| Even tightly packed control rooms show sufficient cooling for each cabinet. |
| The Honeywell Experion R300 system has over 30 process-control I/O and controller modules stacked in columns. Stacking modules at an 18° tilt directs cooling air from right to left across each module and upward toward the exhaust vent. Heated air bypasses the modules' inlets. |
| Coolit analysis shows that tilting the modul es delivers cool air to each one. |
Simulation takes little investment and turnaround is faster than with prototypes. For example, take the Honeywell Experion R300 process-control system. Engineers developing it saw that vertically stacking I/O and controller modules would let it hold the most modules with the smallest cabinet footprint.
But the vertical arrangement created a heat problem. Cool air entering the bottom of the cabinet would get progressively warmer as it drifted upward from one module to the next. By the time air reached the upper portion of the stack, it would be hot enough to overheat and fail the modules.
A simple solution, Honeywell engineers thought, would be to tilt modules so each would receive only unheated air. But would it work? Physical testing could cost thousands, so they turned to simulation.
Experion contains over 30 modules stacked in three columns. Design specs call for ambient air at no more than 50°C. Many components inside the modules are rated at 70°C. To keep from exceeding that limit, air entering any module could be no warmer than 60°C. Experience suggested that without a design change, exiting air temperatures would approach 80°C.
The traditional fix adds fans and vents to increase airflow. But this was not an option. "Even marketing people know that fans reduce reliability and pull in contaminants," says Honeywell Development Engineer Rod Boer.
"As a result, we use only two 4-in. muffin fans atop each cabinet." Adding more side vents was also not acceptable because cabinets are often joined side by side.
Thermal specialists reasoned that if each module were tilted at the correct angle, room air could enter the bottom right side of each module, flow across it and exit the top left side. Each module, regardless of position in the stack, would be cooled by unheated air entering from the bottom. Two potential problems remained.
"We were concerned that heated air leaving one column might eventually reach the column to its left," says Boer. "We didn't have the time to physically build a model to prove the concept, so the only way to meet our deadline was to simulate the system using CFD."
Boer's design team tested the concept using Coolit CFD thermal and flow-analysis software from Daat Research Corp., Lebanon, N.H. (daat.com). Analysis predicted that slanting the modules 18° and stacked with about 0.25-in. vertical spacing would let cooling air reach each of the 36 modules. And air from one column would not mix with air entering another, and the modules would not exceed the 60°C limit.
The second problem presented a greater challenge. One I/O module with 16 field-effect transistors (FETs) dissipated almost twice as much heat as others. Heat from the FETs moved along copper traces across the length of the circuit board to the lowertemperaturerated devices, reducing their reliability.
Engineering thought it could fix the problem by spreading the FETs evenly over the board surface, an approach that would minimize the heat at any one spot. A simulation, however, predicted that low-temperature devices would still see excessive heat.
Another idea was to thermally isolate the FETs from the rest of the board. The design packed all FETs on one-half of the board, lower-temperature-rated devices on the other, and placed a thermal barrier between the two. Only essential traces crossed the barrier.
It worked. Heat from the FETs was significantly reduced. Thermal analysis verified that components on both sides of the barrier remained within operating limits.
"Solving these problems with thermal simulation saved us at least six months," says Boer. "We make hundreds of boards when checking a design, and it takes months to get them fabricated and tested. It would have taken another six months to determine there was a problem and design a solution."