This is a feature that I would love to have in Ekos Polar alignment assistant. Once its taken the 3 images and you have selected a star, it would be good if there is some way for Ekos to highlight the star, so when it gets moved about you know exactly which star you were trying to get into the red circle.
Not only do I agree, but I've already implemented this and it is in v3.5.2. Are you two using that?
It is opt-in, though, so perhaps you haven't enabled it.
If you look in the screenshot below, you can see the align star circled (it's currently just starting, on the yellow/purple vertex of the triangle).
To enable that feature, "Update PA Error" must be checked. Then, once you click "Refresh", the star will be circled.
The reason it's an option instead of always on, is that in order for the system to be able to find the star, I needed to implement a guiding algorithm (I actually used a lot of the internals of SEP MultiStar), which involves star-detection, and all that takes compute time, especially when refreshing the image every second on a full-sized image. Therefore, on machines like Raspberry Pi4s or slower, I recommend binning when polar aligning at least 2x if not 3x3 to speed things up. In my experience that does not degrade polar alignment performance, and makes the refresh cycle much faster.
I've been testing the "Update PA Error" and the behavior is at least "intermittently": the circle and the PA new estimation sometimes it shows and at the next frame disappear. My equipment is an ASI120MC at 1x1 binning running on a RP4 (Astroberry).
Reading your message it seems that I should activate at least binning 2x2. But to use this binning, should I change the focal length of the guidescope or the alignment algorithm is able to get this binning information directly?
I've never used binning successfully because it always interfere with my astrometric reliability, so, some general advices will be more that welcome
You don't need to adjust focal length, the system takes the binning into account when performing its calculations.
With respect to binning, since you likely have a pixel size on the order of an arc-second, and the noise in polar alignment is much much larger than that (I imagine the polar-align error bars are on the order of 1/2 or 1/4 arc-minute), I don't think binning 2x2 or even 4x4 would affect you. I have no hard science to back that up, but, I did experiment with that a bit on my own setup. Of course, you can also try and measure with several binning amounts (without correcting) and repeat several times and see (and let us know).
The speedup with binning was dramatic for me. Using an RPi4 with an ASI 1600 mm camera (~20M pixels), when binning 1x1 the polar align refresh cycle took ~5s even though it was set on 1s refresh, since it doesn't start the next cycle until it's finished with the current one. That's pretty slow feedback. Changing binning improved things: 2x2 --> 2.8s, 3x3 --> 2.3s, 4x4 --> 2.1s. By 4x4, other things (e.g. capture and download times) are dominating the cycle time. Here's a pointer to the merge request for this feature where I described some of that: invent.kde.org/education/kstars/-/merge_requests/209