Pulse-Width Modulation Techniques and Harmonics

Pulse-width modulation is the cornerstone of modern power electronics, enabling efficient control of everything from motor speed to the brightness of an LED. While its core principle is simple, the specific technique used to generate a PWM signal has profound implications for performance and efficiency. By analyzing the harmonic content of these signals, you can design effective filters that reduce electrical noise, minimize energy loss, and ensure your system meets stringent electromagnetic compatibility standards.

Fundamental Operation of PWM

At its heart, pulse-width modulation (PWM) is a method of encoding the amplitude of an analog signal into the duty cycle of a fixed-frequency rectangular wave. The duty cycle is defined as the ratio of the pulse "on" time to the total period of the signal, expressed as a percentage. For a desired output voltage $V_{out}$, if your DC supply voltage is $V_{DC}$, the required duty cycle $D$ is given by $D=V_{out}/V_{DC}$. This relationship forms the basis for voltage control.

A PWM signal is generated by comparing a low-frequency modulating signal (your desired output waveform) with a high-frequency carrier signal, typically a triangle wave. The points where these two signals intersect determine the switching instants. The output is high whenever the modulating signal is greater than the carrier signal. By varying the amplitude of the modulating sine wave, you vary the width of the output pulses, effectively creating an average voltage over each switching period that tracks the modulating wave. This average voltage is what drives a motor or creates an AC waveform from a DC source.

Key PWM Generation Techniques

The method used to sample the modulating signal for comparison with the carrier wave significantly impacts the resulting harmonic spectrum. The three primary techniques are natural sampling, regular sampling, and space vector modulation.

Natural Sampling is the classical analog method. Here, the actual, continuous modulating wave is compared directly with the triangular carrier. The switching instants occur precisely at the points of intersection. This method produces the theoretically optimal harmonic performance for sine-wave modulation, placing the dominant harmonics in sidebands around multiples of the carrier frequency $f_c$. However, it is computationally intensive to implement digitally, as it requires solving for the exact intersection points in real-time.

Regular Sampling was developed to make digital implementation feasible. Instead of using the continuous wave, the modulating signal's amplitude is sampled at a fixed point in each carrier period—typically at the peak or trough of the triangle wave. This sampled value is then held constant for the entire carrier period and compared with the carrier. This creates a staircase approximation of the modulating wave. While simpler to implement on a microcontroller, regular sampling introduces additional low-order harmonics not present in natural sampling, slightly increasing total harmonic distortion.

Space Vector PWM (SVPWM) is an advanced technique primarily used in three-phase voltage source inverters. Instead of treating each phase leg independently, SVPWM considers the inverter as a whole, representing the three desired output voltages as a single rotating vector in a complex plane. The technique selects specific combinations of the inverter's eight possible switching states to synthesize this reference vector. The key advantage is a 15% higher maximum fundamental output voltage compared to sinusoidal PWM and a more optimal distribution of harmonic energy, often pushing harmonics into a higher frequency range where they are easier to filter.

Harmonic Analysis and Spectral Content

Understanding the spectral makeup of a PWM waveform is not academic; it dictates the design of your output filter and predicts potential interference. Harmonic analysis involves breaking down the PWM output into its constituent frequency components: a desired fundamental frequency and a series of unwanted harmonics.

The harmonic spectrum differs markedly between techniques. For natural-sampled sine-triangle PWM, harmonics appear as sidebands centered at integer multiples of the carrier frequency ($f_c,2f_c,3f_c...$). There are no harmonics at the fundamental frequency or its low-order multiples. With regular sampling, low-order harmonics (e.g., 3rd, 5th) can appear alongside the carrier-based sidebands. SVPWM typically eliminates all harmonics below a certain order and clusters the significant harmonic energy around $f_c$ and $2f_c$.

To perform a basic analysis, you can use the double Fourier integral method, which expresses the PWM output as an infinite sum of Bessel functions. The amplitude of a harmonic at frequency $m{f}_c+n{f}_o$ (where $f_o$ is the fundamental output frequency, and $m$ and $n$ are integers) can be calculated. This analysis directly reveals the harmonic distortion you must mitigate. The goal of filter design is to allow the fundamental frequency to pass to the load (like a motor) while attenuating the high-frequency switching components to prevent overheating and electrical noise.

Filter Design and Carrier Frequency Trade-Offs

The results of your harmonic analysis define your output filter requirements. A simple first-order LC low-pass filter is common. Its cutoff frequency $f_{cutoff}=1/(2\pi \sqrt{LC})$ must be placed strategically: high enough to pass your fundamental frequency without attenuation, but low enough to significantly attenuate the lowest significant harmonic, usually the sidebands around the carrier frequency $f_c$. The filter's size, cost, and power loss are direct consequences of the PWM spectrum.

This leads to the critical engineering trade-off in carrier frequency selection. A higher $f_c$ spreads the harmonic energy into a higher frequency band. This allows for a smaller, less expensive filter because the cutoff frequency can be higher. However, every switch of a power transistor (IGBT, MOSFET) incurs a small energy loss. Doubling the switching frequency doubles these switching losses, reducing system efficiency and requiring more expensive cooling. Furthermore, if $f_c$ falls within the range of 20 Hz to 20 kHz, it can produce audible noise from magnetic components or piezoelectric effects, which is undesirable in many consumer and industrial applications.

Therefore, the selection process is a balancing act. For a high-power motor drive, switching frequency might be limited to a few kHz to minimize losses. This necessitates a larger, bulkier output filter. For a compact audio amplifier or LED driver, switching frequencies in the hundreds of kHz are common, allowing for tiny filter inductors and capacitors, but requiring transistors optimized for low switching loss.

Common Pitfalls

1. Ignoring Harmonic Currents in Loads: Even with a good output voltage filter, harmonic voltages can cause significant harmonic currents in inductive loads like motors. These currents do no useful work but increase $I^{2}R$ (copper) losses, causing overheating. Always consider the load's impedance at harmonic frequencies, not just at the fundamental.
2. Under-filtering due to Regular Sampling: If you design a filter based on natural sampling theory but implement regular sampling in your microcontroller, the unexpected low-order harmonics may not be attenuated by your filter, leading to distorted output and potential load issues. Always analyze or simulate the spectrum of your chosen implementation method.
3. Overlooking Dead-Time Effects: In a practical bridge circuit, a small dead time is inserted between turning off one transistor and turning on its complement to prevent shoot-through faults. This dead time introduces voltage errors and generates additional low-frequency harmonics, especially at high output frequencies and low modulation indices. Compensation algorithms are often required.
4. Choosing Carrier Frequency Based on Filter Alone: Selecting $f_c$ solely to minimize filter size can backfire. You must verify that your power switches can operate efficiently at that frequency and that the resulting switching losses and thermal design are acceptable for your application's cost and size constraints.

Summary

• Pulse-width modulation controls average output voltage by varying the duty cycle of a fixed-frequency switch, generated by comparing a modulating signal with a carrier wave.
• The choice of technique—natural sampling, regular sampling, or space vector PWM—directly determines the harmonic spectrum, affecting distortion and filter complexity.
• Harmonic analysis is essential to quantify unwanted frequency components, which informs the design of output filters to meet performance and regulatory standards.
• Selecting the carrier frequency involves a critical trade-off: higher frequencies allow smaller filters but increase switching losses and may cause audible noise, while lower frequencies have the opposite effect.
• Practical implementation must account for real-world effects like load harmonic currents, sampling method distortions, and dead-time, which are common sources of deviation from ideal theoretical performance.