Usrp +/-32767 integer values to actual voltage conversion?

Hey all,

I record signals to files from USRP, then read the file in matlab and
I want to compute the actual received power from FFT computation.I think
the maximum received power expected is equal to the ADC output,16dbm. Is
that correct?

The values read in matlab are float values converted from the 16 bit
usrp integer values,+/-32767.That means,they are not actual voltage
values.So, how can i correct/compensate the computed powers in FFT to
make them equivalent to the power output of the ADC?

I try to dig into scope_sink and fft_sink to see how the computation is
done. In scope_sink, I see that the actual integer values,not voltages
are plotted. In fft_sink, power is computed, but compensation is done
for FFT size and window size.So, where is these conversion from integer
values to actual voltage values done?