An Order-Statistics Based Matching Strategy for Circuit Components in Data Converters

Random mismatch errors in matching-critical circuit components cause variability in component parameters, which in turn degrade linearity performance and parametric yield of data converters realized by these components. Utilizing results from the area of order statistics, this paper introduces a novel random mismatch compensation theory called ordered element matching. By pairing a small parameter component with a large one, it effectively reduces the standard deviation of the mismatch errors by a factor of at least 6.5 in a reasonably sized component population. An order-statistics based outlier elimination strategy further improves the standard deviation and more importantly reduces the differential nonlinearity by a factor of 2 or more. Building on these, a new technique called complete-folding is developed, which selectively converts a unary-weighted array to a binary-weighted array and consequently reduces the differential nonlinearity and integral nonlinearity by another factor of more than 3 and 8, respectively. Monte Carlo simulations are performed in a high-resolution data converter design to quantitatively compare the matching performance achieved by complete-folding technique with state of the art. Significant improvements on linearity performance and design cost are obtained.