Strange shapes in the stacked output from DSS

[almcl]

Had a bit of disaster the other night and I wonder if the community would care to suggest what's going on?

16 subs all have reasonably round stars (Canon 200mm lens on Canon 700d, guided by finderscope and mounted on EQ6 ALT/AZ Goto) but when stacked in DSS, some really strange shapes show up slightly right of center at the top of the frame. The bottom is not quite as bad.

Now I had poblems with guiding - the normally ultra reliable QHY5Lii was causing PHD2 to trip out with the 'hadn't captured a frame in 17secs' message. The flats had weird bits of dust and squiggly stuff on them.  The ever present vicious amp glow was as bad as ever, the guiding was all over the place when the camera would connect and, of course, cloud rolled in before the required subs had been acquired.

But what's causing the mis-aligned stars in the DSS output at the top but not the bottom?  I can't get the number of detected stars below about 900 (at 91% threshold) if I increase the threshold to 92%, no stars are detected, so I wonder if some of the 'stars' are actually noise and it is this that is causing the mis-alignment?

Below are a top and bottom crop of the output from DSS, stretched to show the shapes as well as the first and last subs and the entire output from DSS. 

[bobro]

Could you post the entire first and last subs (rather than just crops) so that any change can be looked for? Thanks.

[tooth_dr]

Can you apply a median filter to reduce the number of stars detected?

[almcl]

Thanks both.  I'll have a go with a median filter and report back.  In the meantime here are the full size subs:

[almcl]

Tried using the median filter for stacking and while it did greatly reduce the number of detected stars (thanks for pointing me in that direction, tooth_dr) this didn't solve the stacking problem :-(

 

[tooth_dr]

Disappointing! I find DSS sometimes can be temperamental but usually down to a error in the settings on my part lol

[bobro]

Interesting ..... the first and last subs align perfectly and also align with the strange corner at the brightest points.  To my knowledge DSS doesn't do any scale changes though does handle rotation. Examining more subs could help by finding out if there is any difference in subs. I'm happy to do this if you are ok to post them.

Stacking with a different program could also be an idea if this is possible.

[almcl]

1 hour ago, tooth_dr said:

Disappointing! I find DSS sometimes can be temperamental but usually down to a error in the settings on my part lol

Isn't that the truth, TD!  Almost certainly it will turn out to be operator error...

 

45 minutes ago, bobro said:

Examining more subs could help by finding out if there is any difference in subs. I'm happy to do this if you are ok to post them.

Stacking with a different program could also be an idea if this is possible.

That's a most generous offer bobro.  Thank you.

I've put jpeg versions (the raws would be a bit large) of the files here: https://www.dropbox.com/sh/t6g6yu4pe2a2w3q/AABxfTHbqE26hf--LIODL9Vda?dl=0

At the moment DSS is my only stacking program although I am on the lookout for one that doesn't white balance the output which StarTools would rather prefer.

[bobro]

The jpegs all look very similar with no issue I could see, so I tried to stack them with DSS. Similar to your findings the number of stars jumped from 0 to over 900 as the star detection threshold was reduced to 48. Even with this setting I couldn't get the images to stack, so I enabled noise reduction in the star detection threshold window. This allowed #detected stars to smoothly increase as the detection threshold was lowered. Stacking then worked ok and the resulting image didn't show the curious corner. Perhaps you could try the same with the noise reduction?

 

[almcl]

Thanks for that bobro.  I wasn't able to get the stars down to quite that level but with your and tooth_dr's help, I found the source of the issue!

Just in case anyone happens upon this thread in future, it was a problem of registration data and for some reason I originally stacked using the intersection option with far too many (>2000) stars detected.  Once the median filter was applied and the detection threshold set to 85%, plus checking the 'Register already registered pictures' box which habitually I uncheck (can't bear waiting 5 minutes for it re-register files it has registered already) and deleting the previous registration data (.txt files) then rerunning the stack, produced a normal image.

Below is the same area as the 1st image in the thread, after the re-run (no idea why the colour is so radically different...)

[ollypenrice]

You could download a free trial copy of AstroArt 6.0 to see if that could stack and align effectively. You can't save on the trial version but it might be instructive to see if AA can do what DSS can't.

Olly

[almcl]

Sorry, Olly, didn't see this till just now...

I do have a trial copy of Astroart, but cannot get on with it - the interface is just too baffling and the output (on the odd occasion when I've got it to load and align some lights) a monochrome grainy grey with almost no detail present.  

Would dearly like a stacking prgram that doesn't muck around with the white balance, but haven't yet found one my simple brain can understand.