A twin T notch filter is an RC circuit topology that combines a low-pass and a high-pass T-network in parallel to create a deep, narrow attenuation band at one specific target frequency. In a real circuit, it changes the frequency response by introducing a destructive signal path at exactly the center frequency ($f_c$), effectively deleting a single offending frequency—like 50 Hz or 60 Hz mains hum—while leaving the surrounding audio or data signal completely untouched.
The Anatomy and Math of the Twin T Topology
The topology gets its name because it looks like two letter "T" networks stacked on top of each other. One T-network acts as a low-pass filter, and the other acts as a high-pass filter. When you tie their inputs together and their outputs together, frequencies well below $f_c$ pass through the low-pass leg, and frequencies well above $f_c$ pass through the high-pass leg. But exactly at $f_c$, the phase shifts through both legs cause the signals to arrive 180 degrees out of phase and cancel each other out.
Component Ratios: The low-pass leg uses
R, R, and 2C. The high-pass leg uses C, C, and R/2.
The theoretical depth of the null in a purely passive twin-T is infinite, but in reality, parasitic capacitance and load impedance drag the Q-factor down to about 0.3. This makes the notch wide and shallow. To fix this, we use an op-amp to buffer the output and feed a portion of the signal back into the common ground node of the T-networks. This "bootstrapping" prevents the load from dragging down the null and pushes the Q-factor up to 10 or higher, creating a razor-sharp notch.
Worked Numeric Example: Designing a 60 Hz Null
Let’s design an active twin T notch filter targeting exactly 60.0 Hz to clean up a sensor signal. We need to select real-world components from the E24/E96 series.
- Choose R: We want low thermal noise but not so high that op-amp bias currents cause DC offset errors. Let’s pick R = 10 kΩ.
- Calculate C: Using $C = \frac{1}{2 \pi R f_c}$, we get $C = \frac{1}{2 \pi \times 10,000 \times 60} = 265.25 \text{ nF}$.
- Select Real Capacitors for C: 265 nF isn't a standard value. We can parallel a 220 nF and a 47 nF capacitor to get 267 nF. This shifts our theoretical $f_c$ to 59.6 Hz, which is close enough to tune later.
- Calculate 2C: The low-pass shunt needs $2 \times 267 \text{ nF} = \mathbf{534 \text{ nF}}$. We can parallel a 470 nF and a 68 nF capacitor.
- Calculate R/2: The high-pass shunt needs $10 \text{ k}\Omega / 2 = \mathbf{5 \text{ k}\Omega}$. We can use two 10 kΩ 1% resistors in parallel, or a single 4.99 kΩ 1% resistor.
| Position | Theoretical Value | Practical BOM (1% Tolerance) | Dielectric / Type |
|---|---|---|---|
| R1, R2 (Low-pass series) | 10 kΩ | 10.0 kΩ | Metal Film Resistor |
| C3 (Low-pass shunt) | 530 nF | 470 nF + 68 nF (parallel) | C0G / NP0 Ceramic |
| C1, C2 (High-pass series) | 265 nF | 220 nF + 47 nF (parallel) | C0G / NP0 Ceramic |
| R3 (High-pass shunt) | 5 kΩ | 4.99 kΩ | Metal Film Resistor |
Where You Meet This In Practice
You will rarely see a twin T notch filter in modern digital signal processing (DSP) where a simple IIR biquad filter can notch out a frequency in software with zero component tolerance issues. However, in the analog domain, this topology is irreplaceable in a few specific scenarios:
- Biomedical Instrumentation: ECG and EEG amplifiers must reject 50/60 Hz mains interference picked up by the human body before the signal hits the ADC, preventing the analog front-end from clipping.
- High-Gain Audio Preamps: Ribbon microphones and guitar pickups are essentially giant antennas for magnetic hum. An active twin-T in the effects loop or preamp stage removes the hum without sucking the "mud" out of the adjacent bass frequencies like a wide band-stop filter would.
- Geophones and Seismographs: Low-frequency vibration sensors often sit near heavy AC machinery. A sharp analog notch prevents the 60 Hz mechanical vibration from aliasing into the sampled data.
Bench Scenario: The 60 Hz Hum That Wouldn't Die
Theory is clean; the workbench is not. Here is a real-world walkthrough of what happens when component selection undermines your math.
The Setup: I was building a low-noise preamp for a high-impedance piezoelectric vibration sensor. The sensor was picking up a massive 60 Hz hum from a nearby variable frequency drive (VFD). I designed a passive twin-T notch filter and buffered it with a standard TL072 op-amp, aiming for a 60.0 Hz null.
The Numbers: I used standard 5% carbon film resistors and X7R multilayer ceramic capacitors (MLCCs) from my bench bins because they were the right values on the label. The theoretical $f_c$ was 60 Hz.
The Outcome: When I swept the circuit with my function generator and looked at the Bode plot on the oscilloscope, the notch was incredibly wide, and the deepest point of attenuation was sitting at 64 Hz. At exactly 60 Hz, the hum was only reduced by about 4 dB. The VFD hum was still clearly visible on the scope.
What Went Wrong: I fell victim to component parasitics and tolerances. First, X7R ceramic capacitors exhibit a severe voltage coefficient; their capacitance drops significantly when DC bias or AC signal is applied, shifting the high-pass leg's corner frequency. Second, stacking 5% resistors and 10% X7R capacitors meant my actual R and C values were likely 8% off from their nominal labels. Because the twin-T relies on the exact ratio of the two legs to achieve destructive interference, any mismatch ruins the null depth.
The Fix: I tore out the X7R caps and replaced them with C0G (NP0) film capacitors, which have zero voltage coefficient and 2% tolerance. I swapped the carbon resistors for 1% metal film. Finally, I replaced the fixed R/2 shunt resistor with a 10 kΩ multi-turn cermet trimmer pot. By injecting a 60.0 Hz sine wave and tweaking the trimmer while watching the scope, I dialed the null in to exactly -58 dB of attenuation. The hum vanished.
Common Confusions: Twin-T vs. Other Notch Filters
People frequently confuse the twin-T with other band-stop topologies. Here is how to tell them apart and when to choose which.
Twin-T vs. Bridged-T: A bridged-T network uses fewer components (two resistors, two capacitors) and is easier to tune because adjusting one resistor changes the Q without shifting the center frequency. However, the bridged-T cannot achieve the extreme null depths of a well-tuned active twin-T. Electronics Tutorials provides a great side-by-side schematic comparison of the two.
Twin-T vs. State-Variable Filter: A state-variable filter uses three or four op-amps to generate low-pass, high-pass, and band-pass outputs, which you then sum to create a notch. It is vastly superior for prototyping because you can tune the center frequency and the Q-factor independently using two separate potentiometers. The trade-off is board space, power consumption, and noise floor. If you need a fixed, set-and-forget notch in a production PCB, the twin-T wins. If you need a user-adjustable notch on a front panel, use a state-variable.
FAQ: Troubleshooting and Tuning
Why is my active twin-T oscillating instead of filtering?
If the op-amp gain in the feedback loop is set too high (Q > 20), component parasitics and op-amp phase margin limitations will push the circuit into oscillation. For audio applications, keep the Q between 5 and 10. You can calculate the required feedback resistor ratio using the Texas Instruments active filter design guide.
Can I use electrolytic capacitors for low-frequency (e.g., 10 Hz) twin-T filters?
No. Electrolytic capacitors have massive equivalent series resistance (ESR) and loose tolerances (often -20% / +80%). The ESR will destroy the Q-factor, and the tolerance will shift the null frequency unpredictably. For sub-20 Hz filters, use large-value film capacitors (like polypropylene) or switch to a switched-capacitor filter IC like the LTC1068.
Does the op-amp choice matter for a 60 Hz notch?
For 60 Hz, the gain-bandwidth product (GBWP) of the op-amp is rarely the bottleneck. However, the slew rate and input voltage noise matter. If you are filtering a low-level audio signal, a low-noise bipolar op-amp like the NE5532 or OPA1612 is ideal. If you are filtering a high-impedance sensor signal, use a JFET or CMOS input op-amp like the TL072 or OPA140 to prevent bias current errors from generating DC offsets across your 10 kΩ resistors.
Designing a twin T notch filter is an exercise in precision. The math gives you the map, but component parasitics dictate the terrain. Always specify C0G/NP0 dielectrics, use 1% metal film resistors, and include a trimmer potentiometer in the R/2 leg for final bench calibration. When you respect the tolerances, the twin-T remains one of the most elegant and effective analog interference killers in your toolkit.
