For our setup, we can determine the coefficient \(a,b,c\) by measuring the total noise under different laser power (Figure 2e). Here only the audio frequency range of 0.1~10 kHz is considered. From the fitting, it is clear that the shot noise is dominant in our setup (Figure 2e). It means that REAL has obtained its best SNR because shot noise cannot be reduced by signal amplification. SNR model (in power) of REAL can be obtained by combining Eqution 1 and Equation 3.