Calc Star Colours: ...
 
Share:
Notifications
Clear all

2022-08-17: APP 2.0.0-beta3 has been released !

Release notes

Download links per platform:

windows 2.0.0-beta3

macOS x86_64 2.0.0-beta3

macOS arm64 M1 2.0.0-beta3

Linux DEB 2.0.0-beta3

Linux RPM 2.0.0-beta3

Calc Star Colours: Indistinct blob instead of clear linear form


(@mestutters)
Red Giant Customer
Joined: 5 years ago
Posts: 144
Topic starter  

Hi,

On the occasions I've used APP to process subs captured using itelescope systems I observe that at Calibrate Star Colours stage APP always produces colour - colour graphs that have a very distinct linear star pattern, for example:

Screenshot 2022 06 27 16.44.41

However, when processing subs captured with my own backyard system I never really see such a distinct pattern.   The below screenshot is typical of what I see prior to making any adjustments:

image

:  While the lower B-R vs G-R shows the expected alignment (though much less distinct than the previous image), in the upper B-G vs G-R chart the slope is much less obvious and in fact looks slightly negative slope.

Screenshot 2022 07 05 11.32.35

If I play with the sliders I can eventually get a result that I think looks reasonably correct for a RGB image but to do so I am:

a) almost out of range on the lower slope adjustment,

b) B-G vs G-R diagram shows almost spherical  star distribution

c) line on B-G vs G-R chart seems perpendicular to what looks like the most obvious slope.

 

I  am wondering if:

a) Any other APP users see this type of pattern.

b) Is my approach to using this tool incorrect

c)  Is it possible some property of my imaging train is giving this effect:  SX814M CCD with Astrodon filters

d)  Are my subs excessively noise or not registering sufficiently accurately thereby 'confusing' APP's analysis

d)  Is the effect solely an unavoidable result of local LP when imaging under Bortle 4-5 sky.

 

I did for some considerable time suspect my filterwheel was perhaps faulty thereby causing subs to get occasionally mislabelled/mixed up but after watching carefully for this I've found nothing in my imaging logs to support this theory.

Hoping someone can shed some light on these observed results and, even better, suggest a remedy. 

Regards

Mike


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4930
 

Are you using filters on your own setup? It seems likely that something is changing the nature of the broadband spectrum, changing it from a linear correlation to a less defined one. It does happen, in cases like that you may have to do more manual color corrections as the spectrum is not very well matching what the tool is expecting.


ReplyQuote
(@mestutters)
Red Giant Customer
Joined: 5 years ago
Posts: 144
Topic starter  

Hi Vincent,

In answer to your question, yes, I am using filters in my set-up.

I can appreciate that different optical trains can produce for the same object a somewhat different broadband spectum/image as a result of e.g. different filter bandpass charateristics, different sensor chip QEs and maybe lens coatings, thereby not producing quite the spectral characteristics as assumed by APPs algorithms.

However out of curiosity I would like if possible to establish exactly what APP is seeing in my calibrated RGB images that requires (to my mind at least) such extreme corrections.   If I understood this I could maybe slightly revise my approach during data capture e.g capture more data with say Red filter.

In the below two screen shots, one of image following an image following lpc-cbg correction and the second the same image loaded into CSC tool following only a single press of Re-Calibrate button.   I observe what seems to be significant Cyan/Green fringing around blue stars that I don't see on the input.

I am wondering if this is just a result of, for example, an excessively noisy image,  poor star registration or  excessively strong Red/Magenta balance in the source image?

RGB Combine image following LPC/CBG:

Screenshot 2022 07 06 13.52.15

Above image loaded into CSC tool following first press of the calibrate button.

Screenshot 2022 07 06 13.48.52

Apologies, at 400% zoom the images are not at all attractive.

 

Regards

Mike


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4930
 

To have a better answer for you it's best if we can analyze the data as well, maybe we can spot something more obvious going on. If you want you can upload your master calibration files and say 10 lights that. Would be nice indeed if we can help you improve on this;


ReplyQuote
(@mestutters)
Red Giant Customer
Joined: 5 years ago
Posts: 144
Topic starter  

Hi Vincent

Files uploaded.     Here's hoping I've not made some really basic processing error to get this problem!

Mike


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4930
 

Thanks Mike, will have a look later today.


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 5 years ago
Posts: 4930
 

So, what I'm seeing is that there's quite a bit of color gradients happening in the background. When you look at an overstretched image and with saturation turned up, you can see this. I do think this may cause the mismatch in being able to use the color correction tool properly as it may influence the star colors themselves. I'll ask Mabula as well to be sure.

 


ReplyQuote
(@mestutters)
Red Giant Customer
Joined: 5 years ago
Posts: 144
Topic starter  

Hi Vincent,

I buy your point about there being significant background colour noise in the images.  This was a somewhat opportunistic data acquisition as M92 is well-placed for me but just as UK nights achieving astronomical darkness become very short. 

On the other hand the SNR levels being reported in the integration files are significantly better than what I manage to achieve after many more hours of data acquisition of fainter objects during say galaxy season!

I don't regard myself as at all good and judging 'incorrect' or the appropriate corrections, but I percieve that there has been a significant magenta to green shift (plus maybe some blue to yellow) during CSC analysis but this is not untypical of what I experienced when processing other images with APP, so as previously, I am interested in what APP is 'seeing' that causes this to happen and if there is anything I can do to correct this trend in advance.

I would be interested in knowing whether Mabula thinks using the HSL tool to reduce the luminosity of the background colour noise prior to using CSC.   I recall I tried this one time but it did not seem to have an effect.  I reasoned to myself that this was probably because CSC was likely only measuring the colour values of the 'cores' of its selected stars for  comparison with its reference model and therefore background influence would be minimal. I fully prepared for this reasoning to be declared faulty.

I am having my CCD camera serviced at present as it is c 5years old.   I am not sure if the camera itself is especially noisy.   I have generally assumed  that the background colour noise was more related to sky noise than to electronic noise.

regards to you and Mabula

Mike

 

 


ReplyQuote
(@mabula-admin)
Quasar Admin
Joined: 5 years ago
Posts: 3187
 

Hi Mike @mestutters,

Thank you for sharing your issue and questions.

First of all, indeed, if you shoot with good sky conditions with broadband filters (like the iTelescope data) you will see very easy and good star color calibration which is indicated by the clear linear fits of the star population to the Black Body model.

Then, in your data, it is nowhere to be found it seems... Bortle 4-5 makes it a bit less easy to make calibration but it should still show a very clear linear pattern which it doesn't in your case. So the light pollution at your site is not the problem that prevents you to get good star color calibration. APP's star color calibration model accounts for light pollution and thus can and will still work very well even with very strong light pollution.

I have studied your dataset for quite some time today and I see a couple of possible problems:

VOyager is not reporting the sensor gain nor the sensor offset, there is no way to tell if we can properly calibrate the data this way... maybe the camera driver does not allow you to adjust gain and offset? In that case, all should be still okay I would assume.

For all filtergroups, R, G, B the first 3 images are very bad and are clear outliers which I would reject from using

Finally, i have been loading the individual R,G,B frames into the image viewer. If you load a R frame and then a B frame, you would expect to see the different brightness changes between Red and Blue stars. I am not seeing this at all... it almost seems that your filterwheel is not doing anything and thus all your frames were shot with the exact same filter.

Zoomed in on cluster :calibrated, registered, normalised Blue frame Blue_300s_BIN1_-20C_2022-05-08_A0003.FIT:

Blue zoomed in

Zoomed in on cluster: calibrated, registered, normalised Red frame Red_300s_BIN1_-20C_2022-05-08_A0005.FIT:

Red zoomed in

 

Red Left and Blue Right, I see no relative brightness differences between the stars indicating blue and red stars in these images... to me it seems to be shot with same filter:

RedLeft BlueRight

Perhaps you can check using the camera and filterwheel during daylight to check if the filterwheel in fact is working or not?

Mabula

This post was modified 1 month ago 2 times by Mabula-Admin

ReplyQuote
(@mestutters)
Red Giant Customer
Joined: 5 years ago
Posts: 144
Topic starter  

Hi @mabula

Thanks so much for taking time to take a look at my images.

Ever since I started to use APP, which I started using very soon after I started AP as a hobby, I have always felt that CSC tended to correct too much towards Green.  The effect on the above M92 image is by far the worst I think I've seen so I have determined to try to discover the cause.

Posted by: @mabula-admin

VOyager is not reporting the sensor gain nor the sensor offset,

My camera is cooled mono CCD camera.  It does not have a variable gain or offset.  I once computed the gain (using GAS method) and was able to include this as a value in SGP but I cannot find a way to do this in Voyager.  It's not a parameter in the ASCOM driver either. I guess I could insert a value into the FITS header post-capture if you think its absence is a problem?  Does APP use the gain value for anything more than simply matching lights to calibration frames?

Posted by: @mabula-admin

Perhaps you can check using the camera and filterwheel during daylight to check if the filterwheel in fact is working or not?

Having once suspected that the filterwheel (or its connection) was a possible problem I've spent considerable time checking it out. I can hear the FW when it moves and have checked that the wheel is actually turning and not just the motor.  All the feed-back I get from Voyager and SX's FW utility indicates it is performing correctly.

My camera and FW are currently with StarlighXpess for servicing but their feedback to date suggests nothing is wrong.

Posted by: @mabula-admin

Red Left and Blue Right, I see no relative brightness differences between the stars indicating blue and red stars in these images... to me it seems to be shot with same filter

I've not yet closely at these two subs but this observation however reopens the whole FW question!! 

 

I guess for now I must wait for camera to be returned and hopefully some clear skies.

Regards

Mike


ReplyQuote
(@mabula-admin)
Quasar Admin
Joined: 5 years ago
Posts: 3187
 

HI Mike @mestutters,

If you can't adjust gain and offset, then it is safe to conclude that the calibration frames are compatible with the lights in that regard 😉 So we can exclude that as possible cause, it is no problem then that gain and/or offset is not reported by Voyager.

On normal RGB data, the initial star color calibration result can have a green or magenta cast. That is normal and really depends on your optical train and the filter characteristics. Simply use the green/magenta slider to remove that cast. And you can see that cast as well in how the star population is plotted in the graphs.

Please let us know what Starlight Express concludes. To me, it really looks like all data was shot with the exact same filter...

Mabula


ReplyQuote
Share: