Ron,
I designed Linear to help with focusers with backlash (probably most focusers have that). It isn't as fast as polynomial, but is more tolerant to backlash.
The key parameter for you is probably step size, which is totally dependent on the motorized focus hardware you're using.
I'm attaching below the parameters I use for my refractor, but, really they depend on the fact that I'm using a f/5.6 105mm refractor with a moonlight v2 focuser.
The step size parameter is important to get in the right ballpark, but don't sweat the difference between 20 and 18, or something of that nature.
The other parameter is important is tolerance. For me, too tight a tolerance can be problematic, as seeing or whatever can introduce noise into the focus measurements. I use 10%, you can experiment with 5% too, but I'd stay away from very tight tolerances.
I like full-field with an annulus to reduce the HFR noise, but I don't bother with multiple samples.
I use SEP for the star detection. If you don't see reasonable star-detection, you need to play with the SEP profile.
Bottom line, to find a decent step-size parameter, set your system to pretty well focused, then start the Linear algorithm. It will move outside focus and start stepping inward. You want to see a reasonable sampling size. Perhaps each sample gives a small improvement or degradation as it moves near the optimal focus. You don't want to see large jumps near the focus point (too large a step size), and you don't want to see several samples with the same focus (too small a step size). You will see a second inward pass after the system establishes a v-curve where your original step size is halved, and finally some fine-tuning for the final focus at the end once the solution is pending.
Hy