Analog Electronics
Sallen-Key Filter: The Classic 2-Op-Amp Second-Order Active Filter
With just one op-amp, two resistors, and two capacitors, the Sallen-Key topology builds a complete second-order (12 dB/octave) active filter — no bulky inductors, no loading problems, and a passband gain you can set to unity by simply shorting the feedback wire. First described in a 1955 MIT Lincoln Laboratory report by R. P. Sallen and E. L. Key, it remains the most widely taught and deployed active-filter cell on Earth, from audio anti-aliasing front-ends to the reconstruction filters inside nearly every DAC.
Formally, a Sallen-Key filter is a voltage-controlled voltage-source (VCVS) RC network wrapped around a non-inverting op-amp buffer. A single stage realizes one complex-conjugate pole pair, giving you independent control over the corner frequency f₀ and the quality factor Q. Cascade a low-pass and a high-pass and you get a band-pass; cascade several and you climb to 4th-, 6th-, or 8th-order Butterworth, Bessel, or Chebyshev responses.
- TypeSecond-order (2-pole) active VCVS filter
- Invented1955, R. P. Sallen & E. L. Key, MIT Lincoln Lab
- Roll-off−40 dB/decade (12 dB/octave)
- Cutoff (LP)f₀ = 1 / (2π√(R₁R₂C₁C₂))
- Used inAnti-aliasing / DAC reconstruction / audio crossovers
- Practical Q limit≈ 0.5 to 10 (higher Q → component sensitivity)
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What It Is and Where You'll Find It
The Sallen-Key filter is a single-op-amp cell that synthesizes a second-order transfer function — one pair of complex-conjugate poles — using only resistors and capacitors. Because the op-amp acts as a near-ideal voltage buffer (a voltage-controlled voltage source, hence VCVS), the RC network never loads down the previous stage, so cascading many sections is trivial and predictable.
- Anti-aliasing filters ahead of an ADC, to kill signal above the Nyquist frequency.
- Reconstruction (smoothing) filters after a DAC, removing image spectra and switching artifacts.
- Audio crossovers in loudspeakers (2nd- and 4th-order Linkwitz-Riley networks are two cascaded Sallen-Key cells).
- Instrumentation and biomedical front-ends (EEG, ECG) needing gentle, low-noise band-limiting.
You will see it inside op-amp datasheets, TI/Analog Devices application notes, and virtually every filter-design textbook. Its popularity comes from a rare combination: minimal parts count, non-inverting gain, and a design procedure simple enough to do by hand.
How It Works: The VCVS Mechanism
Take the classic low-pass version. The signal passes through series resistor R₁, then series resistor R₂ into the op-amp's non-inverting input; capacitor C₂ shunts that node to ground, and capacitor C₁ feeds back from the op-amp output to the junction between R₁ and R₂.
The trick is that C₁'s feedback is positive. At low frequencies the op-amp output tracks the input, so C₁ sees little voltage across it and does nothing. As frequency rises toward f₀, the output leads the mid-node voltage; the feedback through C₁ boosts the mid-node, partially cancelling the roll-off that the RC network alone would produce. This controlled positive feedback is what creates the resonant peak — i.e., sets Q — without any inductor.
The resulting transfer function is the canonical second-order low-pass:
- H(s) = K·ω₀² / (s² + (ω₀/Q)·s + ω₀²)
where K is the passband gain, ω₀ = 2πf₀ is the pole frequency, and Q is the quality factor governing peaking and damping.
Key Equations and a Worked Example
For the unity-gain (K = 1) low-pass Sallen-Key, the design reduces to two clean formulas:
- f₀ = 1 / (2π·√(R₁R₂C₁C₂)) — the corner (pole) frequency.
- Q = √(R₁R₂C₁C₂) / (C₂·(R₁ + R₂)) — the quality factor (unity-gain case).
A common simplification sets R₁ = R₂ = R. Then f₀ = 1/(2πR√(C₁C₂)) and Q = √(C₁C₂)/(2C₂) = ½·√(C₁/C₂). The capacitor ratio alone sets Q — a beautifully decoupled result.
Worked example — 1 kHz Butterworth (Q = 0.707): Pick R = 10 kΩ. Butterworth needs C₁/C₂ = 4Q² = 2, so let C₂ = 10 nF and C₁ = 20 nF (≈ 22 nF standard). Check: f₀ = 1/(2π·10⁴·√(20n·10n)) = 1/(2π·10⁴·1.41×10⁻⁸) ≈ 1.12 kHz. Q = ½·√(22/10) ≈ 0.74. Both land within component tolerance of the target — trim R or swap to C₁ = 18 nF to fine-tune. Attenuation one decade past f₀ is a full −40 dB.
Designing and Tuning in Practice
Two design styles dominate. In the unity-gain / equal-component approach you fix R₁ = R₂ and set Q purely by the C₁/C₂ ratio — robust and low-sensitivity, ideal for Butterworth and Bessel cascades. In the equal-capacitor approach you set C₁ = C₂ and control Q by the op-amp's non-inverting gain K = 1 + R_b/R_a, where Q = 1/(3 − K). This is convenient but dangerous: as K → 3, Q → ∞ and the circuit oscillates, so it is intolerant of gain error.
- Choose R around 1–100 kΩ — low enough to keep resistor thermal noise and op-amp bias-current error small, high enough to keep capacitors practical.
- Use C0G/NP0 or film capacitors; X7R ceramics drift with voltage and temperature and shift f₀.
- Pick an op-amp with GBW ≥ 100·Q·f₀ so finite gain-bandwidth doesn't distort the response.
- For a DC-accurate front-end, match the source resistance seen by both inputs to cancel bias-current offset.
To cascade for higher order, look up the Q and f₀ for each stage in a Butterworth/Chebyshev table — the stages do not share the same Q.
Sallen-Key vs. Multiple-Feedback and Passive Filters
Its main rival is the Multiple-Feedback (MFB) topology, also single-op-amp but inverting. MFB is generally preferred for high-Q band-pass stages and is less sensitive to the op-amp's own gain roll-off because the inverting configuration keeps the input node near a virtual ground. Sallen-Key wins when you need a non-inverting, unity-gain stage — exactly what DAC reconstruction and simple anti-aliasing want.
- vs. Passive RLC: Sallen-Key eliminates the inductor, which at audio frequencies would be physically large, lossy, and pick up magnetic hum. It also provides gain and zero output impedance.
- vs. MFB: Sallen-Key gives positive gain and easy unity-gain buffering; MFB gives better high-Q behavior and immunity to common-mode input capacitance.
- vs. State-Variable / Biquad: those use 3–4 op-amps but give simultaneous LP/HP/BP outputs and independently tunable f₀ and Q — worth it only when you need that flexibility.
For most gentle, low-Q filtering jobs, the single-op-amp Sallen-Key is the smallest, cheapest solution that works.
Failure Modes, Trade-offs, and Significance
The Sallen-Key's weaknesses appear at the extremes. Because it relies on positive feedback, high-Q designs become acutely component-sensitive: near Q = 1/(3−K) the sensitivity of Q to gain error blows up, so a 1% resistor drift can double the peaking or trigger ringing. Keep single-stage Q below about 10; for sharper responses, split the work across cascaded lower-Q sections.
- High-frequency feedthrough: at frequencies far above f₀, capacitor C₁ couples input straight to the output through the op-amp, so the real stopband floor is limited (often only −40 to −60 dB) rather than rolling off forever.
- Op-amp GBW limit: insufficient gain-bandwidth flattens the peak and shifts f₀; the effect worsens with Q and f₀.
- Capacitor quality: lossy or voltage-dependent caps degrade Q and add distortion.
Despite these limits, the topology's 70-year staying power is remarkable. Sallen and Key's insight — that a buffer's positive feedback can mimic an inductor's resonance — turned bulky LC filters into a two-capacitor cell, and it underpins the analog conditioning in almost every mixed-signal system today.
| Property | Sallen-Key (VCVS) | Multiple-Feedback (MFB) | Passive RLC |
|---|---|---|---|
| Op-amps per 2nd-order stage | 1 | 1 | 0 |
| Gain polarity | Non-inverting (positive) | Inverting (negative) | N/A (passive) |
| Best for Q range | Low–moderate (Q ≲ 10) | Moderate–high Q | Any, but bulky |
| Inductor needed | No | No | Yes (large at audio) |
| Sensitivity to op-amp gain roll-off | High at high f/Q | Lower (inverting) | None |
| Typical use | Low-pass / high-pass, unity gain | Band-pass, high-Q | RF, power filtering |
Frequently asked questions
Why does the Sallen-Key filter use positive feedback?
The capacitor from the op-amp output back to the mid-node feeds a portion of the output in phase with the input near the corner frequency. This positive feedback boosts the response there, creating a resonant peak that mimics what an inductor would do in a passive RLC filter. It is precisely this controlled positive feedback that lets you set Q without any inductor.
How do I set the cutoff frequency and Q independently?
Use the equal-resistor version (R₁ = R₂ = R). Then f₀ = 1/(2πR√(C₁C₂)) depends on the product C₁C₂, while Q = ½·√(C₁/C₂) depends only on the capacitor ratio. So you scale both capacitors together to move f₀ and change their ratio to change Q — the two are nicely decoupled.
What is the difference between Sallen-Key and Multiple-Feedback (MFB)?
Both are single-op-amp second-order filters, but Sallen-Key is non-inverting (VCVS) while MFB is inverting. MFB keeps its summing node at a virtual ground, making it less sensitive to op-amp gain roll-off and better for high-Q band-pass stages. Sallen-Key is preferred when you want positive/unity gain and simple low-pass or high-pass sections.
Why does my Sallen-Key filter oscillate or ring?
In the equal-capacitor design, Q = 1/(3 − K) where K is the non-inverting gain. As K approaches 3, Q approaches infinity and the loop becomes unstable, so even small resistor or gain errors produce excessive peaking or oscillation. Keep K comfortably below 3, or use the unity-gain equal-resistor design where Q is set by capacitor ratio instead.
What op-amp bandwidth do I need?
A common rule of thumb is gain-bandwidth product GBW ≥ 100·Q·f₀, and higher for high-Q stages. If the op-amp's bandwidth is too low, its finite open-loop gain near f₀ flattens the resonant peak and shifts the corner frequency downward. High-Q, high-frequency designs are the most demanding on GBW.
Can a single Sallen-Key stage give a steeper roll-off than 40 dB/decade?
No — one stage is second-order and rolls off at 40 dB/decade (12 dB/octave). To get steeper skirts you cascade stages: two give 4th-order (80 dB/decade), three give 6th-order, and so on. Each stage in a Butterworth or Chebyshev cascade uses its own specific f₀ and Q from a design table, not the same values.