More and more, we see the term master dark flat all over the forums. These are supposed to replace the biases that are usually used in the standard preprocessing of astronomical images. We read that the result is better using a dark flat (shot with the same exposure time as the flats) than with standard biases. But is this really the case? Why such a frenzy for a master image that no one was talking about not so long ago?
Hello there, this is my first post on this forum. It is interesting that you use the 294 for your example. I have the same and found I had to use dark frames matched to my narrowband flats for a good calibration. Both my wideband LED arrays and EL panels do not produce high Ha or SII intensities. In the days of my QSI CCDs, I would only use bias frames to calibrate flats but I would like to make a philosophical point:
There is nothing wrong with using matching darks to calibrate flats. As far as I know, it is the best calibration setup in all cases (all CCDs and CMOS). While it may be possible in some cases to use bias frames as a substitute, since there is no obstacle, other than a little patience, why recommend a process that does not suit all use-cases?
In the case of reducing noise injection by using a synthetic bias, a sense of proportion is useful - if a typical flat frame has a level of 20,000 electrons, its shot noise will be 141 e RMS, far exceeding any read noise… and so seems a third order benefit.
Our article explains it is better to replace dark flat by synthetic bias in fact. Because yes, all the tests we have done show that we MODERN sensor, shooting dark flat is a huge waste of time. Why we recommend that ? In fact we recommend to test, we recommend to think to the process and to not follow what we read everywhere like a robot.
The main goal of the article is not that. But why doing dark flat when it is not needed in many cases? The goal of this article is to push people to understand the process, we have provided a lot of plots and explanations. Now people are free to follow or not. But every user I know that have tested it, have adopted it.
I certainly do not take things at face value. I’m writing my fourth book on Astrophotography and thought to include more on SiriL, as it has changed dramatically since I looked at it a few years ago. It is important, however, when recommending processes, to have them as universal as possible. This hobby is complex enough without introducing extra complexity. This returns to my original point - it is interesting to find these things but as astrophotographers use a variety of cameras and there is no specification that defines MODERN, it is better to recommend something that works in all cases?
Yes of course. This is why we encourage every user to test by himself :).
I do agree that the title is very provocative, but it is assumed and done on purpose ;).
The idea of synthetic bias comes from a discussion with T. Legault (author of many books too) that use it for a long time in a different way. To be honest, he read the article before I release it, and he was totally agree with the contain too.
Interesting - I had marked differences before in calibrated light frames, but I cannot reproduce it in PI right now. That was, however, with a QHY183. There are all sorts of undocumented things going on in sensors and I may have stumbled over another. CMOS and CCD, better the devil you know.
Interesting - I tested median, bias and matched darks on flat calibration with 40 SII exposures. Results were remarkably similar. The main etched line of the amp glow was detectable in all stretched images and there was just a slightly lighter region in the area of the amp glow in the median calibrated version. Its noise was very slightly higher (third significant digit) but that is probably due to less suppression of the amp glow and its shot noise. I cannot recall the equipment or conditions where I had very obvious cal issues - when I do, I will re-run the tests once more.
I was also alerted to another potential cause of mismatches, to do with flat normalisation before stacking. If there are saturated pixels in the reference flat frame, it messes up the flux equalization. PI has apparently added some further controls to mitigate that.
With a rejection algorithm that should be fine, don’t you think?
I’m glad to see that your tests converge with ours! But in fact, if we wrote this article, we had no doubts about it.
But many thx for being interested by Siril and to conduct these tests. Next revision of T. Legault book should talk about Siril too, and I hope he will explain this feature.
I think the jury is still out on the need for dark flat. It undoubtedly works in some cases but I cannot explain why this is not always the case, with the same sensor. Looking around there are conflicting views specifically on the 294, similar to my own.
The flux equalization issue is caused by near saturated non-rejected pixels. If the rejection algorithms is statistically based a consistent hot or warm pixel will not necessarily be rejected. It requires a range clip to avoid the issue - but also, worth experimenting by changing the reference frame for the normalisation. This is less likely to be an issue for me as I use an artificial source and set the median about the mid-point. I think it is more of an issue with skyflats.
I use a self-made flat panel, and my 5nm Ha flat is 52.5 seconds long.
When I tried to use bias, the overall result was terrible overcorrection, i.e. the light falloff at the frame edges as well as dust motes became bright.
Shooting a set of darks to match the flats’ exposure length and calibrating with them created a really clean stack to my immense satisfaction.
The other night the experiment was repeated with OIII data where flats are shorter (22s), but the result is still the same: using bias results in overcorrection.