FWHM doubles

[MikeB838]

Hi Robin,

with the last version (V4.1.13.229) I see FWHM values during live stacking (with sharp stars high in the sky) which are about twice as big, as I used to see before.

Do I miss something or is this a bug?

Thanks a lot!

cheers,
Michael

[admin]

Hi Michael,

I did tweak the code somewhat in the measurement of FWHM at that point - the old code was ignoring too much of the outer, fainter, area of the star, which meant that the FWHM became sensitive to brightness changes in the image. The new approach deals with brightness changes better (ie the FWHM of the same star at different brightnesses varies less with the new code), but it will tend to get bigger measurements. Saying that, I didn't notice anything like a factor of 2 jump in my testing.

In one sense, it doesn't matter greatly - FWHM measurements are only really useful in comparison to each other, so a bigger number across the board doesn't hurt anything as long as the relative values of in/out of focus etc still work right. In another sense, maybe it indicates that something is not quite right with the new code in some cases, so I would like to investigate - if you have a frame or two that I could test on (unprocessed - straight from the camera if possible), please share them with me.

thanks,

Robin

[MikeB838]

Hi Robin,

my answer did not get through, so here is my next try:

thanks for your reply! My FWHM normally is between 5 and 6 and now it was about 11. No problem but the values are over the top of the display and
not easy to see.
I zipped for you 2 raw files, the setting file and a screenshot of the FWHM display (here from the folder monitor, it looks the same when connected to
the camera).

Please see: www.twus.de/rawframes.zip (8MB)

Thanks and cheers,

Michael

[admin]

Hi Michael,

thanks for sharing those images - what's going on is basically the effect that I was trying to solve with my update... Particularly with faint stars, estimating the background brightness around the star is important as it has a big impact on the calculation of FWHM. Older versions of SharpCap were estimating this brightness at too high a level (typically), leading to a reduced FWHM value. Current versions are probably underestimating a bit, leading to an increased FWHM value. PixInsight reports a value of about 8.5 for those files, which is halfway (roughly) between the old and new SharpCap values.

I'm going to try to refine the measurement again in a future update, but for now, you should be good to use the current versions as the FWHM should still respond properly to focus/seeing changes.

cheers,

Robin

[MikeB838]

Hi Robin,

no problem and thanks and good to know

cheers,
Michael

[admin]

Hi Michael,

on further digging, I'm still not happy with how the FWHM is being measured - mainly due to problems estimating the background brightness around the stars - and will be looking into this in some detail over the next week or two.

cheers,

Robin

[Gfischer]

Robin,
I noticed the change in Fhwm as well. What’s interesting is I get a lower number with short exposure 2 sec fwhm is 3 apx. When going up to 30s fwhm goes up to 5 or 6. Probably because of brighter image.

[admin]

Hi,

yes, I discovered a whole can of worms when I dug deeper into this - a lot of it comes down to is the background estimated nicely. If the background is estimated too low (current version of SharpCap is doing this) then the FWHM reads high because some of the background is seen as star brightness a long way from the center. If the background is estimated too low then the opposite happens (SharpCap used to be like that until a few updates ago).

I *think* I am closing in on an approach that delivers better, more consistent results, but still I would say that really you should only compare FWHM figures between frames when they are taken at the same settings, etc - ie the FWHM comparison is good for focus, seeing monitoring, etc, but not always valid in some other circumstances.

cheers,

Robin

[Gfischer]

Robin,
I agree this is not an issue I just put limit higher when live stacking. Just out of curiosity how do you determine the edge of stars? Do you subtract adjacent pixels until a change is indicated?

[admin]

Hi George,

finding the background is a big challenge to do it accuractely - there are lots of ways which trade off differences in speed, accuracy, etc. My current take for star detection is to have a rectangle in the image containing the star that is big enough that the star is well inside it - then you look at the pixel values around the edge of the box, but even then you have to take care in case another star is sitting on that line of pixels.

I now have some updates to be released today that seem to cope much better with finding the background fairly accurately and therefore giving values for FWHM that are in decent agreement with other applications (and are also less sensitive to changes in the star detection settings in SharpCap, although there still is *some* sensitivity).

cheers,

Robin