cz550 Example: the curve goes down to 3.2 and it ends the AF at 3.5...
Why ?
The curve is just a general 2nd order (conic section) least squares estimate from the actual data points. So the minimum could be anything, since the 2nd order curves do not model either Geometric Optics, nor the diffraction effects of the optics near focus. (Geometric Optics calls for constraining the 2nd order curves to the special case of hyperbolas to model the focus -- hyperbolas create V-shaped curves, and not the U-shaped curves that you see in ASIAIR).
https://en.wikipedia.org/wiki/Linear_least_squares
Another reason is that ZWO does not use a fixed star to estimate the HFD. It uses an average of multiple stars, and it uses different stars at different focal positions. It should have kept the same stars, but change the exposure value so they do not saturate, but ASIAIR uses a constant exposure duration and camera gain, so it has to choose different stars to avoid using any star that is saturated (saturated stars will not give a correct HFD estimate).
So, multiple errors by ZWO, as usual.
When used properly, linear least squares is wonderful. Gauss invented it, just so that he can rediscover the position of the minor planet Ceres after it was lost after discovery.
https://math.berkeley.edu/~mgu/MA221/Ceres_Presentation.pdf
ZWO simply used the wrong tools with the wrong data. Garbage-in-garbage-out.
Chen
