Research On Technologies Of Spectral Analysis For Optical Frequency Domain Reflectometry

Optical frequency domain reflectometry(OFDR),which is originally introduced from frequency modulated continuous wave(FMCW)interference,can offer better spatial resolution than optical time domain reflectometry(OTDR)thus has been widely utilized in the field of local area communication network monitoring,optical fiber sensing and 3D imaging,etc.As a frequency domain technology,OFDR uses a continuous light-wave as the laser source whose frequency is linearly swept.The light-wave is then separated into a probe and a local reference.The reflected probe from the fiber under test(FUT)is received and mixed with the local reference.Since the reflected probe beam from FUT is a time-delayed version of the original linearly swept beam,a reflection point in the FUT yields a single beat frequency which is proportional to the round-trip time of reflection.Regularly,FFT is applied to perform spectrum analysis of the beat signal.However,several drawbacks of traditional FFT are: 1)The length of input series must be limited to a power of two;2)No implementation can be started before the completion of data acquisition;3)Large memory is required for long signal to achieve fine resolution.According to the resolution of FFT,one possible approach to higher accuracy is to adopt relatively longer sampled data because the minimum of sampling rate is restricted by the Nyquist sampling law.But this will leads to heavier computational complexity as well as larger storage size.In some applications,people require very high frequency resolution not over the entire band but in several certain areas.One example of this is electromagnetic harmonic in power supply networks where only a few harmonic components sparsely distributed are interested.Another example is imaging processing for FMCW SAR where high resolution is required to precisely measure the moving objects.There are many theories have been proposed to make up the drawbacks of traditional FFT.The Chirp-Z Transform(CZT)are classical methods to gain higher resolution within a small portion of the spectrum that users may “zoom in” to find more details of the frequency structure within a predetermined range.However,CZT suffer from much longer processing time than FFT because of its complexity and inefficiency of implementation.In order to perform CZT with a smaller memory size and,to make it able to deal with a large size of input data,Segmented Chirp-Z Transform(SCZT)algorithm was first introduced in 1990 by T.T.Wang [12].Yet,there are several disadvantages that impeded the future of this algorithm: 1)It cannot work if the zoom range is unknown;2)The operational efficiency needs to be optimized;3)It will suffer severe computational burden when analyzing the whole frequency range.In this paper we propose a combination approach named FFT-SCZT for precise and fast spectrum analysis in a wide range.Such an approach is designed for more efficient and accurate calculation of a specific frequency band.The FFT is used for a coarse analysis of the entire frequency range and quick determination of the zoom bands.An optimized SCZT algorithm is then used,along with parallel computing methods,to find more details of frequency components within the bands.Using this approach,the whole analyzing process dramatically reduces the requirement for memory while dealing with huge and continuously received data.The segment characteristic of SCZT makes it easy to realize parallel computation for our approach.To demonstrate the performance of FFT-SCZT,we apply it to an OFDR system to realize fast analysis of five closely spaced fiber Bragg gratings(FBGs).The experimental result shows that all five FBGs have been detected within 2 seconds.The resolution reaches 2 mm and the processing efficiency is about 4 times better than CZT and 30% better than FFT.Several significant advantages are provided by FFT-SCZT combination approach.Firstly,the spectrum analysis can be real-time processed during the data acquisition.And also all parameters for SCZT are optimized to guarantee maximum efficiency.Furthermore,the segmentation permits big data could be processed using smaller memory.Additionally,the segmented characteristic makes it easy to realize parallel computing to accelerate the process of analysis.FFT-SCZT is a promising approach to deal with a huge amount of signal data for precise and real time frequency analysis.This approach is applicable to any situation that frequency peak detection requires high resolution when components distributed in a broad band.