In section 1, a few general theorems about the Fourier analysis of fluctuating quantities are mentioned and it is shown that it is impossible for the noise intensity to vary inversely proportional to the frequency in the whole interval 0 < f < infinity.

It is then shown in section 2, that a noise intensity which is inversely proportional to the frequency in a wide range can be obtained if a distribution function of correlation times is introduced.

In sections 3 and 4, it is shown that such a distribution function should be expected from the various mechanisms involved.

In section 3, the various sources of noise in a semi-conductor are mentioned and it is shown that the "shot effect", discussed by Brillouin, Bernamont and Gisolf is not sufficient to account for all the noise. A few other mechanisms, viz. flicker effect and resistance fluctuations due to the motion of foreign atoms or lattice distortions, are then introduced which give the required distribution of correlation times.

In section 4, flicker effect is discussed and it is shown that Macfarlane's theory is equivalent to the one discussed in section 2. It is then suggested that flicker effect in oxide-coated cathodes might not primarily be due to the surface of the coating but much more due to what happens inside it. In an Appendix it is shown that Bernamont's and Gisolf's treatment of shot noise are equivalent.

1/f in electronic device section