Newton's rings on astrocam, the glass cover shows itself
Peter: "the glass cover for the camera is very thin and flat. probably the cause of the newton's rings."
How the rings turned up
The deep de-rotated sweep for 10-09 turns every frame so the stars stay put and averages the night. Once the sky-glow model was fixed, the background was flat enough to stretch hard, and it showed five or six concentric rings around the pole, steady all night, plus faint vertical stripes.
Rings around the pole are what de-rotation does to anything fixed to the camera: a blemish at some distance from the pole is swept round a circle of that radius. So the rings said only that something was camera-fixed, not what. Taking the median of 61 clear frames in the camera's own frame (the stars move between them and drop out) showed the real pattern directly: rings in the camera, centred somewhere else entirely.
What the pattern is
- Centre at pixel (973, 850) on the binned frame, 9.4 degrees from the lens axis. It is not the pole (697, 475) and not the lens axis (1156, 639), so it is neither the sky nor the lens.
- About seven bright rings out to 33 degrees from the centre, closing up as they go out: the first bright ring is 10.5 degrees out, the next ones about 5, then 4 degrees apart.
- Faint: about plus or minus 0.3% of the sky's brightness near the centre, up to about 0.8% at the edge. Invisible in any one frame; it took a whole night and a hard stretch to see.
- The vertical stripes are a separate thing: small offsets between sensor columns.
Why a thin flat cover makes rings
The camera is focused on the stars, so the cover glass in front of it is completely out of focus. Each pixel looks through the cover in one direction. Some light bounces twice inside a thin layer of the cover before going on, and when it rejoins the light that went straight through, the two interfere. Whether they add or cancel depends on the angle the light crosses the layer at, so equal angles give equal brightness: rings, centred on the direction square to the layer. Physicists call these fringes of equal inclination; Newton's rings are their cousin, from the same thin layer of air between two pieces of glass.
That reading makes three checkable claims:
- The cover is tilted about 9.4 degrees to the lens. The ring centre is where the cover's surface is square on, and it sits 9.4 degrees from the lens axis, towards the bottom left of the frame.
- The layer is about 10 micrometres thick, which is air, not glass. About six and a half rings from the centre to 33 degrees needs a layer about 11 micrometres thick if it is air. Even very thin cover glass is ten or more times thicker, and would crowd dozens of rings into the same space, finer than sky light can show. The layer doing this is the air gap under the cover: Newton's rings in the literal sense.
- It is faint for a reason. Each glass surface reflects about 4% of the light, so the two-bounce light is weak to begin with, and sky light mixes many colours whose rings fall in different places and mostly blur each other out. About 1% left over is the right size.
One thing does not fit a perfectly even gap: the rings close up more slowly than an even layer predicts, and they are squeezed towards the top right. The gap is probably a slight wedge or bow, as you would expect for a thin cover held at its edges.
Why it matters, and the fix
At 1% it is far below anything you would see in a picture, but it is the same size as the faint structure the deep sweeps are after, and it is a pattern a flat field would need to carry for any careful brightness measurement. It may not hold still: the gap can change with temperature, and 10-09 had a lens moving during the night, which left a trace of the rings at the right edge after correction.
The pipeline now builds this map every night and subtracts it before de-rotation (astro c2f56bb). On 10-09 the rings and stripes are gone.
Still to check
- Colour. Interference rings scale with wavelength: the red rings should sit about 13% further out than the blue ones. A lens or vignetting effect would not do that. This is the decisive test.
- Night to night. Do the rings stay put, drift with temperature, or move when the cover is touched?
- The tilt. Is the cover in fact about 9 degrees off square to the lens?