Comparative evaluation of computer methods for calculating the best-fit sinusoid to the digital record of a high-purity sine wave

Abstract

A comparative evaluation of programs used in the “equivalent bit” test is described. Four methods of determining the best-fit sine wave were tested using computer generated data contaminated to varying degrees with white noise. These data were supplemented with actual records of sine waves obtained using fast digitizers with 6- to 12-bit resolution and record lengths of between 512 to 16 000 words. The programs were evaluated for relative accuracy, absolute accuracy, and efficiency, both in terms of how often each method converged and also of how much computer time each method took to converge.