Plate solving issue

[Al McLean]

Having a problem platesolving in the latest versions (2025.09.04).

An older version (2025.04.30) on my desktop PC solves almost instantly. The newer version reports 'no plate solution found'.

The settings on the stacking tab are identical in both cases but while the desktop finds

'573 stars, 458 quads selected in the image. 383 database stars'

the laptop only finds

'64 stars, 53 quads selected in the image. 94 database sars. 78 database quads required

I guess this failure to detect enough stars is the problem? But what is causing it? The same sub is being used in both cases.

Here's the successful log from the desktop PC:

08:58:01 Using star database D50
08:58:01 573 stars, 458 quads selected in the image. 383 database stars, 306 database quads required for the 0.90° square search window. Step size 0.90°. Oversize 1.00
08:58:02 14 of 19 quads selected matching within 0.007 tolerance. Solution["] x:=-0.044658x+ 1.552384y+ -1550.597518, y:=1.552403x+ 0.044895y+ -2467.873421
08:58:02 Solution found: 20: 56 31.30 +31° 30 05.6 Solved in 0.2 sec. Δ was 26.6". Mount Δα=-17.5", Δδ=-19.8". Used stars down to magnitude: 13.5
09:01:42 457 quads found.
09:14:06 Inspection of: E:\Astronomy\2025\2025-09-07\Single__0041_Bin2x2_120s.fit
09:14:06 Median HFD=5.1 (7.9") Tilt[HFD]=0.37 (7% almost none) Stars=321 Off-axis aberration[HFD]=0.21. Median FWHM=1.6 (2.5")

And here's the unsuccessful one from the laptop:

09:12:04 Using star database D50
09:12:04 64 stars, 53 quads selected in the image. 94 database stars, 78 database quads required for the 1.33° square search window. Step size 0.90°. Oversize 1.48
09:12:16 No solution found! :(
09:13:55 Inspection of: C:\APT_Images\CameraCCD_1\2025-09-07\Single__0041_Bin2x2_120s.fit
09:13:55 Median HFD=3.9 (6.0") Tilt[HFD]=0.49 (13% mild) Stars=63 Off-axis aberration[HFD]=0.46. Median FWHM=1.6 (2.5")

[han.k]

corrected version:
Hi,

The log indicates that at the laptop the image size is 1.33 degrees (height) while at your desktop it is 0.9 degrees. Also the HFD is consequentially 6.0"instead of 7.9" . But that doesn't explain that less stars are detected.

I assume the image does not contain any header with information also ASTAP would behave the same for the same image. Then ASTAP relies on its settings or they are dictated by the program your using like CCDCiel or Nina or something else.

Start ASTAP at the laptop. Load the image. go to the stack menu with CTRL-A and check/set the settings as in attached screenshot. Then the press solve button.

Tell me if this fixes your problem.

Han

[Al McLean]

Many thanks for the prompt reply, Han.

I set those (screen shots) but stiil only found 68 stars and no solution.

I have put the fits file in a dropbox folder, if that helps?

https://www.dropbox.com/scl/fi/gf5vu9i604sbjzwo2qusp/Single__0041_Bin2x2_120s.fit?rlkey=nw90j2734uu235ji0l68jc3wk&st=wbjos7mx&dl=0

Cheers

Al

[han.k]

Hi,

The pattern in the image is strange. It is a binned image from your ASI2600MC camera which should remove any OSC pattern. But it is still there.

Yes an older ASTAP detects much more stars. This is unexpected because I like to make ASTAP better not worse. I will try to find out why this is happening. I do no expect this for other images. It is somehow related to the OSC alike patterns I see in the image.

More later, Han

[Al McLean]

Many thanks for looking into this, Han.

I noticed the chequered pattern - and not just in ASTAP, it also shows on the screen in APT. Something in the camera settings possibly? I'll check.

[han.k]

In ASTAP version 2025.07.25 a hot pixel filter was introduced. This filter sees most of the stars in your image as hot pixels. This because there is dark space between the illuminated pixels. See screenshot. The hot pixel filter works good, so I like to focus why your image shows this weird pattern. Is APT using a special algorithm to keep the OSC colour pattern after binning? It is possible to de-bayer the image and the nebula become red. See second screen shot.

So I suspect CFA-aware binning)

If you force the binning to 2, then it detects 572 stars. So would suggest you force binning at 2 then it will work good.

cs Han

Yes — there is a way to do binning on one-shot color (OSC) sensors that preserves the Bayer pattern, but it requires care.

1. Why standard binning breaks the Bayer pattern

A Bayer CFA (color filter array) arranges pixels in a 2×2 repeating pattern (RGGB, GRBG, etc.).

If you do a naïve 2×2 binning (summing or averaging adjacent pixels), you mix red, green, and blue pixels together, destroying the CFA structure.

That means the result is no longer a valid raw Bayer image — demosaicing algorithms can’t be applied properly.

1. Methods that preserve the Bayer pattern

a) "Same-color binning" (channel-aware binning)

Instead of binning full 2×2 blocks, you bin only pixels of the same CFA color.

For example:

Average all red pixels in a neighborhood,

Average all green pixels separately,

Average all blue pixels separately.

This produces a lower-resolution image that still has a Bayer mosaic structure.

Some astronomy preprocessing pipelines call this color-preserving binning or CFA binning.

b) Skip-pixel binning (decimation)

Another approach is to simply subsample the CFA — e.g., take one pixel per 2×2 block, preserving the Bayer pattern without averaging.

This keeps the CFA structure but sacrifices signal-to-noise benefits.

c) Hardware-supported “Bayer binning”

Some sensors and camera drivers support native binning modes that are CFA-aware.

In this case, the sensor bins same-color pixels before readout, giving higher SNR and smaller file size, while keeping the Bayer pattern intact.

Whether this is available depends on the camera model (common in scientific and astronomy-grade CMOS/CCD cameras, rare in consumer cameras).

1. Practical use in astronomy software

Tools like PixInsight, Siril, MaximDL, AstroPixelProcessor and some vendor SDKs offer CFA-aware binning as an option when preprocessing raw OSC frames.

For example, PixInsight’s IntegerResample process with CFA preservation enabled will downsample while preserving the Bayer pattern.

This is usually preferred if you want to debayer after binning.

✅ Summary:
Yes — you can bin while preserving the Bayer pattern, but you must either:

bin only same-color CFA pixels (CFA-aware binning),

use hardware Bayer binning (if your camera supports it), or

subsample in a way that keeps the mosaic intact.

[han.k]

So this setting:

[Al McLean]

Thanks for the explanation, Han.

I should possibly have mentioned that although the camera is OSC, the image was shot through an H alpha filter. I have done this before without seeing this pattern/issue so I think there's a setting in APT that needs a revisit. In the past all the pixels would show up as red but now thyey don't, not sure why.

And yes, with downsample 2 it solves in 0.1 seconds!

[han.k]

Yes, then you better extract the red channel because the green en blue channel will introduce additional noise in your image. In ASTAP you can do that in the viewer menu TOOLS, Batch processing, Raw colour separation ...

Only why is the image already half size?

[Al McLean]

The image is half size because that option is selected in APT - reduces oversampling, smaller fits files, quicker downloads &c.

[han.k]

That can't be normal binning. I assume this option does none standard binning (CFA aware binning?) .

[Al McLean]

Don't know.
It's an option that the camera provides (it's available in the ZWO capture software as well) but I haven't seen an explanation of how it does it, but when done with normal (i.e. not filtered) lights the colours are preserved so guess it is CFA aware?

[han.k]

If you want good quality H-Alpha images, you should download them in full resolution and then later extract from the raw the red sensitive pixels (so 1/4) only. You can do that in ASTAP or some other programs. This will reduce the resolution from full to 3120 x 2088. So 1/4 of the original pixels. This will produce nice images. So keep from RGGB matrix only R sensitive pixels. The others just introduce noise you want to avoid.