Designing a Bandpass Op Amp Filter: MFB Topology and Bench Testing

A bandpass op amp filter isolates a specific frequency band by combining high-pass and low-pass characteristics in a single active stage, passing your target signal while aggressively attenuating noise above and below it. For most hobbyist, audio, and sensor benches, the most robust architecture is the Multiple Feedback (MFB) bandpass filter. Unlike simpler topologies, the MFB design offers high Q-factor stability and relies on a standard dual op-amp like the TL072 or NE5532, requiring just three resistors and two capacitors to set the center frequency, bandwidth, and gain.

Below, we will break down the exact pinouts, biasing requirements, and component math to build a 1kHz audio bandpass filter, followed by a real-world bench war story about what happens when you ignore parasitic capacitance.

The Anatomy of a Bandpass Op Amp Filter

Before wiring the feedback loop, you need to understand the silicon you are dropping onto the breadboard. A standard dual op-amp in an 8-pin DIP package follows a universal pinout standard. Looking at the chip with the notch facing left (or the dot in the top-left corner):

• Pin 1: Output A
• Pin 2: Inverting Input A (-)
• Pin 3: Non-Inverting Input A (+)
• Pin 4: V- (Negative Supply / Ground)
• Pin 5: Non-Inverting Input B (+)
• Pin 6: Inverting Input B (-)
• Pin 7: Output B
• Pin 8: V+ (Positive Supply)

The MFB topology gets its name because the signal is fed back to the inverting input through two distinct paths: one through a capacitor (creating the low-pass rolloff) and one through a resistor (creating the high-pass rolloff and setting the passband gain). This dual-feedback mechanism is what gives the MFB filter its excellent stability at higher Q-factors compared to the Sallen-Key topology.

Selecting and Biasing the Right Op-Amp

Not all op-amps are created equal. If you grab a random chip from your parts bin, your filter might oscillate, distort, or completely miss its target frequency. Here is how the safe defaults stack up for active filtering, along with their critical operating ratings.

How to Bias for the Job: If you are running a dual power supply (e.g., ±12V), biasing is trivial: tie Pin 4 to -12V, Pin 8 to +12V, and Pin 3 to system ground (0V). If you are constrained to a single supply (e.g., a 9V battery or 5V USB), you must create a virtual ground at exactly half your supply voltage (4.5V or 2.5V) using a buffered resistor divider. Pin 4 goes to 0V, Pin 8 to VCC, and Pin 3 to the virtual ground. The input and output signals must then be AC-coupled through series capacitors to block the DC offset.

The Safe Defaults: For general audio and sensor filtering, the TL072 is your go-to. Its JFET inputs mean virtually zero input bias current, preventing your high-value feedback resistors from generating massive DC offset errors. For low-impedance audio driving, use the NE5532. Avoid the LM358 for AC filtering; its notorious crossover distortion and low Gain-Bandwidth Product (GBWP) will ruin your passband.

Complete Application Circuit: 1kHz Audio Bandpass

Let us build a Multiple Feedback bandpass filter targeting a center frequency ($f_c$) of 1000 Hz, a Quality Factor ($Q$) of 5, and a passband gain ($A_0$) of 10 (20 dB). We will use a TL072 on a ±12V dual supply.

First, we select our capacitors. For audio frequencies, 10 nF (0.01 µF) C0G/NP0 ceramic capacitors are ideal because they exhibit zero microphonic noise and negligible voltage coefficient. Set $C_1 = C_2 = 10\text{ nF}$.

Using the standard MFB design equations (derived from Texas Instruments application notes), we calculate the resistors:

1. R3 (Main Feedback): $R_3 = \frac{2Q}{2 \pi f_c C} = \frac{10}{6.283 \times 10^{-5}} = 159,154\ \Omega$. Use a standard 160 kΩ 1% metal film resistor.
2. R1 (Input Resistor): $R_1 = \frac{Q}{2 \pi f_c C A_0} = \frac{5}{6.283 \times 10^{-4}} = 7,957\ \Omega$. Use a standard 8.2 kΩ 1% resistor.
3. R2 (Grounded Feedback): $R_2 = \frac{Q}{2 \pi f_c C (2Q^2 - A_0)} = \frac{5}{6.283 \times 10^{-5} \times 40} = 1,989\ \Omega$. Use a standard 2.0 kΩ 1% resistor.

Bench Walkthrough: When the Filter Oscillates

Theory is clean; the workbench is not. A few years ago, I was tasked with building a 10 kHz bandpass filter to isolate the row tones for a DTMF (Dual-Tone Multi-Frequency) decoder circuit. The specs called for $f_c = 10\text{ kHz}$ and a narrow bandwidth ($Q = 10$).

The Setup: I grabbed an LM358 because I had a tube of them on the shelf, plugged it into a standard solderless breadboard, and scaled the MFB resistor values down to hit 10 kHz. I fed it a 1Vpp 10 kHz sine wave from my function generator.

The Numbers: Based on the math, I expected a clean 2Vpp output sine wave at exactly 10 kHz.

The Outcome: The oscilloscope showed a massive 4Vpp signal, but it was not a 10 kHz sine wave. It was a jagged, rail-to-rail triangle wave oscillating at roughly 85 kHz. The filter had turned into an oscillator.

What Went Wrong: Two distinct bench realities collided to cause this failure. First, the LM358 has a GBWP of only 1 MHz. To maintain a $Q$ of 10 at 10 kHz, the op-amp needs a GBWP of at least $f_c \times Q \times A_0$, which in this high-gain configuration pushed past the LM358's internal phase margin limits. Second, and more insidiously, the solderless breadboard introduced roughly 5 pF of parasitic capacitance between the adjacent metal clips connecting the inverting input (Pin 2) and the output (Pin 1). At 10 kHz, 5 pF is negligible, but at the op-amp's internal high-frequency rolloff, that stray capacitance created an unintended phase-shift feedback path, triggering high-frequency oscillation.

The Fix: I swapped the LM358 for an NE5532 (10 MHz GBWP) and moved the circuit from the breadboard to a piece of copper-clad perfboard, using point-to-point wiring to keep the Pin 2 and Pin 1 traces physically separated. The oscillation vanished, yielding a pristine 10 kHz passband.

Troubleshooting and Multimeter Testing

When your bandpass filter outputs silence, a DC rail voltage, or pure noise, you need a systematic way to isolate the fault. Here is how to test the circuit using a standard digital multimeter (DMM), ranked from most likely to least likely cause.

1. Verify the Supply Rails: Set your DMM to DC Volts. Measure between Pin 8 and Pin 4. You should read your exact supply voltage (e.g., 24V for a ±12V split supply). If it reads 0V, check your power jumpers. If it reads half-voltage, you have a shorted decoupling capacitor.
2. Check the Virtual Ground (Single Supply Only): If running single-supply, measure Pin 3 relative to true ground. It must sit exactly at VCC/2. If it is off by more than 50mV, your biasing voltage divider is loaded down or a feedback resistor is pulling DC current into the input.
3. Measure the DC Output Offset: With no AC input signal applied, measure the DC voltage at Pin 1 (Output). In a dual-supply MFB filter, this should be < 10mV. If it sits at +11V or -11V (rail voltage), your op-amp is saturated. This usually means a capacitor in the feedback loop is installed backwards (if using electrolytic) or is open-circuit, destroying the DC feedback path and forcing the op-amp into open-loop gain.
4. Test for Dead Silicon (Out-of-Circuit): If the output is stuck at a rail and power is verified, power down and remove the chip. Set your DMM to the Diode Test mode. Measure from the ground pin (Pin 4) to the input and output pins. You should read a standard silicon junction drop (0.6V to 0.7V) in one direction and 'OL' in the reverse. If you read a dead short (0.00V) or open in both directions, the internal ESD protection diodes have blown, likely from a static discharge or an overvoltage event on the input pin.

Building a bandpass op amp filter is a masterclass in bridging theoretical math with physical reality. By selecting the right silicon (TL072 or NE5532), respecting component tolerances, and keeping your high-impedance traces away from parasitic capacitance, you will get a filter that performs exactly as the equations promise.