Analysis on the convergence property of quantized-x NLMS algorithm

The adaptive system design by 16-bit fixed point processing enables to empl oy an inexpensive digital signal processor (DSP). The narrow dynamic range of such 16 bits, however. does not guarantee the same performance that is c onfirmed beforehand by computer simulations. A cause of degrading the perfo rmance originates in the operation halving the word length doubled by multi plication. This operation rounds off small signals staying in the lower hal f of the doubled word length to zero. This Problem can be solved by limitin g the multiplier to only its sign (+/-) like the signed regressor algorithm , named 'bi-quantized-x' algorithm in this paper, for the convenience menti oned below. This paper first derives the equation describing the convergenc e property provided by a type of signed regressor algorithms. the bi-quanti zed-x normalized least mean square (NLMS) algorithm, and then formulates it ., convergence condition and the step size maximizing the convergence rate. This paper second presents a technique to improve the convergence property . The bi-quantized-x NLMS algorithm quantizes the reference signal to +/-1 according to the sign of the reference signal. whereas the technique moreov er assigns zero to the reference signal whose amplitude is less than a pred etermined level. This paper explains the principle that the 'tri-qunatized- x' NLMS algorithm employing the technique can improve the convergence prope rty, and confirms the improvement effect by computer simulations.