chromatic aberration? RASA

[Datalord]

I have my RASA set up in the backyard again and started gathering some data. This time around I want to fix a problem I had in my old pictures. I think it is chromatic aberrations, but how do I fix it? Why is it coming? When I focused the scope, it looked like the collimation was fine.

ignore the egg shape, that's a different problem. 🙂

[ollypenrice]

You mean the blue side and the red side of the star? This does sometimes come up. This discussion https://pixinsight.com/forum/index.php?topic=9513.0 involves a mono camera, filters and refractor but the original problem seems to be the same with a red side and a blue side. Various explanations are put forward and yet, given the similarity with your issue from an OSC, it strikes me that the problem may be unconnected with both channel alignment or filter tilt.

You say 'ignore the egg shape, that's a different problem,' but can you be sure of this? You have a considerable distortion in the stellar image. Might not this distortion be arising from an optical issue which is also sending the blue end of the spectrum to one side and the red end to the other? In principle that strikes me as being perfectly credible. Note that the most intense colour is found along the long side of the star, perhaps implying a connection.

How is this effect distributed across the whole image? Does it follow the same orientation? Is the colour 'sidedness' consistently related to the elongation?

There is a processing 'fix' (correctly described as a cheat by Juan in the PI thread) which can mitigate this problem but I think you're after the root cause?

Olly

[Datalord]

Very interesting. 

You're right, the egg might actually be part of the problem. I have another set from last night with considerably rounder eggs (fixed the chicken), so I'll check it out. I will say this has been a problem on all images I have taken from the very beginning with the RASA/QHY combination (this time I have different spacers than last), I just didn't "care" the same way I do now.

So, I went back and looked at my very first image of the rig, with the same target and with very bad focusing.

What's striking to me is that the colour is wrong on exactly the same sides, on round stars. Part of the story is that the camera malfunctioned AFTER this picture, and I got a full replacement, so the camera can't be the root cause. The rotation of the camera on each image is also different by about 45 degrees.

Another difference: On the original image I used just an IR cut that comes with the RASA, on the one above I used the LP RASA filter.

🙄

[ollypenrice]

1 hour ago, Datalord said:

Very interesting. 

You're right, the egg might actually be part of the problem. I have another set from last night with considerably rounder eggs (fixed the chicken), so I'll check it out. I will say this has been a problem on all images I have taken from the very beginning with the RASA/QHY combination (this time I have different spacers than last), I just didn't "care" the same way I do now.

So, I went back and looked at my very first image of the rig, with the same target and with very bad focusing.

What's striking to me is that the colour is wrong on exactly the same sides, on round stars. Part of the story is that the camera malfunctioned AFTER this picture, and I got a full replacement, so the camera can't be the root cause. The rotation of the camera on each image is also different by about 45 degrees.

Another difference: On the original image I used just an IR cut that comes with the RASA, on the one above I used the LP RASA filter.

🙄

When boosting star colour I sometimes get an effect very like your second image. I don't see it till I've done pretting aggressive boosting but it must be latent in the stack. This is with mono camera and refractors. I'm afraid I don't know how it arises.

Just a thought but I wonder if your OSC anti-aliasing filter might be contributing to this? I've only ever used OSC from a CCD camera and debayered it in AstroArt where the anti-aliasing is adjustable. I don't remember ever adjusting it from default but it just might be playing a rôle here. Do you have access to your anti aliasing settings? If so it wouldn't take long to see if they have any effect.

Olly

[andrew s]

At what altitude were these taken as the atmosphere is chromatic? It gets worse as you approach the horizon. Easy to test by taking images at the same altitude east and west of the meridian assuming an eq mount.

The chromatism is along the parallactic angle.

Regards Andrew 

Edited July 1, 2019 by andrew s

[Datalord]

4 hours ago, ollypenrice said:

When boosting star colour I sometimes get an effect very like your second image. I don't see it till I've done pretting aggressive boosting but it must be latent in the stack. This is with mono camera and refractors. I'm afraid I don't know how it arises.

So, this time around I took a photo from last night, just one frame. background extraction and a nuke in PI:

Very apparent, but this time on a different target, which is a different place on the sky (Eastern Veil vs M31 in the previous). The colour location changed with roughly that same angle as the difference in the target in the sky.

3 hours ago, andrew s said:

parallactic angle

I have no clue where that is, even after trying to read up on what it is? The above picture from last night is taken near zenith.

[andrew s]

Look at this image. The chromatic aberration would be along the the blue dotted arrows. Red towards the arrow head blue on the south pole side.

Regards Andrew

[ollypenrice]

3 hours ago, andrew s said:

At what altitude were these taken as the atmosphere is chromatic? It gets worse as you approach the horizon. Easy to test by taking images at the same altitude east and west of the meridian assuming an eq mount.

The chromatism is along the parallactic angle.

Regards Andrew 

I'm not very good at this! However, would we expect to see the red side and the blue side of the stars lying either side of a line of right ascension passing through them if an exposure were made at fairly low elevation with an EQ mount pointing due south? That would be an easy test to perform. We would firstly align the chip along RA and Dec (which many of us do anyway for ease of reframing) then take the shot pointing at the meridian when the parallactic angle is zero. In this situation the atmosphere, if responsible for the aberration, would create a red side and a blue either side of a line parallel with the chip's RA side.

Olly

Edit, We crossed in the post. I think my notion is confirmed?

 

Edited July 1, 2019 by ollypenrice

[andrew s]

This gives an idea of the scale from here http://www.ls.eso.org/sci/facilities/lasilla/instruments/feros/Projects/ADC/index.html

The left scale is arc seconds and the bottom scale zenith angle.  Blue, green and red are  at 350nm, 500nm, and 920nm

Regards Andrew

[glowingturnip]

doesn't help with fixing the root problem of course, but you can fix this easily in post-processing - split out the R, G and B into individual greyscale files, star-align them to each other (use the green for reference since you have twice as much green on your DSLR), then re-combine them to RGB

[Datalord]

Andrew, just to make sure I don't make assumptions on my fail to understand the geometry, I made an astrometry solve of the exact image (non-egg) of the time above, mapped that to the skychart at the time and drew the frame with "up":

Does this match your parallactic angle theory?

1 hour ago, glowingturnip said:

doesn't help with fixing the root problem of course, but you can fix this easily in post-processing - split out the R, G and B into individual greyscale files, star-align them to each other (use the green for reference since you have twice as much green on your DSLR), then re-combine them to RGB

I'll try that!

Edited July 1, 2019 by Datalord
non-egg image clarification

[Datalord]

1 hour ago, glowingturnip said:

use the green for reference since you have twice as much green on your DSLR), then re-combine them to RGB

Uh, this did help, I think, but it reversed the issue. This is from the original egg shaped stack, where blue was on top and red was below. After this procedure, the sides are reversed. 😖

 

 

 

Edited July 1, 2019 by Datalord

[glowingturnip]

55 minutes ago, Datalord said:

 

Uh, this did help, I think, but it reversed the issue. This is from the original egg shaped stack, where blue was on top and red was below. After this procedure, the sides are reversed. 😖

 

 

 

how does it look with a normal stretch ?  that's a strong stretch you've got there.

Could always layer version one over version 2 and adjust the transparency till it looks good  🙂

[andrew s]

1 hour ago, Datalord said:

Andrew, just to make sure I don't make assumptions on my fail to understand the geometry, I made an astrometry solve of the exact image (non-egg) of the time above, mapped that to the skychart at the time and drew the frame with "up":

To first order I think it does.  The simplest test is to image a star either side of the meridian at the same (low) altitude.  You should get the red/blue in the same place wrt the star field. If it is say due to say "wedge" of the optics it should flip 180 with the optics and camera.

Regards Andrew

[Datalord]

2 hours ago, andrew s said:

The simplest test is to image a star either side of the meridian at the same (low) altitude.

Well, that implies I can find the same target at low altitude on either side of the meridian while it is actually dark. I'll get back to that sometime in November...

14 hours ago, ollypenrice said:

There is a processing 'fix' (correctly described as a cheat by Juan in the PI thread) which can mitigate this problem but I think you're after the root cause?

I think I will take Andrews explanation as the root cause, because everything fits. As such, the "cheat" is probably the only way to fix the atmosphere. I tried it...

before:

After:

If that's a cheat, I'm willing to live with it. 

Thanks gents, this is pretty awesome to find a solution to. Now I just have to go back and reprocess everything I ever made with the RASA... 🙂

[ollypenrice]

I wonder if the very fast optics exaggerate the problem? They do exaggerate most problems but this would need thinking about.

If you use the 'split channels and re-align all to green' method in a program which re-shapes the curvature of an image, rather than just re-aligning in x-y and rotation, you shouldn't need to start again from scratch. Just split the channels, save them separately, align each R and B to G and re-combine. I would do this in Registar but I think that PI can do that, as can APP - though I'm not certain. Registar certainly does it. My own experience with Registar, though, is slightly at odds with theory in that I find it best to make red the definitive channel and align to that. I find exactly the same thing when aligning channels in AstroArt. I make red definitive and G and B 'flexible.'

I'd be happy to run an image through Registar if you don't have it and would like to try it. Just dropbox me an OSC image and I'll see what happens.

Olly

Edited July 1, 2019 by ollypenrice

[Datalord]

I think it does. I've only had this problem on the RASA, not the WO. Then again, the WO is tiny in comparison.

I tried the split above, in several attempts, but it never really got me to where I wanted.

I sent you a DM with a picture to try.

[ollypenrice]

This is my reply to Datalord's PM to keep this interesting thread complete.

I couldn't open your JPEG in either Ps or AstroArt, both of which gave me 'incompatible file format.' However, I took a screen grab, split the channels and realigned G and B to R in Registar. Registar did move the G and B slightly but there was no improvement when they were combined.

Thinking about Andrew's parallactic angle test, I don't think you need to take three images of the same object when east, due south and west. It might suffice to take a single exposure due south at a moderate elevation then slew to the east and then the west in RA only, taking an exposure in these two positions. If you found that the axis separating the red and blue sides of stars rotated according to Andrew's diagram the atmosphere would be confirmed as the villain.

Spacial awareness isn't my strong point so please run a critical eye over this!

Olly

Edited July 2, 2019 by ollypenrice
Typo