Design of chopper amplifier circuit in class D audio system

1 Introduction

In the design of the op amp circuit of the Class D audio amplifier, the signal's total harmonic distortion (Total Harmonic distortion) and noise pose a challenge to the design of the op amp. For audio signals in the range of 20 to 20 kHz, the distortion of the op amp is mainly caused by voltage offset and low-frequency 1/f noise. The relatively high 1/f noise and voltage offset of the CMOS process make this problem particularly serious. When the offset voltage of the circuit is required to be less than 1mV and the input equivalent noise is less than 100nV/Hz. Ordinary CMOS op amps are difficult to meet the requirements. The common static offset cancellation technology, such as trimming, can effectively eliminate the influence of voltage offset, but it cannot reduce 1/f noise. The best way to solve this problem is to use dynamic offset cancellation technology (dynamic offset-cancellation techniques), such as automatic zeroing and chopping technology. The automatic zeroing technology (Auto zero tiechnique) is to reduce the offset and noise by sampling the low-frequency noise and offset, and then subtracting them from the instantaneous value of the signal at the input or output of the operational amplifier. Since the automatic zeroing technology uses the principle of capacitor sampling, it is very easy to fold broadband thermal noise into the baseband frequency during circuit operation. The wider the bandwidth of the op amp, the more noise there is on the sampling capacitor, usually up to 70nV/Hz. The chopper technique uses the modulation and demodulation principles to move low-frequency noise and offset to the high-frequency part and filter it out using a low-pass filter. Since there is no aliasing of thermal noise, the noise voltage of the op amp is lower than that of the automatic zeroing technology. However, the influence of the charge injection and charge feedthrough effects of the chopper switch can still produce a residual voltage offset of about 100uV. Moreover, the use of the chopper switch will increase the thermal noise level of the device.

Therefore, this paper designs a single-power-supply fully differential chopper op amp circuit based on the 0.35-micron N-well process. At the same time, in order to reduce the residual voltage offset, the T/H (track-and-hold) demodulation technology is adopted. When the circuit works at a chopping frequency of 150KHz, the input equivalent noise reaches 31.12nV/Hz.

2 Working Principle of Chopper Op Amp

The principle of the chopper op amp is shown in Figure 1, where Vin is the input audio signal, which is modulated by a chopper switch with a frequency of fch and an amplitude of 1. According to the Nyquist sampling principle, in order to avoid aliasing of the input signal, fch must be much larger than 2 times the signal bandwidth.

Figure 1 Principle of chopper op amp

After modulation, the signal is moved to the odd harmonic frequency of the chopped square wave. This signal is amplified by the operational amplifier with a gain of Av. At the same time, the input noise and input offset voltage of the operational amplifier are also amplified by the operational amplifier. After the output of the operational amplifier is modulated by the chopper switch with an amplitude of 1 and a frequency of fch, the output signal is:

It can be seen from formula (1) that after the second chopping, the input audio signal is demodulated to the low frequency band, while the voltage offset and low-frequency noise of the op amp are moved to the high-frequency odd harmonics of the chopped square wave after only one modulation. After low-pass filtering, the high-frequency component in the output signal is filtered out and the low-frequency component is restored to the audio signal, thereby achieving accurate amplification of the audio signal.

Performing Fourier analysis on the output signal, the final input noise spectral density (PSD) of the op amp is obtained as:

The coefficient K is related to the noise parameters of the process.

3. Design of operational amplifier circuit

The chopper amplifier designed in this paper is a CMOS fully differential circuit structure. It consists of four parts: chopper switch, main op amp circuit, output stage and common mode feedback circuit. The operating voltage range of the circuit is 2.5V~5.5V. The circuit structure of the chopper operational amplifier is shown in Figure 2.

Figure 2 Circuit structure of chopper op amp

The input chopper switch completes the modulation of the audio signal. The chopper switch will introduce residual voltage offset at both the rising and falling edges of the clock. Figure 3 shows the waveform of the residual offset voltage at zero input.

Figure 3 (a) Residual offset voltage (b) Chopping signal

Figure 4 T/H demodulation and control timing

By analyzing the CMOS switch characteristics, it can be concluded that the equivalent input residual offset voltage is Vos,rmts=2Vspiketfch, where t is the time constant of the MOS switch. From this formula, it can be seen that there are three ways to eliminate the residual voltage offset:

1. Reduce the chopping frequency:

2. Reduce input resistance;

3. Reduce the charge injection effect of the chopper switch.

Since the corner frequency of MOS tube 1/f noise is generally above tens of KHz, reducing the chopping frequency cannot modulate the 1/f noise well, and the input resistance is only related to the internal resistance of the signal source. It is difficult to reduce the input resistance in the design, so we can only consider reducing the charge injection effect of the switch. For this reason, the input chopping switch adopts a complementary clock structure and uses the minimum line width in size. On the one hand, it can reduce the on-resistance of the transmission and provide a larger voltage swing; on the other hand, it reduces the impact of charge injection and feedthrough and reduces the residual voltage offset. Considering that the 1/f noise characteristics of PMOS tubes are better than those of NMOS tubes, large-area PMOS tubes are used for input tubes MP1 and MP2, which can not only reduce the voltage offset caused by device mismatch, but also reduce the corner frequency of transistor 1/f noise and improve the noise characteristics of the operational amplifier.

In order to further reduce the residual voltage offset, the output of the fold-cascode op amp uses T/H demodulation technology, and the circuit structure and timing are shown in Figure 4. The working principle of this circuit: when tracking the signal, K1~K4 are closed, K5~K8 are disconnected, and the output signal is maintained on capacitors C1 and C2. When the circuit outputs, K1~K4 are disconnected, K5~K8 are closed, and the voltage values ​​of C1 and C2 are loaded on the load capacitor C3 for summation. Since the voltage on C2 is reversed when it is superimposed on the load capacitor, the residual voltage offset of the amplifier can be effectively offset. Since the demodulator uses high-resistance node chopping. Therefore, a smaller area NMOS tube switch can be used to reduce the impact on the output pole.

The main op amp adopts a fully differential folded cascode structure. In the Class-D structure, due to the frequent opening of the output power MOSFET with large current, the electromagnetic interference (EMI) generated will form a strong ripple on the power supply. In actual applications, it is found that when the chip works at a power supply voltage of 5V, the power supply fluctuation caused by EMI can reach ±2V. The fully differential structure can not only improve the power supply rejection ratio and common mode rejection ratio of the op amp, reduce the influence of power supply noise and common mode noise, but also avoid the mirror pole, so it can still show stable characteristics for a larger bandwidth.

In order to provide higher gain and voltage output swing, a common source op amp output stage is added after the fold-cascode. After using a two-stage op amp, the frequency stability of the op amp is analyzed. Without considering the influence of the chopper switch for the time being, it can be inferred that the circuit has at least three LHP poles, which are the main pole Wp1 introduced by the miller compensation capacitor, the output pole Wpout generated by the output filter capacitor. It is the first non-main pole, and the non-pole Wp3 introduced by the folded-cascode (the drain end of MN1 and the source end of MN3). The relationship between the three is Wp1

The common-mode feedback circuit is composed of MN7~MN10, MP10-MP12, one end of the input is connected to the reference voltage of VDD/2, and the other end is connected to the common-mode output of the main op amp. The common-mode detection circuit is composed of resistors and capacitors. After error amplification, the bias current of the main op amp is regulated.

4 Simulation Results and Layout Design

The circuit designed in this paper was simulated and analyzed using the Cadence Spectre tool under the SMIC 0.35 micron N-well process. The process parameters of each device are typical, the power supply voltage is 5V, the input signal is a standard sine wave with an amplitude of 10uV and a frequency of 1KHz, and the chopping frequency fch=150K. The simulation waveforms are shown in Figures 5 and 6.

Figure 5 Amplitude-frequency to phase-frequency characteristic curve of the op amp

Figure 6 Chopping output waveform

As can be seen from Figure 5, under typical conditions, the main pole of the op amp is within 10HZ, and the phase margin is about 75 degrees. It can fully guarantee the stability of the op amp under various comer conditions. From the input fn waveform, the residual voltage spike caused by chopping has also been significantly improved. Table 1 shows the open-loop simulation results of the op amp.

Table 1 Op amp open loop simulation results

The layout of the circuit is designed and optimized using SMIC 0.35um process rules. The substrate grounding uses a fully enclosed double gardring to effectively reduce the substrate coupling noise. The differential pair uses a dumb gate common centroid matching to reduce input voltage offset. In addition, in order to reduce the interference of the peripheral circuit to the op amp, the filter capacitors of the post-stage are dispersed around the op amp circuit. The optimized layout area is 0.24mmx0.34mm, as shown in Figure 7.

Figure 7 Layout

5 Conclusion

The 1/f noise and voltage offset of the Class-D audio amplifier have a direct impact on the distortion and noise performance of the signal, especially the background noise when the input signal is zero. By adopting the fully differential chopper op amp circuit and T/H demodulation technology, the low-frequency noise and voltage offset of the system are effectively reduced. The test of the chip after tape-out shows that the circuit has greatly improved the noise performance of Class-D.

The author's innovation: The fully differential chopper op amp circuit and T/H demodulation technology are used to effectively reduce the low-frequency noise and voltage offset of the Class D audio system.

Project economic benefits: This project has been successfully taped out. According to Forward Concepts lnc data, the total output value of global Class D audio amplifiers in 2008 reached 800 million US dollars.