How to Design an RL Filter for PWM and Signal Conditioning

An RL filter is a passive electronic circuit that uses a resistor and an inductor to attenuate specific frequency ranges while allowing others to pass. Unlike a simple resistive voltage divider that drops voltage equally across all frequencies, the inductor introduces frequency-dependent reactance ($X_L = 2\pi f L$), fundamentally changing a circuit’s frequency response and phase angle while allowing DC to pass through the inductor with minimal power dissipation. Think of the inductor as a mechanical flywheel in a drive shaft: it resists sudden changes in rotational speed (high-frequency AC) but spins freely once up to a steady cruising speed (DC).

Designers commonly confuse RL filters with LC filters (which use a capacitor instead of a resistor and exhibit a much sharper resonance and Q-factor) or mistakenly assume inductors behave as ideal, purely reactive components without parasitic resistance or capacitance. In reality, component non-idealities dictate whether your filter performs as simulated or fails on the bench.

The Core Mechanics: What an RL Filter Actually Does

An RL filter relies on the voltage divider principle, but replaces one of the resistors with an inductor whose impedance scales linearly with frequency. This gives us two primary topologies:

• RL Low-Pass Filter: The inductor is in series with the signal path, and the resistor is in parallel (shunt) to ground. High frequencies see high series impedance and are blocked, while DC and low frequencies pass through the inductor and develop across the shunt resistor.
• RL High-Pass Filter: The resistor is in series, and the inductor is in shunt to ground. Low frequencies and DC are shorted to ground through the inductor's low DC resistance, while high frequencies are blocked from the shunt path and forced through the series resistor to the output.

The Math: Calculating Cutoff Frequency and Impedance

The cutoff frequency ($f_c$) is the point where the output power drops by half (-3 dB), and the inductive reactance exactly equals the resistance ($X_L = R$). The governing equation is remarkably straightforward:

$f_c = \frac{R}{2\pi L}$

Let us run a worked numeric example using standard, off-the-shelf component values. Suppose you are designing a low-pass filter to smooth a noisy sensor line. You select a shunt resistor $R = 100\ \Omega$ and a series inductor $L = 8.2\ \text{mH}$ (such as a standard Bourns 78FR82K radial leaded inductor).

1. Calculate Cutoff: $f_c = 100 / (2 \times \pi \times 0.0082) = 1941\ \text{Hz}$.
2. Calculate Attenuation at 10 kHz: At 10 kHz, the inductor's reactance is $X_L = 2\pi(10000)(0.0082) = 515\ \Omega$. The total impedance is $Z = \sqrt{100^2 + 515^2} = 524\ \Omega$. The output voltage ratio is $R / Z = 100 / 524 = 0.19$, meaning the 10 kHz noise is attenuated to roughly 19% of its original amplitude (about -14.4 dB).

For deeper theoretical modeling of passive networks, the All About Circuits textbook chapter on filters provides excellent derivations of the transfer functions and Bode plot asymptotes.

Where You Meet This in Practice

While RC (resistor-capacitor) filters dominate low-power signal conditioning, RL filters carve out specific, critical niches in modern electronics:

• Audio Crossovers: RL low-pass networks are routinely used to route bass frequencies to woofers. The inductor handles high speaker currents without the thermal losses or phase-shift anomalies that large electrolytic capacitors introduce.
• High-Impedance Sensor Conditioning: When filtering the output of a high-impedance piezoelectric sensor or a specialized photodiode transimpedance amplifier, adding a large capacitor (for an RC filter) can introduce a pole that causes the driving op-amp to oscillate. An RL filter avoids capacitive loading entirely.
• Power Supply Snubbers: In switching regulators, a series resistor and inductor (often combined with a capacitor to form an RLC snubber) are used to dampen high-frequency ringing on the switching node without shorting the fast $dv/dt$ edges to ground.

Real-World Scenario: Smoothing a 20 kHz PWM DAC

To understand how an RL filter behaves outside of a simulation environment, let us walk through a common bench failure involving load impedance.

The Setup: You are using an ESP32 to generate a 20 kHz PWM signal to act as a digital-to-analog converter (DAC), setting the reference voltage for a linear bench power supply. You need to strip the 20 kHz carrier and leave a clean DC envelope. To avoid the capacitive loading issues that made your previous RC design oscillate, you opt for an RL low-pass filter.

The Numbers: You target a cutoff frequency of roughly 2 kHz to aggressively kill the 20 kHz carrier. Using the formula, you select a series inductor $L = 8.2\ \text{mH}$ and a shunt resistor $R = 100\ \Omega$. Your calculated cutoff is $f_c = 1941\ \text{Hz}$. You breadboard the circuit and connect a standard bench oscilloscope (set to 50 $\Omega$ input impedance for high-frequency accuracy) to the output to verify the DC level.

The Outcome: The scope shows the 20 kHz ripple is completely gone, but the filter's step response—how fast the DC voltage tracks a change in the PWM duty cycle—is three times slower than your calculations predicted. This sluggishness introduces a dangerous phase lag in the power supply's feedback loop, causing the supply to oscillate when a load is applied.

What Went Wrong: You forgot that the cutoff frequency of an RL filter is entirely dependent on the shunt resistance. By connecting a 50 $\Omega$ oscilloscope probe in parallel with your 100 $\Omega$ shunt resistor, the effective shunt resistance dropped to $33.3\ \Omega$ ($100 \parallel 50$). The actual cutoff frequency plummeted to $647\ \text{Hz}$ ($33.3 / (2\pi \times 0.0082)$). The filter became overly aggressive, severely limiting the control loop bandwidth.

The Fix: You have two choices. First, buffer the filter output with a unity-gain op-amp (like an OPA340) which presents a giga-ohm input impedance, preserving your 100 $\Omega$ design. Second, redesign the filter assuming the 50 $\Omega$ load is your only shunt resistor, dropping the inductor to $2.5\ \text{mH}$ to maintain the ~3 kHz cutoff. For precise analog input filtering techniques, the Analog Devices guide on analog input filtering details how to properly terminate passive networks into ADCs and scopes.

Common Pitfalls and Component Non-Idealities

Simulators assume ideal components. The workbench does not. When specifying parts for an RL filter, check the datasheet for these three parasitic realities:

1. DC Resistance (DCR): A cheap 8.2 mH radial inductor might have a DCR of 4.5 $\Omega$. In a low-pass configuration, this DCR acts as an unintended series resistor, forming a voltage divider with your shunt resistor and attenuating your DC signal. Always subtract the DCR from your available voltage headroom.
2. Self-Resonant Frequency (SRF): Inductors have parasitic parallel capacitance between their wire windings. At the SRF, the inductor acts like a parallel LC tank; above the SRF, it acts like a capacitor. If your 8.2 mH inductor has an SRF of 5 MHz, it will filter a 20 kHz PWM signal beautifully. But if you try to use it to filter 100 MHz switching noise, it will bypass the noise directly to the output. Always verify the SRF is at least one decade above your target noise frequency.
3. Core Saturation: If your filter passes significant DC current (e.g., in a motor drive or power rail), the magnetic core can saturate. Once saturated, the inductance drops precipitously, effectively turning your filter into a simple piece of wire. Check the $I_{sat}$ (saturation current) rating, not just the $I_{rms}$ (thermal heating) rating. The Texas Instruments Analog Engineer's Pocket Reference provides excellent quick-look tables for inductor saturation derating.

FAQ: RL Filter Design Questions