Second Order Filter Design: Sallen-Key Topology and Roll-Off Explained

A second order filter is an electronic circuit that attenuates frequencies beyond its cutoff point at a rate of 40 dB per decade (or 12 dB per octave), utilizing two energy-storing components to shape the frequency response. What this changes in a real circuit is the steepness of the transition band and the phase shift (up to 180°), allowing you to aggressively separate a desired signal from high-frequency noise without resorting to impractically large inductors or capacitors. Makers commonly confuse the "order" with the literal component count—an active second-order filter often uses five or more components—or blindly assume a steeper roll-off is universally better, ignoring the group delay and phase shift that can smear audio transients or destabilize a PID control loop.

Think of a first-order filter as a single speed bump that slows traffic, while a second-order filter is a speed bump followed immediately by a sharp chicane—it forces high-speed traffic to a halt much more aggressively, but requires careful navigation at the transition.

The Core Concept and Roll-Off Mechanics

The "order" of a filter defines how many reactive components (capacitors or inductors) are actively shaping the poles of the transfer function. A second order filter introduces two poles. This yields a roll-off of 40 dB/decade past the cutoff frequency ($f_c$), which is twice as steep as a first-order RC filter (20 dB/decade).

Worked Numeric Example: 1 kHz Active Low-Pass Design

Let us design a unity-gain-equivalent Sallen-Key Butterworth low-pass second order filter to clean up a PWM signal or DAC output before it hits a microcontroller ADC. Our target cutoff frequency ($f_c$) is 1,000 Hz.

The Sallen-Key topology is the industry standard for active second order filters because it uses a single op-amp and is relatively insensitive to op-amp gain-bandwidth limitations. For a Butterworth response with equal resistors ($R_1 = R_2$) and equal capacitors ($C_1 = C_2$), the required non-inverting gain of the op-amp stage must be exactly 1.586.

Step 1: Select the Capacitors

Capacitor values dictate the impedance of the filter. Too high, and you invite noise; too low, and you load the op-amp. We will select 10 nF (0.01 µF) for both $C_1$ and $C_2$. Crucial: These must be C0G/NP0 ceramic dielectrics, not X7R.

Step 2: Calculate the Resistors

Using the standard cutoff formula for equal-component Sallen-Key:

$$R = \frac{1}{2 \pi f_c C}$$

$$R = \frac{1}{2 \pi \times 1000 \times 10 \times 10^{-9}} = 15,915 \Omega$$

The closest standard 1% resistor value is 16.0 kΩ. We will use 16.0 kΩ for both $R_1$ and $R_2$.

Step 3: Set the Op-Amp Gain

To achieve the Butterworth Q of 0.707, the gain $A$ must be 1.586. The gain formula for a non-inverting amplifier is $A = 1 + (R_f / R_i)$.

$$1.586 = 1 + \frac{R_f}{R_i} \implies \frac{R_f}{R_i} = 0.586$$

If we choose a standard 10.0 kΩ resistor for $R_i$, then $R_f$ must be 5.86 kΩ. The closest 1% value is 5.90 kΩ.

Where You Meet Second Order Filters in Practice

You will rarely see a first-order filter in professional mixed-signal or audio designs. The 40 dB/decade roll-off of a second order filter is the practical minimum for the following scenarios:

• Anti-Aliasing for ADCs: When feeding a 12-bit ADC (like the ADS1115) sampling at 10 kSPS, Nyquist dictates a 5 kHz maximum frequency. A second order filter at 2 kHz ensures that 10 kHz switching noise from nearby DC-DC converters is attenuated by roughly -28 dB, preventing it from folding back into your measurement band.
• PWM to Analog DAC Smoothing: Converting a 20 kHz microcontroller PWM signal into a clean DC voltage requires stripping the 20 kHz fundamental and its harmonics. A second order LC or active RC filter at 200 Hz drops the 20 kHz ripple by -40 dB, yielding a clean analog control voltage for motor drivers or lighting.
• Audio Crossovers and DAC Reconstruction: Digital-to-Analog Converters output stair-step waveforms rich in high-frequency sampling images. A second order (or higher) reconstruction filter smooths these steps back into a continuous analog waveform without introducing audible phase smearing in the 20 Hz–20 kHz band.

Topology Decision Tree: Which Circuit to Build

Do not default to an active op-amp filter for every application. Use this decision matrix to select the correct second order topology and concrete component picks for your specific constraint set.

Real-World Pitfalls and Component Selection

Theory assumes ideal components; the workbench proves otherwise. When building second order filters, avoid these three common failure modes:

1. The Capacitor Dielectric Trap

If you build the 1 kHz filter above using standard X7R or Y5V ceramic capacitors, your filter will fail. X7R capacitors exhibit a massive voltage coefficient (losing up to 50% of their capacitance at rated voltage) and are microphonic (they generate piezoelectric noise when subjected to mechanical vibration). For any filter in the signal path, you must specify C0G (also known as NP0) dielectrics. They cost a few cents more but maintain stable capacitance regardless of voltage, temperature, or vibration.

2. Ignoring Op-Amp Gain-Bandwidth Product (GBW)

An op-amp's open-loop gain drops as frequency increases. If your filter's cutoff is 100 kHz, and you use a legacy LM358 (GBW ~1 MHz), the op-amp will run out of gain to maintain the feedback loop, and your 40 dB/decade roll-off will degrade into a messy, unpredictable curve. Rule of thumb: Your op-amp's GBW must be at least 50 to 100 times higher than the filter's cutoff frequency. For a 100 kHz filter, you need an op-amp with a minimum 10 MHz GBW, like the OPA365.

3. The Cascaded First-Order Illusion

A frequent question is: "Can I just put two first-order RC filters in series to get a second order filter?" Mathematically, yes, it creates two poles. Practically, the second stage loads the first stage, altering the transfer function. Even if you buffer them with op-amps, cascading two identical first-order filters yields a Q factor of 0.5. This results in a sluggish, overdamped response that starts rolling off long before the actual cutoff frequency. To get the sharp, maximally flat Butterworth knee (Q = 0.707), you must use a true second order topology like Sallen-Key where the components interact intentionally.

When designing your next mixed-signal board or audio module, default to the Sallen-Key second order topology. Select C0G capacitors, match your op-amp GBW to your cutoff frequency, and use the 1% resistor values calculated above to guarantee a stable, predictable 40 dB/decade roll-off on the first hardware spin.