tkworld Chapter 198 Practical Data
Chapter 198 Practical Data
March 8th, Chengdu.
In the inertial navigation laboratory of Zhenxin Microelectronics, test engineer Chen Lichao has been working continuously for fourteen hours.
On the test bench is a complete prototype of an autonomous driving inertial navigation module—a three-axis accelerometer, a three-axis gyroscope, a signal processing chip, a filter circuit board, and the Microwave WL-300 sensor embedded in the core of the module.
This sensor is serial number 04782 out of 150,800. From its birth on the production line at CR Microelectronics at the end of January, to its shipment from Hangzhou on February 26th, and its receipt on March 1st, it has traveled 1,800 kilometers. Now, it lies quietly on the experimental table, awaiting its first real-world test after being manufactured.
"Is the data collection complete?" Liu Ren, director of the inertial navigation business unit at Zhenxin Microelectronics, walked in.
Chen Lichao rubbed his sore eyes: "It's the twelfth round. Mr. Liu, you can see for yourself."
Liu Ren walked to the computer, where the screen displayed the statistical results of twelve rounds of testing. He stared at it for three seconds, then froze.
"This number... are you sure you haven't miscalculated?"
Chen Lichao shook his head: "I verified the results three times. And it wasn't just the result from one sensor; I randomly selected five sensors from this batch and performed the calculations separately. The data consistency among the five sensors was within two percent."
Liu Ren took a deep breath.
The core data on the screen is only one line:
Overall angular velocity deviation of inertial navigation module: ±0.021°/hour
What does this number mean? In the field of autonomous driving inertial navigation, ±0.05°/hour is the threshold for "automotive-grade," and ±0.02°/hour is the standard for "high-precision automotive-grade." After Vilan's sensor was integrated into the oscillator module, the overall deviation reached ±0.021°/hour—almost exactly on the edge of high-precision automotive-grade.
Previously, the same module integrated with Bosch's BMA456 sensor had a comprehensive deviation of ±0.038°/hour.
In other words, Vilan's sensors perform almost twice as well as Bosch's best current commercial products in real-world integration scenarios.
"Mr. Liu," Chen Lichao's voice was a little tense, "the nominal accuracy of this batch of sensors is ±0.018°, and the deviation of our integrated system is only 0.003° higher. This noise level... I've been working on inertial navigation for eight years, and I've never seen such a clean signal."
Liu Ren remained silent for a full ten seconds. Then he took out his phone and dialed a number.
"Mr. Chen, this is Liu Ren. The sensor integration test for Weilan is complete, and the data is excellent. Not just excellent, it's... I think you need to take a look in person."
On the other end of the phone was Chen Jun, chairman of Zhenxin Microelectronics.
"How wonderful?" Chen Jun's voice came through the receiver.
"±0.021°/hour. High-precision automotive-grade. Nearly 45% better than Bosch BMA456. Moreover, the five chips have extremely high consistency, and there are no issues with batch stability."
There was a moment of silence on the other end of the phone.
"Send the complete test report to my email," Chen Jun said, "and then contact Ms. Lin of Weilan."
"What information should we notify her?"
Tell her: We need to increase our second order. From 40,000 to 100,000 units.
Liu Ren took a deep breath: "Understood."
After hanging up the phone, Liu Ren glanced at the prototype module quietly operating on the experimental table. The green indicator lights of the five Vilan sensors were flashing evenly, like five inconspicuous green stars.
What he didn't know was that these five stars would illuminate the entire microelectromechanical system (MEMS) industry in a week.
On the afternoon of the same day, March 8th at 4 PM Beijing time, an email was sent from Chengdu to Hangzhou.
When Lin Wei opened the email, her fingers paused for a moment.
The email body consists of only three lines:
Mr. Lin:
The attached document is the integrated test report for the Zhenxin Microelectronics inertial navigation module (WL-300 sensor, five random samples, twelve rounds of testing). Key data: Overall angular velocity deviation ±0.021°/hour, consistency deviation of the five sensors <2%.
See the attachment for details.
Lin Wei opened the attachment and spent fifteen minutes reading through the forty-seven-page test report from beginning to end. Then she closed her eyes, leaned back in her chair, and took three deep breaths.
±0.021°/hour.
She mentally repeated the number. This wasn't lab data, theoretical prediction, or a formula from a research paper. This was the actual integrated performance measured by a customer using Vilan's commercial sensors in their own product.
This is the answer the market has given.
Lin Wei opened her eyes, picked up her phone, and first sent a message to Su Chen: "The test report for the oscillating core is out. ±0.021°/hour. Consistency of five cores <2%."
Su Chen's reply came three minutes later, consisting of only two words: "Understood."
But Lin Wei understands Su Chen. Behind those two words lies the calm of someone who has done six years of research in microelectromechanical systems seeing their theoretical model validated in the real world—not that they have no emotions, but rather that their emotions are too great, so great that they can't express them.
Lin Wei then sent the same message to Zhou Zhiyuan.
Zhou Zhiyuan's reaction was much more outwardly expressive than Su Chen's: "High-precision automotive-grade? The oscillator uses a standard integration process? Was there any special optimization for the WL-300?"
"Standard procedure, no specific optimizations." Lin Wei sent screenshots of the relevant pages of the test report.
Zhou Zhiyuan was silent for a minute, then sent a long message:
"Lin Wei, do you know what this means? It means that the WL-300 can achieve high-precision automotive-grade performance even without optimization. If customers are willing to perform integrated optimizations based on the characteristics of the WL-300, it can theoretically be improved by another 15 to 20 percentage points. This level of performance... there are only a handful of suppliers on the market who can achieve this. And their prices are at least three to five times higher than ours."
Lin Wei knew.
But what she cares about now is not the price, but the time.
She glanced at the calendar. March 8th. Four days until the opening of the 2021 International Conference on Microelectromechanical Systems (MEMS) online—March 12th. Professor Vogt's panel discussion was titled "Frontier Advances in Microelectromechanical Thermoelastic Coupling," and Wei Lan's third-order model was one of the core topics of discussion.
Four days.
Lin Wei made a decision.
She turned on her computer and began composing an email. The recipient was Chen Jun, the chairman of Zhenxin Microelectronics.
The email was simple: requesting Zhenxin's authorization for Weilan to publicly cite core integration test data (de-sensitized) for academic exchange.
Chen Jun's reply arrived two hours later: "Agreed. But there's one condition—when the data is released, it must be labeled 'Zhenxin Microelectronics Inertial Navigation Module'."
Lin Wei's lips curled slightly. Chen Jun was a smart man; he knew that once this data was made public and labeled with Zhenxin's name, it would be equivalent to a free top-tier cheat code.
"Deal." She replied with two words.
Then she called Zhou Zhiyuan: "Zhiyuan, could you present a new set of data during the panel discussion at the International Conference on Microelectromechanical Systems (MEMS)?"
Zhou Zhiyuan was taken aback: "What data?"
"Integrated test data for the oscillator core. ±0.021°/hour, consistency of five cores <2%. Commercially delivered product, not laboratory data."
There was silence on the other end of the phone for several seconds.
"You mean... to directly present real-world data on commercial products at Vogt's symposium?"
"right."
Zhou Zhiyuan took a deep breath: "Lin Wei, do you know what kind of effect this will have? Vogt's panel discussion has at least three hundred people online, all of whom are top scholars and industry executives in the global microelectromechanical systems (MEMS) field. People from Bosch will be there, people from STMicroelectronics will be there, and people from Infineon will be there too. You're presenting real-world data on your commercial products in this setting—"
"I know," Lin Wei said calmly. "This is the effect I wanted."
Zhou Zhiyuan was silent for three seconds, then smiled: "Okay. I'll prepare the materials. Four days should be enough."
March 12th.
The first day of the 2021 International Conference on Microelectromechanical Systems (MEMS) held online.
Due to the COVID-19 pandemic, this year's International Conference on Microelectromechanical Systems (MEMS) was held entirely online. While the atmosphere of face-to-face interaction was lacking, the online format had an unexpected advantage: the number of participants was actually higher than in previous years. More than 1,200 people registered to attend, and the peak number of people watching the various panel discussions online reached more than 600.
Professor Vogt's panel discussion was scheduled for the afternoon of the first day of the conference, at 2:00 PM Central European Time (9:00 PM Beijing Time). The official title of the panel discussion was: "Frontiers of Microelectromechanical-Thermoelastic Coupling: From Theory to Mass Production".
As the name suggests, this discussion group has a very clear focus.
The list of participants in the discussion group was announced a week ago: the moderator is Vogt himself, and the discussion guests include Professor Wei Zhang from MIT (microelectromechanical nonlinear dynamics), Professor Patel from Stanford University (micro-nano manufacturing processes), Professor Yamamoto from Kyoto University (the one who just signed a cooperation agreement with STMicroelectronics), and — Zhiyuan Zhou.
When Zhou Zhiyuan's name appeared on the guest list, it already caused a stir in the industry. After all, he was the corresponding author of Wei Lan's paper, which was still under review in the journal *Nature Microelectronics*. Participating in a public academic discussion before the paper was even published was itself a signal.
The discussion group started on time. Vogt opened with his signature German accent:
"Ladies and gentlemen, thank you for joining us for today's discussion. Over the past six months, there have been some noteworthy new developments in the field of microelectromechanical thermoelastic coupling. Today, we have invited several colleagues who have conducted in-depth research in this area to discuss the significance and prospects of these developments."
The first forty minutes consisted of brief presentations by the guests. Zhang Wei discussed the latest advancements at MIT in nonlinear microelectromechanical modeling, Patel introduced Stanford University's breakthroughs in micro-nano fabrication precision, and Yamamoto showcased Kyoto University's work on thermoelastic coupling simulation—this presentation garnered considerable attention because everyone knew that Yamamoto had just signed a collaboration agreement with STMicroelectronics.
Then it was Zhou Zhiyuan's turn.
His report was titled: "Third-order nonlinear models: from the laboratory to commercial application".
From laboratory to commercial deployment.
With this title, the number of online viewers jumped from 420 to 560 within three minutes.
Zhou Zhiyuan's presentation style was as concise as ever. In the first five minutes, he briefly reviewed the theoretical framework of the third-order model (this content is all in the submitted papers and does not involve any disclosure), and then he switched to a new slide.
The slide only had one line of large text:
Commercial Validation: Results of Initial Product Delivery and Customer Integration Testing
Commercial validation: Initial deliveries and customer integration results.
The online chat window instantly became lively.
"Wait, they've already launched commercial products?" someone typed.
"This is brand new information," another person replied.
Zhou Zhiyuan calmly presented the next slide. It contained a set of anonymized data:
Initial shipment quantity: 150,800 WL-300 series microelectromechanical sensors
Production line: 300mm mass production line (non-laboratory environment)
Yield: 93.6%
Customer integration test results (authorized reference, Zhenxin Microelectronics inertial navigation module):
Overall angular velocity deviation: ±0.021°/hour
Sample size: Five randomly selected, twelve rounds of testing
Five-piece consistency deviation: less than two percent
Integration conditions: Standard process, no targeted optimization
The online chat window completely crashed.
"To achieve ±0.021°/hour without any specific optimization? That's a high-precision automotive-grade level."
"150,000 units delivered in mass production? They've already achieved large-scale mass production?"
"The entire industry landscape will be rewritten. This is no longer just a thesis; it's a real, tangible product."
"Our partner is Zhenxin Microelectronics? I need to look up this company."
Vogt's voice broke the silence, tinged with obvious surprise: "Professor Zhou, if I understand correctly, you mean the WL-300 sensor, developed based on a third-order model, has been commercially delivered and has been validated by customers in automotive-grade inertial navigation equipment?"
"That's right, Professor Vogt," Zhou Zhiyuan replied. "The first batch of products was delivered on February 26th, and customer integration testing was completed on March 8th. The data displayed on the screen has been authorized by the customer for use in academic exchanges."
"That data point of ±0.021 degrees per hour—is it from the client's own testing lab, not from your company's lab?"
"Yes. The entire process uses standard integration technology, and no customization or optimization was done for this sensor."
Vogt paused for two seconds. Sixty-two years of life experience told him that he was witnessing what might be the most important technological verification in the microelectromechanical system (MEMS) industry in the last decade.
"These results are extremely impressive," Vogt finally said, switching back to his characteristically meticulous German tone, "but I have one question: what stage is your paper, submitted to *Nature Microelectronics*, currently at?"
"It's currently in the peer review stage," Zhou Zhiyuan replied. "We expect to receive feedback from the editorial department within the next few weeks."
"I understand." Fogg nodded. "Even so, in my opinion, the commercial test data you just presented is even more significant than the paper itself. A theory that can be applied to produce commercially competitive products is a theory that can truly stand the test of time."
These words, coming from Vogt, carried more weight than any peer review comment.
The number of online viewers had surged to 680, nearing the maximum number of viewers the conference platform could reach.
Once the discussion group entered the open Q&A session, questions flooded in like a tidal wave.
Professor Yamamoto was the first to ask a question. His voice was calm, but those who knew him could sense the underlying urgency: "Professor Zhou, your commercial data is very impressive. However, I noticed that your sensors are manufactured on a 300mm production line. The core inferences of your third-order model revolve around the edge effect of 400mm wafers. Do you have any experimental data related to 400mm wafers?"
This is a precise question. Tsuyoshi Yamamoto and STMicroelectronics signed a cooperation agreement, focusing their research on 400mm process technology. If Vilan possesses mature, measured data on 400mm processes, the very foundation of ST's entire collaborative R&D plan will be undermined.
Zhou Zhiyuan smiled slightly. He certainly had the experimental data from the 400mm experiment—the goodness of fit for the fifteen sets of data was 0.978, and all the S-shaped inflection points predicted by the model were verified. But this data was the core content reserved for revisions and supplements to the paper, and it wasn't the time to publicly display it yet.
"Professor Yamamoto, we have completed the verification experiment on a 400mm substrate, and the results are quite impressive. However, the relevant data is core supplementary content to this peer-reviewed paper and cannot be disclosed to the public at this time. We hope you understand."
Yamamoto paused for a second, then said, "I understand. Thank you for your explanation."
But everyone understood the meaning behind Zhou Zhiyuan's phrase "quite remarkable." He would never have used such a description if the 400mm experiment had yielded poor results.
The next person to ask a question was from Bosch.
"I am Dr. Stein from Bosch's Sensing Technology division. Professor Zhou, congratulations on your commercial testing results. I have a technically related business question: When purchasing in large quantities, such as orders of over one million units, what is the pricing range for the WL-300 sensor?"
Stein.
It was Markus Stein, the senior R&D manager who had been following up on the Vilan project within Bosch for a long time and submitting monthly tracking reports to Albrecht.
Zhou Zhiyuan paused for a second. This question wasn't academic; it was purely a business inquiry. Being asked about pricing at an academic seminar during the International Conference on Microelectromechanical Systems (MEMS) was enough to show that Bosch's attitude towards Vilan had shifted from "continuous observation" to "comprehensive evaluation of competitors."
"Dr. Stein, thank you for your attention. We suggest communicating pricing-related matters through our business channels. However, I can clarify one point: we utilize GE's standard 300mm production line for manufacturing, giving us an inherent structural advantage in manufacturing costs."
The response was standard and tactful, but everyone present understood the subtext: Vilan sensors are priced low. By relying on established, generic production lines, they don't need to invest 1.8 billion euros to build dedicated production lines like Bosch did; their cost advantage is inherent.
Stein did not press further. But on the other side of the screen, he immediately created a new document titled: "Urgent Update: In-depth Analysis of Vilan Commercial Test Data - Presented by Albrecht".
The discussion lasted two hours. In the final summary, Vogt made a statement that has since been repeatedly quoted in countless industry reports:
"Today's discussion has shown me that a fundamental transformation is taking place in the field of microelectromechanical thermoelastic coupling. The speed at which theoretical innovations are being transformed into commercial products is unprecedented. I would like to highlight the commercial validation data presented by Professor Zhou Zhiyuan's team—in my forty-year academic career, I have rarely seen a theoretical model go from a paper to mass production and delivery in such a short period of time. The entire industry should take this signal seriously."
After the discussion group ended, participants took screenshots of the online chat and forwarded them widely. On the evening of March 12th, the keyword "Vilan Commercial Test Results" trended on Weibo for fifteen minutes. Although it was quickly overshadowed by other trending topics, in the niche circle of the microelectromechanical system (MEMS) industry, this night was destined to be a watershed moment in the industry's development.
On Zhihu, the blogger "Calm Analyst" updated his status at 2 AM:
"I just finished watching the replay of the panel discussion at the International Conference on Microelectromechanical Systems (MEMS). I need to revise my previous judgment. Vilan doesn't just have a theoretical paper awaiting review—they have already achieved mass production and supply, and customer test performance has reached high-precision automotive-grade standards. They achieved ±0.021°/hour without specific optimization, with less than 2% consistency across five samples. The value of this set of test data is even comparable to a paper formally published in *Nature Microelectronics*. The core issue in the industry now is no longer 'Can Vilan do it?' but 'How long will it take for other manufacturers to catch up?' Bosch's €1.8 billion roadmap sets mass production at the end of 2024. Vilan is already delivering in batches, creating a three-year gap in technological implementation between the two companies."
This post received 4,100 likes, making it the most liked post in all of "Calm Analyst's" content.
And the blogger "Semiconductor Veteran 2003" finally broke his long silence, leaving a short three-word comment in the post's comment section:
"I'm impressed."
On the morning of March 13th, Su Chen opened his phone and saw a series of screenshots forwarded by Zhou Zhiyuan—chat logs from the International Microelectromechanical Systems Symposium, Zhihu updates, and industry media news.
He finished reading through each item, put down his phone, and walked to the laboratory window.
On a crisp March morning in Hangzhou, the air still carried the faint scent of osmanthus blossoms. On the lab table, the data analysis framework for the second paper had been built up to Chapter Three. Fifteen sets of measured data from 400mm wafers were quietly stored on the hard drive, awaiting the opportune moment for public release.
Su Chen recalled the measured data transmitted by the oscillator—±0.021°/hour. His theoretical model had predicted that the WL-300's theoretical limit of accuracy under the general 300mm process was ±0.018°. After integration into the customer's complete system, the additional deviation range should be between 0.002 and 0.005 degrees. The actual test increment was only 0.003 degrees, perfectly falling within the model's predicted range.
Theoretical derivation, laboratory experiments, mass production, and customer whole-machine testing and verification—all four stages are closed-loop and self-consistent.
A single paper cannot fully explain this entire technological chain. It is a complete and irrefutable chain of evidence.
The entire chain of evidence is now missing only one link: the peer review results from Nature Microelectronics.
Su Chen turned around and sat back down at the computer, opening the journal submission system page.
The page status still displays those two familiar Chinese characters: Under Peer Review.
But he knew that this status would be updated soon.
It has been two and a half months since the paper was submitted. According to the regular review cycle of Nature Microelectronics, I should receive feedback from the reviewers by early April at the latest.
He wasn't in a hurry.
150,000 sensors have been delivered to major customers; the measured data of ±0.021°/hour of the oscillator has been published at an international academic conference; Vogt commented that it deserves the attention of the entire industry; Bosch's Stein wrote a competitive analysis report overnight; and veteran industry bloggers on Zhihu frankly admitted that they were convinced.
All the facts point to the same conclusion.
Su Chen closed the submission page and reopened the data analysis folder.
The second paper will not take shape on its own.
He lowered his head and went back to work.