Focuser backlash?

Hi!

Is it important to have either zero backlash or backlash compensation to get autofocus to work?
Mine have backlash, but fcusb does not seem to have backlash compensation....

Why is backlash compensation not implemented at a higher level than the driver? Seems to be it should be generic to do if one can move the focuser back and forth. And maybe one could also determine the backlash automatically by doing a forward and a backward v-curve?

Thanks for any help!

Best regards
Søren

Yes, I had asked that before, too. Hy seems to have some code for linear focusing , not sure if that is going to end up in the main code.

Thanks a lot. Could one implement this in ekos, such that it would work for any focuser driver?
I may just need to download the source and see how hard it is to modify and compile.....

Sorry for that delay. I will get back to it at some point (the linear focus algorithm--yes, scanning through in one direction then reversing to find the best point) .

WRT backlash, I did look to see if it was trivial to find (and eliminate) the backlash in my Moonlight focuser, and I will say
(a) it's clear I have backlash, and
(b) the number doesn't just pop out in the V-curve when you go back and forth--there's other noise in the system that hides it a bit.
Hy

If there is a lot of noise, then the first thing I would try was to do a least squares fit of a line to both sides of the v-curve (made with many small steps) and see if the intersection of the two lines matches the optimal fucus. If yes, then make a backward v-curve, do the fitting and see if the step difference between the two focused positions matches the backlash. The backlash should also show up as the shift between the lines of the forward and backward 'linear' v-curves.

Can this be tested with a simulator?
How hard is it to compile ekos? Is it done together with compiling kstars?

hy wrote: Sorry for that delay. I will get back to it at some point (the linear focus algorithm--yes, scanning through in one direction then reversing to find the best point) .

WRT backlash, I did look to see if it was trivial to find (and eliminate) the backlash in my Moonlight focuser, and I will say
(a) it's clear I have backlash, and
(b) the number doesn't just pop out in the V-curve when you go back and forth--there's other noise in the system that hides it a bit.
Hy

Thanks Hy for looking at things!
Some remarks:

• The V-curve isn't a V, nor a parabola, but hyperbolic. If you measure the full curve that might make a difference. If you also read CN, there had been a discussion about that not too long ago.
• Use really long exposures to counter seeing influence (or do with quite good seeing)
• I first had my motor connected to the micro-focuser. But that one has slippage, i.e., with some mild force you can trun it without the focus tube moving.
• an option without backlash correction could be forcing focus moves to always end inwards, i.e., an outward move would move some additional distance outwards and then inwards to the intended value. (I had tried to implement that myself, without success )

Attached is a text file with some data I dumped a while back, when I started looking at focus.
Never worked it hard, but if you want to give it a shot, let me know what you think my backlash is

I generated this data by modifying the code a little to get the focuser in an infinite loop, starting 100 steps below the focus point,
going up by 10 steps twenty times, each time computing the HFR,
then reversing direction and going down by 10 steps twenty times, and repeating forever.
I was using my moonlight focuser, I don't remember where it was pointed, I probably still have the images.
Did about 7 passes through the data.

Re compiling kstars/ekos/indi, it's not hard if you're comfortable with that kind of thing.
The easiest way to get started is to run one of Rob's (rlancaste) AstroPi scripts
see indilib.org/forum/general/5524-astropi3-...revised.html?start=0
depending on your target platform (or just read the script and run the commands manually).

Feel free to ask if you have any questions.
Hy

Thanks for the data. It certainly looks different than the curve shown here:
nighttime-imaging.eu/docs/develop/site/advanced/autofocus/