PeteC When I run the af routine it sometimes gets a v curve and finds the point of focus, but it’s always wrong. Any help appreciated.
Pete, a couple of points to consider if you are refering to the ASIAIR's autofocus procedure.
1) the ASIAIR requires the drawtube to move sufficiently so that the star size can defocus all the way to a star HFD of over 10, go through focus, and back up to over 10 again.
This typically requires a drawtube travel of around 1mm (depends on the camera's plate scale and f-number of the telescope).
If the "81" and "85" that you mentioned are in units of millimeters, then you are covered. If the units are EAF steps, then you probably don't have enough range, since most focusers run between 0.5 µm and 4 µm per EAF step.
2) the autofocus procedure in ASIAIR is flawed.
The proper procedure is to measure the HFD at certain points and then try to estimate the focus by looking for that point with geometric optics (i.e., try to ignore diffraction as much as possible). In geometric optics, the curve would be a hyperbola. An example of using a hyperbolic curve fit is this (a program that I wrote for macOS):
Recall that a hyperbola that is a cross section of a cone that is cut parallel to to cone's axis, is a perfect V that is made up of two straight lines, thus modeling the size of a star with geometric optics.
In real life (diffraction optics), the star will never come to a point, as seen in red spots in the diagram above.
But the idea is that even with noisy samples, you have sufficient samples so that when you take a minimum mean square estimate using all the points, you are reducing the variance of the error.
(ASIAIR does not model with a hyperbola but with a parabola, and I've repeatedly told them that a parabola does not model anything in optics, neither geometric optics nor diffraction optics, but they wouldn't listen.)
Now, that is not the major flaw. The major flaw in the ASIAIR procedure is this: after getting a perfectly good curve, ASIAIR goes ahead and ignore the minima of the curve, and proceeds to use a second pass where it looks for the moment when a data point has increased in HFD.
Just watch the ASIAIR autofocus procedure when it is doing a second pass. So, all that work of collecting all the data to reduce the varaince is all thrown away (yes, I kept telling them that too). Instead, they stop when a single HFD data point has increased in size. This is completely prone to noise once you enter the critical focus zone (CFZ).
So, what you will see is that the ASIAIR will often stop once it hits the CFZ (ie., barely in focus). I have almost never seen it reach the minima of their parabola.
OK, now comes the coup de grâce... depending on whether the EAF values shrinks when the drawtube shrinks with colder temperature, once the night gets colder, the OTA will fall immediately outside the CFZ since it started at the edge of the CFZ to start with!
You can mitigate this problem by setting the EAF "reverse" state so that as temperature drops, the actual focus point would move towards the EAF setting that the ASIAIR has chosen. But they don't tell you that. So, it is 50-50 if your OTA will keep focus after a 0.5ºC of temperature change.
Anyway, knowing this (using that "reverse" switch) might help somewhat with your focus problem.
The real solution is to stop using the ASIAIR "autofocus" procedure and use some other method, such as the old reliable Bahtinov mask, that is sensitive enough that it detects focus shifts indide the CFZ.
When it is not freezing outside, I would use the Bahtinov mask. When it is too cold for my liking, and if I am using ASIAIR, I would do my own V-curve sweep (above program) taking star sizes from the Star Detect tool of ASIAIR.
Chen
