I've read the previous discussions on this matter and find the common answer seems to be:

0.707 mVpp which corresponds to +7 dBm based on a 50ohm system.

Also, commonly cited is a note about Sigma-Delta Converters exhibiting a soft overload characteristic, which is only important near maximum inputs.

We are using the AD9375 EVM connected to a custom FPGA EVM. We have a RSSI calculation block built into our RX Signal Chain. This block does a sum of I^2 + Q^2 of the received samples and we then use that to calculate RSSI of our input signal.

For testing and calibration we use a CW tone, 1MHz offset from the tuned frequency. Applying the +7 dBm Full Scale Power, the calculation seems to be about 16.4 dB off expected.

To troubleshoot, I performed the following test. Using a calibrated cable connected between the Sig Gen and the 9375 EVM RX1 port at 680 MHz, Rx tuned to 679 MHz, RX1 Gain index set to 255 (0 dB of attenuation), Logic Analyzer connected to capture the raw 14 bit JESD I/Q samples from the 9375, I see a very nice sinewave with peak amplitude of 2414. Since the samples are 14 bits, this is 2414/8191 in amplitude. 20 * log10(2414/8191) is -10.6 dB. This says my signal is -10.6 dB below full scale. My input is -20.0 dBm, so this says Full Scale Power is actually -9.4 dBm.

I understand the recommended max input to the 9375 is -14 dBm, but I was able to get up -11 dBm with no noticeable distortion in the observed sinewave samples, got a peak sample value of 6840, which using the same math as above yields -1.56 dB or a SP of -9.4 again.

Can you please advise where I went wrong in my assessment of Full Scale Power?