The error analysis and the design of a 16-bit binary digital-to-analog converter are discussed. This method of conversion switches binary weighted constant current sources into a summing junction, in this case the yoke of a CRT. A conversion linearity of 0.00025% and a stability of 0.0004%/h were measured for a full scale current of 1 A showing that a converter linear to better than 16 bits can be constructed requiring no trimming of components.