www.ukastroimaging.co.uk

Images Section => Deep Sky Images => Topic started by: Ian Straton on 16:24:50, 10 February, 2018

Title: Horsehead gone weird
Post by: Ian Straton on 16:24:50, 10 February, 2018
Anyone know what has gone wrong here?

This is a stack of 20x5 minute exposures at iso 400.  The processing has been done in pixinsight.  I ran the sub frames through the blink comparator to get rid of gross aberrations,  then through the subframe selector script to check for good quality frames and basically selected the 20 best frames in terms of star shape eccentricity and fwhm having done that I ran the remaining frames through star alignment and image integration and this is the result. 

The stars are these odd bow tie shapes... now each sub frame is pretty good in terms of tracking but there is overall quite a bit of movement between the first and last frames which if you traced it out would be pretty similar to this shape but I would have expected the star alignment to have removed this?
Title: Re: Horsehead gone weird
Post by: chris.bailey on 16:34:50, 10 February, 2018
Ian

Did you do any calibration (Darks/Flats/Bias)? Whats settings did you have for Interpolation in Star Alignment and what rejection method did you use in Image Integration and what Sigma settings?

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 17:14:11, 10 February, 2018
No calibration just the light files. Interpolation was left at the default settings of Auto / 0.30 clamping threshold.  Rejection method was Winsorized Sigma Clipping sigma low was 4.0 high 3.0 (again default settings)
Title: Re: Horsehead gone weird
Post by: chris.bailey on 17:38:50, 10 February, 2018
Mmm very strange then Ian. Not sure what has happened I'm afraid.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 17:59:38, 10 February, 2018
And here is m101 basically done through the same process...

I'd swear that the stacking hasn't translated the positions of sub frames...
Title: Re: Horsehead gone weird
Post by: PaulB on 21:54:47, 10 February, 2018
Try re-setting Registax back to default. Try again.

BTW. You don't have comet stack, ticked to you?
Title: Re: Horsehead gone weird
Post by: Ian Straton on 22:12:09, 10 February, 2018
Paul, this was stacked in Pixinsight not registax... But the comet alignment was a good suggestion... unfortunately that is a separate function in pixinsight so no cigar on this occasion :(
Title: Re: Horsehead gone weird
Post by: PaulB on 22:29:53, 10 February, 2018
Sorry. I ment to say Deep Sky Stacker.
Title: Re: Horsehead gone weird
Post by: chris.bailey on 08:41:39, 11 February, 2018
Ian

That M101 is more illuminating. During Star Alignment are you getting any Ransac errors popping up in the console window? If so try try increasing the RANSAC tolerance to 3. If not try reducing Log (sensitivity) in the Star Detection to -2.00. It can also help to use the frame with the highest S/N ratio reported in SubFrame Selector as the reference image.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 13:43:42, 11 February, 2018
No there were no errors at all.  Someone on the pixinsight forums suggested that since the images are uncalibrated the star align might be picking up noise and dead pixels as stars... So I am going to start again and calibrate with some darks from my library and see what happens..
Title: Re: Horsehead gone weird
Post by: chris.bailey on 16:18:25, 11 February, 2018
Ian

Yes I saw Robs comment. Worth try calibration first and if not fiddle with the settings I listed.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 19:56:46, 08 March, 2018
I am really starting to despair here so it looks like the original problem was indeed that the alignment routine wasn't detecting stars properly I have calibrated my original lights with flat and dark images following the tutorial here: http://www.lightvortexastronomy.com/tutorial-pre-processing-calibrating-and-stacking-images-in-pixinsight.html (http://www.lightvortexastronomy.com/tutorial-pre-processing-calibrating-and-stacking-images-in-pixinsight.html) and at the end of the process got a stacked output that was clearly mosaicked as I would expect and the stars were nicely lined up.

The problem is that the calibration and debayer process appear to have completely wiped out all the detail except for a few of the brightest stars.  The attached image show the same frame, the top one was calibrated then debayered, the lower one debayered direct from the camera raw.  Both were then stretched with the screen transfer function to make them visible... as you can see the only colour left in the calibrated image is blue and all the detail is gone.

This is really frustrating as I have a lot of data, over 80 subframes at 5 minutes each but I just can't get anything out of the data :(
Title: Re: Horsehead gone weird
Post by: chris.bailey on 20:09:28, 08 March, 2018
Ian

Any chance of Dropboxing a handful of image and I'll take a look.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 20:59:35, 08 March, 2018
Thanks Chris, I have put a selection of the lights, darks and flats here: https://drive.google.com/open?id=1UJQX1qU0wRvETFXjaO8JQgC0pEvBzKm3 along with the master dark and flat I produced.
Title: Re: Horsehead gone weird
Post by: chris.bailey on 22:40:08, 08 March, 2018
Ian

A couple of things. Your Master Dark has 14 bit ADU readings higher than your lights so light-masterdark will result in negative values which will be rounded to zero. Could there have been some light leak in taking the darks. Second you dont seem to have any bias frames. Flats won't work correctly without them

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 00:02:02, 09 March, 2018
I don't think there could have been any light leakage, the dust cap was on the scope, the obs roof was closed and the lights out... hard to get it much darker... I could reshoot them or try something from a previous session?  how have you measured this value?

as for the bias, I have never used bias before,  I seem to remember reading they weren't required as the dark frame also contains the bias signal, is that not correct?
Title: Re: Horsehead gone weird
Post by: chris.bailey on 08:52:28, 09 March, 2018
Ian

Your MasterDark has 14 bit ADU values of 512 pretty constantly. Your Lights vary obviously but are typically in the 530 range so not quite zero but very low.

Yes the Darks contain the Bias but so do the Flats. The (slightly simplified) calibration formula is Light-Dark/Flat-Bias. That gives a different answer to Light-Dark/Flat. Now if the Bias signal is very low the error will be small too BUT the error will be larger in the darker parts of the light i.e. where the dust motes and vignetting are. As such calibrating with flats that do not have the bias removed tend to under correct.

I'm now downloading all the files and will have a play.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 09:21:04, 09 March, 2018
Ok, I guess its pretty easy to get bias.. I'll shoot a load when I get home tonight.

It would be interesting to have a look at the ADU values of images I have previously managed to process, what tool are you using to measure that value?

I look forward to seeing what you get, I have only uploaded a limited selection of the lights and darks, I figured uploading/downloading 160 odd frames would take us far into the diminishing returns calculation! but if you would like more lights or darks let me know.
Title: Re: Horsehead gone weird
Post by: chris.bailey on 09:26:58, 09 March, 2018
Ian

I'm just mousing over the image in PI and reading off the value at the bottom of the screen.

I have downloaded all the images and though the signal is low, I don't get that solid blue image you do. In PI Format Explorer for RAW do you it setup as pure Raw - you just click on the Pure RAW button in Edit Preferences.

Chris
Title: Re: Horsehead gone weird
Post by: chris.bailey on 09:32:04, 09 March, 2018
Ian

This is what I get out of it (using just darks - there isn't a lot to the flats so I get very similar if I do use the flats)

Just colour calibrated quickly and stretched.

Chris
Title: Re: Horsehead gone weird
Post by: Carole on 10:01:35, 09 March, 2018
I am glad Chris managed to get something out of your data.  I tried, but they wouldn't even load into DSS and DSS kept crashing, so I tried Astroart even though I don't yet know how to use that for coloured files, and it kept telling me the images were different sizes, so would not stack.

I thought the flats looked over exposed. 

Sorry I couldn't help.

Carole
Title: Re: Horsehead gone weird
Post by: chris.bailey on 10:43:00, 09 March, 2018
Astro Pixel Processor does a reasonable job with it on default settings. This is what I got entirely within APP other than resize for web.

Chris

Title: Re: Horsehead gone weird
Post by: Ian Straton on 12:35:11, 09 March, 2018
Ian

I'm just mousing over the image in PI and reading off the value at the bottom of the screen.

I have downloaded all the images and though the signal is low, I don't get that solid blue image you do. In PI Format Explorer for RAW do you it setup as pure Raw - you just click on the Pure RAW button in Edit Preferences.

Chris

Yes, this was a setting I changed recently (in-between my first and second efforts on this thread), currently set to Pure RAW
Title: Re: Horsehead gone weird
Post by: Ian Straton on 12:44:00, 09 March, 2018
Chris, I am pleased to see you get a result out of my data but I still don't really understand why I'm getting such weird results.   Could you run me through your work flow so I can try and replicate your steps and see if I get the same result?

I have to say I am rather disappointed by the low SNR in my lights... I thought 5 minutes at iso 400 should give a reasonable signal without generating lots of noise, guess I'll have to try iso 800 instead..
Title: Re: Horsehead gone weird
Post by: chris.bailey on 14:06:38, 09 March, 2018
Ian

Nothing at all special. Image Calibration (Darks only). Debayer (Auto, VNG). Star Align (Default Settings). Image Integration (Winsor Clip [4,3], Large and small Scale Rejection Enabled. I then just applied the unlinked channel screen stretch to the histogram tool.

With Astro Pixel Processor I ran it through with flats and darks with all default settings other than enable BPP in calibration tab and some local normalisation (1 and 3 I think).

This is a pretty faint area (other than Alnitak!) so with an OSC CCD I have used 10 minute subs and if doing it again in RGB may well go longer still.

Chris
Title: Re: Horsehead gone weird
Post by: Nomis Elfactem on 14:15:07, 09 March, 2018
What camera is it Ian ?

Anything younger than a EOS600d and you'll be fine with ISO 800 and possibly 1600 (the later chips are a whole lot better than they used to be).

When I imaged with my 600d I'd hardly get anything at 5 mins on ISO 400.... ISO 800 always felt optimum  ;)

S.
Title: Re: Horsehead gone weird
Post by: allyman on 17:12:50, 09 March, 2018
Hi Folks,

I've not been following this thread but I had this or a very similar problem in Pixinsight so I recognise the weirdness. Basically I had screwed up during integrating as I was experimenting with Local Normalisation at the time. For the life of me I couldn't work out what I had done but the results were very similar, repeatedly. I fixed it by deleting my PI processing folder and starting again.

GrahamM.
Title: Re: Horsehead gone weird
Post by: Ian Straton on 18:37:00, 09 March, 2018
What camera is it Ian ?

Anything younger than a EOS600d and you'll be fine with ISO 800 and possibly 1600 (the later chips are a whole lot better than they used to be).

When I imaged with my 600d I'd hardly get anything at 5 mins on ISO 400.... ISO 800 always felt optimum  ;)

S.

550D, commonly quoted sweet spot is "between 400 and 800" since you can only select *either* 400 *or* 800 I have generally gone with 400 for lower noise... looks like I . might need to boost up to 800 for the higher gain...
Title: Re: Horsehead gone weird
Post by: Annie on 14:43:54, 10 March, 2018
I've used a 550D for 3 years now and it's always performed well re: noise at ISO800.

Annie
Title: Re: Horsehead gone weird
Post by: Ian Straton on 22:12:24, 10 March, 2018
Back to the original problem now, butterfly shaped stars after stacking...

Star alignment didn't report any errors.  it was reporting lots of RANSAC matches for each frame... Should I decrease the RANSAC tolerance so that it finds fewer matches - forcing it to match only the brightest stars?
Title: Re: Horsehead gone weird
Post by: chris.bailey on 08:44:03, 11 March, 2018
That is very strange then Ian as I ran your data through PI Star Alignment on default settings and it stacked fine. You are doing Calibration - Debayer - Star Align - Integration?

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 22:06:57, 11 March, 2018
Hi chris, yes, calibrated (dark, flat and bias), debayer, star align then integrate... I did notice that although the alignment never errorred the report offsets were very low, subpixel low... I have no idea what it is aligning on but it isn't stars...

The pictures I uploaded were the middle 20 of the first nights run, the full set of 80 was captured over 2 different nights and a meridian flip so framing is slightly different.  The gross rotation of the meridian flip has been handled fine but the x - y translation seems entirely absent..
Title: Re: Horsehead gone weird
Post by: chris.bailey on 22:16:55, 11 March, 2018
Ian

I think therefore that the frame chosen as the reference may be the issue, as may some of the frames have a very low S/N ratio. I am in the habit of running all frames through the SubFrameSelector tool and assigning the quality from that as a keyword and then use that as a weighting in stacking. The SFS tool will also give information about what is the best frame to use as a reference.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 22:41:27, 11 March, 2018
I did the same thing, used the subframe selector and used the frame with the highest weighting as the reference... :(
Title: Re: Horsehead gone weird
Post by: chris.bailey on 07:50:45, 12 March, 2018
Ian

VERY strange then. Without seeing all the sub-frames, I don't know what else to suggest.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 10:09:23, 12 March, 2018
Chris, I have uploaded the full set of lights and the bias files I created to the same google drive folder I linked to earlier.

For me the subframe selector showed img_5043 as having the highest snr, lowest eccentricity and lowest fwhm so this is the frame I used as reference.

I was thinking about what I changed in between stacking attempts and came up with the following: When I got the blue frame I had run the cosmetic correction applet to remove hot and cold pixels, given the pervasive H-alpha in the region its possible that I set the tolerance to a point where it simply removed every red pixel?  However of course this was the run where registration and integration produced the expected outcome, so it does suggest that without the hot pixel removal the registration process is still picking up on hot pixels or noise instead of actual stars despite the improved calibration of the subs...
Title: Re: Horsehead gone weird
Post by: chris.bailey on 16:00:50, 12 March, 2018
Ian

I re ran all 80 light frames and got them to properly register in Star Alignment with Detection Scales of 6 and Log Sensitivity of -1.5. Stars are quite big and lots of them are very faint but these setting did the trick.

If you ran cosmetic correction, be sure to tick the CFA button. I ran it with a Hot Sigma of 5.0 and a Cold Sigma of 4.0 and it did clean the frames up nicely.

Given the very low S/N of the individual subs, the stack ended up pretty good I think. This is just with the default unlinked screen stretch applied.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 22:56:32, 12 March, 2018
Chris, that's amazing, thank you for your efforts, I'll check my settings and see if I can replicate what you got :) fingers crossed!
Title: Re: Horsehead gone weird
Post by: Ian Straton on 18:34:22, 14 March, 2018
Great news!  I have got a decent result out of them at last!  Thank you so much for the help and advice :)
just one question, the tweaks to the image registration settings, did those make the algorithm more selective or less selective regarding the detection stars?
Title: Re: Horsehead gone weird
Post by: chris.bailey on 19:51:19, 14 March, 2018
Great news Ian! Detection Scales 6 = bigger stars. Log Sensitivity  -1.5 = more stars. These are the two settings I play with when things dont register.

Chris
Title: Re: Horsehead gone weird
Post by: Ian Straton on 11:07:30, 15 March, 2018
good to know,  the bigger stars thing might be a common issue for me, the Vixon scope has really thick spider vanes which generate big diffraction spikes which make the stars big and square instead of small and round...
TinyPortal © 2005-2012