Simultaneous EEG-fMRI studies require an understanding of the noise characteristics of the acquisition environment so that appropriate pre-processing steps may be taken to remove known artifacts from the data stream. Using a phantom approach, we have developed a general methodology for characterizing non-physiologic noise in EEG signal and demonstrate the use of this methodology for a specific MR scanner and EEG data acquisition system configuration. Our results show the δ frequency band is significantly impacted by baseline drift or baseline correction algorithms while the β and γ bands are impacted by residual gradient artifact and gradient corrections.