I have a question for the more experienced in here.When I first started learning about astrophotography the first thing I learned was "You need darks,bias and flats to calibrate your lights" and I thought that was it.Then things like "You don't need bias frames as they are included in your darks" or "You MUST NOT use bias frames if you have dark frames" or "dont use anything as dark level is suppressed in sensor (which is true),read noise is already low (for some cameras anyway) and flats can be replaced with lens profiles (well if you are using a telescope that is not going to work).Now these people they all create amazing images and are involved in scientific projects or astrophotography software development so I trust they all know exactly what they are talking about, the problem is they dont agree with each other and that leads to confusion for a mere mortal that I am!So my question is what is your standard approach?Do you choose your method depending on your imaging conditions (for example temperature)?
There is no single answer that works for every setup. I know Roger Clark frequently tells people they "don't need to use darks because the camera has on-sensor dark current suppression", for example. However, that is only true for much newer DSLRs, and in the cases where dark current is actually suppressed to a truly effective level that you can actually ignore dark current entirely is...well, basically ONE camera, and only in the winter months. However, that is usually not made very clear, and I know a lot of beginner astrophotographers who hear such advice, such as yourself, and end up confused and frustrated with their own astrophotography as a result.
There is a LOT of confusing rhetoric out there like that. The simple fact of the matter is, it really, really, really does depend on the camera you are using. However, here is one fundamental rule about darks that is true for pretty much every camera:
A dark frame contains more than just the "dark current". Dark current is one of many aspects of the "dark signal" that every digital camera contains, and it is often the least important aspect. The more important aspects of "dark signal" include the fixed bias pattern (i.e. fixed banding), the fixed dark current pattern (i.e. hot and cold pixels), semi-random dark current pattern (i.e. signal interference from an external source, clock signal interference, etc.), and glows (i.e. amplifier glow). Subtraction of a dark, either a single dark frame or a properly constructed master dark frame, is about eliminating pattern noise, not about reducing the random dark current noise.
As a matter of fact, subtracting a frame of pure random noise from another frame of pure random noise actually results in MORE random noise...however, a small increase in random noise is usually VERY preferable to leaving the fixed pattern noises in place, as FPN is something our senses latch onto far more easily, as FPN is very unsightly and unnatural.
Dark current is technically a signal, that follows Poisson counting statustics, meaning that it effectively behaves the same as photons of light...it's noise is random. The random noise in dark current from most modern cameras is very low when the camera is operating at lower temperatures, low enough to not be of major concern. Additionally, if you stack many dark frames together, you average down the random noise, making it even less of a problem. Dark current grows consistently with time, and can introduce a growing offset in the total image signal as more dark current accumulates. Since dark current itself is Poisson, removing it is actually quite simple...you simply subtract the offset. That will remove the signal part, leaving behind only the extra dark current noise (which is usually very small).
Bias is also a signal, however it is a fixed signal for the most part. It is the offset in each pixel that occurs due to the voltage applied to the pixels during readout. The purpose of bias is to ensure that there is enough of an offset from ZERO voltage to give room for noise to be noise, which can result in pixel values ending up smaller than they should be, as well as larger. This avoids clipping. The thing about bias is that every pixel and every row or column (usually column, due to how most sensor readout systems work) will have a different response
to the applied voltage. This is where most vertical banding comes from...and this form of noise is a fixed pattern. The handy thing about true fixed pattern noise is it subtracts out easily. All you need is a clean master bias, or a clean master dark, and the bias pattern will be removed.
A note about the sensor bias. EVERY frame created by the camera will always have the bias signal in it. Bias is an intrinsic trait of the sensor. A bias frame contains basically just the bias signal as well as read noise, and nothing else. A dark frame contains the bias signal, read noise, and the dark signal, which includes the dark current noise, any dark fixed pattern like hot and cold pixels or any fixed or semi-random banding in the dark, as well as any glows. A flat frame contains the bias signal, read noise, the dark signal, as well as the photoelectric flat signal. A light frame contains the bias signal, read noise, the dark signal, as well as the photoelectric image signal. The bias is in EVERY frame! Therefor, if you subtract only a master dark, you will be subtracting the dark signal as well as the bias, and all of the aggregate fixed pattern represented therein.
The only time when you need to use a discrete master bias is when you are using flat calibration without flat darks. There are several ways to calibrate...depends a lot on what you are doing. In my experience, narrow band imaging only needs a single master dark. There isn't enough light with narrow band imaging to really create huge problems with the field
....so, you don't usually see vignetting, and if you are careful and keep things clean, you also usually won't have issues with dust motes. That means you can eliminate flat calibration in this case.
Now, there are two other calibration options. In these two cases, flat calibration IS necessary. However, you can handle the flats in one of two ways. You can do a bias/dark/flat calibration, or you can do a light/dark and flat/flatDark calibration. The former can be used if you have minimal dark signal and/or are using very short exposures for your flats. You use a master bias to calibrate EVERYTHING...every dark frame, every flat frame, every light frame. You then further calibrate the lights by subtracting the calibrated master dark and dividing the calibrated master flat. This only works if you do not have any kind of glow...and glows, particularly amp glow, have become much more common these days (sadly...they used to be a thing of the past, however with all the video capabilities, it seems that amp glow has been reintroduced and seems to be a fairly endemic feature with video-capable cameras). If you have glows, then the other calibration path is usually better. Because the bias is included in every frame, you can create a master dark for your lights (and this master dark is usually reusable for two to six months as long as you can keep the temperatures well-matched), and you create another master dark for your flats. You then just calibrate with these two master darks, one for each type of frame, then further calibrate the lights by dividing out the master flat. You do not need a master bias in this situation.
In the end, calibration is about reducing the noise sources in each of your lights to just read noise and photon shot noise, as much as possible. You want to eliminate patterns, first and foremost...banding, hot/cold pixels, glows, any form of non-random, non-image structure, especially unnatural structures that the eye can latch onto. This is very important with stacking, as anything fixed in nature, vs. random, will actually be reinforced
with stacking. If you do not eliminate fixed pattern noise, then you will hit a brick wall in regards to how many subs you can stack and still benefit from the noise reduction inherent in stacking (random noise is reduced as the SQRT(SubExposursStacked)).
Fixed pattern noise will eventually show through the fainter background areas of an astrophoto extremely well, and it looks hideous. Calibrating with bias/dark/flat can increase random noise...however, if you use sufficient numbers of frames for each of your masters...such as 100 biases, 25-49 darks, and 36-49 flats, then you will average out the noise in those frames as well, so the increase in random noise in each calibrated light becomes quite small. If, for any reason, the contribution appears to large, you can always stack more frames into your masters...again, the random noise in the masters drops as the square root of the number of frames stacked.