We're designing a 32-channel SSR board, in a 6U VME Eurocard format. The relays will be home-made, a pair of self-protecting mosfets back-to-back. These things need too much gate drive for a PV optocoupler to provide, so each gate driver will be a tiny flyback power supply.
The switches will net be a tad under 1 ohm ON, so our ability to dump heat will determinate the current rating. 1 amp per switch might sound nice, which would dissipate 32 watts spread over most of the surface of the board, a little over 1 square inch per SSR maybe. I did some googling on this and, as usual, got a lot of blather without hard numbers.
So I made a fake VME board:
which is a 5x5 array of 10 ohm, 1206 resistors. The board is about 59 square inches.
Standing up vertically on my test bench, with no air flow, and 20 watts dissipated, I got a temp rise of 25C at the bottom of the board,
38 in the middle, 40 at the top, presumably influenced by convection. Call that 38C rise, 1.9 K/w, which is 110 K/W per square inch.If I lay the board flat on a terrycloth wash cloth, my whiteboard eraser, it rises to 160 K/W per square inch in the middle, a bit less at the corners.
In a VME crate with fans, the cooling varies quite a bit in different slots; a simple tray of muffin fans tends to be very non-uniform. Theta per square inch in three slots was 10, 11, and 16 K/W. With fans switched off, I got 119.
So it looks fine to dissipate 32 watts, in a crate with fans. We will include a temperature sensor and maybe issue a warning, then shut down, if the board gets too hot, like if the fans quit.