Thermal Design paper

[Mosaic]

**broken link removed**

 

[Tony Stewart]

Pretty pictures and pretty high temperatures.

Although there was no discussion on convection , conduction or forced air flow.

When I designed a 1U high rack for AVAYA ( nee Lucent) ,while Design Services Mgr for C-MAC in Winnipeg, I had a thermal design problem that expanded my awareness of solutions. I chose to power this DS1 VOIP power over ethernet distribution box with a 180W open frame PSU's from Lambda and PowerOne and the latter unit was slightly cheaper so Purchasing went with this for 10k units/yr. The problem was 2 forced air cooling fans 1.5" high did very little to cool the hottest parts. After performing a smoke air flow test at nite, I found the air flow basically followed the path of least resistance and flew right over the top although the CFM volume flow rate was high. The Power transformer was the hottest part. The Mylar lid was cut from a bulk large roll of plastic with a band saw and blade to length with a hinge jig to fold the flap. in seconds.

Overall the design was met the cost target and 1st prototype with sheet metal , PCB and cable/connectors was delivered 1 week early in 7 weeks instead of 8 from project start date.

Also I needed a solution for UL's coke spill and sledge hammer test so the top lid could not be a weak open vented lid.

Then I had a Eureka moment, if I lower the lid with a spoiler flap on the intake the hotspots at max load, reduced from 40'C to 10'C with the fans still at full speed. So I decided to custom make a mylar lid to serve both as an insulator from the crush test and reduce the gap between the lid and open frame with a 45 deg flap bent down in the intake. I imagined this would setup Eddy currents and create a turbulent air flow rather than a Laminar Air Flow over the top.

THis something they don't take into consideration on PC case designs. So all the CPU heat gets circulated in the box and cooling anything onboard is very ineffective. The trick is to generate about 1~2m/s linear air speed at the surface, then the CFM volume does not matter. This improves cooling by 5x over laminar flow far above the hotspot.

I wanted to servo control the fan voltage based on temperature at 50~55'C for $2 and could not find any decent solutions. So I modified an LM317 with a transistor driver to adjust the voltage and amplify thermistor temp voltage with a Pot to set the threshold for driving the transistor. The LDO was a TO-220 which disspated as much as the fans at half speed, but since they only used a few watts each, that was ok. I then designed a waste FR4 tab with a hole fabricated by the board shop as the washer to press the insulated TO-220 to the inside fan bracket, so it also got forced air cooling. The FR4 washer only cost a penny in bulk with a screw, so the total cost was around $1. THe Thermistor was epoxied onto the power transformer.

You can buy these $1k boxes which cost $250 to manufacture for $35 now on **broken link removed**

 

[Tony Stewart]

If you buy one , look for my innovative Origami 50wire ribbon cable fit in this case to the front cable pair of connectors.