SGP 2.5.1.0 had problems in field with few stars

Sequence Generator Auto Focus
1

Tried the new 2.5.1.0 beta last night. Had significant problems with a field with few stars (2000mm f.l. galaxy shot). Here’s the field (note: I was collecting some Ha frames to highlight Ha regions in the galaxies):

Pasted image872×719 405 KB

Here’s the focus view. Lot’s of non-stars with Hfr of 0.0 selected:

Pasted image839×626 454 KB

Adjustment sliders didn’t seem to help (disabled?). Had to roll back to 2.5.0.23 to finish my run. Clearly not enough stars, but it blew up pretty bad in choosing non-stars for focusing. By the way, this is a 190mm Mak-Newt and I didn’t read the message about unclicking “Disable smart focus” so this box was still checked when this run took place.

I think this is the right AF Pack: Dropbox - File Deleted
…Keith

2

That is just a display issue for debug, HFR of 0 is never used in the mean calculation.

You might want to take a read through the post I made for this beta:

I added a temporary adjustment to allow users to set the size of the smallest viable star (in pixels). As you noted, this run is pretty data starved, but I was able to get it to perform quite a bit better by setting the minimum star size to a 6 pixel diameter.

3

Here is an autofocus run on my RC12" + SBIG8300.
I extended the usual 80 unit step size which keeps me out of donuts to 150 which has donuts on the ends.
As you can see from the pic of the screen, it does not recognize my donuts. They are not completely round which may be the problem. Best focus is around 18037, which is where I started the run.
All the HFRs I see in this image are good stars. Why would it compute HFR=0?

Autofocus_RC12_8300M_A.JPG962×731 190 KB

4

@jmacon

Having trouble with zipx. Not sure exactly what it is… 7-zip “kind of” knows how to interpret it, but it still hangs on this:

Any chance you can compress with standard methods?

5

OK… Actually, I can pull the FITS files out of the zipx… just had to do them without the jpg or 7-zip hangs.

By Adjusting the minimum star star size to something a bit higher than what 2.5.1 shipped with, I was able to get this from your data:

Looks like we will likely need to permanently add the minimum star size lever to the release… unless I can figure a way to auto calculate it. It is not so simple as determining scale of your images… the minimum star size seems to be having an effect on being able to detect “less than circular objects”. You are right… the new AF is a bit sensitive to circular deformities…

6

First trys I got perfect V curves 7 out 8 times.
The eighth was ok too. The point of the V looked a little funny on that one.
The doughnut stage at ~3 - 5 HFR performed very well.

7

Is this in the 2.5.1.0 beta? If not, will you release another beta with it included?

I’ll be trying the new routine tonight.

Thanks.

Tim

8

I am very impressed at the first draft. It did not fail. I am using a astrograph with 50% obstruction.
If you need the logs for the successful tests let me know.

Max

9

The Canon 6D on sv102 refractor gave good V curves for each hourly autofocus except the one at midnight which is attached. Went progressively out of focus in the same direction. This may be the same thing that has happened several times in recent weeks with this same hardware combo.
Star poor region so it likely is the same issue as above finding small non-stars.