The paper describes a new method for measurement and correction of errors associated with DACs. Very small errors of a few μV can be measured and corrected. The method enables us to correct commercially available monolithic DACs, so that they can be used in nuclear ADCs where high integral accuracy and low differential nonlinearity (DNL) are required. This cuts down the circuit size and cost for DACs applied in nuclear ADCs. The paper proposes a “theorem of DAC code correction” which is based on the relationship existing between the errors associated with different bits of a DAC. Complete mathematical proof is given to establish the correctness of the method. The “theorem” can be implemented if the correct value of any bit or of the full scale voltage is known. The method also provides a check on whether errors are correctly measured and whether they are perfectly compensated. The illustration of the relationship between errors associated with different bits also reveals that the conventional ways of adjusting DNL for 16-bit and 18-bit (0.5 LSB) DACs followed by commercial manufacturers are not accurate. These corrections can be implemented more effectively if the procedure described is followed.