Existing methods for improving the accuracy of digital-to-analog converters

Abstract

The performance of digital-to-analog converters is principally limited by errors in the output voltage levels. Such errors are known as element mismatch and are quantified by the integral non-linearity. Element mismatch limits the achievable accuracy and resolution in high-precision applications as it causes gain and offset errors, as well as harmonic distortion. In this article, five existing methods for mitigating the effects of element mismatch are compared: physical level calibration, dynamic element matching, noise-shaping with digital calibration, large periodic high-frequency dithering, and large stochastic high-pass dithering. These methods are suitable for improving accuracy when using digital-to-analog converters that use multiple discrete output levels to reconstruct time-varying signals. The methods improve linearity and therefore reduce harmonic distortion and can be retrofitted to existing systems with minor hardware variations. The performance of each method is compared theoretically and confirmed by simulations and experiments. Experimental results demonstrate that three of the five methods provide significant improvements in the resolution and accuracy when applied to a general-purpose digital-to-analog converter. As such, these methods can directly improve performance in a wide range of applications including nanopositioning, metrology, and optics.

Large-Amplitude Dithering Mitigates Glitches in Digital-to-Analogue Converters

“…The ADC input was filtered by the anti-aliasing filter (15), and a second-order low-pass Butterworth filter with cut-off frequency at 10 kHz was used to filter the measured response in order to achieve the desired averaging of the glitches.…”

“…with k g1 = 5 · 10 −5 and w(t) = L −1 [W ](t), with W (s) given by (15). It can be seen that the glitch is converted from a large amplitude disturbance with a short duration, to a smaller amplitude disturbance with a longer duration, and the maximum slope and value of the averaged glitch response is affected by A, which can be considered a design parameter.…”

“…with k + g1 = −5.0 · 10 −5 , k − g1 = 5.1 · 10 −5 , and w(t) = L −1 [W ](t), where W (s) is as given by (15). Again the glitch is converted from a large amplitude disturbance with a short duration, to a smaller amplitude disturbance with a longer duration.…”

“…As described in Sec. II-C, the USB-6289 contains two first-order passive low-pass filters (15), with f c = 62.5 kHz. Fig.…”

“…Furthermore, in segmented topologies [34], using e.g. binary weighing, DEM has demonstrated significant deterioration of performance [15]. Hence, the applicability of these methods is limited.…”