nRF52832 large RSSI errors vs. input power

Hello,

I am currently checking the RSSI measurement precision, and we got very large errors (up to 10dB error) on the RSSI vs input power.

To avoid any PCB / matching issues, we use the nRF52832 DK board, connected to an RF source via the coax connector (which disconnects the antenna).

We have checked the RF source precision with a spectrum analyzer and we don't have issues here.

We send a CW signal on 2440 MHz. The nRF52832 under-estimates the RSSI up to -10dB (depending on the input power ; for some powers there is no error).

We of course make the mesurement on the -90dBm -> -20 dBm where the accuracy should be +/-2dB.

Is there something that we are missing? Does the RSSI need a modulated signal instead of a CW signal?

Thanks

Related