In daily life, we have the weather to serve as a universal means of starting and maintaining conversations. In the electronics world, noise can play a similar role. It’s always present and always causing some sort of problem. If a design review meeting is slowing down but no one is willing to relinquish the conference room just yet, all you need to do is mention the n-word and the discussion should be lively for at least the next half an hour. If your newest PCB is working fantastically well but you don’t want to sound overconfident when someone comes and asks you about it, just complain about the noise. If your PCB is failing miserably and you’re not sure why, blame it on noise, which is always guilty until proven innocent.
What Is Noise?
This question is somewhat difficult to answer if you try to go really deep, because you end up in the philosophical realm confronting issues like “why is life so complicated?” or “why can’t total entropy just decrease every once in a while?” But if you’re satisfied with a more superficial treatment of the topic, electrical noise is straightforward.
Here is my attempt to define electrical noise from the perspective of circuit design: Noise is a generic word that refers to variations in voltage or current that are often random, usually of relatively low amplitude, and always undesirable.
The following comments should help to unpack that definition:
- Electromagnetic noise is a ubiquitous and often significant factor in circuit design, but I limited the definition to voltage and current because in the context of a circuit the effects of electromagnetic noise are manifested by means of current and voltage variations.
- I say “often” random to account for predictable or periodic noise signals such as 60 Hz interference. However, this article focuses on fundamental noise-generating phenomena, which are random (or to say it more precisely, they appear random to human beings).
- The amplitude of noise signals can certainly be quite large, but in this context I perceive things like large transients, lightning strikes, or severe EMI as belonging to a separate category.
- There are times when we might intentionally generate noise (dithering comes to mind), but I think that the above definition is more coherent if we restrict it to undesired signals. In my opinion, noise that is intentional and useful is called “noise” simply because it resembles noise; intentional noise belongs more to the “signal” category, at least if you judge according to its origin and purpose, because it is created by the designer in accordance with the needs of the system.
I think that we now have a good idea of what noise is, which means that we are prepared to explore the next topic: Where does it come from?
The Causes of Noise
The objective here is to look at the root causes. Noise can “come from” anywhere: the air, the power supply, an LDO, a switching regulator, a resistor.... We want to go deeper, i.e., to the origins of the noise itself, rather than to the components or pathways by which noise enters a circuit.
Thermal Noise, AKA Johnson Noise
This is a fundamental reality associated with resistance to the flow of electrons. Unless we start designing circuits out of superconductors, we’ll always have thermal noise, because everything has at least a little bit of resistance.
Thermal noise is manifested as random voltage variations; it is related to temperature, resistance, and bandwidth. Higher temperature and higher resistance lead to higher noise amplitude. “Bandwidth” here refers to the range of frequencies that are relevant to the circuit. If you include more frequencies in your analysis, you’ll see more thermal noise.
Electrons don’t actually “flow” through a conductor. They sort of bump along, with potential energy accumulating and then being converted into kinetic energy each time the electron has to cross a barrier. (Think of a ball rolling over a series of bumps—velocity constantly changes as energy moves back and forth between the potential realm and the kinetic realm.)
These random variations in electron motion lead to corresponding random variations in current. In other words, noise. Shot noise is more prominent in semiconductors than in conductors because semiconductors have more barriers. Higher current leads to more shot noise, and so does wider bandwidth (again, if you include more frequencies, you see more noise).
This plot shows the noise density for a general-purpose op-amp (the ADA4666-2) made by Analog Devices. The noise density has a general decreasing trend as frequency increases, but to calculate the actual noise amplitude you have to multiply the noise density by the square root of the circuit’s bandwidth.
1/f Noise, AKA Flicker Noise
As far as I know, scientists still don’t understand flicker noise. I don’t try to stay up to date with the latest research, so maybe they’ve figured something out by now, but I think there is still at least partial mystery surrounding the underlying physical phenomena.
The bottom line is that flicker noise is generated by most electronic components and decreases in amplitude as frequency increases. The name “1/f” (i.e., “inversely related to frequency”) reminds us that the relationship between amplitude and frequency is a prominent characteristic of flicker noise. As with shot noise, higher current leads to more flicker noise.
Flicker noise decreases with frequency, such that at a certain point thermal noise becomes dominant. Plot taken from this article published by Analog Devices.
Burst Noise, AKA Popcorn Noise
This type of noise occurs only in semiconductors, but that doesn’t help us much since semiconductors are everywhere nowadays. Imperfections in the semiconductor material lead to abrupt voltage or current transitions. The rapid transitions contain high-frequency energy, but the frequency of the pulses resulting from these transitions is actually rather low:
Measured burst noise, taken from a Maxim Integrated app note on dealing with noise in the signal chain. The horizontal scale is 0.4 seconds per division.
You can hear burst noise if you amplify a contaminated signal and send it to a speaker. It sounds like popcorn popping—or at least that’s what I read; I’ve never heard it (or if I have I didn’t realize what I was hearing), so I can’t verify that it really does sound like popcorn. Burst noise is not much of a problem these days because of improvements in semiconductor manufacturing.
This covers the most significant sources of noise in electronic circuits. I hope you now have a better idea of the physical phenomena that cause so much consternation among electrical engineers. You can’t eliminate noise, but the above information gives you some ideas about how to make it less problematic. For example, to reduce thermal noise you can use smaller resistors or add a filter to limit the bandwidth, and low-current biasing techniques lead to lower shot noise and flicker noise.