<p>Brain-computer interfaces (BCI) are being abundantly used for various applications. Among the methods of BCI, functional near-infrared spectroscopy (fNIRS) is being favored due to its compact size and ability to monitor regional neural activity. As the usage of fNIRS systems has increased, more commercial products and self-built systems are being utilized. The critical components of an fNIRS system are the light source and detector and an abundantly used assumption is that the light sources are composed of a single wavelength and does not have a spectral width. As compact light sources have wide spectral width and compact detectors are incapable of distinguishing between wavelengths, the assumption becomes a source for an error and without careful selection of the light source, large signal distortion may occur during the process of converting detected light intensities into neural activity signals. This research has focused on analyzing the broad-spectrum error caused by the spectral width of the light sources of an fNIRS system, and has conducted simulations to identify wavelength pairs that are relatively robust to the error. The simulation method is simple, and can be easily applied to the light source and detector candidates before building an fNIRS system to assure the system will not cause any large errors.</p>