Allow for multiple consecutive images with a HFR increase before AF

[Former user]

I am wondering if it would make sense to allow for a specified number, 1-3, of frames with an increase in HFR before the AF routine is initiated. I often get a wild HFR frame, maybe due to wind, turbulence, jet stream etc and would not want to initiate AF after each one of these occurrences. But if a couple or three happened in a row then yes, trigger the AF routine. This is especially relevant for shorter exposures, where the HFR may jump every few 120s frames. I think this would greatly enhance this option in the sequencer tab. It would also allow for it to remain at 1 for those who are using that way.

[nickstarboy]

agreed. this is why i dont use it. Some kind of moving average so we could specify 1 or 10 subs, for example.

‌

I use 10sec subs so HFR varies considerably from one to the next, but trends are still clear at the same amplitude as the sub-to-sub variation.

[Ross Walker]

Agreed. A moving average would be a more robust approach to HFR-based refocus. The time variable should perhaps not be the number of frames because a sequence can contain frames of various lengths, so a user-selectable time period might be more robust I believe. Something like an x% increase in HFR over y minutes might be workable?

[Yannick Dutertre]

I'm thinking about something like that:

• Get the all frames post-AF (if # frames post AF >= 4)

• Using those frames, we have point coordinates (frame #, HFR). Could even add the intra-frame HFR StdDev to that.

• Based on the point coordinates, perform a linear regression, potentially weighted by 1/StdDev (like the curve fitting in the AF routine) - Now we have a line. If HFR on that line at current Frame # exceeds HFR on that line at first post-AF Frame # by X%, trigger the AF run. Otherwise keep going

I think that would be far more robust than a simple sliding window running average. Negative part is that it would stay idle for at least four frames - but I assume that people using that feature wouldn’t have 15 minute long frames.

[nickstarboy]

sounds good Yannick!

[Yannick Dutertre]

The enhancement is in nightly #58. Hopefully it will help.

[nickstarboy]

Thanks man, looks forward to trying it….

‌

[Ross Walker]

Excellent! Thank you Yannick.

[Ara Jerahian]

I’m not able to download #58 at this time since I don’t have access to my imaging PC, but can this enhancement be enabled/disabled?

[Yannick Dutertre]

This is an auto-on enhancement - it can’t be disabled. Should be quite robust though.

[nickstarboy]

Hi Yannick - I had a quick look and cant see a way to change it from 4 points, is that possible? I would want to use at least 20-100 as I use 10sec subs and eposides of poor seeing occur which i would like to disregard.

[Yannick Dutertre]

It is a minimum of four points. If it triggers at 4 points, it’s that it detected an increase trend above the percentage value specified. Do you have data you can provide in terms of HFR progression?

‌

[nickstarboy]

Ah I see what you mean…its a line fit to all the data since the last focus, not a ‘running’ fit. Assuming it ‘drifts’ out of focus, rather than a jump?

‌

Will see what I can do about HFR data!

[nickstarboy]

HFR history at right. This is an extreme example, usually it climbs up a but for a dozen or so subs then down again. This is the start of the jet stream coming in. I gues it illustrates that sometimes you just have to make compromises…theres no way the software could distinguish this from a system creeping out of focus. Until it got better again.

‌

The guiding was fine throughout.

[Yannick Dutertre]

Yeah there’s not much that can be done for that use-case - the software doesn’t know it’s going to go down again, it can only assume a trend that goes up.