Dynamic goniometers (DGs) are used to calibrate optical angle encoders with high accuracy over a wide range of rotation speeds. The dynamic mode of operation of the goniometer causes an additional error. A current task is to research this dynamic error. The systematic component of the DG’s error can be eliminated by introducing corrections to the measurement result. Thus the main factor limiting the accuracy is a random error, including non-stationary components. For its analysis, the Allan variation method, Fourier, and wavelet analysis were used. Experimental studies of the DG with ball bearing and reference optical angle encoder at various rotation speeds have been carried out. Arrays of random variables characterizing the DG’s random error were studied; they are non-stationary in terms of the average value. Using the Allan variation, the sinusoidal and white noise were determined. Fourier analysis revealed the presence of frequency errors in the random error, which are caused by the processes of movement of the bearing components, as well as a defect in the separator. Wavelet analysis made it possible to clarify the nature of the processes, to trace the change in the frequency composition over time. When constructing a DG for calibrating encoders, special attention should be paid to the quality of ball bearings. It has been determined that dynamic errors make a significant impact on the measurement error. The ball bearing, namely the defect of the bearing cage, is the major reason that causes non-stationary effects at low frequencies. As studies have shown, ball bearings increase the random error of the DG by an order of magnitude or more compared to its potential accuracy, which is significantly <10−2″.