How to deal with EMI issues in Class D audio applications

Electromagnetic interference (EMI) refers to the interference of unintended electromagnetic radiation from the outside world. This interference can interrupt, hinder or degrade the performance of the circuit. In today's portable consumer electronic device design, space has become the first factor. Designers often need to remove the casing or shielding and suppress EMI and noise through more rigorous circuit isolation. Of course, the smaller space and more functions increase the density of the circuit board. In addition, wafer-level packaging and microcircuit design specifications need to be considered, so EMI issues are more worthy of attention.

EMI includes two aspects: radiation and electromagnetic tolerance. Emission refers to which devices generate radiated noise. Electromagnetic tolerance refers to which devices are affected by electromagnetic waves from other devices. In a later section, we will discuss more about electromagnetic tolerance. Because if electromagnetic radiation can be effectively controlled, it becomes relatively easy to deal with subsequent electromagnetic tolerance. Generally speaking, radiation is roughly divided into two categories: radiated radiation and conducted radiation. Radiated radiation comes from the circuit board, traces or wires, and propagates through the atmosphere in the form of electromagnetic waves to affect nearby receivers. It should be noted that "receiver" can be generally referred to as any circuit whose operation is affected by external electromagnetic energy interference. For example, PCB traces or IC leads. Conducted emissions are energy that escapes or is conducted through wires or cables. Conducted emissions can directly affect circuit performance or be converted into radiated emissions.



Figure 1: Physical relationship between wavelength and frequency

To understand the two types of emissions, we must have some understanding of antennas. Figure 1 shows the physical relationship between wavelength and frequency. An effective antenna must be one-quarter of the wavelength long. If it is in the atmosphere, its dielectric property is 1. Then in FR4 or glass epoxy circuit boards, its dielectric property will be reduced to 4.8. Therefore, once the signal reaches the dielectric gradient of FR4, its propagation trace will slow down. This will cause a "wavelength reduction" effect. For example, a 200MHz signal has a quarter wavelength in the atmosphere of 16.7cm. If it is on the inner layer of the circuit board, the wavelength becomes 16.7/4.8(1/2)=7.6cm. Even if

the length of the PCB trace is shorter than one-quarter of the wavelength, it can still be an effective antenna, which can enhance both radiation and electromagnetic tolerance. In addition to the inner layers, the traces on the surface can also show the wavelength reduction effect. Because one side of the dielectric is enough to change the entire dielectric property of the transmission.

Unintended antennas such as PCB traces can be said to be the culprit behind the radiated noise in digital systems. From the perspective of radiated emissions, we can find that the Class D audio amplifier can be regarded as a digital system in nature. A key principle in electromagnetics is electromagnetic reciprocity, because the flow of current can generate an electric field, and the change of electric flux can induce the flow of current. According to this principle, an antenna can be used to receive electromagnetic signals as well as to send electromagnetic signals. If the unintended antenna is stimulated by noise current and its length is close to one-quarter of the wavelength, radiated emissions will be generated at this time.



Figure 2: Common antenna designs

As shown in Figure 2, there are two common antenna designs: dipole antenna and whip antenna. An interesting fact is that the whip antenna itself is a half dipole antenna. After horizontal grounding and induction, the whip antenna can become another half dipole antenna. As we all know, the function of an antenna is to send and receive signals by radiating electrical energy. However, as shown in Figure 3, unintentional antennas in PCBs can include: long traces; vias; leads and pins of components; connectors and sockets of unpopulated boards.



Figure 3: Unintentional Antennas in PCBs
Some unterminated surface traces or buried traces on PCBs can become unintentional whip antennas. Trace segments at different potentials can become dipole antennas due to poor layout. At the same time, the conductive layer of the PCB can act as the other leg of the dipole antenna, and the board itself can be coupled into the electric field.

Class D Audio Amplifiers

Due to their high efficiency, Class D audio amplifiers have quickly gained widespread application in consumer electronic devices. Class D audio amplifiers modulate a high-frequency square wave with an input analog signal. The frequency of the square wave can be fixed, variable, or even random pulses. A low-pass filter is used to filter out the high-frequency content in the signal and restore the original audio signal. In topologies without filters, the inductance of the speaker itself is incorporated as part of the filter. Pulse width modulation (PWM) is a common Class D topology technique that uses a fixed-frequency waveform and produces a moving average signal after the low-pass filter by varying the duty cycle (Figure 4).



Figure 4: PWM is a common Class D topology technology

The benefits of using a switching topology are obvious, such as high efficiency, low power consumption, and easy heat dissipation. However, increasing efficiency does not come without a price. In order to improve efficiency, a sharp and rapidly changing square wave is required. However, since the spectral energy is highly concentrated on the edges of the square wave, this will cause problems in digital systems to reappear. At the same time, some overexcitation may occur, causing the waveform to exceed the highest and lowest voltages for a short time. Overexcitation produces additional high-frequency content in the output spectrum and has a negative impact on EMI and audio performance.

Fighting EMI

To eliminate EMI, it is necessary to integrate the power of electrical engineers, circuit board layout engineers, and manufacturing engineers during circuit design to jointly develop an optimal PCB design. To deal with EMI problems, you should usually pay attention to the following when designing the PCB:

1. Place decoupling capacitors between the power supply and ground where voltage fluctuations occur. 1. If capacitors are placed haphazardly, EMI problems will be exacerbated;

2. The power plane should be kept a certain distance from the edge of the circuit board;

3. Avoid cutting traces in the ground or power plane, otherwise it may cause unintentional pinholes;

4. Provide adequate terminations for all high-frequency clock lines;

5. Provide appropriate filtering for circuit board connectors;

6. Good PCB design can avoid loop antennas. Loop antennas can conduct both forward and reverse currents on a defined path.

In addition, radiation can be stopped by suppressing the current of the antenna.

For audio designers, the following two points must be considered:

1. Minimize the length of the trace from the audio amplifier to the speaker. Because once the trace reaches a quarter of the wavelength, significant radiation will occur, and the trace or wire will become an antenna.

2. For filterless Class D systems, the trace or cable connecting the amplifier output and the speaker will be the largest source of RF radiation.



Figure 5: Placing ferrite beads near the amplifier is an effective way to suppress EMI

Placing ferrite beads near the amplifier and in series with the speaker can be a very effective way to suppress EMI. To further understand the suppression method of ferrite beads, we can separate ferrite beads into frequency-dependent resistors and inductors, as shown in Figure 5. To suppress EMI, the ferrite bead needs to act as a resistor, but because Rdc=0, there is no DC voltage drop there. This method works well for applications below 1 MHz. In addition, as shown in the figure, a binary voltage divider needs to be considered. Both Z1 and Z2 are frequency-dependent, and in order to achieve the required low-pass filter function, the following relationship must hold: Z2>>Z1 at the required frequency and Z1>>Z2 at the noise frequency.

Ferrites are usually used as series elements, and capacitors are used as shunt elements. The capacitors here can be physical capacitors or lumped capacitors. The transfer function shows that Z1 and Z2 will increase and decrease with frequency (1/jωC), respectively. The system will have a certain degree of damping to significantly reduce the resonance effect.



Figure 5: Placing ferrite beads near the amplifier is an effective way to suppress EMI

As can be seen from the figure, the most basic difficulty when dealing with the inherent periodic square wave of Class D is the concentrated energy that occurs at the resonant interval. To design a "quiet" low-EMI Class D amplifier, one method is to jitter the frequency back and forth, or spread the switching spectrum, reducing the energy at all points in the spectrum. Compared with traditional Class D amplifiers, spread spectrum modulation schemes have several important advantages: in addition to maintaining high efficiency and low THD+N, more importantly, radiated noise and EMI are greatly reduced, as shown in Figure 6.



Figure 6: In addition to maintaining high efficiency and low THD+N, spread spectrum modulation schemes can also significantly reduce radiated noise and EMI
The LM48511 is a spread spectrum modulated Class D audio amplifier that integrates a built-in boost regulator that can increase the voltage to 7V, thereby enhancing the amplifier's output power and audio sound pressure level. In addition, the boost regulator allows the amplifier to maintain a fixed output level even when the battery is degraded.

The LM48511 features a logic-selectable spread-spectrum modulator that reduces EMI and eliminates the need for output filters or chokes. As shown in Figure 7, the spread-spectrum modulator feeds a standard H-bridge that drives a bridge-tied load speaker. In spread-spectrum mode, the switching frequency varies randomly by 10% around 330kHz, reducing radiated EMI emissions from the speaker and associated wires and traces. In this mode, a fixed-frequency Class D amplifier would exhibit spectral energy several times higher than the switching frequency, and the LM48511's spread-spectrum architecture spreads this energy over a larger bandwidth, reducing peak noise power in the circuit.



Figure 7: Spread-spectrum modulator feeds a standard H-bridge that drives a bridge-tied load speaker

Electromagnetic interference is a system-level issue. For audio designers, EMI must be considered when planning the design and selecting components and materials.