I hope you are still keeping well in these strange time?
I have been doing some auto focus testing with APT and discovered that I was getting a strange focus graph from SkySimulator. Instead of being a nice "U" shape, the graph would form a "W". That is, at point of best focus, and either side of it, the HFD/FWHM values increase rather than be at a minimum (see attached "focuscraft.jpg")
Thinking it might be a problem with APT's new focus analysis tool FocusCraft, I tried it's old FocusAid. This produced the same result. So then I tried SharpCap's Focus "Assistant > FWHM measurement". it also produced this strange curve (see attached "test 14 sharpcap.jpg")
I then found that my SkySim version was a bit out of date (V0.1.14) so upgraded to the new version (V0.1.22). This has made the situation worse! Instead of a "W", I now get an "n", the complete inverse of what you would expect (see attached "test 20 sharpcap.jpg"). Visually the stars appear to become more and more focused, but the Focus measurement doesn't.
What I suspect is that they measure the total star flux rather then the half flux diameter. Since the images have to go through 8 bit images, I don't keep the total flux constant. Normally the star flux is constant and the star intensity/pixel drops to the sqr of the diameter since the surface increase with sqr of the diameter. But to prevent they disappear in the background noise too soon, I reduce the flux per pixel linear. So the total flux increases. I have to make an mod or option to avoid this.
Is there a possible solution, perhaps with an option to do it one way or other?
I don't know how Ivo does his star analysis, so is it worth getting him involved to check this is the problem?
cheers, Jim
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Well I managed to get it barely working with SharpCap. The "black threshold" should be set at 1. But it is very unreliable. It doesn't work with multiple stars only with single star FWHM measurement. And it doesn't work with the brightest stars. The Sharpcap routine is not robust. Maybe because the images are 8 bit.
Changing the star algorithm back to the original setup didn't help, even worse.
I can confirm that the Sharpcap multi-star focus routine is useless. I couldn't get it to register anything but a flat line, even with my scope and real stars! I'm sorry, because I could have saved you time, by mentioning g it before.
Cheers, Jim
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
The APT focusing doesn't work well for me. The HFD seems to work better. I see clearly in the APT histogram for out of focus the star curve getting wider and the peak drop. But it looks like APT can't handle small stars. Subpixel calculation missing? Maybe it works better with fat stars. The star detection should be robuster.
Note the histogram for the classic ideal star is symmetrical but weird. It indicates a problem in the HFD calculation. Maybe something to discuss with Ivo.
I had the same problem 3 years ago with APT in the field.
I have tried it during testing but it did not improve APT detection. Anyhow I will make it optional tomorrow.
Other programs have no problem with it since HFD is not related to the flux amount, only the diameter. A faint star can have the same HFD as a bright star.
Cheer, Han
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I just wanted to give you an update on this. I think I found the problem.
I discovered that making sure the simulator didn't add any additional data or graphics to the image, APT didn't see the Best Focus "hump". I can only assume that the APT focus algorithm was seeing the text, object outlines, etc. as possible stars. Now the whole system is working perfectly.
Thanks as ever for you patience with me! :-)
All the best, Jim
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Han,
I hope you are still keeping well in these strange time?
I have been doing some auto focus testing with APT and discovered that I was getting a strange focus graph from SkySimulator. Instead of being a nice "U" shape, the graph would form a "W". That is, at point of best focus, and either side of it, the HFD/FWHM values increase rather than be at a minimum (see attached "focuscraft.jpg")
Thinking it might be a problem with APT's new focus analysis tool FocusCraft, I tried it's old FocusAid. This produced the same result. So then I tried SharpCap's Focus "Assistant > FWHM measurement". it also produced this strange curve (see attached "test 14 sharpcap.jpg")
I then found that my SkySim version was a bit out of date (V0.1.14) so upgraded to the new version (V0.1.22). This has made the situation worse! Instead of a "W", I now get an "n", the complete inverse of what you would expect (see attached "test 20 sharpcap.jpg"). Visually the stars appear to become more and more focused, but the Focus measurement doesn't.
All seems very strange! Any thoughts?
Thanks, Jim
Using CCDCiel I get nice hyperbolic curves.
What I suspect is that they measure the total star flux rather then the half flux diameter. Since the images have to go through 8 bit images, I don't keep the total flux constant. Normally the star flux is constant and the star intensity/pixel drops to the sqr of the diameter since the surface increase with sqr of the diameter. But to prevent they disappear in the background noise too soon, I reduce the flux per pixel linear. So the total flux increases. I have to make an mod or option to avoid this.
Last edit: han.k 2020-12-14
Hi Han,
Thanks for answering so quickly!
Is there a possible solution, perhaps with an option to do it one way or other?
I don't know how Ivo does his star analysis, so is it worth getting him involved to check this is the problem?
cheers, Jim
I will try to test it here and if okay make a new version.
Hi Han,
You are a hero! 🙂
I pinged Ivo and I think he thinks the same as you. anyway, he said he will run some tests too.
Since I haven't seen a clear sky since Sept, I use SkySim all the time to help me test and learn. it us a brilliant tool.
Thanks, Jim
Well I managed to get it barely working with SharpCap. The "black threshold" should be set at 1. But it is very unreliable. It doesn't work with multiple stars only with single star FWHM measurement. And it doesn't work with the brightest stars. The Sharpcap routine is not robust. Maybe because the images are 8 bit.
Changing the star algorithm back to the original setup didn't help, even worse.
I will now try APT.
Last edit: han.k 2020-12-14
How can you set in APT the exposure time in Auto Focus Aid?
Last edit: han.k 2020-12-14
Hi Han,
I can confirm that the Sharpcap multi-star focus routine is useless. I couldn't get it to register anything but a flat line, even with my scope and real stars! I'm sorry, because I could have saved you time, by mentioning g it before.
Cheers, Jim
You just set the exposure and gain (if you've set SkySim driver to support it) on the Camera tab
The APT focusing doesn't work well for me. The HFD seems to work better. I see clearly in the APT histogram for out of focus the star curve getting wider and the peak drop. But it looks like APT can't handle small stars. Subpixel calculation missing? Maybe it works better with fat stars. The star detection should be robuster.
Note the histogram for the classic ideal star is symmetrical but weird. It indicates a problem in the HFD calculation. Maybe something to discuss with Ivo.
I had the same problem 3 years ago with APT in the field.
Hi Han,
On the image the graph is for the FWHM, double click on that area to switch to the HFD graph. Definitely the HFD is the better metrics.
Regards,
Ivo
Hi Han,
Thanks for looking in to it. I think Ivo is following this, so perhaps we can wait for his input. Otherwise, I'll ping him in a while.
Thanks again.
Cheers, Jim
Last edit: Jim Hunt 2020-12-17
Hi Han,
A friend who is also involved has made this observation:
So, is there anything you can do (or are prepared to do :-) )
Thanks, Jim
Last edit: Jim Hunt 2020-12-17
Hello Jim,
I have tried it during testing but it did not improve APT detection. Anyhow I will make it optional tomorrow.
Other programs have no problem with it since HFD is not related to the flux amount, only the diameter. A faint star can have the same HFD as a bright star.
Cheer, Han
Hi Han,
I just wanted to give you an update on this. I think I found the problem.
I discovered that making sure the simulator didn't add any additional data or graphics to the image, APT didn't see the Best Focus "hump". I can only assume that the APT focus algorithm was seeing the text, object outlines, etc. as possible stars. Now the whole system is working perfectly.
Thanks as ever for you patience with me! :-)
All the best, Jim
Hello Jim,
Good you get it working.
Regard, Han