Yet another ADC question

[Valer_I]

Hello, My device is based on nRF51822 SD110 v8.0. Softdevice is enabled, but no BT functionality used. I measure 2 different voltages with ADC each second with two channels. It works, but after each 4096 seconds it measure incorrect (lowered) values for next 4096 seconds. And latches back to correct measurement...
(Grid divide by 4096s). Device is powered from line adaptor and supply voltage is constant. The only thing I found that could be participal is NRF_RTC1 timer that I use to count seconds.

 NRF_RTC1->CC[1] =     4096;
 NRF_RTC1->PRESCALER   = (8-1);

and take to count seconds - NRF_RTC1->COUNTER/4096

Any idea?

[Ole Bauck]

Do you do anything in the code when the counter reaches CC[1]? Possible reason for the wrong measurements could be that system draws more current and the diode between the supply and Vdd has higher voltage drop. Have you measured Vdd to see if this is also constant?

[Valer_I]

Hi, the logics is simple, at least I didn't find anything current consuming. I configured ADC for Vbg reference and 1/3 prescalling. So it should measure up to 3.6v input voltage. What link with protection diode drop? I will measure VDD to check, but only thing I could imagine is if it heating much and Vbg zenner drifts.

[Ole Bauck]

Sorry for confusing you, I meant the diode on the nRF51 DK. On the nRF51 DK there is a diode between the supplies (pin, battery or 3.3V regulator) and the Vdd, see the schematics in the nRF51 DK Hardware Files. Are you using nRF51 DK or other hardware?

[Valer_I]

No, we developed our own device. It is battery powered, but for the test I used wire connection to AC-DC adaptor + 3.3V LDO regulator. So I continue investigation)

[Ole Bauck]

Ok, then that is not the case.

Could you try to not put the device to sleep (not call sd_app_evt_wait()) and see if the result is the same. You can also try to keep the HFCLK on all the time with:

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));
}

And see if the result is still the same.

Just wondering: You say that you use line adaptor, but the graph looks very much like the discharge curve of a battery (if you don't look at the erroneous measurements). You are sure that the graph is not when measuring battery?