Flats Fail to Correct Luminance

[niteman1946]

Hi Gents,
My optical path is:
1.  12"LX200 OTA (Meade Classic 1997)
2.  2" Feathertouch Focuser (no glass)
3.  Starizona SCT Corrector F7.5  (effective scope/corrector FL is 2182mm)
4.  Hutech OAG5 Off Axis guider (no glass)
5.  Atik EFW2 Filter Wheel (1 1/4" threaded Astronomik LRGB, Ha 12nm, OIII CCD, SII CCD filters)
6.  Atik 383L+mono CCD Camera

My location is in central Texas and the skies are somewhat polluted.  Views to the north overlook the lights of the town of Burleson.

There is something (or a combination of somethings) in the optical path that produces a circular ring in the calibrated/aligned/integrated image.  This ring feature sits just about center of the FOV and has an outer diameter of about 50% of the image height.  The feature is not noticeable in individual subs, but shows up once they are integrated.  It is most prominent in Luminance, while less in RGB and least in NB.  The Luminance feature is so dominant that I'm not able to remove it with processing.

For processing, I use Flats, Darks and Bias with Pixinsight.
FLATS are produced with the Tee-Shirt technique in daylight.  These are 30 subs at 10s each, and usually come in  around 30,000ADU
DARKS are 30 subs at the same duration as the Lights (i.e. 600s)
BIAS are 100 subs at 0.001s.
All Flats, Darks and Bias are shot at the same temperature (-10C) as the Lights.

I've included a DropBox folder showing the following:
1.  Screen Print NGC2146 & M63
2.  Screen Print Combo Bias, Dark, Flat Integration & Luminance (600s) Calibration (PI)
3.  Screen Print Luminance Integration (PI)
4.  Luminance Master Flat (XISF) (PI)

Two separate vendors with an interest in this have suggested appropriate Flats should remove this ring feature.  My settings as shown have been in place since around 2012, so I accept that they may be in need of change.  Also, they could have been incorrect from the start.

Any help on this would be greatly appreciated.

Mark

https://www.dropbox.com/sh/tq0bv39b2erjduc/AACTC24-m5xI_Y89nQLau-pXa?dl=0

[pfile]

i don't have a solution necessarily, but every rig i've built which uses a correcting element, both refractor and reflector, has suffered from this problem.

is it easy to try removing the corrector and see what you get?

i suppose you might also try twilight sky flats with no t-shirt to see if this helps any.

beyond that i guess you need to look for anything in the optical path that might be reflective in even the slightest way.

rob

[ngc1535]

This kind of result can happen when the sky has additive light (a bright sky) that is a different color than the flats that are used to calibrate the images.
If you are interested in ruling out my idea- take many many images of the night sky in the direction you were at over the city. Have the tracking off (or dither the telescope) for each exposure. Then combine these (perhaps median combine) to create a night sky flat. Apply this flat to your raw data and see if the issue is greatly reduced (or is corrected). If it happens, it confirms this particular possibility... otherwise it rules it out.

Concerning the night sky flat... don't be concerned the counts will be low. A few thousands of counts will suffice. You will never easily reach a normal flat level (in a reasonable period of time).

-adam

[1DegreeN]

I've had a similar issue and found that using scaled flats has helped a lot. Another option to explore: if you dithered your lights for NGC2146, you could try creating a synthetic flat from those subs.

[GJL]

Perhaps, the pictures are undercorrected by the Flat. Try to change the flat with PixelMath, e.g. "Flat + k". Factor k must be tried, in this case it will be positive, e.g. 0.1 or 0.2 or more.
Recalibrate with the changed flat.

Gerhard

[niteman1946]

Rob, Adam, 1Degree, Gerhard,
thanks a lot for your replies.  I really appreciate it.
FWIW, I shot and processed a few images of M63 thru an open port in the filter wheel.  The feature persisted, although I did use the existing lum flat to process.  So I may have proved nothing.

Rob,
 I did take a stab at twilights last night. The resulting background of M63 was coarse or grainy.  However the circular feature was greatly diminished.  It was the first time I tried twilight flats so it can’t get any worse.  Any suggestions on twilight tutorials?

Adam,
 as I mentioned to Rob, I think I saw improvement even with my clumsy attempt. I do like your suggestion of night flats and will give them a try.

1Degree and Gerhard,
 I need to do some reading on your suggestions.  “Flat + k” sounds easy enough if I understand it correctly.
Synthetic is going to take some research for me.

I’ll followup with results.

Mark

[pfile]

maybe your twilight flats were just too dim and have lessened the SNR of the calibrated result too much?

twilight flats are kind of a pain to do right; there is a spot in the evening/morning sky that is as gradient-free as possible. it does move as the sun goes down. software like ACP will compute where this is and automatically point/repoint the telescope (and adjust the exposure as the sky brightness changes.) if you have to do these flats without automation, you need to babysit them and try to point to the right place in the sky as well as steadily increase the exposure (for twilight flats of course.)

anyway to adam's point, the idea is to illuminate the sensor as closely as possible during flat acquisition as during light acquisition. to that end the night-sky flats might be more accurate (and easier to do.)

rob

[niteman1946]

Gents,

Based on my limited skill set, the best result I got was with the twilight flats.
I've attached a screen print showing the three different method attempts.
1.  Upper Left:  Existing Old Flat
2.  Upper Right:  Existing Old Flat + 0.2 in PixelMath
3.  Lower Center:  New Twilight Flat

Not a large improvement with the Twilights, but some.  Note the "bullet holes" smoothed out when I added more flats.

I also tried Night Flats, but could not get them to Integrate into something usable.  I'll accept the blame for that, although I've caught PixInsight going senile at times when it's heavily flogged.

Lastly, my Twilight technique was not the best.  I resorted to setting the exposure at 10s and just started capturing around 40,000 ADU and stopped around 6000 ADU.  My usual approach is to shoot for around 30,000 ADU +/- 2,000 so....

Thanks Again,
Mark

[GJL]

I have to mention one more point when correcting the flats with PixelMath. Transfer the FitsHeader from the old to the new flat, otherwise it will not be recognized by the BPP script. So open File/FitsHeader, drag triangle onto new flat.

[niteman1946]

Hi Gerhard,
thanks for the follow-up.

FWIW, I do all my processing manually, i.e. no BPP.

Would your recommendation still be relevant for my technique?

Mark

[GJL]

I think it is. At least the filter reference in the Fitsheader is needed. Besides, it´s not a big deal to transfer the header and it is done fast.