- 學位論文
以即時掃描頻率萃取演算法實現立體頻譜儀
An Approach to FPGA-based Time-Frequency Spectrogram by Real-Time Sweep Spectral Extraction Algorithm
摘要
我們已經採用即時掃描頻率萃取之演算法，替代快速傅立葉演算法來縮減大量的記憶體使用需求。藉著固定掃描頻率範圍的本地震盪器和中心頻率的FIR濾波器，我們可以更有效的利用現場可程式閘陣列 (Field Programmable Gate Array : FPGA)的硬體資源，以及得到可靠的頻譜分析效能。由於採用時間頻率漲縮演算法(time-frequency scaling algorithm)使我們可以量測到頻率範圍從sub-Hz至MHz的任意時變信號。我們利用Matlab與Simulink弁鄐頞蘁c建即時掃描式頻譜分析儀之訊號處理流程，且搭配了Xilinx公司為Matlab所發展的System Generator DSP Block Library模組。這些模組不但可與Simulink弁鉏珩桴膃X在一起作數位處理系統之軟體弁鉏甡嚏A而且可以產生現場可程式閘陣列(FPGA)晶片組態檔來作硬體驗證，同時也解決此種系統除錯不易的問題。採用高速、高運算效能的FPGA晶片為硬體平台，使我們能夠透過時間演化過程累積音頻信號的頻譜數據，以時間為Y軸來建立一張三維的即時頻率頻譜圖(X-Z)。
並列摘要
We have proposed an approach to build up an FPGA-based time-frequency spectrogram for time-varying signal. Instead of short time-interval FFT algorithm, we have adopted real-time sweep spectral extraction algorithm to reduce large memory usage. By fixing the sweep frequency range of local oscillator and the center frequency of Finite Impulse Response (FIR) filter, we can obtain more effective usage of FPGA hardware resource and stable spectrum analysis performance. The time-frequency scaling algorithm has been used to guarantee the measurement of wide frequency range (sub-Hz to MHz) for arbitrary input time-varying signal. The high speed FPGA performance enables us to build up a three dimension time-frequency spectrogram for audio signal by accumulating the spectral data against the evolution time.