Method of raising quantizing precision of pulse width modulation

Abstract

The invention discloses a new method for quantifying pulse width modulation, in order to provide higher quantifying precision at the condition of lower clock frequency. In the invention, an input data 'Data' with P bits is decomposed as a high position data H with m bits and a low position data L with n bits. Encoding the two data H, L forms data X and Y that are inputted to two PWM quantizers. The clocks of two PWM quantizers are fn and fb respectively. Expressions of output data passed of two PWM quantizers are X / k* (2 to the n power -1) and Y / k* 2 to the n power respectively. The above data conform to following relational expressino (1) m+n=p; (2) Data=2 to the n power *H+L; (3) fn=k* (2 to the n power -2) *fr, fb=k*2 to the n power *fr, here k is a constant, fr is data sampling frequency, n is a positive integer and its value equals to number of bits of data L. (4) Data / k * 2 to the n power * 2 to the n power -1)=X / k* 2 to the n power -1)-Y / k* 2 to the n power.

Description

technical field [0001] The invention belongs to the pulse width modulation of digital signals, and in particular relates to the pulse width modulation of digital signals by adopting a double-mode quantization method. Background technique [0002] The analog audio signal input by the class D amplifier is pulse-width modulated by the internally generated triangular wave to form a pulse signal output. Because the signal is continuous on the time axis, it is an analog power amplifier. [0003] When the current all-digital audio power amplifier performs pulse width modulation of digital signals, due to the limitation of semiconductor technology, the quantizer has low precision, resulting in a large output quantization noise. Therefore, the existing schemes all use increasing the oversampling multiple and increasing the order of noise shaping, so that the noise components in the audible frequency band are migrated out of the audible frequency band. But the following problems are...

AI Technical Summary

Problems solved by technology

But the following problems are: excessively increasing the oversampling multiple will make the requirements for power devices more stringent, and the efficiency of the amplifier will be reduced; and increasing the order of noise shaping will produce high-frequency oscillation and limiting distortion, and make high-frequency Poor sound quality

Images