Fixed-Point DSP Implementation of FM Demodulation/Decoding

This paper addresses the fixed-point DSP implementation of FM demodulation and decoding for the digital radio application. The demodulation module involves division and arctan for which there is no dedicated hardware on a fixed-point DSP. Division is normally achieved either by using the Newton-Raphson method or by using conditional subtraction instruction. It is normally more efficient to use a table lookup approach. However, in a table lookup, if the denominator value is very small, the size of the table lookup will be very large for high accuracy. In this paper, we present a new method, which utilizes the in-phase representation of the denominator overcoming the table lookup problem. We have shown that this method reduces the instruction cycle time by 40% for the demodulation module. In addition, in the decoding module, we have devised an efficient implementation of pilot frequency computation, which reduces the instruction cycle time by 30%. Example implementation timings on the TI TMS320C6201 DSP are reported.