Jin,
I have found what I believe is your issue: you are not actually feeding 10mVpp to the AFE input. Is there a way that you can confirm this? The AFE 50ohm activate termination does not work with LNA Gain = 12dB ( please see DS page 32).
Therefore, unless you have an external 50ohm resistor on the board, the AFE input is high impedance with LNA=12dB. Assuming that your signal source has a 50ohm source resistance, the signal into a high impedance will be twice the nominal value that you set. In other words, when you set 10mVpp on the source, it will look like 20mVpp at the AFE input. Please see if you can verify this.
With LNA = +12dB, VCA = -36dB, PGA = +24dB, there is zero gain. Also, The LSB resolution is 0.122mV. Therefore, the input to the ADC should also be 20mVpp, when should theoretically result in (20/0.122) =164codes. This still doesn’t match your 200 codes, however.
Upon repeating your experiment in the lab here, we were able to match your results, ≈196.72codes. We also notice, however, that the gain is not exactly 0dB, but more like 1.75dB. We are investigating this gain error.
Hope it helps
Fabio
Applications Engineer
TI Healthtech/Imaging