As a final exercise on the last clear 'summer' night, I wanted to capture a favourite from my teenage years - Vega/Altair/Deneb in the summer triangle. Hoping that I could capture some of the milky way alongside it with long enough exposures. 

Capturing data at 24mm was easy enough: single 30s exposures looked fantastic, though the wide angle kept part of the house and the tree in frame. This turned out to be a bigger pain in post-processing than I thought, as the leaves reflected light causing varying backgrounds for each subexposure, causing me to throw away half of the data where the leaves were too close to Altair (on the right of the image). This was also an excuse to learn the manual stacking process in Siril, and perform per-frame background extraction. The result was still poor and highly colour distorted - the flats showed extreme vignetting even when I took all data stopped down to f/4.5. This caused the corrected lights to be over-compensated and have a james bond-like circular effect, where only the middle looked OK. I ended up not using flats at all, and relying on the background extraction for it. Not a great lens at 24mm.

Stacking (registration) gave strange results initially as well, turned out the leaves were interfering with finding stars. But cropping a sequence pre-stacking messes with the bayering, I learned. Siril plate-solving did not work, even with manually adjusting star finding (dynamic PSF) or selecting a sub-region. I tried solving with astrometry.net and using the solution as starting point for Siril, to no avail. (Apparently plate-solving is more difficult with shorter focal lengths - next time I should try a local astrometry.net solver). Hence, photometric colour calibration was not possible. The dynamic range was mostly gone after the violent background extraction above, so it was hard to extract some colour. I attempted using starnet++ for the first time to help reduce the many stars in this wide field and bring out the milky way a bit more.

It took me multiple processing attempts over the past month to produce a mediocre result. At least I learned quite a bit. And hey, look at that: finally managed to capture North America in its entirety... (bottom left)