Home / Blog / Testing. One, Two, Three Testing

Testing. One, Two, Three Testing

There was an interesting article by Anil Khitolia on EDN a few weeks ago on LTE Stress Testing.  He started out with a fast fact that I found pretty shocking – a claim that 15-20% of smartphones are returned.  Anil attributes this to lack of rugged testing:

Current wireless industry standards require device certification or conformance testing prior to launch. This type of testing requires the device to pass a minimum performance standard that focuses on device functionality, but it does not test the device the way the consumer will ultimately use it. Certification testing guarantees a device meets industry standards, but it tests only one operation at a time and was never intended to establish how well that device works when left running for many hours.

He goes on to talk about the need for stress testing, especially as “LTE has added another dimension and sense of urgency to the equation.” (He also gives a nice example of the difference between a simple conformance checklist test and a stress test that you might find worthwhile. His article is definitely worth a read!)

My initial reaction, which I posted in a comment, was:

Stress testing is one of the important steps in fielding a successful product. With the rapid pace of the constant 2-year upgrade cycle that cell phone customers are addicted to it seems that providers have perhaps been rushing — or skipping all together this important step towards reliability and ultimately customer satisfaction.

For Critical Link, stress testing has always been, well, a critical link in our board development process. Unlike the average smartphone, our system-on modules (SOMs) are expected to have a very long life.  They’re often embedded in complex (and expensive) equipment with a life expectancy of ten, or even twenty, years.  These sorts of applications are pretty much the exact opposite of consumer products, which quickly become throwaway. Scientific instrumentation, industrial automation, and like applications don’t get replaced just because something new and flashy appears on the scene.

For our SOMs, we do many types of design verification testing at the board level, on the prototypes, to make sure that the design is solid and robust. Generally, we run them across the full temperature range in a temperature chamber, running the temp up and down to make sure the board supports the range that it’s rated for and can withstand and function properly through many thermal cycles. We also perform shock and vibration testing, which I wrote about here a couple of weeks back.

If we’re involved in a project at the application stage, where we’re working with a customer to finalize the entire hardware and software package, we also perform stress testing. We accelerate the life cycle of the product, running it for many days, if not weeks, under highly stressful situations and conditions. If the product (and its associated application software), for example, is  expected to receive ten messages per second over Ethernet, we may send 100 or even 1,000 messages per second to try to force any issue with memory leaks and other potential longevity problems to the surface quickly. Once the product passes a full set of functional tests to ensure all the intended requirements are met properly, and the stress testing is complete, the customer will move to production. But that’s once we’ve ensured that the design works under more stressful conditions than it’s ever likely to experience once it’s put into use.