Company Profile
Suzhou NOVOSENSE Microelectronics Co., Ltd. (referred to as "NOVOSENSE", Stock Code: 688052.SH / 02676.HK) is a high-performance, high-reliability analog and mixed-signal chip company. It focuses on three major product directions—sensors, signal chains, and power management—and provides full-chain semiconductor solutions for industries such as automotive, renewable energy and power, industrial automation, and smart terminals.
Editor's Note
Behind the fluid arm movements and turns of humanoid robots lies an extreme demand for "angle" precision in thousands of joints. The component that allows humanoid robots to "know how many degrees they have turned" is precisely the angle encoder hidden within those joints. As "traditional angle measurement" is reignited by robotics, new demands and challenges follow: Why must encoders be developed as a product matrix? Can automotive-grade products be directly reused for robots? What is the non-negotiable red line for customers when selecting components? Compared to overseas brands, what advantages and opportunities do domestic products possess? In this episode of Chip Unveiled · Experts Talk Chips, we interview Xie Ben, Sensor Marketing Director at NOVOSENSE, to delve into the "protractor" of robot joints and dissect the technical breakthroughs and selection logic of encoders.
*The following is a text version of the podcast, confirmed and authorized by the guest.
The "Protractor" in Joints:How Do Robots Sense Rotation Degrees?
Huan Shi: When NOVOSENSE appeared on Chip Unveiled last year, you discussed your technological layout in magnetic sensors, which are core components for robotic joint perception. With the robotics sector booming this year, I am curious whether NOVOSENSE had long targeted the robot scenario. Could you reveal any recent progress in this business segment?
Xie Ben: You may have recently observed that we have launched multiple encoders covering new products utilizing vernier algorithms, supporting off-axis installation, and based on eddy current sensing technology. These products are specifically designed for the robot market. The domestic robot market is developing rapidly, with many players entering. Particularly after our merger with MagnTek, we acquired a large portfolio of products suitable for robots, such as encoders and small servo systems. As the industry evolves, we have introduced encoder products in various forms, including on-axis and off-axis configurations, based on market research.
Currently, most domestic robot manufacturers have established partnerships with us, covering structural design, simulation advancement, through to project mass production. Overall, our encoder products have achieved significant market breakthroughs in the robot industry.
Huan Shi: It is evident that you have achieved phased results. The industry generally believes that precise control in robots relies heavily on angle encoders. Could you explain in simple terms the installation position, working principle of angle encoders, and how they ensure robot control precision?
Xie Ben: As the name suggests, an angle encoder is a product for encoding angles. Like a protractor used in school, an encoder is essentially a "protractor" on the robot's body. It is installed in arms, joints, hips, elbows, knees, etc., collecting angular information generated by mechanical rotation and displacement through principles such as Hall effect or eddy current sensing. Taking a robotic arm as an example, the joint contains a motor and an encoder. The "protractor" collects data on the arm's swing posture and movement position, transmitting signals back to the control system to complete closed-loop control.
Huan Shi: So the encoder tells the joint how many degrees to move, and the joint executes accordingly.
Xie Ben: For instance, if the system issues a command to grab a cup, the robotic arm might need to rotate 35° from a 180° position. The encoder provides real-time feedback on angle changes. The system combines target commands with actual feedback data to perform calculations, driving the mechanism to execute the corresponding action. This requires the entire control process to perceive angular positions with extreme precision.
Huan Shi: How large can the measured angle be? Is 360° possible?
Xie Ben: This depends on the product's structural design. Our products generally support full-range angle detection from 0° to 360°, but some structures may have mechanical limits, allowing only 180° or 90° of rotation.
Huan Shi: Angle measurement is actually a very traditional application field; I didn't expect robots to make it popular again.
Xie Ben: Correct. A robot is essentially a complex mechanical structure equipped with numerous small servo motors. The servo field is well-known; a servo control system includes encoders, motor controllers, control boards, etc. A robot simply integrates massive numbers of encoders and motors into a single system to execute various preset functions.
Huan Shi: Thinking about it, it seems complex. Human actions come naturally, but controlling machines using human-like logic is highly difficult.
Xie Ben: The technical implementation is indeed complex, especially since embodied intelligent robots carry numerous joint modules. The coordination and response consistency of each joint affect the overall posture control of the machine.
Huan Shi: There are many details to consider. This explains why some robot prototypes exhibit uncoordinated movements—the entire system is too complex.
Achieving High-Precision Angle Detection in Confined Joints
Huan Shi: NOVOSENSE has launched its first magnetic encoder based on Hall principle and dual-track vernier algorithm. Please explain: What does this "first" signify, what problems does it solve, and what value does it bring to the industry?
Xie Ben: The term "first" has two aspects. First, it is the domestic first—NOVOSENSE is the first company in China to launch an encoder based on differential Hall and dual-track vernier algorithms. Second, it is NOVOSENSE's own "first" device of this type. Although we have deep accumulation in the robot track and encoder industry, this product was timely introduced along with the development of the robot market to solve customer application issues. Its main characteristics are miniaturization, light volume, compact structure, high precision, and even low cost.
This product primarily tackles technical challenges in off-axis installation scenarios: reducing device size while ensuring installation flexibility and solving the structural issue of hollow wiring through motor shafts. It supports harnesses passing through the center of the motor shaft, which is a relatively new structural approach.
Huan Shi: This sounds like an innovative breakthrough. During development, where were the difficulties in R&D or mass production introduction?
Xie Ben: Sensors themselves are difficult because they need to sense extremely weak signals. This product first uses Hall principles to detect magnetic field signals from the code disk. The code disk itself has consistency deviations, and the size of 1-pole motors is also very small. Therefore, we must detect highly sensitive and high-precision magnetic field rotations and changes despite potential assembly fluctuations. Beyond identifying the signal source, we must also handle internal signal processing. Robots contain various interferences and noise internally; we use differential methods to address magnetic interference, electromagnetic compatibility requirements, etc.
Differential Hall Array: A sensor array composed of multiple Hall elements serving as the magnetic sensing unit. Through differential measurement principles, it suppresses common-mode interference such as external uniform stray magnetic fields and extracts the differential mode signal of the target magnetic field. This structure improves anti-interference capability, accuracy, and reliability, and can offset temperature drift and magnetic field offset errors. Differential Hall arrays are widely used in current sensing, angle measurement, and position detection, suitable for high-noise environments like motor control, automotive electronics, and industrial automation.
In the subsequent backend signal chain, including ADC, CORDIC computation, and angle resolution modules, everything relies on algorithmic IP accumulated by NOVOSENSE over many years. We integrate signal processing and angle resolution IP previously accumulated across various products onto a single chip to achieve high-precision angle detection.
Why a Product Matrix is Essential, and What is the Customer's Only Red Line?
Huan Shi: Outsiders can see that NOVOSENSE's encoder product layout is rich, covering on-axis, off-axis, Hall, and eddy current technical routes. What is your logic for product selection and definition? Is it driven by customer needs or technology trends?
Xie Ben: Our product layout is a comprehensive result of market demand, technological accumulation, and industry trend judgment. The robot industry's demand for encoders is not static—robot dogs have lower requirements for device size and precision; whereas embodied intelligent robots require small volumes for finger joints, high precision for elbow joints, and special structures like hollow wiring. No single solution fits all scenarios in the industry. Therefore, we meet most customer needs through different product combinations, technical routes, and structural designs.
Structurally, products are divided into off-axis and on-axis categories. If the encoder is installed at the motor shaft end, customers might choose on-axis products, supporting magnets paired with encoder control boards to work collaboratively. For example, our NOVOSENSE MT6835 is an encoder chip based on AMR (Anisotropic Magnetoresistance) principles; both NOVOSENSE MT6826S and MT6835 offer measurement resolutions up to 21 bits.
AMR (Anisotropic Magnetoresistance): Refers to the phenomenon where the resistance of strongly ferromagnetic metal materials changes with the angle between the magnetic field direction and current direction under an applied magnetic field, reaching maximum resistance when current and magnetic field are parallel. AMR sensors enable direct, non-contact, and wear-free angle measurement in a wide range of motor applications.
If customers insist on through-shaft installation without drilling holes in the housing, they may need off-axis products like NSM350x or MT6901. Different product features correspond to vastly diverse industry needs. NOVOSENSE provides diverse solutions for the industry relying on a complete product matrix.
Huan Shi: So, NOVOSENSE's ultimate goal is to provide diversified solutions for customers.
Xie Ben: Yes. No single solution solves all problems, but we conduct targeted product design for each application to resolve specific scenario challenges.
Huan Shi: Having interacted with frontline customers for a long time, could you share what points they struggle with during product selection and what point they cannot compromise on?
Xie Ben: Customers often seek small size, compact dimensions, hollow structures, easy installation, while demanding high reliability and controlled costs, hoping for a solution that excels in all aspects.
Huan Shi: This is a typical "want it all" demand.
Xie Ben: Exactly. But ultimately, regardless of route or product, system requirements must be met. The customer's non-negotiable red line is that the product must provide reliable and stable position feedback under various operating conditions. This is the red line. Structural considerations are addressed during the early design phase. However, once the product enters mass production delivery and terminal usage, as the robot's angle protractor, stable and reliable position feedback capability is a hard requirement that cannot be compromised.
From Automotive to Robotics: How Domestic Encoders Compete with Overseas Brands
Huan Shi: In our previous interview, we focused more on NOVOSENSE's layout in the automotive sector, but this year the topic centers on robotics. How do you view the changing demand for encoder products in these two markets? What is the degree of reuse, and how do their R&D requirements differ?
Xie Ben: The robot and automotive markets share commonalities but also have significant differences. Both relate to human safety: automotive chips ensure passenger travel safety; collaborative humanoid robots operate in close proximity to humans, with some products entering home scenarios. Many robot customers in the industry have raised extensive requirements for long-term lifespan, reliability, and stability. In fact, many robot brands choose automotive-grade products, such as those meeting AEC-Q100 Grade 0 standards, covering the -40°C to 150°C full temperature range, and possessing high reliability and stability.
AEC-Q100 Standard: An automotive-grade integrated circuit reliability test standard formulated by the Automotive Electronics Council (AEC), aiming to ensure chips operate stably for long periods in harsh automotive environments. It is a key threshold for entering the global automotive supply chain. Its core classification basis is the chip's operating temperature range, divided into four grades (Grade 0 to Grade 3). Grade 0 covers -40°C to 150°C, suitable for chips requiring the most stringent temperature tolerance.
Differences mainly manifest in product design and application. Products in robot joints are smaller and more compact, sharing space with magnets and motors. Installing rotating targets, encoders, motor driver boards, and motor controllers in such confined spaces poses huge environmental challenges, including motor heat generation and electromagnetic interference.
Additionally, the iteration speed of domestic robots is far faster than the automotive industry. Traditional automotive product iteration cycles typically take two to three years. Even with accelerated rhythms today, there remains a clear gap compared to the robot market due to frequent changes in mechanical structures and control logic updates.
The robot market demands high reliability, high immunity to interference, and high stability. NOVOSENSE has experience operating in the automotive industry and places great importance on its development. Experience in automotive-grade product design, testing, and validation can be applied to robots. Leveraging NOVOSENSE's diverse product portfolio of sensors, drivers, MCUs, combined with deep understanding of industry scenarios and automotive design experience, we are positioned to develop distinctive products based on existing expertise to meet market needs.
Huan Shi: Thus, all your past accumulations are useful.
Xie Ben: Yes, experience is reusable.
Huan Shi: Currently, how many product directions are being laid out simultaneously in your existing product lines?
Xie Ben: Within the sensor category alone, there are 5 product directions, indicating a substantial product planning scope.
Huan Shi: Domestic encoders inevitably face direct competition with leading overseas manufacturers. Are current robot customers willing to try domestic brands? Do they impose stricter requirements on domestic devices compared to overseas ones?
Xie Ben: We do face competition from overseas brands. Frankly, customers choose NOVOSENSE because our products solve their application problems. Based on data released by the globally renowned institution Yole Group, NOVOSENSE's shipment volume ranked fourth globally in 2025, and we hold a leading position in the domestic magnetic sensor market. Many of our products are already used in home appliances, robots, servo systems, photovoltaic inverters, and other fields. Every product exists to meet genuine customer needs.
Customers select us based on our prior technological accumulation and strong market performance. Currently, we can provide full solutions in most robot application scenarios. We have high-resolution products for on-axis solutions, as well as devices supporting off-axis hollow perforation. We offer high-resolution products at 21-bit and 22-bit, as well as 14-bit and 15-bit products for varying cost needs. We have market-tested solutions in every niche.
After years of industry accumulation, there is no substantive technological generational gap between domestic encoders and overseas products. Overseas manufacturers have longer histories, but domestic devices possess prominent competitive advantages in overall dimensions such as market performance, cost-effectiveness, and comprehensive performance, particularly for NOVOSENSE's encoder products.
Behind Thousands of Mass-Produced SKUs: How to Decide Whether to Develop a Product?
Huan Shi: NOVOSENSE has become a local semiconductor enterprise with certain influence in the industry. Facing customer needs, how do you judge whether to develop a product, and what trade-off logic underlies this decision?
Xie Ben: Our judgment on whether to develop a product is based on a comprehensive assessment of market and customer needs. We evaluate whether the needs proposed by customers are representative of the industry, then use system-level understanding to determine the background reasons, distinguishing whether the issue stems from insufficient product capabilities or a unique application direction. If customer demands represent common pain points across the industry, such products definitely have good development potential.
Huan Shi: But if the market hasn't formed a unified consensus, is judging demand more difficult?
Xie Ben: The company has many roles; Marketing looks at the market and needs, while colleagues conduct system analysis. We have many products, possibly thousands in mass production. Some meet current market needs, while others target future development needs. We do not simply chase market opportunities to initiate projects; the decision to develop or not depends on internal priority assessments. Layout falls into two categories: one based on judgments of future trends, and another based on current needs, continuously providing mature solutions. These two routes run in parallel.
Huan Shi: Could you make some predictions for the sensor industry? What technical directions and application scenarios are currently overlooked but possess broad implementation prospects in the future?
Xie Ben: Including dual-vernier encoders and eddy current products, our R&D teams in every product direction continue to validate new technologies, assess demos, and utilize FPGAs, constantly exploring differentiated sensor technologies. We always have many projects in R&D, some targeting the future, with related achievements to be announced later.
Huan Shi: It is understandable that industrial layout must match one's own development rhythm. NOVOSENSE's growth journey is a typical sample of the striving development of domestic sensor enterprises.
Xie Ben: Besides observing changes in industry demand, we must also track the progress of customer projects. The decision behind whether to develop a product is not purely a technical issue but requires a comprehensive evaluation of actual market conditions to complete priority assessments.
Huan Shi: One needs to distinguish between genuine industrial demand and pseudo-demand.
Xie Ben: Some demands may materialize ten years from now. We still need to combine technological accumulation with the advancement of customer market projects. If a product is made but there is no market demand, it can only wait for implementation or degrade into a low-value product; conversely, if there is market demand but the product isn't ready, it is a waste of timing. Therefore, we combine both factors comprehensively for product development and release.
Huan Shi: Each year we talk, we see NOVOSENSE making progress. I look forward to seeing more groundbreaking new products unveiled at next year's interview.
Xie Ben: Definitely.
Interview | Huan Shi Editing | Xiao Qie Review | Huan Shi