Bob, I went back over this thread - an excellent one - and did not find what led you to determine that you were underexposing your flats. Can you expand a bit? What is your target exposure for them now?
Did you mean _this_ thread?
That wasn't a flat. The DSLR light above was underexposed based on its ADU, which I glanced at. The calculation is a bit complicated, and I didn't do it (no point), but my experience told me that ADU was too low. Back then I had very little experience, was pretty much just winging subexposure.
Flats exposure should be easier than most people make it. You want the _linear_ histogram to be about in the middle. Exact doesn't buy you anything, you just need to keep away from the edges, where sensors can get non linear. With a color camera, the 3 channels will generally be separated some, also not a problem.
I have a problem, I can't not answer a question <smile> The ADU of the light was too low, because I generally set my subexposure so that the bias corrected ADU of a light, converted to electrons, is 5-10 X the read noise squared. If you want to know why, the book I cited in #9 has a terrific (but lengthy) explanation. Warning - math ahead.
That's very similar to the method Glover uses. We're both trying to be sure read noise is minimal compared to sky noise.
Hi again. Sorry, I messed up a bit, it was
Astrolamb's post, not yours....confusion from embedding - my bad.
"I also learned I have been severely under exposing my flats thanks to your advice!
It was incredible to see just how much more detail actually came out and how much better the star color was overall."
You will have to bear with me because this is a relatively new concept for me, but explaining it in my own words and having bobzeq25 correct any mistakes I might make will definitely help solidify the concept in my mind.
What led me to the realization was finding empirical data for what a proper exposure is for subs.
Because all the data is captured linearly, and any sub you see on the back of the camera or within an acquisition software is actually stretched and no longer linear, you can use the average values of the pixels to estimate a proper exposure. This value is measured in ADU.
If this is your first introduction to ADU, imagine your histogram as an x, y chart(which it is). The horizontal axis is measured in ADU and the vertical axis is measured in Density. So in other words ADU would be the pixel value and the density would be the number of pixels with that value.
If a proper exposure should be capturing just enough data that the read noise in a bias frame can be subtracted away and leave true data across the entire frame, then my flat my flat frames as I saw them in image previews were actually extremely dark on the edges if left in the linear format.
Stacking software only works with the linear data, it does not apply any stretch or alteration to linear data, if it did it would alter our photos far too much and it would make post processing much more difficult. Since it only works with the linear data and my flat frames were so dark linearly, the correction it applies would be clipped at the edges of the frame and possibly less accurate across the entire frame as well.
I happened to find another thread as I was thinking about all of this that was discussing the proper exposure for a flat frame in a linear format measured in average ADU across the frame. So I tested this with my own setup (happened to be really good timing since I just received my new led tracing board to take my flat frames with).
Since my Nikon D5300 is natively a 12bit camera, it can capture up to ~16000 ADU as a maximum. A bias frame registers as ~600 ADU. To capture a data across the entire frame with my camera I should aim for about 1000 ADU, This prevents me from washing out the stars as I stack my data together and it preserves the level of details in the shadows through the highlights of nebula and galaxies.
With my flat frames that I thought were properly exposed (1/2way through the histogram on BYN, and showed a nice even gradient across the entire frame) viewed linearly they are entirely black. I measured that the data starts below 1000ADU and peaks near 1400ADU. For effective flat frame calibration you should be utilizing as much of the linear data as possible right?
Since you want your flats to have the best possible effect across the entire frame, you should be aiming for an average value right in the middle of the possible range. Since that range is 0-16000ADU for my camera I aimed for 8000 ADU. This means that the brightest sections still have a perfect gradient without being washed out and the darkest sections have a perfect gradient without being clipped.
On BYN the preview photo looks completely washed out and the histogram is almost all the way to the right without being clipped off(probably close to 5/6 if I gave it an estimate). But the result of using the flats that were exposed to this level was blatantly apparent.
As far as numbers go, my flats before only took up a range of 800 ADU out of the available 0-16000ADU where my new flats take up a range of 7200 ADU out of the available 0-16000ADU. That alone should indicate how much more accurate the gradients in the flat are and how much better they actually correct my sub exposures.
I hope that was somewhat clear and bobzeq25 please correct me anywhere I am wrong or too ambiguous in my explanation.
**I must give credit where credit is due and thank bobzeq25 again for sharing this information on so many different threads, every time I searched for more information about this topic bob was in the comments sharing a wealth of information on the subject. I can really credit my understanding of the subject almost entirely to him.**
I almost forgot to suggest a free program called Iris to examine your ADU if you are interested. I also found this because of bob. lol
Edited by Astrolamb, 13 April 2021 - 03:59 PM.