The Core Concept: What an Active Notch Filter Actually Does

An active notch filter is an op-amp-based circuit that sharply attenuates a specific, narrow band of frequencies while passing all others with gain or unity buffer, without the insertion loss of passive LC networks. In a real circuit, it changes your signal chain by surgically removing a single interference frequency—like 60Hz mains hum or a 25kHz switching spike—without loading down the source impedance or rolling off your high-frequency audio or data.

Beginners commonly confuse an active notch filter with a passive RLC notch (which suffers from insertion loss and requires bulky, EMI-prone inductors) or a low-pass filter (which acts as a sledgehammer, rolling off all high frequencies rather than acting as a surgical scalpel to excise a single band). By incorporating an operational amplifier, the active variant provides high input impedance, low output impedance, and can even provide passband gain to make up for upstream signal attenuation.

The Golden Rule of Notch Filters: A notch filter only works if the noise is stationary. If your interference frequency drifts (like a variable-frequency drive sweeping from 40Hz to 80Hz), a fixed notch filter will fail. For drifting noise, you need an adaptive filter or a wider band-stop design.

Where You Meet This in Practice

You will rarely see active notch filters in high-power electrical installations, but they are ubiquitous in low-voltage signal conditioning. Here is where they earn their keep:

  • Biomedical Sensors (ECG/EEG): The human body acts as an antenna for 50/60Hz mains fields. An active notch filter strips this out while preserving the critical high-frequency edges of the QRS complex in an ECG.
  • Audio Preamps and DI Boxes: Ground loops introduce a persistent 60Hz (or 120Hz) hum. A high-Q active notch removes the fundamental hum without dulling the high-end cymbal crashes like a low-pass filter would.
  • Industrial Sensor Lines: When running 4-20mA or 0-10V analog sensor signals near Variable Frequency Drives (VFDs), specific switching harmonics can alias into your ADC. A notch filter targets the exact VFD switching frequency.

Worked Numeric Example: Killing 60Hz Mains Hum

Let’s design a classic Active Twin-T Notch Filter to eliminate 60Hz hum from an audio line. The Twin-T topology uses two RC networks (one low-pass, one high-pass) summed at the op-amp input, with a feedback loop to control the Q-factor (sharpness of the notch).

Target Center Frequency ($f_0$): 60 Hz
Target Q-Factor: 10 (Narrow enough to preserve 50Hz and 70Hz audio, sharp enough to kill 60Hz)

The center frequency formula for a Twin-T network is:

f_0 = 1 / (2 * π * R * C)

Step 1: Pick a standard capacitor value.
We need high stability, so we select a 270nF (0.27µF) C0G/NP0 film capacitor. Avoid X7R ceramics, as their piezoelectric effect introduces microphonic noise and voltage-dependent capacitance shifts.

Step 2: Calculate R.
R = 1 / (2 * π * 60 * 270e-9)
R = 9,824 Ω

Step 3: Select standard resistor values.
9.824kΩ is not a standard value. We use a 9.76kΩ (1% tolerance, E96 series) resistor. If 1% resistors aren't available, use a 10kΩ resistor in series with a 1kΩ multi-turn trimmer potentiometer to dial in the exact null depth.

Step 4: Set the Q-Factor.
In an active Twin-T, the Q is controlled by the feedback ratio from the op-amp output back to the Twin-T ground node. For a Q of 10, the feedback attenuation factor must be precisely tuned. According to Analog Devices Tutorial MT-223, achieving a deep null requires the resistors in the feedback divider to be matched to at least 0.1% tolerance, or the notch depth will degrade from -60dB to a disappointing -20dB.

Decision Tree: Choosing Your Topology and Op-Amp

Not all active notch filters are built the same. Use this decision matrix to select the right topology for your bench or PCB layout.

If your requirement is... Choose this Topology Pros & Cons
Simple 50/60Hz hum rejection, minimal component count Twin-T (Dual-T) Pro: Only 1 op-amp needed.
Con: Tuning $f_0$ and Q interact with each other; requires precision matched resistors for deep nulls.
Precision audio, independent tuning of frequency and Q Fliege Notch Pro: $f_0$ and Q are completely independent; easy to tune with trimmers.
Con: Requires 2 op-amps and more matched passive components.
Variable/swept notch (e.g., parametric EQ or lab instrument) State-Variable (Biquad) Pro: Highly tunable via a single potentiometer; provides simultaneous LP, HP, and BP outputs.
Con: Requires 3 or 4 op-amps; complex PCB routing.
The Default Recommendation: For 90% of audio, sensor, and bench hum-rejection tasks where you need a reliable, deep null without pulling your hair out tuning interacting variables, build a Fliege topology using a Texas Instruments OPA2134 dual op-amp. The Fliege allows you to tune the center frequency without destroying your Q-factor, and the OPA2134’s low noise (8 nV/√Hz) and high slew rate prevent intermodulation distortion in audio bands.

Component Gotchas: Dielectrics and Bandwidth

The math only gets you 80% of the way there. The physical components dictate whether your filter actually works on the bench. Keep these hardware realities in mind:

1. Capacitor Dielectric Matters

Never use Y5V or X7R ceramic capacitors in the frequency-determining network of an active filter. These Class II dielectrics exhibit severe capacitance drop-off with applied DC bias and temperature. Furthermore, they are piezoelectric—mechanical vibrations on the PCB will modulate the capacitance, injecting microphonic noise directly into your signal path. Always specify C0G (NP0) ceramics or polypropylene/film capacitors for the timing network.

2. Op-Amp Gain-Bandwidth Product (GBWP)

An op-amp's open-loop gain drops as frequency increases. For a notch filter to achieve a deep null, the op-amp must have massive excess gain at the notch frequency to correct the feedback loop. A good rule of thumb from All About Circuits filter design guidelines is that the op-amp's GBWP must be at least 100 times the product of your center frequency and Q-factor.

Example: If $f_0 = 10 ext{kHz}$ and $Q = 20$, your required GBWP is $10,000 imes 20 imes 100 = 20 ext{MHz}$. A standard LM358 (GBWP ~1MHz) will completely fail here; you need an NE5532 (GBWP ~10MHz, borderline) or an OPA2134 (GBWP ~8MHz, but check the specific high-frequency variant like the OPA1612 for >40MHz GBWP).

3. Resistor Thermal Noise

When designing for high-Q (Q > 20), the resistors in the feedback network can become quite large (in the megaohm range). High-value resistors generate significant Johnson-Nyquist thermal noise. Keep timing resistors under 100kΩ wherever possible by scaling up your capacitor values.

Frequently Asked Questions

Can I just use a digital (DSP) notch filter instead of an analog one?
Yes, if your signal is already in the digital domain. However, if you are dealing with a massive 60Hz hum on an analog sensor line, it might clip your ADC's input stage before the DSP ever sees it. An analog active notch filter conditions the signal before the ADC, preserving your dynamic range and preventing clipping.

Why is my Twin-T notch filter only attenuating by -15dB instead of -60dB?
The depth of the null in a Twin-T filter is entirely dependent on component matching. If your '2R' resistor is actually 1.95R and your 'C/2' capacitor is actually 0.48C due to 5% tolerances, the low-pass and high-pass paths won't cancel perfectly at the summing node. Switch to 0.1% precision resistors or add a multi-turn trimmer to balance the network.

Does an active notch filter introduce phase shift?
Yes. While the amplitude response shows a sharp dip, the phase response shifts rapidly around the notch frequency (typically sweeping through 180 degrees). In audio applications, this phase shift around 60Hz is generally inaudible. In control loops or biomedical timing applications, this phase distortion can be critical and must be simulated in SPICE before committing to silicon.