| Magnetotelluric sounding method (MT) is one of the main methods of geophysical exploration, it is a effective way for oil exploration, geothermal survey and exploration of mineral resources. In recent years with the theoretical research and field practiced, MT becomes more and more be applied to the target of exploration. With the increase of the depth of exploration, it which put forward higher requirements on the quality of MT data. Because MT’s frequency range is very wide and the signals are very weak, and its resolution is low. These problems troubles the development of the MT. But the performance is limited by the instrument, the low frequency signal is calculated by manual processing results, which greatly restrict MT data later.MT signal is very complex, contains a lot of useful information. The MT signal source come from the sky, so the signal is very weak, very vulnerable to noise interference, which including interference, human interference and interference of Geology, so the time domain signal denoising is a very important link.In terms of temporal filtering predecessors have done a lot of work, including 2D median filtering and remote reference treatment, although the two-dimensional median filter can filter out a lot of spikes, and the curve becomes smoother, but also it filter out a lot the useful signal. Remote reference technique have good suppression effect for electromagnetic noise and geological noise between reference point and measuring point, but suppression effect is small for the humanities noise, and far away the choice of the reference point is very important for reference technology, it also increased the difficulty of field work. The general Fourier transform have a good effect for processing stationary signals, but magnetotelluric data is generally not stationary random signal. Wavelet transform which developed on the basis of the Fourier transform have a very good effect on non-stationary signals, and it use wavelet transform to filter acquisition signal in the time domain, which can be very good in noise suppression.Power spectral estimation is very important in magnetotelluric signal processing and is an indispensable part of application of geophysics. The power spectrum contains a lot of the spectrum information, including the electromagnetic wave frequency, amplitude, phase. Power spectrum is the comprehensive reflection of underground geologic body when the electromagnetic wave propagate in the ground, haying a good estimate of the power spectrum can play a significant effects for the late processing and interpretation of data.Currently power spectrum estimation method is divided into two kinds of classic method and modern method. Classic method is divided into two kinds of direct and indirect methods, both of which are based on the Fourier transform, but this approach is not accurate to get the power spectrum, because the Fourier transform is currently processing the limited length signal, and then have a signal extension for the cycle of finite length. According to the Fourier transform theory to obtain the full spectrum of information, it need to process an infinitely long signal processing, but this is not possible with current computer technology, so in information processing, it is necessary for an infinitely long signal processing, generally we use truncation processing, but the truncation process prone to truncation effect, causing leakage spectrum so that is not accurate for power spectrum estimation. Modern method to estimate the power spectrum is divided into many kinds, such as parametric modeling, and maximum entropy method.Magnetotelluric impedance elements estimating commonly use least squares estimation and robust estimate, MT is very complex, and the signal contains much more, a single power spectrum does not reflect the true subsurface information, so magnetotelluric power spectrum calculation is a statistical process. In the process, it use Fourier transform to get the field values and magnetic values, and it use the multiple sets of power spectrum to do the least squares estimates, producing an estimate and calculating an error, the least square method must meet three conditions, first is that the signal is not related to the noise, and second is that the noise between the various components is unrelated, the last is that the noise is a Gaussian distribution.However, in the actual acquisition signals, it often do not satisfy the above three conditions, they are often affected by "flying spot" when using the least squares estimate the power spectrum, affecting the quality of data processing. Robust method is proposed to address the problem of "flying spot", Robust estimated use the weighted average method, and a point with good quality is given heavier weights, for the "flying spot" is given relatively small weights, so it can effectively remove the influence of "flying spot", and to provide a better quality of data for the latter part of the apparent resistivity inversion and calculation.At present, the magnetotelluric instrument and processing software are imported from abroad, first it is not useful for data processing in a certain demand, on the other hand it also restricts the development of the MT instrument and data processing software in china. For the analysis of MT data, although there are many popular methods, but the actual application effect is not good; and it is similar to the data processing, domestic independent research and development level is not high.V5-2000 system is a high-precision electrical equipment from Canadian Phoenix, it get MT time series data, due to the high integration of software, we could not get the instrument format information and inside information about its file structure. Foreign imports expensive equipment, which also limits the level of development of magnetotelluric instruments, domestic equipment acquisition criterion has not yet reached the world advanced equipment acquisition accuracy, but magnetotelluric data processing method has been very mature, so it is very necessary to start from V5-2000 with the original file system, then analysing their original file format and content, based on the original acquisition file from V5-2000 system, appling current processing methods to process data, first it can guarantee the high quality of the original collection data, then it can develop a domestic magnetotelluric processing software by collecting original documents of V5-2000 systemThe Fourier coefficients of frequency domain is calculated by Fourier transform from time series, this process comes with Phoenix software has done a very mature, and its time-frequency conversion method uses a cascade of sub-sampling method, although this method speed up the data processing speed, but the same it results in a lot of waste data.With the rapid development of modern computer speed, they can already handle large amounts of data. Based on the above analysis, this paper is based on raw data collected by the V5-2000 system, and it research data flow which is handled by SSMT2000 software, using program to read instrument calibration files and collected TBL file, and to identify the physical meaning of the parameters represented, using wavelet transform to filter the time domain signal, then using DFT to process the filtered signal, and then we estimation the power spectrum of electric field and magnetic field value, statistical error and estimating impedance elements to prepare MT time series processing software. Finally, combined with examples, to verify this paper. |
