Butterworth Filter

Butterworth, Chebyshev, and Bessel are three different answers to the same question — where exactly should a filter's poles sit in the s-plane — and each answer optimizes for a different one of three properties that can't all be maximized simultaneously: Butterworth chooses the flattest possible passband amplitude, Chebyshev trades a ripple in that flatness for a steeper rolloff at the same filter order, and Bessel gives up both of those to protect phase linearity instead, which is why picking a filter type is really picking which one of amplitude-flatness, rolloff-steepness, or phase-linearity matters least for a given application.

Same filter order (same pole count)
e.g. a 4th-order lowpass, four poles to place
Pole placement is the only variable
Where in the s-plane each pole sits
Three placement strategies, three different winsEach optimizes one property at the expense of the others
Butterworth: poles on a circle
Maximally flat passband amplitude, moderate rolloff, moderate phase
Chebyshev: poles on an ellipse, pulled inward
Steepest rolloff for the order, at the cost of passband ripple
Bessel: poles spread for delay flatness
Gentlest rolloff, but preserves waveform shape through the filter
Same circuit topology implements any of them
Sallen-Key or multiple-feedback stages, just different per-stage Q and f0

Butterworth's defining property is amplitude flatness, achieved by placing every pole at an equal distance from the origin on a perfect circle in the s-plane. This specific placement makes the first 2N-1 derivatives of the magnitude response equal to zero at DC, which is the mathematical definition of "maximally flat" — no other pole arrangement of the same order can produce a flatter passband; the cost is that Butterworth's transition from passband to stopband is gentler than what the same number of poles could achieve if flatness weren't the priority, and its phase response, while not as linear as Bessel's, is noticeably less linear than a casual look at its clean amplitude response might suggest.

Chebyshev deliberately allows a controlled ripple in the passband in exchange for a dramatically steeper rolloff at the same pole count, and the ripple amount is a design choice, not an accident. By pulling the poles inward onto an ellipse rather than a circle, a Chebyshev filter trades a small, specified amount of passband amplitude variation — commonly 0.1 dB, 0.5 dB, or 1 dB ripple, chosen explicitly by the designer — for a transition band that drops off noticeably faster than a Butterworth filter of the same order; this makes Chebyshev the default choice whenever a design needs to reject an unwanted frequency as close as possible to a wanted one and can tolerate some passband amplitude variation, such as channel-select filters where adjacent-channel rejection with minimal pole count matters more than amplitude flatness.

Bessel sacrifices both amplitude flatness and rolloff steepness specifically to protect group delay — how long every frequency component takes to pass through the filter — from varying across the passband. A filter with non-constant group delay distorts the shape of a pulse or a transient signal even if its amplitude response looks acceptable, because different frequency components making up that pulse get delayed by different amounts and arrive smeared relative to each other; Bessel filters are designed explicitly to keep group delay nearly constant across the passband, which is why they show up in applications like oscilloscope input stages, pulse and square-wave shaping circuits, and anywhere preserving a signal's time-domain shape matters more than achieving a sharp frequency cutoff.

The same circuit topology can implement any of these three response types, which is the direct link back to this site's Sallen-Key filter article — the pole-placement math is a design-table lookup, not a different circuit for each type. A Sallen-Key (or multiple-feedback) stage's gain and component ratios determine where its particular pole pair sits, and a filter design table or software tool simply hands a designer the per-stage gain and frequency values needed to place poles at the specific locations a Butterworth, Chebyshev, or Bessel polynomial of the desired order calls for; cascading several such stages, each contributing one pole pair from the chosen polynomial's table, builds up the complete higher-order filter — the circuit topology stays identical across all three filter types, only the numbers plugged into it change.

Same Pole Count, Three Different Priorities flatness, rolloff, and phase linearity trade against each other Amplitude response Butterworth — flat, moderate rolloff Chebyshev — ripple, steep rolloff Bessel — gentle rolloff frequency → Group delay response Bessel — nearly flat delay Butterworth — moderate delay variation Chebyshev — worst delay variation frequency → the filter that looks best on an amplitude plot often looks worst on a group-delay plot — and vice versa

Read these three filter types through a *pick-your-tradeoff* lens rather than a *best-filter-wins* lens: there is no filter design that simultaneously achieves the flattest passband, the steepest rolloff, and the most linear phase at a given order, because the three properties pull pole placement in genuinely different directions in the s-plane — choosing Butterworth, Chebyshev, or Bessel is choosing which one of those three competing goals the application actually needs, with the Sallen-Key (or any other second-order active filter) stage standing ready to implement whichever pole table the choice produces.

Take butterworth filter further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app