Date of Award

Spring 1968

Document Type


Degree Name

Doctor of Engineering Science in Electrical Engineering


Electrical Engineering

First Advisor

Raj Pratap Misra

Second Advisor

Frederick A. Russell

Third Advisor

Leonard Salzarulo

Fourth Advisor

Joseph J. Padalino


The feasibility of choosing noise-voltage spectral density as a prediction parameter for the degradation of p-n junctions has been examined both theoretically and through a series of life-tests.

Experimental facts show that the noise-voltage spectral density, Sv(w), observed in a p-n junction under the breakdown condition tends to be "white" (which contradicts the 1/f-noise theory), and Sv(w) is inversely proportional to the breakdown current (which contradicts the shot-noise theory). Furthermore, some p-n junctions display one or more multiple peaks of Sv(w) at different current levels which can not be explained by any of the existing noise theories.

In the derivation of a new theory, four factors are proposed to explain the behavior of Sv(W). They are: (1) the number of primary carriers entering the multiplication zone during the given time interval is a random variable, (2) the entry times of primary carriers into the multiplication zone-is a random variable, (3) the number of impact ionization events caused by a primary carrier is a random variable, and (4) the carrier transit times are statistically distributed.

The noise current due to the above mentioned multifold random processes is approximated by an ensemble of a triangular current pulses, and the Wiener-Khintchine theorem is applied to the autocorrelation function of the noise current to obtain Sv(w)ccV2bI-1r where Vb is the breakdown voltage, and Ir is the breakdown current. This new theory can explain both inverse proportionality between Sv(w) and Ir as well as white noise spectrum of Sv(w).

It has been shown that the multiple-peak phenomenon found in some units is due to the microplasmic breakdown channels in a p-n junction. To demonstrate this fact, a microplasma-free, and a microplasmic breakdown channel were simulated by two diodes in parallel, and the difference in the breakdown voltages of two channels was externally inserted in series with a diode. Experimental evidence was obtained that one can generate artificially multiple-peak phenomenon from the above two-channel model.

Life-test experiments indicate that units whose Sv(w) have multiple peaks are highly correlated with degradation of leakage characteristic over a life-test period.

It is recommended that some unreliable units be eliminated by inspecting the multiple-peak phenomenon which indicates the existence of microplasmic breakdown channels in a p-n junction.