m36

[alacant]

Hi everyone

New OAG  with a -life saver- helical focuser. It takes the misery out of focusing oags. Prism to sensors: 65mm.

Recommended. Also an exercise in using dark frames to rescue an old eos450d. 

Thanks for looking. What about the star colours?

eos450d on nt150l, 90 min, ISO400

 

 

Edited January 30, 2020 by alacant
colour

[celestron8g8]

Nice capture ! Color does seem to lean a bit towards aqua but I’m looking from my iphone 11 so on a big screen perhaps it’s not so much . 

[Marvin Jenkins]

Any further capture info? I note your info 90mins at ISO400. how many subs/ darks anything else. I have plans to try for a cluster for the first time next clear skies.

M

[alacant]

15 minutes ago, Marvin Jenkins said:

capture

3 minute light frames, dark frames to remove banding, bias and flat frames. Take the bias from everything before you start. Dither. Or use a modern DSLR which doesn't produce banding and lose the dark frames.

HTH and good luck. Do please post your image.

[Aramcheck]

With star clusters is it best to use a lower ISO? (I've only used ISO 800 so far)

Ta
Ivor

Edited January 29, 2020 by Aramcheck
Brain not engaged

[alacant]

2 minutes ago, Aramcheck said:

lower ISO

Hi. We don't. Probably best to stick to the ISO which gives the lowest noise. In the case of the 450d, it's 400. 

[celestron8g8]

I have an Xsi 450D and if i image CGs’ i will use 1600 ISO always cause I’m using a C8 , f/10 . I need at least 10-50 good 2 min exposures or at least 5 good exposures at 5 mins . Then maybe 1/2 dozen dark frames . Trick  is the more you stack the less noise you’ll have . 60x2 min exposures and 20 darks will practically illiminate most or all noise . 

Edited January 29, 2020 by celestron8g8

[alacant]

9 hours ago, celestron8g8 said:

1600 ISO

Hi

I think the problem at iso1600 is that you lose some contrast. Your c8 is quite a bit faster than our f8 reflector so I think you'd be OK with 3 minutes and iso400 on this cluster. The main reason for using dark frames with our 450 is not to lower the noise (it increases it!) but to remove the horizontal banding. Agreed the number of frames: the more the merrier:)

Cheers and thanks for your input.

[alacant]

16 hours ago, celestron8g8 said:

Color does seem to lean a bit towards aqua

Hi. I had a go at the colour, but only had the .jpg to play with as I deleted the rest of the evidence to save disk space DUH! I took the M of CMYK up a bit, that's all. It seems to have removed the pastel blue a bit at least. Dunno...

[fwm891]

Your background looks a bit dark - maybe my monitor but to me needs to be taken to a darkish grey rather than black...

[vlaiv]

Nice capture.

To my eye, this image shows typical color balance that one would get from StarTools processing.

If you aim for more realistic color in stars you really need to do proper color calibration of your image. For reference here are few good graphs:

First color range:

Second - frequency of stars by type:

Mind you, this second graph is too saturated - that happens if you don't gamma correct your colors for sRGB standard and yet use image that implies sRGB standard. I think that first scale is more accurate but does not go as deep as this scale (which is to be expected as O type stars are about 0.00003% of stars so above range maybe stopped at B rather than going all the way to O type).

Match those two with your image above and you will see that you are too cyan/aqua, and possibly over saturated in your star colors.

[alacant]

57 minutes ago, vlaiv said:

StarTools

Hi

Thanks @vlaiv.

I just thought that the colour could be rescued. We used Siril, but forgot to pass it through the photometric colour calibration. Now I only have the .jpg, it won't solve. It finds the stars but doesn't apply the colour. It only works on the linear .fits stack.

We also used superpixel debayering. I wonder if this has bearing? I know, too many variables. Time to give up on this one. Must get more disciplined!

Some really good feedback in this therad. Thanks everyone for your for your input.

Edited January 30, 2020 by alacant

[Galen Gilmore]

18 hours ago, alacant said:

3 minute light frames, dark frames to remove banding, bias and flat frames. Take the bias from everything before you start. Dither. Or use a modern DSLR which doesn't produce banding and lose the dark frames.

HTH and good luck. Do please post your image.

Should you dither and take dark frames, or just one or the other?

[alacant]

Just now, Galen Gilmore said:

Should you dither and take dark frames

Depends on the dslr. Which do you have?

[alacant]

3 hours ago, vlaiv said:

do proper color calibration of your image

OK. I found the stack in the trash. Oh and the spin off with superpixel is that you have to half the focal length. You learn something new every...

This is what it gives. Little colour, but I assume accurate:

Edited January 30, 2020 by alacant

[Galen Gilmore]

31 minutes ago, alacant said:

Depends on the dslr. Which do you have?

Modified T3i/600D.

[ollypenrice]

We've discussed your backgrounds before and I maintain they're clipped. The JPEG screen grab certainly is but, whatever the cause, I see Francis agrees with me. I do find them a bit dark. Your rework looks much nicer to me. The sky looks lighter, a better colour and not so shiny.

I found your original blues rather cyan so I used Ps Selective Colour in blues and cyans, to lower the cyan and increase the magenta. I felt it looked better. Your reds look good to me. I used to find them rather yellow. In the rework I think the blue stars should now be a touch more cyan. (I know, I know... Where's that 'Headbang' emoticon gone?) A by-product of the colour calibration is a reduction in blue bloat.

4 hours ago, vlaiv said:

Nice capture.

To my eye, this image shows typical color balance that one would get from StarTools processing.

If you aim for more realistic color in stars you really need to do proper color calibration of your image. For reference here are few good graphs:

First color range:

Second - frequency of stars by type:

Mind you, this second graph is too saturated - that happens if you don't gamma correct your colors for sRGB standard and yet use image that implies sRGB standard. I think that first scale is more accurate but does not go as deep as this scale (which is to be expected as O type stars are about 0.00003% of stars so above range maybe stopped at B rather than going all the way to O type).

Match those two with your image above and you will see that you are too cyan/aqua, and possibly over saturated in your star colors.

This is a good bit of graphics. Thanks. I normally use the B-V index and a colour graph linked to it but this is easier to use.

Olly

 

 

 

[alacant]

10 minutes ago, ollypenrice said:

We've discussed your backgrounds before and I maintain they're clipped.

Hi

Really? Then I don't understand!

[alacant]

45 minutes ago, Galen Gilmore said:

Modified T3i/600D

Ok. That's good news, so I'd go with bias only and dither a few pixels between frames. Lose the dark frames.

HTH

[vlaiv]

15 minutes ago, alacant said:

Hi

Really? Then I don't understand!

I've also noticed that often background looks funny in the images you are producing. I don't think it is necessarily background clipping - in this example it is more to do with distribution of the background that gives it such feel.

You are right to say that your histogram is not clipping - but it is not bell shaped either:

This is green channel - histogram in range 0-32 binned to 33 bins (each number one bin) in first image you posted:

Same thing in second image that you posted - one with color correction and background that looks better:

Although your background is not clipping - histogram of it shows left side to be very steep in first image - it is almost if histogram was indeed clipping but not at 0 value but rather somewhere around 3-4.

Second image shows histogram more resembling to bell shaped curve as you would expect from natural looking background - and this shows in the image - background looks better indeed. At least to my eye, and as far as I can tell @ollypenrice agrees with me:

36 minutes ago, ollypenrice said:

Your rework looks much nicer to me. The sky looks lighter, a better colour and not so shiny.

 

[alacant]

Ok. Thanks. The background looks ok/the same to me, but I'll try to make it more like the rework. Maybe too much or not enough fuzz to smooth it. Dunno...

Give up!

 

Edited January 30, 2020 by alacant

[Xplode]

It looks like it was clipped and then "brought back from the edge"
I've seen people use it as "noise removal", but it can be unintenional in cases too with automatic processes.
 

My own image to show how this works.
Here it's easy to see how this affects the histogram and the background, in my opinion it looks "fake" and lifeless


Color calibration and black level adjustment in astrophotography is pretty hard and for manual adjusting  it depends a lot on the gear too, quality of the screen and how well it's calibrated, the light level in the room where the work is done.
For automatic adjustments it's easier, but can often be problematic too, if calibrating on stars the saturation will matter.

The image in this post seems to suffer from some kind of color separation that i bet would affect an automatic color calibration process(artifact of superpixel debayering?)

 

 

[alacant]

Wow... Apart from moving the slider to stretch the image, I don't think we've any control over the shape of the histogram, have we? Or rather you have, but it's way beyond anything we could do ATM.

Good point about the debayer. I don't understand how it works. The reprocessed image doesn't have it though. It must however have used the same algorithm because it was the same stack we used for the reprocess.

Anyway, thanks again everyone. This thread has helped us move forward a lot and helped us realise that there's no quick fix. At least we're clearer on the star colour now. You're a great team and we really appreciate your posts.

To conclude, I think that with a 12 year old 450d and a bashed around 6" f8, we've reached the limit of what we can achieve. A lot of fun though and we'll certainly keep trying:)

Cheers

Edited January 30, 2020 by alacant

[vlaiv]

5 minutes ago, alacant said:

Wow... Apart from moving the slider to stretch the image, I don't think we've any control over the shape of the histogram, have we? Or rather you have, but it's way beyond anything we could do ATM.

You are using Gimp as far as I can see? It is really easy to get that sort of histogram shape that has been discussed above, here are steps (on generic bell shaped noise image):

Step 1: do levels to do initial linear stretch and get background to be visible (I made random noise image to resemble background of astro image - I added one pixel star to get contrast range of astro image):

Step 2: Now we have nice bell shaped histogram after first linear stretch, in second we do the curves like this:

Most left point will be raised a bit - so our output is limited to the bottom of not going all the way to the black. Same point on "x" axis is starting to "eat" into histogram. Next point is just "pivot" point so we get nice smooth rising curve and next two points are just classical histogram stretch. This configuration of curves is often seen in tutorials and it produces histogram looking like this:

flat on left side and almost bell shaped on opposite side - a bit more "ease out" because we applied gamma type histogram stretch in that section.

Just one round of levels and one round of curves with more or less "recommended" settings and we have produced that effect.

 

 

[alacant]

7 minutes ago, vlaiv said:

You are using Gimp

Hi. Yes but only ever used levels. 

Thanks, will give it a go.