Study on fast convergence method of IIR low pass filter

Abstract

In wireless communication, such as 4G LTE,5G NR and next G modem design, it is necessary to average the measured values such as noise power, time domain correlation of frequency offset estimation, frequency domain correlation of time offset estimation and signal-to-noise ratio (SNR) over a period of time in order to reduce noise and make them converge to stable estimation. These measurements have the characteristic that they are stabilized around a constant value for a relatively long time, they can be assumed nearly zero-frequency or low frequency signal with additional Gaussian noise. This paper discusses two kinds of IIR low pass filter to filter the above measured metrics. The one is nonlinear phase low pass filter, the other is nearly linear phase low pass filter. A new method is proposed to reduce the convergence time of IIR filter forced response, including adaptive determination of filter coefficient, selection of appropriate initial system value, and update long term value when the compensation occur.