Signal-to-noise Ratio Enhancement Research Articles

A digital hardware correlation system for fast ultrasonic data acquisition in peak power limited applications

The theoretical and experimental evaluation of a digital hardware correlation system for low-power ultrasonic applications is presented. The system, which incorporates dual Golay code matched filtering, is capable of 20-MHz processing rates with a signal-to-noise-ratio enhancement (SNRE) of 23 dB over a conventional pulse-echo system operating at the same peak power levels. The effects of digitization have been investigated, and a TTL (transistor transistor logic)-based hardware correlator has been developed. For many applications, low-voltage driving followed by differential detection is sufficient, permitting the system to be used in a number of power-limited environments. Sample tests conducted on three different transducers have demonstrated that the system is operational over a wide variety of probe configurations.

Optimum SNR enhancement of narrow‐band signals in surface reverberation

Theoretical performance capabilities of the optimum filter for maximization of signal‐to‐noise ratio (SNR) and a suboptimum approximation, easy to realize in practice, are compared to the conventional matched filter (designed for white noise only) in the case of low Doppler signals masked by surface reverberation. Spectrum models for surface backscattering as a function of local environmental parameters are based on recent results describing scattering from a subsurface bubble layer. Significant improvements in SNR in the low Doppler region as well as a dramatic dependence of performance on windspeed and direction, and the ratio of reverberation to uncorrelated noise power are predicted. Results serve as average performance limits to be compared to those obtained by algorithmic implementations of optimum and adaptive filters.

Precision in measuring resonance spectra

The precision with which the ith parameter of a resonance spectrum profile can be measured by least-squares estimation is shown to be c iK 1 2 R , where c i depends on the line shape, K is the number of observed points per profile linewidth, and R is the profile signal-to-noise ratio (SNR). The maximum useful value of K is twice the product of spectrometer output bandwidth and the time taken to sweep through a linewidth. The product c i R is invariant under linear operations; consequently data processing such as numerical filtering, SNR enhancement, smoothing, etc., does not improve the potentially attainable measurement precision.