Darks library for DSLR

[Pryce]

So, to maximize my imagingtime on the clear nights I'm considering creating a library of darks in different temperatures.

I figure that its a good project to do when I have clouds.

 

So my question is,  how often do I have to change my darks if I do go for a library? 

Would it be a good idea to get 50ish in different temperatur ranges and then taking 2-4 each imaging session and replace the oldest ones in the library? 

Or should I just do this project every now and then in different temperature ranges?

 

And lastly.  Should I have separate folders for each degree or can I create ranges from say 5 to 1, 0 to -4, -5 to -9 and so on? 

[alacant]

Hi

I wouldn't bother with dark frames. Even the best dark optimisation algorithms (the one in Siril is the best we found) can't match the noise reduction offered by modern clipping routines.

Instead, use bias frames to calibrate your light and flat frames. Dither between the lights and stack using a clipping algorithm. Beats dark subtraction. For us.

But of course try both methods as YMMV;)

Cheers

Edited January 16, 2021 by alacant

[vlaiv]

Dark calibration is not about removing noise - it is about removing signal, one that should not be there and one that can mess up your flat calibration and such.

I think that it is good thing to use darks - even with cameras where you can't control darks. Before you do that, and to make your life easier - you could do couple of tests.

- see if your camera has temperature sensor and if you can get it to record it.

- if not - think about getting external temperature sensor that you will use to record ambient temperature

- in controlled ambient temperature (does not need to be cold - it just needs to be stable enough - like not more than 0.5-1°C of change during testing) - test if you can achieve stable and repeatable darks and if your bias is working properly

- change ambient temperature (again, make it stable if you can) and do another set of darks. See if dark scaling is feasible and maybe determine your doubling dark temperature

If all above checks out, and you find your dark doubling temperature - then you only need one duration of dark subs at one temperature to do dark calibration.

[Pryce]

My camera has a temperature sensor and records the temp in the file name! 

 

But I'm not sure I understood that last part!

[bottletopburly]

Just dither if you can less hassle having to take Darks as alacant says .

[vlaiv]

9 hours ago, Pryce said:

My camera has a temperature sensor and records the temp in the file name! 

 

But I'm not sure I understood that last part!

In order to see if you can use dark subs and bias subs - you need to check several things.

First thing to check is if camera bias is stable and are darks repeatable. Sometimes sensor/camera manufacturers create internal self calibration that gets done each time you power on the camera.

This leads to different offset levels each time - which is a bad thing for our use case.

You need to check following:

- Take set of bias then power off camera, power back on and take another set. Stack both sets and subtract them. Resulting image should have 0 average ADU value and ideally be pure noise (no patterns). In reality, you'll most likely see horizontal bands in resulting image and FFT of it will have two distinct dots

- Next, take a set of darks at certain temperature (can be room temperature, but important thing is that ambient temp does not change), power off camera, power on camera, take another set of darks at same temperature but twice the exposure length. Stack first set of darks, stack second set of darks, remove bias from both (you can use bias subs from previous point) and then multiply first image with 2 and subtract from second.

Again, you should get image that has 0 mean ADU value and not visible patterns (in image or in FFT).

This means that darks and bias are working as they should - at given temperature, dark current linearly depends on time.

Next you want to know your dark current doubling temperature. Dark current depends on temperature in exponential way. If you increase temperature by certain constant - dark current value increases by factor of 2. This is around 6°C for most sensors. You need to establish this value because that will let you manually scale your master darks.

There are algorithms that try to automatically scale your darks for you, but I guess it is better if you measure things and apply correct factors yourself.

For this measurement - you need two sets of darks at different temperatures. Stack both, remove bias and see how much mean ADU changed for certain temperature change - calculate doubling temperature (if you want to be really precise with this - take at least 5-6 readings at different temperatures, plot result and do curve fitting). For example, here is one such curve for set point cooling camera:

Note that Y scale doubling with each mark and line being almost straight line most of the time above -10°C (it does have small kink at very low temperatures).

Once you have all those things - you can shoot your master dark library - choose temperature and duration, say 3 minutes at 20°C.

Now you have taken 2 minute lights on particular evening and they were taken at 14°C (and your dark doubling temperature is about 6°C). How do you apply your dark?

First you take your master dark (with bias removed of course) and divide it with 180s to get dark current per second at 20°C. Next you divide that with x2 - because you want to get to 1s at 14°C, and in the end you multiply it with 120s to match duration of your light.

If you don't want to go thru all that trouble - just make sure your bias is stable by doing point number 1 (take two sets of darks with different exposure lengths and one bias and try to do dark-dark calibration) and later on, rely on dark scaling to do its thing. That is algorithm that tries to figure out proper scaling factor to match darks to lights.

(Dark optimization is automatic thing, but there is also check box to the right - dark multiplication factor - that one is for manual multiplication factor that we calculated in our example above - use one or the other depending on your approach).