7
\$\begingroup\$

I've been looking at a few power input circuits and quite a few of them suggest an inductive choke with a large bypass capacitor followed by smaller bypass capacitors.

In theory this is what I interpret the target effect is:

Bypass capacitors provide low impedance at high frequencies, and the inductor prevents current surges from entering/leaving.

Essentially, in using an inductor inline and a decoupling capacitor it's creating essentially a series LC circuit. Now one of the properties of LC circuits is they have a resonant frequency. I modeled a LC circuit with an attached purely resistive load to simulate what the circuit did under different kinds of loads:

LC filter with load

VG1 is a voltage source to simulate noise and VM1 is a product of the simulation software for measuring voltages (I'm using TINA-TI).

Here is the resulting Bode plot for this circuit:

bode plot

As expected high frequency noise is very effectively reduced (if I'm not mistaken, outside the peak region it should decay at 40 dB per decade). However, there is that ominous spike at around 20 kHz where noise is actually being very effectively amplified from the resonant frequency.

Is this typically not a factor when it comes to regulating power to a board? Why isn't the RC filter created using bypass capacitors enough? Is it because the capacitor size required for an effective RC filter is too high? Or am I missing something in my model?

The example recommended power filtering schematic comes from Atmel for their AVR micro controllers.

\$\endgroup\$
  • \$\begingroup\$ It depends on the application where such a filter is needed. If it has switching frequencies around the peak of the LC, then it will have negative effects. As for the RC version, the power dissipated by R will be (again, depends) greater than the one on the L. \$\endgroup\$ – Vlad Sep 7 '12 at 19:56
7
\$\begingroup\$

I'm more for an RC-filter than an LC, and the resonance is one reason. You have a high noise suppression above the resonance frequency, but hardly anything below. If you then want a low cutoff-frequency the inductor may become impractically large.

I've seen circuits where EMI ferrite beads are used, but these are almost useless. They do have a near zero impedance at DC, but their peak impedance (often a few hundred ohms) often lies above 50 MHz, so much noise will hardly be filtered.

But the RC-filter isn't ideal: the resistor will have a voltage drop, and if you choose a low resistance you'll need a quite large capacitor to keep the low cutoff-frequency. An RC-filter may be acceptable if the microcontroller won't need much power (don't forget what it sources to its I/Os!), but at 5 mA a 100 Ω will already drop 500 mV, which is fine if the input voltage is chosen a bit high especially for this, but which you maybe can't afford.

\$\endgroup\$
0
\$\begingroup\$

The proper way to do this is to pick a proper value and ESR for C1 in your schematic to damp the resonance.

Example:

  • Power supply output resistance: 10 mOhm
  • Total inductance 100µH (including wires)
  • Capacitance 10µF

Since damping factor Eta = R/2 sqrt(C/L) and we want Eta>1, then we easily get:

R > 2 sqrt(L/C)

In this case R>6 ohms, so put a resistor in series with your cap.

A better scheme would be 100µF and 3 ohms. Add a 1µF local cap to take care of high frequencies. It will not ring with the inductor.

Simulate the circuit, you'll see...

Of course, this will worsen HF rejection slightly, but you won't get any peak.

\$\endgroup\$

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.