Do you get perfect frames after calibration or have to do DynamicBackgroundExtraction (DBE)?

Calibration frames can only correct for the things in our control, which is generally the things from the OTA entrance to the sensor.

The are things like atmospheric or light pollution effects that can still create gradients or other artifacts that will still need various tools and tricks to fix.

I use the Wipe module in Startools for this.

Not quite sure what the question is here.

If you still have vignetting or circular gradients of some kind after calibration then your calibration frames do not work, simple as that.

Usual suspects for failed flat frames are things like an insufficient focuser that sags under gravity, a wobbly tube (such as an OOUK VX aluminium tube) that buckles under gravity, a reflector with mirror cell stability/flop issues, or a reflector that is not sufficiently flocked to deter internal reflections. Light leaks thru the gaps in the focuser or the back of the scope are another flat-ruiner for newtonians. With refractors most of the above is a non issue, but focuser sag can still ruin flats.

Yuxuan:
I was only thinking about removal of vignetting for long focal length shots, for which light pollution does not cause any recognizable gradients.

Not really true, there will be a gradient in any image that is pointed to a non uniformly lit part of the sky. Which is basically everywhere on earth except under perfect wilderness darkness to the zenith. Maybe that last part was just a guess, but the spot i use sometimes goes down to SQM 21.3 and there is still a gradient even to the zenith with an image as small as 20x20 arc minutes so lets call it an educated guess.

So to answer your initial question about what the effectiveness of calibration frames should be: They should work perfectly, and if they dont then you have to work on your kit to make it so.

Jared Willson:
Calibration, as far as I can tell, perfectly removes vignetting and dust spots from my images. Obviously, it does nothing whatsoever for sky gradients, so I always need to run something like DBE to address those. I usually include local normalization into my pre-processing so that the gradients, while still there, are as simple in structure as possible. Then DBE or whatever tool you like for gradient removal does a really, really good job.

As to the trick? There isn't one beyond making sure you get really good calibration data. The scope needs to be set to infinity focus (or reasonably close) for flats. The camera can't have been removed between flats and lights or dust spots may move. I always include dark subtraction from both my lights and my flats so the math works correctly. The light source I use for flats is very consistent and even. As far as I know there are no "tricks" in getting good calibration--just care such that you have good quality flats. You can't, for example, hold an iPad up to your telescope and wave it around and expect that to give you nice even illumination. If you use sky flats, you need to make sure you don't have any passing clouds when taking flats. So, care in taking the flats is really all you need.

- Jared

Thanks for the detailed answer. I am using a 24"x20" dedicated flat light panel, taking darks, flats, and dark flats, but still end up with a small residual vignetting. The darker the object, the residual vignetting becomes more obvious after stretching.

I only came across this problem when I started using a 12.5" CDK and a full frame camera. Initially, the relative high degree of vignetting of the scope, the large size of the objective and the sensor did pose a lot of challenges. I upgraded my flat panel and started using NINA's flat wizard. I am doing everything you mentioned above. The only thing is that the flat and dark flat frames are both taken at room temperature without cooling, so the temperature drifts a bit and they differ by a few degrees.

Jared Willson:
The flat and dark flat being taken at warmer, unregulated temperatures should not matter at all. Dark current is negligable with such short exposures, and shot noise will completely dominate at the brightness levels typical of a flat frame. Honestly, not sure why you would still be seeing vignetting. It's possible that your flat panel isn't quite as flat as it should be, but I would be surprised. It's also possible that you are getting some stray light in your flats from some source other than the light panel. Are you taking your flats during daytime or night time? I usually take mine on cloudy nights when my observatory is closed up--no light sources other than the panel. If you are taking daytime flats and just assuming the flat panel will dominate, I might try flats in the dark. Not sure whether you are in an observatory or not--the use of a large panel suggests you are. Use a cloudy night when the observatory is closed. Ideally, I'd regulate the temp when taking the flats and flat darks also--it shouldn't matter and almost certainly has nothing to do with your vignetting issue, but why not get the best quality data you can? You can set a higher temp than you would at night, but at least regulate the temp so it's consistent.

I am not using an observatory, but I park my scope buggy in a dark garage at night when I take flats. 

To be clear, the residual vignetting/unevenness is small. When I do I DBE on the calibrated and integrated frames, the background model only has a <2% variation. They do not look like gradients as toward the corners the boundaries of dark regions are curved. 

For brighter targets like M31, the image appears quite even. However, for fainter objects (against sky glow) when aggressive stretching is needed, the residual vignetting shows up. Do you have similar experience, or are your flats always perfect no matter under what amounts of stretching?

For vignetting, you need to use flat frames taken with exactly the same optical train - no changes to focus, rotation etc. I typically do these the following morning.

Getting the correct exposure / ADU can be difficult. I now use standard 2089 (20890) for my Altair CMOS cameras, and Stellarmate EKOS does a great job of automating calibration. In the past it was a matter of some trial and error and a great deal of frustration.

For light gradients, I now use GraXpert - the DBE model works well as does Siril, but their AI model is fantastic (if slow on my PC). Start with good calibration or you will never get a good result.