nrf 51822 IO values not as expected

[Francois]

Good morning all,

We have a circuit based on 51822 chip.

We are currently trying to monitor some value from 0v to 24v from the analog inputs. It's working, however, we have found 2 issues.

1. The values we get are between 74 (0V) and 496 (24V) The precise formula is Y = 17,554X + 74,443 with Rquare = 1 where X is the voltage input value and the analog value we get from the chip.

2. Sometimes we need to call the function 2 times in order to get the right analog value for a given voltage. (could it be a buffer problem?)

See image below. I have traced this curve from several measurements.

Where this could come from?

To get the 24v down to a reasonable voltage we are using this voltage divider (370k, 2.2M)

To get the value from an input, we are using this formula:

 U16 AINS_GetValue(U8 innum)
 
 {
  //Clear Event flags
 nrf_adc_conversion_event_clean();

 //Read Sample
 AIns.ADC_Channel[innum].ADC_Buf[AIns.ADC_Channel[innum].State.Bits.Index++] = nrf_adc_result_get();

 //Find average
 AIns.ADC_Channel[innum].ADC_Avg = 0;
 for(U8 i = 0; i < 4; i++) AIns.ADC_Channel[innum].ADC_Avg += AIns.ADC_Channel[innum].ADC_Buf[i];
 AIns.ADC_Channel[innum].ADC_Avg >>= 2;

 //Return value
 return AIns.ADC_Channel[innum].ADC_Avg;

 }

where both functions (nrf_adc_conversion_event_clean(); nrf_adc_result_get(); ) are from the Nordic Semi Ble UART application code.

Could it be my the //Read Sample part of the code the problem?

ADC initialization seems to be OK with the right scaling/resolution/reference but maybe I'm wrong... Inputs # are good since these values are the same for the 4 inputs.

 void AINS_Init(void)
 {
 U32                      pCompareEvent;
 U32                      pADCTask;
 nrf_ppi_channel_t        PPIChannel;
 ret_code_t               error;
 U32                      TimerTicks;
 nrf_adc_config_t         ADCConfig = NRF_ADC_CONFIG_DEFAULT;

 //Initialize AINS structure  
 AIns.Current.Number = 0;
 AIns.ADC_Channel[AIN1].hwChannel = AIN1_HWCHN;
 AIns.ADC_Channel[AIN2].hwChannel = AIN2_HWCHN;
 AIns.ADC_Channel[AIN3].hwChannel = AIN3_HWCHN;
 AIns.ADC_Channel[AIN4].hwChannel = AIN4_HWCHN;
 
 for(U8 i = 0; i < 4; i++)
 {
 AIns.ADC_Channel[i].ADC_Avg = 0;
 AIns.ADC_Channel[i].State.Raw = 0;
 for(U8 j = 0; j < 4; j++) AIns.ADC_Channel[i].ADC_Buf[j] = 0;
 }

 //Initialize PPI
 error = nrf_drv_ppi_init();
 APP_ERROR_CHECK(error);

 //Initialize Timer
 error = nrf_drv_timer_init(&Timer, NULL, Timer_EventHandler);
 APP_ERROR_CHECK(error);

 //Initialize ADC
 ADCConfig.scaling    = NRF_ADC_CONFIG_SCALING_INPUT_ONE_THIRD;
 ADCConfig.resolution = NRF_ADC_CONFIG_RES_10BIT;
 ADCConfig.reference  = NRF_ADC_CONFIG_REF_VBG;
 nrf_adc_configure(&ADCConfig);
 nrf_adc_input_select((nrf_adc_config_input_t)AIns.ADC_Channel[AIns.Current.Number].hwChannel);
 nrf_adc_int_enable(ADC_INTENSET_END_Enabled << ADC_INTENSET_END_Pos);
 sd_nvic_SetPriority(ADC_IRQn, NRF_APP_PRIORITY_LOW);  
 sd_nvic_EnableIRQ(ADC_IRQn);

//Configure Timer to generate an event every XX us
TimerTicks = nrf_drv_timer_us_to_ticks(&Timer, AIN_ADC_SAMPLE_PERIOD_US);
nrf_drv_timer_extended_compare(&Timer, NRF_TIMER_CC_CHANNEL0, TimerTicks, NRF_TIMER_SHORT_COMPARE0_CLEAR_MASK, false);

 //Allocate PPI Channel
error = nrf_drv_ppi_channel_alloc(&PPIChannel);
APP_ERROR_CHECK(error);

//Get Event and Task Pionters
pCompareEvent = nrf_drv_timer_event_address_get(&Timer, NRF_TIMER_EVENT_COMPARE0);
pADCTask = (U32)nrf_adc_task_address_get(NRF_ADC_TASK_START);

//Assign and Enable PPI channel
error = nrf_drv_ppi_channel_assign(PPIChannel, pCompareEvent, pADCTask);
APP_ERROR_CHECK(error);
error = nrf_drv_ppi_channel_enable(PPIChannel);
APP_ERROR_CHECK(error);

//Enable timer
nrf_drv_timer_enable(&Timer);
}

Maybe someone see right away what we are doing wrong?

Thanks everyone!

EDIT: Excel file for capacitor calculation CapacitorCalculationforVoltageDivider20160913v1.xls

EDIT BIS: OP AMP to isolate ADC

Regards,

Francois

[Stefan Birnir Sverrisson]

Hi Francois

Connecting nRF51 ADC input directly to a voltage divider will cause both ADC offset and gain error. A common method of correcting the gain/offset error is to use a capacitor with the voltage divider, see this blog post. There is also available a ADC code code on Nordic's Github page that samples on multiple input pins, i.e. the ble_app_uart_adc_scan_mode example.

[Francois]

Thank you, looking at it right now!

[Francois]

Stefan,

Thanks a lot, exactly what I'm looking for!

I've read this document: devzone.nordicsemi.com/.../428b530562fded0e484c166e76d88129

I'm clear on everything except the U ADC: It's always 0.6V?

Also, I'm adding an Excel file doing the calculation described in the PDF. It could be practical for others... but I can't find the number from my curve with my values... Any idea why?

For example entering 20V, resistor values, pre-scaling value (1/3), U BATT @ 20V and resolution 10 bit. (at 0V, it's plausible. But at 20V, I'm far from it using the pre-scaling values I'm using).

Also, If I want to sample from 0 to 24v and frequency is not an issue, I must use the biggest capacitor right? I guess I don't have any options...

Thanks again

[Stefan Birnir Sverrisson]

Thanks a lot Francois for attaching your excel sheet. Ill attach it in my blog post if you don't mind for convenience for others that link to the blog post.

[Francois]

No problem. Could you please just help me figure out two things:

Where the 0.6v is coming from? Also any idea why I'm not finding the valued given by the Excel sheets in reality? I'm getting the right value for 0v, but not for 20v... Are you 100% of your ADC resistance values?

Thanks!

Francois