Imx6ull ADC functioning

[RiccardoC_DLK]

Good morning,
I have some doubt about ADC functioning because we read always an offset (about 10%, not fixed in absolute value) when we read an external sensor.

We configured ADC1 in our DT to use gpio1.00 and gpio1.01.

&adc1 {
	num-channels = <10>;
	vref-supply = <&reg_module_3v3_avdd>;
	pinctrl-names = "default";
	pinctrl-0 = <&pinctrl_adc1_dlk>;
	status = "okay";
};

&iomuxc {
    imx6ull-colibri {
		pinctrl_adc1_dlk: adc1grp-dlc {
			fsl,pins = <
				MX6UL_PAD_GPIO1_IO00__GPIO1_IO00        0x3000 /* SODIMM 8 */
				MX6UL_PAD_GPIO1_IO01__GPIO1_IO01        0x3000 /* SODIMM 6 */
			>;
		};
    };
};

HW circuit is composed of on 150 ohm resistor connected to gpio pin to measure the voltage.
Then, if we simulate an external sensor generating 4mA / 8 mA / 12mA / 16mA / 20mA, we read values in figure:

But we are expecting values like:
current - raw value (current * 150 ohm)
4 mA - 600
8 mA - 1200
12 mA - 1800
16 mA - 2400
20 mA - 3000

We measure the resistor and it is 150 ohm for sure.

Do we need to calibrate the ADC? or change some parameters?

Other unresolved question is what is the ‘in_voltage_scale’ value which is stored into the file in figure.
We understood it is related to reg_module_3v3_avdd declared in DT, but it seems it is not used. We tried to change reg_module_3v3_avdd = <4096000> to obtain in_voltage_scale = 1. But ADC raw values remain the same.
So, which is the functioning of in_voltage_scale value?

Thanks.

I work with:

• imx6ull trdx module;
• colibri evaluation v3 board;
• our custom carrier board;
• Linux BSP 5.6
• Kernel 5.4-2.3

[henrique.tx]

Hi @RiccardoC_DLK !

Could you please share the output of tdx-info (Getting Device Information with Tdx-Info | Toradex Developer Center)?

@RiccardoC_DLK

	num-channels = <10>;

Seems like the ADC you are using is the i.MX6ULL SoC one, right? If yes, from where you got this num-channels property?

Also, please share the HW schematics that you are using to generate the voltage. The ground connection is really important: which ground are you using?

Best regards,

[RiccardoC_DLK]

tdx-info

unfortunately this command works from BSP 6. I am using 5.6.

Seems like the ADC you are using is the i.MX6ULL SoC one, right? If yes, from where you got this num- channels property?

I found it into: imx6ull-colibri.dtsi. https://git.toradex.com/cgit/linux-toradex.git/tree/arch/arm/boot/dts/imx6ull-colibri.dtsi?h=toradex_5.4-2.3.x-imx
At the moment we need only two ADC inputs. Should I change it to 2?

HW schematics

In our schematic AI1_4-20mA is directly connected to external connector, while AI1_0.6-3V is connected to pin sodimm8 of colibri SoM. GND is standard for the entire board. In between there is a component for protecting the uC. You see the 150ohm resistance.

[DaveM]

• Do you have a separate supply for your ADC going into pins 10/12 of the colibri? You’ll need something filtered going in there (not the standard 3.3 from the entire board).
• I’ve also never seen a design without some sort of opamp between your output voltage and the colibri adc input.
• You could also back up and just feed a direct voltage into the pin without the external components and see what you get.

[henrique.tx]

Hi @RiccardoC_DLK !

@RiccardoC_DLK

unfortunately this command works from BSP 6. I am using 5.6.

Actually, that is not true. Please read the Prerequisites section of the article.

@RiccardoC_DLK

I found it into: imx6ull-colibri.dtsi. imx6ull-colibri.dtsi « dts « boot « arm « arch - linux-toradex.git - Linux kernel for Apalis, Colibri and Verdin modules

My mistake here. Sorry. You said BSP 5.6 before, but I mistakenly checked in Linux Kernel version 6.1 instead of 5.4. You are correct.

@RiccardoC_DLK

At the moment we need only two ADC inputs. Should I change it to 2?

I don’t think this would cause problems. But you can try anyway.

And thanks for sharing the schematics. We will take a look into it and get back to you.

Best regards,

[Edward]

@RiccardoC_DLK

But we are expecting values like:
current - raw value (current * 150 ohm)
4 mA - 600
8 mA - 1200
12 mA - 1800
16 mA - 2400
20 mA - 3000

These are expected millivolts. in_voltage_raw are raw ADC counts, so you should expect
mA*150 / Vref[mV]*4096 (2^12)
And then you get less than expected. One of reason is how you set up IOMUXC. 0x3000 - is enable 100 kOhm pull down, which using your filter resistor gives 100/(10+100) voltage divider. So for 16mA you should read 2978 without divider and 2708 taking divider into account. Pull resistance values aren’t very precise, so 2708 vs 2650 is pretty close.
Try disabling pull device with 0x0 instead of 0x3000 in pad settings.

@RiccardoC_DLK

what is the ‘in_voltage_scale’

It is the number to multiply raw value by to get voltage in mV units.

Regards
Edward

[RiccardoC_DLK]

These are expected millivolts. in_voltage_raw are raw ADC counts, so you should expect
mA*150 / Vref[mV]*4096 (2^12)
And then you get less than expected. One of reason is how you set up IOMUXC. 0x3000 - is enable 100 kOhm pull down, which using your filter resistor gives 100/(10+100) voltage divider. So for 16mA you should read 2978 without divider and 2708 taking divider into account. Pull resistance values aren’t very precise, so 2708 vs 2650 is pretty close.
Try disabling pull device with 0x0 instead of 0x3000 in pad settings.

Perfect, I was expecting a voltage raw value. Also disabling PullUp resistor I obtain values closer to expected voltage values.
Thank you.