STM32 ADC Scan Mode – Getting Unrealistically Values from Multiple NTC Channels

[durna]

Hello,

I’m trying to measure temperature using NTC sensors on four different ADC channels of my STM32 Nucleo‑H7 board. I could previously connect potentiometers to the same pins and get flawless ADC readings with the same code and CubeMX settings. I only added a function to convert NTC readings to temperature and didn’t change any other configuration.

I connected the NTCs to PA6, PA7, PC4, and PC5. In my loop, I call HAL_ADC_Start(), then four times HAL_ADC_PollForConversion() and HAL_ADC_GetValue() to retrieve the raw ADC values. When I measure the sensor output directly with a multimeter, I see about 1.65 V, but the ADC reads ~0.03 V, resulting in completely unrealistic temperatures.

I never had this issue with potentiometers on the same pins and code. Why am I getting such low readings with the NTCs? Is my channel assignment wrong, the pin mode incorrect, or could hardware jumpers/shielding be affecting it?

Has anyone experienced something similar or have suggestions? Thanks!

[Saket_Om]

Hello @durna 

Did you performe ADC calibration before measuring the temperature? 

[durna]

Hello @Saket_Om 

yes as it appears in the code

[Ozone]

Input impedances and sample times play an important role as well.
What impedances and currents/voltages we are talking about here ?

[durna]

The following values are valid in my setup:

Divider impedance (Thevenin): 10 kΩ (NTC) ∥ 10 kΩ (fixed) ≈ 5 kΩ

Divider current I = 3.3 V / (10 k + 10 k) ≈ 165 µA

Node voltage (multimeter): ~1.67 V ⇒ ADC raw value ≈1.67/3.3-65535 ≈ 33 000

Sampling circuit time constant:

S/H capacitor ≈ 4 pF, switch R_ON ≈ 100 Ω ⇒ τ = R-C ≈ 5 kΩ-4 pF ≈ 20 ns

ADC sampling times:

I actually ran the ADC at a 15 MHz clock with a ÷10 prescaler, so the ADC clock is 1.5Mhz

so the S/H capacitor must be fully charged. Despite this, the raw reading is ~600 (≈ 0.03 V)

[durna]

Everything seems to be correct both hardware and software wise, but the temperature readings are really ridiculous. I've been trying for 2 days and still can't figure it out

 

[TDK]

Do values change if you increase sampling time to max?

Use DMA to convert multiple channels, but maybe get a single channel working first.

[Ozone]

> Use DMA to convert multiple channels, but maybe get a single channel working first.

I would definitely second that.
Perhaps simplify it even more, using continuous conversion for the single channel.

 

> Node voltage (multimeter): ~1.67 V ⇒ ADC raw value ≈1.67/3.3-65535 ≈ 33 000

What do you mean here ?
The part marked in red is mathematically not correct, there is something like operator precedence in mathematical equations, too.

You probably mean "1.67/3.3 * 65535", i.e. scaling the voltage to a unsigned 16-bit value.

[durna]

I've solved the problem. It was in CubeMX settings. Thanks for everything