At the Almost Heaven Star Party in August 2023,, I was one of a group of NOVAC imagers that decided we would start on a collaboration on Lower's Nebula (sh2-261). It would just be imageable at the end of the night in early September and we could give people three months to collect data. Our aspiration was 500 hours (100 H and 200 each O and S) but we didn't know whether we could get close to that.  We didn't want  to limit who could participate so we didn't put any real restrictions on field of view or image scale (though we did give some broad guidelines). This was great for encouraging participation but it made things challenging when it was time to combine the data.

Then we spent the next three months collecting data and periodically calibrating and sending calibrated subs to the group. We used the open source application SyncThing to distribute all the subs to each person so everyone had all the data.

We ultimately had seven people and 12 telescopes contribute to the project. The acquisition period lasted long enough that two of use had equipment upgrades (a camera for me and a telescope for @Chris Kagy) so we accounted for two scope/camera combinations each. We also had the participation of two telescopes in remote observatories. Chris and I are team members on a Planewave CDK14 at SRO and I'm also fortunate to be able to use a telescope belonging to a friend at an observatory in Texas. The instrument there, a Ceravolo 300mm astrograph, is amazing and it's a privilege to be able to use it. 

Here's how the submitted data broke down:

The session information listed in astrobin is just for my ASI6200 subs. There is no practical way to enter all that data for everything, unfortunately. 

Amazingly, we surpassed our goal!

Now came the challenge of processing the data. We had such a difference in FoV, image scale and rotation angle that getting it to combine seemed a hopeless challenge. Most had decided to build a master from the integrated masters from each scope. I was hoping to do a truly Big Integration. I had done some tinkering with a small amount of data and by registering everything to a master image from my scope with the ASI6200 camera, I could produce a pretty seamless integration thanks to local normalization. It all had to be done manually because WBPP couldn't combine such different sources. It took several days of registering and running local normalization. The next step was to do the actual integration. I started with O III. The oxygen signal was very faint and I was really curious to see what a combined integration would look like. I loaded everything into WBPP and kicked it off only to find out that even on my 64GB system it required more memory than I had available. Those ASI6200 subs are big! And when you are registering everything to that size then you end up trying to integrate over 2,000 very big subs. Too big for my system. I also tried Fast Integration (even though I didn't think it would produce a result I could use) with the already registered subs but it would get stuck about 80% of the way through.

Instead, I did a complete integration of each scope/camera combination producing a fully integrated master from each one. Then I did some "post pre-processing" before combining those masters. I ran DBE, BXT (in correct-only mode) and NXT to produce as clean a master for each scope as possible. The I registered them all to my ASI6200MM master as that was the FoV I wanted to use. I ran local normalization on them all, again using my masters as the reference. Then I combined them in image integration using "no weights" and "no rejection". I tried the various weighting options but they all looked worse than the straight combine and the same was true of rejection algorithms. The result was a fairly clean combination. There was lots of signal but the noise levels clearly changed based on the overlap of the various masters. This is how they intersected:


We had good coverage of the nebula, but increasingly sparse coverage the further out we got. The obvious choice would be to crop to the best SNR region in the center but I didn't want things framed that tightly. Thankfully an additional run of NXT (a bit down the road) cleaned up those transitions.

The signal level was great, the faint oxygen showed good signal in the center though it didn't go out very far and H and S were terrific. I processed each master this way:

DBE (even though I had run DBE on the masters there were still some gradients after combining)
BXT
SXT (extracting stars)
NXT
GHS (GH, linear, GH)

The backgrounds were fairly cruddy thanks to the varying SNR levels so I knew I wouldn't be able to go after the faintest details. That was disappointing as it somewhat defeated the purpose of all that integration time.

Now I had to process the image:
channel combination (SHO)
histogram transformation (using mask to get colors approximately where I wanted them)
curves (hue with mask to push the green towards orange)
background neutralization (with inverted mask)
histogram transform (with inverted mask to lower background crud)
curves (contrast)
LHE (masked)
SCNR (additive mask, green) - this was an aesthetic choice that pushed the outer parts of the nebula toward gold)
NXT (inverted mask - small amount - to clean up some noise in the fringe)
MLT (mild sharpening)
pixel math in stars
fina crop

stars:
channel combination (SHO)
GHS (GH)
SCNR (to remove green tinge on larger stars)
unsharp mask (very mild sharpening)

Was this worthwhile? In terms of image quality gained for acquisition time expended I think the answer is no. I suspect I could have produced an image of equal quality from 1/10th the image time. The free-for-all nature of the collaboration meant that detail suffered at the image scale I chose to use as a reference. The variation and odd overlap in rotation angles limited the amount of faint fuzzy stuff we could hope to find in the background. And, finally needing to integrate the masters rather than a full integration also limited the faintest single we could detect. I'm not sure what the math says on that distinction but that's what the background said to me.

I had to massage the data quite a bit to get it to combine in a usable way and it's possible, perhaps probable, that some of my steps ultimately limited what I could get out of it. So, in that sense, it was a bust.

However, working with the group on acquisition and talking through the challenges and working through them was fun and challenging. Sometimes it was a "hurts so good" kind of challenging and it would not have been hard to give up on this, but I was stubborn and wanted to see what I could get out of it and overall, I'm fairly pleased with the result. There's lots of subtle detail in the nebula itself and I was able to get a nice sense of depth in it. I think it's one of the more eye catching images I've done but I also know where every single wart is and so it's hard to be objective about it.

I'm glad I was able to participate and I'm looking forward to seeing what the other collaborators produce!

Here are versions from the other collaborators:
Chris: https://www.astrobin.com/q7o9rk/
Gowri: https://www.astrobin.com/c8fi0m/