The problem you are going to run into in asking about options is that you are only going to get valid feedback from a very small group of users who understand all the options. I certainly don't understand them. And that's a really, really difficult issue to solve. For reference, I manage a team of over 100 engineers doing software and systems engineering, and I am faced with the exact same issue every day. How do I take a complex problem and make it usable and understandable for my users who are not aware of the underlying complexity. So I feel your pain.

The key will be getting a set of defaults that works for the majority of the users. That was case prior to stellarsolver, and a month or two down the road we will hopefully be back there.

I do think the missing piece here is guidance. This is a big change, and there is nowhere to go that I am aware of to even begin to understand how to tweak things to improve matters, and from that provide feedback. What is All Stars? What is Parallel Small Scale & Parallel Large Scale? I don't expect you to answer that here BTW, it's somewhat rhetorical.

If there was a wiki that described the changes and describe the impact of the changes along with some guidance on how to debug and experiment, I think you may have gotten better feedback.

Overall, this is a good change, and I see the validity of it from an architectural perspective. I will continue to play with the nightlies (although the weather does not look like it will cooperate much over the next week or so)

J.

Yes, and they are not working for me on my SCT. The classic SCT donuts were throwing it way off, and even close to focus I was not picking up many stars. Reverted to an earlier build and all is now well and getting data around 0.7"/px. with guiding RMS at 0.79". SEP Multistar autoguiding is fantastic.

One of the other profiles may have worked better, but this change was unexpected and there seems to be little background reading to help figure out what to do. Before using a nightly I usually do some testing with the simulators and sometimes my actual mount along with the camera simulator. The two things you can't really test this way are autofocusing and guiding because there are so many optical and mechanical dependencies.

J.

In all fairness, I have never had much luck with autofocusing on a Reflector. The donuts are making things really difficult for the star recognition algorithms. SEP never worked for me there, I always used Centroid.
Have you tested Centroid with the new Stellar Solver? Also, as I recall, exposure time makes a lot of difference. Less is sometimes more. If your pixels are saturated, HFR will not be accurate.

Rob, I would have loved to test earlier in the year, but I did not know about it. Sorry, I am always up to playing around a bit.

I think I found the sweet spot for my telescope now. I love the speed at which it solves now!

Clear skies,

Jo

"What is All Stars? What is Parallel Small Scale & Parallel Large Scale?"

These are my best shot at starting to make some options profiles that might be good. They are based on some tweaks and settings that I made. Maybe they are good, maybe not. They probably will change in the future as we get good profiles. I don't claim they are perfect yet, this is definitely where I need help.

"I love the speed at which it solves now!"
Yes, My main focus was on developing a better plate solver, a system that would be easy to set up and use, where the plate solves are much faster, optimized, and work on all systems the same way. The old system was a real pain to set up on anything except Linux and I was constantly trying to fix issues for users. StellarSolver works on all 3 operating systems out of the box, if you try out the tester, but the integration into KStars has been a little challenging, and then it broke when we had to make a whole bunch of changes when we integrated it. But now I think I mostly have them fixed. Getting close to having it work nicely again.

"Have you tested Centroid with the new Stellar Solver?"
StellarSolver will only affect SEP, not any of the other focus or guider algorithms.

I have tested it with the simulator and a jpg images I had on hand. At first the simulator images around the NCP weren't solving and it crashed with Load all indexes to memory. It was able to blind solve a jpg frame in EKOS Load and Slew function in a reasonable amount of time. Subsequent simulator frames solved very fast
Something still missing is the ability to load a raw image for Load and Slew. Is there a reason it is not available when KStars is capable of converting Raw to FITS?
If it isn't too much trouble, perhaps now is the time to work it in too?
It's looking very promising Thank you.
Raspberry Pi4, 4gb. Raspbian Buster 18.04.

I have to applaud the developers here ... thank-you for taking on the initiative to make plate solving
a cross-platform improvement and, at least from my prelim tests on linux, a significant speed boost.
I have two pi boxes: one SM OS I don't mess with and one for compile and test. How can I help ?

This is a reproducible error:

SEP | StellarSolver | 4-ParallelSmallScale
I unchecked 'Auto DownSample' just to see that I am still on a working profile, but:

I try to solve M57 and it fails.

I try another one (M65), and get an eyefull:



Another run, same result:

So it is expected that new features can go the nightly. It's 3.5.0 Beta. The only real way to find if the new SEP profiles really work is to test them on all possible configurations. Btw, this is exactly the SAME SEP as it was before in the stable build. The difference is that the previous SEP "options" were hard-coded in KStars, and now they're configurable. A friend of mine actually has issues with focus as well, and today I'm going to monitor remotely the autofocus operation on his RC to fine-tune any parameters and perhaps create a "reflector" SEP profile.

Can you please share a link to the image so I can try to reproduce this crash?

You've got mail!

Jasem, I don't know whether this is important or not: the M65 file has no CFA pattern because it was (as far as I remember) acquired as a colour pattern in a fits file (?).

Right I can't figure out what's going now here, I can see the crash when it downsamples. Sending it to Robert to check out.