This seems to be quite consistent because last fall I was seeing a similar problem. Convergence is OK in the western sky but when targeting the eastern sky (SE), Ekos consistently diverges during "capture and solve" / "slew to target"

The setup is Kstars 3.1.1 with StellarMate 1.3.1 driving HEQ5 Pro

Attached are the plots in a screen capture and logs.

I've noticed the n-star alignment in EQMod doesn't work very well with a lot of sync points (and inevitable backlash) in small area (and does some weird jumps otherwise too) so I've switched it to use nearest point alignment with my HEQ5 and that works very well, so worth a try. The option can be found on the bottom of Align page in EQMod config.

Seems like a bug but I'll give your solution a try. Thanks

This sometimes happens to me as well. For me the solution is:

1) Park the mount
2) Clear the alignment model in the Mount tab
3) Clear all alignment solutions in the Alignment tab
4) Unpark the mount
5) Slew to the object. Pointing will not be accurate
6) Start the alignment procedure using "Slew to target"

So far it then always has centered the target within 30" of the center of the FOV.


HTH, Wouter

Yes, definitely it would be nice to have n-star alignment fixed, but there doesn't seem to be anyone with understanding of the algorithms and time to fix it, fortunately nearest point works quite well.

@wreeven I tried it again on the weekend. This time the converge failed in the west (after finding the first target in the east). It appears you are correct. the Alignment model needs to be deleted to make the next target - I thought I did that but now I see there were two places for the alignment model - deleting just the alignment solutions didn't work.

It seems to me this convergence thing is done wrong. Convergence and Alignment model are two different things in my view. You don't need a model to converge. You just have to move the scope to close the error. Once that's done, then a sync point can be saved for the model. It seems that the alignment is trying to build the model in order to point the scope to the target. This is backwards in my view. You only need that to do goto's. Not to converge.

There is a setting called "use differential slewing" which I thought would do the trick but it doesn't iterate to converge on a target, so I don't know what its doing.

Hi
Just clear the mount pointing model after any flip. HTH.

I see. Thanks for the tip @alacant. However, I still think the convergence should happen without the use of a model. The model should be built from syncing to the points where convergence is successful, which should be every time.

Yes I agree with that.

I agree with pentaxian23. A sync point should only be added once the convergence has settled to within set tolerances.

That would be ideal, but would require some kind of hybrid of differential slew and current default algorithm which works so that it mount is synced after each plate solve and a new slew to target coordinates is made until tolerance is met.

My personal setup is fixed and has only a few (around 10) sync points spread over the sky saved that get loaded by EQMod driver each time, but I don't save the points after session, so the extra sync points added by slew to target don't stay polluting the model. I haven't had issues with convergence anywhere in the sky after changing to nearest point alignment, but n-star alignment did indeed sometimes go totally off the tangent.

I tried nearest point without success. I may not have cleared the model properly but that’s the issue. It should work regardless without such workarounds.