A bandpass filter circuit passes a specific range of frequencies while attenuating signals above and below that window. While passive LC filters work for high-frequency RF, active op-amp topologies are the undisputed standard for audio, sensor conditioning, and sub-MHz instrumentation. If you need high selectivity without bulky inductors, the Multiple Feedback (MFB) active topology is your workhorse. This guide breaks down the MFB node behavior, compares it to the Sallen-Key alternative, and walks through a real 1 kHz audio design with exact bench-tested component values.
The Multiple Feedback (MFB) Topology: Nodes and Behavior
The MFB topology is an inverting active filter that uses two capacitors and three resistors to create a second-order (12 dB/octave roll-off) bandpass response. To understand how it tunes, we need to map the circuit nodes:
- Vin: The AC signal input.
- Node A (Summing Junction): The inverting input of the op-amp. Due to the virtual ground principle, this node sits at 0V AC (assuming a split supply or a biased Vref).
- Node B (Feedback Junction): The intersection between the shunt capacitor (C1), the series feedback capacitor (C2), and the Q-setting resistor (R2).
- Vout: The op-amp output, which drives the load and feeds back through R3.
- GND / Vref: The non-inverting input reference point.
Tuning an MFB filter is an exercise in managing interacting variables. Changing one component rarely affects just one parameter. Here is the behavior matrix you need to memorize for bench troubleshooting:
| Component Changed | Effect on Center Freq (Fc) | Effect on Q (Bandwidth) | Effect on Midband Gain |
|---|---|---|---|
| Increase R1 (Input) | Decreases | Decreases (Wider BW) | Decreases |
| Increase R2 (Q-Setting) | Slight Decrease | Increases (Narrower BW) | Slight Increase |
| Increase R3 (Main Feedback) | Decreases | Increases | Increases |
| Increase C1 (Shunt to GND) | Decreases | Decreases | No Change |
| Increase C2 (Series Feedback) | Decreases | Increases | No Change |
Why MFB Over the Sallen-Key Alternative?
The Sallen-Key is the other dominant active filter topology, but it is fundamentally a non-inverting, unity-gain (or low-gain) architecture. When designing bandpass filter circuits for high-Q applications (Q > 5), the MFB topology wins decisively. According to Analog Devices' Linear Circuit Design Handbook, the Sallen-Key bandpass suffers from severe component sensitivity issues at high Q values, meaning a 1% resistor tolerance can shift your center frequency drastically. The MFB topology distributes the feedback paths, making it far more stable for narrow bandwidths.
| Criteria | Multiple Feedback (MFB) | Sallen-Key (SK) |
|---|---|---|
| Phase Shift | Inverting (180° shift) | Non-Inverting (0° shift) |
| High-Q Stability (Q > 5) | Excellent (Low sensitivity) | Poor (High component sensitivity) |
| Op-Amp GBWP Requirements | Moderate | High (Requires massive GBWP for high Q) |
| Gain Configuration | Inverting gain set by R3/R1 | Non-inverting, usually unity or low gain |
The Verdict: Choose MFB when you need high selectivity, inverting gain, and stable Q-factors. Choose Sallen-Key only when you strictly require a non-inverting output and a wide bandwidth (Q < 2).
Design Walkthrough: Building a 1 kHz Audio Bandpass
Let's design a practical bandpass filter circuit for an audio sub-system. We need to isolate a 1 kHz test tone from broadband noise.
Target Specs: Center Frequency (Fc) = 1 kHz, Quality Factor (Q) = 5 (Bandwidth = 200 Hz), Midband Gain = 10 (20 dB).
First, select your capacitors. For audio, avoid high-K ceramic capacitors (like X7R or Y5V) because their capacitance drops significantly under DC bias and they introduce microphonic noise. Use 5% tolerance NP0/C0G ceramics or metalized polypropylene film caps. We will set C1 = 10 nF and C2 = 2.2 nF.
Using the standard MFB design equations (or a verified tool like the Texas Instruments FilterPro desktop utility), we calculate the required resistances. To achieve exact performance on the bench, you must use the E96 1% resistor series rather than standard 5% E12 values. Here are the real component values:
- R1 (Input): 3.16 kΩ (Sets input impedance and gain)
- R2 (Q-Setting): 6.34 kΩ (Sets the damping factor)
- R3 (Feedback): 63.4 kΩ (Sets primary gain and Fc)
- C1 (Shunt): 10 nF (C0G)
- C2 (Series): 2.2 nF (C0G)
- Op-Amp: TL072 or OPA2134 (JFET input, low noise, sufficient 3 MHz GBWP for this Q and Fc).
With these exact values, the theoretical Fc lands at 1.002 kHz, and the Q sits at 4.98. This is the level of precision you should aim for before touching a breadboard.
Breadboard Testing and Failure Mode Analysis
Breadboarding high-Q analog circuits is notoriously frustrating due to stray capacitance between the breadboard's internal metal clips. At 1 kHz, this isn't a massive issue, but if you scale this to 100 kHz, the 2-5 pF of stray breadboard capacitance will completely detune your filter. Follow this exact sequence to test the circuit:
- Power and Decouple: Wire your +/- 15V (or +/- 9V) rails. Place a 100 nF MLCC and a 10 µF electrolytic capacitor directly across the op-amp's V+ and V- pins (Pins 8 and 4 on a standard DIP-8). Skipping this will cause high-frequency oscillation that looks like noise on your scope.
- Establish the Virtual Ground: If using a single supply, create a low-impedance Vref (Vcc/2) using a buffered voltage divider. Tie the non-inverting input (Pin 3) directly to this node.
- Minimize Lead Lengths: Keep the leads of C1, C2, and R2 as short as physically possible. Node A and Node B are high-impedance summing junctions; long leads act as antennas for 50/60 Hz mains hum.
- Inject and Sweep: Connect a function generator to Vin. Set it to a 1 Vpp sine wave. Connect your oscilloscope to Vout.
- Verify the Passband: Sweep the frequency from 100 Hz to 10 kHz. You should see the output peak at exactly 20 Vpp (Gain of 10) at 1 kHz. At 800 Hz and 1.2 kHz (the -3dB points for Q=5), the output should drop to roughly 14.1 Vpp.
What Breaks at the Extremes: Failure Mode Contrast
When a circuit fails on the bench, you need to know what a specific open or short fault looks like on the oscilloscope. Here is the failure-mode contrast for the MFB topology:
- R1 Opens: No AC signal reaches Node A. The op-amp acts as a unity-gain buffer for its own offset voltage. Symptom: Vout is a flat DC line (usually a few millivolts offset from GND).
- C1 Shorts to GND: Node A is effectively shorted to ground for AC signals. The virtual ground is destroyed, and the input signal is shunted directly to earth. Symptom: Vout drops to zero; massive loading on the function generator.
- R3 Opens (Main Feedback Loss): The op-amp loses its primary negative feedback path. It operates open-loop for DC and low frequencies. Symptom: Vout instantly slams into the positive or negative supply rail (clipping/saturation).
- C2 Shorts: DC feedback is lost, but AC feedback via R3 remains. Symptom: The AC bandpass response might look mostly intact, but the DC output offset will drift wildly until it hits the rail due to the op-amp's input bias current charging the stray capacitances.
Frequently Asked Questions
How do I cascade bandpass filter circuits for a steeper roll-off?
To achieve a 4th-order (24 dB/octave) or 6th-order roll-off, you cascade multiple 2nd-order MFB stages. However, you cannot simply build two identical 1 kHz Q=5 filters and put them in series. Doing so will cause the -3dB points to compound, narrowing your overall bandwidth and dropping your peak gain by 6 dB. Instead, you must use staggered-tuning or standard Butterworth/Chebyshev pole tables. For a 4th-order Butterworth bandpass, the first stage is tuned to a slightly lower Fc with a lower Q, and the second stage is tuned to a slightly higher Fc with a higher Q. The Electronics Tutorials filter guide provides the exact pole-Q pairing tables for cascaded designs.
Why is my active bandpass filter oscillating at high frequencies?
If you see a high-frequency (100 kHz+) sine wave superimposed on your output, your op-amp is oscillating. This is almost always caused by inadequate power supply decoupling or capacitive loading on the output. If your oscilloscope probe (which has ~15 pF of capacitance) is connected directly to Vout, it can introduce enough phase lag to break the stability margin of the op-amp. Fix this by adding a small series isolation resistor (typically 33 Ω to 100 Ω) between the op-amp's Vout pin and your load/probe point.
Can I use a single-supply op-amp for an AC bandpass filter?
Yes, but you must bias the circuit correctly. An op-amp cannot output a negative voltage if its V- pin is tied to 0V GND. To run an MFB bandpass on a single 5V or 9V supply, you must create a 'virtual ground' at Vcc/2 (e.g., 2.5V) using a precision voltage divider buffered by another op-amp channel. You then tie the non-inverting input (Node GND/Vref) to this 2.5V node instead of 0V. All AC coupling capacitors at the input and output must then be added to block this 2.5V DC offset from reaching your source and load. Ensure your op-amp has rail-to-rail input and output (RRIO) capabilities if you are operating on a tight 3.3V or 5V single supply.
