Yet Another Focus Question/Problem - Need Help

Sequence Generator Auto Focus
1

So, after many arguments with my auto-focuser and subsequent make-ups, I think we have come to some agreement of collaboration. With my EdgeHD 8 + 0.7x FR, I am now able to get a relatively nice V-curve consistently with every auto-focus run in SGP, yay! My focuser is attached to my focus knob directly with a belt-driven Pegasus Astro Dual Motor Focus Controller, and my focus values for all 7 filters are somewhere in the range of 166,000 to 168,000. My focus routine runs at 1x1 binning (it was 2x2, but recently changed it to 1x1 in an attempt to solve the main problem I’m about to describe) with 12s focus frames for broadband filters, 20s for Ha and SII, and 15s for OIII. I have my auto-focuser set to refocus on every 1 degree change, every filter change, and every 30 minutes (this last one I added to my workflow most recently). My auto-focus routine uses a step size of 500 over 11 data points, and I have my focuser backlash compensation set to “IN” 500 in SGP.

Those are all the givens for context. Now to my problem (let’s just focus on any one filter…pun intended)…

When I start the evening session for one filter, say Red, I focus with my Bahtinov to set a nice sharp focus. Things look pretty good, though I do have the minor issue of my HFR values swinging by 0.5-1.0 between subs (can this much swing be due to seeing alone). The following image shows my image history over 30 minutes for the same filter, temp only 0.25C difference, same focal position, with the HFR values fluctuating between 5.42 low to a high of 6.56. NOTE: The red box highlights the frames in the table above:

image

But, that is not my main problem. My main problem is that when the auto-focuser runs and I get a nice V curve, everything seems to go swimmingly and SGP calculates what seems to be a nicely calculated focal position. BUT, it’s usually about 1000, sometimes 2000, focal position units or so off from what is visually a focused frame. I had been focusing using 2x2 binning, to make my exposure length a little shorter (8s for LRGB), but I thought maybe using 2x2 binning for focus when I’m actually imaging at 1x1 is the problem. Today, I’m focusing using 1x1 binning, hence the slightly longer 12s exposure for LRGB, and I’m still getting the same issue. When I go to sleep for the night and let auto-focus do it’s thing, I find most of the unattended frames are blurry. Here is an example of what I mean.

Autofocus was just triggered because 30 mins were up…

image

I let it take a couple of frames to check that it indeed autoblurred instead of autofocused, and then go back in and visually refocus using frame and focus…

image

Here is a frame after it autofocused…

image

…and a frame after I went in and visually reset the focal position…

image

So, regarding my main problem, any ideas what may be going on? Any suggestions of what I should try or do differently? Basically, I have an auto-focus setup that does not focus :confused: I don’t understand how a nicely generated V curve would be so off over 11 data points? For those of you using SGP, are you using HFR for your auto focus metric, or is FWHM via PinPoint the way to go?

Bonus points if you try to tackle my “minor issue” as well. I know we have pretty bad jet streams here in southern Maine, is the HFR fluctuation due to the effect of that on seeing?

The fate of my relationship with my auto-focuser is in your hands…

-Ara

NOTE: Some feedback I have received already is to up my backlash to 1500, change my step size to 625, and reduce my data points back to 9. I have not tried this yet due to NEAF and now a New England storm :frowning:

2

Hey Ara,

As suggested on the CN thread, the “flat ear” on the right side of your v-curve indicates that you still have a backlash issue (i.e. backlash compensation is likely not big enough).

A cleaner v-curve will move your auto-focus point to the right (based on the image you posted) , which hopefully better matches your optimum focus point.

Steve

3

Ara, I agree with Steve on this one… backlash seems the reason you don’t get the initial movement of the focuser that you expected and end up with that fish hook patten rather than a V, so you get a poor line fit on the right side of the graph. I also had a bit of a struggle getting my focus boss autofocuser working right in SGP, and my problem was too small a step size (and probably a bit of backlash). Once I increased my step size my curve shapes became very good and my plots became a sharp V. I also had the fish hook. While I don’t know what your actual step sizes are (in um or mm) mine were less than 1um per step. Also a larger step size will somewhat wash out the effect of backlash (in my simple thinking).

A nice feature on the autofocus page would be to offer an auto backlash set. So maybe the first time you use an autofocuser, the focuser to move out, or in, some large number of steps, change direction and move inward in smaller steps (to be defined based on step size) until it sees a difference in FWHM (knowing backlash has been taken up) and then store this backlash number in the profile. It could then reset or rechecked as the user wants. Come to think of it, it would also be nice to have something that calculated a good autofocus step size as well. Almost like an interrogation of your system so it determines optimal step size and backlash.

Terri

4

Thanks, Steve & Terri! Since I posted this on CN, it’s been stormy here in Maine except the days when I was attending NEAF, so haven’t tried the new configuration settings yet. I posted here to see if I would get more/different thoughts on the matter, because when I spoke to the Pegasus Astro guys at NEAF (they are fantastic, by the way), they were surprised I was having to consider a backlash as high as 1500, even for a SCT. Also, when I spoke to Jared at the booth (or was it Ken), he seemed to wonder if it was backlash as well, and I mentioned that I would post on this forum to see if any SGP specific recommendations came about.

Anyhoo, thanks again for taking the time to respond. I look forward to trying out the new settings, with high hopes I can rely on my autofocusing to work <fingers_crossed>!!

-Ara