Star color fringing

[cs_pixinsight]

I recently purchased an Astronomik CLS-CCD clip in filter for my Canon DSLR to use when imaging from home in light polluted suburbia.  It reduces my sky glow by a full 2 stops, so it's definitely working as expected in this regard.  However, there seems to be a side effect when using the filter - my red channel has larger stars than my green and blue channels.  This doesn't happen with the same lens without the filter, so I'm wondering if this is common with light pollution filters, especially the CLS-CCD.  I was using my Canon 28-70 2.8L @35mm F4 during the filter testing which is not an APO lens, but since the lens without the filter is fine there much be something going on that makes the red channel bloat while using the filter.

Secondarily, I'm wondering about ways to fix the issue in PI during post processing.  I've tried MorphologicalTransform on just the red channel using a star mask to protect the background with some success, but an unfiltered sub is still better.  Perhaps I just need to fine tune this process a bit more, but I'd like to hear if others have this same issue and how you go about fixing it - either during capture or in PI after imaging.

Craig

[pfile]

that is interesting - i have noticed the same thing but never ascribed the behavior to the filter. i pretty much always use the filter so i don't have anything to compare it with.

i should look more carefully at my later images taken at f/7. i know i definitely have this problem at f/4.

the problem might be more pronounced at lower f/ ratios. i know that with the astronomik Ha filters the transmission is lower at lower f/ratios, and i wonder where the extra energy goes (haloes/reflections?).

i have in fact used MT just as you are describing to pull in the star shapes in the red channel... it should work. on one image the red channel was not only out of focus, but the star centroids were actually shifted! no idea how that happened but what i did was to realign the R channel with the G channel using StarAlignment.

[astropixel]

I see the same using a Canon 200mm prime with or without a filter and attribute it to focus. Similarly, I have images in which fringing is bluish green - that is, the other side of focus. I put it down to the visual limitations of using live view to obtain focus. A purpose built focus mask improved things significantly. This particular lens begins to perform at f/6.3 in terms of focus and field flatness. It is tempting to want to use these lens wide open. At f/2.8 - f/5.6 coma / fringing is present in diminishing degrees.

[pfile]

i thought it was focus as well... i made a bahtinov mask by laser-cutting a 72mm lens cap; ever since then pretty consistently the red channel is out of focus after using the b-mask.

this could be a result of having a modified camera and starting to see the effects of near IR not coming to focus in the same place that G and B do.

i've heard more than one person claim the 200L is good at f/2.8, but i've always used f/4.

actually if it turns out to be less prevalent at f/7 it's probably because the critical focus zone is a lot wider at that focal ratio, rather than having something to do with the angle of incidence on the filter.

[cs_pixinsight]

Thanks for the replies.  It appears I'm not alone in this regard.  I use Nebulosity for focusing and I'm fairly certain it's about as good as I can get with a camera lens. 

@astropixel:  I tried live view with a friends camera using a 22x loupe on the LCD, that seems to work well, but I didn't take any pics to really determine if it was a viable method.  How do you focus in live view?

@pfile:  My camera is unmodified, so I don't think IR is causing this.  If you get a chance I'd like to know if you had better red star sizes at F7 vs F4 apertures. 

I'll try to fine tune my MT settings to see if I can get rid of this fringing entirely.  Have you tried heavier deconvolution in the red channel to help get the star sizes down?

Thanks,
Craig

Craig

[astropixel]

My solution was to extract all 3 channels of the integrated image and increase the smaller star size (R channel in this case) to fit the other channels, using Deconvolution - Midpoint: (Min+Max)/2 Amount as required, then recombine.

Increasing star size is not optimal, but it was easier and less destructive than eroding the other channels, even with a mask. I followed up with one or two applications of Erosion to the newly combined image.

@Craig - I don't focus in live view anymore - I found it too unreliable.

@pfile - I found the b-mask not accurate enough for camera lenses - the adjustment is 1/10mm in and out of focus - on my lens anyway. I removed the grating and retained the main obstructions - a combination of the Carey mask and Chris Lord's dissertation on Fraunhofer diffraction/mask design.

[troypiggo]

I'm watching this thread with interest but from a different viewpoint.  I have been attempting narrowband images lately, and with the Hubble palette getting those magenta halos around stars because the R and B channels are being stretched so much.  I'd like to be able to have another image with stars closer to RGB colours and blend those into the narrowband nebulosity image.  Any tips on doing that?  I'm assuming star masks should be used, but not sure how to actually achieve the mixing of 2 images.