Autofocus packs for 2.5.1.0 - didn't do too bad for me


#1

I tried it out a couple of days ago but it was a total bust and I didn’t save the packs. I couldn’t get a good curve and smart focus kept moving outwards.

Tonight I tried again but increased my AF exposure time to 10 sec (from 5). All focusing is done with the Lum filter. Scope is a TPO 10" f/4 Newt with Feather Touch focuser. Camera is an Atik 383L+. I got better curves and saved some packs for you:

Dropbox AF Packs

What I did was move out 5 steps from my normal (25) for the first pack and then increased 5 more for pack 2 and again for pack 3. The curve for pack 2 seemed best so packs 4 and 5 are done with the same step size as pack 2. Curves are saved in each folder but I lost the curve for pack 5.

What I notice is that at the outer edges of the curve, there are only 3 or 4 stars detected when I can see a lot of them. When it is close to focus it detects a lot of stars. I was also not getting rings until the farthest inward movement. On the outward side, the stars were more blobs.

I’m letting it run for now to see how it does through the night. Last night on the release version was a bust because it lost focus somewhere in the night and everything after was out of focus. This time of year is a bit of strain for autofocus because my temps are dropping a full degree C every 15-30 minutes until about midnight before they stabilize a bit.

Hope this helps,
Chris


#2

Well, it failed on me as I finished that post. It lost it and started moving the focuser out. The stars are now large rings and it found a new “good” focus much farther out. I don’t have the packs for those, sorry. I did include some screen shots in the dropbox folder (image07 and image08) which are sequential in the focus run.

Chris


#3

Thanks for the report.

With the new routine I don’t feel that this matters much (since the probability of selecting false stars and unfairly weighting the mean is drastically decreased)… 3 or 4 real stars should be just fine.

With the AFPacks you provided 2.5.1.1 seems to pretty well with them, BUT I had to increase the minimum diameter to 5, 6 or 7 pixels… all seemed to provide similar results… Good selection of donut stars (in general):

Doing this provides pretty reliable results across all your AFPacks. We will work on better auto selection of minimum star size… right now it is a manual setting. Here is AFPack3…


#4

my timing could not be worse - just finished a bunch of tests on 2.5.1.0 which are now obsolete. i found that at the outside edges of the focus range that SGP was ignoring all the donuts and picking hot pixels. i assume setting the minimum star size to 3-4 pixels will take care of that problem?


www.mainsequencesoftware.com