21.3 A -106.3dB THD+N Feedback-After-LC Class-D Audio Amplifier Employing Current Feedback to Enable 530kHz LC-Filter Cut-Off Frequency

Class-D amplifiers (CDAs) are used in various audio applications thanks to their high power efficiency. However, they produce high-frequency switching energy that poses EMI challenges. For applications such as automotive, the stringent EMI requirement necessitates the use of an LC filter, which can add significant bulk and cost. To reduce these, not only the LC component values but also their linearity requirements should be reduced. Moreover, a certain LC tolerance should be allowed for practical usage. In [1], high LC filter cut-off frequency (f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> ) (580kHz) is enabled, resulting in small LC component values. However, relatively expensive and bulky linear LC components are still required to achieve low THD+N due to a lack of LC nonlinearity suppression. Using a voltage feedback-after-LC architecture to suppress LC nonlinearity, the CDA in [2] enables the use of more nonlinear, smaller, and cheaper LC components and allows ±30% component tolerance. However, f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> is greatly reduced (106kHz max.) due to a fundamental tradeoff between f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> and LC nonlinearity suppression. Alternatively, current feedback can be applied to CDAs [3, 4]. However, they employ bulky LC filters (e.g., f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> »20kHz in [3]), and analysis on LC nonlinearity suppression and tolerance is absent. In this work, current feedback is exploited in a feedback-after-LC CDA to maximize f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> , achieve sufficient LC nonlinearity suppression, and allow good component tolerance. Implemented in a 180nm BCD process, the prototype enables a maximal f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">LC</inf> of 530kHz while achieving -106.3dB peak THD+N, 12.8× inductor volume reduction compared to [2] and ±30% LC tolerance. The CDA can deliver 14W into an 8Ω load with 90% efficiency, measured at 10% THD.