| Abstract: | A new method of random data analysis has been developed with special implications for membrane noise. The integral spectrometer uses overlapping broad-band filters of simple design, whose bandwidth increases linearly with center frequency. A random two-state process, which has a Lorentzian-shaped spectral density, results in an integral spectrum whose maximum value occurs at the mean frequency of the events, and which is symmetric about that frequency on a semilog plot. 1/f noise is flat and does not distort the symmetry on the frequency axis. The integral spectrum exchanges resolution on the frequency axis for accuracy in the amplitude. The expected statistical error in amplitude has been calculated for three types of membrane noise assuming finite data. The integral spectrum compares favorably with conventional spectral densities and may be a reasonable alternative for random data analysis. |
