Design and application of new green energy-efficient Class D audio amplifier

0 Introduction

In the multimedia era, traditional Class A, Class B, and Class AB linear analog audio amplifiers can no longer meet the new development trends of green energy saving, high efficiency, and small size of electronic audio-visual LCD/PDP/OLED/LCOS/PDA due to their low efficiency and high energy consumption. Class D amplifiers, which are nonlinear audio amplifiers, will be incorporated into more and more new product designs due to their advantages of energy saving, high efficiency, high output power, low temperature rise effect, and small footprint. Class D amplifiers are divided into half-bridge asymmetric type and full-bridge symmetric type in terms of architecture. The full-bridge type has an output power of up to 4 times that of the half-bridge type and is more efficient. In terms of signal adaptation, they are divided into analog type and I2S full-digital type. Because the full-digital type is still in the development stage and has high cost, the analog type will be in the mainstream of application in the next few years due to its cost advantage. This paper focuses on the design elements of full-bridge analog Class D amplifiers, realizes a circuit design based on NXP's new green energy-efficient analog Class D amplifier TFA9810T, and summarizes and analyzes several aspects of green energy saving and high efficiency, high output power, low temperature rise effect, PCB layout, and EMI suppression.

1 Principle and characteristics of Class D power amplifier

1.1 Class D amplifier system structure

The Class D amplifier consists of integral phase shift, PWM modulation module, G gate driver, switch MOSFET circuit, logic auxiliary, output filter, negative feedback, protection circuit, etc. In terms of the process, the analog input signal is first modulated into a PWM square wave signal. The modulated PWM signal drives the power output stage through the drive circuit, and then the high-frequency carrier signal is filtered out through a low-pass filter. The original signal is restored and drives the speaker to sound, as shown in Figure 1.

1.2 Modulation level (PWM-Modulation)

The modulation stage is the A/D conversion, which samples the input analog audio signal to form a digital PWM signal in the form of high and low levels. In Figure 2, the comparator's in-phase input is connected to the audio signal source, and the reverse input is connected to the triangle wave signal generated by the internal clock of the power amplifier. When the signal level at the audio input is higher than the triangle wave signal, the comparator outputs a high level VH, otherwise, it outputs a low level VL, and converts the input sine wave signal into a PWM wave whose width varies with the sine wave amplitude. This is one of the cores of the Class D power amplifier. It must require good triangle wave linearity, stable oscillation frequency, high comparator accuracy, fast speed, and steep rising and falling edges of the generated PWM square wave. For in-depth modulation measures, please refer to the literature [2].

1.3 Full-bridge output stage

The output stage is a switching amplifier with an output swing of VCC. The circuit structure is shown in Figure 3. MOSFET is equivalent to an ideal switch. When it is turned off, the conduction current is zero and there is no power consumption. When it is turned on, the voltage at both ends is still close to zero. Although there is current, the power consumption is still close to zero. During the entire working cycle, MOSFET basically has no power consumption, so theoretically the conversion efficiency of the Class D amplifier can be close to 100%, but considering the auxiliary circuit power consumption and MOSFET conduction loss, the overall conversion efficiency can generally reach about 90%. Because the conversion efficiency is very high, the chip itself consumes little heat energy and the temperature rise is also very small, so poor heat dissipation can be completely ignored, so it is called a green energy-efficient Class D amplifier.

For the full bridge, to further reduce the conduction loss, the on-resistance RON of the MOSFET drain-source should be as small as possible. Select MOSFET with low switching frequency and small gate-source capacitance to enhance the driving capability of the pre-driver.

1.4 LPF low-pass filter stage

LPF filter can eliminate electromagnetic interference and switching signals in PWM signal, improve efficiency, reduce harmonic distortion, directly affect amplifier bandwidth and THD, and must set appropriate cutoff frequency and filter roll-off factor to ensure audio quality. For audio-visual products, 20 Hz ~ 20 kHz is audible sound; below 20 Hz is infrasound; above 20 kHz is ultrasound. In applications, the cutoff frequency is generally set to 30 kHz. The lower the frequency, the narrower the signal bandwidth, but too low will damage the signal quality, and too high will cause noise to mix in. Commonly used LPF filters are generally Butterworth filter, Chebyshev filter, and Cauer filter. Butterworth filter has good maximum flat amplitude characteristics in the passband BW and is easy to implement. Therefore, audio-visual products mostly use LC second-order Butterworth filter with small equivalent internal resistance and large output power as shown in Figure 4.

1.5 Negative Feedback

Negative feedback is an LPF circuit that feeds back the detected output audio components to the input stage, compares them with the input signal, and compensates, corrects, and noise-shapes the output signal to improve the linearity of the power amplifier and reduce the ripple in the power supply (power supply rejection ratio, PSRR). Negative feedback can reduce the noise generated in the passband due to pulse width modulation, output stage, and power supply voltage changes, so that the low-frequency components in the output PWM can always be consistent with the input signal to obtain a good THD, making the sound richer and more accurate.

1.6 Power efficiency analysis

When THD is less than 7%, the efficiency of Class D amplifier can reach more than 85%, which is much higher than the commonly used maximum theoretical efficiency of 78.5% of linear amplifiers. The fundamental reason is that the output stage MOSFET is completely working in the switching state. Theoretically, the efficiency of Class D amplifier is:

Assuming that the on-resistance of the Class D power amplifier MOSFET is RON, all other passive resistors are RP, the filter resistor is RF, and the load resistor is RL, the efficiency without considering the switching loss is:

Where: fOSC is the oscillator frequency; tON and tOFF are the MOSFET on and off frequencies respectively. At this time, the efficiency is:

From the above formula, we know that the larger the ratio of load RL in class D power amplifier to other resistors, the higher the efficiency; MOSFET as a freewheeling switch, the power consumed is almost equal to the sum of I2RON loss and quiescent current on MOSFET on-resistance, which is almost negligible compared to the power output to the load. Therefore, its efficiency is much higher than that of linear power amplifier, as shown in Figure 5. It is very suitable for the current green energy-saving requirements and is suitable for large-scale use in digital audio-visual products such as flat panels.

2 Key points to note about Class D amplifiers

The following points should be noted in Class D design applications:

2.1 Deadtime (Dead Zone Correction)

The full-bridge MOSFETs are turned on in pairs in turn. Ideally, one pair is turned on and the other pair is turned off. However, in reality, there is a process for the power tubes to be turned on and off. During the transition process, there must be a moment, as shown in Figure 3, when IN1/IN3 has not been completely turned off, IN2/IN4 has already started to turn on. Because all MOSFETs are connected across the power supply, there may be a large voltage and current applied to the four MOSFETs at the same time in an extreme time, resulting in large power consumption, reduced overall efficiency, and increased device temperature rise, burning out the MOSFETs and reducing reliability. In order to avoid the two pairs of MOSFETs being in the on state at the same time, causing a potentially threatening large short-circuit current, it should be ensured that there is a very short stagnation dead time (Dead-time) between the conduction of one pair of MOSFETs and the cut-off of the other pair of MOSFETs. This time is controlled by the Logic controller to effectively ensure that after one set of MOSFETs is turned off, the other set of MOSFETs is turned on in time, reducing MOSFET losses and improving amplifier efficiency.

However, if Deadtime is not set correctly, the following problems may occur:

(1) Glitches will be generated in the output signal, causing electromagnetic interference, that is, IN1/IN3 are both turned off during the dead time. The completely out-of-control output voltage will be affected by the body diode current in Figure 6(a) (the formation of the body diode current, see the EMI section below), and glitches will appear in the output waveform.

(2) If the deadtime is too large, the energy contained in the glitches in the output waveform will continue to be consumed in the body diode in the form of heat, seriously affecting the chip's operating stability and output efficiency.

(3) Deadtime is too long, which affects the linearity of the amplifier and causes crossover distortion of the output signal. The longer the time, the more serious the distortion.

2.2 EMI (Electro-Magnetic Interference)

EMI is mainly caused by the reverse recovery charge of the MOSFET body diode. The specific generation mechanism is shown in Figure 6.

In the first stage, MP1-MOSFET is turned on, and current flows through MOSFET and the subsequent LPF inductor; in the second stage, when the full bridge enters the dead-time period, MP1 itself is turned off, but its body diode is still turned on to ensure that the subsequent inductor continues to flow; in the third stage, when the dead-time period ends and MN1 is turned on, if the residual charge stored in the MP1 body diode has not been completely released, the stored charge that was not released during the last turn-on period will be released instantly, resulting in a surge in reverse recovery current. This current tends to form a sharp pulse, which is finally reflected in the output waveform, as shown in Figure 6(b). Therefore, the output spectrum will contain a large amount of spectrum energy at the switching frequency and the switching frequency multiple, forming EMI to the outside.

In order to suppress EMI, some new technologies of internal EMI elimination circuits are applied to new products with the purpose of reducing the output square wave frequency and slowing down the top pulse of the square wave:

(1) Dither. Spread spectrum technology, that is, within the specified range, periodically adjust the triangular wave sampling clock frequency, so that the fundamental wave and higher harmonics avoid sensitive frequency bands, making the output spectrum energy flat and dispersed;

(2) Add an active radiation limiting circuit to actively control the output MOSFET gate when the output changes transiently to avoid high-frequency radiation caused by the subsequent inductive load.

2.3 PCB layout design rules

(1) Since the output signal contains a large amount of high-frequency square waves, the low-distortion, low-insertion-loss LC filter capacitor and ferrite inductor low-pass filter components need to be placed close to the power amplifier to minimize the loop area carrying the high-frequency current and reduce transient EMI radiation.

(2) Due to the large output current, the audio output line diameter needs to be wide and the line length needs to be shortened. Therefore, it is necessary to reduce the passive resistance RP and the filter resistance RF, increase the load resistance RL ratio, and improve the output efficiency.

(3) The bottom of the PCB is the heat dissipation channel with the lowest thermal resistance. The exposed heat dissipation copper area at the bottom of the amplifier should be large. As much copper as possible should be placed between the copper block and the adjacent pins with equipotential and other components. The copper block connected to the exposed pad is connected to other copper blocks on the back of the PCB board with multiple vias. The copper block should have as large an area as possible while meeting the system signal routing requirements to ensure that the chip core has the best heat dissipation characteristics through these copper areas with the lowest thermal resistance.

(4) Add more vias near the ground terminal of high-current devices. If the signal is connected between two layers of the PCB, add more vias to improve connection reliability and reduce on-resistance.

(5) The signal input component pads and signal lines should be kept at an appropriate distance from the output end, and the key feedback network components should be placed in the middle of the input/output PCB layout module to prevent EMI radiation from the output end from affecting the small signal at the input end.

(6) Keep the ground wire and power wire away from the input/output stage and use a single-point grounding method.

3 Design and application of green energy-efficient Class D amplifier TFA9810T based on the above factors

3.1 Internal structure of TFA9810T

TFA9810T is a dual-channel high-efficiency Class-D power amplifier with a rated output of 2×12 W launched by NXP. It is mainly composed of two sets of full-bridge power amplifiers (Full-Bridge), driver front end, logic control, OVP/OCP/OTP and other protection circuits, fully differential input comparator, power supply module, etc., as shown in Figure 7.

[page]

It has the following features: the heat sink can be eliminated, it has high reliability, it is powered by a single power supply of 8-20 V, it has adjustable external gain, the power supply current in the standby energy-saving state is in the micro-ampere level, and it consumes very little energy, etc. It is very suitable for flat-panel TV products, multimedia systems, and wireless audio fields.

3.2 Analog Input Stage Design

The input of TFA9810T adopts a fully differential input circuit that can suppress common-mode interference. Taking the AMP-Rin input of Figure 8 as an example, RA128/RA133/CA139 constitutes a negative feedback low-pass filter, which is used to attenuate the high-frequency carrier component in the feedback signal. Increase the feedback amount of low-frequency components, especially the DC component. It effectively improves the duty cycle error caused by the difference between the DC level of the input signal and the comparator threshold voltage at zero input. Adjusting RA128 can also realize TFA9810T gain control, so that Au (dB) = 20log (VOUT/VIN) ≌ 20log (RA128/RA132). Devices CA153/RA132/RA133 and the internal resistance of TFA9810T constitute a high-pass filter for buffering the input signal. If the capacitance of CA153 is too small, it will affect the low-frequency response. The theoretical determination formula is:

This design takes 1 μF and determines the low-end frequency to be 16 Hz. If the frequency is set too high, the low-end input reactance (such as at 20 Hz) will be too large, which may cause large noise and DC offset noise (plop-noise) at the output. After the feedback signal is compared with the buffered input audio, it enters the TFA9810T through RA133 for PWM modulation. In order to avoid the Rin/Lin input signal frequency in Figure 8 generating sum frequency and difference frequency due to semiconductor nonlinearity, resulting in howling at the output, the two carrier frequencies are adjusted by about 50 kHz by adjusting the capacitors CA123/CA145. In this design, CA123=22 pF and CA145=47 pF are used to achieve a 50 kHz difference in the Rin/Lin carrier frequency.

3.3 Output stage LPF low-pass filter design

The low-pass filter at the output end of TFA9810T adopts the second-order Butterworth filter method. The actual Butterworth second-order filter is composed of RCA-type electronic components CA135/RA145/CA136/LA5/CA137/CA138/RA148/CA159/CA140/CA141/RA152/LA6/CA142/CA144 in Figure 9. It exhibits a pass-through effect on the 15 Hz to 20 kHz audio components in the PWM square wave, and exhibits a -12 dB/octave roll-off rate for high-frequency components above 20 kHz that exceed the audio range.

In the simplified model, Lse and Cse, R, and C1 form a basic Butterworth filter, and R and C1 form a peak elimination circuit with a Zobel network to remove high-frequency spike pulse interference.

3.4 Temperature rise test

The designed power amplifier TFA9810T is powered by a 15.2 V DC power supply, and its operating environment temperature is 20°C. The audio system input is a 2Vp unmodulated 1 kHz single audio signal, and the matching load is an 8 Ω speaker. The audio output power is adjusted to 21 W and it works continuously for 30 min. The maximum temperature in the center of the TFA9810T shell is 45°C measured by temperature testing equipment, and the temperature rise is only 25°C, so there is no need to add a heat sink.

3.5 Audio A/D/A Test Analysis

Figure 10 tests the waveform of a 1 kHz 2V single audio signal at the audio input of the TFA9810T amplifier, and the voltage between the output speaker and GND is 12.84V. The PWM waveform of the amplifier output before the LPF filter in Figure 9. Figures 11 to 13 expand the A/B/C area in Figure 10 respectively.

As shown in Figures 10 to 13, the input waveform is superimposed with high-frequency clutter. This indicates that the front end introduces adverse interference, which needs further analysis and improvement; the output waveform is smooth, without crossover distortion, and has good deadtime characteristics; the input/output sine waves are in opposite phases, and a closed-loop negative feedback path is directly formed by resistors RA128, etc., which reduces noise interference and performs gain control. The expansion diagram of the A, B, and C areas shows that the PWM frequencies at the peak, trough, and S area of ​​the output sine wave are 238.8 kHz, 224.9 kHz, and 626.4 kHz, respectively, with different duty cycles, which conform to the triangular wave sampling characteristics. The PWM pulse fall falling edge and rise rising edge at the peak and trough in the figure are steeper, and contain a large number of high-frequency harmonics compared to the S-shaped area, which is easy to cause EMI radiation. However, after filtering by the Butterworth second-order filter, the output sine wave is good, without obvious high-frequency clutter superposition, and the EMC test also has no obvious external radiation frequency, which meets the design requirements.

3.6 Power and efficiency test

FIG14 tests the actual waveforms of the power supply voltage and current of the power amplifier TFA9810T under the state of FIG10.

As shown in Figure 10, the single-ended output power of the power amplifier is:

From the parameters in Figure 14, we can see that the total power carried by the power amplifier power supply system is:

From this, the efficiency of TFA9810T is:

4 Conclusion

The analog full-bridge Class D amplifier topology is introduced, and the EMI caused by Deadtime is suppressed by the second-order Butterworth filter design and amplifier PCB layout in detail. Finally, a new green energy-efficient dual-channel Class D audio amplifier design is realized based on the NXP Class D amplifier TFA9810T. The simulation and test results show that when the supply voltage is about 15 V, the amplifier can provide 10 W×2 output power to two 8 Ω speakers, with a conversion efficiency of 90%, a total harmonic distortion of less than 7%, a 1 kHz sine wave audio output without crossover distortion, no obvious EMI interference, and a relative temperature rise of 25℃ for the amplifier housing. With the requirements of energy conservation in today's society, this type of green energy-efficient design will be more widely used in the next few years.