Autofocus needs Data Quality factor


Under good conditions and particularly with refractors, the autofocus routine works great, at least for me. I have had good success with it with my 2 refractors (FL 300mm aluminum tube, and 714mm carbon tube). My RC 8" (FL 1624 mm carbon tube) has been more of a challenge. I have checked the central obstruction box. It is well collimated, nice round stars to the edges, so that is not the problem. It also has a new high quality 2.5" focuser. Both issues previously were a problem, and it took me a long time to sort them out. Slow learner I quess. Not sure what the problem is. May be the longer FL or sub-par tracking. I think from other posts that in general, mirrored scopes are harder for autofocus routines. Please correct me if I am wrong on this, I’m really a newbie to this hobby.
However, regardless of what might be causing my difficulties, under less than good conditions, say light clouds or wind or equipment issues or whatever, I have noticed that when my focus numbers are jumping all around, the focus routine still produces a linear fit line and acts as if it is good, consistent and valid data. In fact, often these data points are not much better than random numbers, and the routine still shifts way to the right and starts over. Now it’s going to get even less reliable data. The end result is that it often stops much farther out than where it started, very much out of focus, and never finds the correct focus.
Since my RC 8" has a carbon fiber tube, (carbon fiber has 1/20 the temperature expansion coefficient of aluminum), my focus position changes very little during the night. Since I always manually set my initial focus, and that rarely changes much from the night before, it is very disheartening to have the autofocus routine put me out of focus. As a result of this I usually turn off autofocus on the RC8.
So I propose the following enhancements to autofocus:
(1) Add a Minimum Data Quality factor. This could be something like the standard deviation of the data points from the linear fit line. This would tell the focus routine to do something different if the data quality is poor, such as try again once or twice, or move to the left, or quit.
(2) A maximum allowable change from the starting focus point. For a given set of hardware, if you have originally started with good focus, there is a pretty set maximum change in focus point that might occur. If the focus routine comes up with a value outside this range, it has clearly made a mistake. In other words, it should have the same motto as the medical profession: do no harm.


Hello @Ken ,
I am bumping up this thread, as this could be a great addition, considering all the efforts on the current AF routine. Is it still a considered feature?

Maybe, at first, a AF quality factor could be just indicative (like running a quick standard deviation calculation), and a good way to debug/enhance the routine in light of the current stream of Betas.

I should be saying that more often, but you guys are doing a great job on SGP overall and the AF routine specifically! Keep it up :slight_smile:

Clear skies.