Is there such a thing as too many Darks?

[Earl]

I have 80 Darks for my -20 300s.

Now due to our delightful weather I have plenty of time to gather Darks :|

But where do you draw the line?

Also, what sub lengths?

I currently use 300s and 600s and have darks to match.

[JimStan]

The important thing about darks, as I understand it, is that they be taken under the same conditions as your light subs. This way, the noise and other artifacts will be balanced out, and can cancel each other in the stacking process, leaving only the data from the desired photons to create the picture. That is the theory, anyway.

I have been somewhat successful in using a Photoshop technique of balancing a "subtraction" layer with the base layer, and getting rid of a lot of "amp noise" etc. without resorting to darks and bias frames. But as I have mentioned before, I am still climbing that steep learning slope and I know my images will be better once I have mastered the use of darks, flats, and bias frames!

Good luck! Jim S.

[Ben Ritchie]

Also, what sub lengths?

I currently use 300s and 600s and have darks to match.

I generally use 6-10, enough to get a decent median combine. You don't really need more than that. You don't need bias at all if your dark length matches your light length.

As for lengths, there are two options. You can shoot the same length as your lights, or you can shoot shorter and use bias to scale them. In either case they must be the same temperature - with a non-setpoint CCD that usually means doing them every session, but with temperature control you can use predefined temperatures (e.g. -25, -20, -15) and build up library darks.

The reason that short darks with bias work is that any light frame contains (to a first approximation) three things, signal (your target), time-dependent thermal noise, and time-independent bias current. A dark frame of the same length will contain the same level of thermal noise and the bias, so subtracting will (in theory) leave just the signal. However, as thermal noise is linear with time, you can take a light of 600s, and use a library dark of 300s by multiplying the thermal noise by 600/300 = 2. That would give you double the bias too, i.e. you'd oversubtract it, so you first subtract the bias, multiply the remaining signal, then add the bias back - your software does this for you automatically. But it only works if the light and library dark are at the same temperature, hence the need for setpoint.

Make sense?

[swag72]

I understand that for darks to be useful they are as much about having the same temperature as the same exposure time. I started to create a darks library based on exposure length AND temperature. So if you are taking your darks at a different time and, for example, your temperature varies widely from your lights, I'm not sure they'd be as useful as they should.

Doesn't answer your question, Sorry!! But just gives you another headache!

[Stephen]

As above, same exposure length and temperature as the lights.

I believe I've read on here that due diminishing returns you start to get little benefit above 20 darks. (I may be wrong!)

[RobH]

I generally use 6-10, enough to get a decent median combine. You don't really need more than that.

I disagree with that Ben....you need to treat darks in the same way as you would lights, and a stack of 6-10 lights is nowhere near enough to get a good S/N ratio. A decent S/N ratio stars at about 20-25, and gets really decent at around 50. Beyond that, you'll see a gradual improvement up to around 100, after which it's almost impossible to see much improvement.

Using darks with a bad S/N ratio will add noise to your imasges, so (and this also includes flats and bias), a minimum of 20 is needed, and preferably more.

As you can shoot darks anytime (use a fridge tio get the correct temperature, or your camera may have set-point cooling), there is no reason not to have enough.

Rob

[FraserClarke]

Using darks with a bad S/N ratio will add noise to your imasges, so (and this also includes flats and bias), a minimum of 20 is needed, and preferably more.

That is very true Rob -- the data are only as good as the calibrations It's also worth considering *where* the noise in your final image is coming from though. For most situations, the overall contribution to the noise from the darks is pretty small, so having a really really high S:N dark doesn't improve things a lot. That depends on the camera though, and it may be that the dark noise is dominant in some situations, and then it is worth having loads of them. I guess for things like DSLRs it is much more important than for 'proper' CCDs.

I'm with Ben though -- I usually take 10 darks. I notice no benefit from taking more than that with amateur type CCD camera. Out of interest, most professional instruments probably take 3--5 darks; but the detectors here are usually lot higher quality, so the dark is better behaved.

As for the scaling of S:N, it goes as the square root of the number of exposures. So to double your S:N, you need to take four times as many exposures. It pretty quickly gets into diminishing returns as the number of exposures goes up.

[Cloengaa]

Question. How do I take darks with my SPC900.

Also why do I need darks and lights ?? how do you take lights ?

[RobH]

Lights are the pictures you're taking.

[Cloengaa]

Ahh ok..but how do I take darks then and why are they necessary ?

[Starflyer]

I understand that for darks to be useful they are as much about having the same temperature as the same exposure time. I started to create a darks library based on exposure length AND temperature. So if you are taking your darks at a different time and, for example, your temperature varies widely from your lights, I'm not sure they'd be as useful as they should.

Doesn't answer your question, Sorry!! But just gives you another headache!

The camera's internal temperature can vary quite a lot and I've found that the temperature of darks taken on different evenings at the same ambient temperature vary significantly.

Kuso Exif viewer will show you the camera's internal temperature. I take a whole bunch of darks at the required exposure and go through them labelling them with the temp, iso and exposure for use at a later date. There is a piece of software called Dark Library, which automates this in conjunction with DSS, but I've yet to get acquainted with it properly.

To answer the original question, I try and use 20 to 30 temperature matched darks. If the darks are warmer than the lights you get spots of dark pixels where the dark has removed something that wasn't there in the lights. If the darks are colder than the lights you'll get thermal noise and hot pixels left in your image.

[Whippy]

Ahh ok..but how do I take darks then and why are they necessary ?

People take darks to take out noise from their camera in the final image. In your case, unless you're doing long exposure work with your webcam, there's not much point.

Tony..