DSS alignment puzzle

[Tommohawk]

Odd one this - never had problems like it before with DSS.

I've done a 300mm FL image of the WEst and East Veil, and was initially pretty pleased. But then i noticed that a portion of it seems to have misaligned. I repeated the stack, just in case DSS had a funny five minutes, but same result.

Here's a crop of the stacked image with the approx problem outlined. This is basically the top right area - the rest of its fine. Howzit possible for only a bit to be misaligned??

BTW I've looked at the individual images with ++ mag and they all look fine. I did have to slightly reframe at one point but I cant see how that would make a difference. I cant think of a way of comparing the images all together - if I could put them in PS in different layers that would work, but no easy way of putting CR2s in PS, and they wouldnt align perfectly anyhow unless I repositioned them manually I supose.

Any ideas please?

Edited September 28, 2016 by Tommohawk

[Jonk]

In photoshop, you can load all of your frames as individual layers - as long as you have the latest Camera Raw installed, so PS can read the CR2 files.

You can then do an automatic layer alignment, with 'median' filter - this should at least line them all up, and you can turn all layers off, then on 1 by 1 to see what's going on.

Look up loneyspeck.com, he has some video tutorials, which should work for what you're ttying to find out.

[Thalestris24]

Hiya

Strange dss problem! Maybe something to do with the large number of stars? Just speculating! What calibration frames and dss settings are you using? Can we see the whole image?

Louise

[Tommohawk]

OK guys, thanks for that.... need some ZZZs and busy tomorrow til late.

Quick thought - I only have olde worlde PS CS2 - doesnt seem to read CR2 but could convert to tiff or similar. Will try when have mo

[Tommohawk]

OK, further weird goings on.

First I imported each image into PS (CS2) as a separate layer, moved them all around to line up ... took forever! Looking quickly I discovered that a few were rotated with respect to the others but noting else obvous amiss - meant to look in mmore detail but it failed to save properl;y and now I have a 2 GB file with no layers showing. Grr.

So, I went back to DSS and it seemed to me that the stack list had no Dx  Dy or rotation values. Couldnt be sure if I'd just done something strange, so srated again with just 3 frames, no darks/flats/bias, and all looked OK with Dx Dy and rotation values showing, and a normal image. 

Then I added in darks flats and bias and got the same effect as above - a strange blurring in the centre only. So I repeated without flats - having deleted the txt data files to ensure a clean start - and got this:

whats that all about!!?? 

Edited September 30, 2016 by Tommohawk

[StargeezerTim]

I had sort of similar problems a while back. I changed the stacking setting from auto to bicubic and it worked. 

[Tommohawk]

36 minutes ago, StargeezerTim said:

I had sort of similar problems a while back. I changed the stacking setting from auto to bicubic and it worked. 

Thanks thats worth knowing for sure - but weird that its only occurred with that combination ie lights bias darks no flats. Plus I've used DSS with loads of other DSOs and various cal frames and never had that before.

I just use the recommended settings - I'm running it again now to see if its repeatable, then I'll check the settings 

I wonder if Louise is onto something with the exceptional number of stars - maybe I should change the threshold?

[Tommohawk]

Just found another thread re the auto v. bicubic issue - so Ill try that next.

BTW Ive got 16000 stars per frame - I'm guessing that's enough?

[Thalestris24]

1 hour ago, Tommohawk said:

Just found another thread re the auto v. bicubic issue - so Ill try that next.

BTW Ive got 16000 stars per frame - I'm guessing that's enough?

You could try increasing the threshold to lower the number of stars to maybe 100-200

Louise

[Tommohawk]

Wow this has turned into a total nightmare!!

Having had problems with the weird images as posted above, I tried to simply replicate the successful 3 sub stack witohut  cal frames that I had achived previously - and it wasnt working. All sorts of odd renditions - stripes, a very dark frame with no detail at all. This was with everything set as it had been previously so somethin out of kilter.

I tried increasing the threshold to lower the star count - no diffrence. I tried using bicubic mode - no difference. I tried both of these - just the same wacky images.

 So - something mangled in DSS, or possibl;e mangled subs, but the subs render in PS just fine. Likewise the cal frames look normal.

So I removed the registry entries - no difference. Then I removed and reinstalled DSS 3.3.4 , to find it wouldnt run at all. DSS live worked OK but not the ormal version - it wouldnt open.

So I reinstalled the last stable version of DSS (3.3.2?) and then re-installed 3.3.4. ... and now it opens OK. 

BUT having run the stack again I have the same weird blurred area. It looks like the colours have separated. Maybe I should find a non-bio version!

I'll try some other experiments to narrow down where the problems arise

[Jonk]

300mm fl - Lens creep?

[Tommohawk]

29 minutes ago, Jonk said:

300mm fl - Lens creep?

Its a fixed FL not zoom so hopefully not! But thanks for the idea. BTW the subs all look fine.

[Owmuchonomy]

This very thing happens to me occasionally and I call it 'melt'.  In my case it is always solved by re-doing the stack without flats. However, reading your thread it looks like you didn't use flats so I am a bit stumped. Also, as Louise says I always set the threshold to around 100 stars. That's plenty.

[Tommohawk]

I'll try resetting the threshold. TBH in the past Ive always just left it as the default 10% and never had a problem.

This time there is a shed-load of stars - over 16,000 - but when I upped threshold to 65% star count reduced to ~2000 but didn't make any difference.

Threshold is v twitchy - if I up it to 70% it find no stars at all. 

ATM I'm on another theory - I use intersect mode rather than normal  cos it saves cropping after. MAybe that's relevant. Stacks just finishing now and....

... hmmmm. Same again.

Tried the threshold again: 70% finds no stars, 69% finds 1400!! I'lll try that!

1 hour ago, Owmuchonomy said:

This very thing happens to me occasionally and I call it 'melt'.  In my case it is always solved by re-doing the stack without flats. However, reading your thread it looks like you didn't use flats so I am a bit stumped. Also, as Louise says I always set the threshold to around 100 stars. That's plenty.

OK thanks for that Chris - so at least someone else has seen this!

But not to do with calibration I think.

BTW whan is calibration done - when registering or when stacking.. or both?

 

[Owmuchonomy]

From what I understand stacking is the last process but I'm no expert. Just double checking you have carefully checked every sub for an issue. A rogue sub has tripped me up before. I will go and get my notes as I keep diaries of what I do to see if that throws anything up.

[Owmuchonomy]

Ok, I had this issue in October 2014. This is how I solved it:

-Make sure sigma clipping median combination method is selected for both flats and lights

-Stacked everything EXCEPT flats

-Stacked again with flats this time

came out fine.

[Tommohawk]

Well I'm just about run out of options and still have problems.

As in previous post, I've set the threshold for star detection to 69% and get 1400 stars. if I go to 70% I get no stars. this gave a small improvement - the dodgy area is a bit smaller.

I've set stacking alignment to bicubic, and lights and flats to median Kappa sigma clipping. Darks and Bias are set to median.

I still have the same problem - I'm stumped and TBH getting a bit cheesed off with the whole thing. I've only managed 3 decent DSO efforts this year due to rotten weather, and now I cant do anytihng with the subs.

I could try converting CR2s toTIFFS before stacking, or some other software I suppose.

Any more ideas welcome.. please!!!

PS - sorry Chris, I meant to say thanks for taking the time to check that out.

Edited October 2, 2016 by Tommohawk

[Steve Ward]

Have you just tried stacking the light frames on their own , as it's all worked for you before then surely there must be something amiss with what you've fed DSS this time round.

I'd reset DSS to the default settings do a simple elimination exercise first adding each set of Dark/Flat/Bias in turn until you find the culprits.

[Tommohawk]

1 minute ago, Steve Ward said:

Have you just tried stacking the light frames on their own , as it's all worked for you before then surely there must be something amiss with what you've fed DSS this time round.

I'd reset DSS to the default settings do a simple elimination exercise first adding each set of Dark/Flat/Bias in turn until you find the culprits.

Ive done this Steve - in fact Ive tried to methodically go through everything but been thwarted to some extent by some variations in DSS performance, culminating in all the wacky output shown above. I'm on autosave 34!!

Question: If varying whether cal frames are added, do I need to register again? Put differently, at what point does cal occur - in registering or in stacking? Or both?

[Jonk]

Which frame have you chosen as the reference? Pick the best by eye, then stack say the first half of your subs, then same again with the other half. It might weed the issue out.

[Tommohawk]

6 hours ago, Jonk said:

Which frame have you chosen as the reference? Pick the best by eye, then stack say the first half of your subs, then same again with the other half. It might weed the issue out.

One of my first tests showed the same problem when only using 3 obviously very good subs with minimal difference - and in that serieis of experiments adding the flat seemed to cause the problem. The difficulty was that when I removed the flat a whole new series of problems arose which threw me completely.

The latest is.... Having got DSS back on an even keel,  it looks OK with Bias and Darks, and interestingly I can see now that without the flat whilst there is no "melting" effect, there is a bright central area which corresponds to the problem area found previously with the flats included.

So its adding the flat thats the problem after all. I tried stacking with lights set to entropy weighted average and that gave a hideous green colour to everything. I'm trying a few other things and will post back in a bit.

Thanks again for the suggestions.

[Steve Ward]

Could well be a gradual build-up of dew on the lens that's causing the issue then ... ?

Check the first sub against the last , this will show the greatest difference if it's the case.

[Tommohawk]

19 minutes ago, Steve Ward said:

Could well be a gradual build-up of dew on the lens that's causing the issue then ... ?

Check the first sub against the last , this will show the greatest difference if it's the case.

I did get a little condensation at one point and I wondered the same - which is why I did my initial tests with only the first 3 subs - no dew for sure. Also the problem only arises when I add the flat. With no flat it looks OK, so it doesnt seem to be a problem with the subs.

I'm trying again using the individual flats rather than the master flat. Cant see how that will make a difference, but gotta try something. Next step will be to redo a different image Ive done previously without problems.

I've also downloaded a trial of nebulosity.

[jazza]

If you like you could upload three subs that cause the issue, along with your master flat/dark somewhere and I/we could have a look to see if we can reproduce the problem? 
I could at the very least align them by hand and 'blink' them to see if they show noticeable difference.
Just an idea. Bit of a strange one!

[Tommohawk]

13 hours ago, jazza said:

If you like you could upload three subs that cause the issue, along with your master flat/dark somewhere and I/we could have a look to see if we can reproduce the problem? 
I could at the very least align them by hand and 'blink' them to see if they show noticeable difference.
Just an idea. Bit of a strange one!

Thats a very kind offer! I've just repeated the original experiment ie with just 3 subs and I'm definitely getting the same problem. Will be a while loading cos my broadband is super-slow...

... some time later....

IMG_3718.CR2

IMG_3712.CR2

IMG_3717.CR2

MasterDark_ISO800_240s.tif

MasterOffset_ISO800.tif

MasterFlat_ISO800.tif