How many dark frames?

[Moonshed]

I understand that dark frames should be taken using the same settings as the light frames and shortly after. What I haven't yet come across is how many dark frames to take. Is it a percentage of the lights, say 1 dark for every so many lights? Or is it good enough to take a fixed number, say 5 every time and that would be good enough? Also what would be a sensible exposure time for the light subs, for say the Orion Nebula? I guess the more subs the better?

Any advice would be appreciated.

Keith

[Marky1973]

Hi Keith, what camera are you using? Some suggest that they are not necessarily useful for DSLR imaging as they can introduce more noise due to the inherent temperature issues. Others have suggested that with a decent CCD chip you don't need to take them!

If you are taking them, get as many as you can! You can always do them before you take the light frames - I have done that before. Set everything up and then set some draks up while I wait for the night to get truly dark - just make sure the sensor is in complete darkness. I don't think "when" you take them is key - the temperature is the important thing, that shoudl be the same for the DARKS and for the LIGHTS - which is why it can be tricky with an uncolled sensor. If you have a cooled CCD camera, you can set the cooling and then take a library of darkas of various lengths, to use in the future.

Do you mean exposure time for the LIGHT frames or total exposure? For total exposure then the more the merrier, but there is always a point of diminishing returns I believe. The length of individual exposures will vary for the target you chose. ORION for example, might only need a couple of minutes with an astro-modded camera to get some data as the core is quite bright. But you might want longer to get the finer detail in the outer regions. Its getting the balance right to get the detail but not blow out the core - hence the reason why many people take two sets of exposures to capture the outer regions and the core separately and then blend in Photoshop... Orion is a great target, but I think it is trickier that it looks to get it right....

But then I'm still a novice, so learning as I go.

Best thing to do is have a play and try a different range of exposure to see what you come up with - but keep an eye on the Histogram - the general advice there is to expose to about 1/3 from the left. You can bring that up in the DSLR or in camera software like APT if you use it?

Good luck!

[Moonshed]

Thanks for all the info. My camera is a Canon EOS 1100D. The exposure time I was referring to was for each light frame sub. I was reading somewhere that a 4 min. exposure with an ISO of 800 would be a good starting point. As you say, the best thing really is to get stuck in and play around with it. Thanks. 

Keith

[Marky1973]

Hi Keith

This is one of my efforts at Orion - which was a combination of 90 second and 20 second exposures using an un-modded 1100d. There is a lot more data to be found, but the core was starting to blow out at 90 seconds, so 4 minutes will make it very bright. Of course, you will then be getting more of the finer detail I missed.

https://markwalkerscreenwriting.wordpress.com/astro-images-2/orion-m42-24th-january-2015/

As you say, have a play! :-)

[Moonshed]

5 minutes ago, Marky1973 said:

Hi Keith

This is one of my efforts at Orion - which was a combination of 90 second and 20 second exposures using an un-modded 1100d. There is a lot more data to be found, but the core was starting to blow out at 90 seconds, so 4 minutes will make it very bright. Of course, you will then be getting more of the finer detail I missed.

https://markwalkerscreenwriting.wordpress.com/astro-images-2/orion-m42-24th-january-2015/

As you say, have a play! :-)

I am amazed that you got so much detail from such short exposures, I must make a note of that. Combining exposures of different lengths is clearly the way to go to get both the bright core and finer detail. What ISO were you using?

[Marky1973]

Not half as amazed as I was! :-) I may have had a light pollution filter in the mix somewhere, I forget now, and it most likely would have been on a night when the moon was more scarce, but you will be amazed at what appears after you start stacking. I used 800ISO for that one I can't remember exactly without going back through a load of files, but suspect that was also at least 1 to 1.5 hours of LIGHT frames - with them being fairly short, I probably took about 15-20 darks and maybe 30 FLATS/BIAS. I'm not very scientific when it comes to the other frames, and just take as many as I can, when I can!.

[billyharris72]

I don't really see the benefit taking loads of darks, since you are not getting any extra information, simply trying to even out the background noise. With median stacking that should happen very quickly, and it's hard to see much benefit in going beyond, say 9 darks, maybe 15. With median stacking you may as well shoot an odd number of frames, as at least you'll get an actual pixel value returned (though in reality it makes very little, perhaps no, difference).

Sort of the same for flats, though I stend to stack more because:

1)  Flats are quicker anyway; and

2) I move the camera slightly each time. With more images and median stacking, this should remove any minor blemishes or motes in the screen I am shooting the flats from. Even if there are quite a few of them, provided they make up only a timy proportion of the area they should be completely removed.

[ollypenrice]

Beware of darks with uncooled DSLRs. They will not be at the right temperature because the camera temp is not controlled. The best DSLR users (and be aware that I'm not one of them) use 'bias as dark' and a large scale (12 pixel) dither between subs. I'm a set-point cooled CCD imager and I don't use darks either, despite the fact that they ought to work. Since I switched to 'bias as dark' and a bad pixel map, and aggressive hot pixel filtration, my stacks have been so much cleaner that it beggars belief. I used to spend 2 hours cleaning hots on a single panel final image. I now spend 10 minutes needlessly checking 10 panel mosaics!

Olly

[spillage]

Hi Moonshed,

I too have the 1100d and normally take about 10 darks at the end of my session. I am very new to this so have allot to learn. I have included a photo I took last night with the heavy light pollution from the moon. Please don't complain too much about the detail. I am posting it in another post to get some advice.

 

The canon is not modded and I normally use iso 1600 but this was on iso 800 to keep to pollution to a minimum.

 

 

[RichLD]

Hi Keith

If you haven't already tried it, I'd highly recommend the Cosmetic Correction process in PixInsight. I haven't bothered with darks since discovering this - it's quick, easy and very effective. A fully featured trial licence can be requested here. 

HTH

 

Rich 

[Stub Mandrel]

Don't forget that darks will also get rid of hot pixels - I have about 40 or 50 of these on my 450D that would otherwise show up as extra stars.

I have done an experiment on ISO levels and it shows advice here appears to be right while some respected websites are wrong. The experiment shows ISO800 is much less noisy than ISO1600 for my camera on long exposures although theory (based on short-exposure noise measurements) suggests the opposite should be true.

Same I did a simple experiment, processing the same image twice once with 60 darks (probably over the top) and once with no darks. I know which is which and this might bias my feeling about which is noisier. Which one do YOU think is the noisier image - I will then reveal which has no darks.

 

[Ouroboros]

The second one looks noisier to me. 

[spillage]

I really do not know why but I think the second image has more to offer and could end up the better photo.....but that is why I am here asking the questions no answering them :)..

 

Cheers

Spill.

[alacant]

Hi. I would not recommend darks with a canon. Its difficult if not impossible to match the light and dark temperatures. They introduce more noise than they remove. The only way to remove hot pixels I've found is to use bias frames and around a 10 pixel dither. When you see what dither does for the noise, I doubt you'll go back to darks. HTH

[wimvb]

For my setup, darks do not improve the image quality enough for me to be bothered taking them.

Darks are supposed to clean up hot pixels and dark current signal (amp glow), but for me they don't work. I agree with Olly, alacant and Rich; bias as dark, dithering (I aim at 15 pixels), cosmetic correction during preprocessing, and abundant pixel rejection.

But I'm also in favour of Neils method: test and see what works best for your situation. (That's how I came to my workflow.)

But if you use them, I would suggest taking the same number as light frames. You want the darks to cancel bad signal, not add noise.

You can always do the extreme test: take many light frames and process them with just one dark frame, and see what you get.

BTW, taking darks with the camera in the fridge/freezer didn't work for me either, I rarely do my imaging at +5 or -18 C

Just my 0.02 €

[Stub Mandrel]

Second one is without darks, but also stretched a bit more. If you look at the bottom right (image spun 1980 degrees) you can see traces of amp glow. Here's a single dark at ISO800, look at the amp glow (this is stretched far more than I would stretch any astro pic).

[wimvb]

10 minutes ago, Stub Mandrel said:

Second one is without darks, but also stretched a bit more. If you look at the bottom right (image spun 1980 degrees) you can see traces of amp glow. Here's a single dark at ISO800, look at the amp glow (this is stretched far more than I would stretch any astro pic).

 

That's 5.5 turns  

[Stub Mandrel]

1 minute ago, wimvb said:

That's 5.5 turns  

So am I wrong? ;-)

[wimvb]

No, just making me dizzy (sorry, just couldn't resist)

Seriously though, it's hard to tell when one image is stretched more. But I agree that dark frames can correct amp glow, just not enough in my case. And for darks to remove hot pixels efficiently, the temperature has to be a close match.

To make a fair comparison, you would have to take two single calibrated frames, one with dark subtraction, and the other with bias subtraction (and possibly cosmetic correction, since that's the alternative). Personally, I have never done that. I took darks out of the equation, when I discovered that I had bad pixels left in my stacked images, and the amp glow wasn't removed properly. Cosmetic correction and careful pixel rejection during preprocessing did a better job.

As I've said before, it is always best to test what works for a certain setup and imaging conditions, and then make the decision if taking darks is worth the extra effort.

 

[RSM]

On 15 August 2016 at 16:21, Moonshed said:

Thanks for all the info. My camera is a Canon EOS 1100D. The exposure time I was referring to was for each light frame sub. I was reading somewhere that a 4 min. exposure with an ISO of 800 would be a good starting point. As you say, the best thing really is to get stuck in and play around with it. Thanks. 

Keith

Hi Keith. Like you mention, try different settings on ISO and exposure time. I've settled on ISO 800 and depending on the subject and conditions I vary exposure time between 3 and 6 minutes for my image run. I use the histogram feature on my 500D to choose an exposure time which gives me the histogram between 30-50% of maximum. I always take darks and typically 10-20. Generally take these at the start of the image run whilst waiting for it to get dark, or mid run if the clouds roll in. Dithering really helps reduce noise - I've not tried processing without darks (might try based on some of the comments on this thread) yet. Interested also to look into Cosmetic Correction in Pixinsight - not used that yet. Good luck. Richard. 

[Moonshed]

Thanks Richard. I note your comment about using Pixinsight, lots of people here recommend it. I checked out their web site and was a bit taken aback by the price, €230 + VAT. not sure but think the VAT would be 20%? That would bring the price up to €276 which is outside my budget. I am disappointed because the programme looks so good, looks brilliant in fact. Oh well, I shall just have to make do without it as I am sure many others do, after all it is a lot of money.

[MattJenko]

If you decide to commit to astrophotography long term, I am sure you will come to see PixInsight as a bargain in an expensive hobby. It has so many well thought out processes that are just not available anywhere else and for the price of a good planetary camera or less than a premium eyepiece you get a premium piece of well supported software which is regularly updated and runs on multiple platforms for a one off fee. I am a complete fan!

[Marky1973]

Yeah, it's not cheap, but well regarded. You can get a free trial for 45 days if you want to have a look, and there are plenty of tutorials out there. BUt people also manage perfectly well with Photoshop, Paintshop Pro and the free GIMP. Chek out StarTools as well - a much cheaper alternative that people get some great results with as well.

[Marky1973]

1 minute ago, MattJenko said:

If you decide to commit to astrophotography long term, I am sure you will come to see PixInsight as a bargain in an expensive hobby. It has so many well thought out processes that are just not available anywhere else and for the price of a good planetary camera or less than a premium eyepiece you get a premium piece of well supported software which is regularly updated and runs on multiple platforms for a one off fee. I am a complete fan!

Yeah, just to reiterate - I have only had the free trial to date, but found it excellent for me to work in as i found the use of Levels and Curves (to any great effect) in Photoshop trickier to get my head around. I am very tempted to buy it - as Matt says, when you have spent thousands building up an imaging rig, it is peanuts in comparison.

[alacant]

Hi. Maybe worth a mention. If you don't mind using dss for stacking, I believe StarTools does the same. If you've already learned Pixinsight via the free trial, there's an easy to follow switchover guide: http://forum.startools.org/viewtopic.php?f=7&t=447 HTH