Race condition in starting the ADC following calibration

RE: Race condition in starting the ADC following calibration

Malcolm — Fri, 24 Apr 2020 07:41:07 GMT

Many thanks

RE: Race condition in starting the ADC following calibration

Simonr — Fri, 24 Apr 2020 05:27:59 GMT

Hi Malcolm

It won't suffice to just set PSELN, you'll also have to change mode in the channel config. Please see this GitHub link for reference. I don't see that you do this in the snippet you've posted. Can you show us how you configure the ADC in your project?

Best regards,

Simon

RE: Race condition in starting the ADC following calibration

Malcolm — Wed, 22 Apr 2020 14:49:41 GMT

In my project I am trying to determine temperature by measuring resistance of an NTC. I have set up the following circuit

            /*
                           VDD
                           |
                       ---
                       |       |   RVdd
                       |       |
                       ---
                           |____Vdd (AIN4)
                           |
                       ---
                       |       | R2
                       |       |
                       ---
                           |_____Vo (AIN2)
                           |
                       ---
                       |       | R1
                       |       |
                       ---
                           |_____Vg (AIN1)
                           |
                       ---
                       |       | Rgnd
                       |       |
                       ---
                   ____|_____

           R1/R2 = (Vo - Vg) / (Vdd - Vo)

           R1 = R2 * (Vo - Vg) / (Vdd - Vo)

           We make R1 unknown and R2 known, so find R1 as Rt
           */

This is my current code

            /* configure each of the channels to be used */
           channel_config.gain = NRF_SAADC_GAIN1_4;
           channel_config.acq_time = NRF_SAADC_ACQTIME_3US;

           /* input for Vdd */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN4;
           err_code = nrf_drv_saadc_channel_init(0, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* input for Vo */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN2;
           err_code = nrf_drv_saadc_channel_init(1, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* input for Vo */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN1;
           err_code = nrf_drv_saadc_channel_init(2, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* initalise buffers */
           err_code = nrfx_saadc_buffer_convert(temp_sample_buffer, BUFFER_SAMPLES);
           APP_ERROR_CHECK(err_code);

           /* start sampling */
           nrfx_saadc_sample();
           APP_ERROR_CHECK(err_code);

        case NRF_DRV_SAADC_EVT_DONE :
           thermometer_sample(p_event->data.done.p_buffer);
           break;
   }

In this example R1 = R2

When I break on sample done I get

temp_sample_buffer[0] = 3523

temp_sample_buffer[1] = 2411

temp_sample_buffer[2] = 1303

I use Rgnd as the analogue ground in this chip is rather poor as is takes all the digital current.

This approach gives me extremely good results as I average 1000 samples of each, effectively giving 20 bit accuracy (1 in 1000), and averaging reduces the noise.

I change the code to

            /* configure each of the channels to be used */
           channel_config.gain = NRF_SAADC_GAIN1_4;
           channel_config.acq_time = NRF_SAADC_ACQTIME_3US;

           /* input for Vdd */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN4;
           channel_config.pin_n = NRF_SAADC_INPUT_AIN2;
           err_code = nrf_drv_saadc_channel_init(0, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* input for Vo */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN2;
           channel_config.pin_n = NRF_SAADC_INPUT_AIN1;
           err_code = nrf_drv_saadc_channel_init(1, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* input for Vo */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN1;
           err_code = nrf_drv_saadc_channel_init(2, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* initalise buffers */
           err_code = nrfx_saadc_buffer_convert(temp_sample_buffer, BUFFER_SAMPLES);
           APP_ERROR_CHECK(err_code);

           /* start sampling */
           nrfx_saadc_sample();
           APP_ERROR_CHECK(err_code);

           break;
       case NRFX_SAADC_EVT_LIMIT :
       /* this is not used */
           break;
       case NRF_DRV_SAADC_EVT_DONE :
           thermometer_sample(p_event->data.done.p_buffer);
           break;

I get

temp_sample_buffer[0] = 3491

temp_sample_buffer[1] = 2391

temp_sample_buffer[2] = 1289

buffer[0] should be equal to buffer[1], but isn't and voltages are similar to the first case with single ended.

Regards

Malcolm

RE: Race condition in starting the ADC following calibration

Simonr — Wed, 22 Apr 2020 12:58:39 GMT

Hi Malcolm

Can you upload your input signals and the output you're getting, as we'd like to take a look, to see if we can spot what's actually going on in your project? As far as I know, there shouldn't be any problems with using differential measurements like this, so I'm at a loss of what this could be.

Best regards,

Simon

RE: Race condition in starting the ADC following calibration

Malcolm — Tue, 21 Apr 2020 06:56:47 GMT

I have multiple channels working in SCAN mode, that is not my problem, rather I wanted to make a different differential measurement on each channel in the scan (see code), and this appears not to work. Instead it appears each channel works as single ended measurement and PSELN is not used in SCAN mode.

I have workaround by using single ended for AIN2, AIN4 and sampling AIN1 and then subtracting.

RE: Race condition in starting the ADC following calibration

Simonr — Tue, 21 Apr 2020 06:40:22 GMT

Hi Malcolm

Please take a look at this thread, where my colleague explains how to set up sampling multiple channels in SCAN mode. Keep in mind that you can't have oversampling enabled on any channel while in scan mode, as these modes are incompatible with one another.

Best regards,

Simon

RE: Race condition in starting the ADC following calibration

Malcolm — Mon, 20 Apr 2020 13:28:56 GMT

I need further help. I am using SCAN mode to sample 3 separate channels. I tried to sample each of these in differential mode and so set PSELP and PSELN (see code) but when looking at the data, it appears to be single ended, and SCAN mode does not change the PSELN pin.

RE: Race condition in starting the ADC following calibration

Simonr — Mon, 20 Apr 2020 12:50:06 GMT

Hi Malcolm

Is this issue related to your other ticket at all? Do you still need support in this thread or are you happy with the results from that thread?

Please note that PSELN should be used when you want to measure negative values or you use differential measurements.

Best regards,

Simon

RE: Race condition in starting the ADC following calibration

Malcolm — Mon, 20 Apr 2020 08:57:58 GMT

Further on this topic. It is not explicit in the documentation if SCAN mode also changes PSELN when the channel is changed. It appears not.

Malcolm

RE: Race condition in starting the ADC following calibration

Malcolm — Mon, 20 Apr 2020 08:32:31 GMT

Thank you, this was helpful. I therefore modified my code to the following. However note that the **** line no longer works as intended as the interrupt routine now checks for a STOPPED event and set the SAADC to IDLE state and the second buffer is not passed in the following interrupt routine..

Regards

Malcolm

void saadc_callback(nrf_drv_saadc_evt_t const * p_event)
{
   ret_code_t err_code;

   switch (p_event->type) {
       case NRFX_SAADC_EVT_CALIBRATEDONE :
           /* workaround for errata 86 */
           nrf_saadc_task_trigger(NRF_SAADC_TASK_STOP);
           // Wait for ADC being stopped.
           bool result;
           NRFX_WAIT_FOR(nrf_saadc_event_check(NRF_SAADC_EVENT_STOPPED), 1000, 0, result);
           NRFX_ASSERT(result);

           /* configure each of the channels to be used */
           channel_config.gain = NRF_SAADC_GAIN1_4;
           channel_config.acq_time = NRF_SAADC_ACQTIME_3US;

           /* differential input for Vdd */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN4;
           channel_config.pin_n = NRF_SAADC_INPUT_AIN2;
           err_code = nrf_drv_saadc_channel_init(0, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* differential input for Vo */
           channel_config.pin_p = NRF_SAADC_INPUT_AIN2;
           channel_config.pin_n = NRF_SAADC_INPUT_AIN1;
           err_code = nrf_drv_saadc_channel_init(1, &channel_config);
           APP_ERROR_CHECK(err_code);

           /* initalise buffers */
           err_code = nrfx_saadc_buffer_convert(temp_sample_buffer1, BUFFER_SAMPLES);
           APP_ERROR_CHECK(err_code);

****          err_code = nrfx_saadc_buffer_convert(temp_sample_buffer2, BUFFER_SAMPLES);
           APP_ERROR_CHECK(err_code);

           /* start sampling */
           nrfx_saadc_sample();
           APP_ERROR_CHECK(err_code);
           break;
       case NRFX_SAADC_EVT_LIMIT :
       /* this is not used */
           break;
       case NRF_DRV_SAADC_EVT_DONE :
           thermometer_sample(p_event->data.done.p_buffer);
           break;
   }
}

RE: Race condition in starting the ADC following calibration

Andrew Lentvorski — Sun, 19 Apr 2020 22:10:38 GMT

You might want to check the Errata as the 52832 has several for the SAADC. Errata 86 may apply.