Functional near-infrared spectroscopy (fNIRS) is widely used in brain science. The sensitivity and signal-to-noise ratio (SNR) are core parameters during the measurement. However, previous studies have not systematically analyzed the parameters in wholebrain measurements which limits its further improvement. In this paper, a noise model was established, including electronic noise, shot noise and speckle noise. Considering different brain regions and different source-detector(S-D) separation, a Monte Carlo simulation tool was utilized to simulate the photons migration inside a realistic human-brain template (Colin27). Based on the simulation, we analyzed the light fluence-rate distribution, detection-sensitivity, noise composition and SNR. The results show that sensitivity is positively correlated with S-D separations and the parietal region has the lowest sensitivity. In terms of noise, the equivalent noise power in different brain regions were consistent, and the noise proportions varied greatly under different S-D conditions and the SNR decreased by about 45 dB in the S-D range 10 mm∼45 mm. Based on the results, a fNIRS system optimization strategy is proposed and in our simulation conditions, we suggest an optimal S-D in frontal, temporal, parietal and occipital lobes are 30 mm, 25 mm, 30 mm, 30 mm respectively. The systematical analysis method will contribute to guiding the optimal design of the fNIRS system.