Tuesday, January 22, 2013

Working on autofocusing with @Focus2

I posted on the Bisque forum and asked about my problems with autofocusing. Apart from the manual, someone pointed me to a good post - explaining it in more details.

Trying to determine a better step size:
  1. Critical focus zone
    With the Lepus reducer, I have a focal ratio of 6.2. That gives a critical focus zone of 6.2*6.2*2.2 = 84.568 microns
  2. Focuser Travel
    I measured how much the Robofocus focuser moves the focus tube: 9000 steps moved it ~7mm. That means I have a movement of about 0,778 microns per step.
The Range paremeter for @Focus2 should be about 40 times the size of the critical focus zone: 3382.72 microns ( = 84.568 * 40). That's equivalent to 4349 steps (3382.72 / 0.778). That's quite different from the default value of 1000. Though, I wonder what this will do to the focusing sequence. When I ran it with the smaller default value, the star blew up quite a lot...

I also increased the number of averages that @Focus2 takes to 5 - yes, that will take longer, but in my light polluted backyard, that's probably a better choice. At least for the beginning, I will also increase the Samples to 10 (5 on each size). Then I can analyze the logs data better. Once I have this working, I will try to decrease both values again to make this faster.

Finally, I have to set the star magnitude that I want to use and the exposure time. I'll try on the brighter side (mag 4). For a star of this magnitude I'll need ~0.2 secs exposure time.

But now I'm running into a different issue. If I set the range to 4500, SkyX tries to move from the initial focuser position to 2250 steps inward... and times out. I guess, Robofocus doesn't give any updates during the movement and then takes too long :-(

After lots of tries and error, I found out that moving Robofocus by 1400 steps still works (higher values timeout again). So, I had to use that x2 (=2800) as the range parameter.

With this setup, @Focus2 ran without complaining - good!


These are the first 2 v-curves:

Neither looking very good. But seeing conditions were pretty bad tonight: bright moon, light pollution and one of our neighbours had his fireplace going.

So, I inserted the Ha narrowband filter and tried again:







Looks better. But I noticed that the subframe was too small to show the defocused star. So, I turned off subframing, and now I got:
Now, this looks like a good v-curve. And focus was perfect!!!Now, if I could only figure out how to use the full range - that would make the fitted curve even more accurate. And then I can start optimizing the average and samples parameter and fit the subframe to speed the process up.