Gain oscillations about libcamera HOT 4 CLOSED

Interesting observations, Laurent has posted a viewfinder scaling patch, which when integrated (small merge conflicts on the rpi branch) and when scaling the oscillations go away. That might be some sort of timing feedback issue perhaps? I.e. when the performance is reduced it affects the control loops.

from libcamera.

from libcamera.

That's interesting. We have seen oscillations in the past, but I think they should all be fixed (particularly with the imx219, we had some recent issues with the imx477).

The observation that it happens when system is loaded suggests that we might be dropping frames and upsetting the exposure/gain pairs of writes somehow. Although with group hold, etc. I can't see how that can happen.

Any chance you can reproduce this and record a video with a phone and upload for @davidplowman and me to have a look at?

from libcamera.

This isn't visible on the latest firmware/kernel/libcamera - so likely I had a version mis-match and something wasn't happy.

from libcamera.