I have acquired the ZWO EAF recently and learning to use it with my SW130PDS (130/650mm) and ASI533 camera. I dont have an observatory so I setup and dismantle every night.
I have managed to get it working but I realised that as soon as the stars start showing as even slight donuts, Ekos is unable recognise the stars. And this means that everytime I start I have to manually adjust until stars are no longer donuts.
Is there a parameter that I can set to improve this recognition?

I use in the Focus tab in the tab "Process" for "detection" "Gradient" instead of SEP.

The other thing is to start Auto Focus with a fairly good focused star (which means to be not too far out of focus at the beginning) and dont go too far away from the focus point (to be set with "out step multiple" and as far as I remember with "initial step size" in the "Mechanics" tab). This prevents to get "donuts" which are difficult to handle for the algorithm. I try to go out that the HFR is about 2x or 3x the HFR at focus point.

Thanks. Will give it a shot with gradients option. I have been using SEP so far, as I thought that was the superior option. Also the context help says it will use SEP for initial or full field analysis.
How did you work out the step size for 2x/3x HFR ?

The first time I looked at the steps of the focuser at good focus the I go out (or in) with the focuser and looking at the HFR of one star (using the "loop framing" button) until it is about 2x or 3x the focused size. the difference of the step numbers I divide by 5 and use this value as "initial step size" and use 5 for "out multiple steps". You can also use www.goldastro.com/goldfocus/ncfz.php for the critical focus zone to use as "initial step size", but I think it doesnt matter which way you go.

I'm using Linear1pass. It workes quite well, but sometimes with my dual band filter (l-eXtreme) I have problems, maybe due to the fainter stars...

You can also play with the "initial step size" and "out multiple steps", because it depends a little on seeing conditions (at least my experience with high focal length of >1000mm)

I had a try using Gradient but thats struggling so opted for SEP with Polynomial.
BTW is there a way to reverse the way it counts the steps? As I swapped the focuser around it thinks 0 is when its fully extended.