Telescope Reviews: Creating a "dark" library

Hi Roy,

You are making this un-necessarily complicated. (He said just as he proceeded to make it REALLY complicated...)

BIAS

Bias is a low-level charge that is applied to a sensor in the form of a fixed offset voltage value. It is present in every frame, light and dark.

So if you subtract a dark from a light, you have also subtracted the bias.

In this simple case, where the thermal signal matches exactly, and the exposure for the dark is the same as for the light, then you don't need to subtract a separate bias frame.

SCALABLE DARKS

Now... thermal signal scales linearly with exposure time. That means if you have a 2 min dark, and a 1 min light, you can still subtract the dark from the light... you simply let the software divide the dark by two.

The problem is, you're also dividing the bias by two. So you don't correctly subtract the bias then.

The way around this problem for scaled darks is to subtract the bias separately from both, and get it out of the equation at the start.

Then only the thermal signal is left in the darks, which technically at this point are called "thermal frames" and not "darks".

So, once you subtract the bias from both, you can scale a thermal frame down to a shorter exposure light frame, and remove the thermal signal.

Pretty cool deal... shoot 10 minute darks, and you can use them for any exposure that is less than 10 minutes long taken at the same ISO and temperature.

MASTER FRAMES

Now, moving on to master frames... the reason you create a master frame is to reduce the noise... let me explain... There is a difference between signal and noise. But everyone confuses this. What we remove with dark frames is thermal signal, not thermal noise.

However, every signal has noise associated with it. It's called photon, or shot, or Poisson noise. It's the square root of the signal.

So when you subtract a single dark frame, you are actually doing two things... removing the thermal signal, but you are also adding the thermal noise. Yes, this sounds confusing and counter-intuitive.

To reduce the amount of this real noise that is added to the lights when you subtract a dark, you create a master dark. A master dark has less noise than a single dark, because the noise is random and it averages out.

So, to get really complicated... a

light frame = photon signal + photon noise + thermal signal + thermal signal noise + bias signal + bias signal noise

If you subtract a dark frame, you have

Image frame = light frame - dark frame. This removes the thermal signal and the bias signal (because bias is also contained in the dark). The original light frame photon noise and bias noise remain, but you have added the thermal signal noise and bias signal noise from the dark frame now. (The bias signal noise should be insignificant).

So, even though we have vastly improved our image frame by removing the hot pixels and color blotches that come from the thermal signal, we have really increased the noise. To reduce this, use a Master dark and master bias, which add less noise when the light is calibrated.

Even still, the light frame technically has more noise now after calibration.

But the thermal signal is usually much more objectionable if there is a lot of it.

DEALING WITH NOISE

NOW, the way to deal with this noise is by gathering more signal from the photons that you really care about... those from the deep-sky object.

The only real way to deal with true noise is by increasing the signal in the signal-to-noise ratio.

FLATS

Flats are a completely separate subject. They correct for things like uneven pixel response, vignetting, and dust spots.

BUT... you have the equation wrong here... flats are not subtracted, they are divided. You can only do this with an astronomical image processing program.

Personally, I never use flats... I believe that uneven pixel response is small, or it is probably being dealt with already in the sophisticated image processing taking place by the image processor in the camera before the raw file is written.

I don't have much vignetting, and the little that I have I can correct with software. If your vignetting is severe, then flats are really a poor solution. You will never recover the lost signal-to-noise ratio in the corners where the vignetting is. If the vignetting is bad, you need to fix it at the source. Flats are not going to fix the kind of vignetting you get on a full-frame sensor from using an f/6 telecompressor on an f/10 SCT.

I find it is easier to keep the sensor clean than to shoot flats.

Jerry

--------------------
A Beginner's Guide to DSLR Astrophotography
http://www.astropix.com