How to Design a Passive Low Pass Filter for PWM and Sensor Noise

A passive low pass filter is an electronic circuit built entirely from passive components—resistors, capacitors, or inductors—that allows low-frequency signals to pass through while attenuating frequencies above a specific cutoff point. In a real circuit or installation, it changes a jagged, noisy, or pulsed waveform into a smoother, cleaner signal by shunting high-frequency energy to ground or blocking it via reactance, all without requiring an external power supply. Beginners commonly confuse it with active filters (which require op-amps and external power rails to provide gain) or assume any resistor-capacitor (RC) pairing is a low-pass filter (an RC network actually becomes a high-pass filter if you take the output across the resistor instead of the capacitor).

The Core Math: Calculating Cutoff Frequency

The most common topology you will build on the bench is the first-order RC low pass filter. It consists of a series resistor and a shunt capacitor to ground. The cutoff frequency ($f_c$)—the exact point where the signal power drops by half, or -3dB—is determined by a straightforward formula:

$f_c = \frac{1}{2 \pi R C}$

Let's run a worked numeric example with standard bench components. Suppose you need to filter high-frequency switching noise from a 5V analog sensor line. You select a 1.5kΩ resistor and a 100nF (0.1µF) capacitor.

• $R = 1500 \, \Omega$
• $C = 0.0000001 \, F$
• $f_c = \frac{1}{2 \times 3.14159 \times 1500 \times 0.0000001}$
• $f_c = 1061 \text{ Hz}$

At exactly 1061 Hz, your output voltage will be 70.7% of your input voltage ($1/\sqrt{2}$). Frequencies below this pass with minimal loss; frequencies above it are rolled off at a rate of -20dB per decade. For a deeper look into the phasor math behind this attenuation, the Electronics Tutorials low-pass guide provides excellent vector diagrams.

Where You Meet This In Practice

You will rarely see a standalone 'filter' module in commercial gear; instead, passive low pass networks are embedded into almost every subsystem. Here is where they do the heavy lifting:

• PWM to Analog Conversion: Smoothing the 20kHz pulse-width modulated output from an ESP32 or Arduino into a steady DC voltage to drive 0-10V industrial LED dimmers or analog motor controllers.
• Audio Crossovers: In passive speaker cabinets, a series inductor and shunt capacitor form an LC low pass filter, directing only bass frequencies to the subwoofer while blocking treble that could damage the voice coil.
• Sensor Signal Conditioning: Stripping 60Hz/50Hz mains hum and high-frequency RF interference from sensitive thermocouple or strain gauge amplifier inputs before the signal hits an ADC.
• Snubber Networks: A variation of the filter (often an RC series network placed across a switch or relay contact) that absorbs high-frequency inductive voltage spikes when a motor is turned off.

Real-World Scenario: Smoothing a 20kHz PWM Signal

Theory is clean; the workbench is not. Here is a classic scenario that trips up both hobbyists and junior engineers.

The Setup: You are using an ESP32-WROOM-32 to output a 20kHz PWM signal. You want to convert this 3.3V logic-level PWM into a smooth 0-10V analog signal to control an industrial Mean Well LED driver. You use a level-shifter to get a 10V PWM square wave, then feed it into an RC filter to smooth it to DC.

The Numbers: You want the ripple to be virtually invisible, so you choose a very low cutoff frequency. You select R = 10kΩ and C = 1µF. Your calculated cutoff is a highly aggressive 15.9 Hz.

The Outcome: You hook up your oscilloscope directly to the capacitor. The scope shows a beautiful, flat DC line that rises and falls smoothly as you change the PWM duty cycle. You connect it to the LED driver.

What Went Wrong: The LED driver maxes out at 80% brightness, and you notice visible flickering at low dimming levels. Two distinct bench realities ruined your design:

1. Impedance Loading: The LED driver's 0-10V input isn't infinite impedance; it has an internal pull-up/pull-down network totaling about 100kΩ. Your 10kΩ series resistor and the driver's 100kΩ input impedance form a voltage divider. Your maximum 10V signal is now dropping to 9.09V ($10V \times \frac{100k}{110k}$), capping your brightness.
2. Capacitor DC Bias Derating: You used a standard 1µF X7R ceramic capacitor (MLCC). Murata's technical documentation on MLCCs explicitly warns that high-dielectric ceramics lose massive amounts of capacitance under DC voltage. At 10V, your '1µF' capacitor actually measures closer to 0.3µF. This shifts your cutoff frequency up to 53 Hz, allowing the 20kHz PWM ripple to bleed through and cause the flickering.

Component Selection: The Hidden Trap of DC Bias

Choosing the right capacitor is where most passive low pass filters fail in real-world installations. Here is a quick reference for selecting your shunt capacitor based on the application:

Step-by-Step: Building and Verifying on the Bench

Follow this sequence to ensure your passive low pass filter actually performs as calculated before soldering it into a permanent PCB or enclosure.

1. Calculate and Derate: Calculate your target $f_c$. If using MLCCs for DC voltages above 3.3V, check the manufacturer's DC bias curve and select a capacitor with 2x to 3x the required nominal capacitance.
2. Breadboard the Network: Place the resistor in series with your signal line, and the capacitor from the output node directly to a clean, dedicated ground point. Keep lead lengths short to avoid introducing parasitic inductance.
3. Inject a Known Signal: Use a function generator to feed a sine wave into the input. Start at 10 Hz and sweep up to 100 kHz. Set the generator's output impedance to 50Ω (or account for it in your math if your series resistor is very low).
4. Probe Correctly: Connect your oscilloscope probe to the output node. Crucial: Ensure your probe is set to 10x attenuation. A 1x probe adds ~100pF of capacitance and 1MΩ of resistance in parallel with your filter, which will skew your high-frequency roll-off measurements.
5. Verify the -3dB Point: Set the function generator to 2V peak-to-peak. Find the frequency where the oscilloscope reads exactly 1.414V peak-to-peak. This is your true, physical cutoff frequency. Compare it to your math.