James J. Brophy

Low‐Frequency Variance Noise

Variance fluctuations in 1/f noise examined with apparatus capable of responding to zero frequency signals are found to have the same magnitude as those in band‐limited 1/f noise signals. The distribution of variances in identical sample lengths is skewed to small values in the case of current noise in a carbon resistor and is symmetrical in the case of current noise in a germanium p‐n junction. The probability amplitude is normally distributed in all samples investigated, from 10 to 2400 sec long. Both the average variance and the variance of the variance are found to increase as the logarithm of the sample duration.

Variance Fluctuations in Flicker Noise and Current Noise

Variance fluctuations in band‐limited vacuum‐tube flicker noise, current noise in a germanium p‐n junction, and current noise in a carbon resistor are measured by determining probability amplitude distributions of the noise signals with a multichannel pulse‐height analyzer. The variances of 30 noise specimens, each 100 sec in duration, for the three noise sources are obtained from the width of the probability amplitude distributions, all of which obey a normal‐distribution law. Fluctuations in variance are largest in the case of resistor current noise and least for current noise in a p‐n junction. Even in the latter case, however, the variations are well in excess of artifacts resulting from sampling errors, as demonstrated by examining Nyquist noise signals. It is concluded that all three 1/f noise sources exhibit variance fluctuations to varying degrees.

Influence of Lower Cutoff Frequency on the Measured Variance of 1/f Noise

Previous experimental comparisons of 1/f and Nyquist noise in carbon resistors reveal that both are normally distributed in any time interval but that 1/f noise appears to have a nonstationary variance. The present paper examines the possibility that this fluctuation is caused by insufficient sampling of the 1/f noise signal. It is hypothesized that the 1/f noise is wide‐sense stationary and the number of independent samples per measurement is found and shown to be far less than that for Nyquist noise. An experiment is then described in which the number of samples is sufficiently large to eliminate this source of variance fluctuation. The results exhibit the same spread in measured 1/f noise variance reported earlier, suggesting that an intrinsic nonstationarity does exist in this noise process when produced in carbon resistors.

Current Noise in Thermistor Bolometer Flakes

The characteristics of the additional electrical noise generated by iron oxide thermistor bolometer flakes due to direct current passing through the units is investigated. It is found that the noise is similar to the current noise observed in other semiconductors. The current noise power density is observed to depend inversely upon frequency, to be approximately proportional to the direct current, and to be independent of temperature. It is indicated that the major source of the noise is at the electrodes which make electrical contact to the semiconductor. Electrodes produced by evaporated gold applied through a potential difference are found to yield relatively low noise flakes.

Zero‐Crossing Statistics of 1/f Noise

Probability density distributions of the interval spacings between zero‐crossings of band‐limited Nyquist and 1/f noise signals are examined using an interval‐to‐pulseheight converter and a multichannel pulse‐height analyzer. The Nyquist noise distributions follow a simple exponential law, as expected. In the case of 1/f noise, the distributions are approximately proportional to the inverse square of the spacing interval. The probability density is statistically stationary, that is, independent of sample length and independent of the sample selected. For signals in a 1 Hz to 5 kHz band the most probable spacing interval is 1.2×10−4 sec, which occurs with a maximum probability density of 2.2×103 sec−1. The most probable value is inversely proportional to the highest signal frequency present and the peak probability density is proportional to this frequency.

Magnetic Fluctuations in Molybdenum Permalloy

Considerations of generalized noise and magnetic viscosity suggest the existence of magnetic fluctuations in ferromagnets. Magnetic fluctuations in molybdenum Permalloy tape ring cores have been detected by observing noise voltages at the terminals of a coil wound on the sample. For thin tape cores and at low frequencies, the observed magnetic noise spectrum is in excess of Nyquist noise and is characteristic of a slow relaxation phenomenon with an activation energy of 0.04 ev. For thick tape cores and at high frequencies, core losses yield noise voltages in agreement with Nyquists law and provide a means of studying ferromagnetic loss with zero magnetic excitation.