L2+ Autonomous Cars: Discrete IMU Module or Integrated Solution?

Overview

As auto and Tier-1 suppliers begin the design of the next generation of ADAS domain controllers, one of the most critical architectural decisions for the L2+ highway assist function is whether it is better to use a separate discrete 6-axis IMU module or to take a more integrated approach and mount the raw IMU sensor die directly onto the main domain controller PCB.

This report offers a detailed analysis of this choice, in particular of the trade-offs between Bill of Materials (BOM) cost, integration complexity, development effort and total risk for each approach.

1.0 Introduction: The Important Function of the IMU in L2+ Systems

The movement towards increasing levels of driving automation in the automotive industry is unstoppable and the development of Level 2+ (L2+) Advanced Driver Assistance Systems (ADAS) has emerged as a crucial differentiator in the passenger vehicle market. L2+ systems, especially those that provide “Highway Navigation Assist” or “Navigate on Autopilot” features, do not stop with simple lane keeping and adaptive cruise control. They can do automated lane changes, bypass interchanges, and give a better semi-autonomous experience. The reliability of these systems depends on an accurate, continuous, and high-integrity knowledge of the state of the vehicle, i.e. its position, orientation, and dynamics.

While Global Navigation Satellite Systems (GNSS) offer the main source of absolute positioning, they are prone to signal degradation and signal outage in urban canyons, tunnels and dense foliage. In order to fill such gaps and to supply the high-frequency motion data that is needed for smooth and safe vehicle control, the Inertial Measurement Unit (IMU) is invaluable. A 6-axis IMU (which includes a 3-axis accelerometer and a 3-axis gyroscope) provides important information on linear acceleration and angular velocity which is the backbone of the dead reckoning and sensor fusion engine within the vehicle.

2.0 The Discrete IMU Module A Path of Predictability and Lower Risk

The discrete module approach means that one purchases a pre-packaged, pre-calibrated and fully qualified IMU component from a specialized supplier. This is a module considered a “black box” system which is integrated on the main ADAS domain controller board.

2.1 Specification of the Discrete Module

An automotive grade discrete IMU module represents a highly complex subsystem by itself. Its salient features usually consist of:

• MEMS Sensor Dies: This sensor is a 3-axis accelerator and a 3-axis gyroscope co-packaged to ensure maximum alignment.
• ASIC/Microprocessor: An onboard processing unit which reads raw sensor data, implements factory calibration and temperature compensation algorithms and possibly makes the first sensor fusion.
• Firmware: Software in the ASIC that controls the sensor as well as provides a standardized data output.
• Packaging: A strong package that is hermetically sealed and is meant to guard the delicate MEMS components against the environment and mechanical forces.
• Interface: A typical communication interface like the SPI or the I2C to connect it to the host processor (the ADAS SoC).

Mainstream suppliers such as STMicroelectronics include automotive grade modules such as the ASM330LHB designed specifically to be used in vehicle dynamics and qualified with functional safety standards including ASIL-B.

2.2 Cost Analysis of Discrete Module Approach

Discrete module approach cost structure is skewed towards the initial price of the component with much lower non-recurring engineering (NRE) and integration costs.

• Bill of Materials (BOM) Cost: The most visible cost is the per-unit cost of a discrete module. In the case of an automotive-grade 6-axis IMU module to be used in L2+ applications, prices may be different depending on performance, temperature range, and functional safety certification. Basic consumer-grade IMU chips may cost a few dollars but full-qualified automotive modules are much more¹. The approximate range of high-volume pricing of high-accuracy, AEC-Q100 qualified modules with ASIL-B certification, such as the STMicroelectronics ASM330LHB, is in the range 8 to 15 USD . Even higher prices can be charged on premium performance modules. This price is a summation of the costs borne by the supplier of the sensor die, the development of the ASIC, elaborate packaging as well as complex testing, and the final key step of factory calibration².
• Integration and Development Cost (NRE): This is where the discrete module approach has a great advantage.
  • Hardware Integration: The amount of effort is minimal. The hardware team of the ADAS domain controller has to simply reserve space on the PCB as well as provide a clean power supply and route the standard communication bus (eg SPI) to the ADAS SoC. The “black box” nature of the module abstracts away all the complexity of the inside.
  • Firmware Development: The effort is not that high. Calibrated and temperature-compensated data are output from the module. The main duty of the ADAS SoC firmware team is to write a driver that will communicate with the module through its normal interface and parse the clean data stream. This removes the need for the OEM/Tier-1 to design complex and low-level sensor correction and calibration algorithms.
  • Validation and Certification: This is a major cost and time saver. The module is supplied pre-certified by the supplier to automotive standards such as AEC -Q100³. The OEM/Tier-1 validation effort is lowered to system level integration testing as opposed to an arduous and expensive process of qualifying a raw sensor component from scratch.

In a summary, the discrete module provides a forecastable, but more expensive, BOM price with a significantly shorter development cycle, less NRE investment and less risk with the project.

3.0 The Integrated Solution: Pursuing BOM Savings at the Cost of Complexity.

The integrated approach, which involves soldering a raw MEMS IMU sensor chip directly on the PCB of the ADAS domain controller like any other component. The major reason this strategy is pursued is to eliminate the price of packaging of the module, internal processor, as well as the margin the supplier demands, thus reducing the cost of the per-unit BOM. However, this imposes an enormous amount of complexity and responsibility on the integrator.

3.1 Complexity of Integration: A Complex Engineering Challenge

Integrating a raw IMU sensor onto a complex, high density digital board such as an ADAS Domain controller is full of challenges.

• Hardware Design and PCB Layout: This is not an easy task that can be done without extensive knowledge.
  • Sensor Placement: MEMS gyroscopes and accelerometers are ridiculously sensitive to mechanical stress, temperature gradients as well as vibration. Putting a small sensor on a large PCB filled with powerful and heat-generating SoCs and switching power supplies is a crucial challenge. It needs to be in a thermally stable and mechanically isolated position so that its performance is not corrupted⁴.
  • Power Supply Integrity: IMU sensor needs an extremely clean and stable power supply. Noise on the sensor analog front-end can be due to any noise on the digital logic or power management ICs (PMICs) on the domain controller, and this reduces the noise floor and performance of the sensor. This requires the use of a dedicated, low-noise LDO, careful filtering and careful power plane design to separate the power rail of the IMU⁵.
  • Signal Integrity: SPI/I2C high-speed communication lines connecting the IMU sensor and the main SoC need to be properly routed in controlled-impedance and shielded in order to avoid crosstalk and pickup of noise along other tracks which is a major problem on a dense, multi-layer board⁶.
  • Mechanical Stress: Soldering activity may cause stress on the sensor package which causes a change in its bias (zero-offset). Moreover, the difference between the Coefficient of Thermal Expansion (CTE) of sensor package and PCB material may result in stress in over-temperature cycling and thus in performance drift during the lifetime of the vehicle. This involves special PCB design and control of manufacturing process.
• Firmware Development and Calibration: This is a titanic engineering process and is probably the single biggest obstacle.
  • Raw Data Handling: The primary ADAS SoC is now charged with the role of reading raw, uncompensated data at high rate made available to the sensor.
  • Development of Algorithms: The integrator should develop, code and test advanced algorithms on the ADAS SoC to execute roles that are currently performed by the in-house processor of the module. This involves the compensation of:
    • Bias and Scale Factor Errors: Correcting the inaccuracy of the sensor.
    • Temperature Dependency: This is done to implement complex models when the bias and scale factor changes with the temperature of the board.
    • Axis Misalignment and Cross-Axis Sensitivity: Compensating against the effect of an imperfect sensor die.
  • Production Line Calibration: This is one of the major costs of operation. Since each raw sensor and each PCB assembly is slightly different, each individual domain controller board must pass a calibration procedure at the end of the manufacturing line . This generally involves setting the board on a multi-axis turn-table or an accuracy of motion fixed to have the sensor driven in order to calculate its individual calibration parameters. This is a significant added cost, complexity, and cycle time of the manufacturing process.

3.2 Economical Analysis of the Integrated Approach

While the savings of the BOM are the main driving force, a total cost analysis shows a different picture.

• Bill of Materials (BOM) Cost: The unit cost of raw automotive grade 6-axis MEMS sensor chip is indeed much cheaper than the complete module, possibly in the range of 2.50 to 5.00 USD. This is the most important attraction of this approach.
• Non-Recurring Engineering (NRE) Cost: The Non-Recurring Engineering (NRE) cost is very high. It includes thousands of person hours of engineering for:
  • Specialized hardware engineers to design and validate the complicated layout of the PCB.
  • Sensor fusion and signal processing experts to manufacture and test the calibration and compensation algorithms.
  • Manufacturing engineers for the design, procurement, and implementation of the production line calibration station.
  • Massive validation and functional safety qualification campaign of the complete integrated system.
• Lifecycle and Risk Cost: The risk is entirely shifted from the IMU supplier to the OEM / Tier-1. Any long-term performance drift, reliability problems or flaws in the calibration algorithm are the responsibility of the integrator. A design flaw that was found late in the development of the vehicle could cause a costly redesign of an entire domain controller, resulting in extensive program delays. The trend toward providing domain controllers with positioning hardware is recognised but the inherent challenges are significant⁷.

4.0 Comparative Analysis and Recommendation

The decision of whether to use discrete module or an integrated solution is a conventional trade off between unit BOM cost and total cost of ownership, which encompasses the cost of development, risk, and time to market.

Although there are sources that report the incorporation of the IMU as part of the domain controller is a growing-rapidly trend in order to fulfill the high-precision and stability requirement. This probably refers to a slow evolution and is only possible for organizations with huge scale and special expertise. Even case studies that boast of huge cost-reductions of integrated solution refer to extremely optimized modules or system-level packages, not necessarily raw chip-on-board designs.

Recommendation of L2+ Highway Navigation Assist:

For the vast majority of automotive OEMs and Tier-1 suppliers developing L2+ systems the discrete IMU module is the unequivocally superior option.

The justification is based on holistic consideration of cost and risk. The potential savings of a few dollars on the unit BOM pales in comparison to the enormous NRE investment, the extended development timeline, the added complexity in the manufacturing process and the tremendous technical and financial risks associated with the integrated approach. In the highly competitive world of ADAS, speed, reliability, and predictability are of the utmost importance. The discrete module approach provides all three at a cheaper lower-risk way of achieving a high-performance, certifiably safe system.

The integrated chip-on-board strategy should only be considered by vertically integrated players with multi-million unit per year volumes for a single platform who have world-class, in-house MEMS sensor and algorithm expertise and are willing to make a multi-year, multi-million dollar NRE investment to amortize the BOM savings over a very long production run. For everyone else, it is a high-risk gamble with a questionable return.

5.0 Conclusion

In the architectural decision of L2+ highway assist systems, the option to choose either discrete IMU module or integrated IMU solution is a clear cut. The integrated route has a temptation of a unit BOM cost that is lower but hides an entire mountain of hardware design, firmware development, and manufacturing costs and risk. It puts the responsibility to make sensors perform and be reliable in the hands of the integrator completely.r.

On the other hand, discrete IMU module approach offers a strong, predictable and de-risked path. With the specialization of a supplier at IMU, automakers and Tier-1s would be able to purchase a performance-guaranteed, pre-calibrated and pre-certified component. This enables them to devote their own precious engineering capabilities to system-level integration and application-level software functionality which are the actual distinguishing features of their ADAS product. This increased unit cost of the module is a small and prudent price to pay in order to have a quicker time-to-market, less risk in the development, and the confidence that there will be a trustworthy and safe system of vehicle dynamics sensing.

1. “一文了解IMU原理、误差模型、标定、惯性传感器选型以及IMU产品调研(含IMU、AHRS、VRU和INS区别),” 2021
2. “Bosch Sensortec BMI160 6-Axis MEMS IMU，” P. 1, Distributed by Yole, 2015
3. “ASM330LHHTR 汽车 6 轴惯性测量装置 (IMU) 模块,” Digikey
4. TDK InvenSense, “IMU PCB Design and MEMS Assembly Guidelines for ICM/IAM/IIM-4xxxx, 2xxxx and MPU-6xxx Products,” 2025
5. WZK, “PCB和信号完整性,” P. 13
6. “ADAS 域控制器硬件设计】TJA1102 硬件设计指南,” 2020
7. ResearchInChina, “Automotive High-precision Positioning Research Report,” P. 9, 2023