The third sub-division relates to what we have termed the Thermal Limit of Output, that is, the maximum output with safe heating. It can be fairly said that while the theory of all the losses in a commutating dynamo are understood, yet, with the exception Of the C2 R losses, it is Still a matter of practical experience to determine what relation the actual losses bear to what may be termed the predicted losses. It is invariably found that the iron losses are in excess of those which may be predicted from the tests made upon the material before construction. The hysteresis loss in the armature core is generally found to be greater, owing to the mechanical processes to which the material in the core has to be sub jected during the process of construction. Owing, probably, in a large measure to a Species of side magnetisation, the eddy-current loss is found to be greater than is indicated by calculations based upon the assumption of a distribution of magnetic lines parallel to the plane Of the laminations. If the armature conductors are solid, the losses therein by foucault currents may often be considerable, even in projection type armatures, especially when the projections are run at high densities. Under load losses, not including friction, there have to be considered the foucault current loss in the conductors due to distortion, and the increased loss in the armature projections from hysteresis and eddy currents likewise due thereto. There is also the loss brought about by the reversal of the current in the armature coil under commutation. It is apparent, therefore.