[PATCH 0/3] Reduce rkisp1 flicker on first start
Kieran Bingham
kieran.bingham at ideasonboard.com
Sun Oct 20 19:33:26 CEST 2024
Quoting Mikhail Rudenko (2024-10-20 14:39:45)
> Hi Kieran,
>
> On 2024-10-18 at 23:26 +01, Kieran Bingham <kieran.bingham at ideasonboard.com> wrote:
>
> > Hi Mikhail,
> >
> > Quoting Kieran Bingham (2024-10-17 14:35:40)
> >> Hi Mikhail,
> >>
> >> Quoting Mikhail Rudenko (2024-10-17 13:47:04)
> >> > The chart from camshark (see attachment).
> >> >
> >>
> >> That looks like a good result indeed!
> >>
> >> I'll test out the patches.
> >
> > I've just tried this out - and it seems to work well on the IMX283 I
> > have - but caused quite noticible AnalogueGain oscillations on the
> > IMX335 ... I would 'bet' that's not the fault of this series - but the
> > fact that we don't have per-sensor delayed-controls set up correctly
> > which I already know about - but I haven't seen that get fixed up yet.
>
> I did a few more tests myself (I use OV4689 btw), and also observed
> analogue gain oscillations on darker scenes when AGC tries to steer the
> gain above 1.0. It looks like the hardcoded gain delay value of 1 is
> wrong at least for my sensor. When I increased it to 2, gain
> oscillations were gone. I have also tweaked a few things here and there
> and will send v2 soon.
>
> > I think your series probably highlights the fault that already exists...
>
> I feel like we have reached the moment when we should switch from the
> delays hardcoded in the pipeline handler to sensor specific values like
> rpi does. I could try doing that, but in a separate series. What do you
> think?
Oh - absolutely. If it's something you're willing to work on - I think
that would be a massive help. Indeed, I think that's a separate series.
--
Kieran
>
> --
> Best regards,
> Mikhail
More information about the libcamera-devel
mailing list