Sony A350 its own Darks ?

[johnb]

Hi

I’m currently use an unmodded Sony A350 for DSO imaging, if I take say a 30 second exposure then camera then process the image for 30 seconds, I think its creating a dark and subtracting from the exposure ?

Am I right in which case I don’t think I need to create my own darks ?, of course I could be so way of

Regards

John B

[jgs001]

John, that's correct. If you have ICNR (In Camera Noise Reduction) active, then the camera will shoot a dark per light frame and auto subtract. You therefore don't need to take darks. But.. You will take twice as long to capture the same exposure time of lights.

Let's say you capture for 2 hours... with ICNR on, that will give you 1 hours worth of dark corrected exposures. With ICNR off, that will give you 2 hours of exposure. Then you cap the scope, and shoot 10 to 15 darks (at 30s, that's only 8 minutes)... It won't be as efficient as ICNR, but you will capture far more data and still have a useable dark set to work with in stacking.

[Alestrom]

I believe it is better to take separate darks and stack them, this is because the darks themselves have 'noise' in addition to the dark current, which can be averaged out by stacking. If you use the INCR you are only using one dark per sub, rather than subtracting an averaged set, and this would cancel out the effect of stacking the subs in the first place.

I hope this makes sense, I'm sure someone else will be able to explain it better than I have

[jgs001]

Actually, I was thinking the other way... that using ICNR produces a better result as the temperature of the sensor and operating conditions will be as near to identical for each frame as you can get (I could well be wrong though )

[Alestrom]

As could I...

There was a similar thread on this forum about using in-camera darks a while ago, where a couple of people pointed out that darks have to averaged as well. But then I suppose the noise would get averaged out by the stacking process anyway? I'm confusing myself now...

I take your point about the temperature though

[Alestrom]

See here

http://stargazerslounge.com/imaging-discussion/95491-darks-continued.html

and here

http://stargazerslounge.com/imaging-discussion/95357-why.html

[jgs001]

Ah... I do recall one of those threads... but I didn't fully appreciate the content about the multiple darks and random noise elements...

[johnb]

thanks for the advice i will turn off the INCR because that way I can get twice the amount of subs in as im unguided i feel thats quite important, however i will experiment as well in both modes

thanks again

John B

[narrowbandpaul]

hmmm glad i could have helped!

yes. darks have two components (3 actually but 2 that are unique to darks). There is dark shot noise, and dark fixed pattern noise.

First what creates dark signal?

well because the silicon lattice that makes your sensor (CCD and CMOS) is above absolute zero, some electrons can emitted thermally from the silicon lattice. This is dark current. The thermal emission of electrons per second.

most pixels are random in their generation of electrons. So for these guys the variation in electrons emitted is random. We call variation, noise. Random noise is reduced by averaging frames. If you could average an infinite amount of frames, these variations would average to a single number. Each pixel would contain the same amount of charge.

but there are some rogue pixels, which always like to emit more, or less than the statistical average. Since as it is the same pixel everytime, this variation is fixed. We call it Fixed Pattern Noise. But we cannot reduced this FPN by averaging several frames. Nope. This noise is a fixed source.

If your sensors pixels emit electrons randomly, then darks are useless......seriously. If every pixel emits electrons randomly, then taking a million darks and averaging will have no effect.

heres why: you cannot remove random noise. It can be reduced by averaging, but subtracting a dark with only random noise will not increase the quality, it will reduce it.

but we can very nicely subtract FPN. Yes, because it stays right where it is. Whaat we need then, is a frame where the random noise is very low, but the FPN is pretty obvious. How might we get this.

well if we take lots of darks, then average them, the random noise will be reduced, but the FPN wont be. So the FPN will be obvious. Then we subtract this off out data images. hey presto!

your image, which contains some dark shot noise and some dark FPN will now have the DFPN component removed. The image is said to have been 'despiked'

when people take darks, they are NOT removing the random, what we call shot noise, they are removing the DFPN.

you cannot remove random noise by subtraction. Only fixed pattern noise.

so i suggest you get lots of subs, then take a bunch of darks at the end. Then average, and go and despike your image!

good luck

cheers

paul