Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Pre- and post-calibration Fwhm, eccentricity, etc, values

  • Please log in to reply
16 replies to this topic

#1 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 06 December 2018 - 01:07 PM

     Has anyone else out there noticed the fwhm, eccentricity, snr, etc of their subs going from particular values to significantly worse values after calibration? For example, a sub's fwhm going from 4.140 to 4.725 (as measured by the newest SFS process). I've checked using multiple data sets of various DSOs from different times of the year (with the correct cal frames for the particular object and date), and have found that the values of fwhm and SNRweight almost always go up and down, respectively, (get worse) after calibration, eccentricity might get better or worse, and other statistics can also get better or worse. I would assume that most statistics would improve with calibration.

     I'm thinking that this might indicate that my cal frames could be MUCH better. Any thoughts?

 

(I am using PI for my pre- and post-processing, and I did a futile search here on CN for pertinent posts)


Edited by twidget, 06 December 2018 - 08:51 PM.


#2 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 06 December 2018 - 08:55 PM

I checked more data sets, and found that what I commented on happens less often than I thought, but still over 50% of the time. I used good and poor data, long and short sets, NB and LRGB, and could determine no pattern.

 

Is this an expected result of calibration? Only occurs with good (poor) data? I fear I am showing real ignorance here, but this has me baffled, and concerned that I am doing something incorrectly, or missing something, or.....


Edited by twidget, 06 December 2018 - 08:56 PM.


#3 pfile

pfile

    Aurora

  • -----
  • Posts: 4650
  • Joined: 14 Jun 2009

Posted 07 December 2018 - 03:08 AM

calibrated only or calibrated and registered?

 

rob



#4 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 07 December 2018 - 09:06 AM

Calibrated. I haven't compared against any calibrated and registered subs. yet. I will check that today, if possible. Thanks for the suggestion!



#5 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 07 December 2018 - 09:10 AM

When I use Nebulosity 4 to check individual stars' fwhm, I get the same result - single star fwhms tend to get worse post-calibration compared to pre-calibration.



#6 pfile

pfile

    Aurora

  • -----
  • Posts: 4650
  • Joined: 14 Jun 2009

Posted 07 December 2018 - 12:00 PM

well i was asking because thing can change due to the interpolation happening to a registered frame. i guess i wouldn't expect the FHWMs or other properties to differ wildly after only calibration. but i have never run the experiment you have done.

 

rob



#7 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 21596
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 07 December 2018 - 03:24 PM

This seems pretty odd. Calibration with darks and flats should not actually be changing the small scale nature of the signal. A change in eccentricity would be pretty odd after simply subtracting out a dark. I wouldn't expect FWHM to change much either, however it may depend on how the signal floor is identified, and calibration may have an effect on that.

 

A change in SNR weight may occur if you are clipping data. I would make sure you do not have very many pixels clipped to black after calibration, and if you do, add an output pedestal to prevent that.



#8 AtmosFearIC

AtmosFearIC

    Viking 1

  • -----
  • Posts: 852
  • Joined: 10 Dec 2015
  • Loc: Melbourne

Posted 07 December 2018 - 05:38 PM

I wonder if it has something to do with the flat frames correcting for the vignetting. It is possible that once the vignetting has been corrected for it is picking up more stars in the outer parts of the image.

 

Your best way of doing a comparison would be to compare the same image before and after calibration. If they stars aren't changing then it is how Nebulosity is checking FWHM rather than the images becoming worse.



#9 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 7821
  • Joined: 12 Oct 2007

Posted 07 December 2018 - 05:40 PM

Two things to consider are the interpolation method - as mentioned above - and how well the images are registered.

 

I don't use PI - but if you can create calibrated and shifted versions of each frame prior to stacking - and view them as an animation - you should get a sense of how well the stars are stacking on top of each other in the stack.  If you see them jittering around - that suggests an imperfect alignment is the cause of any bloating in the stack.

 

Frank

 

addendum:  Another thing is to make sure the same stars are being measured in all cases.  If you are using a tool that measures many stars and gives a mean result - then if more stars are being used in the stacked version it may pull in stars that are slightly bloated.  So you might want to measure individual stars before and after stacking to make sure it isn't an artifact of how the measured stars are selected.


Edited by freestar8n, 07 December 2018 - 05:42 PM.

  • rgsalinger, DavidHickey and 42itous1 like this

#10 Dimperev

Dimperev

    Lift Off

  • -----
  • Posts: 23
  • Joined: 25 Feb 2017
  • Loc: Saint-Petersburg

Posted 07 December 2018 - 06:44 PM

I think that the fwhm measurement process operates with certain contrast values to highlight the star above the background.

When calibrating, we reduce the overall background level and increase the contrast of the stars.

This should lead to an increase in the measured fwhm due to the capture of a weak halo of stars.

Try instead of calibration, simply increase the subframe pedestal. I expect a decrease in measured fwhm. Although we understand that the real characteristics of the image have not changed.


  • NaNuu likes this

#11 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 08 December 2018 - 10:08 AM

Lots of possibilities...I'll need a little bit to go through them all, though I do have exact replies to several already; I'll get back to you all...

lots to chew....

lots to do



#12 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 08 December 2018 - 10:08 AM

and,

lots of THANKS smile.gif


Edited by twidget, 08 December 2018 - 10:20 AM.


#13 spokeshave

spokeshave

    Surveyor 1

  • *****
  • Posts: 1614
  • Joined: 08 Apr 2015

Posted 08 December 2018 - 11:29 AM

This seems pretty odd. Calibration with darks and flats should not actually be changing the small scale nature of the signal. A change in eccentricity would be pretty odd after simply subtracting out a dark. I wouldn't expect FWHM to change much either, however it may depend on how the signal floor is identified, and calibration may have an effect on that.


I suspect you're right about the noise floor. Calibration will clean up the background quite a bit, allowing the star selection routine to pick up more of the tails of the star profile, increasing the reported FWHM. If there is eccentricity, it will also pick up more of that. It's probably important to recognize that the actual FWHMs and eccentricity are not increasing, it's just that calibration allows them to be more accurately reported.

Tim
  • twidget likes this

#14 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 09 December 2018 - 08:24 AM

A slight illness stopped me for a day or two. Onward...

 

well i was asking because thing can change due to the interpolation happening to a registered frame. i guess i wouldn't expect the FHWMs or other properties to differ wildly after only calibration. but i have never run the experiment you have done.

 

rob

Rob, I still haven't compared registered subs statistics against their non-registered stats, since the PI subframe selector script ( and now the new process) wants calibrated, unregistered subs to work on. I will get around to comparing some calibrated registered subs' statistics to their raw statistics in the next day or two, as I feel better. BTW, the initial observation was due strictly to serendipity, I accidentally used raw versions of the subs I was interested in when using SFS, and just happened to notice the stats were better than the cal'ed sub stats. Not really an "experiment." lol.gif

 

This seems pretty odd. Calibration with darks and flats should not actually be changing the small scale nature of the signal. A change in eccentricity would be pretty odd after simply subtracting out a dark. I wouldn't expect FWHM to change much either, however it may depend on how the signal floor is identified, and calibration may have an effect on that.

 

A change in SNR weight may occur if you are clipping data. I would make sure you do not have very many pixels clipped to black after calibration, and if you do, add an output pedestal to prevent that.

 

Jon, I also find it odd, and agree with everything you stated (I'm NOWHERE near as sharp as you on this stuff, but I try to keep up). A while back, I thought I saw that I was clipping some pixels to black, so I studied up a bit, and started adding an output pedestal. In the Output Files section of the IC process, I have 1000 entered in the Output Ped (DN) box, and it appears that the ped is added to all my subs. 

 

I wonder if it has something to do with the flat frames correcting for the vignetting. It is possible that once the vignetting has been corrected for it is picking up more stars in the outer parts of the image.

 

Your best way of doing a comparison would be to compare the same image before and after calibration. If they stars aren't changing then it is how Nebulosity is checking FWHM rather than the images becoming worse.

 

Atmos, I would have to think that what you stated accounts for some of the (negative) change in stats I've observed between raw and cal'ed subs. However, I've used Neb4.1 to check the same individual stars (by placing the pointer on a single star, and reading its HFR in the Pixel Stats window) on raw and cal'ed copies of the same sub, and seen the reported HFR go up after calibration. The Subframe Selector gives an "average" of the image star fwhms, if I'm not mistaken. My short investigation has so far revolved around checking single subs, before and after calibration, with PI SFS (for image "averages"), and with Neb4.1 (for individual stars) . In any of the cases where the stats got worse after cal, the images always appeared the same or better post-cal than pre-cal; of course, my eyes aren't what they were even ten years ago.


Edited by twidget, 09 December 2018 - 08:27 AM.


#15 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 09 December 2018 - 08:37 AM

Two things to consider are the interpolation method - as mentioned above - and how well the images are registered.

 

I don't use PI - but if you can create calibrated and shifted versions of each frame prior to stacking - and view them as an animation - you should get a sense of how well the stars are stacking on top of each other in the stack.  If you see them jittering around - that suggests an imperfect alignment is the cause of any bloating in the stack.

 

Frank

 

addendum:  Another thing is to make sure the same stars are being measured in all cases.  If you are using a tool that measures many stars and gives a mean result - then if more stars are being used in the stacked version it may pull in stars that are slightly bloated.  So you might want to measure individual stars before and after stacking to make sure it isn't an artifact of how the measured stars are selected.

 

Frank,

 

The issue I'm looking into here occurs prior to registration. It is the difference in sub statistics between raw and calibrated subs, before stacking. What you say makes perfect sense, but I'm checking pre-stack stats. As far as measuring the same stars in the raw and cal'ed subs, we have no ability (as far as I know) to specify certain stars to measure when using SFS (either the script or the newer process); as I stated above, in Neb4.1 I do use the pointer with Pixel Stats to check the same stars in the raw and cal'ed versions of subs.



#16 twidget

twidget

    Vostok 1

  • -----
  • topic starter
  • Posts: 121
  • Joined: 12 Jul 2016
  • Loc: SW Missouri

Posted 09 December 2018 - 08:55 AM

I suspect you're right about the noise floor. Calibration will clean up the background quite a bit, allowing the star selection routine to pick up more of the tails of the star profile, increasing the reported FWHM. If there is eccentricity, it will also pick up more of that. It's probably important to recognize that the actual FWHMs and eccentricity are not increasing, it's just that calibration allows them to be more accurately reported.

Tim

 

Tim,

Your last sentence is exactly the kind of thing I was hoping to hear. waytogo.gif   However, I still feel I need to dig far enough into this to determine if this is a common issue, and illuminate and eliminate any problems I can, either with my eqpt, software, or processes/procedures, that might be causing or contributing to it. 



#17 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 7821
  • Joined: 12 Oct 2007

Posted 09 December 2018 - 04:34 PM

Frank,

 

The issue I'm looking into here occurs prior to registration. It is the difference in sub statistics between raw and calibrated subs, before stacking. What you say makes perfect sense, but I'm checking pre-stack stats. As far as measuring the same stars in the raw and cal'ed subs, we have no ability (as far as I know) to specify certain stars to measure when using SFS (either the script or the newer process); as I stated above, in Neb4.1 I do use the pointer with Pixel Stats to check the same stars in the raw and cal'ed versions of subs.

The fwhm values should be fairly insensitive to where the noise floor is placed.  This isn't the case for HFD - half flux diameter - though because changing the floor a bit could pull in a lot of flux in the outer part of the star spot.  If you look at the geometry of how fwhm is measured - a small shift in the baseline won't affect the "waist" measurement much.

 

Are the calibrated frames also aligned - or are the stars in the same places?  If the frames are aligned then there will likely be some bloating of the stars in the process due to interpolation.

 

If you're getting significant bloating on individual stars after calibration and the stars are not shifted in preparation for stacking - something doesn't sound right.

 

Frank




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics