Recent focusing bugs - crash in 1.10 beta003, odd behaviour in 005

[nickstarboy]

Using 003 at the start of this evening, I noticed a bug in the focusing which also explains (i think) why my sequence crashed last night 90 minutes into a 6 hour sequence.

After fitting a good parabolic curve to the datapoints, which involved (maybe significant) a final point at the right hand end of the curve (ie 1000,990,980,970,960,950,1010) it found a good fit, possibly moved to that focus, and then crashed. As attachment. This was also how I found the system this morning, having been in that state for 6 hours.

I have upgraded to beta 005. Now that problem hasnt been seen yet (but possibly due to the ‘final point’ issue not occurring) but a new peculiarity.

For each focus point, instead of a 2 second capture, a very short exposure appears to be taken which is analysed using the inner and outer rectangles, and only 1-2 stars are detectable on a very dark background. Then the proper exposure length is shown, with NO rectangles, so the full frame is analysed for focus with many stars. This works, though not ideal. So thats what its doing now!

[nickstarboy]

just realised the short exposure comes AFTER the long one, and is the one used for focus. This makes the focus procedure highly unreliable, and its not working properly….the curve is very messy and it finds a poor focus.

‌

so do i go for the beta003 which crashes, or the beta005 which does this??!!!

[Stefan B]

Neither one is a beta, because 1.10 is in nightly build state - meaning a lot of changes are done without deep tests done already.

For imaging you should go with released (1.8) or beta versions (1.9)

‌

Nightly versions can contain bugs and might be unreliable for imaging runs.

[Stefan B]

Can you also provide a screenshot containing your AF settings?

[nickstarboy]

I know, I understand there can be bugs, but the parabolic focus fit is such an important thing for me that Im keen to try the new nightlies, and hopefully help to troubeshoot. Im not angry! The software is great and I appreciate what you’re all doing.

[nickstarboy]

Ive returned to 003, switched to hyperbolic fit (as an experiment - my god, it works so well!) and changed the number of ‘autofocus initial offset steps’ up to 4, so its less likely it finishes on a jump to the right hand side of the curve again - hopefully that was the problem. anyway,working beautifully right now.

‌

[Stefan B]

Issue should be fixed in #006

[nickstarboy]

Thanks Stefan! Still working perfectly as we speak…..

[nickstarboy]

FYI, I was wrong about it hanging after moving back to the right of the curve. It didnt do so in this case, and hung anyway.