Abstract

A multi-tone measurement can be used to characterize the noise and distortion of a device or a system. Such assessment can be accomplished by a least-squares sinusoidal fit, which is widely analyzed in the prior art. The IEEE standard sinusoidal least-squares fit (sinefit) can be quite easily expanded to multiple tones, real or complex. From a designer point-of-view, the challenge is to generate a proper stimulus. The design questions in such case are the following: ”How do I design the tone spacing?”, ”What is the minimum record length of the stimulus with n tones?”, ”Can I use the 3-parameter fit (known frequencies)?”, ”How many iterations do I need in the 4-parameter fit (unknown frequencies)?” and ”How accurate the initial guess of the frequencies have to be for the 4-parameter fit?”. This study strives to answer the aforementioned questions with reflections to the analytical results from prior art.