Re: [PATCH v2 3/6] arm64: dts: qcom: glymur: Add camera MCLK pinctrl

[Jesse Casco]

On Thu, Oct 15, 2026 at 04:22:28PM +0530, Nihal Kumar Gupta wrote:
> +			cam_mclk4_default: cam-mclk4-default-state {
> +				pins = "gpio100";
> +				function = "cam_asc_mclk4";
> +				drive-strength = <2>;
> +				bias-disable;
> +			};

I've been running this series on an ASUS Zenbook A16 (UX3607OA), which
has an OV02C10 on MCLK4 at 19.2 MHz. With 2 mA here the sensor shows
horizontal banding, and its frame period stretches by about 9%, whenever
the internal DMIC clock (LPASS LPI GPIO6, 2.4 MHz) is running. The bands
are already in the RAW10 frames, and the sensor test pattern is clean,
so it looks like the MCLK pad itself rather than CSI or the ISP.

Changing only this pin to 6 mA fixes it. A/B/A across three boots, same
kernel, fixed exposure/gain, row high-pass RMS / frame period:

  MCLK4   mic off          mic on            after
  2 mA    0.96 / 33.28 ms  20.44 / 36.17 ms  1.04 / 33.27 ms
  6 mA    0.89 / 33.27 ms   0.90 / 33.27 ms  0.90 / 33.27 ms
  2 mA    0.95 / 33.27 ms  15.33 / 34.37 ms  0.93 / 33.28 ms

At 6 mA the frame timing holds steady at 33.27 ms and the image stays
clean in every phase, including while the microphone is recording,
which is exactly the case that couples interference into the 2 mA
clock. Normal camera use with the mic on is also clean at 6 mA, so the
fix holds up on real hardware, not just in the capture script.

Could cam_mclk4_default use 6 mA? If you'd rather keep 2 mA in the SoC
dtsi, I'll override it in the A16 board DT instead.

Thanks,
Jesse