autofocus failure: brightest pixel too close to edge of the image

Calibration tools, stacking, aligning,..
Post Reply
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Fri Jul 12, 2019 1:41 pm

I haven't been able to get autofocus to work in prism... I can use SGP to focus but why I can't focus with in prism is very confusing, seems like the error isn't correct.

The error is "brightest pixel (pixel) analysis sub-window (80x80) is too close to the edge of the image (50x50)" - it happens consistently on the first exposure's analysis.

I've tried changing the method of HFD and FWHM, I don't think it changes the error , it definitely is not successful.

astrometry.net reports this hyperstar v4 + c11 hd focal length as 518 mm, datasheet states 540mm - I've tried with both of those numbers. I use an optec directsync feathertouch focuser with the focuslynx controller.

I am using an asi 183mm - it's not clear if asi 183mm has really much of anything to do with the issue but there this thread: http://www.prism-astro.com/forum_us/vie ... +asi#p1067 - what was the issue/resolution on this thread?

My driver is maximum of 3 months out of date, please let me know if there are any other diagnostics needed.

I also get an error with direct auto focus, but I'm not sure what it was.

Also - the user interface I am seeing for autofocus is different looking from the linked thread and https://youtu.be/JIQuVZ05Lfg?t=282 - it's in steps for one and it looks almost feature wise less than those instances. There's no optimum step button and no optimum gap button, the focus points profile manager region is also missing.

Pics
autofocusfailed1.png
autofocusfailed2.png
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Tue Jul 16, 2019 5:19 am

FWIW @Cyril Cavadore , I tried this under 8 bit mode using the ascom drivers today (I normally I use 16bit) and it didn't change the result any. I still have no autofocus solution with prism working. I'm out of ideas. manual focus isn't a bad experience but I mean I have optec focusing hardware... it's good stuff and prism isn't playing ball in a really bizarre way - any ideas?

Trying to grind through and trying to get it working many times, I noted via ascom that the zwo ascom camera 1 and 2 get into confusing states in prism. Sounded like from @Hamza Touhami that this was a known issue. I think I've only observed this in prism. I changed to the zwo direct driver and it seems issue is worked around but it is worth noting this bug bit me and it's really confusing to seems to leave subtle errors here and there. I
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Wed Jul 17, 2019 4:13 pm

any next steps or things I can do to help further the issue along @Hamza Touhami / @Cyril Cavadore ?

I think one thing I need to try is to see if the mm vs step size software option influences this at all but that shouldn't be the case at first glance.
Hamza Touhami
Site Admin
Posts: 500
Joined: Sat May 20, 2017 6:05 pm
Location: Phoenix,AZ
Contact:

Wed Jul 17, 2019 5:15 pm

I'd like you tp perform a test for me. When you autofocus next time, double the size of the window in the autofocus function. in the presence of hot pixels in the frame and when the star is defocused, it happens that other "pixels" become brighter. But let's start by increasing the window size
Hamza Touhami | Site Administrator
Hyperion-Astronomy.com
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Wed Jul 17, 2019 8:36 pm

Hi @Hamza Touhami ! You'd like to double the size of the analysis window, correct? I believe I'm operating at maximums for both methods thinking along those lines. I've tried many random locations and gotten the issue so I'd be very surprised if it's hot pixel and not showing up well visually. I can also make available a dark so you can see where they are, bad pixels were iirc < 2%.

The screenshot shows hdf analysis radius pixel of 40, it won't let me set higher than that. For FWHM I'm running at 101, can't recall if that's maximum, but I do recall it being larger than defaults. I will update this post later tonight with the settings occluded by the popup and a dark/bias frame. Looking at the log messages and X1, Y1, X2, Y2, I have

For the first image:
>>> 4043-3840+1
204
>>> 2251-2048+1
204

Brightest pixel @ x=26, Y=25

For the second image
>>> 1563-1360 + 1
204
>>> 3095-2892 + 1
204

Brightest pixel @ X=27,Y=29

To confirm, the window sizes you are talking about are the FWHM and HDF analysis window sizes behind the popup, yes?

Is X=0, Y=0 the top left of the analysis window?

Would any of the calibrations remove the hot pixel before it can have an effect in focus analysis or is there some median filter somewhere?

Unfortunately several days of cloudy nights now, I may have to use city lights at a distance as a proxy and at least see if we can get some changes.
Cyril Cavadore
Site Admin
Posts: 61
Joined: Sat May 20, 2017 6:15 pm

Thu Jul 18, 2019 1:19 pm

Dear,
We have spotted an issue two month ago, like a regression with the .56 release, and this has been fixed since.
Please try out with this ->
https://www.dropbox.com/s/v7l420sam68sm ... e.zip?dl=1
Our apologies.
Please check and if OK I'll transfer it to the official release.
Rgds,
Cyril Cavadore
Principal Prism Author,
Alcor-System
www.alcor-system.com
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Fri Jul 19, 2019 12:40 am

Hi @Cyril Cavadore

I tested today... without a mount so you'll have to rely on some manual positioning and not really being tracked. Fun exercise - I just don't have a ton of options to setup easily at home.

I'm attaching some interesting screenshots. I don't think the problem is fixed - is it an asi camera specific issue? Could you explain it maybe? I'm a programmer so I'd like to know what the camera is doing wrong.

I managed to get it to crash with the analysis window open, I believe it always failes to identify the star on the first exposure's analysis

I'm attaching an out of focus shot just for a spot check,@ 4x binning.

I unpacked the dropbox zip archive and replaced all dlls/exes/enus into the program installation directory (32 bit) prior to running the tests - about reads 10.3.57.467
Attachments
fail-manualtracking-directdriver.png
fail-manualtracking-ascom-crashvisanalysis.png
fail-manualtracking-ascom.png
Hamza Touhami
Site Admin
Posts: 500
Joined: Sat May 20, 2017 6:05 pm
Location: Phoenix,AZ
Contact:

Sat Jul 20, 2019 8:27 am

thanks Jason, I will test the same version tonight fro Chile. we will get through this, stay tuned
Hamza Touhami | Site Administrator
Hyperion-Astronomy.com
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Sun Jul 21, 2019 5:06 am

I tested in a more proper setup tonight, it most definitely still doesn't work.

Can anyone describe to me the nature of the problem? I am familiar with lower level details of cmos sensors, ascom, drivers, image processing. I can read the error prism is presenting but I don't think it is the real issue being debugged maybe?

Is this just broken for me or asi cameras?

Unfortunately still too expensive to focus all my filters manually so I only had ha manually focused tonight... and even then no where near as well as my optec would allow with an automated global focusing procedure.

Screen shot showing failure with a tracking stable system.
Attachments
newfailure.png
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Wed Jul 24, 2019 9:41 pm

Hi @Hamza Touhami / @Cyril Cavadore

Is there any update.

Please let me know what if anything is going wrong.

The worst part of this is not clarifying what's wrong or why. I've watched the prism tutorials on youtube on all the focusing features several times now... I'll be trying the relevant settings used in the video in my next test - likely this weekend.

Is further trial and error on my part going to resolve the issue? This answer alone makes the issue a whole lot less painful on my time and experiences with prism and support.
cioc.adrian.dan
Posts: 33
Joined: Mon Feb 18, 2019 1:03 pm

Thu Jul 25, 2019 1:13 am

Hey, I some times have the same error. I usualy try 4-5 times to autofocus until I succeed. The direct autofocus also gives me the same error constantly. Actually I was going to ask the same question as you today. Tonight I will install the update and hopefully it will fix the problem
Hamza Touhami
Site Admin
Posts: 500
Joined: Sat May 20, 2017 6:05 pm
Location: Phoenix,AZ
Contact:

Thu Jul 25, 2019 7:01 am

I understand,, we will have to revisit this immediately.
Hamza Touhami | Site Administrator
Hyperion-Astronomy.com
cioc.adrian.dan
Posts: 33
Joined: Mon Feb 18, 2019 1:03 pm

Thu Jul 25, 2019 4:25 pm

I installed the update provided in the link and both direct focus and automatic focus work now ;)
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Thu Jul 25, 2019 8:39 pm

@cioc.adrian.dan what camera/driver were you using (doubt it, matters but it could)

Is it working like 5/5 times per your previous sporadic issues?

Can you confirm the version prism reports? My screen shots say 10.3.56.442, the zip archive version.txt file says 10.3.55 - maybe I failed to overwrite a program (file in use pitfall on windows) or something. I believe I tried this a few times to be sure but it's one of the reasons I called out the version in my post, to zero in on user error /version mixups.

To install, you just closed prism, and copied all the files into the 32bit program directory, yes?
cioc.adrian.dan
Posts: 33
Joined: Mon Feb 18, 2019 1:03 pm

Sat Jul 27, 2019 11:58 am

Yes I copied the files from the zip folder to prism folder. I have a moravian g3 16200 ccd camera with a 12 inch f8 acf and an optec focuser. I tested 1 night several times and it worked. I like the direct auto focus tool better because it selects an appropiate star automaticky for focusing
cioc.adrian.dan
Posts: 33
Joined: Mon Feb 18, 2019 1:03 pm

Sat Jul 27, 2019 12:04 pm

Tomorrow I will see the verion reported and post the image.
It should work. I don’t know why only we had the problem. Maybe others didn’t updated Prism. There was an update maybe 2 months ago, when I noticed the sharplock continuos focus capability was removed, but I found a workaround for my on axis guider anyway.
cioc.adrian.dan
Posts: 33
Joined: Mon Feb 18, 2019 1:03 pm

Sat Jul 27, 2019 12:10 pm

Also if you have an older prism instalation file you could uninstall current version and install the older one just to test, or if you have a virtual machine you could install the older version there so you dont have to reconfigure anything again.
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Thu Aug 01, 2019 2:07 am

ok so this does seem to work, I didn't change anything from last time I tried it on a perfect night. I don't have a good polar alignment going right now though I got a good 2-5 minute window I can run the procedure within tolerance on.

It does seem to be giving me random failures still of brightest or too close to edge to start with but it is close to 50/50 odds that it'll work without issue or crashing.
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Fri Aug 02, 2019 6:07 am

issue train not stopping... restarted prism several times tonight. I had a few instances without issue and then it started only finding noise/absense of stars in the search window - like it was extracting the wrong region of the image. I've had many times where it seemed there was some sort of offset between the mouse click and the region of interest selected but it was really bad tonight. Once it started doing this, I wasn't able to get it working again despite restarting prism a few times.
jason
Posts: 21
Joined: Thu Apr 18, 2019 3:53 am

Tue Aug 06, 2019 5:47 pm

@Hamza Touhami

Saw the TAIC after the fact, good job, looking forward to more of those. I learned more there as a global workflow and things that the isolated tutorials are missing, There's just so much functionality, you need to spend a ton of time in prism experimenting to have a chance to notice some of it ( I don't believe the documentation calls out much of this.... which is kind of understandable) . I'll have to go back a few times to make sure I try out everything you called out in the session. Looking forward to the next one. I think one thing I and other people were noticing while you were talking was being able to graphically or more contextually change the angle of the sensor previews, plugging in the numbers by hand can be backwards in the workflow. The panoramic could also be streamlined but I'll have to think about what changes work there better - interactivity of the sky chart /redraw speed would also be very nice if improved - I run on an intel compute stick. Another drive-by thing I noticed is the usage of "sorting" verbiage was more of a filtering criteria, maybe it sorts somewhere sometimes though.

Couldn't help but notice you had no issues running automated focusing using the same version... per my last message, I seem to be experiencing times, which I'll record next time where prism doesn't seem to be extracting the region of interest that contains the star - have you seen a bug like that in any other context, especially that of auto-focusing? I may be open to remote desktop based control to let you have a go at it or something if my next video doesn't capture the issues good enough.
Post Reply