by Robert B. Tomer (Excerpt from ‘Getting the Most Out of Vacuum Tubes’ published in 1960 but is no longer in print)
Noise is defined as an output signal originating within the tube and is not specifically hum or microphonics. There are several possible sources of noise signals. Whether they constitute a defect depends upon how the tube is used. Hence, noise is considered to be a subjective failure since the actual application of the tube is the significant factor in determining whether or not it actually becomes classified as a failure.
The most common form of noise within a vacuum tube is that caused by intermittent shorts or opens. The nature of these intermittents is such that it makes a big difference how much voltage is applied between elements and at which source impedance, as to whether they exist or not. Also, it is important to define what kind of detection device is going to be used before you can decide whether you have an intermittent or not.
The classic short tester is a neon lamp in series with a voltage source and a resistance to limit the current flow. It is assumed that if an intermittent short occurs, a circuit will be formed and a current will flow. At the same time, the lamp will flicker and the user will know that he has an intermittent short. But will he? The neon bulb is often used in a self -rectifying circuit with an AC voltage applied to it. The voltage required to cause the lamp to ionize is not present during the entire cycle. During a very significant portion of each cycle, the lamp cannot light even if a dead short occurs because the voltage across the circuit is insufficient. Suppose that the voltage is just sufficient, but falling, when the short occurs. The neon lamp requires a finite time to ionize and fire; hence, if the short occurs at the time the voltage is approaching the critical value there will be insufficient time to record it and it will pass unnoticed.
There is also the matter of nonrepeatable shorts and intermittents. Particles sometimes shake loose from within the tube structure and drop down between the elements. These occasions produce momentary shorts which in all probability will not occur again. Nevertheless, in some pulse-triggered applications, as for instance, in counting circuits, these random noise pulses can cause false operation.
Repeated testing of tubes in an endeavor to eliminate this failure has been unsuccessful. An exact explanation for this involves problems in statistical probability that are not within the scope of this discussion. However, it can be mentioned that the use of a statistical approach to this problem can be quite successful. For instance, if 100 tubes are measured and the number showing momentary shorts is noted and compared with another lot of 100 tubes similarly tested, the lot having the significantly lower number of random intermittent shorts will always show this characteristic, no matter how many times it is remeasured. This means that statistical sampling can be used to indicate the probability of the occurrence of intermittent shorts in future use.
The second and most common form of noise in vacuum tubes is sometimes referred to as “frying noise.” It is most often the result of leakage paths across the micas. The measurement of this characteristic is complicated because it has impedance and frequency characteristics that make a universal test very impractical. For example, there are tubes which will produce considerable noise when tested in a high-gain RF amplifier, but none when tested in a high-gain audio amplifier. The converse is also true. Tubes which produce extraneous noises in an audio amplifier may produce no noise at all in a high-gain RF amplifier. The reasons for this observed phenomenon are not well known. Neither is it thoroughly understood why the relation between the amount of noise detected and the sensitivity of the amplifier used to detect the noise is not a linear function.
Noise in vacuum tubes is a problem which depends almost entirely upon the users’ requirements. It does not lend itself to accurate definition and there are very few methods, if any, by which the user or the designer can protect himself against it. About the only method is to make use of the laws of probability in some way or another. In fact, so many vacuum-tube characteristics resolve themselves into a matter of statistical probability, it is felt that some understanding of this subject is essential before proceeding into some of the other areas of vacuum tube knowledge. For this reason, the subject of characteristic variables and their normal ranges and limits is discussed in Chapter 4.