Beginning with formal and informal education and extending into almost all projects, engineers learn that noise is an ongoing potential problem, even for digital designs. There are many noise types and characterizations. That includes Gaussian, white, Rician, pink, 1/f, impulse, shot, Johnson-Nyquist, cyclostationary, thermal and quantization noise, to cite just a few overlapping types (Figure 1).
Figure 1 White noise, one of the many noise types, is popular since it occurs widely in nature and is also among the easiest to analyze and understand. Source: www.real-statistics.com
In general, the “entry” paths for noise fall into three broad categories:
- Noise that is embedded into the signal of interest; for instance, static on a modulated signal.
- Noise that enters the system independent of the signal of interest, such as noise created due to conducted or radiated EMI/RFI.
- Internal self-created noise of the circuitry and its components.
Noise becomes a limiting factor in many designs. It can degrade signal-to-noise ratio (SNR) and increase bit error rate (BER), Which can force retransmits and can limit the sensitivity of an analog channel or system. As a project manager once quipped to me, “if it wasn’t for noise, half our problems would just disappear.”
As a consequence of this important and repeated lesson that “noise is a problem,” it’s hard to imagine that noise can ever be a good thing. However, the reality is that it can be beneficial in a sort of “judo-like” move, where the “bad” of noise can be leveraged to create “good.”
How so? I can think of these situations:
- By adding random noise “dither” to a signal being digitized, performing multiple analog/digital conversions, and then doing some basic statistical analysis on the results, it is possible to increase the resolution for the A/D converter beyond its physical rating. Of course, there is a limitation and a tradeoff, as the signal being digitized must be stable during the conversions, and the time to do a higher-resolution conversion is the base A/D conversion time multiplied by the number of conversion cycles.
- Mechanical noise—also called dither—is often added to devices such as servo-valves to prevent static friction (stiction) from impeding the start-up motion of the valve spool. This dither is a low-amplitude signal added to the valve control signal, typically between 100 Hz and 1000 Hz, which keeps the spool moving at all times. The challenge is to add just enough dither to prevent stiction, yet not so much that it adds significant error to the valve position and thus control of fluid flow.
- Noise sometimes needs to be deliberately added to a channel to let users know the system is turned on and working, rather than being disconnected. When the digital cellular GSM standard was introduced in the 1990s to replace the existing analog cellular system, users found it unsettling that there would be times when the audio channel was absolutely silent while neither person was speaking. They would actually hang up and try to call again.
The solution was to add “comfort noise” at the receiving end, a synthetic background noise to fill the “sound of silence” in a transmission. Comfort noise is also used in some voice-over-IP (VoIP) links for the same reason. This comfort noise is presented at a low but audible volume level, and is usually varied based on the average volume level of received signals to minimize jarring transitions. There’s even a recommended standard—ETSI TS 100 963—covering how to create this noise and at what levels.
- Noise can also be used as the encoding key in encryption systems as long as the sender and receiver have access to the same random sequence and timing. This encryption noise can be a pseudorandom sequence which is relatively easy to create and recreate, or it can be a truly random one using phenomena such as Johnson noise, electron motion, quantum properties of photons, or even nuclear decay.
Using random noise this way is not a new idea. The SIGSALY system was used in World War II to scramble phone calls between President Roosevelt and Prime Minister Winston Churchill; the noise source was a 78-rpm record with recorded vacuum-tube random noise of which only two copies were made, one for each side of the link. Synchronizing the record-player turntables at each end was a major problem (Figure 2). There’s a very detailed discussion about SIGSALY here, but it is hard for us to graph the hardware it involved. Each terminal node consisted of more than 30 full-height 19″ racks plus four turntables; weighed 50,000 kg, consumed 30 kW, and had special air-conditioning requirements.
Figure 2 These are two turntables used in the mammoth, super-secret SIGSALY scrambler system in WWII to play the records on which truly random noise had been recorded and then used for encoding and decoding the audio signal. Source: Crypto Museum
- When added to electric and hybrid vehicles, one of the characteristics of EVs and of HEVs in electric mode is that they are almost noiseless at low speeds. This can be disconcerting and even dangerous to nearby pedestrians, bicyclists, and especially those who have vision impairment. As a result, these cars in the United States and worldwide are now required—via a phase-in schedule which has been delayed—to emit noise of specified loudness and spectrum when traveling below 30 km/hour or 18.6 mph (Figure 3). Details are available here and here; then there is the 137-page Environmental Assessment from the National Highway Traffic Safety Administration (NHTSA). A cynic might say that making very quiet vehicles is an example of the maxim that “no good deed goes unpunished.”
Figure 3 Here is an example of an A-weighted, one-third octave plot of noise emission from a vehicle passing at 10 mph. Source: NHTSA
These are just a few examples, of course. Gaussian, white, pink, and other noise spectra can be used to mask offending sounds, cloak and cloud images, and hide other information. Have you ever deliberately used acoustic, electrical or other noise to solve, mask, or deal with a problem? Did you have to go against your initial feelings that “noise is a bad thing”? Did you have to convince others of the validity of the approach?
- Which filters are noisier – analog or digital?
- Optimize mixed-signal Circuit/PCB design using noise modelling
- Managing power, ground and noise in microcontroller/analog applications
- “Stop” in the Name of Noise: Do I Shut Off That Switching Supply?
- Reducing Noise in DC/DC Converters with Integrated Ferrite-bead Compensation