Autofocus seems suspect

Questions, bug reports, requests for enhancements, etc.
User avatar
Peter Polakovic
Posts: 2742
Joined: Thu Apr 21, 2016 9:38 am

Re: Autofocus seems suspect

Post by Peter Polakovic »

Never mind, what is good for CloudMakers is also good for INDIGO and vice versa :)
User avatar
Peter Polakovic
Posts: 2742
Joined: Thu Apr 21, 2016 9:38 am

Re: Autofocus seems suspect

Post by Peter Polakovic »

Seems to work, I'll rebuilt AstroImager/AstroDSLR tomorrow...
Screenshot 2021-09-20 at 21.24.13.jpg
Screenshot 2021-09-20 at 21.24.13.jpg (395.89 KiB) Viewed 300 times
rumen
Posts: 156
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

Guys we have some focus test data from 2 telescopes:
1. 2meter RCC at NAO Rozhen Observatory - star rich region of the sky
2. Astro-Physics 130mm f/6 - region with only one not very bright star (very challenging image).

Each data set contains 15 frames taken with a step of half the critical focus zone (CFZ / 2).

And I compared the results visually and to another method called Variance method considered to be really accurate, but way more complex and slow.

The results from Rozhen are spot on for both methods and visually:
Variance: 80.13 mm
RMS contrast: 80.13 mm
With CFZ = 0.10mm
Visually: 80.10, 80.15 and 80.20 are hardly distinguishable.
both methods agree.

For the AP130 (the image set is more challenging only one visible star in the field)
Variance: 23832 steps
RMS contrast: 23820 steps
Visually: 23600, 23800 and 24000 are hardly distinguishable.
With CFZ of 400 steps.
and both methods disagree with 12 steps which is 1/33 of the CFZ. (Absolutely invisible).

Here is how it looks:
Sequence of 3 frmes with 200 focuser steps difference (23600, 23800 and 24000), same star Taken with !P130, The one in the middle is only 20-30 steps from the predicted perfect focus. Left and right are about 200steps off
Sequence of 3 frmes with 200 focuser steps difference (23600, 23800 and 24000), same star Taken with !P130, The one in the middle is only 20-30 steps from the predicted perfect focus. Left and right are about 200steps off
Screenshot from 2021-09-22 19-28-29.png (33.14 KiB) Viewed 285 times

Very promising results so far!

Rumen

Update: Just received a result from a friend: FWHM Vcurve yields 23794 to be the best focus for the AP130 set. Which is 38steps from the Variance and 26 from RMS contrast method.
User avatar
elmore
Posts: 79
Joined: Tue Dec 04, 2018 2:34 am

Re: Autofocus seems suspect

Post by elmore »

How can I make sure that a Sequence performs an auto focus with the "RMS contrast estimator"? Will it pick up the last method used for manually performing an autofocus?
User avatar
Peter Polakovic
Posts: 2742
Joined: Thu Apr 21, 2016 9:38 am

Re: Autofocus seems suspect

Post by Peter Polakovic »

For now, sequencer always set HFD estimator, so you can test it just with second "autofocus" button. But we'll change it as soon as it will be matured enough :)
rumen
Posts: 156
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

I had a chance to test the RMS Contrast focus and It works excellently as long as there are no saturated stars :)
I think soon we will release it.
iMustBcrazy
Posts: 339
Joined: Sun Jun 02, 2019 3:52 pm

Re: Autofocus seems suspect

Post by iMustBcrazy »

rumen wrote: Fri Oct 08, 2021 7:59 pm I had a chance to test the RMS Contrast focus and It works excellently as long as there are no saturated stars :)
I think soon we will release it.
I tried it on Wednesday night and it worked very well!

I think previously I may have had optical problems with my FF. I bought a new WO FF recently and it made a world of difference. The focus has been working well but Wednesday the contrast method to start really made the focusing user friendly. Thanks!

In fact, on Wednesday, I had no issues at all. The whole suite worked beautifully :D
rumen
Posts: 156
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

I think previously I may have had optical problems with my FF. I bought a new WO FF recently and it made a world of difference.
good optics always make difference :)
The focus has been working well but Wednesday the contrast method to start really made the focusing user friendly. Thanks!

In fact, on Wednesday, I had no issues at all. The whole suite worked beautifully
I am happy to hear that :)
But wait a bit more and I will make it more user friendly: I will take backlash out of the equation - you will only need an approximate value for it :)
rumen
Posts: 156
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

INDIGO focusing is enhanced with backlash overshooting algorithm - it does not require precise backlash measurement. an approximate value will do provided the system backlash is symmetric (backlash in = backlash out). it is described here:

https://github.com/indigo-astronomy/ind ... _TUNING.md
rumen
Posts: 156
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

I have made some changes to the RMS Contrast estimator and now saturation is not an issue any more, as long as there other stars in the field which are not saturated. Tony (Stoyan) just did some tests and it seems that the concept is working. RMS Contrast is getting even better. The code is still pretty rough and just a proof of concept but once stabilized it will become part if the next indigo release...
Here is a bachtinov mask test of the achieved focus of a frame with a bright saturated star.
Attachments
RMS_saturation.jpg
RMS_saturation.jpg (373.84 KiB) Viewed 8 times
Post Reply