Some maintenance to the camera I use for astrophotography in this month included using epoxy to glue the shake reduction system in place, upgraded with water cooling to gain 20C of temperature drop, and a new standard exposure to limit problems from flexure.
I'm still using the same Pentax K10D camera, the one with the broken shutter which is the reason why the shake reduction system needed to be epoxied in place. Even when "turned off" the SR will "park" the sensor at each exposure. This causes smearing with bright stars.
I've upgraded the cooling method for the camera to use larger peltier devices and put a CPU-type water block to manage the waste heat. With the new system in place, the cold finger going into the camera shows a 20C drop from ambient. Alas, the Pentax K10D has bad amp glow that will always be visible no matter how cold the camera gets (I've put it into the freezer for an hour to check). Cooling the inside of the camera does help improve the thermal noise on the areas of the image that are not affected by amp glow. Power budget for this upgrade is pretty steep - 7 A at 12V.
Lastly, I've been battling flexure with the gear. I've decided to throw in the towel and dropped the exposure duration down to manage the trailing stars. No longer using 1200 second subexposures, now using 600 second. Bumped the ISO setting from 100 to 200 at this change as well. This new calibration method requires a new collection of darks which is pretty easy to collect with the shorter duration.
Here are the results from work done in the back yard over the last few weeks. Worked under moonlit suburban skies using a Baader Ha filter (35nm bandwidth) and IDAS LPS-D1 filter.
Ha data gathered from September 10 & 11 with 52 subexposures of 10 minutes at 200 ISO.
RGB data gathered from September 15 & 16 with 25 subexposures of 10 minutes at 200 ISO.
Total integration time was 12 hours and 50 minutes.
Calibrated each data set separately and then stacked with DSS. Used PixInsight's NBRGB script to merge the two images.
Also on Astrobin
For comparison, here is the Ha data. This was just the Red channel after registration to the color stack, crop, and DBE to manage gradients.