Reducing Electromagnetic Interference in Class D Amplifiers

In recent years, there has been a rapid growth in the number of portable devices that contain powered speakers—including cell phones, MP3 players, GPS systems, laptops and notebooks, tablets, game consoles, toys, and more. In these applications, the type of audio amplifier that is often chosen to drive the speaker is called a Class D (or switching) amplifier because it dissipates less heat (important in compact products) and is more efficient (extending battery life) than traditional Class AB amplifier designs. One possible disadvantage of the Class D amplifier switching topology is that it tends to emit electromagnetic radiation that may interfere with other nearby electronic devices. This interference can be mitigated to some extent through external passive filtering methods, but this increases the cost, footprint, and complexity of the final product. This article will explore some of the internal circuit design methods used to mitigate EMI problems.

Edge Rate Control

To amplify audio signals, the output (or various outputs, in different configurations) of a Class D amplifier alternately switches between two power rails (usually positive and ground) at a frequency that is 10 times or more the highest audio frequency to be amplified (which may be 300kHz or more). The switching signal is modulated so that the audio signal is recovered by a simple low-pass filter, sometimes included in the speaker itself. This switching transition is typically very fast—perhaps 2ns or less—and therefore contains significant high-frequency energy. This can cause EMI radiation from the interconnecting conductors, especially if there is no low-pass filter in the signal path and the length of the conductors between the amplifier and the speaker is significant (perhaps more than 1cm).

One method used to mitigate EMI radiation is to reduce the slew rate of the amplifier output. Figure 1 shows an example in the time domain, where the upper trace has a rise and fall time of 2ns and the lower trace has a rise and fall time of 20ns. The reduction in slew rate

(here by a factor of 10) has a significant effect on the amount of radiated energy produced by a Class D amplifier. Figure 2 shows the spectrum of the two waveforms, when the Class D output is silent (no audio, duty cycle = 50%) and the switching frequency is 333kHz. It can be seen that throughout most of the spectrum between 30MHz and 1GHz, the high-frequency (HF) content is reduced by about 20dB. In systems that include FM broadcast reception electronics (88MHz ~ 108MHz), cell phones, or wireless Internet circuits (700MHz ~ 2.7GHz), this can significantly reduce EMI, thereby reducing the risk of potentially impacting system performance.

Spread Spectrum Clocking

While the edge rate control (ERC) discussed above is an effective method for attenuating EMI generated in the frequency range above 30MHz (also subject to FCC regulations), the fundamental carrier frequency of the Class D amplifier switching output and its associated odd harmonics (square waves) that fall below 30MHz are not well addressed by this technique. Figure 3 shows the energy generated by a conventional, unmodified Class D amplifier output in this frequency band.

To reduce the height of the fundamental and overtone peaks in the Class D output spectrum, a small amount of frequency modulation can be added to the amplifier's clock circuit—perhaps with a modulation index of around ±5% that will not affect the quality of the amplified audio signal. There are many choices for the characteristics of the modulation signal source, but a common approach is to use a pseudo-random pattern with a repetition rate (the full-pattern repetition rate) that exceeds the highest expected audio signal frequency (usually 20kHz) by a suitable margin. This prevents the generation of tones that might fall into the audio band.

Figure 4 shows the same Class D output as Figure 3, but with ±5% modulation, implemented by a pseudo-random sequence at a 40kHz pattern repetition rate.

Figure 5 shows a color overlay of Figures 3 and 4, showing more clearly the difference made by the spread spectrum clock. It can be seen that the odd harmonics of the reference clock frequency are suppressed by nearly 10dB across the entire spectrum.

Single-Sided Modulation

An additional approach to reducing EMI can be employed by modifying the modulation scheme to allow a single-sided differential or bridge Class D output pair to stop switching when the audio baseband signal amplitude becomes large enough (Figure 6). This essentially allows the output to reverse direction until it switches so that full modulation can be applied, keeping the output signal at its highest peak for the remaining interval. With this scheme, only one output is switching for a large percentage of the time (depending on the audio source material), so EMI (during that time) is reduced by half. This has the added benefit of reducing fixed switching losses due to charging and discharging of power device gate and other parasitic capacitances. It also reduces the time that the output is in the ERC transition, which, as mentioned above, has a small efficiency penalty. The downside of this technique is that the overall forward gain of the amplifier is slightly reduced, and likewise, total harmonic distortion (THD) and noise are slightly increased. The Class D output spectrum with and without single-sided modulation is shown in Figure 7.

Conclusion

Class D amplifiers are commonly used in portable devices because they are more power efficient than traditional Class AB amplifiers. The main disadvantage of Class D technology is its inherent EMI, which can adversely affect surrounding electronic devices. Some effective IC design techniques have emerged that can greatly alleviate EMI problems without the burden of additional external components.