High accuracy is achieved by employing, in conventional DAC architectures, very accurate current sources. To create these high-accuracy current sources, the outputs of several smaller, less accurate, nominally equal current sources are summed. A procedure is taught for selecting the number of current sources to achieve an arbitrary degree of accuracy with a desired level of confidence. Assuming the current sources are taken from a population whose output currents deviate from a design value according to a normal distribution, the minimum number of constituent current sources, n, required to provide an accurate total current is given by the formula n=(Z.sigma./E).sup.2, where Z is a number which corresponds to the probability that the output will fall within an error band E (i.e., a predetermined accuracy level) with a predetermined level of confidence, .sigma. is the standard deviation of the population.