In some converter architectures, unary digital-to-analog (DAC) converter elements generate an analog output which represents the digital input signal. Thermometer codes trigger an appropriate number of DAC elements to generate the analog output. The DAC elements are not all perfectly weighted, and mismatch shaping is often used to dynamically equalize the usage of each DAC element during data conversion to average out the mismatches. Unfortunately, mismatch shaping adds additional switching and can worsen the effect of switching errors. Switching errors which are non-linearly dependent on the input causes a second order distortion if the sum of the switching errors corresponding to a set of DAC elements is not zero. Prior to data conversion, calibration can select a subset of DAC elements having a lesser sum of switching errors for data conversion. Other (redundant) DAC elements are not used at all or shut off permanently.