Nanopores and submicrometer pores have recently been explored for applications ranging from detection of single molecules, assemblies of nanoparticles, nucleic acids, occurrence of chemical reactions, and unfolding of proteins. Most of these applications rely on monitoring electrical current through these pores, hence the noise and signal bandwidth of these current recordings are critical for achieving accurate and sensitive measurements. In this report, we present a detailed theoretical and experimental study on the noise and signal bandwidth of current recordings from glass and polyethylene terephthalate (PET) membranes that contain a single submicrometer pore or nanopore. We examined the theoretical signal bandwidth of two different pore geometries, and we measured the signal bandwidth of the electronics used to record the ionic current. We also investigated the theoretical noise generated by the substrate material, the pore, and the electronics used to record the current. Employing a combination of theory and experimental results, we were able to predict the noise in current traces recorded from glass and PET pores with no applied voltage with an error of less than 12% in a range of signal bandwidths from 1 to 40 kHz. In approximately half of all experiments, application of a voltage did not significantly increase the noise. In the other half of experiments, however, application of a voltage resulted in an additional source of noise. For these pores, predictions of the noise were usually still accurate within 35% error at signal bandwidths of at least 10 kHz. The power spectra of this extra noise suggested a 1/f(alpha) origin with best fits to the power spectrum for alpha = 0.4-0.8. This work provides the theoretical background and experimental data for understanding the bandwidth requirements and the main sources of noise in current recordings; it will be useful for minimizing noise and achieving accurate recordings.