A Butterworth filter is a type of signal filter designed to have the flattest possible frequency response in its passband, meaning it lets through the frequencies you want without boosting or weakening any of them unevenly. This “maximally flat” characteristic is what sets it apart from other filter designs and makes it one of the most widely used filters in electronics, audio processing, and biomedical engineering.
What “Maximally Flat” Actually Means
Every filter has a job: let certain frequencies through (the passband) and block others (the stopband). The challenge is that most filter designs introduce some degree of ripple, small bumps and dips in the signal strength across the frequencies they’re supposed to pass cleanly. A Butterworth filter is specifically engineered to eliminate this ripple entirely within the passband.
Think of it like a perfectly level table. Other filter designs might give you a surface with small waves or ridges, which can distort the signal passing through. The Butterworth filter keeps things as flat as mathematically possible, so a 100 Hz tone and a 500 Hz tone (assuming both are in the passband) come through at essentially the same strength. This is why the design is called “maximally flat”: no other filter of the same complexity can achieve a flatter passband response.
How Filter Order Changes Performance
A Butterworth filter’s behavior depends heavily on its “order,” represented by the number N. The order controls how sharply the filter transitions from passing frequencies to blocking them. A first-order Butterworth filter has a gentle slope at the cutoff point, rolling off at 20 decibels per decade. A second-order filter doubles that to 40 dB per decade, a third-order hits 60, and so on. Each increase in order makes the wall between “pass” and “block” steeper.
The tradeoff is complexity. A higher-order filter requires more components in a hardware circuit or more computation in software. It also introduces more phase shift and group delay, meaning the signal takes slightly longer to pass through and different frequencies experience slightly different delays. For many applications, a fourth-order Butterworth filter strikes a practical balance between sharp cutoff and manageable complexity.
Types: Low-Pass, High-Pass, and Bandpass
The Butterworth design can be applied to any standard filter configuration. A low-pass Butterworth filter passes frequencies below a chosen cutoff and attenuates everything above it. A high-pass version does the opposite, blocking low frequencies while passing high ones. A bandpass Butterworth filter passes only a specific range of frequencies, blocking everything above and below that window.
The cutoff frequency is the point where the output signal drops to about 70.7% of its original amplitude (a 3 dB reduction). Below that point in a low-pass design, the response stays flat. Above it, the signal falls off at a rate determined by the filter order.
How It Compares to Other Filter Types
The Butterworth filter sits in a middle ground between two other common designs, and understanding those differences helps clarify when to pick each one.
- Chebyshev filters trade passband flatness for a steeper rolloff. They allow ripple in the passband (Type 1) or stopband (Type 2), and the more ripple you tolerate, the sharper the transition becomes. This makes them useful when you need aggressive frequency separation and can accept some unevenness. The downside is a noticeably non-linear phase response, which can distort pulse-shaped signals and cause problems for demodulators.
- Bessel filters prioritize phase linearity above all else. They preserve the shape of signals passing through them better than any other common design, making them ideal for applications where timing relationships between frequency components matter. The cost is a very gradual rolloff, the gentlest of the group.
- Elliptic (Cauer) filters offer the sharpest rolloff of any standard design but introduce ripple in both the passband and stopband. They’re the best choice when you need the tightest possible transition band and can tolerate some signal distortion.
The Butterworth filter offers better selectivity than a Bessel filter and a cleaner passband than a Chebyshev or elliptic filter. Its phase linearity is not as good as a Bessel filter’s, but it’s considerably better than a Chebyshev’s. This balanced profile is exactly why it shows up so often as a default choice.
Where Butterworth Filters Are Used
Audio equipment relies heavily on Butterworth filters because the flat passband means no coloring of the sound. Crossover networks in speakers, for instance, use them to split audio signals between woofers and tweeters without introducing audible bumps in frequency response.
In biomedical engineering, Butterworth filters are a standard tool for cleaning up physiological signals. Heart monitors typically use a fourth-order Butterworth bandpass filter set between 0.5 and 40 Hz to remove baseline drift and high-frequency electrical noise from ECG recordings while preserving the shape of the heartbeat waveform. The flat passband and relatively linear phase response are critical here because even small distortions could shift the apparent timing of heartbeat peaks, throwing off heart rate variability measurements. Research comparing filter types for cardiovascular signal processing has consistently shown that Butterworth filters outperform alternatives for this purpose due to their absence of passband ripple and predictable peak timing.
Wearable fitness devices processing pulse signals from optical sensors (PPG) also rely on Butterworth filters. These sensors are prone to motion artifacts and baseline wander, and the filter’s characteristics help clean up the signal without introducing the kind of timing errors that would make heart rate data unreliable.
Beyond medicine and audio, Butterworth filters appear in telecommunications, radar systems, image processing, and control systems. Any application that needs clean frequency separation without signal distortion in the passband is a natural fit.
The Group Delay Problem
No filter is perfect, and the Butterworth filter’s main limitation is its group delay behavior. Group delay measures how long different frequency components take to pass through the filter. Ideally, all frequencies would experience the same delay, keeping the signal’s shape intact. The Butterworth filter’s group delay is reasonably flat across most of the passband but becomes uneven near the cutoff frequency, especially at higher orders.
This means that frequencies close to the cutoff get delayed more than frequencies in the middle of the passband, which can smear or distort transient signals like sharp pulses. For most applications this effect is minor, but in situations where precise timing of signal features matters, engineers sometimes pair a Butterworth filter with an all-pass equalizer circuit that compensates for the uneven delay without changing the frequency response. Alternatively, they may choose a Bessel filter if timing accuracy is the top priority and a gentle rolloff is acceptable.
Digital vs. Analog Implementation
Butterworth filters were originally designed as analog circuits using resistors, capacitors, and inductors (or operational amplifiers in active designs). Today they’re just as commonly implemented digitally in software or on microprocessors. Digital versions use mathematical transformations to replicate the analog filter’s behavior on sampled data.
In practice, designing a digital Butterworth filter is straightforward in most programming and engineering environments. You specify the filter order, the cutoff frequency, and the type (low-pass, high-pass, or bandpass), and the software calculates the necessary coefficients. This accessibility is another reason the Butterworth filter remains so popular: it’s easy to implement, well-understood, and predictable in its behavior across a wide range of applications.