Real Time Frequency Domain Filtering Maximizing Vertical Resolution While Minimizing Noise

Log measurements are recorded in the presence of noise statistics, etc. The industry has for decades suppressed the noise and statistics through the use of averaging schemes. Averaging effectiveness is derived from the principle that over an interval, noise or statistics will average to zero and effectively cancel. The processing applied by the industry simply gathered information over a depth or time interval and averaged the values. The averaging could be adjusted through the number of levels selected and the corresponding weights applied at any given level. A major drawback of this technique is that high resolution measurements suffer a severe degradation of resolution when averaging is applied over an interval exceeding the resolution of the device. The industry has attempted to compensate for this problem by increasing the sample rate and applying the averaging over similar sample increments, but shorter physical lengths. However, the increase in sample rate frequently increases both the frequency and severity of noise (especially in the case of radiation tools) resulting in exacerbation of the exact problem one is trying to overcome. In recent years, numerous signal processing schemes have been experimentally applied to well log for the purposes of filtering noise and statistics as well as to deconvolve the spatial response. These techniques were designed for post processing application, requiring multiple passes or analysis techniques not practical for real time application. A technique has been designed and applied where noise and signals are separated on the basis of tool response characteristics. The vertical resolution of any given measurement can be characterized by frequency content in the depth domain, where frequency is described as cycles per unit of depth. Noise or statistics can then be selectively discriminated from signal on the basis of frequency content. Through careful selection of the cutoff frequency, signals can be separated from noise which is relatively high in frequency content. When the filter's frequency is adjusted to the vertical characteristics of the measurement, maximum vertical resolution can be achieved while rejecting higher frequency noise. Filter design, a method for characterizing tool resolution and matching that resolution to the filter, and results of application of the filters are presented and compared to conventional processing.