NRF51822 ADC float up 0.2V

[Songplus]

Hi, Now I use the NRF51822 ADC to sample VCC , 1.2V internal reference，internal prescaling 1/3.The PCB connection show as the pictrue .My application develop base on soft device-6.0 .The ADC starts to sample VCC when the application is waked up from sd_app_evt_wait(). When connect the J-Link to debug the application,the ADC samples VCC correct ,once removing ,the ADC samples VCC voltage is higher 0.2V than the reality voltage,(eg. The VCC is 3.3V ,in debug model,ADC sample value is 3.29V,but in removing model, the result is 3.5V ).Is in the debug model the CPU running at full speed? I have tried testing the ADC works at CPU full running ,but the result is still wrong.

What does the picture mean?Is it saying that the AIN port voltage never can't be over 2.4V when using internal or external prescalling? By the way,use external crystal oscillator the accuracy is better than before,but the error still exsit,especially sampleing a high lever voltage(3.25V is sampled as 3.50V).

[Ole Bauck]

This may be because of the HFCLK is running of RC oscillator. Then the accuracy will be as bad as the accuracy of the RC oscillator. Try to sample with the crystal running. Turn it on with this code (can be done once):

uint32_t p_is_running = 0;
sd_clock_hfclk_request();
while(! p_is_running) {                             //wait for the hfclk to be available
    sd_clock_hfclk_is_running(&p_is_running);
}

[Songplus]

Thanks.It is better than before,but error is beyond acceptable.The NRF5122's ADC is too bad to minitor a battery charging.

[Ole Bauck]

The voltage at AIN after prescaling should not be more than 2.4V, so for example if you use prescaler 2/3 the maximum voltage can be 2.4V / (2/3) = 3.6V, see here.

What are the resistor values of the voltage divider? Using an external voltage divider is not that straight forward, see this blogpost.

The ADC of nRF51 should be adequate for battery measuring.