Square Stars

Juno16

Active member
Dec 29, 2020
25
2
Hi Pixinsight Forum,

I have been using PI for about 9 months now and still have quite a bit to learn! I am enjoying PI and learn new things especially when I run into trouble.

When I just started PI WBPP pre-pocessing, I noticed that some of the brighter stars were "square" or "pointed". I read online (I think SGL forum) to try using Bicubic Spline Pixel Interpolation and a Clamping threshold of 0.1. It seemed to work (somewhat) and I have processed all of my sets of data with those settings since then.

I recently set up a Sharpstar 61/ASI533 and my resolution with this combination is high at 2.89"/pixel, but I also have a Explore Scientific ED102 that I use with the ASI533 with a resolution of 1.36"/pixel and have the same "squarish star" issue.

I set the Clamp threshold back to 0.3 today (I believe the default) and left the Pixel Interpolation set to Bicubic Spline (I believe that the default is "Auto").

Most of the bright stars are pointed or "square" looking again (no surprise!).


PI Image Registration.JPG

The images below are a zoomed in portion of a M31 dist lane..On the left and center are masterlights with an STF stretch applied. The image on the right is a DSS stack also with an STF applied.



PI Image Registration1.JPG

The image on the left below is with Image Registration settings of:
Pixel Interpolation: Bicubic Spline
Clamp threshold 0.3

Bright stars are very "squarish".

The image in the the middle below is with Image Registration settings of:
Pixel Interpolation: Bicubic Spline
Clamp threshold 0.1

Same bright stars look more rounded.

The image on the right was stacked and registered in DeepSkyStacker

The stars look good.

By my eye, the leftmost image with the square stars has the best detail, followed by the center image, and then the DSS stack.

Is the Clamp threshold setting of 0.3 resulting in the better detail? Is there something that I can do with my data to alleviate the square star issue and use the 0.3 setting?

If there is a previous thread concerning this issue, I could not find it. If so, please point me to it.

Thanks a bunch for any assistance that you can offer!

Jim
 

Attachments

Juan Conejero

PixInsight Staff
Sep 2, 2004
8,156
344
57
Valencia, Spain
pixinsight.com
Hi Jim,

I cannot see any problems. Your "PI square stars" are perfectly fine and demonstrate perfect guiding and focus. The leftmost image of the three is also a fine result with very good resolution. I'd recommend drizzle integration to work with better sampled data.
 
  • Like
Reactions: Juno16

dld

Well-known member
Sep 24, 2017
209
35
I think one should say here that PI produces "diffraction-limited" stars and not "square stars". You should be proud for your optics and guiding, and happy that PI can reveal to you the effect of diffraction by a square aperture (your sensor's photosites a.k.a. pixels).

Edit: I am not sure if my statement regarding diffraction is correct. I will feel much better if someone else can confirm my statement :).
 
Last edited:
  • Like
Reactions: Juno16

Juno16

Active member
Dec 29, 2020
25
2
Hi Jim,

I cannot see any problems. Your "PI square stars" are perfectly fine and demonstrate perfect guiding and focus. The leftmost image of the three is also a fine result with very good resolution. I'd recommend drizzle integration to work with better sampled data.

I think one should say here that PI produces "diffraction-limited" stars and not "square stars". You should be proud for your optics and guiding, and happy that PI can reveal to you the effect of diffraction by a square aperture (your sensor's photosites a.k.a. pixels).

Edit: I am not sure if my statement regerding diffraction is correct. I will feel much better if someone else can confirm my statement :).
Wow!
I didn't expect that answer at all! Actually, I thought that I had problems with my optics!

When I first started using PI, I noticed the "square stars" and thought that I had a major issue. I started using DSS for stacking until I read the post on SGL concerning that issue and to try Bicubic Spline and a clamping threshold of 0.1. I have used those settings ever since.

I just modified the WBPP settings to return them to where they belong. Pixel Interpolation "Auto" and Clamping Threshold at 0.30.

Thank you Juan and dld for your happily received responses and for the compliments!

Jim
 
  • Like
Reactions: Juan Conejero

dld

Well-known member
Sep 24, 2017
209
35
You are welcome Jim,

If we only compare the resolution of the hydrogen region towards the top border of the image, I think we can all agree that the PI images are superior.

Now, why the brighter stars have a "plus sign" (+) shape? The shape reminds me the diffraction pattern by a square aperture, and a sensor is just an array of square apertures (the photosites).
 
Last edited:

Juno16

Active member
Dec 29, 2020
25
2
You are welcome Jim,

If we only compare the resolution of the hydrogen region towards the top border of the image, I think we can all agree that the PI images are superior.

Now, why the brighter stars have a "plus sign" (+) shape? The shape reminds me the diffraction pattern by a square aperture, and a sensor is just an array of square apertures (the photosites).
Thanks dld. The explanation is very much appreciated!
Yes, most definitely, the PI pre-processed image on the left is superior in resolution.
I have been using the WBPP settings that produced the center image until now.

Very glad that I posted to the forum and I appreciate your and Juan’s help very much!

Jim