San Francisco, CA. Test Vision 2020 convened in conjunction with SEMICON West this week, with presenters commenting on such topics as RF test, microwave-millimeter wave radar, Kanban for test development, big data for manufacturing test, known-good-stack testing, WLP test handling, and WLCSP final test.
Kicking off the day-and-a-half event was general chair Steve Tilden, who explained that the goal was to take an eyes-wide-open view of test—the 2020 refers to perfect vision, not the year.
Following Tilden’s remarks, Ben Brown of Xcerra presented the 2014 award for best ATE paper. Lead author Chun-Hsien Peng of MediaTek Inc. accepted the award for “A novel RF self test for a combo SoC on digital ATE with multi-site applications,” presented at the 2014 IEEE International Test Conference.
Kaivan Karimi, vice president and general manager of wireless solutions at Atmel, then delivered a keynote address on how the Internet of Things will impact the semiconductor and cloud markets. “You are experts in semiconductor test and packaging,” he said, “so I won’t bore you with what you already know.”
There is some hype surrounding the IoT, he said, but credible forecasters such as ABI Research are predicting 30 billion devices deployed by 2020. Atmel, he said, estimates there will be about 100 connected devices per household. Whatever the exact figures, he said, “A lot of widgets are going to be talking to each other.” And the question an engineer will face is, “How do I provision connectivity?”
That connectivity will take the form of machine-to-machine (M2M), machine to nature (or the environment), machine to infrastructure, and machine to human (for biometric monitoring and disease management via implantable and wearable electronics).
Karimi cited several examples. Water is at a premium for Arizona farms, he said, and simple timers don’t constitute a particularly effective way of controlling irrigation. But if you deploy moisture sensors in the field, you can cut water use by 37%.
In addition, he said, the western world has an aging infrastructure. If we can’t muster the will to pay to fix all of it, we can at least determine which structures are most likely to fail by adding little 30-cent microphones. It would take about $12k to monitor the Golden Gate Bridge, he said.
He also said that the casualty rate due to natural disasters is dropping. Previously, people didn’t receive timely warnings about a tsunami, for example. Now with the vast improvements in communications, people do get timely warnings, but there is still a problem in that they choose the wrong escape route. He said that water-level detectors could indicate that a particular bridge, for instance, is in danger of failing, so people could be directed to another route.
As for machine-to-human connectivity, Karimi didn’t seem too impressed with Apple Watch and Google Glass, suggesting they will end up in a museum of technical history. More promising, he said, are categories of wearables that could revolutionize the healthcare system—from disease management to prevention.
Atmel, he said, is addressing many markets: buildings, energy, consumer homes, healthcare and life sciences, industrial, transportation, and retail. These divide into two larger categories, he said—the industrial IoT and the consumer IoT, each of which will require different test specifications. The Industrie 4.0 initiative in Germany, he said, is prompting the development of test specifications for the industrial IoT.
It will also be necessary to identify the appropriate IoT communications topology. For some categories, your cellphone is gateway; for others, a box in your house.
Thing-to-cloud communications will play a role, but it shouldn’t be overused, and it’s important to determine where intelligence should reside. If your thermostat determines your house is too warm, for example, it needs to talk to the air conditioner—not the cloud. Conversely, if it is very, very warm, your smoke detector might need to call the fire department. It’s an “exception event” that should prompt communications with the cloud, he said.
On a related note, he said he lives in Austin, and he would appreciate a machine-condition-monitoring solution that would give him a two-week warning that his air conditioner is about to fail—another exception event.
Karimi described a typical IoT device as an ocean of software surrounding an MCU/MPU, analog circuitry, nonvolatile memory, I/O, antennas, RF circuits, sensors (perhaps MEMS), a PMIC or PMU, and security. The fact that it needs to cost less than $2 means you are not going to have seven different packages. That calls for “more than Moore” and a diversity of heterogeneous technologies integrated in one device. That presents EDA challenges as well as semiconductor process challenges. Unfortunately, the EDA companies are focusing on DSM, he said, and the high-volume IoT area is being ignored.
He then discussed security. He said he’s all for open systems, but the only secure system is a closed system. “It’s amazing how insecure web is as a medium,” he said, despite a façade of security.
He concluded with some comments on test and standards—such as those of the OIC. He questioned whether test should shift from unit test to system test and how to have test embedded within the layers of fanout packaging.
In any event, he said, the IoT is here to stay, concluding, “You need to understand the system well before you test it.”
