Noise is the enemy of all receivers, obliterating the desired signals and rendering attempts to eliminate it problematic, but for more than 70 years, it has also been used beneficially, helping to test receivers and many other types of systems. The process relies on deceptively simple devices—noise sources—that provide the basis for many types of measurements by generating and injecting Additive White Gaussian Noise (AWGN) into the device or system under test. The noise source has come a long way from its origin in the 1940s, so it is helpful to know about modern noise sources, how they work, and the tests they can enable.
At the time of World War II, RF and microwave receiver technology was advancing rapidly, and they were able to detect weaker and weaker signals buried in the noise. It was thus necessary to create a reference standard whose noise characteristics were known and could lead to more effective noise factor, signal-to-ratio (SNR), and other tests.
Three approaches were investigated: “hot” resistors, vacuum noise diodes, and gas discharge tubes. Hot resistors were a possibility but as they heated to 2900° F to measure a noise factor of 100 they were not practical. Vacuum tube noise diodes (Figure 1) also seemed promising but could only reach a frequency of about 300 MHz—and required an ignition voltage of 6000V. Nevertheless, they were used for lack of something better, which turned out to be gas discharge tubes (Figure 2).
Back in the 1940s, it was realized that these tubes, which contain noble gases such as argon, helium, or neon, produce a steady-state gas plasma containing light radiation as well as wideband microwave noise up to about 3 GHz when excited with the proper DC voltage. They were a significant improvement over previous methods, but building on a phenomenon revealed in the 1930s, they were soon to be faced with a better, smaller, and generally more effective solution.
In 1934, theoretical physicist Clarence Melvin Zener described the breakdown of electrical insulators, one of his many accomplishments, which resulted in what is called the Zener Effect. His work was later implemented by Bell Labs scientists to form an electronic device, the Zener diode. A Zener diode has many of the basic properties of an ordinary diode, conducting in the forward direction with the same turn-on voltage, for example.
However, in the reverse direction, the operation of a Zener diode (Figure 3) is quite different because at low voltages, it does not conduct as would be expected. When a specific reverse (Zener) voltage is reached, the diode “breaks down,” and current flows because of electron quantum tunneling in the small space between the p and n regions (i.e., the Zener effect). In this condition, the diode generates shot noise, which is temperature and frequency independent, in contrast to Johnson-Nyquist noise that is proportional to temperature. Both are considered white noise.
The Zener diode’s characteristics are similar to those of avalanche diodes, as both generate noise, except that a Zener diode has a breakdown voltage of below 5 VDC, while avalanche diodes are used for breakdown voltages above that voltage. The Zener diode’s overall characteristics are very appealing because of their ability to generate noise over a very wide range of frequencies, making them well suited for use as noise sources.
Interestingly enough, not all Zener diodes have the “right” characteristics for this purpose, so in any given batch, individual devices will exhibit somewhat different characteristics. Only those suited for broadband noise output and flat spectral response are selected by noise source manufacturers. So, although a Zener is a special type of diode, those suitable for use as noise sources are very special indeed.
Notably, in the higher echelons of noise generation that require the highest possible precision, such as national standards laboratories like the National Institute of Standards and Technology (NIST), esoteric types of noise sources are used. For example, NIST recently advanced the accuracy of its reference standard used to define kelvin, the international unit of temperature, using its Quantum Voltage Noise Source (QVNS) based on arrays of Josephson junctions (Figure 4).
These superconducting circuits operate with quantum accuracy and generate a precisely controllable voltage fluctuation equal to thermal noise except that the signal is not random. As a result, the QVNS can produce an output of perfectly quantized integer units to produce a calculable noise source reference. Other commercially available types are liquid nitrogen-cooled subsystems capable of producing extremely high accuracy at frequencies up to 400 GHz.
The Color of Noise
All noise is not the same and is broadly referred to by color, such as pink noise and white noise. While people use so-called “white noise” generators to drown out the noise around them, they actually generate pink noise because its frequencies produce a more soothing sound. If they used white noise, it would sound like static from an FM radio tuned to a space between channels because white noise covers a broad swath of frequencies. It draws its name from white light that has uniform emissions at all frequencies in the visible spectrum. As this applies to electromagnetic energy below about 300 GHz, it is equally dense at a wide range of frequencies, resulting in a constant power spectral density.
RF, microwave, or millimeterWave noise sources produce AWGN, which is additive because when injected into a device under test, it adds to wherever noise is present in the system, with the result being equal to the transmitted signal plus noise. It is white because, ideally, it has the same power spectral density with frequency, and Gaussian because its probability distribution is Gaussian with a mean of zero. A Gaussian noise distribution (Figure 5) can be represented as a bell-shaped curve symmetrical about the mean with no left or right bias. Adding AWGN into a system makes it desirable for a wide range of applications.
AWGN noise sources are used for a broad array of measurements, from calibrating communications, electronic warfare, and radar systems to noise figure, gain-bandwidth, noise power, carrier-to-noise-radio (CNR), Eb/No (normalized Carrier to Noise ratio for digital systems), bit error rate (BER) testing, and as a simple broadband signal generator for fault isolation testing (FIT), built-in test (BIT) within a system. They are also used for calibrating instruments, testing DOCSIS 3.0 for cable systems, evaluating satellite communication links, and even increasing the dynamic range of analog-to-digital converters (ADCs) by dithering and reducing correlated noise.
From Simple to Sophisticated
In addition to its inherent benefits, what makes a Zener diode-based noise source so versatile is that it can provide these benefits while requiring only a minimal amount of space. For example, a noise diode alone can be soldered onto a PC board or integrated within a surface-mount chip to deliver wide bandwidth AWGN from about 1 GHz to 110 GHz. It can be mounted within an aluminum enclosure with connectors appropriate for its frequency range and ruggedized to meet many military and aerospace standards. For use as laboratory standards, these connectorized (or waveguide) noise sources are calibrated in 1 GHz increments and accompanied by a calibration report (Figure 6).
Noise sources can also be combined with other devices to produce a benchtop instrument. For example, their output can be increased with an amplifier or reduced with an attenuator, and their bandwidth narrowed with a bandpass filter to cover a specific frequency range. For example, RF and microwave products supplier Pasternack offers the PE85N1018, which is a calibrated benchtop noise source covering 10 MHz to 3 GHz (Figure 7). This unit includes a variable attenuator with 1dB steps, an RF power amplifier, and a 120 VAC power supply. In all cases, the goal for noise source manufacturers is to ensure that none of these components modify the original AWGN characteristics.
In references to Zener diodes, their use as noise sources is perhaps the least mentioned and sometimes not mentioned at all, which is surprising considering their importance. Without them, evaluating the performance of receivers, subsystems, and systems, both analog and digital, would be much more difficult. There is simply no easier, smaller, less complex, or more inexpensive means for performing these functions. Consequently, it seems unlikely that other solutions will be proposed in the foreseeable future.