Number 90 (Story #1), August 4, 1992 by Phillip F. Schewe and Ben Stein 1/F NOISE HAS 1/F NOISE. Certain signals, such as the light coming from quasars or the current flowing through a resistor, exhibit small fluctuations which, although seemingly random, actually possess an underlying structure. When studying this problem, scientists often transform their data from the time domain---the raw signal, consisting of fluctuations around some mean value, as a function of time---into the frequency domain, in which the fluctuation "power" (essentially the fluctuation likelihood) is plotted as a function of frequency. Such a plot typically shows that the fluctuation power is proportional to the inverse of the frequency; hence the name "1/f noise." The data is Fourier-transformed in this way because it better reveals the fact that all time scales are involved in the physical process at work. In other words, fluctuations occurring now are related to fluctuations at various times in the past, and are anything but random. Scientists at the University of Minnesota (contact James Kakalios, 612-624-9856) have carried this process one step further and have shown that a 1/f noise spectrum can itself be characterized by 1/f noise. Studying the fluctuations of current flowing through hydrogenated amorphous silicon, the Minnesota researchers formed a power spectrum, which they monitored as a function of time. This time dependence in turn furnished a "second spectrum," which proved itself to be 1/f in nature. (C.E. Parman, N.E. Israeloff, and J. Kakalios, 17 August 1992 issue of Physical Review Letters.)