OnStep always has a star model when you boot up. It knows your GPS location, UTC Offset and Date time. It knows how many steps per degree it takes to move the mount.

When you do a star alignment you are then changing the model by telling it how far off you are from each target object.

Thus you will get your error of polar alignment as you add alignment stars telling onstep how far off each star is from its default model.

Chad

Thanks Khalid
your'e right, it is the unique way OnStep confirms successful alignment.
Now it does not explain why his goto accuracy doesn't improve and here it is on OnStep side to search I believe.

As it turnout, after doing alignment and press save to NV memory, I got this error message in INDI. I did not notice it before.

2023-10-26T12:27:44: [INFO] Getting Alignment Status: response Error, response = 1965>
2023-10-26T12:27:44: [INFO] Sending Command to Finish Alignment and write
I am trying to do a 5 stars alignment. If I try doing a 3 star alignment then the error code turns to 1963

You would be best off posting to the OnStep group for OnStep errors and questions.
onstep.groups.io/g/main/topics

Doing a reset causes OnStep to forget all alignment data.
That, and the NV save failure is probably why you see no improvement.

I see what you mean:

in fact there are some changes in how OnStep replies to commands "AW#" and "!MP#"
before it was no return so we did use "sendOnStepCommandBlind" but OnStep returns n (0 ..9) that then is read by the following "A?#" command and shows this error message

I fixed it and code is available here : github.com/azwing/indi/tree/OnStep_Work

let(s confirm it works before issuing another pull request

testing now

It worked! Thanks! The goto accuracy still show no improvement though. I tried 9 point align on Sky Planetarium and there are significant improvement of goto on RA. (always in the +_30 arcsec). Keeping the same model in INDI does not show that improvement.

The challenge with OnStep is to follow the little changes like this .... and I am always behind

Your goto accuracy definitively is not Indi side (even before my changes the commands were send correctly, only the response was missinterpretated

How do you run alignmenet with Ekos?

The process should be like this:
1) Start Ekos
2) Select number of stars in control panel
3) Start Alignment in control panel
3) open Ekos and select "Sync" as action after plate solving
4) open the mount model
5) select same number of stars as in (2)
6) generate stars (or choose your own)
7) start mount model
Ekos will now slew to each star, wait for "settle time" shoot a picture, plate solve and "Sync" to the solved coordinates
9) when Ekos has finished the sequence for all stars, the Align should be defined as Finished in the Control Panel
10) OnStep then calculates the delta and Indi show it in bottom of alignment tab
11) Then you have to click "Write Align to NVRAM / Flash" (this will ask OnStep to sore the model (not Ekos Model, the Model calculated by OnStep)

Indi Mount Model is "NEVER" written to OnStep, indi only initiates the sequence by telling OnStep
Number of stars (selector), validate the coordinates for each star the (issue align), confirms to OnStep that the model can be applied ( Write Align to NVRAM / Flash")

All the remaining is done by th OnStep firmware and indi just shows the status (Max Stars, Current Star, # of ligned Stars)

I am using OnStepX version: = 10.19.b and it works so far I cannot test it on the stars with the deluge we have currently

PS: there is a video for that

I think you forgot let the mount return to home position, pointing to the NCP before press start align. I did exactly as your steps. After writing to NVRAM, the information does show, but after a reset, I do not see any sign of improvement on the goto accuracy. Anyway, do you have the problem of the mount do not remember the park position? I slew the mount to the parked position, press use current posstion as park then save in INDI, then go to the SWS interface, press save park position on the web interface. But everytime I resets, the pointing is reseted to the NCP, not the parked position, I have to manually slew out of the place then do a blind plate solve.
Also, there is a weird behaviour with my onstep. after I press parked in INDI and then unpark, the mount just refuses to slew. I issue goto command and the mount just won't slew. I have to do a power cycle to the Onstep controller, and then the mount slews ok again (but the pointing resets to NCP). This is really annoying since I am running everything remotely. Any thoughts on this?
P/S: cool avatar, I have always thought you used a Gandalf image as your avatar but today I use my laptop and see that it's not Gandalf )

Just for info,

I made some cosmetics changes to the Alignment and fixed a typo :GX02# > :GX03# (which displayed errors for both axes the same)
makes no changes in the process and verified that mount model is correctly written in NVRAM with the Write button.
Corrections stay after cold reboot of OnStep and therefore OnStep is aware of the mount model after alignment.

Any tests / reports are wellcome

Pull request issue and lates changes on github.com/azwing/indi/tree/OnStep_Work