At the 2023 Almost Heaven Star Party, a group of NOVAC members decided to again collaborate to produce a very deep image of a deep sky target.  In the past we had done the Tadpoles Nebula and the Rice Hat Nebula, and this year we selected Lower's Nebula. Lower's Nebula (Sh2-261, is a supernova remnant on the boarder between Orion and Gemini, and met a number of criteria for selection: It would be well placed through the fall, would be approachable for members with different telescopes and cameras, and seemed like it would give us potential for a nice image when done. Our hope was to collect an ambitious 500 hours of data on the target (100 Ha, 200 each O3 and S2).

Seven people and 12 telescopes contributed data to the project in the end and we gathered an impressive 578 total hours!  Note that the data listed above only includes my time, but here is a full accounting of the contributions of each member: 

As we collected our data, we each calibrated our frames and put them in a standardized directory structure which was synchronized to all team members using SyncThing. Processing this amount of data from such a variety of telescopes was quite a challenge, and I decided to focus on integrating a master of masters.  I used Siril for initial stacking because of its speed.  I wrote two scripts that automated stacking.

One generated quick & dirty masters so we could keep an eye on progress. It registered each person's data by filter and did a stack using the Sum method. It was able to produce masters for each person's filters in approximately 20 minutes on my 8th gen Intel Core i7 system. 

The second script was more thorough and intended to produce higher quality masters. This process calculated and subtracted sky background from each subexposure using a 1st degree polynomial, did a 2-pass registration to select the best frame of each group and aligned to it, and then stacked using an Average methodology with 3 sigma pixel rejection calculated using a 3x3 matrix and the Generalized Extreme Student Deviant method. This second script routinely completed processing all data and generated master in approximately 2.5 hours.

To create the master of masters I first pre-postprocessed each submaster by lightly cropping and using GraXpert to extract the background. To determine where to place sample points in GraXpert I loaded each frame, ran the AI background process and then placed sample points throughout the identifiable background. I then changed the extraction method to RBF, the smoothing factor to 0.7, and extracted the background.

To create the final H, S, and O masters, I then loaded these cleaned submasters into AstroPixelProcessor and processed them as if they were panels of a mosaic. I turned up multiband blending to 30% and used a maximum of 10 iterations of 1st degree local normalization.  The analysis, normalization, registration and stacking completed very quickly.

Finally, for further processing I opened the final masters in PI and proceeded as one normally would. At this point, things (finally) became routine: Crop, light DBE, BXT, SXT, GHS, Channel Combination, etc.  Early in our effort we identified an interesting structure in the O3 frames, and I wanted to process my image to focus on and emphasize that structure.


I deliberately understretched the Ha and S2, and lowered their saturation in an attempt to let the blue O3 structure take center stage.

This was a beast of an effort.  We collectively learned lessons about data management and exchange and, of course, about approaches to process data that exceeded the capacity of our standard techniques. At the start we deliberately decided to try to make this project open to NOVAC imagers of any skill level, so we didn't specify a camera rotation, field size, etc - in the end this limited the quality of image data as we expanded the FOV beyond the overlapped are of the nebula itself. Also, the amount of data we generated was decidedly unfriendly to many people.  Lessons learned!