Filters and Frequency Selective Circuits
Electronic filters are circuits that selectively pass or attenuate signals according to their frequency content. They are among the most fundamental building blocks in analog signal processing, appearing in nearly every electronic system: radio receivers use them to select one station while rejecting the rest, power supplies use them to strip alternating-current ripple from a direct-current output, audio systems use them to shape tonal balance, and data converters depend on them to band-limit signals. This page surveys the core ideas and points to detailed treatment of each topic.
A filter is described by where it passes signals and where it rejects them, by how sharply it makes that transition, and by what trade-offs its underlying mathematics imposes on flatness, phase, and component count. The sections below introduce these concepts at an overview level. For design procedures, circuit topologies, and worked component-level detail, follow the in-text links into the Filter Design and Implementation subtree.
Filter Functions
Filters are classified first by the shape of their magnitude response, that is, by which frequency ranges they pass and which they reject:
- Low-pass filters pass frequencies below a cutoff and attenuate those above, useful for removing high-frequency noise or limiting bandwidth.
- High-pass filters pass frequencies above a cutoff and block those below, commonly used to remove direct-current offsets or low-frequency interference such as 50/60 Hz hum.
- Band-pass filters pass a defined band while rejecting frequencies on either side, as in radio channel selection and audio equalization.
- Band-stop (notch or band-reject) filters reject a defined band while passing frequencies on either side, often to suppress a single interfering tone or power-line harmonic.
- All-pass filters pass every frequency at equal gain but introduce a deliberate, frequency-dependent phase shift, used for delay and phase equalization.
A filter's order sets the steepness of the transition between passband and stopband. Each added order contributes roughly 20 dB per decade (6 dB per octave) of asymptotic rolloff, so a first-order section rolls off at 20 dB/decade, a second-order section at 40 dB/decade, and so on. Order equals the number of reactive elements in a passive network, or the number of transfer-function poles in an active design. Higher order sharpens selectivity at the cost of more components, greater phase shift, and tighter tolerance demands.
Response Families
For a given order, several mathematical approximations distribute the inevitable imperfections differently, letting the designer optimize for flatness, sharpness, or phase fidelity:
- Butterworth: maximally flat passband with no ripple, at the cost of a gentle transition. A good default when flat magnitude matters.
- Chebyshev (Type I): equiripple in the passband in exchange for a steeper transition than Butterworth of the same order. Chebyshev Type II (inverse Chebyshev) keeps a flat passband and places the ripple in the stopband instead.
- Elliptic (Cauer): equiripple in both passband and stopband, achieving the steepest transition possible for a given order. Where Butterworth might need tenth order and Chebyshev fifth, an elliptic filter may meet the same mask at third order.
- Bessel: maximally flat group delay (near-linear phase), preserving waveform shape for pulse and step signals, at the cost of the gentlest magnitude rolloff of the common families.
The choice is a trade-off: sharper magnitude selectivity generally degrades phase linearity, and vice versa. Selecting the right approximation, then translating it into poles, zeros, and component values, is the subject of Filter Synthesis and Design Methods.
Implementation Approaches
The same response can be realized with different circuit technologies, each with a distinct niche.
Passive Realizations
Networks built only from resistors, capacitors, and inductors require no power supply and add no active-device noise. Simple RC sections handle non-critical low-pass and high-pass filtering, while LC resonators provide the steep rolloff and high quality factor (Q) needed for band-pass and band-stop functions. Inductors, however, are bulky and lossy, which makes them impractical at audio frequencies but valuable at radio frequencies and high power. Ladder, pi, and T networks, along with crystal and ceramic resonators that reach Q values in the thousands, are detailed in Passive Filter Networks.
Active Realizations
Active filters add operational amplifiers to provide gain, buffer each stage against loading, and synthesize inductor-like behavior without physical inductors. This makes them the preferred choice below several megahertz. Workhorse second-order building blocks include the Sallen-Key and multiple-feedback topologies, while state-variable and biquad structures offer independent control of frequency and Q for high-Q work. Switched-capacitor designs replace resistors with clocked capacitors, giving a clock-tunable cutoff that suits integrated circuits. These topologies, with their sensitivity and tuning trade-offs, are covered in Active Filter Topologies.
Passive Versus Active: Choosing
The practical decision turns on frequency, power, and precision. Passive filters dominate at radio frequencies and where signals carry significant power or no supply is available. Active filters win at low frequencies, where they avoid impractical inductors, and where gain or buffering is needed. Digital filtering becomes attractive once a signal is already sampled and perfect repeatability or otherwise unrealizable responses are required, though analog anti-aliasing and reconstruction filters remain mandatory at the boundaries of any sampled system.
Applications and Specialized Filters
Beyond the canonical four responses, many applications call for purpose-built filters. Anti-aliasing and reconstruction filters band-limit signals at the input and output of data converters. Notch and twin-T networks remove a single interfering frequency such as power-line hum. Audio crossover networks split a signal among woofer, midrange, and tweeter; Linkwitz-Riley designs are favored because their outputs cross at -6 dB and sum to a flat, in-phase response. Tracking filters, adaptive filters, and biomedical signal filters address still more specialized needs. These cases, with their distinct constraints, are treated in Specialized Filter Applications.
Conclusion
Frequency-selective circuits give the analog designer direct control over a signal's spectrum, separating wanted information from noise and interference. The essential decisions are the same across every application: choose the response shape (low-pass, high-pass, band-pass, band-stop, or all-pass), select an approximation that balances flatness against transition sharpness and phase, fix the order, and pick a passive, active, or hybrid realization suited to the frequency, power, and precision at hand. From those choices follow the detailed synthesis methods, circuit topologies, and specialized variants explored throughout the Filter Design and Implementation section.