This work consists of a theoretical part and a practical part conducted in the Laboratory for precise measurements of length (LFSB) of the Faculty of Mechanical Engineering and Naval Architecture. The theoretical part of the work deals with the measurement of the deviation of roudness, the methods used in measuring roundness, the uncertainty of measurement and the methods used to evaluate the measurement uncertainty. All measurements were carried out on the device Mahr MMQ3, the method of testing roundness with an external reference measurement with rotating table. A method used is the numerical method of least squares circular (LSC) and it is outlined in detail in the work in order to explain in detail to the reader. The methods used for the determination of measurement uncertaintie are GUM method and Monte Carlo method. GUM method is a method that is commonly used for the determination of measurement uncertaintie and with it first need to determine a mathematical model, then choose which component comprises measuring and ultimately determine the measurement uncertainty. Monte Carlo method is a statistical simulation that is done by computer and that uses random variables in its calculation. The practical part of this work consists of calibration devices for measuring roundness Mahr MMQ3. Meter Calibration-Round consisted of calibration of probe on ULM uniaxial device, and inaccuracy of determining the rotation of the rotary table by measuring hemispherical RET 227-509. Determining probe correction is carried out on ULM device in a way that probe is placed on the device to try to avoid a linear error of probe. Then, the three repeated measurements is carried out for all measuring ranges (3 microns, 10 microns, 30 microns, 100 microns), and the necessary correction is entered into the software. After the correction was carried out, the calibration of probe is performed with 3 repeated measurements for all of these ranges. Calibration of the rotating table was conducted on basis of 30 repeated measurements using the method of least squares. For the process of calibration devices for measuring roundness the calculation for estimate measurement uncertainty is conducted. A mathematical model of measurements was made and there are defined and estimated impacts of the dominant component of the measurement uncertainty.