Startup Validates its MEMS Scanning Mirror Process for LiDARs
Setting out to overcome the drawbacks of today’s LiDAR mirror alternatives, Omnitron Sensors has crafted a 3D MEMS technology fit for low-cost, high-volume markets.
The LiDAR subsystem market continues to gain momentum, with forecasts of $2.3 billion by 2026 according to Yole Développement. Aiming to capitalize on that, today, Omnitron Sensors announced that it has validated its process for a MEMS (microelectromechanical systems) scanning mirror.
The scanning mirror is a new optical subsystem designed to meet the rigorous requirements of the LiDARs used in automotive advanced driver assistance systems (ADAS), drones and robotics.
Omnitron claims that its MEMS mirror will produce a 2-3x larger field of view (FoV) than other MEMS mirrors used in long-range LiDARs. The step-scanning mirror is designed to handle rugged high-vibration automotive and aerial environments—applications that aren’t suited to spinning mirrors in LiDARs produced by other vendors, says the company.
Compared to the innovation seen in semiconductors, over the ten years, MEMS technology advancement has failed to keep pace.
In this article, we describe the problem the company’s MEMS process solves for LiDAR, the approach Omnitron Sensors takes with its process, and we share insights from our interview with Eric Aguilar, co-founder and CEO of Omnitron Sensors.
MEMS Topology Innovation Has Been Stagnant
From his point of view, Aguilar says he saw an opportunity to apply MEMS to solve problems with today’s LiDAR technology. To do so would mean coming up with a new approach with MEMS, a technology area that’s been stagnant for years.
Aguilar recalls that, in the ‘90s, MEMS was deployed in airbag sensors, and ten years ago, Steve Nasiri, founder of Invensense, integrated the first ASIC with a MEMS in a low-cost package. “But there hasn’t been anything notable in MEMS in terms of high-volume capability or disruption since then,” he says.
“I noticed that there really hasn't been any significant movement in the MEMS industry from a technology perspective for the last decade or so.”
Aguilar says the problem with today’s LiDARs is that they don’t meet the needs of next-gen system designs. At the heart of the issue is that LiDARs suffer from a number of problems with their optical subsystems—their mirrors and scanners. The optical components are too large and too fragile, while also being too expensive to manufacture and maintain.
Meanwhile, mirrors that are cost-effective and reliable are possible using MEMS technology, but current MEMS technology hasn’t quite been a fit for MEMS, says Aguilar. “LiDAR customers are saying ‘We need a beam diameter of 10 mm in order to hit a range of 200 meters or so,” he says, “But MEMS mirrors today are 1- or 2-mm in diameter. That's why you don't see a lot of MEMS in LiDAR today, because you’ve got to build a very big mirror to do that, and you can't do that with today's MEMS processes.”
Today’s LiDARs All Have a Mirror Problem
Aguilar says that Omnitron’s MEMS mirror technology contrasts with older LiDAR optical subsystems, including voice coils, spinning polygons, and Galvos. Those alternatives are all slower, bulkier, more expensive, and subject to failure.
For their part, voice coil technology has some advantages like its good feedback mechanism. It’s a robust technology that can deal with vibration and deal with temperature cycles. That’s why voice coil mirror technology is used in space communication systems.
The downside is a voice coil subsystem can cost $5,000, says Aguilar. That’s prohibitive for automotive system designers that need a solution for under $500. Worse, you would need multiple voice coil units to cover the full field of view because they only move a few degrees— less than ten degrees each.
Today’s LiDAR mirror technologies each have cost, reliability, and field-of-view (FOV) trade-offs.
Other LiDAR mirror alternatives include SCALA and spinning polygon technologies. These essentially put a mirror on a motor and spin it. A challenge with LiDARs in general is that you’re dealing with accuracies and precision in centimeters. That means less than 5 cm of error over a 200 meters of range. That translates into a surface planarity for these mirrors to be down to the nanometer level, says Aguilar.
Achieving that surface planarity requires extensive polishing so that the mirrors are highly accurate at initial calibration time. Unfortunately, that calibration can get lost if the motor bearings of these spinning systems start to wobble. It only takes a few microdegrees of error to lose your calibration. And, because the wobble is random, there’s no way to easily recalibrate.
A 3D MEMS Approach for LiDAR
With all the problems of today’s LiDAR mirror alternatives in mind, Omnitron developed a 3D MEMS process technology that enables a scanning mirror suited to next-gen LiDAR needs. “We've developed a 3D MEMS processor and it is, in a sense, a new topology in MEMS,” says Aguilar. “Just like you have a new technology node for semiconductors, this is a fundamental new topology in MEMS.”
Omnitron’s approach was to try to build an electrostatic motor that can move a MEMS mirror and achieve much more force-per-unit-area than what's available in the market today. Omnitron leveraged a 3D MEMS topology to achieve this, but it was important to make sure it was a manufacturable process, says Aguilar.
“I've done third party verification that this is achievable,” he says. “Meanwhile, I didn't want to have to spend millions of dollars on a new machine that could do this.”
“We're using very standard tools and processes that are available at any industrial or large scale semiconductor foundry. That was very important to our approach. I didn't want to have to invent a new machine or a whole new process to do this.”
In order to ensure a simple, manufacturable process, Aguilar says they didn’t go with metal springs for their MEMS. “We went with silicon-based springs because they're a thousand times stiffer and they don't wear out,” he says.
“This is why airbags sensors are built in semiconductor MEMS types devices because they don't fatigue over time,” says Aguilar. “You need to be able to do hundreds of thousands of cycles in operation and billions of cycles over the lifetime. That’s why we went with a semiconductor process and used bulk silicon to etch our mirror and device out of that.”
Omnitron’s 3D MEMS design approach is based on a simple and robust wafer manufacturing process.
Accuracy and reliability are also key aspects of Omnitron’s technology. “If I'm saying that the mirror is pointing at three degrees, it really is pointing at three degrees,” says Aguilar. “We've created isolation layers in the plane that allow us to decouple the drive energy from the sensing mechanism.That means I don't have the drive signal coupling into my sense signal, corrupting the data.” Those aspects have all been verified through fabrication, he says.
Enabling Reliable Calibration
Another feature of Omnitron’s 3D MEMS sensor technology is that it ensures reliable calibration. This is enabled by the company’s choice of drive scheme. “For the motor that we're building, the actuator is driven electrostatically,” says Aguilar.
Aguilar explains that there are three approaches with MEMS in terms of how you drive it. You can drive it electrostatically, electromagnetically, or with piezoelectric scheme. “We went with electrostatics primarily for linearity across temperature,” he says.”The great thing about electrostatic systems is they expand and contract over temperature in a very linear fashion. It's very deterministic.”
Aguilar says that the problem with the other alternatives— electromagnetic or piezo technology—is you have a hysteretic response over temperature. That means that the temperature curve that the device is on is different going up in temperature versus going down in temperature. That creates a significant error in your performance, which impacts calibration.
“The reason why we went with this approach in terms of an electrostatic motor was to make sure that you don’t lose calibration over time. So if the mirror says it is at three degrees, it will stay at that three degrees forever unless something breaks.”
A New MEMS Topology for Next-gen Designs
It’s true that much of the innovation in today’s embedded systems are due to software-based functionally running on ever-faster microprocessors and MCUs. Omnitron’s new MEMS topology is a reminder that hardware innovations can still have major impacts.
“As good as software currently is, we won't be able to realize the future which we imagine without fundamental changes in hardware,” says Aguilar. “We had to go back to first principle thinking in order for us to develop this fundamental new MEMS process that we know will change the sensor world.”
All images used courtesy of Omnitron Sensors