Study On The Filter Algorithm For Orthogonal Detector In Nuclear Magnetic Resonance Logging

Inversion of nuclear magnetic resonance (NMR) logging echo signal can acquire the information unrelated with rock properties. It can also estimates permeability and determine the characteristic and type of pore fluid. The result of the echo signal peak decay curve directly affects the accuracy of the inversion, so the effective detection of the echo signal peak is one of the key issues of nuclear magnetic resonance logging measurement.Under the premise of introducing the principle of nuclear magnetic resonance echo signal detection, this paper analyzes the characteristics of echo signal which with variable AM signal bandwidth design by the characteristic of fluid, signal to noise ratio is low, detection of real-time requirements is high. On this basis, the paper described the principle of orthogonal detection. Combination of Matlab simulation, this thesis analysis the performance of average filter in digital phase sensitive detection algorithm. The results show that the pass, stop band characteristics of filter is poor, in the case of ensuring signal energy that the signal sampling period is less so the equivalent noise bandwidth is larger, the ability adapts to the echo center offset is poor.To solve the problems, the thesis proposes to design low-pass filter form wavelet orthogonal detection means. By analysis the principle of wavelet, wavelet transform has a variable time-frequency window, the output signal filtering without delay and so on. Through the Matlab simulation, compared to the average-type filter, the pass, stop band characteristics of filter is better, the equivalent noise bandwidth is smaller in the case of ensure the same signal energy, the equivalent noise bandwidth of low-pass filter can adjust by setting the decomposition level, the ability adapts to the echo center offset is better. Finally, to meets the high requirement of real-time echo detection, this paper estimated the computation of wavelet orthogonal detection algorithm, the paper assumed the hardware computing speed rely on the operating frequency of DSP 2812 of TI company. Optimized by parallel computing, under the assumption, the wavelet detector meets the requirement of real-time.