Skip to Main Content
In this paper, we consider the problem of quantization noise suppression in the general family of direct digital-to-RF conversion circuits, including digitally-segmented polar-modulated power amplifiers, or digital PA's. In such power amplifier architectures, the PA is designed to have N parallel segments (or PA subsections), where N equals the number of bits representing the base-band amplitude component of the baseband IQ signal. The PA segment gains are designed to follow a "power-of- 2" weighting pattern. That is, the segments are weighted as 2n; n = 0,..., (N - 1). The outputs of the PA segments are then RF-combined in a proper manner to result in an RF carrier that is amplitude-modulated by a quantized representation of the base-band amplitude signal. The digitally-segmented PA architecture faces an important challenge, which is how to reduce the output noise resulting from quantizing the base-band envelope signal. In this paper, we address such a problem by introducing a novel signal-processing solution, which we refer to as "segmented-filtering (SF)". Our proposed approach has demonstrated the ability to preserve the required signal quality and PA efficiency, while providing more than 35 dB attenuation of the quantization noise. Hence, eliminating the need for substantial post-PA pass-band RF filtering, which is known to be costly and inefficient.