2.5.1 AF some (very) good runs, some bad


Hello guys,

I spend some time last night to test this new AF routine, using my SCT scope Celestron C8 + .63 reducer (roughly 1280mm f.l. and 64mm obstruction vs. 203mm aperture) and QSI 683 camera (5.4um pixel size) and Moonlite focuser (from 0 to 5900 steps)
On fields with a lot of stars, it does work quite well. Actually much better and further out of focus that it use to do with 2.5.0. unfortunately it is not perfect on the ‘wings’ of my Vcurves.
On fields with fewer stars, it performs poorly, apparently not recognizing stars. I even tried to push the frames to 15s, with better luck, but still it eventually went awry.

A lot of AF pack in my dropbox (one big 400Mo file, but I can break it down by packs).

I increased significantely the step size so I could test the doughnuts! ‘Disable Smart Focus’ was unticked.
From pack 157 to 173 it was manual runs. I apologize I did not save the Vcurves… A lot went well, but often the ‘wing’ points are bad. It does not recognize stars that are obvious on the field.
From 174 to 180 these were from a run (AF every 300s frame) but it did not save the pack after a few frames because the AF failed, as it wanted to go outside the 5900 steps for my focuser. (I am actually surprised it did not save any frame at all??)The focus went from the 2900 range for the first few frames to the 5400 range afterwards!

An example of a ‘doughnut’ field but stars not all recognized?

NB 1. I always trim my field by 10% for star recognition, but it seems the new routine does not take this and measures stars outside of the box.
NB 2. It would be great to have a tool that could re-run the AF pack through a different set of parameters (# of stars, nebula filtering, etc…) and produce the Vcurve. (maybe you already have that :wink: )




Thanks for the feedback. Are you, by chance, able to identify AFPacks that did not work well (ones with “wings”)?


I knew you would ask this :wink:
I managed to recreated this one (in excel) from the log file
(log file: https://www.dropbox.com/s/ilafz1wbxv3hma6/sg_logfile_20160403182000.zip?dl=0 )
This is from the pack 159 in the previous file
(also here with only this pack :wink: https://www.dropbox.com/s/pxcnwy2d7egzdsl/AFPack_159.zip?dl=0 )

A good (tiny) request feature: save the V-curve in the AF pack :smiley:


This is an example of a AF failure. This should be the right part of the Vcurve, but the smart focus wanted to go further right, which is out of range of my focuser.
This is the AFpack167: https://www.dropbox.com/s/efjo0f3htopoegs/AFPack_167.zip?dl=0



So the good news here is that simply providing the “minimum star size” adjustment to the user, I was able to produce this curve, bumping the minimum pixel size up to 5 (from 4):

I will take a look at 167 tomorrow… good progress so far I think. Now I need to see if calculating the minimum star size automatically is viable (or at least suggesting it).


Hi Ken,
I last used the AF routine in October of 2015 (v2.4.3.8), and was satisfied with the autofocus perfomance at the time. I’ve just now recommissioned my 12"LX200 (using SGPro v2.5.1.0) with the same optical setup as last year, but so far no luck with the AF.
FWIW, my problems mirror closely what Oguerrea has described. But I’m going to give it one more night of fiddling, before sending in the files for your review.

A question regarding your reference to minimum pixel size; is this adjustment available on for me to try, or is this internal and automatic?

Thanks, and keep up the fantastic work,


This is not available yet… It will be in the beta


This is a good idea. will also have a JPG screenshot of the graph in order to help with debugging and general improvement of the new AF routine.


I ran many packages the other night, and some got a bad jpg image (screenshot?) of the graph, 2 out of 33 runs.

an example jpg (timestamp says 11:07pm, which is slightly later than the log?) for AFpack187 and I have a very short 5s close time for the autofocus dialog box:

log file:sg_logfile_20160423215916.zip (153.5 KB)