What is the notch filter equation for a twin-T network?
The notch frequency is f0 = 1 / (2πRC). One T is two resistors R with a capacitor 2C from their junction, and the other is two capacitors C with a resistor R/2 from their junction.
Work out the notch frequency, the R and C values and the Q of passive and active versions with this twin-T notch filter calculator. Enter the parts you actually have to see the real notch frequency and how deep it is once they stop matching.
Notch frequency f0
59.61 Hz
Figures assume an ideal source and no load. A source resistance or a load does not move the null, but it changes the notch width and the passband gain, so drive the network from a low impedance and buffer its output.
With Q = 0.25 the notch is wide: it is 3 dB down from 14.1 Hz to 253 Hz. Bootstrapping narrows it without moving f0.
Use f0 = 1 ÷ (2πRC), with two series resistors R, a shunt capacitor 2C, two series capacitors C and a shunt resistor R/2. For 60 Hz with C = 100 nF, R = 26.5 kΩ and R/2 = 13.3 kΩ. The passive network has Q = 0.25, so its −3 dB width runs from 14.2 Hz to 254 Hz; feeding a fraction k of the output back to the common node raises Q to 1 ÷ (4(1 − k)).
A twin-T is a low-pass T (R, 2C, R) and a high-pass T (C, R/2, C) in parallel. At f0 the two paths have equal size and opposite phase, so they cancel at the output.
With the common node grounded, an ideal source and no load, the balanced network has this notch filter transfer function, with s normalized to 2πf0:
Driving the common node from a buffered fraction k of the output (National LB-5) keeps the same zeros and moves the poles:
TI’s active twin-T instead returns R3 to the output of a non-inverting stage of gain G = 1 + Rf/Rg (TI labels those two resistors R2 and R1) and leaves C3 at ground. G is also the passband gain, so gain and Q are tied together:
The −3 dB points of any of these are at f0 × (√(1 + 1/4Q²) ∓ 1/2Q).
R = 1 ÷ (2π × 60 × 100 nF) = 26.5 kΩ, so R/2 = 13.3 kΩ and 2C = 200 nF. The nearest E96 resistors are 26.7 kΩ and 13.3 kΩ, which put the notch at 59.66 Hz with 66.6 dB of rejection if the capacitors are exact. Q is 0.25, so the −3 dB band runs from 14.2 Hz to 254 Hz.
R = 1 ÷ (2π × 50 × 100 nF) = 31.8 kΩ. For Q = 5, k = 1 − 1 ÷ (4 × 5) = 0.95: a divider that feeds 95 % of the output through a follower into the common node. The −3 dB band narrows to 45.2 Hz to 55.2 Hz.
With C = 10 nF, R = 15.9 kΩ and R/2 = 7.96 kΩ. G = 2 − 1 ÷ (2Q) = 1.75, so Rf = 0.75 × Rg, for example 7.5 kΩ with Rg = 10 kΩ. The passband gain is also 1.75 (4.9 dB).
LB-5’s Figure 1 uses R1 = R2 = 10 MΩ, R3 = 5 MΩ, C1 = C2 = 270 pF and C3 = 540 pF and labels the result 60 Hz. The formula gives 58.95 Hz, within 2 % of that label. With C3 just 1 % high, the depth of the passive network falls from a theoretical null to 58.1 dB.
Band edges are multiples of f0. k is the bootstrap fraction, Q = 1 / (4(1 − k)). G is the gain of TI’s active twin-T, which starts at Q = 0.5 because G cannot be below 1. Ideal Q of the passive network is 0.25.
| Q | −3 dB band (× f0) | Width (% of f0) | k (bootstrapped) | G (TI active) |
|---|---|---|---|---|
| 0.25 | 0.236 to 4.236 | 400 % | 0 (passive) | n/a |
| 0.5 | 0.414 to 2.414 | 200 % | 0.50 | 1.00 |
| 1 | 0.618 to 1.618 | 100 % | 0.75 | 1.50 |
| 2.5 | 0.820 to 1.220 | 40 % | 0.90 | 1.80 |
| 5 | 0.905 to 1.105 | 20 % | 0.95 | 1.90 |
| 10 | 0.951 to 1.051 | 10 % | 0.975 | 1.95 |
| 25 | 0.980 to 1.020 | 4.0 % | 0.990 | 1.980 |
Passive network, E96 resistors, capacitors assumed exact. The last two columns are the notch with those rounded resistors, computed from the circuit equations.
| Target | C | 2C | R exact | R (E96) | R/2 (E96) | Actual notch | Depth |
|---|---|---|---|---|---|---|---|
| 50 Hz | 100 nF | 200 nF | 31.8 kΩ | 31.6 kΩ | 15.8 kΩ | 50.37 Hz | deeper than 120 dB |
| 60 Hz | 100 nF | 200 nF | 26.5 kΩ | 26.7 kΩ | 13.3 kΩ | 59.66 Hz | 66.6 dB |
| 100 Hz | 47 nF | 94 nF | 33.9 kΩ | 34 kΩ | 16.9 kΩ | 99.74 Hz | 62.7 dB |
| 120 Hz | 47 nF | 94 nF | 28.2 kΩ | 28 kΩ | 14 kΩ | 120.9 Hz | deeper than 120 dB |
| 400 Hz | 10 nF | 20 nF | 39.8 kΩ | 40.2 kΩ | 20 kΩ | 396.4 Hz | 64.1 dB |
| 1 kHz | 10 nF | 20 nF | 15.9 kΩ | 15.8 kΩ | 7.87 kΩ | 1.008 kHz | 66.5 dB |
| 10 kHz | 1 nF | 2 nF | 15.9 kΩ | 15.8 kΩ | 7.87 kΩ | 10.08 kHz | 66.5 dB |
Depth of a balanced passive twin-T when one part is high by the amount in the first column and every other part is exact. Several parts off at once can add or cancel.
| Error | R3 (R/2) | C3 (2C) | R1 or R2 | C1 or C2 |
|---|---|---|---|---|
| 0.1 % | 78.1 dB | 78.1 dB | 84.1 dB | 84.1 dB |
| 0.5 % | 64.1 dB | 64.1 dB | 70.2 dB | 70.1 dB |
| 1.0 % | 58.1 dB | 58.1 dB | 64.2 dB | 64.1 dB |
| 2.0 % | 52.1 dB | 52.2 dB | 58.3 dB | 58.0 dB |
| 5.0 % | 44.1 dB | 44.5 dB | 50.6 dB | 50.0 dB |
Match matters more than absolute value: a common error across all six parts moves the notch but leaves it deep, while a mismatch between them fills it in. LB-5 suggests silver mica or polycarbonate capacitors and precision resistors to limit drift with temperature; the usual modern choices are C0G/NP0 ceramic or film. Make 2C from two of the same C in parallel and R/2 from two of the same R in parallel, so each ratio tracks. Drive the network from a low-impedance source and buffer its output: source and load resistance do not move the null, but they change the notch width and the passband gain. At high Q the notch width depends on the buffer’s gain error as well as on k. With k = 1 exactly the network has no notch at all: the output just follows the input, and the Q of a follower-bootstrapped twin-T comes from the follower’s gain being slightly under one. The model here uses ideal amplifiers, so simulate a very narrow notch with the real op amp model.
The notch frequency is f0 = 1 / (2πRC). One T is two resistors R with a capacitor 2C from their junction, and the other is two capacitors C with a resistor R/2 from their junction.
A notch filter is a band-stop filter with a narrow stop range: it removes a narrow range of frequencies and passes everything else. You enter one notch frequency here, not two band edges, so this page works as a band-stop filter calculator only for a twin-T centered on that frequency. A passive twin-T is wide, with Q = 0.25, unless feedback narrows it.
With C = 100 nF and 2C = 200 nF, R = 1 / (2π × 60 × 100 nF) = 26.5 kΩ and R/2 = 13.3 kΩ. The nearest 1 % values are 26.7 kΩ and 13.3 kΩ, which move the notch to 59.66 Hz.
Q = 0.25, so the notch is very wide: a 60 Hz notch is 3 dB down from 14.2 Hz to 254 Hz. That is why a passive twin-T also takes a slice out of the signal on both sides of the frequency you wanted to remove.
Connect the junction of R/2 and 2C to a buffered fraction k of the output instead of ground. Q becomes 1 / (4(1 − k)), so k = 0.95 gives Q = 5. The notch frequency stays where it was, because the feedback changes only the denominator of the transfer function.
With perfect parts the null is infinitely deep. In practice part matching sets it: one capacitor 1 % off limits the notch to 58.1 dB, and National's LB-5 suggests 0.1 % resistors and 1 % capacitors to minimize the trimming needed for a 60 dB notch.
Not the null: with balanced parts the transfer admittance of the network is zero at f0, whatever is connected to it. The rest of the response does change. Source and load resistance widen or narrow the notch and set the passband gain (at DC, RL / (RL + Rs + 2R)), so a low-impedance driver and a buffered output keep the response where the formulas put it.
National LB-5 reports that raising C3 by 1 to 10 % raises Q but degrades the notch depth, 10 to 20 % makes the circuit a tuned amplifier with voltage gain, and more than that makes it oscillate. Enter your parts in the Check real parts tab of this twin-T calculator to see where a design sits.
See all free electronics tools →
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.
Free to start. No card required.