Open AccessCalibration of pipeline ADC with pruned Volterra kernels
Author(s) -
Centurelli F.,
Monsurrò P.,
Rosato F.,
Ruscio D.,
Trifiletti A.
Publication year - 2016
Publication title -
electronics letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.375
H-Index - 146
ISSN - 1350-911X
DOI - 10.1049/el.2016.1601
Subject(s) - volterra series , calibration , pipeline (software) , effective number of bits , nyquist–shannon sampling theorem , algorithm , kernel (algebra) , computer science , sampling (signal processing) , mathematics , linearity , cmos , electronic engineering , nonlinear system , statistics , filter (signal processing) , engineering , physics , programming language , quantum mechanics , combinatorics , computer vision
A Volterra model is used to calibrate a pipeline ADC simulated in Cadence Virtuoso using the STMicroelectronics CMOS 45 nm process. The ADC was designed to work at 50 MSps, but it is simulated at up to 125 MSps, proving that calibration using a Volterra model can significantly increase sampling frequency. Equivalent number of bits (ENOB) improves by 1–2.5 bits (6–15 dB) with 37–101 model parameters. The complexity of the calibration algorithm is reduced using different lengths for each Volterra kernels and performing iterative pruning. System identification is performed by least squares techniques with a set of sinusoids at different frequencies spanning the whole Nyquist band. A comparison with simplified Volterra models proposed in the literature shows better performance for the pruned Volterra model with comparable complexity, improving linearity by as much as 1.5 bits more than the other techniques.