voltage divider and adc input impedance (Gain bug)

How do I take in consideration the "adc input impedance" to proper mesure the voltage on a voltage divider?

What are the recommended values for the DUE?

Read the bellow to understand the problem...

I'm calling the "Gain bug", the Gain or the multiplier that is used to correct the measure voltage from the voltage divider.

I have set 6 adc ports to read the same input, in order to have a more accurate measure.

I have tested 3 combination of voltage divider, 100k + 10K, 10k + 1K and 4.7k + 1K

The maximum voltage in this case is 12v but I wanna to use up to 18v.

The problem I'm having is that the Gain multiplier that I have to use is not a fixed value.

For example in the 100 + 10 or 10 + 1 the Gain is around 11 (I have properly mesure the resistor to make shore what the real value is on my case 10.985). But on my test I need to set it as low as 7.5 to measure a input voltage of 0.5v and as high I go I need to make the Gain bigger and bigger even above the 11 that I calculate.

Here is a table with the values from the 10 + 1 voltage divider, the 100 + 10 are very similar and the 4.7 + 1 is better in the relation that the diference from the lower and the higher gain is lower (5.1 to 5.738).

So I assume I have a issue in relation to the adc input impedance here, I read the section "45.7.2.1 Track and Hold Time versus Source Output Impedance" of the datesheet but I don't get how can I apply those values to understand what I have to do to calculate this in the proper way

I use a simple spreadsheet to calculate a formula for a "trend line" and use that to adapt on my code, but is not very precise as I wish it to be...

Here is my code in case helps is a simple DC to DC buck converter

#define PERIOD 420 // 84MHz/420 = 200Khz // 84MHz/336 = 250Khz // 84MHz/280 = 300Khz // 84MHz/240 = 350Khz
#define Vref 10.25 // the target Vout of the sistem

#define OFFSET 3.3 / 4096 // ADC is 3.3V base with resolution of 0 to 4095 bits

// This is a approximation function for the gain, will give a + or - 0.045 accuracy for the output voltage
#define GAIN (Vref / ((109.68036 * Vref + 25.99648) * OFFSET))
#define MULTIPLIER (OFFSET * GAIN) / 6

//#define CLOCK (84000000 / PERIOD)

void setup () {
  //Serial.begin(115200);

  // ADC configuration
  PMC->PMC_PCER1 |= PMC_PCER1_PID37;   // ADC power on
  ADC->ADC_CR = ADC_CR_SWRST;          // Reset ADC
  ADC->ADC_MR |=  ADC_MR_FREERUN;      // Enable free run
  ADC->ADC_ACR = ADC_ACR_IBCTL(0b01);  // Only if 1 MHz > conversion frequency > 500 KHz

  ADC->ADC_CHDR = 0xFFFF; //Disable all adc channels
  ADC->ADC_CHER |=  ADC_CHER_CH2 | ADC_CHER_CH3 | ADC_CHER_CH4 |
                    ADC_CHER_CH5 | ADC_CHER_CH6 | ADC_CHER_CH7; // Enable Channels 2 to 7 physical port A0 to A5

  // PWM configuration
  REG_PMC_PCER1 |= PMC_PCER1_PID36;                    // Enable PWM
  REG_PIOC_ABSR |= PIO_ABSR_P24;                       // Set the port C PWM pins to peripheral type B (PWM7 physical port is PWM6)
  REG_PIOC_PDR |= PIO_PDR_P24 ;                        // Set the port C PWM pins to outputs           (PWM7 physical port is PWM6)
  REG_PWM_CLK = PWM_CLK_PREA(0) | PWM_CLK_DIVA(1);     // Set the PWM clock A rate to 84MHz (84MHz/1)

  PWM->PWM_CH_NUM[7].PWM_CMR =  PWM_CMR_CPRE_CLKA;     // Enable single slope PWM and set the clock source as CLKA
  PWM->PWM_CH_NUM[7].PWM_CPRD = PERIOD;                // Set the PWM period register 84MHz/(PERIOD)= Freq
  PWM->PWM_CH_NUM[7].PWM_CDTY = 0;                     // Duty recommended starting at 0 or (Vref/Vin) * PERIOD as the Vin is variable value not using here

  PWM->PWM_IER1 = PWM_IER1_CHID7;                      // Interrupt on end of counter period (CPRD)
  NVIC_EnableIRQ(PWM_IRQn);                            // Enable PWM interrupt

  REG_PWM_ENA = PWM_ENA_CHID7; // Enable PWM CH 7
}

void loop() {}

void PWM_Handler() {
  //static uint32_t Count;
  //static float Bits;

  static float Vout, uk, uk_1, uk_2, xk, xk_1, xk_2;
  static int DUTY;

  PWM->PWM_ISR1; // Clear PWM status register

  //Reads and convert adc values to a voltage reading
  Vout = (ADC->ADC_CDR[7] + ADC->ADC_CDR[6] + ADC->ADC_CDR[5] + ADC->ADC_CDR[4] + ADC->ADC_CDR[3] + ADC->ADC_CDR[2]) * MULTIPLIER;

  //For testing
  //Bits = (ADC->ADC_CDR[7] + ADC->ADC_CDR[6] + ADC->ADC_CDR[5] + ADC->ADC_CDR[4] + ADC->ADC_CDR[3] + ADC->ADC_CDR[2]) / 6;
  //Vout = Bits * MULTIPLIER;

  //Type II Compensator, a feedback duty calculator for DC-DC converter, using the bilinear (Tustin) discretization method to generate the time function of the PID transfer function
  xk = Vref - Vout; // Error

  // Bilinear (Tustin) discretization method of the PID function of this buck
  uk =  44.8983097883 * xk + 2.7027784604 * xk_1 - 42.1955313279 * xk_2 + 0.4714280152 * uk_1 + 0.5285719848 * uk_2;

  DUTY = uk * PERIOD / 8;

  if (DUTY > PERIOD) DUTY = PERIOD; //Prevent DUTY bigger then PERIOD
  else if (DUTY < 0) DUTY = 0; //Prevent DUTY lower then Zero

  PWM->PWM_CH_NUM[7].PWM_CDTYUPD = DUTY; //Set the DUTY

  //Set the previous values
  xk_2 = xk_1; //xk(t - 2)
  xk_1 = xk;   //xk(t - 1)
  uk_2 = uk_1; //uk(t - 2)
  uk_1 = uk;   //uk(t - 1)

  // To print testing values
  //  Count++;
  //  if (Count == CLOCK) {
  //    Serial.print(" Vout: ");
  //    Serial.println(Vout);
  //    Count = 0;
  //  }
}

Go on. It is simple maths.

18V means that the divider must be 1 : 5.5 e.g. 18V gives 3.27V which is below 3.3V

Using 10k and 2k2 will give a suitable ratio.

The ADC input impedance is in parallel with the 2k2. For example 1M0 input impedance is not going to make much difference.

Typical divider values:
10k w 2k2
47k w 11k
100k w 22k
The ADC sees "source impedance" of 2k, 10k, 20k. I would advise using less than 20k.

David.