Re: [PATCH v2 1/1] media: i2c: ov02c10: Keep power on and use reset for power management

[Bryan O'Donoghue <bryan.odonoghue@linaro.org>]

On 26/01/2026 11:23, Saikiran B wrote:
> "Where do you get this conclusion from ? Are you inferring it from
> what you see on the platform or can you point to some known
> data-source for this ?"
> 
> This is determined on the Lenovo Yoga Slim 7x (X1E80100). I tested
> extensively and found that if I attempt to power-on the sensor less
> than ~2.3 seconds after power-off, it fails to identify or times out
> on I2C (brownout behavior). If we wait >2.3s, it works reliably 100%
> of the time.

I don't think we've established the regulator is at fault. That's the 
feedback I'm giving you here.

I think it is far, far, far more likely the power-on sequence of the 
sensor needs tweaking.

> 
> "2 seconds to discharge ? These regulators are PM8010 anyway - so
> you're saying the PMIC takes two seconds to discharge ?"
> 
> Yes. I checked the regulator driver
> (drivers/regulator/qcom-rpmh-regulator.c) and found that unlike other
> Qualcomm regulator drivers (e.g., spmi/glink), it currently lacks
> active_discharge / pull-down support. Without active discharge, the
> voltage rails float and decay very slowly via leakage current when the
> load (sensor) is in reset/high-Z.


Right so looking at the power for this part we have:

&cci1_i2c1 {
	camera@36 {
		compatible = "ovti,ov02c10";
		reg = <0x36>;

		reset-gpios = <&tlmm 237 GPIO_ACTIVE_LOW>;
		pinctrl-names = "default";
		pinctrl-0 = <&cam_rgb_default>;

		clocks = <&camcc CAM_CC_MCLK4_CLK>;
		assigned-clocks = <&camcc CAM_CC_MCLK4_CLK>;
		assigned-clock-rates = <19200000>;

		orientation = <0>; /* front facing */

		avdd-supply = <&vreg_l7b_2p8>;
		dvdd-supply = <&vreg_l7b_2p8>;
		dovdd-supply = <&vreg_cam_1p8>;

		port {
			ov02e10_ep: endpoint {
				data-lanes = <1 2>;
				link-frequencies = /bits/ 64 <400000000>;
				remote-endpoint = <&csiphy4_ep>;
			};
		};
	};
};

// qcom standard RPMh -> PMIC LDO regulators
// these are not the droids you are looking for
vreg_l7b_2p8: ldo7 {
	regulator-name = "vreg_l7b_2p8";
	regulator-min-microvolt = <2800000>;
	regulator-max-microvolt = <2800000>;
	regulator-initial-mode = <RPMH_REGULATOR_MODE_HPM>;
};

// this OTOH
vreg_cam_1p8: regulator-cam-1p8 {
	compatible = "regulator-fixed";

	regulator-name = "VREG_CAM_1P8";
	regulator-min-microvolt = <1800000>;
	regulator-max-microvolt = <1800000>;

	gpio = <&tlmm 91 GPIO_ACTIVE_HIGH>;
	enable-active-high;

	pinctrl-0 = <&cam_ldo_en>;
	pinctrl-names = "default";
};

Dell has used - likely reused - part of the x86 design in the qcom 
implementation - and toggles 1v8 via a GPIO directly.

If your theory about brown-out is correct then

vreg_cam_1p8: regulator-cam-1p8 {
	// add this
	off-on-delay-us = <20000>;
};

Then please let us know how she goes.

---
bod