Thermal engineering used to be a job for thermal engineers. In large companies, or companies where thermal design is absolutely crucial, they may still have an important role. They may deploy their talents during every stage in the design process, right from analysis of the concept in order to draw up initial design specifications. However, smaller design teams may have to accept that such expertise may not be readily available.
Even big companies get it wrong. The highest profile case is probably the overheating Xbox 360 that cost Microsoft somewhere upwards of $1bn when initial models had to be recalled in 2007.
Tom Gregory is the 6SigmaET product specialist at Future Facilities, and he commented: "They didn't release the exact reasons for the problem but I wonder if they didn't really consider how people might place the Xbox. It can be quite useful, for example, if there are things stacked on top of it or if there are vents that are blocked. It was certainly an embarrassing failure."
There are three consequences to consider when looking at the thermal management. Firstly if a component exceeds its specified operating temperature then its operation may be compromised.
Secondly, most specifically with respect to consumer items, it can cause the product to be uncomfortably hot to handle. And lastly it can actually be a safety risk if it is in danger of catching fire.
So who is responsible for making sure this doesn't happen?
Gregory commented: "Some big companies will still have dedicated thermal engineers but many companies cannot afford that. Some work can be done at board level by the electronic engineer but at system level the mechanical engineer, the person responsible for the enclosure, will often be responsible for the making sure the electronic equipment inside the box doesn't overheat."
Clearly thermal engineering is a discipline in its own right and while such issues as experienced by Microsoft are an anomaly in the world of consumer electronics, Gregory believes that practice in other sectors is not always up to scratch.
"I would say in some companies it is lower priority than other design aspects," he said. "There is a lack of understanding and over-reliance on rules of thumb – like if we put a big enough heat sink on it or use a large enough fan it should just about work. And perhaps an over-reliance on measurement after the case."
Measurement after the case is a familiar practice - that recurring loop of taking a design to prototype, taking measurements, fixing problems and then re-prototyping.
"Respinning a PCB or having to change the tooling for the enclosure can be incredibly expensive," claimed Gregory, "so you need to find mistakes before the prototype stage. The cost of fixing problems goes up the further down the design process you go."
What Future Facilities promotes is the use of thermal simulation right from the outset, starting with creating a thermal simulation at the design concept stage. By creating a simulation based just on a conceptual board, with certain components already on it, it is possible to get a good idea if the mechanical enclosure will be able to dissipate the right amount of heat in certain environmental conditions. This simulation can be done even before a mechanical or electronic engineer has got anywhere near their CAD packages and can help form some of the initial specifications for the project before they do.
As soon as CAD data has been developed for either the electronic or mechanical design, it can be fed into the model. This model can undergo continuous – daily if necessary – simulations to make sure that as it evolves its thermal performance is not being compromised by design iterations. At the end of this cycle a prototype can be made that when tested, theoretically, validates what was predicted by the simulation. This 'right first time' prototype can add speed to the time to market and reduce the cost of getting there.
Thermal simulations are not new. It is a process that has been around for around 25 years, but has predictably moved on a lot in that time. Chris Aldham, product manager at Future Facilities, commented: "25 years ago doing a thermal simulation required making a lot of assumptions. You were obliged to simplify the models, but there was always a danger in simplifying these models you would throw out something that is important. What has happened over the past few years is that the combination of more powerful computers and better software has meant that you can put more detail into the model."
That issue of detail forms the basis of one of the two major developments in the latest revision, R9, of 6SigmaET. Most thermal modelling systems are built up using a structured cell grid. Having uniform sized cells in this structured grid makes for a system that is relatively simple to run calculations on. However, it means that cells exist at non-critical parts of the model and equally the critical parts lack the necessary detail. And, said Gregory, it was limiting the potential of the system.
"So we went for the unstructured approach," he said, "which people have done before, and means only placing grid cells where required. It can be computationally quite an inefficient way of doing it and people who have done it in the past have had performance issues. So we spent a number of years developing 'the solver' that can solve an unstructured grid in a really fast and efficient way with none of the inefficiencies or limitations of previous attempts."
It means that fewer cells can better represent the important parts of a thermal simulation so it gives the option of either increasing model complexity without making extra demands on the computing environment, or simply running the simulation faster.
The other main improvement is in the user interface, which has become critical in making thermal engineers' software accessible to mechanical engineers. Aldham observed: "Sometimes people would only use the software once a week, or month or even three months. So if it is a complicated tool they will forget all the details that are needed, but if it is intuitive then they can leave it for three months and get straight back into it."
A simulated future?
So could simulation ever completely replace prototyping? "That is where people are going," claimed Gregory. "Already people, like Volvo, are trying to do as much virtual prototyping as possible."
But we are not there yet warned Aldham: "In order to do a very accurate thermal simulation you need a lot of data and that data has to be correct. This includes what the power dissipation of the components is and that sometimes is difficult to get a handle on in the theoretical stages – electronic engineers don't really know what power that component is going to be putting out until they have built and measured it, particularly with things like processors whose output depends on what software it is running. Until you can tie it all together there will always be a need for a bit of verification."