Module 11: Filters and Impedance Matching

Almost every signal in electronic equipment passes through filters. An anti-aliasing low-pass filter in front of an analog-to-digital converter keeps out frequencies the converter would misread. A low-pass filter after a square-wave oscillator or a class-D amplifier removes the harmonics that would otherwise cause interference. The band-pass filter in a Wi-Fi or Bluetooth module lets its own band through and keeps out everything else, and the EMI filter on a power inlet stops a switching supply's noise from escaping onto the mains. And wherever one impedance meets another — an amplifier output driving a 50-ohm load, a chip driving a filter — a matching network makes sure the power or signal transfers efficiently.

This module covers filter design from first principles, starting with a survey of all filter types and responses, moving through the three classical filter families (Butterworth, Chebyshev and elliptical), and then into crystal filters for very narrow bandwidths. The second half of the module covers practical impedance matching with L and Pi networks, and an introduction to the Smith chart — the graphical tool that brings every impedance matching problem into a single diagram.

Module Overview

Filters and impedance matching are two sides of the same coin. A filter selects which frequencies pass and which are blocked. An impedance matching network transforms one impedance to another so that maximum power can transfer between stages at the desired frequency. In practice most matching networks — L, Pi, T and Pi-L — are also filters: they transform impedance and suppress harmonics and out-of-band signals at the same time. Understanding both subjects together gives you a complete picture of how signals are shaped and delivered in a circuit.

The module begins with a survey of filter types and families that sets the vocabulary for the rest of the module: the four basic types (low-pass, high-pass, band-pass, band-stop) and the three classical response families. Each family involves a different trade-off between passband flatness, rolloff sharpness, component count and group delay.

Classical Filter Families

The next three lessons cover the families in depth, with component value tables and worked design examples. The Butterworth filter is the simplest: it has the flattest passband of any filter but the gentlest rolloff. A five-pole Butterworth low-pass filter with a 30 MHz cutoff provides about 28 dB of attenuation at 60 MHz. The Chebyshev filter trades a small amount of passband ripple for a much steeper rolloff, and the elliptical filter adds notches in the stopband to achieve the sharpest transition of all — at the cost of more components and more sensitivity to their tolerances.

Crystal Filters

When the bandwidth required is a tiny fraction of the center frequency, LC filters cannot do the job: their Q is far too low. Crystal filters exploit the extremely high Q of quartz resonators — typically 20,000 to 100,000 — to achieve bandwidths of a few hundred hertz to a few kilohertz at frequencies of several megahertz. Their ceramic cousins provide the 10.7 MHz and 455 kHz filters found in broadcast receivers, and the surface-acoustic-wave (SAW) filters in phones and wireless modules use the same principle at hundreds of megahertz and above.

Impedance Matching Networks

L and Pi networks are the standard way to match an RF amplifier, an oscillator or a wireless chip to a 50-ohm load. The Pi network transforms the low output impedance of a transistor stage to 50 ohms while simultaneously acting as a low-pass filter to suppress harmonics. The loaded Q of the network sets both the bandwidth of the match and how much harmonic suppression it provides. The lesson derives the component values from first principles and covers the Pi-L combination, which adds extra harmonic suppression.

The Smith Chart

The final two lessons introduce the Smith chart — the graphical tool invented by Philip H. Smith at Bell Telephone Laboratories in 1939, which remains in daily use in RF design today and appears on the screen of every vector network analyzer. The Smith chart represents all possible values of normalized impedance on a single diagram bounded by the unit reflection coefficient circle. Any impedance transformation — adding a series or shunt component, or moving along a transmission line — becomes a simple movement on the chart.

Lessons