Back to forums

What reasons could affect Flat Frame performance?

Luc Coiffier DeepSkyStacker (DSS) Pleiades Astrophoto PixInsight Acquisition Calibration Deep Sky
8 replies•376 views
Menelaos avatar
Menelaos avatar
Dear Friends,

I require assistance with flat frame acquisition. It looks like my flat frames overcorrect the images and specifically in the green channel. I dont understand what the problem is! I had this issue before but usually the problem was light leaks. This time i've sealed every possible little hole there is on the telescope and took 10 minute dark frames in broad day light and i am confident there are no light leaks now! Flat frames still overcorrect! i tried exposing all the way from the left of the histogram to all the way to the right! I've tried calibrating with bias, dark flats, different exposure lenghts it just doesnt work! I used DSS and PixInsight for stacking and both give me the same result. It gets me scratching my head. I dont understand what is going on! 

When i looks at the Histogram of the flat frames the channels are not aligned. does this matter? it shouldn't because in the past my flats used to work perfectly. Nothing changed in my setup or the way i capture flats and they used to work ok! Any ideas what is happening? 

What other reasons could there be for flat frames not to work?

Very intriguing is that the flats work ok when i use dualband filters. They are only problematic unfiltered. Strange i guess
andrea tasselli avatar
andrea tasselli avatar
When we mention "light leak" it doesn't necessarily follow the hydraulic analogy. A light "leak" could be as well a reflection, even a glancing one, from any part of the imaging train (e.g., filter edges, filter holder edges, focuser end sections, you name it). In short any additive (additional) signal that gets on the camera's sensor. And it does NOT correct with flats since it is additive (and highly variable on top of that). Strip down your imaging train to the barest, and I mean really naked bare barest, set (ideally just the camera) and see how it goes.
Well Written Helpful Insightful Concise
andrea tasselli avatar
andrea tasselli avatar
And NBs don't show anything because there is very little to none of background illumination that comes with them.
Tony Gondola avatar
Tony Gondola avatar
Flat frames will have issues if the interaction with the optical system changes for whatever reason between shooting the object and shooting the flats. Can you lay out exactly what your procedures are? Also, are the defects that you see in your calibrated stacked frames the same or does it vary from object to object or filter to filter?
Well Written Insightful Respectful Concise Engaging
Robert Žibreg avatar
Robert Žibreg avatar
The only interference I've encountered is when flat exposure time was too short and flat panel banding was visible. Can you explain how you take your flats, what software and flat panel?
Well Written Engaging
John Hayes avatar
John Hayes avatar
Difficulty with flat correction falls into 4 general categories:

1) Stray light.  Stray light is additive and flat correction only corrects for multiplicative signals.  Stray light can come from a light leak or a stray reflection inside of the optical train.  It does not take very much stray light to screw up flat data.  Taking darks or flats in the daytime is risky.

2) Correction limits.  Correction limits happen when there is too much vignetting within a system.  Flats will not correct all the way down to 100% vignetting.  I can’t say for sure what the limits might be but in general if you have more than about 30% vignetting you may start to run into trouble with the accuracy of the correction.

3) Nonlinear or unstable data.  Flat data should be exposed so that the peak of the histogram is roughly 0.5 - 0.75 of the maximum signal level so that the data is recorded in the linear response region of the sensor.  Some cameras are not repeatable at very short exposures so flats should be exposed using a minimum exposure time not shorter than 2-3 seconds.

4) Improper calculation.  Improper calculation happens when you don’t include bias offset, flat darks, or if the data is taken outside of the linear response region.

To insure no stray light, you should try taking your calibration data at night in the dark.  After that, you need to post some data so that we can see what you are talking about.  If you want a good answer, you need to provide more details about precisely what you are doing and what you are measuring.  The more information you provide, the better the answer you’ll get.

John
Well Written Helpful Insightful Engaging Supportive
Mark Petersen avatar
Mark Petersen avatar
John Hayes:
Difficulty with flat correction falls into 4 general categories:


4) Improper calculation.  Improper calculation happens when you don’t include bias offset, flat darks, or if the data is taken outside of the linear response region.


John

Hi John,  

Can you expand on the use of flat darks?  I know many CMOS users no longer use bias frames because their electronic pattern noise is minor, but in theory the only difference between a bias frame and a flat dark is the flat dark contains a small amount of dark current in addition to the bias data. 

I bring this up because I was having a slight under correction issue  in WBPP when I used bias, flats and darks.  This mostly was only a problem when I used a lot of stretch.  I tried everything including not using bias frames, using different histograms, using a flat panel vs sky flats, etc.  and the under correction remained.  The one thing that fixed the problem was adding dark flats.  I get well calibrated data now.  This is with a IMX455 sensor at -10C, the flats and dark flats were 25s exposures.  A 25s exposure has negligible dark current with an IMX455 at -10C.  I visually inspected both the master bias and master dark flat and both looked identical.  Both had a signal floor that matched the camera plateau/offset.  So all I could conclude is that WBPP's calibration math when using dark flats is a little more accurate in this situation.  

Just curious if you've come across something like this.  

Best Regards,
Mark
Well Written Helpful Insightful Respectful Engaging
John Hayes avatar
John Hayes avatar
Mark Petersen:
John Hayes:
Difficulty with flat correction falls into 4 general categories:


4) Improper calculation.  Improper calculation happens when you don’t include bias offset, flat darks, or if the data is taken outside of the linear response region.


John

Hi John,  

Can you expand on the use of flat darks?  I know many CMOS users no longer use bias frames because their electronic pattern noise is minor, but in theory the only difference between a bias frame and a flat dark is the flat dark contains a small amount of dark current in addition to the bias data. 

I bring this up because I was having a slight under correction issue  in WBPP when I used bias, flats and darks.  This mostly was only a problem when I used a lot of stretch.  I tried everything including not using bias frames, using different histograms, using a flat panel vs sky flats, etc.  and the under correction remained.  The one thing that fixed the problem was adding dark flats.  I get well calibrated data now.  This is with a IMX455 sensor at -10C, the flats and dark flats were 25s exposures.  A 25s exposure has negligible dark current with an IMX455 at -10C.  I visually inspected both the master bias and master dark flat and both looked identical.  Both had a signal floor that matched the camera plateau/offset.  So all I could conclude is that WBPP's calibration math when using dark flats is a little more accurate in this situation.  

Just curious if you've come across something like this.  

Best Regards,
Mark

Flat darks are generally only needed for flats taken using a very long exposure--typically at dusk or at night (yes, that is possible).   Bias doesn't just remove "pattern noise", the bias signal may also contain offsets, which is important for the correct computation of the calibrated result.  Some cameras have very little offset but not all do.  In my experience, the calibration routine either works or it doesn't...so no I've never seen it "partially" work.

John
Well Written Helpful Insightful Concise
Mark Petersen avatar
Mark Petersen avatar
John Hayes:
Mark Petersen:
John Hayes:
Difficulty with flat correction falls into 4 general categories:


4) Improper calculation.  Improper calculation happens when you don’t include bias offset, flat darks, or if the data is taken outside of the linear response region.


John

Hi John,  

Can you expand on the use of flat darks?  I know many CMOS users no longer use bias frames because their electronic pattern noise is minor, but in theory the only difference between a bias frame and a flat dark is the flat dark contains a small amount of dark current in addition to the bias data. 

I bring this up because I was having a slight under correction issue  in WBPP when I used bias, flats and darks.  This mostly was only a problem when I used a lot of stretch.  I tried everything including not using bias frames, using different histograms, using a flat panel vs sky flats, etc.  and the under correction remained.  The one thing that fixed the problem was adding dark flats.  I get well calibrated data now.  This is with a IMX455 sensor at -10C, the flats and dark flats were 25s exposures.  A 25s exposure has negligible dark current with an IMX455 at -10C.  I visually inspected both the master bias and master dark flat and both looked identical.  Both had a signal floor that matched the camera plateau/offset.  So all I could conclude is that WBPP's calibration math when using dark flats is a little more accurate in this situation.  

Just curious if you've come across something like this.  

Best Regards,
Mark

Flat darks are generally only needed for flats taken using a very long exposure--typically at dusk or at night (yes, that is possible).   Bias doesn't just remove "pattern noise", the bias signal may also contain offsets, which is important for the correct computation of the calibrated result.  Some cameras have very little offset but not all do.  In my experience, the calibration routine either works or it doesn't...so no I've never seen it "partially" work.

John

Thanks John, if I recall Adam Block mentioned using dark flats in the context of long flat exposures too, I think his guideline was 35s+.  But weird that flat darks at 25s worked for me, while bias, flats, and darks alone did not.  I suspect it might be a difference in the calculations WBPP uses when flat darks are involved.
Concise