James E. Galford

A Method for Enhancing the Vertical Resolution of NMR Logs

The vertical resolution of continuous, multi-frequency NMR logs largely depends on the length of the stacking interval. Several factors determine the stacking length, such as the number of Carr-Purcell-Meiboom-Gill (CPMG) echo trains that are stacked to improve signal-to-noise ratio, the time interval between individual CPMG events, and logging speed. Standard processing requires stacking of phase-alternate pairs of CPMG events to remove ringing contributions. Ringing lumps together the non-formation signals present in the acquired CPMG echo trains. This requirement imposes a minimum stacking length for any combination of CPMG sequences. If additional stacking is required to improve signal-to-noise, the simplest filter design requires multiples of the minimum stacking interval. This may produce a better signal-to-noise-ratio, and a larger degradation of vertical resolution than is necessary. An alternative single echo-train offset removal method (SETOR) has been developed to improve the vertical resolution of NMR logs. In this method the ringing contribution is estimated and removed from the individual CPMG echo trains before stacking. In many cases, results with significantly better vertical resolution, compared to the standard processing, can be obtained because the minimum stacking interval is reduced to a single CPMG event. This allows the stacking interval to be tailored to a desired signal-to-noise ratio. The application of the technique is particularly important for the latest nine-frequency tools that are capable of logging at greater speeds than previous tools. The theory and current implementation of SETOR is described. In addition, a log example is presented to demonstrate how the method can significantly improve vertical resolution. In this example, stacking length could be reduced from 8.5 feet to 3 feet producing a porosity log that closely matches the finer scale features of a conventional density porosity log. Station data are also presented to help validate the method.