Print

What kinds of noise does one encounter in electronic circuits ?

An electronic circuit consists of a maze of electronic components linked by conductors, and the physical quantities that matter (the reason why the circuit has been engineered in the first place) are currents and voltages.   As we consider information systems, the function of these currents and voltages is to be information sources.  There is of course a whole branch of electronics where one is interested in the energetic effects of these currents and voltages - power electronics - but this is not the kind of electronics we consider here.  We are interested in the information content of the currents and voltages in the circuit, not in their power effects (which are most of the time, undesired side effects).   As such, there is an 'ideal functioning' of the circuit, which would result in certain currents and voltages, and then there is the real behaviour, which will result in somewhat different currents and voltages.  Of course, part of the difference will be systematic, and due to tolerances, approximations and non-ideal behaviour of the components making up the circuit.   But a part will be "random".  This random part is what we call noise.  We mean that, with identical input signals and an identical state, repeated several times, the circuit will have voltages and currents which differ systematically from the desired values, on average, but there will also be fluctuations which are different from trial to trial.  The deviation of the average from the desired average is what we call the systematic error ; the fluctuations with average 0 is what we call noise in the circuit.

Noise being a random phenomenon, it has an ensemble description, which is a statistical description of the random fluctuations of the currents and voltages.

There are 3 main potential contributions to the noise in electronic circuits:

  1. pick up noise
  2. technological noise
  3. fundamental noise

External versus internal noise sources

Pick up noise finds its origins external to the circuit.  Technological and fundamental noise find their origins internally in the circuit.   This aspect makes that the engineering considerations of pick up noise are totally different than those that have to do with internal noise.

Pick up noise is what is usually called electromagnetic compatibility issues.  Pick up noise is the result of the effect of nearby electromagnetic phenomena which have nothing to do with the local functioning of the circuit at hand.  For instance, if you switch on the light in the room, the sudden current in the wires may propagate an electromagnetic wave which may induce currents and voltages in the circuit ; or it may induce a fluctuation in the mains, which, through the power supply of the circuit, is transmitted to certain currents and voltages in the circuit.  Or it may induce a fluctuation in the ground potentials of different, linked systems. The statistical description of pick up noise is difficult.  Pick up noise can be seriously diminished by good design practices and by extra protection and shielding measures.  Actually, pick up noise should never be an issue in a final product or design although it may take a lot of expertise to eliminate pick up noise if one is confronted with it.  In principle, pick up noise can be made arbitrarily small except in some exceptional situations where one is obliged to use very sensitive electronics in highly perturbing environments in such a way that one is limited in one's possibilities to shield it.  That said, it can be an economical choice to accept pick up noise, and to deal with it.  For instance, air power lines can be said to have been designed to accept pick up noise (lightning).  One could have shielded the lines, but economic optimisation makes people accept the statistics of lightning impact over the cost of shielding.  However, in sensitive signal processing electronics, the cost of avoiding pick up noise is almost all the time far inferior to the price of accepting pick up noise.  Design practices and shielding against pick up noise are a whole business and domain of expertise.

Fundamental noise corresponds to statistical fluctuations in current or voltage inherent to the physics of the device action, and is as such unavoidable.  The two principal fundamental noise effects are Johnson noise (thermal resistor noise) and shot noise (pn junction noise).  These two noise sources are moreover "pure entropy" noise sources, in the sense that they are fundamentally unpredictable.  The Johnson noise is directly related to the second law of thermodynamics.  Shot noise is an unavoidable quantum noise effect.  Fundamental noise only depends on a few physical parameters and not on many specific characteristics of the device.

Technological noise is noise that occurs because of technological imperfections of components.  There is burst noise and flicker noise, which tend to be low-frequency noises.  These noise sources are purely due to technological imperfections such as ions trapped in insulation layers that move under the influence of electric fields, crystal imperfections and the like), which can be reduced by the state and art of technology.

Finally, there's a kind of mixture of fundamental and technological noise: Zener noise and avalanche noise.  It has a fundamental aspect, as the Zener effect, as well as the avalanche breakdown effect has an unavoidable statistical nature ; on the other hand, the amount of noise that is generated is technology-dependent.  So although the state of the art of technology will never be able to eliminate it, the amount of noise is dependent on technology. 

Noise or signal ?

There are two cases in which what we call noise here, is actually useful.  In that case, talking about the phenomenon as "noise" is an abuse of language, as in fact, the phenomenon is the signal we're after, and our source of information.  These two applications are:

  1. electromagnetic sensor
  2. noise generator

The "pick up noise" being a coupling of electromagnetic external influences to our circuit, if those electromagnetic influences are what we're after, then pick up noise is the signal we're after.   If the "perturbation" we want to capture is an electromagnetic wave, then we're constructing an antenna.  But pick up noise can also be due to other electromagnetic influences, such as ground potential differences, quasi-static electric or magnetic fields and so on.  So even if we're constructing an antenna and we want to receive radio waves, we wouldn't like our voltages and currents that are influenced by the radio wave we're after, be perturbed by totally other sources of pick up noise we're not interested in.  In that case, the pick up phenomenon has both signal and noise aspects.

The noise from internal sources can also be considered a useful signal.  The two aspects of this signal are:

  1. its statistical properties
  2. its "true randomness"

 Depending on the application, one or the other aspect can be important in the useful application of "noise".  The statistical properties of the noise can be interesting, for instance in a measurement set up.  If one has a noise source with a white spectrum, then, applying this noise source to a linear circuit allows one to do a spectral estimation of the output, and hence of the passband of the circuit.   By adding noise to a system, one can test its robustness to noise.  Noise can be used in a technique called "dithering".  There are many applications of the statistical properties of the noise.  In many cases, this kind of noise can also be replaced by a digital equivalent which is called a pseudo-random generator.  However, especially at high frequencies, the digital version would be expensive as compared to the noise of a component.

The "true randomness" aspect of electronic noise can have several applications. The basic property of (fundamental) electronic noise is that it is a true source of entropy.  The outcome of "the next drawing" is totally unpredictable apart from its ensemble.  This comes about because fundamental electronic noise finds its origin in true physical entropy, thermodynamic or quantum-mechanical.  As such, electronic noise can be a source of "pure randomness", which has several applications by itself.  Most applications are cryptographic, but there are two others that we can mention: fair drawing (the "innocent child's hand") and limiting a communication channel's capacity.

How to handle electronic noise ?

In most applications, electronic noise is noise, that is, a nuisance that will limit the circuit's performance, or even render it useless.   Mastering the noise is hence an essential part in electronic engineering.  Very often, this aspect is overlooked and it is only during the test phase, or worse, at the customers' place that the problem is evident.   As such, expertise in "noise engineering" can save you from a lot of costs, delays and difficulties or loss of reputation if handled earlier on in the design process. 

As we saw, the differences between external and internal noise are such, that they actually are two different fields of expertise.  It is absolutely essential to find out whether one is confronted with pick up noise, or internal noise, as the ways to handle them are totally distinct.  Fighting pick up noise will help nothing when the noise is internal, and engineering internal noise will not help at all when the noise is pick up noise.  In fact, the last case it most often encountered.

In any case, one has to have a good estimation of the internal noise of a circuit.  In fact, internal noise calculations and simulations are rather well mastered with sufficient expertise.  In the case the "noise" is the signal, of course the correct estimation of the noise is essential.  Not only an "upper limit" but a correct estimation of the noise signal is then needed.  In the case of a sensitive circuit, one might think that just an upper boundary estimation of internal noise is enough.  However, it is much better to have an accurate estimation of the internal noise, because this allows one to see experimentally whether the observed noise is only due to internal noise (which is good), or whether there are unknown noise sources to which one is sensitive (such as pick up noise).  The latter case is problematic, because even if under experimental conditions, the circuit is working still properly, it means that we don't master the noise situation, and are hence vulnerable to unknown sources of error.  However in order to find out whether a circuit only suffers from known internal noise, one has to calculate it.

Pick up noise on the other hand, is much, much harder to estimate.   For sensor (antenna) design, one can use finite-element field calculations, but general circuit pick up calculations are most of the time, useless, because one cannot model the (unknown, external) origins of noise reliably.  In most cases, pick up noise has to be handled by good design practices, even some "over design" and experimental testing.  Most of the time, the cost of over design is small compared to the potential costs of a pick up problem, unless one plays in extremely competitive markets where any small saving on production cost is essential.   If your product is a relatively expensive quality product, ask yourself what's more expensive to you and your reputation: a few unnecessary shielding components and a few capacitors too much in the product, or a lot of unsatisfied customers who found your product under performing or outright useless because of pick up noise ?