There is a new HFR routine for single stars that should be able to detect stars that were not properly detected by the previous method (e.g. donuts). I only gave this a test today, and it "seems" to be OK, but please test and check if the autofocus solution gets the correct critical focus point or not. Should be in PPA tomorrow. If it breaks things, I'll just disable and we'll stick to the usual method.
So when I wrote that post last night, I was extremely sleepy and fell asleep several times on the keyboard! So now that it's morning here, I'd like to know if it reaches a better focusing position or same or worse than before. Now you can measure HFR during framing if a star was selected. So you can manually try to find the minimum HFR then try to check if the autofocus will get a similar result.
I have just tested the focus routine with my Canon 550D, it detects stars just fine and it gets the focus much better than before. I got 1.70 manually (10 steps with 0.60% tolerance) and switching to auto it gave me 1.65 so It's pretty close. Just had to be careful with square size, the best results where with box size 32. Although it's not the best night for testing but it just gave me better results. I hope it's better for others too.
I will continue the tests the rest of the night and report later if any changes occur.
The following user(s) said Thank You: Jasem Mutlaq
The new routine is still problematic under real tests today so I reverted back to old algorithm. When I resolve the issues, will let you guys know so you can test.
Well, I tested it today and it's "passable" but not as consistent as the older algorithm. At any rate, give it guys and go and let me know how it goes.