There were no safeguards in place to
expect
a potential false alarm. Three minutes in, humans determined that alarm was false…but it took another 35 minutes to notify the public. During those 35 minutes, there was widespread panic
while officials knew full well that it was a false alarm.
Why did this happen? We have come to trust computers and electronics more than humans - and common sense. As a test engineer, I can tell you that
computers and other electronic devices are only as trustworthy as the tests they undergo
.
We must anticipate that things WILL go wrong - and design products that way:
Design for Testability
. Test technology needs to make a monumental change. I believe the future cannot survive without it. Only then can we hope to save ourselves from technology.
Tests are woefully inadequate; they have been marginal through the decades.
When I first started in test, we could sometimes cover 95% of all possible faults. Which means 5% of the circuit was
un
tested: roughly 1 of the 20 transistors could have been faulty, and we wouldn't know it.
Today, we test 20 billion transistor parts running at 10 GHz (10 million times faster than when I began testing) and cannot reach 95% fault coverage. Even then, we would have 1 billion untested transistors - that are switching 10 million times as fast as they did decades ago. Many of these 1 billion transistors could be faulty, and we would not know it.
Now that I have painted a bleak picture, let me tell you: the situation will become even worse.
The Internet of Things is a huge network of devices talking to each other: giving instructions and executing commands
without human interaction
. Here is an everyday example…
Your refrigerator weighs the milk carton, decides you need more, and commands your phone to call for milk delivery by drone to your house at 5:30 PM, when you are expected to be home from work; it will charge the purchase to your account.
With just one glitch…
1) you have plenty of milk but just bought more
2) your phone fails to make the call - and you run out of milk
3) your phone gives the wrong address - and somebody else gets your milk
4) drone delivers to wrong address - or not at all
5) drone delivers milk to your door at 1:30 PM, to spoil in the hot sun until you arrive at 5:30
6) you come home at 8:30 PM to spoiled milk – because you didn't tell any of your devices that you would be late
7) your account will be charged $5,000 instead of $5
Remember, just 1 of the 1 billion transistors in your Internet of Things could have failed - and you wouldn't know it. In fact, it could have failed only momentarily.
There will soon be 50 billion items working interactively and continuously - with each having 1 billion untested transistors. We are heading for chaos.
Progress in electronics increases the possibilities of failures. At 4-nanometer technology (already in use), the distance between these connections is less than 10 atoms wide. What is the likelihood that two "wires" so close to each other will not touch, and create a short circuit? While the possibility of faults increases, test comprehensiveness is compromised: it is much harder to detect faults at such density and high speeds.
Small inconveniences will become annoyances… that will become problems…which will eventually lead to "debilitating denial of service." We were able to buy our own milk without the Internet of Things… but after relying on it, will we be capable of coping with mistakes after mistakes? How much will these mistakes cost in dollars - and sanity?
Technical solutions to the testing problem are available but seldom approved - because of costs. It seems more cost-effective to hope that lack of good tests will go unnoticed - as it has for years. As thinking goes today, the false alarm in Hawaii really wasn't all that terrible: no one was injured or killed. (Miraculously, considering the running of red lights and other panic-stricken behaviors).
Now stop to consider a slightly different scenario: What if there was an
actual
missile attack, with no warning because the alarm system was malfunctioning?
Louis Y. Ungar is president of A.T.E. Solutions, Inc., a testability consulting and educational firm.
www.BestTest.com
Mr. Ungar is the founding president of Testability Management Action Group, a Consultant to The American Society of Test Engineers, has served as Testability Chair for the Surface Mount Technology Association and on committees for IEEE standards, , and is in the working committee of the IPC Design for Excellence Standard. He has taught at UCLA and other universities, and his next class is in San Diego on March 1:
Design for Testability and for Cost-Effect Test Strategies with JTAG/Boundary Scan
. For more information, go to
www.ipcapexexpo.org
> Education & Technology >Professional Development.
Share an experts provocative perspective: forward this message to your team members, customers and suppliers.