Only using Flats and Bias for image calibration viable?

[Robert York]

Some CCD manufacturers are now claiming that the dark current of their sensors are so low, there's no reason to take darks. And that in fact, trying to dark subtract will add more noise to the images than it will take away. Even some of the newer DSLRs that come out are getting very good performance when it comes to noise. Dithering can be automated relatively easily into a guiding setup, which can help tremendously with patterned noise.

So my question is, if one uses one of these newer sensors, is dark subtraction really needed any more? Is there some solid evidence to support the idea that dark subtracting could add more noise, or at the least, not significantly remove noise, if dithering is used?

Flats are a pain in the butt to get. Darks are almost as bad. It would be nice if I didn't have to worry about darks, but I'd like some more solid evidence than anecdotal comments and marketing pitches.

(I left bias frames out of the question because they're straight forward, simple to acquire, and don't take a lot of effort. I have noticed that usually subtracting bias makes my images more noisy, but perhaps I'm doing it wrong)

[Carlos Milovic]

How many bias/darks/flats do you usually take?

[Robert York]

For Bias and Flats, either 20 or 30. Sometimes 10, depending on how quickly twilight is falling. Lately I shoot flats indoors at my wall. Works out better anyway, so I can get a lot of flats.

[Carlos Milovic]

For bias, don't use less than 50-60. I would go over 100. For flats, that should be enough, if you are near half deph of the electron well.

[Robert York]

I've recently read more about the difference between the histogram in some tools and the one in PixIsight. In short, no, the flats aren't near the half well depth. I'll have to do better next time.. but they seem to work out okay for flattening. They may be adding some noise, since they're more around the 10% mark

I'll have to double check my bias frames. I recall taking an arbitrarily large number, but they're always easy to redo.

However, my main question is, with modern (last year or three) sensors, do dark frames really help that much, assuming you're doing dithering on your acquisitions? (or maybe even if you're not?)

I just did an image up twice, one with darks and one without, and I can't tell the difference. Everything else should be the same.

[pfile]

i think it really depends on the camera. some of the sony chips are supposed to have very, very low dark current. i suppose it then comes down to your subexposure length.

you might be able to get away with only CosmeticCorrection on such a camera just to get rid of hot/cold pixels. i find that with the new Lancosz scaling algorithm in StarAlignment, any hot pixels will cause ringing artifacts which are a little harder to reject during integration than just single pixels. getting rid of hot/cold pixels with CC usually takes care of that problem.

[oldwexi]

Robert,
you are correct, to get perfect flats is already a pain, if you need darks the pain is nearly doubled.
With my QHY8 i need only flats and bias. It is not the newest chip but it is a Sony.
Darks are absolutely NOT needed with QHY8, the few warm Pixels are easy to remove.

The calibration therefore is an easy game, compared to my friends SBIG,
which is much more noisy but also much more sensitive.
Calibration with the ST11K is a everytime a drama...

Aloha
Gerald

[georg.viehoever]

Robert,

- Flats and darks dont add much to the noise if you take enough of them - its all in the math.
- You may be able to answer your questions for your camera: just take a series of darks, lets say 15, 30, 60, 120, 240, 480 seconds, and have a look at the images and their statistics. If you want to be really careful, you do the same at different temperatures as well. If they are different, then darks matter. If they are all identical, you can indeed forget about them (or shoot them once - since they are all the same).

Georg

[Robert York]

When you say identical, what exactly are you referring to? It seems like if they had a different scale factor, you could still remove it from the light images by simply averaging the light images, as long as the noise was normally distributed. And maybe I'm thinking about it wrong, but really, unless it's not normally distributed noise, you can still average it out. If it isn't normally distributed, you can randomly dither to help make the noise fit a more normal distribution.

Is it that the amount of averaging you'd have to do to remove this really high? It just seems that unless there was a pattern to your dark noise, if you took an image of white noise, and then took another image of white noise, you'd just get more white noise. I'm not a statistics wiz, so please help me out if I'm off.

I guess my point is, is the big reason to subtract darks because it's not random noise? Or is there another philosophy I'm missing?

[pfile]

dark signal and bias signal are non-random. the problem here is that people refer to dark current and bias signal as noise... which leads to misunderstandings. it's not noise, it's unwanted signal.

suggested reading is the Handbook of Astronomical Image Processing - that's the bible and it goes into calibration in depth.

rob

[Geoff]

dark signal and bias signal are non-random. the problem here is that people refer to dark current and bias signal as noise... which leads to misunderstandings. it's not noise, it's unwanted signal.

suggested reading is the Handbook of Astronomical Image Processing - that's the bible and it goes into calibration in depth.

rob

Good points Rob. Adding to the misunderstanding is the fact that, as with any signal, there is noise in the dark current and bias signals, which is why we average many darks and biases. Definitely recommend HAIP.

[boris sputnik]

 This is a picture taken with 24 x 5minutes lights ,30 flats ,200 bias. In Pixinsight the only thing I did was a Stretch/debayer/stretch.
That's it ,no darks....I will add Darks and probably process it more ,but I think it's not bad as is !
Picture was taken with a Canon T3 ( 1100D ) in a red zone via BackyardEOS with dithering on ,and PHD .