A differential scanning calorimeter was calibrated by Joule effect and by measuring the temperature of fusion of several pure (99.999%) metals. Repeatability and reproducibility tests were performed to access the accuracy of the instrument, the stability of the calibration constant and the correction to the temperatures of fusion was obtained experimentally. The results demonstrate that this calorimeter is capable of producing data of enthalpies of fusion with an uncertainty of 1.5%, temperatures of fusion between 0.1 and 0.2 K, and heat capacities with an uncertainty of 1.5%, all at a 95% confidence level. In order to calculate the uncertainty of the measurements performed, a discussion about the traceability in thermal measurements and its key elements (relation to 'stated references', unbroken chain of comparisons and stated uncertainties of those references) is done. A possible structure of the traceability chain in DSC and DTA measurements is presented.