Walsh-Hadamard Transform

Key Takeaways

• Walsh-Hadamard encoding in NMR logging tools enables simultaneous multi-frequency or multi-phase RF pulse sequences that improve the efficiency of T1 and T2 relaxation time measurements in formation pore fluids, reducing the total measurement time required to acquire the NMR data needed for porosity and permeability estimation while drilling at commercial penetration rates: conventional NMR logging uses a Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence in which a series of 180-degree refocusing pulses generates an echo train whose amplitude envelope decays at the rate of the transverse relaxation time T2, with the distribution of T2 values across the pore size spectrum being inverted from the multi-exponential echo train amplitude decay; the limitation of sequential CPMG acquisition is that each T2 measurement at a single CPMG echo spacing takes time proportional to the longest T2 component in the formation (up to several seconds in large pores), and multiple measurements at different echo spacings or wait times are needed to characterize the full pore size distribution and distinguish capillary-bound from free-fluid water; Walsh-Hadamard encoding allows multiple CPMG sequences with different echo spacings or polarizations to be applied simultaneously in orthogonal pulse patterns, with the receiver applying the inverse WHT to separate the individual responses, effectively halving or quartering the total acquisition time for the same NMR data quality; this acquisition efficiency improvement is particularly valuable in LWD NMR logging where the logging speed is limited by the formation's T1 relaxation time and the tool's depth resolution requirement, and any reduction in the required measurement time per depth point directly translates to higher logging speed or higher data quality at the same speed.
• Hadamard spectroscopy in downhole fluid analysis tools applies the WHT encoding principle to improve the signal-to-noise ratio of spectroscopic measurements made on formation fluids sampled from the formation during wireline formation testing (MDT, RFT) or LWD fluid sampling operations: conventional single-channel spectroscopy illuminates the fluid sample with one wavelength at a time and measures the absorption or fluorescence at that wavelength sequentially across the spectral range of interest (a scanning or dispersive spectrometer approach), with the signal-to-noise ratio per spectral channel being limited by the measurement time available at each wavelength position; Hadamard spectroscopy replaces the single-wavelength illumination with a series of Hadamard-encoded multi-wavelength illumination patterns (using a spatial light modulator or a programmable light source), with the spectrometer detector measuring the combined intensity of multiple wavelengths simultaneously and the WHT applied to the sequence of detector readings to recover the individual wavelength responses; the Fellgett advantage (also called the multiplex advantage) of Hadamard spectroscopy states that if N wavelengths are measured simultaneously under equal noise conditions, the signal-to-noise ratio improvement over sequential measurement is the square root of N, providing a factor of 5-10 improvement in SNR for 25-100 wavelength channels; this SNR improvement allows weaker spectroscopic features (such as the near-infrared absorption bands of oil, water, and gas components in live formation fluid samples) to be detected with higher confidence, improving the discrimination between oil types, quantification of gas-oil ratio, and identification of contamination by OBM filtrate that must be corrected in fluid PVT property estimation.
• Fast Walsh-Hadamard transform algorithm implementation in real-time downhole electronics demonstrates the computational advantages of the WHT over the DFT (discrete Fourier transform) for signal processing applications that must be performed inside the constrained hardware environment of a downhole tool: the butterfly computation of the WHT involves only additions and subtractions (no complex exponential multiplications as required by the FFT), making the WHT implementable in fixed-point integer arithmetic without the floating-point arithmetic units that are expensive in power consumption and silicon area; the in-place computation of the WHT (which processes an N-point sequence in log2(N) stages using only the N storage locations of the input sequence without additional memory) is well-suited to the limited memory resources of downhole tool digital signal processors; the WHT's perfect energy preservation property (the WHT of a sequence has the same total energy as the original sequence, distributed among the transform coefficients) makes it straightforward to verify the correctness of the WHT computation by checking energy conservation, providing a diagnostic capability that is valuable in the harsh and intermittently noisy downhole environment; these computational advantages make the WHT preferred over the FFT for downhole applications where the Walsh function basis (square wave functions) is equally appropriate to the measurement physics, such as in the binary coding of NMR pulse sequences where the pulses are inherently square wave modulated.
• Seismic data processing applications of the Walsh-Hadamard transform exploit the sparse representation of certain seismic signals in the Walsh domain as an alternative to frequency domain filtering for specific noise attenuation operations: seismic data typically has excellent sparse representation in the Fourier frequency domain for noise that is separated from signal in frequency content (random noise, which spreads across all frequencies while signal is concentrated in a specific band), but for non-stationary noise events (ground roll with dispersive propagation, air wave arrivals, coherent noise from surface sources) the Fourier representation is less sparse and frequency filtering less effective; Walsh domain processing of seismic data has been explored as an approach to remove certain types of coherent noise that are better separated from signal in the Walsh domain than in the frequency domain, based on the observation that some noise types are more efficiently represented by square wave basis functions than by sinusoids; while Walsh domain seismic processing has not become a mainstream industry practice (Fourier transform methods continue to dominate due to their well-developed theoretical foundations and commercial software implementations), the WHT-based seismic applications illustrate the general principle that the choice of mathematical basis for representing a signal should match the underlying physics of the signal and the noise, and that alternative transforms to the FFT may provide advantages for specific signal types in specific geological contexts.
• Error correction and data compression using Walsh-Hadamard codes in downhole telemetry systems provides reliable data transmission through the noisy communication channel of mud pulse or electromagnetic MWD telemetry, where the low data rate and high error rate of the transmission channel require efficient coding to maximize the useful information content of each transmitted bit: Hadamard codes are a class of error-correcting codes in which each transmitted codeword is a row of the Hadamard matrix (the N-by-N matrix whose entries are plus or minus 1 and whose rows are mutually orthogonal Walsh functions), providing the maximum Hamming distance (N/2) between any two codewords in a code of N codewords; this maximum Hamming distance makes Hadamard codes optimal for the binary symmetric channel (where each bit is flipped with equal probability) in terms of the minimum distance-to-length ratio, ensuring that the maximum number of errors can be detected and corrected for a given code length; in mud pulse MWD telemetry where the bit error rate can be 5-20% in noisy conditions (compared to 10^-9 in a wired data link), error correction coding is essential for reliable data transmission, and the efficient encoding and decoding of Hadamard codes using the Walsh-Hadamard transform algorithm makes them computationally practical for the processing speeds available in downhole tool electronics; the combination of Hadamard error correction with data compression (reducing the number of bits needed to represent each measurement value) and adaptive bit rate control (adjusting the telemetry rate based on real-time noise estimation) provides the multi-layer communication strategy used in modern MWD systems to maximize the useful data rate delivered to the surface despite the inherent limitations of the mud column as a communication medium.