In watching auto focus runs, I see instances where the second frame ends up with an HFR calculation that seems inconsistent with the previous frame; that is, the second frame produces a larger HFR calculation even though I know for sure that the second frame is better focused than the first. In some extreme cases, the HFR calculations produce an almost horizontal line across all nine frames.
When an auto focus frame is processed, I assume an algorithm is used to select a batch of stars from the frame and those stars are analyzed to produce an average HFR for the frame. When the next auto focus frame is taken the same routine is followed. However, it seems to me that the second frame is likely to select a different set of stars to be analyzed. My concern is that the second HFR calculation will basically be doing an apples to oranges comparison and the two HFR values will be based on different stars and will not reflect the true change in focus.
If the first auto focus frame not only selected the batch of stars to be analyzed but also recorded their positions (x,y coordinates), then the second and subsequent frames could be guaranteed to do their HFR calculations on the exact same set of stars. It seems like that could produce a more consistent and accurate HFR calculation across all the frames.