How to process UHC frames

[SteveBz]

Hi Guys,

Some help needed, my frames look green and monochromatic.

I just bought a new SvBony UHC filter from Aliexpress for about 30 quid.  While the reviews have been mixed, in the end I decided it was good enough to experiment with.  The response is not so dissimilar to the Optolong UHC.

After fitting it, I realised that my scope wasn't well balanced, so I rebalanced it leading to my best guiding ever.  An unexpected benefit.  So I left it running for 3 hours to get 36x5 mins exposures at ISO400 on a Nikon D5000 DSLR and I got some really quite good data.

I don't yet have new calibration frames to match so it's still uncalibrated, never-the-less I stacked them and pumped the result into GIMP for stretching.  Really I was unable to find any colour at all.  I'm not sure what to do next.

Here are the before and after stretching (uncalibrated, remember):

What to do next?

Regards,

Steve.

[MarkAR]

Calibration frames will help a lot.

Post processing would be a crop to remove border artefacts, noise reduction, gradient removal etc.

[SteveBz]

 

19 hours ago, MarkAR said:

Calibration frames will help a lot.

Post processing would be a crop to remove border artefacts, noise reduction, gradient removal etc.

Well, I need a full set, biases because I changed the ISO, darks because I extended from 4 mins to 5 mins to compensate for the filter and new flats because my flats were never very good.  I'll update here as soon as I can.

S.

Edited May 10, 2020 by SteveBz

[almcl]

On 09/05/2020 at 13:17, SteveBz said:

Hi Guys,

Some help needed, my frames look green and monochromatic.

I just bought a new SvBony UHC filter from Aliexpress for about 30 quid. 

Isn't green and monochromatic what you's expect from a UHC filter?

Not sure about  the filter you mention, but the transmission curve from one I bought a while ago for around the same price looks like this:

and certainly when I used it, a dark green tint covered everything.

Do you have a transmission curve for your filter?  It might help explain the result.

 

[SteveBz]

29 minutes ago, almcl said:

Isn't green and monochromatic what you's expect from a UHC filter?

Not sure about  the filter you mention, but the transmission curve from one I bought a while ago for around the same price looks like this:

and certainly when I used it, a dark green tint covered everything.

Do you have a transmission curve for your filter?  It might help explain the result.

 

Hi,

This is whats on their website, and it's what I thought it was before I bought it.

None-the-less, how does that produce a colour image?  I'm no longer sure what post-processing I should carry out to get an RGB image.  It looks green, but it does have a bit of blue and red in it, which I thought to amplify. I'm just having difficulty doing it.

Regards,

Steve.

[SteveBz]

It's looking a little better.  Still needs darks to get rid of those dark currents.  But little colour variation.

Steve.

[almcl]

Well that explains the histogram I got from your original (below): you've got red and green/blue. 

Not entirely dissimilar to Ha and Oiii so perhaps processing it as a bi-colour, as you would a two channel narrow band image might work?  Separate the two channels, process each separately and then recombine.

[SteveBz]

9 hours ago, almcl said:

Well that explains the histogram I got from your original (below): you've got red and green/blue. 

Hi ALMCL,

How did you get that histogram?  When I run it, either on DSS or with GIMP, I get a small red peak and then pretty much equal blue and green. 

9 hours ago, almcl said:

Separate the two channels, process each separately and then recombine.

This is my first foray into filters, so I have little idea how to split and recombine.  I've tried two mechanism, one was to split into .fits, stretch each colour, stretch each colour and then recombine.  It was quite painful and I don't really see how to automate it.  I'm on Linux, so I used SIRIL (which is IRIS for Linux) to convert the RAW files to FITS, register, calibrate and stack, then I opened the resulting FITS colours one at a time in FITS Liberator to give it an arcsinh stretch while saving as TIF then I opened each TIF as layer in gimp, coloured each one and merged the layers.

The other method was not to use FITS, instead I stacked in DSS output a colour TIF, opened in GIMP and stretched each colour in levels to amplify the red.  There was little difference except that DSS is a littler more user friendly.

Last night was cold and cloudy, so I took the time to get three hours of darks.  I'll process today and we'll see what it does.

Tx

Steve

 

[almcl]

Morning Steve

I got the histogram from Corel Photopaint.  Unfortunately, I don't think anybody else on SGL uses this, so probably not a lot of help 😞

Also am only a Windoze user, so not much useful advice to offer about Linux based processing.  I did try FITS Liberator but found it very difficult to get reliable stretches (and the interface was extremely buggy) so ended up as a StarTools user, but one suggestion from the StarTools forum which might be worth a look, is to use your filtered data (from the UHC stack) to create a Luminance dataset and then to take some unfiltered images and uses these as RGB data after a suitable wipe to remove any light pollution.  Process the Luminance stack for max detail and then combine it with the RGB to restore the colour.  I barely know how to do this in ST and have never used GIMP but guess it must be possible?

I think stretching the filtered data alone in whatever program may not yield 'natural' colours as so much of the spectrum is missing - my 'City Lights Suppression' filter (Astronomik CLS CCD clip filter) while great at getting rid of sodium lamp skyglow does also mean that I never get any yellow stars!

[SteveBz]

1 hour ago, almcl said:

Morning Steve

I got the histogram from Corel Photopaint.  Unfortunately, I don't think anybody else on SGL uses this, so probably not a lot of help 😞

Also am only a Windoze user, so not much useful advice to offer about Linux based processing.  I did try FITS Liberator but found it very difficult to get reliable stretches (and the interface was extremely buggy) so ended up as a StarTools user, but one suggestion from the StarTools forum which might be worth a look, is to use your filtered data (from the UHC stack) to create a Luminance dataset and then to take some unfiltered images and uses these as RGB data after a suitable wipe to remove any light pollution.  Process the Luminance stack for max detail and then combine it with the RGB to restore the colour.  I barely know how to do this in ST and have never used GIMP but guess it must be possible?

I think stretching the filtered data alone in whatever program may not yield 'natural' colours as so much of the spectrum is missing - my 'City Lights Suppression' filter (Astronomik CLS CCD clip filter) while great at getting rid of sodium lamp skyglow does also mean that I never get any yellow stars!

Do I take it from your signon, that your name is Al?

Anyhow, I've re-run the DSS route, and stretched in GIMP.  Still not much in variation of colour, it's all a bit white.  It also doesn't look as if any calibration has been done.  When I stretch, I can still see vignetting (no flats) and a strong texture to the background noise that implies the new flats (taken last night) haven't calibrated.  See below.  When I have a sec, I'm going to experiment with the SIRIL route and see if it calibrates any better.  The final nag, is that if you look carefully above each star, there is a slight shadow, like wobble, except that I have carefully removed any such wobbles from the input.  Almost as if some of the frames have not been stacked properly.

Regards

Steve

[almcl]

9 minutes ago, SteveBz said:

Do I take it from your signon, that your name is Al?

yes, sorry - my sig has lost a bit somewhere - it's Al.

I see what you mean about calibration, and now I've looked a bit closer, I think there may be some walking noise (the yellow rings around some artefacts):

I can see the wobbly stars, are you sure you got all the rogue subs out of the stacking list (my favourite is missing one satellite trail...) if you are sure, then try stacking with a Sigma Clip routine and see if that helps?

 

[SteveBz]

5 minutes ago, almcl said:

I think there may be some walking noise (the yellow rings around some artefacts):

Hi Al,

The walking noise seems to come from a dead pixel moving slightly in each frame.  Since my guiding is the best I've ever had it, I totally don't understand why it moves, but it does.  I'm also going to try to check my darks.  The interesting thing is that if I apply my flats AGAIN in gimp, it improves the image a lot.  That's why I think that maybe none of my calibration files have been applied.

Since there are only 35 frames, I could potentially, manually stack and calibrate my image.  I might try that too and see what it reveals.

So, quite a bit of work here.  I'll get back when I can.

Thanks very much for your interest, I really appreciate it.

Regards

Steve.

[almcl]

Good luck with those, Steve - quite a bit of work.  The moving artefact may be caused by mirror or guidescope movement ( a few microns is all it takes), but dithering and a Sigma clip stacking routine may help.

[SteveBz]

34 minutes ago, almcl said:

mirror or guidescope movement

Ohhh kay... so is that what is called flexure? I see from your signature that you use an OAG with a ZWO ASI120, I use the QHY5, but I think they're broadly similar apart from pixel size (yours is finer).  Do you have difficulty finding guide-stars? I have been thinking about getting a low profile OAG for my Newtonian f5 (is yours similar?), but it seems like a vanity rather than a necessity - and I particularly don't want to buy a new cam at this stage.

I can try the dithering quite easily, how big a dither should I make (I think from memory it's .5x to 5x)?

I can probably also try the Sigma clip, again, let me ask, what parameters do I need?

Thanks again,

Steve.

[SteveBz]

1 hour ago, almcl said:

Good luck with those, Steve - quite a bit of work.  The moving artefact may be caused by mirror or guidescope movement ( a few microns is all it takes), but dithering and a Sigma clip stacking routine may help.

So, here's step 1) Retry with the FITS route.  Quite nice, but still a lot of field curvature.  Secondly it's clear that the dark artefacts are largely in the red.  Interesting. Where the signal was the smallest and therefore the ratio of the red signal to the dark signal would have been most equal.

S

[almcl]

6 hours ago, SteveBz said:

Ohhh kay... so is that what is called flexure? I see from your signature that you use an OAG with a ZWO ASI120, I use the QHY5, but I think they're broadly similar apart from pixel size (yours is finer).  Do you have difficulty finding guide-stars? I have been thinking about getting a low profile OAG for my Newtonian f5 (is yours similar?), but it seems like a vanity rather than a necessity - and I particularly don't want to buy a new cam at this stage.

Yes, flexure is the term.  My rig suffered from it quite badly (had a thread about it somewhere but can't find it at the moment). I switched to a Canon specific OAG and the improvement in star shape was dramatic.  Originally I used a QHY 5Lii m but the USB socket on the back wasn't strong enough for the heavy and inflexible QHY supplied cable and it became unreliable and eventually broke (there's a thread about that somewhere, too).  I then switched to the ASI 120 mini and found it to be very similar to the QHY.  It has never failed to find a guide star (YMMV) in the two years I have been using it. If your QHY is OK, I think it would work well with an OAG.  

Yes, my Newtonian is (or was) an F5.  I have since mucked it about a bit (shortened the focus tube to stop it intruding in the light path, moved the mirror cell when a Baader click lock was added to better centre the Coma Corrector, baffled the main mirror to remove the mirror clips from view, added a bottom fan to stop the main mirror dewing up, replaced the single speed focus knob with a Lacerta 10:1 geared version &c &c).

 

6 hours ago, SteveBz said:

I can try the dithering quite easily, how big a dither should I make (I think from memory it's .5x to 5x)?

Opinions vary a bit on how big a dither to use, some say big (>20 pixels) others say 5 or 6.  Getting your capture program and PHD2 to co-operate is the trick. I am currently using between 5 and 12, but again YMMV.

 

6 hours ago, SteveBz said:

I can probably also try the Sigma clip, again, let me ask, what parameters do I need?

I think I left DSS at the defaults (Kappa 2, iterations 5) but Vlaiv had a thread in which he suggested other values (with convincing mathematical arguments) but they didn't work as well for me so I reverted.

By the way, your latest stack is beginning to look quite nice - hours of processing fun to had during the light nights of the summer :-)

[SteveBz]

3 hours ago, almcl said:

Yes, flexure is the term.  My rig suffered from it quite badly (had a thread about it somewhere but can't find it at the moment). I switched to a Canon specific OAG and the improvement in star shape was dramatic.  Originally I used a QHY 5Lii m but the USB socket on the back wasn't strong enough for the heavy and inflexible QHY supplied cable and it became unreliable and eventually broke (there's a thread about that somewhere, too).  I then switched to the ASI 120 mini and found it to be very similar to the QHY.  It has never failed to find a guide star (YMMV) in the two years I have been using it. If your QHY is OK, I think it would work well with an OAG.  

Yes, my Newtonian is (or was) an F5.  I have since mucked it about a bit (shortened the focus tube to stop it intruding in the light path, moved the mirror cell when a Baader click lock was added to better centre the Coma Corrector, baffled the main mirror to remove the mirror clips from view, added a bottom fan to stop the main mirror dewing up, replaced the single speed focus knob with a Lacerta 10:1 geared version &c &c).

 

Opinions vary a bit on how big a dither to use, some say big (>20 pixels) others say 5 or 6.  Getting your capture program and PHD2 to co-operate is the trick. I am currently using between 5 and 12, but again YMMV.

 

I think I left DSS at the defaults (Kappa 2, iterations 5) but Vlaiv had a thread in which he suggested other values (with convincing mathematical arguments) but they didn't work as well for me so I reverted.

By the way, your latest stack is beginning to look quite nice - hours of processing fun to had during the light nights of the summer :-)

So here's the completely fits route with SIRIL and Fits-Liberator and GIMP.

The vignetting seems to be gone, but the flat is not perfect.  There still seems to be some texture to the background, from which I guess the darks weren't perfect. However there is still no colour!!

What to do?

Regards

Steve.

[almcl]

Forgive me if you already know this, but FITS Lib treats each channel separately and you have to stretch and then save them separately and then combine them in other software. Confusingly it refers to them as plane1, plane2 &c.

Your image looks as if you may only have one of the channels?

Edited May 11, 2020 by almcl

[SteveBz]

1 hour ago, almcl said:

Forgive me if you already know this, but FITS Lib treats each channel separately and you have to stretch and then save them separately and then combine them in other software. Confusingly it refers to them as plane1, plane2 &c.

Your image looks as if you may only have one of the channels?

You're right, it does look exactly like that.  On the way it's been all over the place.  When I was processing this morning there was some confusion over channels, but this does, I think, have three separate channels, stretched and saved separately.  I spent most of today converting RAW to FITS, calibrating and recalibrating, stacking and restacking and editing.

I'll check through, but I'm pretty sure it's so, because the red green and blue have calibrated differently, ie they look different after calibration.

So I've just set off a 2-hour run for some more data and I'll check again tomorrow.

Thanks for your help,

Regards,

Steve.

[SteveBz]

12 hours ago, almcl said:

Forgive me if you already know this, but FITS Lib treats each channel separately and you have to stretch and then save them separately and then combine them in other software. Confusingly it refers to them as plane1, plane2 &c.

Your image looks as if you may only have one of the channels?

Do I need to 'Demosaic'?  I'm not really sure what that does, but maybe I need to do it.  I'm also not sure which of the three images in my fit is the R or the G or the B, is it just 1 2 3 = r g b?

Regards

Steve.

[SteveBz]

Here's the most recent one. Same flats, but applied through Siril not DSS, then demosaiced stacked, stretched in Fits-liberator and combined in GIMP.  Still nearly no colour!

S

[almcl]

6 hours ago, SteveBz said:

Do I need to 'Demosaic'?  I'm not really sure what that does, but maybe I need to do it.  I'm also not sure which of the three images in my fit is the R or the G or the B, is it just 1 2 3 = r g b?

The only stacking program I use (DSS) doesn't use the term demosaic, so not absolutely sure about this, but if it's synonymous with debayering, and you are using a DSLR then, yes it has to be done. DSS will do this automatically with RAW files but the correct Bayer pattern needs to be selected (for the ones I've encountered it's RGGB).

The FITS planes in Fits Liberator are, as you say, Red, Green and Blue. 

Your latest image has got colour, although it's quite hard to see until it's been stretched (image below is a crop of your central area which I stretched and colour balanced in StarTools).  As expected from the UHC filter, you've got blue/green and red, but of course no yellow.  Nice round stars and several of the smaller faint fuzzies, too.  

 

Edited May 12, 2020 by almcl

[SteveBz]

Yes, I looked it up, demsaic and debayer appear to be synonymous.  But what you have done is amazing.  I think I don't understand 'stretch' yet.  I'll look it up some more.  I might download a version of StarTools, it appears that you can get an evaluation version.

Tx.

Steve.

[almcl]

29 minutes ago, SteveBz said:

 I might download a version of StarTools, it appears that you can get an evaluation version.

If you decide to try StarTools, the Release Candidate versions (RC) are stable and have all the latest bells and whistles.  You can do everything a paid-for version does, except save (but you can take screen grabs).  There is a bit of a learning curve but several SGLers use it rather successfully.

[SteveBz]

Hi Al,

Do you have a view about colour spaces? When I convert from NEF to TIF I get a choice of raw, sRGB, Adobe, Wide, ProPhoto, XYZ, ACES and I choose raw.  Is this a good choice, do you think?

Regards

Steve.