An electronic filter is a frequency-selective circuit that passes signals within a designated bandwidth while attenuating frequencies outside that range. In a real circuit, a filter changes the spectral composition of a waveform—stripping high-frequency switching noise from a power rail, isolating a specific audio band for a tweeter, or converting a digital PWM square wave into a smooth analog DC voltage. Makers commonly confuse the filter type (Low-Pass, High-Pass) with the filter topology (Butterworth, Chebyshev, Bessel), or they mistakenly assume a passive RC filter will maintain its calculated cutoff frequency regardless of the load impedance attached to it.
The Four Core Filter Responses
Before selecting components, you must identify which frequency band you need to keep and which you need to reject. Every analog filter falls into one of four fundamental response categories:
- Low-Pass Filter (LPF): Passes frequencies below the cutoff ($f_c$) and attenuates higher frequencies. Used for smoothing, anti-aliasing, and bass isolation.
- High-Pass Filter (HPF): Passes frequencies above $f_c$ and blocks lower frequencies. Used for AC coupling (removing DC offset) and tweeter crossovers.
- Band-Pass Filter (BPF): Passes a specific "window" of frequencies between a lower and upper cutoff. Used in RF receivers and audio midrange extraction.
- Band-Stop (Notch) Filter: Attenuates a narrow, specific frequency band while passing everything else. Used primarily to eliminate 50/60 Hz mains hum from sensitive audio or sensor lines.
Worked Example: Smoothing a 1 kHz PWM Signal to DC
Let's look at a common bench scenario: you are using an ESP32 to generate a 1 kHz PWM signal, and you need to smooth it into a clean DC voltage to drive an analog gauge. A simple first-order passive RC low-pass filter is the standard approach.
The formula for the cutoff frequency ($f_c$) of an RC filter is:
f_c = 1 / (2 * π * R * C)
To effectively smooth a 1 kHz square wave into DC, your cutoff frequency needs to be at least one decade (10x) lower than the fundamental frequency. Let's target an $f_c$ of 15.9 Hz.
- Choose R: We select a 10 kΩ resistor. This is high enough to avoid drawing excessive current from the ESP32's GPIO pin, but low enough to minimize thermal noise.
- Calculate C: Rearranging the formula:
C = 1 / (2 * π * 10,000 * 15.9). This yields exactly 1 µF. - Verify Attenuation: At the 1 kHz fundamental frequency, the ratio of $f / f_c$ is $1000 / 15.9 = 62.89$. The attenuation factor is roughly $1 / 62.89 = 0.0159$, which translates to -36 dB of attenuation. The 1 kHz ripple will be virtually invisible on a standard multimeter or analog gauge.
Where You Meet Filters in Practice
You will encounter filter networks across almost every domain of electrical and electronics work:
- Mains EMI Suppression: Heavy inductors and Y-capacitors form low-pass filters at the AC inlet of switching power supplies to prevent high-frequency switching noise from radiating back into the grid.
- ADC Anti-Aliasing: A strict low-pass filter must be placed immediately before any Analog-to-Digital Converter to block frequencies above the Nyquist limit, preventing high-frequency noise from "folding back" into your digital readings as false low-frequency data.
- Audio Crossovers: Passive LC networks inside speaker cabinets route low frequencies to the woofer (LPF) and high frequencies to the tweeter (HPF).
- RF and IF Stages: Ceramic and SAW band-pass filters isolate specific channels in WiFi receivers and software-defined radios.
Active vs. Passive: The Loading Effect Trap
The most common mistake hobbyists make with the types of filter circuits built from passive components (resistors, capacitors, inductors) is ignoring the loading effect.
In our 1 kHz PWM example above, we calculated $f_c$ assuming the filter output was connected to an infinite impedance (an open circuit). But what if you connect that 10 kΩ / 1 µF filter directly to a 1 kΩ load, like a small motor or a low-impedance analog meter?
The 1 kΩ load sits in parallel with the 10 kΩ resistor from the perspective of the AC signal, effectively altering the circuit's impedance and shifting the cutoff frequency unpredictably. Furthermore, the load will form a voltage divider with the 10 kΩ resistor, dropping your maximum DC output voltage from 3.3V down to roughly 0.3V.
The Solution: Use an active filter. By placing an operational amplifier (configured as a unity-gain buffer or in a Sallen-Key topology) between the RC network and the load, the op-amp's high input impedance prevents loading, while its low output impedance easily drives the next stage. For deep dives on active topologies, the Texas Instruments SLYT164 application note is an essential bench reference.
Decision Tree: Picking the Right Filter for Your Build
Stop guessing component values. Use this decision matrix to select the exact filter architecture and part number for your specific application.
| Scenario / Application | Filter Type Needed | Architecture / Topology | Concrete Part Pick / Implementation |
|---|---|---|---|
| Mains AC EMI suppression on a 10A SMPS | Low-Pass (Common Mode & Differential) | Multi-stage LC Passive | Schaffner FN2090-10-06 (Chassis mount, integrated bleed resistors) |
| High-frequency RF noise on a 5V DC rail | Low-Pass (Feedthrough) | Pi-network SMD Passive | Murata NFE31PT222Z1E9L (SMD EMI filter, 2200pF, handles up to 10A) |
| Precision anti-aliasing for a 16-bit SAR ADC | Low-Pass (Strict cutoff) | 2nd-Order Active Sallen-Key (Butterworth) | TI OPA211 Op-Amp + 1% C0G Caps (Calculate via AAC Filter Designer) |
| Audio subwoofer crossover (variable 40-160Hz) | Low-Pass (Adjustable) | Active State-Variable | TI UAF42 (Universal Active Filter IC, requires only external tuning resistors) |
FAQ: Filter Topologies vs. Filter Types
Q: What is the difference between a filter type and a filter topology?
A "type" defines the frequency band (LPF, HPF). A "topology" or "alignment" defines the mathematical shape of the roll-off curve within that band.
Q: When should I use a Butterworth topology?
Use Butterworth when you need a maximally flat passband. It has no ripple in the passband, making it ideal for audio and DC-accurate sensor signals, though its roll-off is more gradual than Chebyshev.
Q: When should I use a Chebyshev topology?
Use Chebyshev when you need the sharpest possible transition from passband to stopband and can tolerate some amplitude ripple in the passband. It is common in RF channel selection.
Q: When should I use a Bessel topology?
Use Bessel when preserving the shape of a time-domain pulse is critical. It offers linear phase response, meaning all frequencies are delayed equally, preventing "ringing" on square waves.
When designing mixed-signal PCBs or conditioning sensor data on the bench, default to a 2nd-order active Butterworth Sallen-Key low-pass filter using a precision op-amp like the OPA211 and C0G capacitors. This specific combination guarantees a flat passband, eliminates the passive loading effect, and provides a predictable -40 dB/decade roll-off that will reliably protect your ADCs from aliasing without requiring complex tuning.
