Article Blog

Make Class-D amplifiers sound better

Due to their small physical size and high energy efficiency, Class D amplifiers have become fixtures in smartphones and battery-driven artificial intelligence (AI) speakers, where space and power are severely budgeted. They have supplanted Class A and Class AB amplifiers, both of which offer excellent linearity, high gain, and low signal distortion levels but are power-hungry.

Class D amplifiers are among the key technical constituents that have enabled the exponential growth in consumer demand for next-generation products. These non-linear switching amplifiers can, theoretically, reach 100% efficiency. Unlike Class A and Class AB, a current is only drawn through the transistor that is turned on. However, there is a flipside to Class D amplifiers’ significant benefits, which is the potentially-problematic noise generated due to their high-speed switching (Figure 1).


diagram showing how high-speed switching can generate noise in audio devicesFigure 1 High-speed switching can generate noise in audio devices like smart speakers. Source: TDK

Therefore, deploying effective noise countermeasures is a must for design engineers. At the same time, however, these measures must not undo the space-saving benefits nor affect signals and audio quality. It means that the properties of the noise-suppression filters inserted in speaker lines are significant. Speaker sound quality is also affected by the properties of inductors used in low pass filters (LPFs) at the Class D amplifiers’ output stages.

Additionally, measures against electrostatic discharge (ESD) are often found to be necessary in speaker lines. Here, ESD notch filters featuring multilayer chip varistor technology provide both protection from transient over-voltages associated with ESD as well as filter noise sources associated with wireless communications.

The ESD notch filters also play a critical role in suppressing the radiation noise. The multilayer chip varistors have a specific parasitic capacitance designed to work together with the filtering noise sources associated with cellular, Bluetooth, or Wi-Fi band interference (Figure 2).

graph of MAF and AVRF series filters insertion lossFigure 2 The MAF and AVRF series filters work in unison to lower the insertion loss. Source: TDK

The basic application approaches to achieve these functions are summarized below.

100 mW to 2 W class audio

In speaker lines for smartphones and other devices whose speaker output is relatively small—100 mW to 2W—Class D amplifiers without LPFs are commonly used. The sound distortion level is generally expressed in numerical form as total harmonic distortion plus noise (THD+N); the lower its value, the better the sound quality.

If generic chip beads are used in the speaker lines, the output causes the THD+N value to rise, degrading the sound quality. However, with noise suppression filters, the THD+N characteristics are equivalent to that of using no filter. Even if the output power is increased, there are no effects on the signals, and no sound distortion is generated. Using the frequency spectra for the output signal (1 kHz), the harmonics level is significantly higher when chip beads are used; this harmonic component is what is heard as distortion. In contrast, with noise suppression filters, high frequencies are left unaffected, so only the clean 1-kHz signal is heard.

Noise suppression filters’ effectiveness in radiation noise suppression compared to that of using no filter is equivalent to that occurring when the Class D amplifier is switched off. TDK’s MAF series filters, for instance, are multilayer chip components that use a new ferrite material to achieves low distortion while retaining its noise elimination characteristics.

2 W to 20 W class speaker lines

For devices with a speaker output in the 2 W to 20 W class, such as AI speakers, tablet PCs and other audio equipment, it will be necessary to provide inductors for LPFs externally to accommodate the large current. Since these inductors are inserted into the speaker lines, they must not affect the signals in those lines.

Metal inductors have cores made of metallic magnetic materials. They can accommodate large currents; however, the THD+N value increases as the output power becomes larger. The inductors that have a wire-wound shielded magnetic structure using ferrite can provides benefits including low DC resistance, Rdc, and the ability to accommodate large currents. So, when inserted into speaker lines, THD+N value changes only slightly.

As with the lower output class smartphone speakers, adding noise suppression filters into 2 W to 20 W class speaker lines will avoid any possible degradation of audio output induced by radiation noise.

Again, the THD+N characteristics of wire-wound noise suppression filters are equivalent to those using no filter, and the level of harmonics is also nearly the same. These results clearly show that replacing chip beads with a wire-wound noise suppression filters in a speaker line is highly effective at reducing distortion and improving sound quality (Figure 3).

graphs show that Noise suppression filters are effective at reducing distortionFigure 3 Noise suppression filters are effective at reducing distortion and improving sound quality in smart speaker designs. Source: TDK

A comparison of noise intensity versus frequency characteristics for speaker lines shows how noise intensity in the 100 MHz to 400 MHz band can be significantly reduced. It also shows that the high impedance of wire-wound noise suppression filters in this band makes it suitable as a countermeasure against noise for Class-D amplifiers, comfortably meeting the CISPR Class B standard with some headroom (broken red line).

photo of Matt Reynolds Matt Reynolds is director of marketing at TDK’s Piezo & Protection Devices Business Group.



photo of Rafael Pérez Rafael Pérez is product marketing manager at TDK’s Magnetics Business Group.



Related articles:

0 comments on “Make Class-D amplifiers sound better

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.