Designing a Third Order Butterworth Filter: Theory, Math, and Circuit Examples

A third-order Butterworth filter is an electronic signal-processing circuit that provides a maximally flat frequency response in the passband while attenuating unwanted frequencies at a steep rate of -60 dB per decade. If you need to strip out high-frequency switching noise or RF interference before a signal hits a sensitive analog-to-digital converter (ADC) or audio amplifier, this filter prevents aliasing and harsh harmonic distortion without altering the amplitude of your target signal. Unlike other filter topologies, it achieves this aggressive roll-off without introducing any ripple in the passband.

The Math and Mechanics Behind the -60 dB Roll-Off

Every "order" in a filter corresponds to one energy-storing element (a capacitor or inductor) that contributes a -20 dB/decade (or -6 dB/octave) roll-off. Therefore, a third-order filter uses three reactive components to achieve a combined -60 dB/decade attenuation slope. Mathematically, the transfer function of a normalized third-order Butterworth low-pass filter is defined by three poles: one real pole and a complex conjugate pair. This specific pole placement is what guarantees the "maximally flat" magnitude response, meaning the first five derivatives of the gain with respect to frequency are zero at DC.

When designing or specifying filters, engineers commonly confuse the Butterworth response with other standard approximations. Here is how they actually compare in a real circuit:

If your primary goal is to preserve the exact shape of a square wave or digital pulse, a Butterworth filter will round the edges due to its non-linear phase shift. In that scenario, you would choose a Bessel filter instead. But if you simply need to guarantee that no high-frequency noise above a specific threshold reaches your microcontroller's ADC, the Butterworth's flat amplitude response is the superior choice.

Worked Numeric Example: Designing a 1 kHz Audio Low-Pass

Let’s design a practical 3rd-order active low-pass Butterworth filter with a cutoff frequency ($f_c$) of roughly 1 kHz. We will use a cascaded topology: a 1st-order passive RC stage followed by a 2nd-order active Sallen-Key unity-gain stage. This avoids the need for precise gain-setting resistors while maintaining stability.

For a 3rd-order Butterworth, the 1st-order stage handles the real pole, and the 2nd-order stage handles the complex conjugate pair, which requires a quality factor ($Q$) of exactly 1.0. Using the Analog Devices Filter Wizard or standard transfer function equations, we can derive the component ratios.

Stage 1: First-Order RC Network

We need an $f_c$ of ~1 kHz. Using the formula $f_c = \frac{1}{2\pi RC}$:

• Choose $R_1 = 16\text{ k}\Omega$ (Standard E24 value)
• Choose $C_1 = 10\text{ nF}$ (Standard value)
• Calculated $f_c = \frac{1}{2\pi \times 16,000 \times 10 \times 10^{-9}} \approx 994.7\text{ Hz}

Stage 2: Second-Order Unity-Gain Sallen-Key

For a unity-gain Sallen-Key to achieve $Q = 1.0$ with equal resistors ($R_A = R_B$), the capacitor ratio must be $C_A = 4 \times C_B$. We keep the resistors at $16\text{ k}\Omega$ to maintain consistent impedance levels and minimize thermal noise.

• $R_A = 16\text{ k}\Omega$, $R_B = 16\text{ k}\Omega$
• $C_B = 5\text{ nF}$ (Feedback capacitor to ground)
• $C_A = 20\text{ nF}$ (Series input capacitor)

Verifying the math: $f_c = \frac{1}{2\pi \sqrt{R_A R_B C_A C_B}} = \frac{1}{2\pi \sqrt{256 \times 10^6 \times 100 \times 10^{-18}}} = 994.7\text{ Hz}$. The $Q$ factor calculates to $\frac{1}{2}\sqrt{\frac{C_A}{C_B}} = \frac{1}{2}\sqrt{4} = 1.0$. Perfect.

For the operational amplifier buffering the Sallen-Key stage, a dual op-amp like the MCP6002 is ideal for 3.3V microcontroller systems (like an ESP32 or Raspberry Pi Pico), while the TL072 or OPA2134 is better suited for ±12V or ±15V analog audio racks.

Where You Meet This in Practice

You will rarely see a 3rd-order Butterworth filter built with discrete inductors on a modern PCB, but the active (op-amp based) version is a workhorse in several specific domains:

• ADC Anti-Aliasing: When sampling audio or sensor data with a microcontroller, any noise above the Nyquist frequency (half your sample rate) folds back into your measurement as false low-frequency data. According to the Espressif ESP-IDF documentation, the internal SAR ADC can be susceptible to high-frequency RF rectification. A 3rd-order Butterworth set just below your Nyquist limit provides the -60 dB/decade brick-wall attenuation needed to keep your digital readings clean.
• Active Audio Crossovers: In powered subwoofer amplifiers, a 3rd-order (18 dB/octave) low-pass Butterworth filter is frequently used to route bass frequencies to the woofer while aggressively rolling off the midrange and treble, preventing the subwoofer cone from attempting to reproduce directional vocals.
• Motor Control Current Sensing: When measuring phase current in a BLDC motor driver using inline shunt resistors, the PWM switching frequency (often 20 kHz to 50 kHz) creates massive voltage spikes. A 3rd-order low-pass filter set around 2 kHz smooths the current feedback signal for the PID control loop without introducing the severe phase lag that a higher-order Bessel filter might require to achieve the same noise rejection.

Frequently Asked Questions