Shaping Quantization Noise and Clipping Distortion in Direct-Detection Discrete Multitone

Abstract

This paper introduces a method for boosting the signal-to-noise ratio at the transmitter of a discrete-multitone system through reshaping of the spectrum of quantization noise and clipping distortion. This is of particular importance for high-bandwidth transmission, such as 30 GHz, where digital-to-analog converters have a very limited resolution of only 5-8 bits. This limitation of the data converters is thus significant and reducing their noise contribution has a pronounced effect on the overall system's performance. The proposed technique resembles the quantization noise shaping of delta-sigma data converters but is a more general technique that shapes both quantization and clipping noise and it additionally does not require oversampling. The technique relies on the existence of an unused band of subcarriers, as would be the case, for example, when tone reservation is applied. The proposed system is additionally well-suited for low-power transceivers as its most computationally-expensive element is a short finite impulse response filter. In a practical setting, the noise contribution of quantization and clipping can be reduced by 6 dB.