A novel high-fidelity analog class-D amplifier based on self-oscillation for earphones/headphones

Abstract

The specifications of an earphone/headphone audio amplifier typically include high-fidelity (including Signal-to-Noise Ratio >90dB, Total Harmonic Distortion + Noise > 80dB and Power Supply Rejection Ratio > 70dB), in part due to the psychoacoustic effects of donning earphones/headphones and to the requirement of the single-ended output. Despite the power-efficiency (n) advantage of Class D amplifiers (CDAs) and their increasing acceptance in general audio applications, they remain largely unacceptable as earphone/headphone amplifiers because their fidelity are insufficient. Not unexpectedly, earphone/headphone amplifiers, at this juncture, are presently predominately Class AB amplifiers, and to the best of the author’s knowledge, there is no reported commercial CDA appropriate for earphones/headphones (for mobile applications). We propose the design of a novel single-ended output Class D earphone/headphone amplifier based on the complete-feedback self-oscillation architecture. The complete feedback approach is advantageous over conventional (‘incomplete feedback’) CDAs because the LC lowpass filter now constitutes part of the feedback loop, and the nonlinearities thereof are mitigated by (Loop Gain + 1). The self-oscillation approach exploits the phase lag of the LC lowpass filter, thereby reducing the hardware complexity of the CDA. The novelty of the proposed design includes an integrator-cum-control block that provides high loop gain and a well-defined oscillation frequency. These attributes effectively mitigate the nonlinearities, including the high nonlinearities at high modulation indexes of reported CDAs based on the same design architecture. We further analytically derive the parameters pertinent to the proposed CDA to depict the mechanisms of and the pertinent parameters that affect the nonlinearities. This analytical work is useful as it provides insight to the design of the proposed CDA. By means of simulations on the proposed design based on a commercial 0.35um CMOS process and physical measurements on the same CDA built with discrete components, we show that the proposed CDA meets the specifications required of earphone/headphone applications. We also show that its power-efficiency is significantly higher than its competing (commercial) Class AB amplifiers (nCDAmax = 90% vs nClassABmax = 65%), a very worthy advantage in portable audio applications. When benchmarked against reported CDAs for earphone/headphone applications, the proposed design is advantageous in terms of lower nonlinearity, lower cost (a cheaper inductor may be employed) and lower load variance. The shortcomings of the proposed design are the higher switching frequency, higher sensitivity to variations of the passive components and higher component count. These shortcomings are discussed and are not serious, particularly when viewed in terms of the merits of the proposed design.