Flats adding to images instead of subtracting in PixInsight WBPP - Forums

Topic starter

9 days ago

Astro Friends,

I have been imaging for about 5 years and have been successful using refractors and mono imaging. I have branched out this year and now have an SCT and mono setup. I am seeking help with my flats which seem to add themselves to my images in pixinsight rather than subtract using WBPP. This problem has me stumped and I am sure it is something simple. Below is more information that could help in solving my problem.

My setup:

C9.25 - Standard Celestron

Starizona SCT 0.63 Corrector

ZWO EFW and EAF

Chroma 1.25” LRGB

ZWO 533MM Camera

Below is a raw luminance image:

📷 L Raw Stretched.jpg

Here is my master flat for L:

📷 L MasterFlat.jpgHere are the statistics from my flat:

count (%) 100.00000

count (px) 9048064

mean 2.512512e-01

median 2.625194e-01

avgDev 4.611702e-02

MAD 4.279369e-02

minimum 4.761905e-02

maximum 3.344004e-01

Here is my calibrated L file:

📷 L Calibrated Stretched.jpgI use an LED panel on its dimmest setting diffused with a t-shirt and white paper to take my flats. This process has always worked for me with my refractors. This particular flat was 3 seconds long with an ADU of around 27k. I used WBPP to process, but as you can see it seems to add the flat rather than subtract it.

In WBPP I loaded my master darks for my frames, the raw flats (35 each channel), and my lights. I did not use dark flats and did not check optimize darks or flats. I used the standard settings in WBPP.

My thoughts are:

Could it be my LED panel or the way the flat is taken?
User error in WBPP?

Any thoughts or have you encountered this issue before?

Thank you in advance,

Scott.

Well Written Helpful Engaging

AstroRBA

9 days ago

Are you sure that you didn’t mis-enter darks/bias with the flats?

Well Written Respectful

9 days ago

You need to use dark flats, otherwise WBPP cannot correct properly. I had this same issue, taking dark flats solved it.

Scott Chambers

Topic starter

9 days ago

Thank you! Ironically this is what AI suggested as well. I did try adding dark flats to WBPP but it strangely added them into one category and didn’t seem to do anything with them. The result was essentially the same. I usually image with my 2600MM and don’t use them with my refractor. How did you add them in WBPP and what settings did you use?

Well Written Respectful Engaging

pfile

9 days ago

with that sensor, bias frames will also work to calibrate a flat. but one way or another the flat subs need to be calibrated or you get exactly what you see in your miscalibrated light.

9 days ago

Scott Chambers · Jan 2, 2026, 04:15 AM
Thank you! Ironically this is what AI suggested as well. I did try adding dark flats to WBPP but it strangely added them into one category and didn’t seem to do anything with them. The result was essentially the same. I usually image with my 2600MM and don’t use them with my refractor. How did you add them in WBPP and what settings did you use?

While it pointed you in the right direction, generally speaking I wouldn’t trust AI for anything in this hobby since corps like OpenAI train off of generic internet slop. Niche hobbies suffer the most from this.

I would take 100ish bias frames, 25 darks of equal length of the flats. Can’t go wrong with full calibration ;)

Add them in the same way you add bias, darks, flats etc in WBPP, then you go to the calibration menu and assign them to the flats by clicking on the flats, pressing optimize master dark, and then selecting a dark frame to use on the flats. AFAIK optimizing the master dark automatically applies the bias to the calibration for both flats and lights.

Helpful

Peter Hannah

9 days ago

Just to clarify some terminology, flats are not subtracted from lights, the lights are divided by the flats after each has been dark- and/or bias-subtracted.

Vignetting makes the corners of your frames darker than the centre, both lights and flats. When you divide your lights by your flat, you’re dividing the corner values by a smaller ADU count than the centre. Dividing by a smaller number gives a bigger result, so the dark corners of your lights get boosted.

What should happen is that the process restores balance to the overall illumination. In your case the flat-fielding process is over-correcting, boosting the corners too much.

It might be worth checking that your flat panel really is giving equal illumination over your scope’s full aperture. Also I assume that you’re taking flats at the same temperature as the lights.

It shouldn’t be necessary to take flat-darks - as pfile says above, a master bias frame should be adquate for calibrating flats of 3s exposure.

Well Written Helpful Insightful

Jan Erik Vallestad

9 days ago

To me the flats looks mis-matched. The dark areas and dust donuts seem to be much darker than in the lights, which may be why they over-correct and end up brightening those bits in the final image. It seems like a lot of vignetting for such a small sensor though.

I’m using a SCT too but not with a reducer since I struggled with lots of internal reflections within the Celestron reducer I own.

Helpful Insightful Respectful

Francis C

9 days ago

I've also had an LED panel with a dimmer for three years.

I stopped using it because I think the PWM dimming module is faulty.

To reduce brightness, these tablets don't lower the LED voltage, but make them blink very rapidly.

I use an ASIAIR, and the flat-field light distribution on the histogram wasn't at all what one would expect.

Since then, I've only been shooting flats with a white t-shirt before considering buying a proper flat-field box with a real dimmer, or even without one.

One test would be to leave the LED panel at maximum power.

I don't know if you have the same problems, but it might be worth looking into.

Helpful Insightful Respectful

Larry Cable

9 days ago

interesting, as I just purchased a Starizona x0.63 Apex-L reducer and its 1st light results with a 2600MCP produced a similar issue…

John Hayes

8 days ago

With 3 second exposures, flat darks are not your problem. That looks like a lot of light falloff for such a small sensor. I think that you might simply have too much vignetting to get good correction. How big are your mounting tubes?

John

Well Written Insightful Respectful Concise

Larry Cable

8 days ago

I’m assuming with a 30mm image circle with a 48mm barrel diameter any vignetting would occur as a result of the reducer clipping the native image circle of the scope …

Scott Chambers

Topic starter

8 days ago

Hello all,

I want to thank you all for your responses. After reading your suggestions and doing a bit of my own homework I figured out the following:

Flat Darks are indeed necessary with the 533MM sensor. For whatever reason pixinsight cannot properly calibrate the flats without them.
When you enter the flat darks in WBPP you must reduce the exposure tolerance to zero in order to separate them from your regular darks. Also, you must assign them (or at least I did) to each specific flat for them to process in WBPP.
I appreciate the lesson in flats dividing not subtracting - that makes perfect sense.
I think for now my LED panel will work, but I do agree that the pulsing nature of LED’s is probably not the best for flats.

Below is the final result from my initial data set for this telescope.

📷 NGC2403 Final 3.jpg

I appreciate your responses and the discussion!

Scott

Helpful Respectful Engaging Supportive

Larry Cable

8 days ago

@Scott Chambers did you try stacking in Siril just to compare results?

pfile

8 days ago

Scott Chambers · Jan 3, 2026 at 12:02 AM
Hello all,
I want to thank you all for your responses. After reading your suggestions and doing a bit of my own homework I figured out the following:
Flat Darks are indeed necessary with the 533MM sensor. For whatever reason pixinsight cannot properly calibrate the flats without them.

sorry about that, i thought we were talking about a 2600MM for which bias frames will work fine for flats. i dont know about the 533 but there are definitely sensors out there that need darks for flats for one reason or another

rob

John Hayes

7 days ago

Scott Chambers · Jan 3, 2026 at 12:02 AM
Flat Darks are indeed necessary with the 533MM sensor. For whatever reason pixinsight cannot properly calibrate the flats without them.

Unless that sensor has extreme dark current, with 3-second exposures, the only reason that flat darks would be necessary would be if there’s a large bias offset in the flat data. If that’s true and there isn’t an offset being added by the acquisition software, it shouldn’t matter whether you use dark data or bias data for the flat dark files.

John

Willem Jan Drijfhout

6 days ago

Using bias or flat darks to calibrate your flats is unlikely to cause the differences you demonstrated. Either should work well. What is important is to assign the right darks to the right frames. You seem to have taken a more cautious approach with that in WBPP, which is good and may have just solved the problem. Sometimes it just makes sense to switch from 'auto' in the calibration settings to a manually chosen channel. That way you are sure the right calibration files are used for each flat/light channel.

It is a lot of vignetting though for such a small and square(!) sensor. Anything else that could be in the way? John mentioned width of the tubes between filter and reducer as an example? Have you calibrated the EFW? The ZWO EFW is known to occasionally get out of calibration, resulting in the filters not straight in front of the sensor. Often this is fixed by just running the calibration routine in the driver once.

It will be hard to find a LED panel that does not use pulsing LEDs (PWM). If you want something non-pulsing, you can go the Electroluminance route, but those panels are usually less bright and age noticeably. It causes especially the red channels (Ha, SII) to become weaker. But PWM's are not an issue. When the electronics is carefully selected and specifically tailored for use in Flat panels, you should never have a problem with LED panels. I have dug deep into this subject a while ago (if interested, see here), and it is all about the use of good electronics and proper design choices. The Flatpanels from DeepSkyDad for example have the best control that I have ever tested, from the lowest level (can't see it is on) to full brightness. In any case, the flat panel is not the source of your issues.

Helpful Insightful Respectful Engaging

Larry Cable

6 days ago

I believe the image circle of the SCT reducer (27mm) is smaller than that of the 9.25 itself…