How do you design a multiple feedback band-pass filter?
Choose equal capacitors C, then R2 = Q ÷ (π f0 C), R1 = R2 ÷ (2|Am|) and R3 = |Am| R1 ÷ (2Q² − |Am|). For f0 = 1 kHz, Q = 10, a gain of −2 and C = 100 nF: R2 = 31.8 kΩ, R1 = 7.96 kΩ and R3 = 80.4 Ω. The gain magnitude cannot exceed 2Q², and the output is inverted.
Active band-pass filter formula (MFB topology)
R1 feeds the middle node, R3 goes from that node to ground, one capacitor goes from the node to the output and the other from the node to the inverting input, and R2 closes the loop from the output to the inverting input. With an ideal op amp and both capacitors equal to C:
f0 = (1 ÷ 2πC) √((R1 + R3) ÷ (R1 R2 R3)) Q = π f0 R2 C −Am = R2 ÷ 2R1 B = 1 ÷ (π R2 C)
- f0
- center frequency in Hz
- Q
- f0 divided by the −3 dB bandwidth
- Am
- gain at f0 (negative: the output is inverted)
- B
- −3 dB bandwidth in Hz
For a multiple feedback filter design, solve for the resistors in the order TI’s Op Amps for Everyone uses:
R2 = Q ÷ (π f0 C) R1 = R2 ÷ (2|Am|) R3 = |Am| R1 ÷ (2Q² − |Am|)
The −3 dB frequencies are f0 × (√(1 + 1/4Q²) ∓ 1/2Q). In the parts mode the calculator also accepts unequal capacitors: with C1 from the node to the output and C2 from the node to the inverting input, B = (C1 + C2) ÷ (2π R2 C1 C2) and |Am| = R2 C2 ÷ (R1 (C1 + C2)).
Active band-pass filter design: worked examples
Example 1: 1 kHz, Q = 10, gain −2 (TI Op Amps for Everyone Example 16-5)
With C = 100 nF: R2 = 10 ÷ (π × 1 kHz × 100 nF) = 31.8 kΩ, R1 = R2 ÷ 4 = 7.96 kΩ and R3 = 2 × 7.96 kΩ ÷ (200 − 2) = 80.4 Ω. Rounded to E96 (7.87 kΩ, 31.6 kΩ, 80.6 Ω), the filter centers at 1.002 kHz with Q = 9.95. The op amp works at a noise gain of 201 at 1 kHz, and about 1 MHz of gain-bandwidth holds f0 and Q within 1 %.
Example 2: one octave wide around 1 kHz at unity gain
One octave means Q = √2. With C = 10 nF: R2 = 45 kΩ, R1 = 22.5 kΩ and R3 = 7.5 kΩ. The −3 dB points land at 707.1 Hz and 1.414 kHz, which are f0 ÷ √2 and f0 × √2, exactly one octave apart.
Example 3: what happens without R3
R1 = 10 kΩ, R2 = 100 kΩ, C = 10 nF and R3 left open: f0 = 1 ÷ (2πC√(R1R2)) = 503.3 Hz, Q = 1.581 and the gain is 5.00, which is exactly 2Q². Without R3 you lose the freedom to set gain and Q separately.
Bandwidth, gain limit and op amp need for each Q
Band edges are shown as multiples of f0. Max gain is the 2Q² limit with equal capacitors. The noise gain at f0 is 1 + 2Q². The GBW column is the gain-bandwidth that holds f0 and Q within about 1 %, per kilohertz of f0 (scale it with your f0).
| Q | Bandwidth (% of f0) | −3 dB edges (× f0) | Max gain | Noise gain at f0 | GBW per 1 kHz of f0 |
|---|
| 0.707 | 141.4 % | 0.5176 to 1.9320 | 1.00 | 2.000 | 141 kHz |
| 1.000 | 100.0 % | 0.6180 to 1.6180 | 2.00 | 3.000 | 150 kHz |
| 1.414 | 70.7 % | 0.7071 to 1.4142 | 4.00 | 5.000 | 177 kHz |
| 2.00 | 50.0 % | 0.7808 to 1.2808 | 8.00 | 9.000 | 225 kHz |
| 4.32 | 23.2 % | 0.8909 to 1.1225 | 37.3 | 38.30 | 443 kHz |
| 5.00 | 20.0 % | 0.9050 to 1.1050 | 50.0 | 51.00 | 510 kHz |
| 10.00 | 10.0 % | 0.9512 to 1.0512 | 200 | 201.0 | 1 MHz |
| 20.00 | 5.0 % | 0.9753 to 1.0253 | 800 | 801.0 | 2 MHz |
| 50.00 | 2.00 % | 0.9900 to 1.0100 | 5000 | 5001 | 5 MHz |
Unity-gain, Q = 5 values for common center frequencies
E96 resistors, equal capacitors. The Actual f0 column is the center frequency with the rounded resistors; the last column is the op amp gain-bandwidth that holds f0 and Q within about 1 %.
| f0 | C | R1 | R2 | R3 | Actual f0 | Op amp GBW |
|---|
| 100 Hz | 100 nF | 78.7 kΩ | 158 kΩ | 1.62 kΩ | 100.5 Hz | 51 kHz |
| 1 kHz | 10 nF | 78.7 kΩ | 158 kΩ | 1.62 kΩ | 1.005 kHz | 510 kHz |
| 10 kHz | 1 nF | 78.7 kΩ | 158 kΩ | 1.62 kΩ | 10.05 kHz | 5.1 MHz |
| 40 kHz | 470 pF | 42.2 kΩ | 84.5 kΩ | 866 Ω | 39.99 kHz | 20.4 MHz |
Making an MFB band-pass work on the bench
The gain and bandwidth depend on R1, R2 and the capacitor ratio, and f0 on all five parts, so use C0G/NP0 or film capacitors and 1 % resistors, and tune f0 with R3 if it matters. The op amp is the usual limit: its noise gain at f0 is 1 + 2Q², which is 201 for a Q of 10, so its open-loop gain at f0 has to be well above that. An op amp that is too slow lowers f0 and raises Q; TI SBOA588 works this through for an MFB low-pass and shows how to pre-adjust the parts for a known gain-bandwidth, and the same idea applies here. Above Q ≈ 20 a single stage becomes touchy; two stagger-tuned stages or a state-variable filter are easier to hold. On a single supply, bias the non-inverting input at mid-supply and AC-couple the input.
Glossary
- Center frequency (f0)
- The frequency of peak gain, at the geometric mean of the two −3 dB points.
- Quality factor (Q)
- Center frequency divided by the −3 dB bandwidth. Higher Q means a narrower, more selective filter.
- Bandwidth (B)
- The distance in hertz between the lower and upper −3 dB frequencies.
- Multiple feedback (MFB)
- An inverting active filter with two feedback paths, one resistive and one capacitive. Also called the Rauch or Deliyannis-Friend topology.
- Noise gain
- The gain the op amp itself is closed around. It sets how much open-loop gain the op amp must have for the filter to stay accurate.
- Stagger tuning
- Cascading band-pass stages tuned slightly apart to get a flatter, wider passband with steeper skirts.
Frequently asked questions
What is the formula for the center frequency of an MFB band-pass filter?
f0 = (1 / 2πC) × √((R1 + R3) / (R1 R2 R3)) with both capacitors equal to C. R1 is the input resistor, R2 the feedback resistor from the output to the inverting input, and R3 goes from the middle node to ground.
How do I calculate R1, R2 and R3 for an MFB band-pass filter?
Pick C, then R2 = Q / (π f0 C), R1 = R2 / (2 |Am|) and R3 = |Am| R1 / (2Q² − |Am|). For 1 kHz, Q = 10, gain 2 and 100 nF this gives 31.8 kΩ, 7.96 kΩ and 80.4 Ω, the values in TI’s Op Amps for Everyone Example 16-5.
What is the maximum gain of an MFB band-pass filter?
With equal capacitors the center gain can be at most 2Q², where R3 becomes an open circuit. Above that, R3 would have to be negative. For a Q of 2 the limit is 8 (18 dB); if you need more gain, add a separate gain stage.
What does R3 do in a multiple feedback band-pass filter?
R3 only shifts the center frequency: bandwidth (1 / πR2C) and gain (R2 / 2R1) do not depend on it. That makes R3, or a trimmer in series with it, the one part to adjust when tuning f0 on the bench.
How do I get Q from the bandwidth I want?
Q = f0 / bandwidth, where the bandwidth is between the −3 dB points. A one-octave filter has Q = √2 = 1.414, and a one-third-octave filter has Q = 4.32. The −3 dB points sit at f0 × (√(1 + 1/4Q²) ∓ 1/2Q), so they are not symmetric in hertz, only on a log scale.
What op amp does an MFB band-pass filter need?
TI’s rule of thumb sets the open-loop gain 100 times above the Q of the section. For this circuit a gain-bandwidth product of 100 × (Q + 1/2Q) × f0 holds f0 and Q within about 1 %: 1 MHz for the 1 kHz, Q = 10 example. With a slower op amp, f0 falls and Q rises while the peak gain barely moves.
Is this MFB filter calculator only for band-pass filters?
Yes. This is a second-order active band-pass filter calculator only. For a second-order low-pass or high-pass, the Sallen-Key filter calculator covers the non-inverting alternative.
Is the output of an MFB band-pass filter inverted?
Yes. The signal enters the inverting input, so the output is 180° out of phase at the center frequency, moving toward −90° below it and −270° above it. If polarity matters, follow it with an inverting stage or use the inversion elsewhere in the chain.
Design the whole board, not one number
ProtoFlow is free AI PCB design software: describe a circuit and get a schematic with real parts, PCB layout with an autorouter, SPICE simulation and Gerber files. Native KiCad, Altium and EAGLE import and export. Free on macOS, Windows and Linux.