« Microsoft's entry into tablet hardware a result of partner failure? | Main | "The Windows Marketplace for Mobile for windows mobile 6.x devices is closing" »
August 11, 2012
Performing under pressure
As I am writing this, the London Olympic games are coming to an end. What two weeks of intense competition proved again is that winning means meticulous preparation, at times a bit of luck, and always the ability to perform under pressure. The latter made me think because rugged computers are all about the ability of a piece of equipment to perform under pressure. Pressure as in heat, cold, dust, rain, sun, and whatever else may keep a system from running at peak efficiency.
Ruggedness testing is designed to determine if systems hold up under pressure, but are the tests really meaningful? Many probably are. If, for example, a system is dropped a number of times from a certain height and still works afterwards, chances are it'll survive similar drops out there in the field. But are all tests as meaningful?
A while ago a manufacturer of rugged computers challenged us to test computing performance not just in an office environment, but also over the entire listed operating temperature range. We did, and not surprisingly, the machinery supplied by that company passed with flying colors, i.e. it ran through the benchmarks as fast at freezing and near boiling temperatures as it did at the 72F we usually have in the test lab.
But, as we subsequently found out, that seems to be the exception. We've been doing benchmark testing on some other rugged devices under thermal stress, and the results are reason for concern. If a rugged handheld, laptop or tablet is supposed to be used out in the field, it's reasonable to assume it'll be asked to perform at peak efficiency at temperatures one might likely encounter outdoors or on the job. Depending on where you are, that might easily include temperatures well over 100 degrees. Such work may well include prolonged exposure to the sun where it may heat up beyond ambient temperature. If it is 105 degrees outdoors, temperatures may easily reach 115 or 120 degrees or even higher if you sit the device down somewhere, or even if it's left in a car. So what happens to performance then? Can the device perform under pressure?
Turns out, not all can.
Running our standard benchmarks after leaving rugged systems out in the California summer sun showed performance drops of 50 to 80%. That's pretty serious. Is it acceptable that a piece of equipment that's supposed to be used outdoors then runs only at a fraction of the speed or even at half speed? I'd say not. Think of the potential consequences. Tasks may take between twice to several times as long, potentially affecting critical decisions.
Is it reasonable to expect full performance under extreme conditions? Not necessarily. Extreme conditions can have an impact on electronics, and there may be justifiable, reasonable precautions to limit performance so as to safeguard the equipment and its life. But is it acceptable to see performance drop to a fraction at the limits of a listed operating temperature range? It's not. Customers should know what level of performance they can expect when the going gets tough.
Like at the Olympics, performance under pressure separates the rugged system winners from the also-rans. This really needs to be addressed.
And it's not a simple issue. Complex electronics such as processors have sophisticated internal power management. Boards have sensors that report temperatures to control mechanisms that then may throttle system performance. Firmware and the OS may also monitor environmental situations and then engage fans or throttle performance. The hardware itself may have inherent design limitations. Variables such as Glass Transition Temperature, or Tg, come into play. Tg is the temperature at which polymer materials go from a glassy state to a rubbery state. The types of capacitors used matters. Conformal coating can protect boards. HALT testing can predict real life reliability better than the simple mean time between component failures. And so on.
All of this is standard practice in embedded systems design. It should be fully and universally adopted in rugged mobile system design as well.
Posted by conradb212 at August 11, 2012 4:46 PM