I'm relatively new to EKOS, and brand new to the focusing module, so forgive me if this is a stupid question ..
Basically I just started trying to use focus in EKOS, and everything looks good on the first pass (a nice curve is generated) but the issue arises when it tries to return to the starting focus point to get back to that sweet spot. I decreased step size down to 5, and that looks quite nice, but after the algorithm determines it's time to move back to the starting position, I will see something like focusing outward 75 steps, inward 50 steps. However, I don't believe that it's actually now focuses at the point where it thinks. Rather, it doesn't even make it back to the other side of the curve, and focus just keeps getting worse.
I suspect it has something to do with either the focus-by-wire of the Olympus and possibly the inertial focusing that it also has when using MF. (I'm ignorant when it comes to how these focusing steps/pulses are being sent to the camera).
So the big question is ... is there a way to get the algorithm to just reverse itself and settle back into the peak focus of the curve rather than attempting to return to the starting focus point? I've run out of time to test with the other algorithms, but they seemed to behave in a similar fashion. Something that just steps a discrete amount inward to generate the curve, and then that same amount outward to return to prime focus seems like it would work.
In the attached (and ever so beautiful) screenshot, you can see shots 1-8 look good ... but when it *thinks* it's returned to around the starting focus position, it's not quite there and focus just gets worse and worse ...
I've read all the docs I could find but it feels like each algorithm still suffers the same fate with my Olympus body+lens.
My hypothesis is that moving 5 steps inward 10 times does NOT equate to the same actual focusing distance as moving 50 steps outward 1 time on my body+lens. This *may* have to do with the fact that on Olympus bodies+lenses, when you engage manual focus it is actually focus by wire and it is not linear. I.e., if you turn the focus ring quickly, it will change a great deal of focus. But if you turn the focus ring very slowly but the same distance as when you turned it quickly, it will actually change the focus a great deal less. (Much like the inertial scrolling on touch devices).
Now I presume the reason for moving inward in small increments and then moving back out in one large increment is to minimize backlash on the focusing mechanism, but for me it's not entirely accurate. So possible solutions (at least in my eyes) would be to 1) lock the step size and only move in small steps in one direction (backlash be damned, of course) or 2) have some sort of user configuration option to define how far to move outward again when trying to find the sweet spot on the V-curve.
I hope this all makes sense ... I just don't know if this is an improvement that hasn't been considered because I'm guessing not a lot of Olympus users are running EKOS or if there is actually already a solution/workaround that exists, I just haven't found it yet. Like I said, very new to EKOS because it hasn't always played well with Olympus cameras (not EKOS' fault, but rather Olympus ... the libgphoto2 folks are doing their best to support this proprietary camera).
I see. I think if moving 50 steps is unpredictable (backlash aside) then you're going to struggle with autofocus in general, as all the algorithms assume some sort of predictability in the focuser. Maybe somebody else on here with an Olympus may have some tricks.
You can control the backlash movement by setting the backlash amount. When changing direction, the focuser will add this to the previously calculated movement. So you try playing around with this.
That said... what you were initially trying, the Linear algorithm, might be your best hope as it uses heuristics on the second past to try to stop at the best place rather than moving straight to a pre-calculated point.
What you say about predictability, I think, is the key I'm missing with the Olympus ... I've seen complaints about videographers wanting a more predictable MF, because the whole inertial focus vs. a nice linear focus ring makes MF during a video shoot difficult.
But thanks for the suggestions .. when I get a chance (hopefully this evening!) I'm going to experiment some more with Iterative + Max step size (hoping it doesn't just fail if the max step size is too small) and also Linear/Linear 1 Pass with an explicit Backlash value set.
Just an update .. I tried each algorithm with every set of parameters imaginable. Unfortunately, no success. Of note, `Backlash` was never enabled as an option using GPhoto CCD for the focuser so that shot down my theory that maybe setting some backlash value would help.
So question for anyone who might be paying attention to this thread ... is there any way to sniff the gphoto commands that are being sent to the camera? I hope to do some gphoto testing later but I want to try and determine if there's some issue with gphoto and how the Olympus is handling the commands that can be easily fixed, or if perhaps it's worth while trying to write an incredibly dumb focusing algorithm that just moves a fixed amount in each direction in an attempt to get some reproducibility in focus command to actual optical train focus change.