Designing a 4th Order Butterworth Filter: Theory, Math, and Circuits

A 4th-order Butterworth filter is an analog or digital signal conditioning circuit that provides a maximally flat passband response and attenuates unwanted frequencies at a steep rate of -24 dB per octave beyond the cutoff point. In a real circuit, it strips out high-frequency noise, switching harmonics, or RF interference without introducing the amplitude ripple that plagues other filter types. However, it trades this pristine passband flatness for increased phase shift and a doubled component count compared to a basic 2nd-order stage.

Think of a Butterworth filter like a velvet-rope bouncer at an exclusive club: everyone on the guest list (the passband) walks in smoothly without any hassle or distortion, but once you cross the threshold into the rejection zone, you are turned away aggressively and consistently. Unlike Chebyshev filters, which cause the "guest list" to experience varying delays (passband ripple), the Butterworth keeps the passband perfectly uniform.

The Math and the "Maximally Flat" Mandate

The defining characteristic of any Butterworth filter is its maximally flat magnitude response. Mathematically, this means the first $2n-1$ derivatives of the gain with respect to frequency are zero at DC (for a low-pass filter). For a 4th-order filter ($n=4$), the transfer function polynomial factors into two 2nd-order complex conjugate pole pairs.

You cannot build a stable 4th-order active filter using a single operational amplifier. The standard bench practice is to cascade two independent 2nd-order stages. The damping factors (and consequently the Quality factor, $Q$) for these two stages are strictly dictated by the Butterworth polynomial roots to maintain the flat response. If you use identical $Q$ values for both stages, you will accidentally build a generic 4th-order filter with a peaked response, not a Butterworth.

As the table highlights, the 4th-order configuration is often the "sweet spot" in embedded systems and audio. It provides a steep enough roll-off (-24 dB/octave) to effectively isolate signals near the Nyquist limit without the excessive phase shift (-270° or more) and op-amp noise accumulation of 6th or 8th-order designs.

Worked Numeric Example: 1 kHz Active Low-Pass

Let's design a 4th-order low-pass Butterworth filter with a cutoff frequency ($f_c$) of 1 kHz. We will use the Equal-Component Sallen-Key topology, which is highly favored on the bench because it uses identical resistors and identical capacitors for each stage, simplifying your BOM (Bill of Materials). The trade-off is that the op-amp must be configured for a specific non-unity gain to set the correct $Q$.

The 4th-order Butterworth polynomial dictates two distinct $Q$ values for the cascaded stages:

• Stage 1 (Lower Q): $Q_1 = 0.5412$
• Stage 2 (Higher Q): $Q_2 = 1.3066$

In an equal-component Sallen-Key filter, the required op-amp closed-loop gain ($K$) is calculated using the formula $K = 3 - \frac{1}{Q}$. The gain is set by the feedback network: $K = 1 + \frac{R_f}{R_g}$.

Stage 1 Component Calculation

• Required Gain: $K_1 = 3 - \frac{1}{0.5412} = 3 - 1.8478 = 1.1522$
• Feedback Ratio: $\frac{R_f}{R_g} = 0.1522$
• Choose $R_g = 10 \text{ k}\Omega$. Therefore, $R_f = 1.522 \text{ k}\Omega$.
• Bench Selection: Use a 1.5 kΩ standard resistor (or a 1.5k + 22Ω series combo for precision).

Stage 2 Component Calculation

• Required Gain: $K_2 = 3 - \frac{1}{1.3066} = 3 - 0.7654 = 2.2346$
• Feedback Ratio: $\frac{R_f}{R_g} = 1.2346$
• Choose $R_g = 10 \text{ k}\Omega$. Therefore, $R_f = 12.346 \text{ k}\Omega$.
• Bench Selection: Use a 12.4 kΩ (1% tolerance) standard E96 resistor, or a 12k + 330Ω series combo.

Verification Step: Before soldering, simulate this exact netlist in LTspice. Apply an AC sweep from 10 Hz to 100 kHz. You should see a perfectly flat 0 dB response up to roughly 800 Hz, crossing exactly -12 dB (the sum of two -6dB 2nd-order stages) at 994 Hz, and dropping at -24 dB/octave thereafter.

Where You Meet This in Practice

You will rarely see a 4th-order Butterworth filter used for simple DC power supply ripple reduction; a basic RC or 2nd-order LC is sufficient there. The 4th-order Butterworth shines in applications where signal integrity in the passband is non-negotiable, but out-of-band noise must be aggressively crushed.

2. Audio Crossovers and DSP:
In active loudspeaker management, a 24 dB/octave (4th-order) slope is the industry standard for separating subwoofers from mid-range drivers. While the famous Linkwitz-Riley crossover is technically two cascaded 2nd-order Butterworths (making a 4th-order with a -6dB point at $f_c$), pure 4th-order Butterworth alignments are heavily used in DSP room correction and synthesizer voltage-controlled filters (VCFs) where passband amplitude ripple would color the audio tone.

3. Vibration and Piezo Sensor Conditioning:
Piezoelectric accelerometers generate massive high-frequency ringing when subjected to mechanical shocks. A 4th-order Butterworth low-pass filter removes this ringing without introducing the passband peaking of a Chebyshev filter, which would falsely amplify the primary vibration frequency you are trying to measure.

Common Confusions and Bench Mistakes

Even experienced hobbyists and junior engineers trip over a few specific nuances when moving from textbook theory to physical PCB layout.

Confusion 1: Butterworth vs. Chebyshev vs. Bessel

People frequently confuse the Butterworth with other standard approximations. Here is the quick decision matrix:

• Choose Butterworth when you need absolute amplitude flatness in the passband and can tolerate moderate phase shift. (Best for general instrumentation and audio).
• Choose Chebyshev when you need the absolute steepest roll-off possible to kill a specific nearby interference frequency, and you don't care about passband ripple. (Best for RF channel selection).
• Choose Bessel when you need to preserve the shape of a square wave or pulse. The Bessel has a very slow roll-off, but its linear phase response prevents time-domain ringing. (Best for digital data lines and pulse-width modulation filtering).

Confusion 2: The "Single Op-Amp" Myth

A persistent mistake is attempting to force a 4th-order response out of a single Sallen-Key stage by tweaking component ratios to extreme values. A single op-amp can only provide two poles (a 2nd-order response). To get four poles, you must use two op-amps (or a single dual-op-amp IC like the OPA2134 or TL072). Attempting to simulate a 4th-order response in a single stage will result in an unstable, oscillating circuit.

Mistake 3: Ignoring Op-Amp Gain-Bandwidth Product (GBWP)

For a 1 kHz filter, a generic LM358 works fine. But if you shift this exact same 4th-order design up to 100 kHz, the LM358 (GBWP ~1 MHz) will fail miserably. The second stage requires a $Q$ of 1.3, meaning the op-amp must have sufficient open-loop gain at the cutoff frequency to maintain the feedback loop. Rule of thumb: Your op-amp's GBWP should be at least $50 \times f_c$ for a 4th-order Butterworth. For a 100 kHz filter, you need an op-amp with a GBWP of at least 5 MHz (like the OPA1678).

For deeper mathematical derivations and automated component calculators, refer to the Analog Devices MT-202 Filter Tutorial or utilize the Texas Instruments FilterPro Design Tool to verify your specific E24/E96 component selections before cutting your PCB.