NGC 7000

[darditti]

Here is the North America Nebula taken over 6 nights between 2009 May 24 and June 04 from Edgware using a Celestron 11 SCT fitted with the Starizona Hyperstar imaging system, giving a focal ratio of 2, a broadband light pollution filter (the Orion Skyglow, which I find is very good for a severely light polluted location), and a QHY 8 colour camera. The image was taken in 48x5 minute subs (total 4 hours), unguided, using an Astro-Physics 900GTO mount. 10 flat frames, 10 flat dark frames, and 4 dark frames at two ambient temperatures were used. Processing was with DeepSkyStacker, Photoshop CS4 and GradientXterminator. At this time of year the truly dark part of the night is so short at 51 deg. north (only half an hour either side of 1.00 BST) it was necessary to spread the imaging over this number of nights.

I think this does show what can be done with a fast Schmidt camera-type system even in a very polluted location (limiting mag 4) given time.

David

(click to enlarge)

[Psychobilly]

Plenty of depth to the image David

Peter...

[Tim]

I like that David, an unusual target for an SCT, but as you say, it shows what can be pulled out of even a light polluted sky.

For what it's worth, you'd probably be better if you could stretch to a minimum of 25 flats and offset/dark flats. However just using a few is enough to eliminate dust bunny traces etc, but a far smoother result comes from using more of them. There's a good explanation of why this is so on the DSS website.

Again, very nice image, thanks for posting.

TJ

[ashenlight]

That's a lovely image I feel like I could dive into it, lot of depth as Peter said

Amanda

[barkis]

Nice work. The Hyperstar certainly grabs large chunks of sky, and the NA certainly needs that capability.

Good result David.

Ron.

[darditti]

Thanks all. interesting point from TJ about more flats and dark flats. It's useful to post all the details of one's images, as then you get interesting comments like that. I have been assuming 10 flats is plenty if they are good ones, ie. they fill the histogram. But it would be easy to use more. I looked at the DSS website, and it does claim you should use 50-100 darks, flats and biases for best results. It would take a long time to take 100 5 minute darks...

David

[Psychobilly]

I have started using skyflats i normally grab around 40 in the morning following a nights imaging. I tend to slew the scope around slightly between groups of 5 flats and take a simialr number of bias frames.

I don't use darks anymore with the 1000D. The 350D still needs them though. What's worse is you need to leave the same inter sub period betwwen the darks as you do with the lights... the 350D has largely been "retired" until the winter when it performs much better.

Peter...

[darditti]

I use skyflats as well. I've never found that artificial flats work well. Slewing between them is a good idea. They have to be taken I find a very few minutes before sunrise or after sunset, and the former tends to be easier.

David

[roundycat]

I think, and I caution you all about the word 'think', that taking 100 darks or flats is complete nonsense. The uncertainty in the darks reduces along the lines of a simple arithmetic progression, every time you double the number of darks you reduce the uncertainty by 2. By the time you have taken sixteen darks you have reduced that 'noise' by a factor of five. Trying to push this value below the read noise is pointless as there is no further gain in 'smoothness' to be had. Look at the background standard deviation of the darks combination to verify this.

A similar reasoning applies to flats, basically any more than twenty is not going to improve anything. Providing they are correctly exposed they will be linear and will remove vignettes, dust shadows and fixed pattern signal (often referred to as fixed pattern noise which is a contradiction in terms) and will give you a good, clean result.

Dennis

[daz]

I thought the rule of thumb was to take the same number of darks as lights, but if I read correctly, 16 is the optimum number?

[roundycat]

If there is a rule of thumb it is that you should take enough darks to equal or exceed the light exposure time. If you shoot bias and can do your data reduction using the advanced method (as detailed in AIP4Win) by scaling your darks then you only need to take a set of darks equal to your longest exposure time for the lights. Use that particular master dark for everything provided you have a master bias to use.

Put another way, expose your lights for any time you like just make sure you use a master dark that is longer than your longest light exposure.

16 is a convenient number in terms of the arithmetic but there is no optimum in practice. The uncertainty in your darks needs to be lower than your read noise which may be a significant proportion of the signal in shorter exposures. You therefore need to have a grip on your read noise to really assess how many darks you need.

Dennis

[daz]

Thanks Dennis

[gary1968]

I like that image David, plenty of depth, oodles of stars and the colurs are amazing. Thanks for sharing.

Gary

[Milliways]

Awsome image David.. has a proper 3d effect

[Tim]

Thanks Dennis for the extra info. I dont even pretend to understand the maths involved, or why such and such a figure should be so, I just tend to go along with the accepted norm.

How would one go about empirically determining the optimum number of complementary frames to use? As David mentions, the DSS guy says 50 - 100 of each, but that could be a mammoth waste of time?

[darditti]

This is an interesting discussion. Thinking about it further, I particularly don't understand the justification of the claim that one should take lots of flats. This is because flats, if they fill the histogram, contain a lot of signal, so the s/n ratio in flats is big. Then one divides the light signal by the flat signal. So the difference in result between using a perfect flat and a slightly inaccurate flat is going to be something like the difference between dividing a light frame pixel value by 1000 and dividing it by 1001, ie. it will be a tiny difference, which will not be significant. Wrong?

David

[roundycat]

Tim, I go along with the idea that it is a waste of time. I don't know where Craig Stark gets the idea that 100 darks is a good idea other than from pure theory, which, as often as not, doesn't always sit too happily with practice in the real world. Ten minute subs are quite commom, 10 x 100 = 1000m, that is over 16 hours. How many nights would it take to shoot that lot. At least two or three nights for me this time of the year as I cannot guarantee my ST10 staying at -20 during a summer's day.

What do you gain by having shot 100 darks? I was just working it all out, here are the figures based on the quadratic adding of various numbers of darks. The light frame thermal noise (not hot pixels) is added to by the thermal noise in the dark frame. If you use one dark the answer, basically, is the sq rt of 2 or 1.414. That is an addition of 41.4% extra noise from the one dark. 3 darks add 25%, 5 darks add less than 10%, 10 darks add less than 5%, 16 darks add 3%, 20 darks add 2.5%, 40 darks add 1.2% and 100 darks add 0.5%.

For the record, Ron Wodaski in his first book suggested a median combine of three darks, AIP4Win suggest 10 darks and, for what it's worth, I use 20 usually as a median combine. I have used SD Mask on my darks but do not see any advantage.

As for your empirical determination there are two ways that I can think of. One is to determine your camera's read noise and stay at or below that level with your additional dark noise. The other is to suck it and see. Just make sure that you compare apples with apples. Take a light frame, make a half dozen copies and then subtract a different master dark from each one of the copies. Don't mix them up! Then measure the background standard deviation in each frame centred at the exact same pixel.

David, the argument for lots of flats is much the same as for darks. If your master flat is noisy it will add that noise to your lights. The uncertainty in each flat is equal to the sq rt of the signal and it is that uncertainty that you need to drive down by combining several flats. One flat with an average of 20,000 adu will have a s/n of 141. Multiply that by ten and you have sq rt 200,000 which is 447. Much better. I think the last bit of your comment is irrelevant, it is just about noise, not the figure you illustrated which is just the flat ratio.

Dennis

[darditti]

Thanks Dennis, interesting analysis vis-a-vis darks. Seems to me one is unlikely to notice the difference going above 5-10.

But I still don't agree with respect to flats. Surely it is a fundamental difference that the dark is subtracted, therefore the noise in that goes linearly into the final picture, but the flat is divided by, so doesn't - can't add appreciable noise, even in the case of using only one flat with a s/n of 141.

David

[roundycat]

Flats have noise. For the same reason that every exposure you make has noise, measurement uncertainty. The best way to rid yourself of it is to take many flats (or samples in statistics speak) and combine them. If you don't that noise will translate itself into the light frames.

Before you can get to the light/flat division you must first get your flat ratio. Divide the average flat value by the pixel values in the flat and you will straight away suffer from the noise in the flat pixel values. For this an average of several flats will improve the situation. You then end up with a flat ratio that is minimally affected by the noise in any one flat.

All calibration frames improve with quantity. The improvement is roughly exponential so deciding where to stop becomes an issue. I usually shoot 10-20 flats only because the signal is well above the noise, or should be. I have seen noisier than expected results come from one flat but it is a simple thing to put to the test.

Dennis

[ollypenrice]

Hello David,

I like the picture and am bemused by the discussion! (Arts background showing here...)

Olly

[SteveL]

Might not be worth taking darks with the QHY8, its a very noise free (sony) chip and cooled very well. I have never had to take darks with my QHY8 (or my M25C), even with 10min subs.

[Deneb]

Wonderful Shot !

[Lester]

Fantastic image.