Achieving full synergy immediately after integrationLeveraging Chiplets for Analog-Digital Hybrid ----
[ROHM x LAPIS] Collaborate on Unique MCUs

ROHM, which merged with subsidiary LAPIS Technology in April 2024, is accelerating the development and deployment of products that combine the strengths of the two companies. ROHM cites MCUs as one of the categories with the greatest potential synergy. A lineup of unique new products is ready for market launch, including an MCU for power supplies featuring analog-digital hybrid control that leverages the best of both topologies, a type utilizing chiplets, and a model capable of performing on-chip machine learning and inference.

On April 1, 2024 ROHM merged with wholly owned subsidiary LAPIS Technology, giving rise to the new ROHM. Combining ROHM's strength in analog/power semiconductors with LAPIS Technology’s expertise in logic/MCUs will result in a robust management structure capable of succeeding in the fiercely competitive semiconductor industry which is undergoing significant changes in the business environment. Going forward, we will accelerate the development and deployment of products that leverage the technologies of both companies.

A prime example of such products are MCUs. Hiroshi Fukuyama, General Manager of ROHM’s LSI Business Division and LAPIS’ LSI Business Unit, emphasizes that MCUs are the category that is most likely to benefit from integration. In this article, we will introduce three new MCUs with unique features, with sample shipments slated to begin in 2024.

‘Digital Control’ Power Supplies for the Medium Power Domain

ROHM’s LogiCoA^™*1 MCU for power supply control integrates the necessary analog and digital circuits into a single chip. This makes it possible to achieve the same functionality as digitally controlled power supplies at the lower power consumption and cost levels of analog power supplies.

1: LogiCoA^™ is a brand that embodies a design philosophy of fusing digital elements to maximize the performance of analog circuits. LogiCoA^™ is a trademark or registered trademark of ROHM Co., Ltd.

For power supplies there are two types of control, analog, which uses an analog controller, and full digital that involves an MCU or other digital device. While full digital control offers precise control, a high-speed CPU/DSP is required, which is costly and consumes significant power.

As a result, it is often used in applications with high power demand (1kW or more), such as solar power generation systems and EV chargers where the output power fluctuates greatly and analog control alone may struggle to cope. Aside from these applications, analog control power supplies characterized by low cost and power consumption are mainstream in the low to medium power range (50W to 1kW class), such as PCs, robots, factory automation (FA), and manufacturing equipment. However, circuits configured with analog semiconductors are difficult to modify once the circuit is constructed, which can become a barrier when developing derivative products.
This is because the power supply circuit cannot be used as-is. ‘Analog-controlled power supplies are designed to be optimized for a specific application, so even slight changes in requirements such as the supply voltage or output power necessitate readjustments to the power supply circuit. This is a significant burden for designers,’ says Fukuyama.

Another issue is the efficiency of power conversion (AC-DC). Analog control does not allow for fine adjustment of the output according to the power used. For example, robot arms are typically designed to maximize power efficiency at the ‘maximum required power’, regardless of whether the arm is moving large distances (high power) or small distances (low power). As a result, power efficiency becomes poor when operating at low power levels.

For these reasons, the demand for digital-controlled power supplies in industrial equipment in the low to medium power range is increasing, but as mentioned above, there are challenges in terms of power consumption and cost. The LogiCoA^™ MCU was developed to solve these issues. In addition to a low-cost low-speed CPU, a 3ch analog comparator is built in along with a DA converter for digital parameter control. Changing the parameters of the MCU program on a PC makes it easy to adjust the output power and other settings. In the case of the robot arm, the output power can be dynamically adjusted as needed, improving efficiency.

At the same time, repurposing power circuits becomes easy. Since the analog components remain unchanged, adjusting the voltage and current values can be performed by simply changing the MCU parameters. Reusing analog power circuits makes it easier to develop more derivative products. With the need to shorten development times, this could provide a significant advantage for designers.

ROHM will provide software for the LogiCoA^™ as a reference design after some degree of customization based on the user’s application. ‘Customers can then immediately begin usage by simply changing the parameters to suit set requirements.’

‘Fully digital controlled power supplies require high performance MCUs, which are costly. The LogiCoA^™ MCU is designed to meet the needs of those seeking more precise power control without incurring high costs.

We expect the LogiCoA^™ MCU to expand the range of devices that can benefit from digital control in the small to medium power range.

Mass production shipments of LogiCoA^™ MCUs began in June 2024.

LASCA^™ Chiplets

The second is a new brand of MCU utilizing proprietary chiplet technology called LASCA^™ (LApis Scalable Chiplet Architecture). This technology combines a high-performance MCU chip based on the 22nm process with a high-performance analog chip for specific applications using 130nm/150nm legacy processes in a single package.

LASCA^™is designed to reduce the time and cost of SoC development. Developing a System-on-Chip (SoC) using the same 22nm process as the MCU integrated in LASCA^™ would entail several years of development time and enormous costs.
When using advanced processes, a certain scale of mass production is required to achieve cost benefits, but such applications are limited, particularly in the industrial sector.

By customizing only the analog chip, LASCA^™ can achieve MCUs with functionality equivalent to 22nm SoCs while significantly reducing development time and costs. ‘Since only custom chip development is necessary, the development period could be potentially shortened to within a year. It also carries lower risk than developing an MCU from scratch.’

And as an added value compared to 22nm SoCs manufactured using a single process, we emphasize the ability to incorporate high-performance, high voltage analog circuits. This makes it possible to integrate 5V-operable AD/DA converters, high voltage sensor control AFE (Analog Front End) circuits, and even high voltage process gate drivers. The MCU chip incorporates an Arm Cortex-M33^*2 core featuring an operating frequency of 300MHz and up to 1MB of MRAM. ‘A variety of functions are also built in that adds versatility. The idea is to be able to develop a 22nm SoC while significantly reducing development costs.’

As the first product in the LASCA^™ series, ROHM developed an MCU with an analog chip optimized for motor control in consumer and industrial equipment applications. Support for 5V power supply operation is enabled, along with high-speed feedback control that improves system power efficiency. ‘Designed for motor control, the 5V analog chip features excellent noise resistance. Going forward, ROHM will continue to expand the LASCA^™ lineup by incorporating analog chips optimized for customer applications.’

Chiplet technology offers significant advantages for ROHM as well. Total manufacturing costs can be reduced by using processes suitable for both MCU and analog chips. And in the future, there is a strong possibility that advanced processes beyond 22nm will be used for MCUs.

Samples of the first LASCA^™ products will be available in the fall of 2024, with mass production scheduled for the spring of 2025.

An AI MCU Capable of Performing On-Chip Learning and Inference

The third product, a ‘standalone AI MCU’, is a single-chip solution that combines ROHM’s ultra-low power AI chip with a 32bit Arm MCU. As the name ‘standalone’ suggests, this MCU can perform machine learning and inference independently.

The AI chip is equipped with a hardware accelerator developed by ROHM, based on an ‘on-device learning algorithm’ devised by Professor Hiroki Matsutani of the Department of Information and Computer Science, Faculty of Science and Technology, Keio University. Learning and inference are possible at an ultra-low power consumption of just a few tens of mW. ROHM announced this AI chip as ‘Solist-AI^™*3’ in September 2022. With standalone AI, the AI chip handles machine learning-related processing, while the Arm MCU manages essential tasks like controlling FA equipment.

This standalone AI MCU is initially targeted for predictive maintenance applications in industrial equipment, such as FA (Factory Automation). Depending on the location, FA equipment in factories operate in environments with varying temperatures and vibrations. Retrofitting FA equipment with the standalone AI MCU will allow users to collect data on current, vibration, and other parameters to establish a baseline and issue alerts when unusual behavior is detected. However, many factories are reluctant to connect to the cloud due to concerns over security, latency, and communication. In such environments, the ability to perform not only inference but also learning directly on the device is a significant advantage, states Mr. Fukuyama.

Sample shipments of the standalone AI MCU is slated to begin in the fall of 2024.

The MCUs introduced here are all unique products that leverage the combined strengths of ROHM and LAPIS Technology. ‘The results of the integration of the two companies are already evident,’ says Mr. Fukuyama. While LAPIS Technology offers a broad lineup, including communication ICs, MCUs are the products most likely to yield the greatest synergy with ROHM’s power and analog technologies. Therefore, we will continue to accelerate the development and market introduction of such products in the future,’ he continued.

2: Arm and Cortex are registered trademarks of ARM Limited (or its subsidiaries) in the EU and other countries.

3: LASCA^™ and Solist-AI^™ are trademarks or registered trademark of ROHM Co., Ltd.