A raw file input (CR3 from a canon Ra) returns a different (and wrong) sized image to fits files.
Using a mix of raw and fits files then fails as the image sizes do not match.
If I convert the raw files myself (to tiffs with the correct size) then I can use a mix of tiff and fits files without problems.
I suspect the problem is that siril is using libraw to convert the raw file and is running into this issue.
I am using a Canon Ra via stellarmate. Mostly I capture fits files, but it is vastly faster to capture bias and flat frames just using fast multiple exposures on the camera, this way is not only much faster, but helps keep the camera temperature from rising too fast
This is a know issue and the image size depends on the software used. Siril depend on libraw indeed, so we do what experts on the subject decided to do. The suggestion from Siril would be to take all your images using the same system. I don’t know how you can crop images to the exact same pixels.
Well in this case libraw has decided to behave differently to every other app and the camera manufacturer as well. I have a workaround in a python script which opens the raw file in rawpy (which is a pyton wrapper for libraw), crops it and saves it as a tiff. I then give the tiff to Siril and is is quite happy, just irritating that I have to effectively keep a copy of every raw file.
Btw, it is not exclusive to LibRaw - RawSpeed (used by darktable for all files expect CR3s) also defines its own “maximum data” crops.
One thing to keep in mind - these are just libraries designed to give you access to raw data (and metadata like vendor crops separately), the end product is then responsible for the rest.
Ah! interesting. While there appears to be sensible image data in the area below / right of the canon defined crop - as used by libraw - it seems a bizarre decisison to me. This is the same sensor as the canon 5D mk4, and it too defines the image as 6720 x 4480 as the maximum resolution.
There is the option in the later versions of libraw to set the crop - C++ API | LibRaw, but without burrowing into Siril this won’t help sadly.
The main reason for doing this is that when capturing flats and bias frames, the fastest ekos / indi will go is about 1 frame every 3.5 seconds, and while doing this the camera quickly heats up to around 25C above ambient, which adds a lot of noise. By using fast repeat in camera I can easily capture 16 or so frames in around 1 - 2 seconds and avoid heating the camera up.