Oh No! Not another M31. Yawn.

[x6gas]

That's a very nice image - congratulations - and a very fine job of processing.

But one thing - and please don't take this the wrong way - but why did you choose that framing when you could have got the whole galaxy on your DSLR chip?  Seems a shame with such a lovely image.

[IanL]

That's a very nice image - congratulations - and a very fine job of processing.

But one thing - and please don't take this the wrong way - but why did you choose that framing when you could have got the whole galaxy on your DSLR chip?  Seems a shame with such a lovely image.

Everything is a screw fit from the focuser tube backwards, so the only way to rotate the camera is to undo the grub screws that hold the back and front halves of the camera adaptor together and then rotate.  This presents a challenge as it is tricky to do even in daylight without risking dropping the camera on the floor, plus it is actually quite hard to ensure the adaptor is back together properly without a small tilt in one direction.

Normally I would plan the framing ahead of time and get the angle right before setting up outside, but as I posted earlier on this was really just meant to be a quick run out to shake down the kit after a long summer break.  In a way it is a bit 'annoying' that it has turned out so well, but I can live with the pain

Also I have a spacing problem as the LP filter is between the camera and reducer, which leads to elongation in the corners of the chip (look at the unprocessed sub and you will see it).  At least in this crop I have hidden my shame.  Depending on finances I am looking to get a CCD at some point, but suspect I may need a focuser upgrade at the same time so waiting to see how that pans out.  If it doesn't look likely I will probably invest in a click-lock and the push fit adaptor which will solve the spacing issue and the rotation issue at once.  The cost of that is a fair chunk of the cost of a replacement focuser, so I am on the fence at the moment.

[Mr Bergman]

Remarkable good capture! I like the colours! 

[IanL]

Okay, episode one is now up and ready - all about Bias Frames:

http://www.blackwaterskies.co.uk/2013/09/pixinsight-dslr-workflow-part-1-bias.html#more

Enjoy.

[Earl]

Its a nice start, I think you need to go back to colour calibration as im seeing a lot of pink / violet stars if you can get that under control better it will make it look even better, might be worth a morphological transformation to control the stars too.

[tdmodels]

That is a stunning image.  Well done!

[Ouroboros]

Okay, episode one is now up and ready - all about Bias Frames:

http://www.blackwaterskies.co.uk/2013/09/pixinsight-dslr-workflow-part-1-bias.html#more

Enjoy.

Just a thought .... I would expect the data in your graph to lie on a smooth curve proportional to 1/SQRT(number of bias frames) since signal to noise is proportional to the square root of the number of samples.

[jetstream]

Awesome image,lots of great info too.

[IanL]

Its a nice start, I think you need to go back to colour calibration as im seeing a lot of pink / violet stars if you can get that under control better it will make it look even better, might be worth a morphological transformation to control the stars too.

I have done an MT, as will be covered later in the blog.  The before and after is very noticeable on the small/mid scale stars.  Shrinking the large stars more than I have is difficult without creating obvious artefacts and my preference is not to shrink all the stars to tiny dots as it makes the overall image look a bit flat.  The mid-scale stars are the real problem.  They do have genuine red fringes on the subs and short of upgrading to a triplet or moving to a mono ccd with RGB filters I can't do much about that.  One of my next learning objectives is to create star masks for different sized stars. This may well allow me to selectively apply the MT or saturation to limit the effect you are seeing.  Any pointers on doing that gratefully received!

Just a thought .... I would expect the data in your graph to lie on a smooth curve proportional to 1/SQRT(number of bias frames) since signal to noise is proportional to the square root of the number of samples.

Thanks for that, I'll add it to the post.

[Earl]

On my OSC data from a qhy8l i also have the red fringes too, humm interesting you have the same

[Ouroboros]

Ouroboros, on 18 Sept 2013 - 12:06 AM, said: Just a thought .... I would expect the data in your graph to lie on a smooth curve proportional to 1/SQRT(number of bias frames) since signal to noise is proportional to the square root of the number of samples.

IanL said: Thanks for that, I'll add it to the post.

I'm wondering if Excel will allow you to fit the curve y = 1 / √n. to your data.

[Russe]

30 min is amazing! Great pic.

Say, which coma corrector are you using?

[perfrej]

That's a nice M31, but I am worried about the number of calibration frames. If I remember correctly, 3-4 darks contributewith about 10 percent noise. Increasing it to 8 will reduce noise contribution to 5 percent, and then with 25 you are down at 1 percent.

The returns keep diminishing with calibration frames, just as they do with the light frames. I have found that going much beyond 50 lights is simply not worth the effort.

I never use darks myself but am in the process of confirming to myself that it is of no or little use. I actually shot 30 10-minute darks at -20 the other night and will apply them to my Elephant as soon as I get a chance.

Anyway, very nice M31!

/per

[jackrussell0232]

Must admit that I'm with Per, in that I don't quite follow why using such a huge number of darks and flats should be advantageous.

Once you have enough to produce a good statistical average (surely no more than 10), what does going to 20 or 30 add?

If you've taken say 5 darks which had 20 pixels always hot, and 4 pixels which were hot in just 2 of the 5 darks, the master dark will scale the 4 intermittent pixels as 40% hot.

If you took 100 darks, statistically you'd say 40 of those 100 would have these 4 extra hot pixels.

What's the difference in the master dark?

If you had enough light frames, which would obviously need to be dithered, then you could probably get a great result with just a few flats. The warm/hot/cold/stuck pixels would be taken out with the dither and sigma clipping.

Each to their own though, you've got a really good result here for sure.

Most of my images have a bit more noise than I'd like, but that's because I don't take enough subs, and also don't like to soften rhings too much with aggressive noise reduction.

Might take a ton of darks, just to see what happens, at least I'm quite confident I can take a good dark frame....

Jack.

Jack

[IanL]

30 min is amazing! Great pic.

Say, which coma corrector are you using?

The standard Skywatcher 0.85x reducer for the 80ED.  It's not working as well as it should since I have the IDAS LP filter between the reducer and the camera adaptor so the spacing is a bit too long.  I need a different adaptor so it can go in front of the reducer but funds not available.

That's a nice M31, but I am worried about the number of calibration frames. If I remember correctly, 3-4 darks contributewith about 10 percent noise. Increasing it to 8 will reduce noise contribution to 5 percent, and then with 25 you are down at 1 percent.

The returns keep diminishing with calibration frames, just as they do with the light frames. I have found that going much beyond 50 lights is simply not worth the effort.

Must admit that I'm with Per, in that I don't quite follow why using such a huge number of darks and flats should be advantageous.

Once you have enough to produce a good statistical average (surely no more than 10), what does going to 20 or 30 add?

If you've taken say 5 darks which had 20 pixels always hot, and 4 pixels which were hot in just 2 of the 5 darks, the master dark will scale the 4 intermittent pixels as 40% hot.

If you took 100 darks, statistically you'd say 40 of those 100 would have these 4 extra hot pixels.

What's the difference in the master dark?

If you had enough light frames, which would obviously need to be dithered, then you could probably get a great result with just a few flats. The warm/hot/cold/stuck pixels would be taken out with the dither and sigma clipping.

Well regarding the light frames, there are two things to consider:  Firstly the total length of the exposure (sum of all the subs).  I don't think it is meaningful to talk about the number of subs in this context, i.e. is 50 subs enough - depends on how long they are of course. 6 x 30 minutes is only 180 minutes and going to 360 or 720 minutes would show a noticeable improvement.  Secondly how many subs in total - yes you want more subs to use better rejection algorithms effectively; with 6 subs I have just had to go for a percentile clip to get rid of the hot pixels, with 60 subs I could use linear fit which gives a better result.  On the other hand more subs = more read noise, and for a DSLR with perhaps a 10 or 11e RMS read noise contribution using more subs to get the same total exposure is a big deal.  That is one reason why I think I am getting a better result with a few 30 minute subs vs. previous images where I have perhaps 40 x 10 minute subs; I only have 180 mins of exposure here vs. 400 plus in other images (taken at lower temps too!) and the image cleans up better.

As far as the benefits of more calibration frames, yes there is a diminishing return the higher you go, but again I do not think the conclusion that 10 or 30 is sufficient is bearing out in practice.  You can plot the SNR improvement (ΔSNR) of stacking one more frame as ΔSNR = √ n - √ (n -1) where n is the number of frames (be that bias, darks, flats or lights).  See below:

So there is a clear and rapid improvement in SNR early on (low tens of frames) and it flattens out later, but actually not that much until you get above 100 frames.  Yes the improvement is probably only going to allow you to stretch the faint details that are a few ADU above the read-out or sky noise (depending on which you are limited by in your sky conditions and exposure length), but actually that is the difference between a fairly poor image and a reasonably average one (not kidding myself that this is 'good' compared to some of the CCD images of the same subject).  It seems to me the main reason why really good CCD images are so much better than DSLR images is that the noise floor is just that few e below the DSLR.  It is also worth noting that darks are mainly about removing another source of fixed pattern noise over and above the bias - from what I've seen so far the DSLR bias frames tend to be dominated by a pattern at the 1-2 pixel scale, whereas the darks have a pattern that is perhaps 5 to 10x that in scale.  As with the bias frames, the pattern does get clearer with more subs, but it is definitely harder to make sufficient darks with an uncooled camera, so I have not gone as far as I would like.

[perfrej]

Well, that is all fine, but going above 30 or 40 for a calibration frame seems simply not worth it for the extremely marginal return. Now, there is another factor involved which may be of interest since you are using a DSLR. Craig Stark (author of Nebulosity and PHD) has written a nice litlle paper on dark/long exposure behaviour in Canon's cameras. It seems that all is not what it appears to be. If you haven't read it, give it a shot! It's good stuff!

http://www.stark-labs.com/craig/resources/Articles-&-Reviews/CanonLinearity.pdf

http://www.stark-labs.com/craig/resources/Articles-&-Reviews/DSLRvsCCD_API.pdf

On another note, here's one of my M33s... 15 flats per channel, 30 BIAS frames and no darks... It's here just to show that stuff can be done even without massice amounts of calibration frames.

All the best, and keep up the good work!

/per

[Russe]

http://www.stark-labs.com/craig/resources/Articles-&-Reviews/CanonLinearity.pdf

http://www.stark-labs.com/craig/resources/Articles-&-Reviews/DSLRvsCCD_API.pdf

]

What a great read!!! So I'll stop using ISO 1600 but rather go for 400 or 800.

[perfrej]

Yes, that is correct. Craig knows what he's talking about and I value hos views and opinions. There are more readable stuff on his site, and I know that there is serious debate about the actual usefulness of darks with DSLRs... Keep diggin'!

/per

[Russe]

Yes, that is correct. Craig knows what he's talking about and I value hos views and opinions. There are more readable stuff on his site, and I know that there is serious debate about the actual usefulness of darks with DSLRs... Keep diggin'!

/per

Yeah, I've been scratching my head just now looking at his graphs on darks with DSLR's. If they are really "tick" shaped, then I don't quite see the point in removing that signal. Would only do bias and flats?!?!?!