On one way to improve accuracy and on design of an optimal structure for a digital filter

The paper is concerned with a method to reduce the resultant rounding off error for a digital filter by changing the sequence of the rounding off and addition and by selecting a special sequence of elementary elements of the filter.