Darks

[papaf]

Hi all,
the next week I'll speak at my local group about PI and how a magnificent program it is...
The situation I'm in is a bit difficult, though: I always used PI with my ccd, which is cooled, sony based and b/w. I'm going to show PI to people who has DSLRs or OSC ccd. I wanted to talk about how PI handles darks, among other things, and I just discovered the very clever way in which PI determines if the dark should be stretched/adjusted to match different lenght/temperature of the exposures.
First question is: did I get it right? Is it true that darks can be taken in a different moment and at a different temp and PI will magically adapt the master dark to the light frames?
Secon: if I got it right, what's the better way to shoot darks? I know it would be better to shoot them right in the middle of lights, but let's assume we're lazy and we're going to shoot them in another occasion. Is it better to take them in the worst possible way, so say room temp? And what exposure lenght?

Thanks!

Fabio

[Emanuele]

Fabio:
Darks should be taken at the same temperature and with the same exposure lenght of the light frames.
They don't have to be taken in the middle of taking the lights.

PI and other programs are able, through the use of Bias frames, to scale the darks to match the exposure length and noise of the Light frames. BUT in order to do that the darks need to be longer than the light frames.
Eg. Take darks Of 1800sec. Take Bias frames (150-200). Then use the Bias and the Darks to calibrate light frames (of less than or equal to 1800sec)

Hope this helps. Others will chime in.

[Nigel Ball]

Fabio

When I got my QSI camera the first task I did was to create a library of darks at my target temperature with is -20°C. Emanuelle is correct darks do not have to be taken in the middle of lights with temp controlled CCD cameras

Bias frames will be applicable for around 6 months.

HTH

Nigel

[papaf]

Thanks for the replies. You two kind of seem to throw water to the story, which I read in a page Juan wrote a while ago to explain how PI was handling the darks. From this page, Juan seems pretty sure about his algorithm, but you two sound pretty cautious.
I know darks should be taken every time, and I also know about taking them at fixed temp. But we're talking DSLRs and beginners, with not much time available for shooting. So I was exploring alternate routes to use darks. I don't pretend to have the absolute best performance out of them, but maybe to offer them a viable solution.
In any case, E. kind of answered it. Darks have to always be as long or longer than light subs. That's a start!

Fabio

[Carlos Milovic]

Hi Fabio

What I do (before PI's method, but this works fine) with my Canon 300D is to take darks to match the longest exposition time. I really don't care too much about temperature. Usually I do them at the end of the night, start nearly at dawn, and I leave the camera and laptop taking the darks for one or two hours, well covered.

Since what we are trying to capture with darks is the underlying "signal" due to temperature, more time implies that you'll have better SNR. So, to better represent the true dark signal, I would keep using my rule of thumb. Scaling down a well sampled signal is easy. Doing the opposite may produce ugly results.

[Harry page]

Darks what are they   

Oh a very happy starlight xpress customer

Harry


I know not much help , but it cheered me up 

[DaveS]

I gave up taking darks a long time ago, as with the very low dark current of SX cameras, they just aren't necessary.

That said, I always subtract a Bias frame, which if you do take 'darks', isn't necessary, as the Bias data in contained within the dark frame.

Dave

[Carlos Milovic]

Bias are always necessary, if you are taking flats

[papaf]

Harry, I have a sony based Atik camera so...  hell yeah! High five!

But, I do have to come up with something for my fellow dslrs users! What Carlos said is what I had in mind. Maybe one could simply put the camera outside on a middle week evening and capture just darks, right?
About biases: I thought I read bias are also used with darks. Am I wrong?

Fabio

[Emanuele]

The Bias is already contained in the Darks. Bias are only used to scale darks and for Flats.

[Carlos Milovic]

Yes, as Emanuele said, Bias are needed to rescale darks too. So, at the end, just take them. After all, it is pretty cheap to do

[Nigel Ball]

What I do (before PI's method, but this works fine) with my Canon 300D is to take darks to match the longest exposition time. I really don't care too much about temperature. Usually I do them at the end of the night, start nearly at dawn, and I leave the camera and laptop taking the darks for one or two hours, well covered.

This is pretty much what I used to do with my Canon 1000D. At the end of the nights imaging I would set the DSLR running to take a batch of Darks equal in lenght to the Lights

[astropixel]

Scaling down a well sampled signal is easy. Doing the opposite may produce ugly results.

Hmmm... this may be the answer to a question in another forum and possibly the source of earlier processing issues.

Frankly, I took darks and lights of the same exposure time, and took only 10 bias and 10 darks. Whether it matters or not, I don't know, but I took 10 flats. It seems to work OK.

[Niall Saunders]

Carlos said:

Bias are always necessary, if you are taking flats

Not if you have taken FlatDarks at the same temperature and exposure time as your Flats (as, then, you can use the FlatDarks without worrying about Biases)
(Similarly, Biases are not needed if you take Darks at the same temperature and exposure time as your Lights)

SX users claim that they never need to take Darks - yes they do, if only to prove to themselves that a MasterDark is an ABSOLUTELY IDENTICAL image to a MasterBias - i.e. if you took 30 Darks (at the same exposure time and temperature of your Lights, say 5 minutes at -20C) and 30 Biases (at minimum exposure time, and also at -20C), and then created a MasterDark and a MasterBias using the same steps in each case, and then DIVIDED the MasterDark by the MasterBias (or subtracted one from the other) then the result should be an image with nothing but '1.0000' for the DIV process or '0.0000' for the SUB process. And I have yet to see THAT result demonstrated to me by anyone - fancy giving it a go Harry? It's a nice exercise for a cloudy night!!!

[Carlos Milovic]

Since darkflats are almost the same as bias, I'd call them bias