Differential ADC Anomaly on nRF9160 DK

[bencra]

Hi,

I'm trying to read differential ADC input with inputs configured as:

NRF_SAADC_INPUT_AIN1 (P0.014) -> POSITIVE

NRF_SAADC_INPUT_AIN2 (P0.015) -> NEGATIVE (differential mode)

I'm also using 4x oversampling, gain of 1, +-0.6V reference, 12-bit ADC and 5us acquisition time.

Unfortunately, I seem to be getting very poor results.

By grounding both inputs, I'm receiving an output value of ~350, which equates to around 103mV, which is too poor for my use case.

Any advise would be appreciated.

Regards,

Ben

[bencra]

For grounded inputs, for any given sample set (extra_samplings set to 399 - total samples = 400), there seems to be large alternating errors, such as:

Sample 368: 415
Sample 369: 364
Sample 370: 416
Sample 371: 360
Sample 372: 414
Sample 373: 362
Sample 374: 415
Sample 375: 365
Sample 376: 417
Sample 377: 357
Sample 378: 416
Sample 379: 363
Sample 380: 418
Sample 381: 364

I'm also struggling to find documentation regarding the correct calibration procedure.

At current, I'm performing NRF_SAADC_NS->TASKS_CALIBRATEOFFSET = 1; before all sampling.

Update:

Considering the invalid result (errata) which comes from settting VDD_GPIO > 2.1V, I've changed the VDD_GPIO to 1.8V. The error in the samples now appears to have decreased, and the samples look something like this:

Sample 365: 209
Sample 366: 304
Sample 367: 211
Sample 368: 303
Sample 369: 213
Sample 370: 302
Sample 371: 213
Sample 372: 304
Sample 373: 213
Sample 374: 305
Sample 375: 210
Sample 376: 303
Sample 377: 212
Sample 378: 302
Sample 379: 213
Sample 380: 300

This is still not useful data for my use case.