NGC891

[alan4908]

10 minutes ago, Tom OD said:

that star looks good on my work screen, but a bit rectangular on my phone. What adds to the effect, is that the halo on the star is rectangular, and has a smoother transition to the core of the star. So maybe thats what my eye is picking up, making it look overall that shape.

Again though definite differenecs between the phone and monitor.

Thanks Tom - I'd be interested to know how this star looks on your screen without any or minimal processing (see above post for the info).

Alan

[Davey-T]

They still look elongated to me Alan but once again the crop enlarged looks OK, should've gone to Spec Savers 

Dave

 

[Filroden]

This is really odd. I also saw the elongation from 11 o'clock to 5 o'clock on both the iPad and my 4k monitor. I can see it still when zoomed in.

In part, I think it's not just the core that's contributing to the effect but also the star halo. When you zoom in, your eye can see less of the halo effect and stars start to appear rounder again. I suspect it's a combination of resolution, visual perception/illusion, and age/wine

And just to prove myself wrong, here's PixInsight's analysis of eccentricity across the field:

Pretty round to me

[ollypenrice]

Thanks for the TIFF. It's very instructive. When I open it in Ps exactly as you sent it and zoom it up to 3200%* I can open Curves mouse over it to identify exactly which pixels in the core are saturated. When I do so I find that the saturated core measures 7 pixels top to bottom and only 5 pixels side to side. To my mind this gives objective proof of what we are seeing at normal scale.

What complicates the issue is the difficulty of measuring objectivesly the way this elongation carries on as we move out from the core. However, if I measure pixels lying between 50K and 60K ADU I make it 13 pixels high and 10 pixels wide, the axis between the top and bottom pixels being offset to the upper left by 1 or 2 pixels. Again this confirms exactly the impression reported by others. The elongation is in the data, simple as that - if my test is valid and accurately conducted. (This is a wine-free hour of the day in our house...)

Is this a guiding effect? Could be. The way to eliminate either guiding or optics is to take a short sub on a bright star. If it's there in a 1 second sub it won't be the guiding.

Olly

* Can I have Pixel Peeper of the Year for this, please!!! (I never normally pixel peep since I think a good imager works to produce the optimal result at the chosen scale of presentation. What is optimal at one scale might not be optimal at another - star reduction being a classic case.)

[alan4908]

6 hours ago, Davey-T said:

They still look elongated to me Alan but once again the crop enlarged looks OK, should've gone to Spec Savers 

Dave

Thanks Dave - the unprocessed star definitely looks rounder to me than the processed version !

6 hours ago, Filroden said:

This is really odd. I also saw the elongation from 11 o'clock to 5 o'clock on both the iPad and my 4k monitor. I can see it still when zoomed in.

Yes - I can see this too. 

6 hours ago, Filroden said:

In part, I think it's not just the core that's contributing to the effect but also the star halo. When you zoom in, your eye can see less of the halo effect and stars start to appear rounder again.

Yes, I think the core and halo have different aspect ratios, with the saturated core being worse.

 

6 hours ago, Filroden said:

Pretty round to me

Yes, it is interesting that all the software programs MaximDL, CCDInspector and Pixinsight believe that the stars are quite round. I presume this is because they as looking at both the core and the halo and then they have a weighting about how much each contributes to the roundness.

5 hours ago, ollypenrice said:

Thanks for the TIFF. It's very instructive. When I open it in Ps exactly as you sent it and zoom it up to 3200%* I can open Curves mouse over it to identify exactly which pixels in the core are saturated. When I do so I find that the saturated core measures 7 pixels top to bottom and only 5 pixels side to side. To my mind this gives objective proof of what we are seeing at normal scale.

What complicates the issue is the difficulty of measuring objectivesly the way this elongation carries on as we move out from the core. However, if I measure pixels lying between 50K and 60K ADU I make it 13 pixels high and 10 pixels wide, the axis between the top and bottom pixels being offset to the upper left by 1 or 2 pixels. Again this confirms exactly the impression reported by others. The elongation is in the data, simple as that - if my test is valid and accurately conducted. (This is a wine-free hour of the day in our house...)

Is this a guiding effect? Could be. The way to eliminate either guiding or optics is to take a short sub on a bright star. If it's there in a 1 second sub it won't be the guiding.

Olly

* Can I have Pixel Peeper of the Year for this, please!!! (I never normally pixel peep since I think a good imager works to produce the optimal result at the chosen scale of presentation. What is optimal at one scale might not be optimal at another - star reduction being a classic case.)

Olly

Thanks for performing this analysis, I do like this methodology.

Just to let you know that I repeated this myself with PS and CCDStack and obtained slightly different results but perhaps this is because I measured something slightly different !

In PS, I set the eyedropper to point sampler mode and looked at the information window where it gave a saturated (255/255/255) reading. Using this method, I measured the saturated core at 7 pixels top to bottom and 6 side to side.

On the halo, I programmed CCDStack to reject all pixels between 50k and 60k ADUs, which gave an annulus of rejected pixels within the star, since CCDstack marks rejected pixels in red, I then simply measured the height (12 pixels) and width (11 pixels) of the outer annuls. 

Although the numbers do slightly differ from yours, they do lead to a similar conclusion (at least for this star), that within the processed data, the saturated core roundness is slightly elongated in the generation direction seen by some.  This elongation appears to be lower if you take both the core and part of the halo into account.

You may be correct about the cause being a some form of tracking or optics error, however, I think it is more likely to be due to the way that I've processed the data.  I shall investigate.

Many thanks for your help and assistance. 

Alan

 

 

 

[ollypenrice]

9 hours ago, alan4908 said:

Thanks Dave - the unprocessed star definitely looks rounder to me than the processed version !

Yes - I can see this too. 

Yes, I think the core and halo have different aspect ratios, with the saturated core being worse.

 

Yes, it is interesting that all the software programs MaximDL, CCDInspector and Pixinsight believe that the stars are quite round. I presume this is because they as looking at both the core and the halo and then they have a weighting about how much each contributes to the roundness.

Olly

Thanks for performing this analysis, I do like this methodology.

Just to let you know that I repeated this myself with PS and CCDStack and obtained slightly different results but perhaps this is because I measured something slightly different !

In PS, I set the eyedropper to point sampler mode and looked at the information window where it gave a saturated (255/255/255) reading. Using this method, I measured the saturated core at 7 pixels top to bottom and 6 side to side.

On the halo, I programmed CCDStack to reject all pixels between 50k and 60k ADUs, which gave an annulus of rejected pixels within the star, since CCDstack marks rejected pixels in red, I then simply measured the height (12 pixels) and width (11 pixels) of the outer annuls. 

Although the numbers do slightly differ from yours, they do lead to a similar conclusion (at least for this star), that within the processed data, the saturated core roundness is slightly elongated in the generation direction seen by some.  This elongation appears to be lower if you take both the core and part of the halo into account.

You may be correct about the cause being a some form of tracking or optics error, however, I think it is more likely to be due to the way that I've processed the data.  I shall investigate.

Many thanks for your help and assistance. 

Alan

 

 

 

Best of luck anyway. Do you know the 'Star Rounder' trick as a Ps Action? It only works on one star at once but can be quite quick if you use the magic wand to pick the star and save the action to a function key. You'll need to experiment with the expand and feather values but try 5 and 2 for starters. You can even record two versions of the action for smaller and larger stars, the larger ones having larger values for expand and feather. It goes like this:

- select star using magic wand.

- start recording action saved to a function key.

- Filter, blur, radial blur set to spin, 100% and best quality.

- Repeat the action above for good measure.

- Deselect. (Don't forget this one!)

By running this action on your star with expand value 5 and feather value 2 it came out like this, upsized to be easier to see.

Olly

PS The fact that all these bells and whistles bits of software can't see that the stars are out doesn't surprise me in the least. As I just may have mentioned before  I place far more faith in the human eye and brain. I wonder what the software says about the 'rounded' star?

[Petergoodhew]

A fascinating, intriguing, and bewildering thread. Olly may not be a pixel peeper, but he's damned good at astroforensics.

Sadly none of this will explain away all of my processing failures!

[alan4908]

6 minutes ago, Petergoodhew said:

A fascinating, intriguing, and bewildering thread. Olly may not be a pixel peeper, but he's damned good at astroforensics.

Sadly none of this will explain away all of my processing failures!

In order to track down where the vertical star elongation in the star cores came from, I decided that I needed a software tool, so I decided to program CCDstack to highlight the areas in the image where the pixels where either at or near saturation. Using this technique, I went back to my subs and then tracked these through the various processing stages.

Although the stacked FITS subs do not give perfectly round stars, they are not the source of this vertical elongation effect, which largely eliminates tracking and optics.

Loading the various stages of my processing into CCDstack, I discovered that my experiment with a global star shrink operation introduced the effect. I think the effect is hidden from some people (including myself) and the various software programs (Pixinsight, CCDInspector and MaximDL) because the star halos are rounder than the core - I tend to use separate processing techniques on the halos.  

So, it would appear that this mystery is solved. 

Alan  

 

[Tom OD]

1 hour ago, alan4908 said:

In order to track down where the vertical star elongation in the star cores came from, I decided that I needed a software tool, so I decided to program CCDstack to highlight the areas in the image where the pixels where either at or near saturation. Using this technique, I went back to my subs and then tracked these through the various processing stages.

Although the stacked FITS subs do not give perfectly round stars, they are not the source of this vertical elongation effect, which largely eliminates tracking and optics.

Loading the various stages of my processing into CCDstack, I discovered that my experiment with a global star shrink operation introduced the effect. I think the effect is hidden from some people (including myself) and the various software programs (Pixinsight, CCDInspector and MaximDL) because the star halos are rounder than the core - I tend to use separate processing techniques on the halos.  

So, it would appear that this mystery is solved. 

Alan  

 

Only getting back to this thread now. That's good news Alan. Were you using PI to shrink the stars? I only ask as the selections of the star shapes selection in the Morphological tool  can be an odd way to go about it. I sometimes use PI and later PS, with its Minimum filter. For ease of use though I generally just go with PS.

Tom 

[ollypenrice]

Very interesting and good news!

Olly
 

[sloz1664]

Great news Alan. I'll ad that info to my many lists of "do's & don'ts " in astrophotography.

Steve

[alan4908]

1 hour ago, Tom OD said:

Only getting back to this thread now. That's good news Alan. Were you using PI to shrink the stars? I only ask as the selections of the star shapes selection in the Morphological tool  can be an odd way to go about it. I sometimes use PI and later PS, with its Minimum filter. For ease of use though I generally just go with PS.

Tom 

Hi Tom

No I wasn't using PI but a PS tool that I believe uses the liquify and pucker tools within PS. If used correctly, then it does not cause any problems, if you set it at too high a strength, as I did, then you can get issues.

1 hour ago, ollypenrice said:

Very interesting and good news!

Olly

Thanks Olly !

4 minutes ago, sloz1664 said:

Great news Alan. I'll ad that info to my many lists of "do's & don'ts " in astrophotography.

Steve

Thanks Steve - however, please don't think that you should never shrink stars - just when you do, don't over do-it and double check that everything is OK !

Alan