FWHM

My understanding (and I may be wrong) is FWHM is pixels not arcsec/pixe

My bad...FWHM is arcsec, not arcsec/pix....It can, of course be pixels, but moist people use arcsec as that is universal.  Of course you can compare from different nights.  Images that have over 20 hours of exposure are capotured over multiple nights.  When you look at your subs to see which you will keep based on FWHM, you look at them all.  It does not matter on what night you took them.  Higher FWHM will be on nights of poorer seeing.  But theory is not the point of this thread.  The question what FWHM is a keeper and what is a throw away is the point.

Seeing really is the great equaliser. There are a couple of ways you can go about it for broadband imaging; capturing RGB under poorer seeing and then capturing the Lum under the best conditions. This doesn't work so much when doing narrowband however. Ha, OIII, SII in that order is what we'd consider as being the importance of FWHM. Ha usually has the most detailed fine structures and SII has the worst so you can image under worse conditions without it having as much of a detrimental effect. One of the reasons we've been shooting with a OSC is to have all three RGB channels with the same FWHM and we've found this working well.Something we have found over the years is that imaging at a smaller image scale does better at resolving fine details. A few years ago we were using a very similar setup to what you are with your TOA-130 and ASI1600 and found that the very best seeing we ever got was 1.8", averaged at around 2.5" and only rarely got worse than that.

Since moving to a Mewlon 250CRS it's changed. We now find that our average is below 2", not uncommon to get below 1.5" but still have nights of 2.5" and beyond. We're getting better and picking the seeing conditions in our local environment so on nights we feel that it isn't worth shooting at 2500mm we go down to 600mm for wider field work.

https://www.astrobin.com/hvqcvk/?nc=user
10 minute stack with an average of 1.28"

https://www.astrobin.com/8dv9tk/
NGC 5128; approaching 11 hours with a stack average of 1.65".

For a long time we didn't bother with longer focal lengths as our 130mm refractor didn't consistently give really low FWHM but once we attempted it we came to realise that our local seeing is better than we'd ever thought.

As for a flat guide graph, that doesn't necessarily mean that you're guiding is good, it can at times mean that your guiding is poor but in both directions. Under poorer seeing your guide graph may be flat but the RMS should rise by a fair amount.

very interesting about FL. But the question remains. Should I use my red stack at 3 FWHM. Or reshoot and replace the subs that are above 3.   Time is the great equalizer. If I waited until all my subs were 2 I would finish the image in 6 months

Jeff:
Hey Rodd. Interesting question, but so hard to answer quantitatively. FWHM for me varies so much on the target (eg. fine Milky Way stars versus chunky open cluster), and also on exposure length and camera gain. But as a rule of thumb, I generally turf subs with FWHM about 30% worse than the best, for any given filter. I also look at the star count ....just in case FWHM figures dip (and look better) just because of thin cloud, and eccentricity.So typically I turf about 20% of my subs on a good night, and about 50% on a poor but not disastrous one. I’m a bit lenient on RGB, but stricter on Lum which I find matters most. Still cutting my teeth on NB.

Thanks Jeff.  I guess it is a can of worms.  The other question is what happens when one filter is much better than another.  My red stack has a FWHM of 3.1 arcsec and my green stack has a FWHM of 2.4 arcsec.    I wish I lived in an area with steady seeing!

Kevin Morefield:
Rodd,I don't pay nearly as much attention to FWHM on RGB.  Generally, sharpness is not important on RGB and I often blur the RGB anyway to limit color noise.  My blur is usually about 2.5 pixels so you can assume the resulting FWHM would be pretty high.  I don't lose a lot of information with that blur and since this is just pretty pictures I don't mind.

Where you can run into problems with high FWHM color data is color fringing on the stars.  If the FWHM of one color channel is significantly higher than the others you will get a color fringe.  To resolve that I sometimes deconvolve that color channel master to push down the star size to match the others.  CCDStack is great for this because you can watch the deconvolution iterations as they happen and stop the process when the stars look like you want.  Any noise that was enhanced by the deconvolution can be easily handled via noise reduction or blurring.   If all three color channels have stars that are significantly larger than the Luminance you can do the same deconvolution on the RGB combined image.  Note: I usually create an RGB and Luminance separately and take non-linear RGB and L masters into Photoshop to process them separately and combine them there.

I am very aggressive with culling high FWHM subs on Luminance.  We get lots of variation in seeing in Winter and it wouldn't be uncommon for me to toss out 75% of the Luminance I shoot.  My thinking is that one good picture is better than 4 bad ones.  I might also create a super luminance of the low FWHM RGB and L frames if needed.

Occasionally I will do two Luminance masters - one with only sharp subs and one with all subs.  The great thing about sharpness is that it is most apparent in the highlights.  SNR is determined by sample size of photons and not integration time so those highlights will have good SNR even with few subs.  So...using a luminance mask in Photoshop (or PI) I combine the highlights of the sharp L with the shadows of the everything L for my master.

Rodd Dryfoos:
Is it possible to end up with a great image if FWHM is 3.5 or so?

Yes, but the answer is to shoot a widefield shot with a large image scale.  The larger your object, the smaller your stars and details will seem.

Kevin

But that is not possible without changing somethng--like a camera, OTA or reducer.  If you check your subs and the subs coming off the camera are 4.3 arcsec--your stuck with that scale and framing.

Rodd Dryfoos:
Sounds pretty familiar.  Deconvolution in PI is a bit tricky.  I have it down pretty well for increasing deatil in high signal areas–but not for star reduction–it ends up making thing worse.

I agree about PI Decon.  Like I said, CCDStack is really easy.   But you could do the same sort of star reduction with a minimum filter in PS.  You would do a star selection (many ways to do that) and then apply the minimum filter to the single color channel that is the problem.

This is the most difficult choice, what to keep and what to throw away. Ultimately, this is a choice you will have to make based on your own experience and on your target, no ones situation is the same and each target has its own requirements . You have spent a lot of long nights to collect the data, so how do you decide what to keep?

As a scientist, I am reluctant to throw out data. Any data should be rejected because it is an outlier. So for me, on any one night, I will delete any images that deviate significantly from the mean (might be 1% to 50%, there is no rule, every night is unique. Let the data inform you, there is no way there is a general rule.)  If it is obvious in the raw image, I do this as the images are coming in, otherwise I wait to when I fully process the image. Generally on any one night there is a narrow range, get rid of the clear outliers, first in background (nothing worse than clouds in an image, if not odd shaped stars), shape (almost nothing worse than odd shaped stars, except for clouds), FWHM (my experience this is the least objectionable defect, and many image processing programs can fix halos, deconvolute to improve your FWHM), and of course just avoid stretching your stars too much. But process the images from that night first, look at your result, and if you see issues, try removing what you feel are the outliers that lead to the problem you see. I personally focus on one target for multiple nights. Then I can compare each night and see which are outliers (background, shapes, FWHM). If one night is worse, I then go and try to fix the problem in those poor nights, whatever it might be. If I see the target needs a more transparent sky, or better seeing, I look for those better nights to continue the image.   If I cannot fix these nights, I have, without any feeling of loss, deleted those nights. Don't delete images arbitrarily based on any metric, if removing images doesn't improve the image, keep them. More exposure can more than compensate for a few bad subs.

In the end, it will be up to you. Do you want to be spending all this time to image, and deleting 90% of your images, or accept what you can do. Or perhaps find that you needed to focus your image better or more often. Use a Bahtinov Mask to focus, make sure your telescope is equilibrated to the night time temperature (at least an hour unless you have a small scope), and focus when your images start to deviate from the mean for the night so far. Decide what your tolerance is. There is no right answer, but if you are deleting 50% of your images please figure out what is wrong, your setup or your expectations.

Personally, and I think you will find many in the same situation, my location has too much wind, too much light (clouds must be deleted they are way too bright), and the jet stream is usually parked right over me with bad seeing. My equipment vibrates in the slightest wind, fogs up on humid nights–so work to get the best out of what you have. Once you see where your problem in the final image really lies, over many nights, you know what to do for your next equipment, or what mountain to you need to move too.

Hope this helps,
CS
Rick

FWHM of subs is important, but not only. Eccentricity, SNR and other values are important too. So a mix of criterias while selecting subframes is always good.

There are a lot of good comments in this thread.   I'd like to add my 2 cents.   Among the other factors listed above, there's a definite elbow for FWHM that occurs as a function of elevation.   I have imaged about 12K subexposures this year, and have thrown out just less than half.   I'm in the southwest US, and we have decent dark skies, some wind, and my mount has certain favorite (and less favorite) quadrants of the sky to image in.   So, I definitely look at FWHM, eccentricity, and background level.   I concur with the post that said you need to be careful though.  A high eccentricity on a low FHWM sub may actually be much better than a lower eccentricity on a high FWHM sub.   So, you need to get a feel that balance.  Everyone's site will be slightly different.   I try my best to get the lowest FWHM possible, but I typically get only 2.25 best measure.   I typically try to keep eccentricity below 25 and FWHM less than 3.5 for targets above 50 degrees elevation.   Below that elevation, expect to need to allow for more FWHM.   Here's a chart from my archive's statistics for 6K+ subs.   Not perfect, but I have some reasonable images (look at my home gallery to see if you like or not).   FWIW, the chart shows data I've saved higher, but the worst data is older than newer (I'm learning).   I hope this helps a bit if nothing else, to see what someone else is keeping vs pitching.   These are what I've kept, and I'm trying to move left always to pitch what's on the right.  Cheers,  Doug