Abstract
A new pulsed nuclear magnetic resonance (NMR) logging tool, known as the CMR* Combinable Magnetic Resonance tool, is being used worldwide. During the CMR tool development phase, one challenge was to design a robust and economical data acquisition and signal processing scheme for the hundreds or thousands of spin-echo amplitudes that can be acquired during the Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence. This challenge was met by developing a new signal processing and associated downhole data compression algorithm. Data compression is essential to reducing the processing times so that formation T2-distributions can be estimated in real time. Compression of the digital data is possible, without loss of information, because the linear dependency of the NMR measurement kernels results in gross redundancy of the measured spin-echo amplitudes.
An attractive feature of the algorithm described in this paper is that the compression can be performed in the downhole tool, thus substantially reducing the telemetry requirements. The raw spin echoes can also be sent uphole and made available for additional processing.
Logs of CMR porosity, free-fluid porosity, mean relaxation time and rock permeability are computed from the estimated T2-distributions. The accuracy and precision of the CMR log outputs are demonstrated by repetitive Monte Carlo computer simulations in which noisy, synthetic spin-echo amplitudes are generated from known T2-distributions and then processed to obtain log outputs. Monte Carlo simulations are used to elucidate CMR log responses in typical clean sand, shaly sand, and carbonate rocks. The relative insensitivity of the measurements to short relaxation times (e.g., those less than a few milliseconds) is discussed and used to explain the differences between CMR log porosity and "total" formation porosity in shaly formations. We show that CMR porosity is an "effective porosity" that does not include clay bound water and microporosity having relaxation times less than a few milliseconds.
We give examples of statistical fluctuations that can occur on estimated T2-distributions to assist log analysts in recognizing artifacts that are not indicative of actual reservoir rock properties. Field logs that display many of the features of CMR log responses revealed by the simulations are also presented.
