Link here . It’s from the RAS… is there a function like this in Ekos? Seems like a winner since it doesn’t rely on star detection. I guess they’ve been using it in APT for the last year. 

 

I don't use APT, so I haven't tried out the new algorithm. But I skimmed through the paper, and it does look quite interesting. They have a Github with code here: github.com/bschulz81/hyperbolicfitdll

There's a CN thread about it, much about publishing, but some talk about their algorithm. 
CN thread

This is indeed interesting and was previously brought up by Bob Majewski IIRC. We were missing open source code, but since now we have one, it would be interesting to try out. Bob wanted to use it for guiding so that we can detect shifts in the image overall irrespective of star detection.

The paper and CN comments are interesting. Some cautionary thoughts:

Having a new focus algorithm in Ekos wouldn't necessarily hurt anything, but it may not help much either. The old saying "out of the pan, into the fire" comes to mind.
The paper acknowledges some new algorithm weaknesses, and these are different pains, but on par with existing algorithm weaknesses.
1. Any saturated stars in the image will cause an inverted fitting problem. Until this has been addressed, fast imaging systems (e.g < f/4?) will likely be particularly sensitive to the issue.
2. A quote from the author (CN post): "Starting from nothing is a problem for any algorithm. IMO, a rough focus position is needed to get autofocusing started". This is the same problem ekos currently has (all algorithms), and one of the primary reasons why temperature and altitude focus compensation was implemented (although never integrated).
3. Improperly set step size is a problem. The authors suggest 1/2 CFZ for a step size. I've previously stated that CFZ size is critical (the setting should have been automated). I concur with the 1/2 CFZ size suggestion.
4. The number of images needed is stated from 8-17. That probably excludes any use for guiding assistance (only noted because Jasem raised that question).

In the end, star detection becomes non-problematic, but avoiding star saturation becomes problematic. The same "avoid donuts" starting condition problem prevails, and setting step_size appropriately is equally important.

IMO, integration of temperature based focus compensation (even if using the current config file vs unimplemented GUI) would answer the donuts problem while seriously reducing time consuming focus runs (as focus is updated between exposures...thus tracking longer before another focus run is needed).

See: indilib.org/forum/ekos/7626-temperature-...us-compensation.html

CS... Doug