The design of the joints in a power transformer core has a marked influence on the efficiency of the core as a whole. Two experimental cores have been built such that the lengths of their 45° mitered overlaps could be varied from 0 to 2.0 cm. The larger core, with 1.0 m limbs, was used for investigating flux and loss variations and the second, smaller core was used to study how the results might be affected by changes in dimensions. For the larger core, the power loss was measured in two ways for a range of overlaps: first, using the localized power loss technique, and then by measuring the total power loss with a precision wattmeter. In both cases, a minimum power loss was found when the core was built with an overlap of 0.5 cm. The overlap length was varied in the smaller core and again a minimum power loss was found, but with a 1.0 cm overlap. A change in loss of over 20% was found for the range of overlap lengths used, and the optimum overlap length was independent of flux density over the range from 1.0 T to 1.8 T. The special flux distribution was determined from an array of search coils. The variations in flux distribution enables a qualitative explanation of the occurrence of the minimum in power loss.