基于修正Rife算法的正弦波频率估计及FPGA实现

Rife算法的基础上,通过对输入信号进行频谱搬移,给出了一种修正Rife(MRife)算法。该算法易于并行实现。Monte Caro仿真表明,MRife算法具有频率估计精度高、整个量化频率范围内性能平稳等优点。当SNR(信噪比)大于0 dB时,MRife算法频率估计均方根误差接近克拉美-罗限(CRB,Cramer-Rao bound)。为了提高算法FPGA实现时的系统运行速度,提出使用FFT运算后的实/虚部代替FFT模进行插值,仿真表明对MRife算法性能影响不大。最后,将MRife算法在单片FPGA芯片内进行了硬件设计。布局布线后的时序仿真结果表明,该设计能够对输入数据速率为200 MHz的信号进行实时频率估计,数据不堆积。

Modified Rife algorithm for frequency estimation of sinusoid and implementation in FPGA

Based on Rife's method,a modified-Rife(MRife) algorithm is proposed which is derived from moving the frequency of input signal.This algorithm is easy to be realized with a parallel structure.Monte Caro simulation results indicate that MRife algorithm has good and steady frequency estimation capacity in the whole frequency range.When the input SNR(signal-to-noise ratio) exceeds 0 dB,the RMS error of MRife algorithm can approach CRB(Cramer-Rao bound).In order to improve system speed,only real or image part of the FFT coefficients is used when implementing MRife algorithm in FPGA.Simulation results show that this simplification doesn't modify the efficiency of MRife algorithm too much.Finally,the MRife algorithm is implemented in a single FPGA IC.According to the after-place-and-route simulation results,this design can achieve real-time frequency estimation even though the input data rate is as fast as 200 MHz.There is no data deposit.