The results of nRF52840 ADC conversion are not as expected

[kkuma256]

I'm using PCA10040 for studying nRF SDK, and I tried running the sample 'saadc_pc10040' with the following change in the 'config.h':

#define NRFX_SAADC_CONFIG_RESOLUTION 2

I connected the board to my PC via USB and applied an external power source with the negative terminal connected to GND and +0.6V to P0.02.

At this point, I expected the ADC value (p_event->data.done.p_buffer) to be calculated as follows:

RESULT = [V(P) – V(N)] * GAIN/REFERENCE * 2^(RESOLUTION - m)

 

Where:

• V(P) is the voltage at input P
• V(N) is the voltage at input N
• GAIN is the selected gain setting
• REFERENCE is the selected reference voltage
• m = 0 if CONFIG.MODE=SE, or m = 1 if CONFIG.MODE=Diff.

In this case:

• V(P) = 0.6
• V(N) = 0.0
• GAIN = 1/6
• REFERENCE = 0.6
• m = 0

So, I expected to get the result as:

[0.6 – 0.0] * (1/6)/0.6 * 2^(12 - 0) = 682

However, in reality, I got 62000.

Can you please explain why this is happening?

 

[Bendik Heiskel]

Hi,

Apologies for the delay.

I'm not able to replicate the issue you are seeing, using the saadc sample in nRF5 SDK 17.1.0  with the modified saadc resolution.

Have you made any other modifications to the sample?

 

Best regards,

Bendik

[kkuma256]

I'm sorry. When I tried it again here, the numbers turned out as expected, so I'm not sure what the cause was. If it happens again, please let me know at that time.

[JONATHAN LL]

Do you only experience this on one device? 

And are you sure that you have not done any other changes to the code?

The behavior that you see can come from the Errata related to calibration. 
https://infocenter.nordicsemi.com/topic/errata_nRF52840_Rev3/ERR/nRF52840/Rev3/latest/anomaly_840_252.html?cp=5_0_1_0_1_52 
https://infocenter.nordicsemi.com/topic/errata_nRF52840_Rev3/ERR/nRF52840/Rev3/latest/anomaly_840_237.html?cp=5_0_1_0_1_44 

Try to disable the calibration and see if that changes the behavior that you see. If there is no changes then its likely not the Errata that is the issue.

Regards,
Jonathan

[kkuma256]

How can I specifically disable calibration? I've looked into it briefly, but I'm not quite sure of the process. If you could provide a link to relevant documentation or any guidance, that would be appreciated.

[JONATHAN LL]

Hi, do not call nrf_drv_saadc_calibrate_offset()/nrfx_saadc_calibrate_offset() and don't trigger the TASKS_CALIBRATEOFFSET. 

[kkuma256]

Symptoms occur in 2 out of 4 units.

Initially, during development, I recall that the issue was not present, but there is no recorded evidence, and my memory is the only basis, so it's not definitive.

When running the exact same program on different boards, the problem appears inconsistently for each board.

In the test program (an adaptation of the saadc_pca10056 sample to work with KAGA's board), it seems that calibration is not performed at all.

Just to be sure, I placed a breakpoint in the following line in nrfx_sadc.c:

    nrf_saadc_int_disable(NRF_SAADC_INT_END | NRF_SAADC_INT_CALIBRATEDONE);
=>  if (m_cb.adc_state == NRF_SAADC_STATE_CALIBRATION)
    {
        :

But when I broke into it, the condition (m_cb.adc_state == NRF_SAADC_STATE_CALIBRATION) did not hold.

[Jørgen Holmefjord]

Can you post your current application code and configuration?

The SAADC sample in nRF5 SDK does not use offset calibration by default.

The results from the SAADC is given as 2's complement 16-bit values (corresponding to int16_t type). int16_t have a range between [-32768, +32767]. If you see values in the 62000 range, your data type is incorrect.

You should not enable any other peripherals (e.g., GPIOTE) on the same pins used for SAADC input.