Teradyne Corp., the world’s leading ATE equipment supplier, has long been using
CFD (computational fluid dynamics) software to model the temperature distribution and
cooling air movement in its new system designs. The 2 and 3 D thermal simulations enable
designers to zero in on the best cooling approaches without physically testing each
possible option.
When a problem developed in a subsystem already in production, the company tapped the
same thermal simulation tool to verify the impact of an external design change on the
subsystem’s internal temperature without having to tear the box apart to find out
what was going on inside. The results showed that the simulation could accurately predict
internal temperatures faster and less expensively than physical testing.
The problem had surfaced in the field when service personnel were installing a
replacement power supply in one of the company’s huge memory testers. The heavy and
bulky supply, measuring 6 inch wide by 6 inches high by 24 inches long and weighing almost
30 lbs, rack-mounts in a frame that holds large card cages filled with PC boards, fan
trays, and heat exchangers. With so much gear in the rack, replacing the power supply is a
‘blind’ operation; the service person cannot see the insertion area.
In about 15% of the cases, as the service person plugged in the power supply, the rear
contact fingers did not align with the mating connector and would get bent under the
insertion force. Each supply was worth $3000, so damage was costly in terms of replacement
parts, as well as service personnel time. More importantly, it inconvenienced the user,
who had to wait for a replacement supply.
Power supply engineers devised a spacer that would mount between each power supply
finger. The relatively inexpensive spacer increased finger rigidity and prevented bending
under the insertion force.
But there was one potential ‘gotcha’. While the spacer added stiffness, it
also partially blocked the air flow exiting the power supply. This could potentially mean
serious health problems for the 550 watt power supply, as well as disruption of the
cooling pattern downstream.
The product was in high demand, both for new systems and as replacement parts, so there
was pressure to get an accurate answer and get it quickly. Wind tunnel testing easily
verified that the air temperature at various distances from the power supply exit fell
within the allowable limits. But how does one find out what is going on inside a product
without getting inside; i.e. building a time consuming and costly thermal model?
The solution was to create a thermal simulation of the internal conditions using a CFD
software package, named Coolit from Daat Research Corp., Hanover, NH. Coolit had been
chosen because it is exceptionally easy to use, yields answers quickly and is tuned
specifically for electronics applications.
To build the simulation, the product was ‘sketched’ on the computer screen
using shapes from the extensive Coolit built-in library of electronic devices, fans, heat
exchangers, heat sinks, etc. Once the geometry is specified, the user taps another on-line
library to identify the material properties, such as the thermal conductivity for heat
conducting components and the fan curve.
Material properties come in an array of incompatible units; pressure drop in inches of
water, fan curves in cfm, and thermal dissipation in watts. Trying to convert from one to
another can be a nightmare, but Coolit allows the user to mix and match units. The
software performs all the necessary conversions, thereby eliminating what could be a major
source of error.
To solve the simulation problem, Coolit automatically divides the problem into a
contiguous set of grid cells (finite volumes) and calculates its way from cell to cell.
Once the calculations are completed, the user can display the results, zoom in or out on
the model, rotate it and view cross-sections of the model.
Temperature distribution patterns are color coded making hot spots glaringly obvious,
while velocity vectors indicate the direction and speed of air movement through the
volume. Because the two graphics are overlaid, the interaction between air flow and
temperature becomes immediately apparent.
In the modified power supply, the results were surprising; the redesign actually
delivered better thermal performance! Apparently the spacer increased air turbulence and,
therefore, created a more thorough mixing of the air downstream. The validity of the
simulation data was reinforced by the fact that the downstream simulation results matched
the actual experimental temperatures measured in the wind tunnel.
Once verified, the design change was quickly incorporated into production. To date, 300
new systems have been built with this modification.