[Request] Some help processing Leo Triplet

[jjosefsen]

Hi,

I have been plucking away with PI at this data I have of the Leo Triplet for a few days now, and I just can't get it right. Particularly color is causing me problems..

It seems to come out either very red, or really low on color.

I am hoping someone would give it a whirl and see wha they can get, and maybe point me in the right direction (maybe @wimvb pretty please! :)).

As I use PI it would be preferable if someone with PI would try, but anyone is more than welcome to try.

 

The data was captured with a modified EOS 600D and an Optolong CLS-CCD filter.

Stack consists of 37x240s frames at ISO1600 - calibrated with Bias, Flats and Darks in PI (Total integration: 148 minutes)

https://www.dropbox.com/s/n3nerojediicz7x/integration.tif?dl=0

This is where I am at currently:

 

Is it a matter of an 80mm refractor + DSLR not beign up to the task of capturing something like this?

 

Clear skies!

//Johannes

 

[wimvb]

Ok, I'll have a go. Clouded out anyway.

13 minutes ago, jjosefsen said:

Is it a matter of an 80mm refractor + DSLR not beign up to the task of capturing something like this?

Don't think this is the problem. I'll check the image.

[jjosefsen]

17 minutes ago, wimvb said:

Ok, I'll have a go. Clouded out anyway.

Don't think this is the problem. I'll check the image.

Much appreciated! I forgot to mention that I already cropped out stacking artifacts and some Amp glow.

I didn't do dbe, as I am having problems getting a smooth background to substract, and I'm not sure there even is a gradient.

[wimvb]

The image is very red. PI statistics show the highest values in R, G and B to be:

R: 0.94

G: 0.27

B: 0.23

DBE with 12 samples (one in each corner, one a bit further in from each corner, and four between the galaxies). Main gradient is vignetting (weak), so used division as the background correction method.

Photometric colour calibration with a small preview in the center of the image as background reference. Btw, the platesolving gives a focal length of only 472 mm, which seems a bit short for your scope/reducer combination.

I'm also going to test standard colour calibration, using the galaxies for white reference.

[spillage]

Well its about 10 mins in startools. My monitor is rubbish so may look awful on a better one.

[jjosefsen]

6 minutes ago, wimvb said:

The image is very red. PI statistics show the highest values in R, G and B to be:

R: 0.94

G: 0.27

B: 0.23

DBE with 12 samples (one in each corner, one a bit further in from each corner, and four between the galaxies). Main gradient is vignetting (weak), so used division as the background correction method.

Photometric colour calibration with a small preview in the center of the image as background reference. Btw, the platesolving gives a focal length of only 472 mm, which seems a bit short for your scope/reducer combination.

I'm also going to test standard colour calibration, using the galaxies for white reference.

I assume this color bias is due to a combination of the modified sensor and the CLS filter?

the focal length thing is quite surprising to me, should be around 520mm, looking into this.

do you get a "smooth" background model when doing dbe?

[jjosefsen]

This is the plate solve i get:

[wimvb]

4 minutes ago, jjosefsen said:

I assume this color bias is due to a combination of the modified sensor and the CLS filter?

the focal length thing is quite surprising to me, should be around 520mm, looking into this.

do you get a "smooth" background model when doing dbe?

The colour bias could very well be due to the modding and the filter. The best LP filters I 've seen ar IDAS, since they only block LP light sources, and have a high transmitivity at all other wavelengths. CLS are good for nebulae, but not so good for galaxies.

Some discrepancy between specified and measured fl is to be expected, but 472 mm is excessive. If the reducer had been 0.8x, then the fl would be as expected. Btw, PI's photometric colour calibration accepts the pixel size (3.9 micrometer for the Nikon D7100, as per you signature), and adjusts the focal length to match.

The background is smooth, with only a very small amount of vignetting in the corners.

Since the green and blue channel are so weak, your image is basically underexposed. The problem is that with a DSLR, it's impossible to balance the exposure. If you were to increase exposure time, you'd overexpose the red channel, leading to all kinds of weird colour effects in the stars.

[wimvb]

2 minutes ago, jjosefsen said:

This is the plate solve i get:

Ok. Looks much better. I took the pixel size from a web site:

https://www.digicamdb.com/specs/nikon_d7100/

[jjosefsen]

 

1 minute ago, wimvb said:

Ok. Looks much better. I took the pixel size from a web site:

https://www.digicamdb.com/specs/nikon_d7100/

My bad! I was using my new EOS 600D and not the D7100 from my signature, and it has 4.3um pixels, so thats the discrepancy.

[wimvb]

And signature updated I see. 

[wimvb]

(click on image to see the full size version).

The good news: there definitely is colour. And you also managed to capture a VEERY faint tidal tail (going up from ngc 3628 in this image).

The bad news: the image is underexposed (due to the excessive colour imbalance), leading to a background where the sensor read pattern is starting to show, and with some colour mottle.

Here's the PixInsight process icon (after dbe)

leo_triplet_jjosefsen.xpsm

[Ibbo!]

Had a little go

i put it into PS and adjusted levels  then ran a noise reduction routine and then did a little sharpening and reduced size to 1600 x 1200

 

[jjosefsen]

10 hours ago, wimvb said:

(click on image to see the full size version).

The good news: there definitely is colour. And you also managed to capture a VEERY faint tidal tail (going up from ngc 3628 in this image).

The bad news: the image is underexposed (due to the excessive colour imbalance), leading to a background where the sensor read pattern is starting to show, and with some colour mottle.

Here's the PixInsight process icon (after dbe)

leo_triplet_jjosefsen.xpsm

Thank you so much for taking the time!

your comments about the color unbalance led me to investigate my data a little further, and I have a problem with calibration somewhere. All the red bias is added during calibration.

I will have a look at it tonight, and hopefully I can find what is causing it. Much of that red data is probably noise, so if I can get that cleaned up the final result could be good.

Thanks again!

[wimvb]

Have a look at your flats. I don't exactly know how flats are optimised for each colour, but mathematically, if your flats are weak in red, the calibrated lights could come out strong in red. This is simply because division by a smaller number gives a larger result. In a monocamera this isn't an issue, but in an osc, such as a dslr, this might very well cause a problem.

Just an idea.

[jjosefsen]

8 hours ago, wimvb said:

Have a look at your flats. I don't exactly know how flats are optimised for each colour, but mathematically, if your flats are weak in red, the calibrated lights could come out strong in red. This is simply because division by a smaller number gives a larger result. In a monocamera this isn't an issue, but in an osc, such as a dslr, this might very well cause a problem.

Just an idea.

I think you nailed it.. Here is my masterflat debayered:

Really weak in the red spectrum!

It was shot with APT and the camera in Av mode, using an Aurora flatpanel..

Now I need to figure out how to take them with a more balanced color result.

[wimvb]

Your master flat seems very dim. Even in blue it's only 520 out of 16 k. I would expect it to be at least 10 times brighter.

Btw, your master flat is almost certain 16 bit, or 32 if created in PixInsight. I usually use Image statistics with normalised (0 ... 1) settings. It makes interpretation of pixel values easier.

[jjosefsen]

2 minutes ago, wimvb said:

Your master flat seems very dim. Even in blue it's only 520 out of 16 k. I would expect it to be at least 10 times brighter.

Btw, your master flat is almost certain 16 bit, or 32 if created in PixInsight. I usually use Image statistics with normalised (0 ... 1) settings. It makes interpretation of pixel values easier.

A little more investigation done here.

The source of the color imbalance in the flats is the CLS filter, so it seems the clear answer is - don't use one on galaxies, use it for emission nebulae and not much else.

Maybe the new IDAS-LPS-D2 would be a better all-round filter?

My flats are definately under exposed, I am working on the settings I need now, using this guide: http://www.myastroscience.com/proper-flats-with-dslr

My full well saturation is about 15.000, so aiming for around 7.000

[wimvb]

7 hours ago, jjosefsen said:

Maybe the new IDAS-LPS-D2 would be a better all-round filter?

Very likely. This is the new version of hutec's lp filter. Their previous version (-D1) is more widely used, and generally shows good colour in stars and galaxies. Which one you need depends on your local lp conditions. Comparing filter curves, I would say that the D2 version falls between your current filter and the D1 filter.

[wimvb]

To get better flats, you can try this recipe from @Oddsocks (near the end of the message):

I haven't tried this method myself, but the advice seems sound.

[jjosefsen]

Hmm the synthetic flat approach sounds interesting.. Thanks wimvb