Autofocus seems suspect

Questions, bug reports, requests for enhancements, etc.
Post Reply
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Autofocus seems suspect

Post by iMustBcrazy »

I started using autofocus with the ZWO AEF and it seems like its having a lot of trouble. One thing is the FWHM, HFD and Peak seem to not agree. Minimum FWHM and HFD should occur with max peak but the graphs often move in opposing directions.

Also FWHM is saturating at 9.000 even though its on a relatively good star that is not saturated.

HFD generally around 4 or 5 does not change much and seemingly not of much use.

If I perform the autofocus many times, it never completes with the same or similar setting.

The star profile is also very poor. Sometimes it has a zig-zag pattern.

I get the impression that star analysis is being performed before the image is debayered :o Next time out, I'm going to use 2x2 binning to see if things work any better.
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

Nobody complained about the auto-focus accuracy so far. Even though several people were comparing it with Bachnitnov and said it is always spot on. However if you you have a significant backlash you have to figure out the amount of backlash you have and provide it to the system.

Meanwhile can you provide a Tace log from the server?

This is how it is done on indigosky:
1. configure the setup and prepare to start the focus.
2. In control panel open Server @ indigosky ->log level and select Trace
3. initiate the focusing sequence.
4. once finished set to log level back to error or info
5. send us the /home/indigo/indigo.log

Rumen
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: Autofocus seems suspect

Post by iMustBcrazy »

rumen wrote: Mon Jul 26, 2021 2:56 pm Nobody complained about the auto-focus accuracy so far. Even though several people were comparing it with Bachnitnov and said it is always spot on. However if you you have a significant backlash you have to figure out the amount of backlash you have and provide it to the system.

Meanwhile can you provide a Tace log from the server?

This is how it is done on indigosky:
1. configure the setup and prepare to start the focus.
2. In control panel open Server @ indigosky ->log level and select Trace
3. initiate the focusing sequence.
4. once finished set to log level back to error or info
5. send us the /home/indigo/indigo.log

Rumen
Hi Rumen, I'll try a trace but up to now traces stalled operations to the point where nothing worked, so probably the best I can do is debug.

I've measured and set the backlash - its not much, 17 steps.

What about the de-bayering of the image itself? Maybe those who have good results have monochrome cameras? Because I see this weird zig-zag profile curves and saturated profile curves - red and blue, I don't know what the colour means - yet the peak value is not saturated.

The skies are really bad these days - due to forest fires and high humidity - so I'm going to try to go out and just test. Hopefully come back with some good data.
User avatar
Peter Polakovic
Posts: 2696
Joined: Thu Apr 21, 2016 9:38 am

Re: Autofocus seems suspect

Post by Peter Polakovic »

Algorithm uses the average of all channels for colour images, maybe a better idea than debayering is to use binned raw image.
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

I ma not sure I understand, how can a trace log stall your operations...

And yes images are not debarred and in this case FWHM is not a good criteria (FWHM relays on the stellar profile, which is distorted by the bayer matrix). That is why HFD is taken in to consideration together with max. On the other hand HFD is not so sensitive to bayer pattern as depends on the total flux not the profile. HFD is more robust measurement in real life as stellar profiles are deformed by many factors (seeing, central obstruction, bayer patterns etc).
So i would not and I never look at the FWHM. On the other hand debayering is not one algorithm. There are many methods to do so and each suffering from its own problems ranging from chromatic errors to reduced sharpness.

By the way how do you set your final step value? it should be about 1/3 of the CFZ (critical focus zone) for your setup. As if the value is greater you may never fall in to precise focus. On he other if too small it will take forever to complete.

So my advise is to provide a trace log to see what is going on and separately try to play with minimal step and backlash.
Please note 17steps may be too much it all depends how these 17 steps translate to focuser travel distance. And if you have a backlash you can not expect that the focus will be achieved at the same focuser position. Ideally it can land anywhere in a range with the length of the backlash.

Can you provide some more info about your setup like scope (diameter, Focal length, optical system) and camera (pixel size). What focuser?
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: Autofocus seems suspect

Post by iMustBcrazy »

Peter Polakovic wrote: Mon Jul 26, 2021 8:02 pm Algorithm uses the average of all channels for colour images, maybe a better idea than debayering is to use binned raw image.
Interesting. I will try binning to see if that works but the problem with that is that the camera configuration has to be changed to focus which is not really desirable. I imagine that would start to hinder automation, while star measurement on the fly would be impossible.
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: Autofocus seems suspect

Post by iMustBcrazy »

rumen wrote: Mon Jul 26, 2021 8:20 pm I ma not sure I understand, how can a trace log stall your operations...

And yes images are not debarred and in this case FWHM is not a good criteria (FWHM relays on the stellar profile, which is distorted by the bayer matrix). That is why HFD is taken in to consideration together with max. On the other hand HFD is not so sensitive to bayer pattern as depends on the total flux not the profile. HFD is more robust measurement in real life as stellar profiles are deformed by many factors (seeing, central obstruction, bayer patterns etc).
So i would not and I never look at the FWHM. On the other hand debayering is not one algorithm. There are many methods to do so and each suffering from its own problems ranging from chromatic errors to reduced sharpness.

By the way how do you set your final step value? it should be about 1/3 of the CFZ (critical focus zone) for your setup. As if the value is greater you may never fall in to precise focus. On he other if too small it will take forever to complete.

So my advise is to provide a trace log to see what is going on and separately try to play with minimal step and backlash.
Please note 17steps may be too much it all depends how these 17 steps translate to focuser travel distance. And if you have a backlash you can not expect that the focus will be achieved at the same focuser position. Ideally it can land anywhere in a range with the length of the backlash.

Can you provide some more info about your setup like scope (diameter, Focal length, optical system) and camera (pixel size). What focuser?
It seems that there is so much logging during trace that things slow down. I'll try again... it may be better with V4 logging on the server.

My final step value is quite small - 5 steps - with almost no visible effect. I start with 35 steps which has noticeable effect but not too much. This is all very subjective but there's not much else to go on.

I have a 102mm F7 scope with dual speed (10x) 2" crayford focuser. Pixel scale with ASI294 is 1.34"/px
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

Do you have a bachtinov mask?
If you have one try checking the focus with it.
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: Autofocus seems suspect

Post by iMustBcrazy »

rumen wrote: Mon Jul 26, 2021 10:03 pm Do you have a bachtinov mask?
If you have one try checking the focus with it.
Yup, I start with that.
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: Autofocus seems suspect

Post by rumen »

Excuse me, I do not understand. Maybe it is my English, but do you usually start with it and the results are not OK. Or next time you will start with it?
Post Reply