T113 TCON Top tinting troubleshooting

[Jakob L]

Hi John,

i am pretty sure other devices don't have this. At least i have never
seen it on V3s, A20, A33, H3, A64 with RGB, HDMI, LVDS, or Mipi. This
was the first - and on MIPI its pretty bad, to a point where its not
usable.
With the allwinner u-boot from BSP i had not experience that tint -
but i assumed it to be somewhere in the wrong Mipi packager or DMA
setup.

Since i am not very experience with mainline contributions i am not
sure what is the best approach. While i agree that ideally its done
right on all sunxi devices, the risk of breaking something without
need should be avoided?

All the best
Jakob

Am Sa., 29. Juni 2024 um 23:01 Uhr schrieb John Watts <contact at jookia.org>:
>
> Hi,
>
> On Sat, Jun 29, 2024 at 12:14:39PM +1000, Kirby Nankivell wrote:
> > Tested this on a MangoPI-MQ-R with a RGB lcd and can confirm that this fix
> > removes the green tint I experienced.
> >
> > Disabling the mixer1 in DTS entirely also works.
> >
> > I suspect on the basis that the Mixer/TCON-TOP topology is identical to the
> > R40 but does not seem to experience this issue, that is a
> > silicon errata specific to the T113/D1s as far as we currently know.
> >
> > However, there is no functional reason from a drm perspective to require
> > more than one mixer tied to one output via tcon-top - unless someone has a
> > good argument here?
> >
> > in my opinion:
> > 1) implement this as a quirk on the t113 and fix mixer0 > rgb/dsi and
> > mixer1 > tv0
> > 2) make this a global default for the sun8i tcon top driver where only one
> > mixer is initialised against an output (which would also impact the R40)
>
> The only reason I caught this issue was because I implemented LCD output in
> U-Boot and saw it looked a little better- I thought the tint was the backlight.
> Had you not contacted me and had it not affected the person I'm doing
> work for I would not have bothered looking in to this.
> It's entirely possible that the other supported devices have this issue,
> or that future ones will and nobody will notice.
>
> Going with a preventative approach here by fixing it for all potential devices
> and relying on the behaviour the sunxi code assumes rather than the datasheet
> says seems like the best approach to me.
>
> John.