2
\$\begingroup\$

This question is a follow up on the question about the inductor choice for power supply noise filtering. One of the inductor's parameters was frequency response curve. To use this, I would need to have information about potential noise sources in the system and their magnitude. In general, what are the major "noise makers" in a system, based on which I should select choking inductors (uCs, ADCs, DACs, DSPs, CPLDs, FPGAs, clock oscillators, etc.). How to estimate the noise in the system?

\$\endgroup\$
  • 2
    \$\begingroup\$ Noise/ interference is like an onion. There are lots of layers, but it's only the top layer that you see at one time. So I could list all sorts of potential sources, but it would be a waste of time for you to track them all down. What you want to do is tackle the top layer, and then move in until the noise is low enough such that it doesn't bother you any more. As one example, I was making a filter like yours above with L and C. (To get rid of SMPS crud.) I found that the open coil inductor I first used also picked up a lot of interference and so I used a torroid. (but that may not... \$\endgroup\$ – George Herold Dec 4 '14 at 15:59
8
+50
\$\begingroup\$

As @george-herold said, it is pretty pointless to estimate noise in a circuit, unless, I'd add, you're a circuit engineer who wants to check how much noise your circuit induces in a system. What one generally wants is minimize the impact of noise on a system.

This said, noise sources include almost anything in electronics. Anywhere an electron passes generates noise. Semiconductors do generate noise, especially when electrons cross the gap in diodes or transistors, which happens quite often, to say the least.

Among the most common (internal) sources you have:

  • Switching power supplies : Power switching creates hi-frequency ripples that can spread through wires. Can be narrow-band noise at best.

  • Digital circuitry : Clock signals induce switching noise as well just like switching power supplies. Complex digital circuitry may generate a lot of parasitic noise if you don't take care of ground connections when mixing audio and digital components. Connect analogue and digital grounds in a single point and make all ground lines converge to that point, like a star, is one of the best design practices, for instance. Also avoid ground loops.

  • Semiconductors : A simple diode in reverse polarization generates white noise (IIRC) depending on silicon pureness, voltage and intensity. Signal-to-noise ratio is indicated with amplifier circuits, for instance.

  • High-impedance lines : high impedance inputs are very sensitive to electromagnetic parasites and act like an antenna when not shielded properly, which may as well propagate the input noise to components in the output chain. Sensors with differential inputs are less sensitive to parasites, which induce common-mode parasites, for instance but appropriate shielding is best whenever applicable.

Common external sources include:

  • Electrical environment : if your device is wired to the mains, for instance, in an industrial plant. Industry and automotive environments are known to be "polluted" electrically so any device needs some good insulation against such parasitic noise to work in those environments. Also bear in mind power line communications, which can propagate narrow band noise (at best) if not filtered properly.

  • Friction-based electrical contacts : electric engines with contacts on a mobile element also generate a lot of noise and parasites, due to the inductive nature of the component.

  • Radiation and electromagnetic parasites : with the growth of wireless components comes more electromagnetic interferences, which requires appropriate shielding.

Noise is always present in the working frequency range of an electronic circuit. Therefore all you can do is reduce its energy, i.e. reduce its spectrum by narrowing its bandwidth using L/C filters for instance.

Depending on the circuit type you also have to beware of inductors or anything that can act like one. Use cores whenever applicable.

Again, it's pretty pointless to enumerate noise sources as you rather want to limit noise using all possible means, which the best design practices are all about.

| improve this answer | |
\$\endgroup\$
1
\$\begingroup\$

Main sources digital parts that use a lot of power

  • FPGAs
  • uC and uP
  • DSPs

Analog parts of note

  • Switching power supply
  • RF transmitters
  • Motor drivers
| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Brian, would you be able to provide a more detailed answer for noise estimation and inductors choice? It would be great to know some quick way to guesstimate the noise of the system based on the set of commonly used components that have a dominant noise contribution to the system. Furthermore, how to choose a inductive filter (choke) based on the estimated noise. \$\endgroup\$ – Nazar Dec 4 '14 at 15:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.