Research On Noise Suppression Techniques Of Low Power Audio Subsystem

Audio Subsystem gains more and more popularity and is widely used in manyportable apparatuses such as PDAs, mobile phones and MP3s. The development ofaudio subsystem with multi-function and high density integration trends will inevitablyintroduce serious noise and cause a decline in audio performance. Nowdays, theresearches on noise suppression technique mostly focus on functional unit level but notaudio subsystem level, and the technical performance according to these researchesshould be improved too.The dissertation is researching noise suppression techniques of low power audiosubsystem, comprising the suppression technique of aduible noise, background noiseand power supply noise. After theoretically analyzing the generation mechanism ofabove noises，corresponding noise suppression techniques are proposed through thecombination of the system design, circuit design, layout design and PCB design. Andsuch designs are further optimized and verified by simulation and road testing. Theproposed noise suppression techniques could successfully eliminate the audible nosieand reduce background noise and power supply noise, wherein the specific innovationsare listed as following:1. Dummy switch technology (DST) and zero crossing auto gain control (ZAGC)are proposed for suppressing POP noise and broken sound noise of low power audiosubsystem. The DST introduces a dummy input switch to improve the charging speed ofthe input capacitor and calibrate the error between an input singal and a reference signal.The DST further introduces a dummy feedback switch to simulate the startup of themodulation stage. The two switches sucessfully eliminate the POP noise during startupand greatly reduce the startup time. ZAGC automatically adjusts the system gain basedon the amplitude of the audio signal, so as to avoid the broken sound noise caused byclipped audio signal. Gain change is enabled when the audio signal crosses AC zero, soas to avoid sonic boom noise caused by the transient changes of the audio signal.Implemented via0.6μm Bi-CMOS process, it finds that the class D amplifier withabove two techniques does not generate aduible noise during startup and operation. 2. A class D amplifer with an integrated active filter is designed for suppressingbackground noise of low power audio subsystem. The active filter with a3Hz cut-offfrequency is formed by utilizing a MOS device that works in linear region to achieve a300Meg active resistior, and utilizing a0.1mm2layout area to achieve a160pFcapacitor. The active filter is used to reduce the thermal noise of the reference signal ofthe class D amplifer. Implemented via0.6μm Bi-CMOS process, the class-D poweramplifier with the active filter gets test results of0.08%total harmonic distortion,90dBcross talk rejection ratio,72dB power supply rejection ratio, and88dB signal-to-noiseratio, which gains substantially the same performance compared with a class Damplifier with an external filter.3. NMOS LDO is proposed to suppress power supply noise of headphone driver.The NMOS LDO utilizes an NMOS transistor as conduction element, configered toform a low noise positive supply without external compensation capacitor. A low noise,low static power, high swing class AB amplifier is obtained with the utilization ofchopping technique to reduce the offset voltage and1/f noise, quiescent current controltechnique to reduce static power and rail-to-rail structure to increase input and outputsignal swings. Implemented via1.2μm BCD process validation, the headphone drivershows a test result of88dB power supply rejection ratio, and providing an output signalwith7μV noise and1mV offset.4. A low EMI class D amplifier with random spread spectrum modulation isdesigned. A random clock circuit based on the mixed analog-digital control is designedto provide random sampling signal for a class D amplifier, configured to randomlyspread the energy from a fixed frequency to frequencies nearby, thereby reducing theEMI of switching signals. Implemented via0.6μm Bi-CMOS process, the random clockcircuit occupies a layout area of0.05mm2, which is only about2.2%of the entire diesize. Compared with a class D amplifier without spread spectrum, a more than6dBpeak power reduction is measured. The fliterless class D amplifier utilizing proposedrandom spread spectrum modulation could passes the FCC Class B standards forradiated emissions when using24inches of twisted speaker cable.