I apply the thermodynamics of radiation to Dyson spheres as machines that do work or computation, and examine their observational consequences. I identify four properties of Dyson spheres that complicate typical analyses: globally, they may do no work in the usual sense; they use radiation as the source and sink of energy; they accept radiation from a limited range of solid angle; and they conserve energy flux globally. I consider three kinds of activities: computation at the Landauer limit; dissipative activities, in which the energy of a sphere's activities cascades into waste heat, as for a biosphere; and "traditional" work that leaves the sphere, such as radio emission. I apply the Landsberg formalism to derive efficiency limits in all 3 cases, and show that optical circulators provide an "existence proof" that greatly simplifies the problem and allows the Landsberg limit to be plausibly approached. I find that for computation and traditional work, there is little to no advantage to nesting shells (as in a "Matrioshka Brain"); that the optimal use of mass is generally to make very small and hot Dyson spheres; that for "complete" Dyson spheres we expect optical depths of several; and that in all cases the Landsberg limit corresponds to a form of the Carnot limit. I explore how these conclusions might change in the face of complications such as the sphere having practical efficiencies below the Landsberg limit (using the endoreversible limit as an example); no use of optical circulators; and swarms of materials instead of shells.

To join please click on the Zoom link: https://psu.zoom.us/j/92466964691