Microwave Engineering Europe - December 2007 - (Page 31) THERMAL ENGINEERING 31 lacked any kind of visibility so it was hard to see if the analysis was heading in the right direction. It shot off at tangents and would come up with some wildly inaccurate results.” So when it came to developing an entirely new product range – a group of amplifiers operating below the 1-GHz band – Hodson knew it was time to invest in an up-to-date solution that would deliver reliable upfront analysis. Thermal analysis was deemed necessary because the transistors Hodson planned to use in the new amplifier module had a far greater wattage output than any the company had used before. Output from a 1-2 GHz-unit transistor is approximately 22.7 W working through a surface area of 63 mm2. Output from the RF transistor was estimated to be 96 W working through an area of 60 mm2. The use of only two transistors in the module indicated a design with two very localised heat spots needing to be dissipated across the rest of the module. The initial stage of the analysis examined the heat junction to get a realistic figure for the thermal joint between the transistor base and the aluminium heat sink. The next stage was to optimise the amount of material under the transistor. This was to ensure the right amount of material to spread the heat to as many fins as possible but not so much as to let the temperature build up in the component or to make the module too large. From this the analysis turned to fin spacing. This had to be optimised with the available airflow generated from a fan. Hodson used the performance graphs of Milmega’s commonly used Pabst 5212 fan, which the company uses in all its amplifiers. It became apparent from the flow and pressure graphs that the temperature was much more dependant on pressure slowing the airflow than was expected. The fin profile that Hodson arrived at was with a larger air gap and smaller metal fin than the standard range of module: 4.0 mm of air and 1.5 mm of fin as opposed to 3.2 mm air and 2.0 mm fin on the 1-2 GHz. It was also shown that if the fins were too long the pressure build-up along the length slowed the air and had a detrimental effect on temperature. To reduce pressure Hodson removed 50 mm of fin from the leading and trailing edges of the chassis fining. Reliability After only twelve months of use, Hodson has already identified a range of features within CFdesign that have delivered Figure 1: Analysis of a temperature section (left) and velocity trace (right). considerable benefit throughout the product development cycle, not least of which is the ability to defer physical testing to much later in the process. “Previously we’d get only so far in the design process before having to stop and run physical tests,” explains Hodson. “We’d use our intuition when it came to the CFD, which is fine if you get it right but costly and time consuming if you have to go back to the drawing board and start all over again. With CFdesign we have a lot more confidence that our physical models will work first time.” Following his experiences with the previous system, Hodson is also enjoying using software he can rely on. Despite the fact that CFdesign runs on a normal computer (avoiding additional investment in specialist hardware), Hodson has found it completely reliable and never experiences the frustrating crashes that were such a feature of the previous system. “Complex analyses can run over a couple of days, and to save time during development it helps if you can start the CFD running on a Friday and be able to pick up the results when you come in on Monday morning,” says Hodson. “You don’t want to get back and find it crashed on Friday evening. CFdesign just gets on with it and comes up with the right answer time after time.” Hodson has also found CFdesign to be an intuitive system, a feature which, as he explains, can be of real importance for companies like Milmega that tend to have a few months, break between design cycles: “Like many other manufacturers of our size, we tend to run a lot of analyses over about six weeks, then drop it for up to three months while that design goes into production. You need a system which is easy to use so you can get up to speed again immediately, even after a break.” Design optimisation CFdesign was introduced into Milmega at the start of a project to develop a completely new product range. Twelve months later, the new range is about to go into production. “We haven’t used any of our traditional methodologies at all,” says Hodson. “We’ve designed everything from scratch and used the analyses to get right to the very heart of the module, not just working out the air flow, speed, pressure and temperature throughout the unit, but also calculating the optimum amount of metal necessary to get the correct heat spread.” The accuracy of the results achieved through the CFD simulation also encouraged the team to build the amplifiers with one fan instead of two. Usually, two fans would be used just in case, but when CFdesign predicted that only one would be necessary to do the job, Hodson put his faith in the results. “We certainly wouldn’t have had the confidence to do that before,” he says. “The results turned out to be right, enabling us to reduce our costs and optimise the design even further than we had originally imagined.” CFdesign has enabled Milmega to bring a completely new range to market in only twelve months, and despite the fact that Hodson clearly rather enjoys the more traditional airflow experiments (“I put a smoke bomb in the amplifier outside the factory and see how the smoke flows through it”), upfront CFD simulation has helped to keep physical tests to a minimum and has achieved demonstrable design optimisation. Milmega has built its reputation on achieving powerful results within small spaces, and upfront CFD has proved an ideal tool for helping to deliver just that. Microwave Engineering ● December 2007 ● www.mwee.com 030-031_MWEE.indd 31 23/11/07 12:02:08 http://www.mwee.com
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.