A 3rd-order Butterworth filter is a signal conditioning circuit that provides a maximally flat frequency response in the passband while attenuating unwanted frequencies beyond the cutoff at a steep rate of -60 dB per decade. In a real circuit, implementing this 3rd-order topology eliminates the passband ripple found in Chebyshev designs and provides a sharper transition band than a standard 2nd-order (-40 dB/decade) stage, making it the definitive choice for strict anti-aliasing and precision audio crossover applications.
The Math and Topology Behind the -60 dB/Decade Roll-Off
The defining characteristic of any Butterworth filter is its "maximally flat" magnitude response. Unlike other filter approximations that trade passband flatness for a steeper roll-off, the Butterworth polynomial ensures that the first $2n-1$ derivatives of the gain function are zero at DC (for a low-pass). For a 3rd-order filter ($n=3$), the transfer function features one real pole and one pair of complex conjugate poles.
To achieve the -18 dB per octave (or -60 dB/decade) attenuation slope, the circuit is typically constructed by cascading a 1st-order RC network with a 2nd-order active Sallen-Key stage. The complex pole pair of the 3rd-order Butterworth requires a quality factor ($Q$) of exactly 1.0. This is a critical design parameter, as a standard 2nd-order Butterworth requires a $Q$ of 0.707. If you blindly use a 2nd-order Butterworth table for the active stage of a 3rd-order design, your overall frequency response will sag and fail to remain maximally flat.
| Filter Type | Passband Response | Roll-Off Steepness | Phase / Step Response | Best Application |
|---|---|---|---|---|
| Butterworth | Maximally Flat | Moderate (-60 dB/dec for 3rd) | Moderate overshoot | General purpose, audio, ADC anti-aliasing |
| Chebyshev Type I | Ripple in passband | Steep | Poor (high ringing) | RF channel selection, strict band-edge rejection |
| Bessel | Maximally Flat | Very Slow | Excellent (no overshoot) | Pulse/step signal preservation, time-domain filtering |
Worked Numeric Example: 1 kHz Low-Pass Design
Let us design a 3rd-order low-pass Butterworth filter with a cutoff frequency ($f_c$) of 1 kHz. We will use the cascaded topology: a passive 1st-order RC stage followed by an active 2nd-order Sallen-Key stage with equal resistors and equal capacitors.
Step 1: Calculate the Base Component Values
For the 2nd-order Sallen-Key stage with equal $R$ and equal $C$, the cutoff frequency formula is $f_c = \frac{1}{2 \pi R C}$. Let us select a standard 10 nF film capacitor (C0G/NP0 dielectric for stability).
$R = \frac{1}{2 \pi \times 1000 \times 10 \times 10^{-9}} = 15,915\ \Omega$
The nearest standard E24 resistor value is 16 kΩ. Using 16 kΩ and 10 nF shifts our actual cutoff to 994.7 Hz, which is well within standard component tolerances.
Step 2: Set the Sallen-Key Gain for Q = 1.0
For an equal-component Sallen-Key low-pass filter, the Quality Factor is defined as $Q = \frac{1}{3 - K}$, where $K$ is the non-inverting closed-loop gain. Since a 3rd-order Butterworth requires $Q = 1.0$ for its complex poles:
$1.0 = \frac{1}{3 - K} \Rightarrow K = 2.0$
To achieve a gain of 2.0, the feedback network must satisfy $K = 1 + \frac{R_f}{R_g}$. Therefore, $R_f$ must equal $R_g$. We will use 10 kΩ for both the feedback resistor ($R_f$) and the ground resistor ($R_g$).
Step 3: Final Bill of Materials (BOM)
- 1st-Order Stage: $R_1$ = 16 kΩ, $C_1$ = 10 nF
- 2nd-Order Stage: $R_2$ = 16 kΩ, $R_3$ = 16 kΩ, $C_2$ = 10 nF, $C_3$ = 10 nF
- Gain Network: $R_f$ = 10 kΩ, $R_g$ = 10 kΩ
- Op-Amp: OPA2134 (Dual FET-input, low noise)
Where You Meet This in Practice
You will rarely see a 3rd-order Butterworth filter used just for basic noise reduction; its specific phase and magnitude characteristics solve distinct engineering problems.
- ADC Anti-Aliasing: When feeding an analog signal into a 16-bit SAR ADC (like the ADS1115 or the internal ADC of an STM32), you must attenuate frequencies above the Nyquist limit. A 3rd-order Butterworth provides the -60 dB/decade roll-off needed to push high-frequency noise below the ADC's noise floor without introducing the passband ripple that would ruin DC accuracy.
- Audio Crossovers: In 3-way loudspeaker designs, an 18 dB/octave (3rd-order) Butterworth slope is the industry standard for blending midrange drivers and tweeters. The maximally flat amplitude response ensures the vocal range is not colored by amplitude bumps near the crossover point.
- Load Cell Conditioning: Strain gauges and load cells output microvolt-level signals susceptible to 50/60 Hz mains hum and high-frequency motor noise. A 3rd-order Butterworth LPF set to 10 Hz cleanly isolates the slow-moving weight data without the step-response ringing that a Chebyshev filter would introduce.
Common Confusions and Mistakes to Avoid
The most frequent mistake hobbyists and junior engineers make is confusing the Butterworth response with Chebyshev or Bessel topologies. If your design requires a sharp cutoff and you don't care about a 1 dB ripple in the passband, Chebyshev is superior. If you are filtering digital pulses and need to preserve the square wave's sharp edges without overshoot, Bessel is mandatory. Using a Butterworth for pulse filtering will result in a smeared, rounded step response.
Another critical error is ignoring the op-amp's Gain-Bandwidth Product (GBW). The math assumes an ideal op-amp with infinite bandwidth. If you are designing a 100 kHz 3rd-order Butterworth filter and you use an LM358 (GBW ~1 MHz), the op-amp's internal poles will interact with your filter poles, destroying the Butterworth response and causing peaking near the cutoff. Always select an op-amp with a GBW at least 50 to 100 times higher than your target cutoff frequency.
Frequently Asked Questions
Why use a 3rd order Butterworth filter instead of a 2nd order?
A 2nd-order Butterworth filter rolls off at -40 dB/decade (-12 dB/octave). While sufficient for basic tone control or rough noise filtering, it is often too shallow for anti-aliasing before high-resolution ADCs. The 3rd-order version steepens this to -60 dB/decade (-18 dB/octave), providing an extra 20 dB of attenuation one decade past the cutoff point. This extra rejection is usually the difference between a clean digital readout and one corrupted by aliased high-frequency noise.
Can I build a 3rd order Butterworth filter with just passive components?
Technically yes, but practically no. A passive 3rd-order LC filter requires inductors, which are bulky, expensive, and suffer from parasitic series resistance (DCR) that ruins the maximally flat Butterworth response. Furthermore, passive filters suffer from insertion loss and impedance matching issues; the load impedance will directly alter your cutoff frequency. Active filters using op-amps provide buffering, zero insertion loss, and allow the use of small, precise film capacitors instead of inductors.
How does component tolerance affect a 3rd order Butterworth response?
Butterworth filters are relatively forgiving compared to high-Q elliptic or Chebyshev designs, but tolerance still matters. Using standard 5% resistors and 10% X7R ceramic capacitors will shift your cutoff frequency and introduce a slight dip or peak near $f_c$. For precision applications, use 1% metal film resistors and C0G/NP0 dielectric capacitors. C0G capacitors have near-zero temperature and voltage coefficients, ensuring your filter response does not drift as the board heats up or the signal voltage swings.
What is the best op-amp for an active 3rd order Butterworth filter?
The "best" op-amp depends on your signal type. For low-noise audio applications (20 Hz to 20 kHz), the Texas Instruments OPA2134 or NE5532 are excellent choices due to their low voltage noise and high slew rates. For DC-coupled sensor applications (load cells, thermocouples) where low offset voltage and low 1/f noise are critical, a precision chopper-stabilized op-amp like the ADA4522 is ideal. Avoid the LM358 or LM324 for anything other than slow, single-supply DC filtering, as their crossover distortion and poor high-frequency phase margin will degrade active filter performance.






