If you need to strip out DC offset, 50/60 Hz mains hum, and low-frequency drift while cleanly passing a 100 kHz+ signal, a 2nd-order active high frequency filter circuit based on the unity-gain Sallen-Key topology is your most reliable bench option. For a 100 kHz Butterworth cutoff, using a high-GBW op-amp like the OPA350 paired with 1.5 nF C0G capacitors and 1.5 kΩ / 750 Ω resistors provides a flat passband and a sharp -40 dB/decade roll-off below the knee.
Unlike passive RC filters that suffer from insertion loss and poor Q-factor, an active topology buffers the output and allows you to cascade stages without loading effects. But at high frequencies, parasitic capacitance and op-amp gain-bandwidth product (GBW) limits dictate your success. This guide breaks down the exact component selection, node behavior, and breadboard verification steps to get this filter working on the first try.
Why Sallen-Key Over Multiple Feedback (MFB)?
When designing an active high frequency filter circuit, the two dominant 2nd-order topologies are Sallen-Key (SK) and Multiple Feedback (MFB). While MFB is excellent for high-gain, low-frequency audio applications, Sallen-Key is vastly superior for high-frequency, unity-gain signal conditioning.
| Criterion | Sallen-Key (Non-Inverting) | Multiple Feedback (Inverting) |
|---|---|---|
| High-Freq Stability | Excellent. Non-inverting input has lower parasitic capacitance. | Poor. Inverting node parasitic capacitance degrades phase margin. |
| Gain Configuration | Best for unity gain (buffered). | Requires gain > 1 to maintain stability in some configs. |
| Op-Amp GBW Demand | Moderate. Open-loop gain requirements are relaxed at unity. | High. Noise gain increases, demanding massive GBW headroom. |
| Component Spread | Low. Can use matched C or matched R pairs. | High. Requires precise, wide-ratio component matching. |
For a 100 kHz high-pass filter, the Sallen-Key topology keeps the signal path clean and minimizes the risk of high-frequency oscillation caused by the op-amp's internal pole interactions. As detailed in Analog Devices' guide on Sallen-Key filters, the non-inverting architecture inherently isolates the feedback network from the input source impedance, making it ideal for RF and IF signal chain conditioning.
Topology Mapping and Component Spec Sheet
The unity-gain Sallen-Key high-pass filter routes the signal through two series capacitors, with resistors pulling the intermediate nodes to ground. The op-amp is wired as a voltage follower to provide a low-impedance output.
Node Labels:
- IN: Signal input source.
- N1: Junction between C1, C2, and R1.
- N2: Junction between C2, R2, and the Op-Amp Non-Inverting Input (+).
- OUT: Op-Amp output, fed directly back to the Inverting Input (-).
Below is the exact bill of materials for a 100 kHz Butterworth response (Q = 0.707). Component tolerances and dielectric types are critical at these frequencies.
| Ref Des | Value | Specification / Tolerance | Role in Topology |
|---|---|---|---|
| C1 | 1.5 nF | 1%, C0G/NP0 Ceramic, 50V | Series coupling, sets high-pass knee with R1. |
| C2 | 1.5 nF | 1%, C0G/NP0 Ceramic, 50V | Series coupling, passes HF to op-amp input. |
| R1 | 1.5 kΩ | 1%, Thin Film, 1/10W | Shunt to GND from N1, sets damping. |
| R2 | 750 Ω | 1%, Thin Film, 1/10W | Shunt to GND from N2, sets Q-factor. |
| U1 | OPA350 | GBW = 38 MHz, Slew Rate = 22 V/µs | Unity-gain buffer, provides low-Z output. |
| C3, C4 | 100 nF | X7R, 0603 package | Power rail decoupling (V+ and V-). |
Element Behavior and Extreme Failure Modes
Understanding how each component influences the transfer function is vital for debugging. The cutoff frequency ($f_c$) is governed by $1 / (2\pi\sqrt{R_1 R_2 C_1 C_2})$, while the Q-factor is determined by the ratio of the resistors and capacitors.
| Element | Parameter Change | Effect on $f_c$ and Q | Extreme Failure Mode (Short/Open) |
|---|---|---|---|
| C1 | Increases to 2.2 nF | $f_c$ drops to ~85 kHz; Q increases slightly (peaking). | Short: DC passes directly to N1. High-pass action is destroyed; circuit becomes a 1st-order low-pass. |
| R1 | Decreases to 1.0 kΩ | $f_c$ shifts up to ~122 kHz; Q drops (over-damped response). | Open: N1 floats. C1 and C2 charge via op-amp input bias current, latching OUT to the positive supply rail. |
| C2 | Decreases to 1.0 nF | $f_c$ shifts up to ~122 kHz; Q drops. | Short: N1 couples directly to N2. R1 and R2 parallel, turning the circuit into a 1st-order HPF with a -20 dB/dec roll-off. |
| R2 | Increases to 1.0 kΩ | $f_c$ drops to ~86 kHz; Q increases heavily (resonant peaking at the knee). | Open: N2 floats. The non-inverting input has no DC return path; output saturates to the positive rail immediately. |
The most common breadboard mistake is forgetting the DC return path. In a high-pass filter, capacitors block DC. If R2 is omitted or goes open, the op-amp's non-inverting input (N2) has no path to ground. The tiny input bias current of the OPA350 (typically 3 pA) will slowly charge the parasitic capacitance at N2 until the op-amp output rails out. Always verify R2's continuity before applying power.
Step-by-Step Breadboard Verification
Testing a 100 kHz high frequency filter circuit on a standard solderless breadboard introduces parasitic inductance and capacitance (typically 2-5 pF between adjacent rows). To minimize these effects, follow this exact build and test sequence.
- Prep the Power Rails: Place the OPA350 across the center ditch. Connect V+ (Pin 8) and V- (Pin 4) to your ±5V or ±12V rails. Place 100 nF decoupling capacitors (C3, C4) immediately adjacent to the op-amp power pins, routing them directly to the ground rail.
- Wire the Feedback: Connect the op-amp output (Pin 6) directly to the inverting input (Pin 2) using the shortest possible jumper wire. This establishes the unity-gain buffer.
- Build the Shunt Network: Insert R2 (750 Ω) from N2 (op-amp Pin 3) to ground. Insert R1 (1.5 kΩ) from N1 to ground. Ensure N1 and N2 are separated by at least three breadboard rows to minimize stray capacitance between the nodes.
- Place the Series Capacitors: Insert C2 (1.5 nF) between N1 and N2. Insert C1 (1.5 nF) from your IN rail to N1.
- Probe Setup: Connect your function generator to IN (set to 1 Vpp sine wave). Connect a 10x oscilloscope probe to OUT. Critical: Remove the standard alligator ground clip and use the probe's ground spring. At 100 kHz, a 3-inch ground clip acts as an antenna and will inject switching noise from the room into your measurement.
- Sweep and Verify: Sweep the function generator from 10 kHz to 1 MHz. At 100 kHz, you should measure -3 dB (0.707 Vpp). At 50 kHz, the signal should be attenuated by roughly -12 dB, confirming the 2nd-order -40 dB/decade slope.
Parasitics and the GBW Ceiling
While the math assumes ideal components, real-world physics imposes strict limits on active high frequency filter circuits. The primary bottleneck is the op-amp's Gain-Bandwidth Product (GBW).
The OPA350 has a GBW of 38 MHz. In a unity-gain configuration, the open-loop gain at 100 kHz is roughly 380 (51 dB). This is more than enough loop gain to force the feedback network to behave ideally and maintain a low output impedance. However, if you attempt to push this exact topology to a 2 MHz cutoff, the open-loop gain drops to 19. At that point, the op-amp can no longer correct for the network's phase errors, the Q-factor will artificially inflate, and the filter will exhibit severe peaking or outright oscillation.
If your application requires a cutoff above 500 kHz, you must upgrade the active element. According to Texas Instruments' SLOA024 application note on active filter design, a good rule of thumb is to select an op-amp with a GBW at least 100 times the desired cutoff frequency for a 2nd-order Butterworth response. For a 1 MHz filter, you would need an op-amp like the OPA847 (GBW = 3.9 GHz) or the ADA4817, and you must transition from a breadboard to a solid ground-plane PCB to manage the sub-picofarad stray capacitances that will otherwise detune your carefully calculated 1.5 nF capacitors.
By respecting the dielectric limitations of your capacitors, securing the DC return paths at N1 and N2, and matching your op-amp's GBW to your target knee frequency, you can reliably deploy this high frequency filter circuit to clean up noisy sensor data, strip DC offsets from RF mixers, or condition IF signals in software-defined radio front-ends.
