A Butterworth 3rd order filter is a signal conditioning circuit that delivers a maximally flat passband with zero ripple while attenuating unwanted frequencies at a steep rate of -18 dB per octave (-60 dB per decade). In a real circuit, inserting this filter between a sensor and an analog-to-digital converter (ADC) or within an audio signal chain fundamentally changes the noise floor: it eliminates high-frequency noise and prevents aliasing without altering the amplitude of the baseband signals you actually want to measure. Think of the passband as a perfectly paved highway with no speed bumps (zero amplitude ripple), ending in a steep guardrail (the roll-off).
Unlike simpler 1st or 2nd-order filters, the 3rd-order topology provides the necessary attenuation to protect high-resolution 16-bit or 24-bit ADCs from out-of-band interference, while avoiding the passband ripple inherent in Chebyshev designs. Below, we break down the exact math, component scaling, and practical bench considerations for building one.
Filter Characteristics and Component Scaling
A 3rd-order Butterworth low-pass filter is typically built by cascading a 1st-order passive RC filter with a 2nd-order active Sallen-Key stage. The transfer function denominator is a 3rd-degree polynomial with one real pole and two complex conjugate poles. For the Butterworth response, the 2nd-order stage must be tuned to a quality factor (Q) of exactly 1.0.
To design this on the bench, we start with normalized component values (where cutoff frequency $\omega_c = 1$ rad/s and base resistance $R = 1\Omega$) and then apply frequency and magnitude scaling to hit our target cutoff. The table below provides the exact scaling for a highly practical 1 kHz cutoff frequency, assuming a base capacitor value of 10 nF.
| Component | Stage / Role | Normalized Value | Scaled Value (C_base = 10nF) | Nearest 1% E96 / E12 Practical |
|---|---|---|---|---|
| R1 | Stage 1 (RC Series) | 1.000 | 15.92 kΩ | 15.8 kΩ |
| C1 | Stage 1 (RC Ground) | 1.000 | 10.0 nF | 10 nF (C0G) |
| R2 | Stage 2 (Sallen-Key Series) | 0.500 | 7.96 kΩ | 8.06 kΩ |
| R3 | Stage 2 (Sallen-Key Shunt) | 0.500 | 7.96 kΩ | 8.06 kΩ |
| C2 | Stage 2 (Sallen-Key Ground) | 4.000 | 40.0 nF | 39 nF (C0G) |
| C3 | Stage 2 (Sallen-Key Feedback) | 1.000 | 10.0 nF | 10 nF (C0G) |
Worked Numeric Example: 1 kHz ADC Anti-Aliasing Filter
Let's design a real-world anti-aliasing filter for an industrial vibration sensor. We are sampling an accelerometer using a 16-bit ADC (like the TI ADS1115) at 10 kSPS. The Nyquist frequency is 5 kHz, but to prevent high-frequency switching noise from a nearby 500 kHz buck converter from aliasing into our baseband, we need aggressive attenuation.
We set our -3 dB cutoff frequency ($f_c$) to 1 kHz. A 3rd-order Butterworth filter attenuates at -18 dB per octave. By 8 kHz (three octaves above 1 kHz), the attenuation is -54 dB. By the time we hit the 500 kHz switching noise, the theoretical attenuation exceeds -160 dB, completely obliterating the interference before it reaches the ADC's sample-and-hold circuit.
Op-Amp Selection and GBW Requirements
For the active 2nd-order Sallen-Key stage, the op-amp must have sufficient Gain-Bandwidth Product (GBW). The rule of thumb for active filters is:
GBW > 100 × f_c × Q
With $f_c = 1000$ Hz and $Q = 1.0$, we need a GBW greater than 100 kHz. While a generic LM358 (GBW ~1 MHz) technically meets this, its crossover distortion and high input offset voltage make it a poor choice for 16-bit sensor chains. Instead, select a precision op-amp like the OPA2277 (1 MHz GBW, 20 µV offset, low noise) or the ADA4522 for zero-drift applications. Using a dual op-amp package allows you to buffer the passive 1st-order RC stage if loading effects from the Sallen-Key stage become a concern, though a unity-gain Sallen-Key has high input impedance and usually doesn't require the buffer.
Where You Meet This In Practice
You will rarely see a 3rd-order Butterworth filter used for simple DC power supply ripple reduction (where a 1st-order RC or LC is sufficient). Instead, it lives in precision signal chains:
- ADC Anti-Aliasing: As demonstrated above, protecting high-resolution data acquisition systems from out-of-band RF and switching noise.
- Active Audio Crossovers: In powered subwoofers, a 3rd-order low-pass filter at 80 Hz ensures that vocal frequencies (100 Hz - 300 Hz) are attenuated by at least -20 dB before reaching the driver, preventing muddy midrange reproduction.
- Biomedical Instrumentation: In ECG machines, a 3rd-order low-pass filter at 150 Hz removes high-frequency EMG (muscle) noise and RF interference while preserving the sharp QRS complex of the heart signal without the phase distortion that would smear the waveform.
- PLL Loop Filters: In high-performance phase-locked loops, a 3rd-order passive or active filter is used to suppress reference spurs that a standard 2nd-order loop would let through.
Common Confusions and Topology Choices
When specifying filters, engineers and hobbyists frequently confuse the Butterworth response with other standard approximations, or misunderstand how to achieve the 3rd-order slope.
| Filter Type | Passband Ripple | Roll-Off (3rd Order) | Phase Linearity | Best Use Case |
|---|---|---|---|---|
| Butterworth | None (Maximally Flat) | -18 dB/octave | Moderate | General purpose, audio, ADC anti-aliasing |
| Chebyshev Type I | Yes (e.g., 0.5 dB) | Steeper than Butterworth | Poor | When you need maximum stopband attenuation and can tolerate passband ripple |
| Bessel | None | Gentler than Butterworth | Excellent (Linear Phase) | Pulse preservation, digital data lines, time-domain accuracy |
The Cascaded 1st-Order Mistake
A common beginner mistake is assuming you can build a 3rd-order filter by simply wiring three 1st-order passive RC filters in series. This fails because of loading effects. The input impedance of the second RC stage loads the first stage, shifting the pole frequencies and destroying the Butterworth alignment. To achieve a true 3rd-order response, you must isolate the stages using op-amp buffers, or use a dedicated active topology like the Sallen-Key or Multiple-Feedback (MFB) architectures.
Sallen-Key vs. Multiple-Feedback (MFB)
For a 3rd-order Butterworth, the Q of the active stage is exactly 1.0. The Sallen-Key topology is highly stable and easy to tune at Q=1. However, if you were designing a filter with a higher Q (like a Chebyshev or a narrow bandpass), the Sallen-Key becomes sensitive to op-amp GBW limitations and component tolerances. In those high-Q scenarios, the MFB topology is preferred because it relies on inverting gain and is less sensitive to the op-amp's open-loop gain limitations. For our 3rd-order Butterworth, stick to the unity-gain Sallen-Key for the lowest component count and best stability.
Frequently Asked Questions
Does a 3rd-order active filter require three op-amps?
No. A standard implementation uses one op-amp configured as a 2nd-order Sallen-Key stage, preceded by a passive 1st-order RC network. Because the Sallen-Key in unity-gain configuration has a very high input impedance, it does not significantly load the passive RC stage. If your source impedance is high, you can add a second op-amp as a unity-gain buffer between the RC stage and the Sallen-Key stage, utilizing a single dual op-amp IC (like the OPA2277).
How do I tune this filter on the bench without a network analyzer?
Inject a 1 Vpp sine wave from a function generator. Sweep the frequency up to your target $f_c$ (e.g., 1 kHz). At exactly $f_c$, the output should measure 0.707 Vpp (-3 dB). Next, sweep to $2 \times f_c$ (2 kHz). The output should be approximately 0.125 Vpp (-18 dB). If the -3 dB point is shifted, adjust the series resistors. If the passband shows peaking before the roll-off, your Q is too high—check your capacitor ratio (C2/C3) and ensure you are using tight-tolerance C0G capacitors.
Can I use this topology for a high-pass filter?
Yes. To convert the Sallen-Key low-pass to a high-pass, simply swap the positions of the resistors and capacitors in the active stage. The 1st-order passive stage also swaps its R and C. The normalized values and the Q requirements remain identical, but the component scaling math will invert to target the high-pass cutoff frequency.
For automated component selection and Bode plot generation, the Analog Devices Filter Wizard is an excellent free tool that handles the pole-zero calculations and suggests real-world op-amp part numbers. For deeper mathematical proofs on active filter alignment, refer to the classic Texas Instruments SLOA049 application note on active filter design techniques.
