2
\$\begingroup\$

Suppose I've a circuit that is affected by EMI noises that have a freq. less than 80khz and noises above 2ghz. I want to cut off this noise freq. before they reach my ICs. The circuit is powered by 5 different voltage regulator on the same pcb with a single power supply from a battery.

In the first case I want to filter freq. less than 80Khz so I realize an LC filter with 3.3uH L and a 1uF C.

My problem is, where I need to place this filter?

  • A single filter between BatteryVcc and all voltage regulator is ok?
  • A filter for each voltage regulator Vin?
  • A filter for each voltage regulator Vout?
  • In the last two cases that will rise a lot my BOM. Is necessary?

Another question I have:

  • Since I have decoupling capacitors before my final connections, placing a filter before them, will change my C in my LC filter(more Cap in parallel)?

About 2Ghz noise:

  • Should I use a ferrite bead?
  • A ferrite bead between BatteryVcc and voltage regulators will create me problems whereas some voltage regulators power MCUs?
  • There's alternatives to void ferrite bead use?
\$\endgroup\$
  • 3
    \$\begingroup\$ If you wonder where to put these filters, you need to figure out where the noise currents are flowing. Put the filter there. \$\endgroup\$ – Phil Frost Jan 13 '15 at 13:19
  • \$\begingroup\$ Also beware of asking too many questions in one. \$\endgroup\$ – Fizz Jan 13 '15 at 15:57
2
\$\begingroup\$

Noise coming into an enclosure on the conductors (conducted noise) should be filtered out closest to the source or right at the enclosure wall. Any circuit loop made by the filter components should be as short as possible. Feed-thru capacitors (shaped like short coaxial cables), inductors, and/or ferrites can help here. If there is one input with conducted noise then a single filter set might be enough for the whole system.

If the noise is being radiated into the enclosure (magnetically or electro-magnetically - eg. RF) then you might require filters at each regulator and on each sensitive component. An extra filter for each regulator output might not be needed unless the component being powered is some distance from the regulator, though standard by-pass caps and recommended regulator caps should still be used. (By-pass caps are also used to limit noise or pulse energy coming back from the component itself.)

Adding by-pass caps after an LC low-pass filter will further lower the noise and the frequency points, though this is usually a desirable effect especially on power supply lines. If for some reason you need to have a specific cut off frequency on the line you could add a parallel LC pair in the line. The caps to ground would still give the low-pass effect.

Don't go too far with too many caps or extremely large cap values. Remember that when power is switched on all those caps need to fill, this might put a big strain on the input power system or battery, (or system fuse). When power is switched off all those caps also need to drain.

To reduce RF noise from getting at sensitive circuits you could also consider shielding the whole electrical system within a metal enclosure. This is one alternative that might avoid using a large number of ferrites. Unfortunately if the internal components are producing the RF noise then there may still be the need to have several ferrites in the circuit.

For stubborn RF noise you might need to filter even the ground level power lines with a series inductor or ferrite, perhaps even with a cap connected to an earth ground or to another known quiet ground.

To filter common mode RF noise (equal noise on two opposing lines) a component such as a common mode choke can be used. This looks like a small transformer that tries to reduce RF on the two lines by winding them near each other to actively oppose the noise currents, sometimes with a ferrite core.

Using ferrites can be tricky too. These do not work the same as a standard inductor. They reduce RF by dissipating the energy into the ferrite material. Ferrite materials are also rated by frequency. You need to look over the manufacturer's spec. Ferrite material for 500 MHz might not help as much at 2GHz. A ferrite component also works best when there is actual noise current trying to flow through it.

\$\endgroup\$
  • \$\begingroup\$ At low frequencies, most chip ferrites are primarily inductive. The high loss frequency "kicks in" higher up in the band. You can see this in the graph of total impedance vs reactive impedance that most manufacturers supply. The two graphs are almost identical at low frequencies. \$\endgroup\$ – mkeith Jan 13 '15 at 21:13

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.