The equations for SFDR rely on the assumption that rounding errors from the finite precision of phase and amplitude are incoherent. This assumption is violated for values of phase increment that are not mutually prime with the weight of the phase accumulator. The anomalies, such as spurii, are more obvious for larger common factors between the phase increment value and the weight of the accumulator (2Phase Width). This is because such values might not access every location in the SIN/COS LUT, so the rounding errors are not randomly spread. To avoid this, do not use values of output frequency that are simple rational fractions of the frequency per channel, Fs, such as 3/8 and 1/64.