Hi John!

For now I only have few "autofocus_frame_*" frames. No screenshots and log data yet. I'll try to get everything packaged and have it shared it with you.

Probably tonight (my night @ GMT+2)

Until then, thank you very much!

Hi,

I spent my last night to test the autofocus of Kstars V3.6.6 on my RPI4Stellarmate. I don't use the app, I link to the RPI4 with Nomachine.
I join multiple logs and screenshots.

I don't understand how the program runs. At a time, the focuser is going out again and again. So the stars become bigger and bigger. I should expect the program, noting the drift, would go back to reach a better focus.

I had better result, if I can say, with polynomial algorithm. The Linear 1 pass is very bad. The V-curve is flat and when Iit successes, the stars are football ballon! I am glad to not be alone in this case.

I am ready to test my focuser with NINA and CCDCiel on Windows11 OS this night.

Thank you for your help.

Hi Morelli,

Thanks for the information. I see you are using a myFocuserPro2. In the past folks have had trouble with a couple of things. This is a reasonable trail FYI: www.indilib.org/forum/general/13289-line...n.html?start=0#92263

1. Firmware version needs to be current. I can't tell from the log which version you are using but you'll need to make sure its the latest. If not, you'll need to upgrade.
2. Setting "AF Overscan" to 0. For some reason that I don't understand setting a value > 0 caused problems in the past with this focuser. So you'll need to manage any focuser backlash some other way, e.g. use the driver backlash.

These 2 things hopefully with improve things for you. Please let me know.

Here are a couple of other suggestions. These won't fix your issue but are just suggestions...
I see you are using a curve type of Quadratic. This should work but I'd recommend Hyperbola or Parabola. You are also using the SEP profile of "All stars". Again, this should work but I would normally recommend 1-Focus-Default - unless of course, you have a specific reason for using All Stars.

Hope this helps. Please let me know how you get on.

Hi John,

As I said on my last message, I made tries with NINA and CCDCiel. I didn't succed to operate the AF with CCDCiel and I will try again. Because, there is a focuser training that calculates focus parameters and I would want to see how it runs.

With NINA, no issues with the AF, even if the user interface is of poor quality. The V-curve is quickly made and I could reproduce it.

Compared to these two programs, KSTARS' user interface is by far the best ! And more, I had to dive back into Windows. A real nightmare for me!

So, I think to have eliminate focuser device issue.

Hi,

(sorry, yesterday was impossible)

Attached ScreenShots of my configuration for Autofocus module. I simply manually drove to optical good focus (not perfect) and tryed with Luminance...

I also have bunch (56- compressed about 760MB) of autofocus-frames 32MB each... (iI'll try to batch convert those to 8 bits and try to upload later if need it)

Behaviuor was as described before: It reaches perfect focus but it continues to perform/evaluate on more different positions. it does stop nor recognizes focus) Curious as it could be, I finalized it and manually reach again autofocus which happened to be at - 2xInitialStep (it started at 8.535 ticks, after cancelling the process it was going no where, manually focused again at 7.535). I could of course have backslach in my optical train, but I doubt it could be as much as 1000 ticks. I sincerely do not know.

Hope you find what the problem is.

Many thanks in advance for your help.

Jason

For me, I ran some more updates last night, some of which contained indi stuff. Then I did some imaging and all my focusing ran without issue. Can't really say that there is much usefulness to me replying other than I complained on this thread of the same problems. Im at a fairly long focal length of 2450 (304mm RC/F8), so maybe my issues are seeing related.

Now my guiding, that is another matter. Still haven't dialed that in well.

Jason lower intial step size in mechanic tab. Ekos is not good at picking up too defocused stars. From your screenshots it seems like your step size is 500. Try lower it to 100.

Hi Jason,

As Nou suggests it looks like you are moving the focuser too far from the focus point. So what value to set Initial Step Size to? There are a couple of ways:

1. Set it to the Critical Focus Zone of your Optical Train. The CFZ module will calculate it for your optical train but you'll need to supply 1 parameter which is how far in microns the focal plane moves when the focuser moves 1 tick. Its the highlighted field:


For refractors you can move the draw tube in and out, measure how far it moves with callipers / ruler and note down how many ticks the focuser moved. This will give you the number of microns / tick. For other telescopes where you are moving the primary or secondary mirror you'll need to investigate how to measure this with the telescope specs. If you are struggling with this if you send details of your telescope and focuser I'll see if I can figure it out from the online specs (some manufacturers publish the required info, but some don't unfortunately).

So in the screenshot, if you measured 1 tick to be equal to 1 micron then your CFZ would be 39 steps so try setting the Initial Step Size to that to start with. Once its working you can tweak to optimise it.

2. Start at focus and move the focuser (in or out it doesn't matter) a few ticks and take a frame. Keep repeating. What you are looking for is a step size that makes a difference to the HFR (displayed at the bottom of the v-curve) but doesn't go so far as to get bad values in the HFR. Once you get too far away from optimum focus star detection breaks down and the HFR will be incorrect. So you are after the sweet spot between moving enough to get different values for the HFR but not so far as to start to get star detection breaking down. When you have worked out a range from optimum focus to where star detection starts to break down then you have the operating range for Autofocus. I would suggest aiming to get around 10 datapoints so 5 either side of optimum focus. So divide your previously calculated range by 4 and that's your step size.

Backlash
You have 50 ticks in AF Overscan. Maybe that's fine. Can't tell. There's help in the manual for sorting out backlash if you're unsure. It will need to be sorted out in order for focus to work.

There are some other settings that you have that could probably be optimised but lets get the basic focus working before fiddling with those.

Hi Bryan,

With that equipment your focus is going to pretty sensitive to many things, like ambient temperature, start point, filter, seeing, etc. If you get problems again feel free to post on the forum. There may be some things we can do to make Focus more resilient.

Thank you John, Nou.

I took into account your remarks, and I believe I reached a non-textbook-orthodox-solution, at very least, it seems persistant through various autofocus procedures with different filters. Nevertheless, I'll continue studing it since this is much deeper as I thought, considering John's remarks.

I measured microns/tick and it is 0,7636 microns/tick. My resolution is 0,96 arcsec/pixel with a 203/800 Newtonian using ASI1600MM (sensor with pixels of 3,8microns and 4656x3520) I believe I also introduced wavelengths by filter (roughly)

I'm sorry but I skept what advisor propose me: 18 ticks with x5 multiplier. I went to 75 ticks initial step with No OverScan and no multiplier because I changed the algorithm to use to Polynomial and increased tolerance to 10% (instead of 5%, I tested with 20% and 10%, I can imagine the lower the better focused image) Seeing and temperature difference of the primary and ambient is something that should be seriously considered. I could say before that it was important, now I believe is critical!

You can see all parameters below. Like I said, solution is not orthodox (I ignored advisor) but it seems to work so far.

I'll keep testing and improving my knowledge!
BTW: Among the screenshots there is the solution autofocus found for L, Ha, Oiii & Sii. Next step I'll try the "Offset Generator" to put more time on imaging and less on focusing.



Many thanks for your advices!

Hi everybody,

I am discouraged ! I install the last update of Kstars, after reading the messages on this post. I made tries this afternoon in simulation mode. It seems AF was running well. I made tries on the real sky this night. No way !!!!! No real V-curve, no reproductibility !!! No stars detection I see on the screen !

My setup: ED80 with reducer 0.85, ASI183MM, Focuser DIY based on Robert Brown model. It runs well with NINA !
step size=7,9 micron. I start the autofocus near the focus point, made with Bahtinov.

How is it possible to have this v-curve with two point at the opposite with the same HRF ?

The internal plate solver crashes.

Hi Jason,

Glad things are starting to work for you!

The 18 ticks is probably too small for the step size. I normally would use a tolerance of 1 in the CFZ screen to get the tick size which would turn the 18 into 54 ticks. Its just a starting point and if 75 works for you then fine.

I suspect the Advisor is suggesting the 0.33 tolerance which is why you got 18 ticks. I'll look into changing that.

I would recommend using Linear 1 Pass as I personally believe it to be more accurate than Polynomial but obviously its your choice.

Let us know how things go!