Omnitron’s MEMS Tech Boosts Lidar Reliability

Five years ago, Eric Agilar He was full of throat.

He worked on Lidar and other sensors for years in Tesla And Google X, but the technology always seemed too expensive and, more importantly, unreliable. He replaced Lidar The sensors, when they broke – what was too often, and, it would seem by accident, and developed complex calibration methods and service procedures so that they were functioned, and the cars were controlled.

So, when he reached the end of his rope, he invented more reliable technologies – what he calls “the most powerful micromashine, when he was made.”

Agylar and his team in the startup Omnitron sensors developed new microelectromechanical systems (Mems) The technology that he claims can give more strength per unit area than any other. Providing new levels of power MicromirriersThe technology is able to accurately control the laser rays of the lidar, even during the weathering of dangerous elements, as well as shocks and shocks on an open road. With tested chips conducted by customers, Omnitron is currently changing the technology to reduce the power consumed by artificial intelligence Data processing centersField

Lidar, scanning and detection system that uses lasers determine how far the objects are often accepted Self -driving cars To find obstacles and navigate. Even if it is expected that the market for lidar grow by 13.6 percent per yearAccording to Agilar, the use of Lidar in the automotive industry remained relatively congestive in recent years, partly because the life of the technology is so short.

Vibration from bumpy roads and difficult environmental conditions is the largest reliability killer for automobile lidar, says Mo LiWhich is studying photon systems at the University of Washington. Optical alignment in the Lidar package on autonomous cars is a delicate terrain from poor work by power can physically change where the mirrors are located in the housing, potentially mixing the beam and leading to the failure. Or temperature fluctuations can lead to expansion or reduction of parts or with the same unsuccessful result, he explains.

Agylar wondered which part crashed most often, and discovered that the culprit Scannersparts responsible for the trap of small mirrors that direct Laser beam From the housing of the system. He wanted to make scanners that could withstand the strict conditions of the lynar person, and the bend of silicon stood out as a solution. These structures operate as springs and allow you to carefully control the mirrors in lynar systems without wear, as standard metal springs do, says Agilar.

Design of the best chip

Agilar hoped that the new material would be an answer to the problem in which he was tormented, but even silicone sources did not make lynar systems as reliable as they were supposed to confront the elements that they encountered.

To make Lidar even stronger, the Omnitron team sought to develop a more powerful MEMS chip by increasing the amount of force that the device can use to control the mirrors in the Lidar array. And they argue that they have reached this – their chip can provide 10 times more strength per unit area on the drive that the micro -world or another component of the sensor positions than the current industry standard, they say. This additional force provides extremely valuable control in thin adjustment.

To achieve this achievement, they had to dig deep – literally.

OMNITRON micromirers direct lynar rays and can find use in data processing centers.Omnitron

In this device, the MEMS mirror and its drive are etched in one Silicon plateThe field on its non-Miroic end is covered with tiny, closely located plates, which are placed between the trenches in the plate, like interconnected teeth of two combs. To move the mirror, the use of voltage, and electrostatic forces angry the mirror to a certain position, moving the plates up and down in trenches How Electric field pulls the side walls of the trenchField

The force that can be used to move the mirror is limited by the ratio of depth to the width of the trenches, called the ratio of the parties. Simply put, the deeper the trench, the more electrostatic power can be applied to the drive, which leads to a higher range of movements for the sensor. But the manufacture of deep narrow trenches is a difficult effort. Overcoming this limiting factor was mandatory for Agylar.

Agylar says that Omnitron was able to improve the parties of about 20: 1, which he notes, is typical for MEMS (other experts say that 30: 1 or 40: 1 closer to the average today), reaching up to 100: 1 through experiments and prototyping at a small university Foundry through United States “This is really our main breakthrough,” says Agilar. “It was through blood, sweat, tears and disappointment that we started this company.”

According to Agylar, the startup received more than 800 million dollars of letters about Automotive Partners, and for two months in 18 -month -old terms to prove that he can produce his chips with a full level of demand.

Even after checking production capabilities, the technology will have to face a “very stringent” security test for thousands of consistent hours in realistic conditions, such as vibrations, thermal cycles and rain, before it can come to the market, says Lee.

Energy saving

Meanwhile, Omnitron applies its technologies to solve another problem, which is faced with another industry. By 2030, it is expected that data processing centers will be required About 945 hours Tawardtt function – more than a country Japan Absorbs today. The problem is that “the method moves the data,” says Agilar. When the data is sent from one part of the data center to another, Optical signals are converted into electrical signals, redirected, and then returned to optical signals be sent in their way. This process, which occurs in systems called network switches, burns a huge amount of energy.

Google decision called Apollomaintaining data packages in the form of optical signals during their travels, which gives a 40 percent power savings, The company claimsField Apollo This is done using an array of mirrors to send data. Agilar plans to make the process even more effective, using dense arrays of more powerful Omnitron mirrors. This can four times four times Network switch According to Agilar, it can direct an increase in the number of channels in each switch from 126 to 441.

Omnitron is still at the beginning of the implementation of the data processing center, so it is not yet clear to what extent this technology can really improve on Apollo Google. Nevertheless, after a “critical review of design” in mid-September, “one of the best AI hypershalors in the world asked for our mirrors to switch the next generation,” says Agilar. “This is evidence that Omnitron solves the problem that even the largest companies in the AI ​​infrastructure cannot solve in the house.”

And there may be even more applications. Omnitron received sensations from the defense industry, space companies and groups interested in methane Detection, says Agilar. “It’s pretty good to see people knock on our door for this, because I just focused on Lidar,” he says.