Master flat - alternate dark pixels

[woodblock]

The camera I'm using is a ZWO ASI1600mm - mono camera and I've produced a master flat file from a set of flats and dark flats.  Bin 1x1. I've spoken about this in the camera section but this question is better placed under software. I think I've processed it correctly. I was idly looking at the master flat in MAXIMDL and casually twiddling the zoom looking roughly at the centre of the image when I noticed something strange about the pixel colours. It appears that they are alternately dark and light as you can see in the attached jpeg.  I've attached two screenshot images. The first is a heavily magnified section of the middle of the image and the other is the whole image with what MAXIMDL denotes as medium screen stretch.

 

is there something wrong here or is it just an artifice of the screen stretch?

 

Cheers

Steve

 

[paul]

I don't think you can be processing them as mono images. Perhaps you have some sort of debayering going on, a left-over setting from processing DSLR images perhaps?

 

A flat should look smooth and the variation between individual pixels will be very small. 

 

This is one of my ASI1600 master flats with auto stretch:

Closeup you can see the general smoothness - the stretch exaggerates the differences but probably only 1 or 2 ADU between the pixels.:

 

Edited April 12, 2022 by paul

[woodblock]

Yes. Thanks Paul. I must admit when I saw the chequerboard I immediately thought it looked like a colour image.  I'll have to delve into the settings.

Steve

 

[ApophisAstros]

Try using a different stacking method to see if results are the same?

Roger

[paul]

Yes, good idea  - something like deepsky stacker👍

[woodblock]

The pattern seems to be in the raw data. I've taken the flats straight from the camera and stacked them using "Average".  I've checked manually that it's done the calculation correctly. I've done this by taking actual pixel values and working out the average over 4 images and it's correct.

 

At the moment I'm looking to see if I've got a setting wrong in APT when the frame is taken.

 

[woodblock]

I've checked all the camera settings in APT and can't find anything obvious. I've tried using the direct camera driver instead of doing it through ASCOM. I've also got a set of flats using NINA and stacked them. They all show the same effect.

 

In fact the effect is quite small. The difference between the dark and light pixels is about 1000 on a value of about 28000. For example looking at two adjacent pixels I get 28513 and 27430 roughly in the middle of the frame.

[paul]

If you are getting a difference of 1000 between pixels on an unstretched image then that is a problem. For an average adu of 32000 that is too much. 

I suspect you are directly measuring stretched values. 

[woodblock]

I'm pretty sure I'm not measuring stretched values.   The pixel values don't change when I change the stretch even though the image on the screen changes.

 

This morning I took 10 flats and stacked them using the average method. I checked that the average is calculated correctly by taking a sample of pixels and manually calculating the average value by laboriously getting the value of those pixels from each of the 10 original image files. It works out spot on.  That also gives me confidence that I'm not looking at stretched values. I'm pretty sure that the problem is already in the original image files when the image is collected from the camera. I'm looking for a setting which might do that.

 

By the way, at present there is no filter on the camera. I don't know how that affects things. I didn't think it would matter. I think the pictures you sent me were taken with an H filter.   There is actually a hint of light/dark alternate pixels on the image you sent but that might just be a statistical thing.

 

Thanks for your help too and Roger as well.

 

 

[paul]

 

4 hours ago, woodblock said:

By the way, at present there is no filter on the camera. I don't know how that affects things. I didn't think it would matter. I think the pictures you sent me were taken with an H filter.   There is actually a hint of light/dark alternate pixels on the image you sent but that might just be a statistical thing.

 

I have to eat my words🙂. The variation between pixels is more than I expected (TBH I never looked so zoomed in to the flats before and expected because of the smoothing effect they would be little difference pixel to pixel).

 

I can't see the filter is relevant to the principle. I exported my flat to a 16-bit FIT and used a FIT viewer with auto-stretch and it looked like this:

 

Zoomed in: 

So the software has an impact on how the stretched image looks but there is variation. Nothing like your extremes between pixels though. I'll PM you with a file location for you to download it to compare in maximDL against yours.

[woodblock]

Thanks Paul

[woodblock]

I've contacted the supplier who asked me to take a series of test images using ASICAP which I've done and sent to them. They say it all looks ok. I guess I'll have to accept that.