STM32H723: Remove 1.65 V bias after ADC conversion for DC voltage measurement

[Pandian]

MCU: STM32H723VET6
ADC: ADC1 / ADC2 / ADC3 (12-bit, single-ended)
VREF: 3.3 V
Application: Digital Multi meter (DC / AC voltage measurement)

Problem Description

In my DMM hardware design, the Analog frontend applies a mid-scale level shift (VREF/2 = 1.65 V) before the ADC input.

This is done intentionally to support bipolar AC measurements using a single-ended ADC:

Input signal → amplifier/divider → +1.65V offset → ADC (0–3.3V)

As expected:

• 0 V input → ADC reads ~1.65 V

• Positive/negative AC swings are in around mid-scale

However, for DC voltage measurement, I want the behaviour to be:

• No input connected → display 0.000 V

• Applied DC voltage → display the same voltage

• Mid-scale offset should NOT appear in the final DC reading

---

Observed Issue

If the mid-scale offset is not removed after ADC conversion, the firmware reports:

• ~3 V on 6 V range

• ~30 V on 60 V range

• ~300 V on 600 V range

even when no voltage is applied.

This is because the ADC correctly reads ~1.65 V, but that bias is still present during DC scaling.

Solution Implemented:

float Apply_Voltage_Calibration(float v, Voltage_Range_t range, uint8_t phase)
{
  Calibration_Point_t* cal = &user_calibration.voltage[range][phase-1];
  if (cal->gain == 0.0f)
    cal->gain = 1.0f;
  v -= cal->offset;
  v *= cal->gain;
  return v;
}

used in this below function

float Measure_Voltage_DC(uint8_t phase, Voltage_Range_t range)
{
  // Auto-range
  if(range == VOLT_RANGE_AUTO) {
    range = Voltage_Auto_Range(phase);
  }

  // Configure voltage gain, divider, enable frontend
  Configure_Voltage_Range(phase, range);

  // Select voltage MUX path
  MUX_Select_All_Voltage();

  // Enable correct voltage input channel (phase 1–3 -> channels 4–6)
  Channel_Enable(phase + 3, true);
  HAL_Delay(10); // settling

  // Choose the correct ADC instance & channel
  ADC_HandleTypeDef* adc = NULL;
  uint32_t adc_channel = 0;
  switch(phase)
  {
    case 1: adc = &hadc3; adc_channel = ADC_CHANNEL_10; break; // V1 -> ADC3 CH10 (PC0)
    case 2: adc = &hadc1; adc_channel = ADC_CHANNEL_16; break; // V2 -> ADC1 CH16 (PA0)
    case 3: adc = &hadc2; adc_channel = ADC_CHANNEL_3; break; // V3 -> ADC2 CH3 (PA6)
    default: return 0.0f;
  }

#define DC_SAMPLES 64
  uint16_t samples[DC_SAMPLES];

  // Acquire samples (same logic, but HAL)
  for(int i = 0; i < DC_SAMPLES; i++)
  {
    samples[i] = ADC_Read_Single_Channel_HAL(adc, adc_channel);
    HAL_Delay(1); // ~100us minimum tick
  }

  // Compute average
  uint32_t sum = 0;
  for(int i = 0; i < DC_SAMPLES; i++)
    sum += samples[i];

  float adc_avg = (float)sum / (float)DC_SAMPLES;

  // Convert ADC average -> voltage
  float voltage = ADC_To_Voltage_DC(adc_avg, range);

  // Apply calibration
  voltage = Apply_Voltage_Calibration(voltage, range, phase);

  // Disable channel after use
  //Channel_Enable(phase + 3, false);
  return voltage;
}

This is an issue i am facing, kindly help me to get out of this problem.

---

Edited to apply source code formatting - please see How to insert source code for future reference.

[TDK]

Use a pull resistor to set the voltage to what you want when nothing else is connected. “Not connected” is not a voltage level that can be read. The adc can only read what is on the pin.

[RobK1]

The best solution would be to make your frontend switchable for AC/DC, so that no offset is applied in DC mode. Although all the DMM's I have ever worked with can measure both positive and negative voltages in DC ...

The simplest solution, of course, is to just subtract the half-scale value from your raw ADC readings.

Oh, and please use this option to insert readable code: