Trailing or Coma

[twobleak]

Hi All

I recently picked up a Samyang 16mm f2 for my Canon 90D and I was trying out some test shots at f2.8, iso 1600 and 10 seconds. I believe this length of exposure should be short enough to avoid star trails, but looking at one of my images closer I can see what looks like trailing or possible coma issues (or both) - I'm not quite sure really. Could someone offer a suggestion as to what I might be doing wrong or is the lens not a particularly good copy? I have attached a TIFF - just a straight export out of Lightroom. Thanks.

IMG_8930.tif

[vlaiv]

That is mostly lens aberrations.

You don't seem to have significant trailing at all. Look at stars at center of the frame - that is sharpest part of the lens - they are mostly round.

If you look at MTF of that lens:

It clearly shows that at 30 lpmm (equivalent to 33µm pixel size - gray lines show sagittal and tangential MTF) - there is significant sharpness issue at 13mm away from center -  you are using APS-C sensor with diagonal of 28mm - that is 14mm distance from center to corner.

No wonder that stars in corners are astigmatic (dominant tangential component)

Even at 10 lpmm (red graph) - you'll have some elongation in the corners.

This also affects distant light sources that are ground based (hence not moving like stars):

If you create image with larger pixel size - you simply won't notice that. Lens are not diffraction limited optics like telescopes and have issues even on optical axis - stars are not perfect pin points (within diffraction limit). This simply means that you need to use them to produce lower resolution images than you would otherwise with a telescope - just bin your data to make pixel size much larger and you'll have nice looking image (or alternatively - just resize this image so that it can't be fully zoomed in - if it bothers you).

I think that with lens it is always better option to shoot with higher focal length lens and then do mosaics to achieve wider field. This is problematic when you don't have tracking - so you'll have to settle for this kind of resolution.

 

[alacant]

5 hours ago, twobleak said:

what I might be doing wrong

Hi

Nice shot. Love the foreground terrestrial stuff.

I don't think you're gonna get a 16mm lens corner to corner over aps-c with anything wider than f8, and even then some.

No problem as it's easily corrected in software.

Cheers and HTH.

 

 

 

 

[twobleak]

Thank you both for the replies, that gives me some peace of mind that the lens and my technique are not all that bad! 

38 minutes ago, alacant said:

No problem as it's easily corrected in software.

 Can I ask what you did to correct that in software? 

I did notice that once I had stacked all my shots (around 50 or so) in Sequator, the odd star shapes noticeably looked a lot better.

[vlaiv]

3 minutes ago, twobleak said:

Can I ask what you did to correct that in software? 

Actual correction of optical aberrations in software is very hard.

What above algorithm does is simply replaces what is perceived as star - with nice circular shape. It is called morphological filtering. That would not correct underlying image - only stars. You would not be able to correct daytime blur or nebulosity in the image. However - these are outer parts of the image and when sky is dark - any defects in background go unnoticed so you can use this technique to make your stars nice and round.

@alacant will no doubt explain it better and explain software and process used to correct star shapes.

[alacant]

1 hour ago, twobleak said:

what you did to correct that in software

NP. I used StarTools.

1 hour ago, vlaiv said:

algorithm does is simply replaces what is perceived as star

Mmm. I wonder;)

Please post your solution so that the OP can choose which is best. My method took under a minute. 

Cheers

Edited June 4, 2021 by alacant

[vlaiv]

5 hours ago, alacant said:

Mmm. I wonder;)

Wander what?

5 hours ago, alacant said:

Please post your solution so that the OP can choose which is best. My method took under a minute. 

There is no my solution. There are well known deconvolution algorithms which all have their drawbacks.

In general case, when PSF is not known and in presence of noise - you simply can't solve deconvolution problem. There are only approximate solutions and all of them increase noise - provided that you can guess PSF sufficiently accurately which is not the case here.

Solution is to use different type of optic - one that does not have that level of aberration.

Alternatively - since this is a hobby - one can resort to using morphological transforms to round the stars, and that is fine - as long as one understands what they are doing (that it won't fix whole image just stars) and accept the process as is.

If this problem was readily solvable - then there would be no point in purchasing field flatteners or coma correctors - as software would solve that at much lower cost.