DSS and stacking. I'm doing something wrong, but....

[Delboy_Hog]

I don't know what's causing the problem...although I think it might be the images rather than the stacking.  Either way, I'd be **really** grateful for your help!

Short version of the question: 

How can a series of seemingly ok individual photos without any obvious signs of trailing, come out of DSS in such a calamitous mess!?

Longer version:

I've been trying to capture as much of the Veil nebula as possible in a single frame, with the Canon 6d, Skywatcher ED80, HEQ5.  5 min subs at ISO 1600, guided by a Skywatcher Synguider.  Utterly fabulous patch of the sky, I always think!

All of the individual light frames look reasonable (albeit to my admittedly un-trained eye) - example frame below.  But stacking them results in something that looks like one of those hidden / 3d image things that make you feel a bit sick if you look at them for too long!!  On the plus side, the calibration frames seem to be evening out the illumination across the field of view, and got rid of the dust-bunnies!

My guesstimate of what's going wrong:

1) I gather the 6d is not a great match for the ED80 because the chip isn't fully (evenly?) illuminated.  I may have to accept that the corners will always be ropey, and be prepared to crop significantly?  Can flats do anything about the star shapes, or are they all about the illumination across the field of view?

2) To compound that, in my over-zealous attempt at dithering, I attempted to manually dither the frames (stop the autoguider, tap the arrow on the handset to move the field of view slightly, restart the guider, start the next frame), but I wonder if some of the movements I made were too big.  I read that only the tiniest of movements is required, so have the bigger movements made it harder for DSS to stack those streaky corner stars?  (e.g. if it's trying to stack a seriously sausagey corner star from 1 frame with a non-corner, non-streaky corner star from a different frame, it's going to struggle?

Example of the light frames:

 

And the images: straight out of DSS, no processing...

Lights and calibration frames - check out those corners!!

Light frames only, and took the lowest scoring frames out of the stack...about the best of a bad bunch - though the corners (especially the bottom left) are still a mess...still, I could probably process a crop of this?

Last one - a hideous attempt at using Kappa Sigma process on the light frames - it appears to have focused on aligning the stars in the corner, rather than in the middle....appears to have worked on the satellite trails though!!

 

Some bonus questions for any DSS experts out there:

1) I gather the "Kappa-Sigma" setting in DSS is the best approach to remove satellite trails - is that Kappa Sigma or Median Kappa Sigma?

2) And is that just on the light frames, or do you apply that approach to the calibration frames as well?

3) And do people just use the default settings when you choose one of these, or do you change the Kappa figure or the number of iterations? 

Many thanks for any light you can shed on all of this for me! 

Derek

[pete_l]

Just by eyeballing the left side, there appears to be 6 separate "dithers". Between them they make up about 1/14th of the frame. Normally a dither would be 2 or 3 px per frame.

I would suspect that if the polar alignment wasn't pretty good, there would be some field rotation which is tricky for a stacker to account for, since it is only looking for uniform X,Y displacements across the field. Not ones that vary due to curvature.
How long are you allowing the mount to "settle" back into its tracking after each dither?

[michael8554]

The corners look fine to me in all but the last one, given the combination of gear you used. 

What star threshold did you use ( how many stars used?) 

Dither should be 12 or so pixels for a DSLR.

Michael 

[Delboy_Hog]

17 hours ago, pete_l said:

Just by eyeballing the left side, there appears to be 6 separate "dithers". Between them they make up about 1/14th of the frame. Normally a dither would be 2 or 3 px per frame.

I would suspect that if the polar alignment wasn't pretty good, there would be some field rotation which is tricky for a stacker to account for, since it is only looking for uniform X,Y displacements across the field. Not ones that vary due to curvature.
How long are you allowing the mount to "settle" back into its tracking after each dither?

Thanks for this Pete, that's really helpful.  I'll definitely be going for much smaller dithers next time around.

Your suspicion about my polar alignment could be spot on too.  I guess field rotation over that kind of time-frame would explain why the individual subs had no trailing, but the final product wasn't spot on...  I'm interested in this - if there had been field rotation, that would be reflected in the "angle" column of DSS, is that right?  I noticed there were some differences in the figures in that column when the frames were stacked.  Probably only about half a degree over the time period these frames were taken in - would that be enough to give the stacking a hard time, do you think?

The Synguider appears to be pretty good at continuing where it left off, once you 'resume' from its previous calibration run, so I've only been allowing a few seconds.  But then I hit restart on the intervalometer, which gives it a few more seconds to settle I suppose.  It always feels quite time-consuming at the time, but I've no doubt it's worth it for the noise reduction it brings. 

Thanks again!

Derek

[Delboy_Hog]

14 hours ago, michael8554 said:

The corners look fine to me in all but the last one, given the combination of gear you used. 

What star threshold did you use ( how many stars used?) 

Dither should be 12 or so pixels for a DSLR.

Michael 

Hi Michael, 

Many thanks for coming back on this.  The Star Threshold is an interesting question actually - I think I went with the standard DSS setting, and remember it found several thousand stars, which seemed a bit many - I've read somewhere since that about 100 stars is optimal, does that sound about right?  Do you have a particular threshold that works for you, or would it vary depending on your target?

Thanks again!

Derek

[michael8554]

There's probably an optimum ratio but no idea what it is. 

Too much and you could be stacking hot pixels. 

Your image has loads of stars, another object may have far fewer, so depends. 

Michael 

[david_taurus83]

That single sub looks really good. Is it unstretched? Your data looks good so something is going wrong with your DSS settings. I use the standard settings in DSS but I do adjust the slider to detect just over 100 stars. You could post a link to your data on here and see if anyone else gets the same result in DSS.

[Delboy_Hog]

15 hours ago, david_taurus83 said:

That single sub looks really good. Is it unstretched? Your data looks good so something is going wrong with your DSS settings. I use the standard settings in DSS but I do adjust the slider to detect just over 100 stars. You could post a link to your data on here and see if anyone else gets the same result in DSS.

Thanks David, much appreciated.  Yes, straight off the camera - I was really pleased with the individual subs, so was a bit bewildered with what came out of DSS!

Oh I'd be intrigued to see what anyone else could get from the data - I've uploaded the light frames to Dropbox here (hope this link works?):

 https://www.dropbox.com/l/scl/AACWyRfxTlCwi4kLul4poTuMl6zmhaN8Vew

It took an age so I haven't uploaded the flats and bias frames as well, but if anyone is particularly keen to play with the data, I'd be happy to!

Many thanks,

Derek

[david_taurus83]

I think you've got some really nice data there considering its just over an hours worth. You are probably pushing the limits of the flattener with your 6D but apart from the corners and edges the stars look nice and round. I stacked in DSS with the standard settings and I got something similar to your stack with just the light frames. I set the star detection slider to 89% and ticked the box to reduce noise. This gave me a figure of around 250 stars. It might sound silly but you didn't accidentally mix your flat frames into the light frames tab? I don't know whats happened your stack above with calibration frames!

 

I done a quick process in Pixinsight. The difficulty with this object is its within a dense starfield so it kind of washes out the nebulosity.  We are used to seeing beautiful narrowband images of this but Ha and OIII filters really make the supernova material stand out. I've never been good at processing OSC images but I've managed to reduce the stars a bit. I reckon if you were to add another 3 or 4 hours worth you'd have a brilliant image with a bit of careful processing!

 

 

 

Autosave.tif

[Delboy_Hog]

On ‎11‎/‎08‎/‎2019 at 17:34, david_taurus83 said:

I think you've got some really nice data there considering its just over an hours worth. You are probably pushing the limits of the flattener with your 6D but apart from the corners and edges the stars look nice and round. I stacked in DSS with the standard settings and I got something similar to your stack with just the light frames. I set the star detection slider to 89% and ticked the box to reduce noise. This gave me a figure of around 250 stars. It might sound silly but you didn't accidentally mix your flat frames into the light frames tab? I don't know whats happened your stack above with calibration frames!

 

I done a quick process in Pixinsight. The difficulty with this object is its within a dense starfield so it kind of washes out the nebulosity.  We are used to seeing beautiful narrowband images of this but Ha and OIII filters really make the supernova material stand out. I've never been good at processing OSC images but I've managed to reduce the stars a bit. I reckon if you were to add another 3 or 4 hours worth you'd have a brilliant image with a bit of careful processing!

 

Autosave.tif 171.59 MB · 3 downloads

Hi David,

So sorry for the slow reply, especially after the efforts you went to with the data - I've had to be away with work and it's given me no spare time at all. 

Many thanks for taking the time to run through with the data and seeing what you could get out of it...HOLY MOLY!  That's looking nice!  Especially with limited data, as you say.  You seem to have managed to minimise the tragic corners better than I did, and it has a really nice balance to it, to my eye.  My processing always seems to come out a bit stark and...obvious, if that makes any sense?  I've still not had time to run this through in Photoshop, so I must try it.  Hoping for some clear nights to get some more data in the stack too.

As for whether I included flats in the light frame stack...ha ha, possibly!  I've made more stupid mistakes than that in the past!!

Thanks again for your help with this, and for walking me through what you did in the stacking phase, that's very much appreciated.  I really like seeing the starless version too, really interesting to see - I'm assuming that's a Pixinsight feature - lovely!  Hopefully I'll be back on here before too much longer with a bit more data and having had slightly more success through the stacking!

Thanks again 

Derek

[david_taurus83]

1 hour ago, Delboy_Hog said:

Hi David,

So sorry for the slow reply, especially after the efforts you went to with the data - I've had to be away with work and it's given me no spare time at all. 

Many thanks for taking the time to run through with the data and seeing what you could get out of it...HOLY MOLY!  That's looking nice!  Especially with limited data, as you say.  You seem to have managed to minimise the tragic corners better than I did, and it has a really nice balance to it, to my eye.  My processing always seems to come out a bit stark and...obvious, if that makes any sense?  I've still not had time to run this through in Photoshop, so I must try it.  Hoping for some clear nights to get some more data in the stack too.

As for whether I included flats in the light frame stack...ha ha, possibly!  I've made more stupid mistakes than that in the past!!

Thanks again for your help with this, and for walking me through what you did in the stacking phase, that's very much appreciated.  I really like seeing the starless version too, really interesting to see - I'm assuming that's a Pixinsight feature - lovely!  Hopefully I'll be back on here before too much longer with a bit more data and having had slightly more success through the stacking!

Thanks again 

Derek

Your welcome!

 

Regarding the starless version, I used Starnet++. Its available as an add on for PI but also standalone for windows. I have started using it for creating masks, either nebula or stars. It's very effective. No settings to tweak, just press start!

[vlaiv]

There are couple of things going on here and all are quite "normal".

First, you are using full frame sensor with a scope that does not have such a larger corrected field - hence distortions in the corners. It's the limitation of the gear used and you can't do much about it except crop.

Star trailing in your stack in the corners is another consequence of gear used and imaging method - as noted, you did quite a large dithers between subs. You have fairly large field covered with this setup, and when trying to map a "round" thing onto a "flat" thing there will be distortion - sky is a sphere and imaging surface - sensor, is flat. This leads to slight distortion in star positions over the surface of the sensor. It is very similar to barrel / pincushion distortion that you get from wide lens in regular photography (or some eyepieces in visual).

There would be no problem if you had your frames pointed to same place every time (small dither), but large dither means that different parts of image are stretched / distorted, so DSS is having trouble aligning them as it can't handle image distortion of this sorts. It expects "flat" images - meaning geometrically flat, or each frame having stars the same distance from each other, which is not the case here.

Using different software that is capable of correcting for distortions of this kind will help get better stack. I think that APP can do it as it says among its features: " advanced image registration using true optical distortion correction ".

[Delboy_Hog]

On ‎23‎/‎08‎/‎2019 at 09:13, vlaiv said:

There are couple of things going on here and all are quite "normal".

First, you are using full frame sensor with a scope that does not have such a larger corrected field - hence distortions in the corners. It's the limitation of the gear used and you can't do much about it except crop.

Star trailing in your stack in the corners is another consequence of gear used and imaging method - as noted, you did quite a large dithers between subs. You have fairly large field covered with this setup, and when trying to map a "round" thing onto a "flat" thing there will be distortion - sky is a sphere and imaging surface - sensor, is flat. This leads to slight distortion in star positions over the surface of the sensor. It is very similar to barrel / pincushion distortion that you get from wide lens in regular photography (or some eyepieces in visual).

There would be no problem if you had your frames pointed to same place every time (small dither), but large dither means that different parts of image are stretched / distorted, so DSS is having trouble aligning them as it can't handle image distortion of this sorts. It expects "flat" images - meaning geometrically flat, or each frame having stars the same distance from each other, which is not the case here.

Using different software that is capable of correcting for distortions of this kind will help get better stack. I think that APP can do it as it says among its features: " advanced image registration using true optical distortion correction ".

Thank you Vlaiv, that's a helpful explanation of what's going on here.

I'm currently looking into the various bits of software that I use as compared to some others, so I'll take a look at APP - it would be helpful if there was something out there that could help with this, though I'll certainly be looking to significantly reduce the size of the dither on my next session, so hopefully that'll give DSS a helping hand in the meantime!

Thanks again,

Derek