The barrier to entry in the MEMS market is a lot higher than most entrepreneurs think, said Xie. He rattled off six things anyone needs to know before getting into MEMS business.
- First, each MEMS technology requires a unique manufacturing technology. You can’t just go to any foundry and ask them to produce MEMS. Most foundries don’t have the IPs necessary to produce your own MEMs.
- Second, MEMS requires your own sensor designs.
- Third, MEMS also demand the pairing of a sensor and a low-power-consumption ASIC that works with precision under weak signals.
- Fourth, a sensor and ASIC must be put in a single package. Chances are that most foundries don’t have that experience, either.
- Fifth, MEMS needs to be tested while in motion.
- Sixth, sensors in a handset, for example, need to talk to the main chip in the system. Each sensor, however, is different, and each cellphone module is different. Hence, the way they communicate is also different.
If there’s anything Xie underestimated before getting into the MEMS market, it was the amount of software work his company and field application engineers needed to fine-tune the application software. When Xie discovered that his company had made MEMS that worked perfectly in one phone but wouldn’t work in another, he said, “We realized that software was the killer.”
The push toward the 6D and 9D sensors has been largely driven by sheer complexity and difficulties in mixing and matching different algorithms and sensors integrated with ASIC offered by different vendors. “A one-stop solution makes MEMS integration cleaner, simpler and more reliable,” Xie explained.
Challenges MEMS faces in future aren’t just the increasing number of sensors that must be integrated in a single package. The industry’s new thing is a “sensor hub,” in which vendors add a microcontroller to the sensors-and-ASIC combination in the package. The microcontroller is needed inside the sensor hub, mainly to manage power consumption, Xie explained.
Previously, the main CPU in a handset did that job, according to Xie. But waking up the CPU every time sensors need to measure something turns out to be a huge power drain.
Beyond the company’s first accelerometer just announced this week, QST today has a road map for CMOS integrated 6D motion sensor (such as tri-axial accelerometer + tri-axial gyroscope, and tri-axial magnetic sensor + tri-axial accelerometer technology) and 9D sensors.
But at the advent of the sensor hub, Xie said, “We decided to move up our plan to add a microcontroller to our MEMS sooner.” Previously, it was on the comapny's agenda for 2015 or even 2016.
QST has a choice between outsourcing its MEMS package to an MCU vendor and working with an MCU vendor to integrate it into its MEMS. QST decided on the latter. Xie said, "Considering ARM's ecosystem, ARM core will be the first one we will consider. For specialty, low-power version, we may consider MIPS as well."
— Junko Yoshida, Chief International Correspondent, EE Times