The paper presents a conception of uncertainty calculation of a result obtained in a direct measurement realized in conditions described by random errors. The conception basis on the error denition being an eect of analysis of a quantization process and, rst of all, it permits to determine uncertainty of a single measurement result in measuring and control systems processing signals varying in time. Division of the errors into two types A and B permits elaboration of such a procedure which enables uncertainty calculation for an average value of a series of measurements in the way close to this one proposed by GUM and widely discussed in last years. Theoretical considerations are illustrated by examples showing practical properties of the presented uncertainty calculation procedures.