This thesis first explains the eddy current and the phenomenon of skin effect, where the resultant flux flows near the surface of the metal. A new flux direction perspective is created for steel laminations, from which derivations of the eddy current resistance and power losses in different directions are developed assuming uniform flux conditions. The developed method compares with a proposed theory through experimental data. The results from the comparison support the validity of the developed derivations. Two uniform flux generators and their billets construction are introduced. The power loss between two cubic billets with different orientations is compared. A Finite Element Analysis (FEA) program is used to show the difference between lamination alignments. To prove the validity of the developed theory, two experiments were performed using two different electroheating apparatus. The results give scale factors from which the theoretical values can be matched to the experimental ones. Due to the poorer construction of the first apparatus, the scale factor of measured to computed losses is 1.15. The scale factor for the second apparatus can be taken as unity, revealing a good match between theory and measurements. After verification of the developed equations for uniform flux experiments, the focus of the eddy current loss calculation turned to partial core transformers. The flux background of a cubical core is reviewed. Three key factors ( L', Kec and βa) are introduced into the eddy current power loss model. L' is a length which indicates the region of the flux spreading at the ends of the core. Kec as a ratio indicates how much of the main flux spreads at the ends of the core. βa is the ratio of the winding axial length and winding thickness. Using simulations from the Finite Element Analysis (FEA) program MagNet, a partial core side view with the flux distribution and flux density from two orthogonal angles is created. A flux linkage comparison between the experimental results and the returned values from MagNet verifies the high accuracy of the flux plot in MagNet. The eddy current power loss model is then built up with equations. The relationships amongst the three key factors are studied and confirmed using the experimental results. Normally, a partial core transformer uses a cylindrical partial core rather than a cubical partial core, to reduce the amount of winding material. Therefore, a further goal was to prove the developed model for cylindrical partial core transformers. The construction differences between the cubical and cylindrical core is discussed. The orthogonal flux assumptions for the cylindrical core in two directions are reviewed. The flux penetration between two adjacent blocks is considered and explained. The mathematical core loss model is created for a cylindrical core composing by ten blocks. Three tests were performed using the developed core loss model. The results visualize the power loss from the core by its temperature distribution, and consequently prove the validity of the developed core loss model. An eddy current loss comparison and the discussion are made between the previous method and the developed method. Overall, the results confirm a significant improvement using the developed core loss model, and a generic form of the partial core can be used for designing future models of partial core transformers which have a stacking factor greater than 0.96.