8 set the thresholds for when the correction should refuse to run. Nothing camera-specific ships: the feature fits each image from its own pixels, and these frames exist only to prove the fit is right.

Measurements from the frames already supplied are on Astro test frames, and the pipeline these validate is on Astro render mode design notes.

Lenses deliberately left off the list

  • 500mm Soviet mirror lens — mirror designs have their own unusual falloff and a fixed aperture, so it would confound the fit rather than test it.
  • a7R IIIa in APS-C crop mode — keeps only the sweet spot of the image circle, which tells us nothing the a6000 does not already say.
  • A second lunar frame on the a7R IIIa — _DSC2596 already covers the full-frame guard case.

Clone this wiki locally