RE: Noise in ADC readings of nRF52840

haakonsh — Fri, 08 Jan 2021 12:29:14 GMT

The resistance from a GPIO to GND and VDD should be several Mohms when the device is off, can you double-check your findings?

If they are in the kohm range then the GPIOs are likely fried, possibly do to applied over/under-voltage.

RE: Noise in ADC readings of nRF52840

michaelatom — Thu, 07 Jan 2021 15:14:46 GMT

Thanks for the answer.

Yes, it is important for us to overcome the differences at the 3mV range.
I retested the SAADC example now on P002 and on VDD, got readings between 844 and 856 (diff 12). Is it what is expected.
Also, checked the resistance of AIN2 (p0.04), to GND is 55kohm and to VDD 40kohm.
For AIN0 (p0.02) got 90kohm to GND and 40kohm to VDD

RE: Noise in ADC readings of nRF52840

haakonsh — Thu, 07 Jan 2021 12:34:13 GMT

[quote user=""]The lowest difference we are able to get is 3mV on one of the configurations we are using for the analog read which have oversampling.[/quote]

That's the expected result, you won't get much more precision than that with this SAADC.

With ~9 ENOB you will get a resolution of 600mV/2^9 = 1.172mV, but the accuracy of the internal reference is not exactly 600mV and there's also some gain error, DNL/INL, and input offset that affects the end result.

Is it crucial for your application to get better accuracy/precision than 3mV?

[quote user=""]The same FW configuration that gave us 3mV differences in our system is giving 4.8mV differences in the DK after sampling for a few minutes.[/quote]

The DK's analog pin layout is not ideal, the traces pass close to digital/power traces and will likely pick up some noise.

[quote user=""]

Next we tried the SAADC example from SDK 17.0.2 connected to 2.523 V. We saw a difference of 50 on the readings (max 741, min 691). When shorted to GND on the DK, the readings range is from -1 to 2 and when tied to VDD pin of the DK the ADC reading range is from 772 to 831.

The only change made to the example code is that we are using AIN2 instead of AIN0.

[/quote]

There's definitely something wrong here, 50 codes of variance is way too much.

My first guess is a blown ESD fuse on the AIN2 (p0.04) GPIO. The most common failure mode is a partial or full short to either GND or VDD. Can you measure the ohmic resistance of the AIN0 and AIN2 with respect to GND and VDD and compare them?