When I was design manager for C-MAC in Winnipeg, I won the first outsourced design from Avaya (nee Lucent). We had to redesign a power distribution rack for 16 VoIP (voice over Internet protocol) Lucent phones on T1 RJ45 ports. It was a simple design requirement that could have been written on a paper napkin.
a) Cut the size of the rack in half, from a 2U (rack unit) high to a 1U high 19" rack.
b) Select a 180W 48V DC for internal power distribution.
c) Make 10,000 units for half the price of existing 2U unit, or roughly $250 per rack.
d) Distribute 16x T1 (1.544Mbps) signals from two connectors to 16 VOIP RJ45 connections.
e) Pass Telco crosstalk, UL safety, and FCC requirements.
f) Deliver the first prototype in eight weeks.
g) Then ship 1,000 units a month after approval.
Well, one of the UL requirements (that snuck in later) prevented the use of vertical cooling through the top lid, because UL had some special tests called the "Sledge Hammer" and "Spilled Coke" tests. It had to survive them without a safety issue of fire or sparks. Mind you, if one did both tests at the same time, the Coke might put out the fire.
So the internal open-framed PSU (power supply unit) needed extra insulation and cooling. No problem. Twin 25mm fans should do the trick. Buzz said the sheet metal at max RPM, and a 45°C temperature rise said the thermocouple on the hot spot was not good enough.
So I decided on a goal of spending $2 for a design solution with my BOM (bill of materials) budget, product cost, and margins. That's all I wanted to spend to achieve a 15°C temp rise above ambient for worst-case hot spots in the PSU.
But how? I already had laminar flow from one side to the other in the open frame PSU. It was a nice plenum shape to match the PSU "U frame" with adequate vent size, so I should get good CFM (cubic feet per minute). Whats the deal?
So I looked for more powerful 25mm fans and considered axial and turbine types. Bearing types last longer in heat but are noisier, and the 1U rack design acted like a piano sounding board. Nothing seemed to improve it much. The conducted heat removal was as good as it was going to get.