Less Dithering for Higher Accuracy in Stochastic Harmonic Instruments

This paper examines improvements in the design of stochastic harmonic instruments, which utilize dithering of input signals to enhance the precision of analog-to-digital conversion. Their original design employs a set of uncorrelated random dithering signals to ensure zero summing of their numerical integrals. Unfortunately, these random signals - some of which are embedded into the instrument's memory - are also the main source of the inherent measurement inaccuracy in practical implementations. To address this issue, we explore two potential modifications aimed at improving accuracy while preserving the benefits of high precision: (1) restricting dithering to low-resolution cases, and (2) eliminating dithering of the second signal. The proposed modifications are analyzed theoretically and validated through simulations across a representative space of measurement signals.