Thorp's burner produces a cylindrical flame, like that of the Argand, but without the aid of a glass chimney which is a necessary adjunct to the latter burner. By means of a deflector on the inner side of the flame, the latter is made to curve outwards and assume a somewhat convex form, so as to obviate the shadow which otherwise would be cast by the gas chamber at the bottom of the burner. Above the flame is a cylindrical chimney, divided by a vertical partition into two concentric chambers, which are intersected by a series of metal gills, or projections, continued through both chambers. The outer chamber is for conveying away the products of combustion; the inner one for the passage of air to feed the flame; while down the centre of the inner chamber there passes a tube conveying the gas to the point of ignition. The hot products of combustion pass up from the flame through the outer chamber, and give up the greater portion of their heat to the projections; by which it is conducted into the inner chamber, and transferred to the incoming air. A common imperfection of regenerative burners is that, in consequence of the diminished rate at which the gas flows through the burner when expanded by heat, when starting the burner the gas must be only partially turned on, and the quantity gradually increased as the burner becomes heated; thus necessitating considerable attention. To prevent the need for this attention, there is in Thorp's burner an ingenious contrivance for automatically regulating the quantity of gas admitted to the flame. The central gas-tube, which is referred to above, contains a brass rod, fixed at one end, and at the other connected to a valve controlling the quantity of gas that enters the tube. At first, when the gas is lighted, this valve is almost closed; but as the rod becomes heated it elongates, gradually opening the valve until the full quantity of gas is admitted which the burner is intended to consume. At the Stockport exhibition, Thorp's burner was tested with the following results, as recorded in the Judges' report. After it had burned about two hours, "it gave an illuminating power of 183 standard candles, while burning 27 cubic feet of gas per hour (equal to 6·77 standard candles per cubic foot), with gas of 3·5 candles per cubic foot.... In another experiment with the same quality of gas, after burning half an hour it yielded, under similar conditions, 154 candles with a consumption of 25·29 cubic feet per hour, which gave an illuminating power of 6·02 candles per cubic foot." [Sidenote: Clark's regenerative burner.]