People complain about fan noise in their computers. The same applies to active cooling on high-power LED arrays. It's not CFM (cubic feet per minute) that counts, but the linear air velocity at the surface of the hot spot.
There is a known relationship between air flow velocity, LFM (linear feet per minute) at the surface over a passive heat radiator, and temperature rise. On the other hand, volumetric flow (CFM) is strongly dependent on the design of the air flow. Laminar vs. turbulent -- which is best?
It is essential to minimize the volume of air flow for efficacy of the cooling while providing the maximum linear velocity at the surface. This determines the reduction of thermal resistance. There is an optimal point where the incremental power added to increase velocity is equal to the power removed. This varies with the power dissipation of a device over the power of a fan, and it can be compared with the surface area exposed over the area of the plenum of the flow area. A higher ratio is better.
Typical systems try to achieve 0.5m/s to 5m/s velocity, depending on the power needed where the air pump heat is included with the device heat. Beyond this, it becomes less efficient, and CFM is irrelevant, whereas the benchmark for linear velocity feet per minute or meters per second is relevant.
I used this approach for designing a turbulent flow plenum chamber with a temperature-controlled, three-terminal regulator driving fan and verified it with smoke speed measurements. The results were outstanding, with a 10C rise for active cooling, versus a 50C rise for passive. So the answer to the initial question is: Turbulent flow is the best design to collect the heat, followed by laminar flow to remove it from the enclosed volume. The technical reason is that the eddy currents create higher vortex velocity at the surface, and the air flow velocity is what reduces thermal resistance, not CFM.
This entry was submitted by D. Anthony Stewart and edited by Rob Spiegel.
Tell us your experience in solving a knotty engineering problem. Send examples to Rob Spiegel for Sherlock Ohms.