I have noticed that the temperature profile between a chip and a pcb changes as a function of orientation of the pcb. Unfortunately, we do not have the luxury of having all of our products pcbs in a convenient orientation. If one pcb is orthogonal chances are another one in the same product will not be. The issue I see is that for a 'flat' board the heat flow from a chip into the pcb (pcb is defined having different thermal conductivity across board vs through board) is smooth for the flat case. However for an angled board, the transition is abrupt. I expect the transition to be smooth and then drop off rapidly thru the board but slowly along the pcb.
Things I have verified :
(1) Local mesh density is not a factor
(2) Convergence criteria not a factor
(3) Thermal conductivity is set to 'Orthotropic' with the anisotropic plane set to the face of the pcb (pcb material is set with X-direction k_x is << k_y, k_z.
(4) Confirmed via interference detection (coincident surfaces are interference) that all intended chip surfaces are in thermal contact with pcb.
Does this sound like an issue with the software? Is there a setting I need to verify? Anyone else seen this issue?
Split chip above has heat applied to mid-plane surface. (I expect the thermal gradient to be smooth from chip to pcb)