User:Clintroymkt/sandbox/Full-Stack Microcontroller

Full-Stack Microcontroller

Full-stack MCU The term “stack” was interpreted early as a collection of the entire technology required in network development, the concept of “full-stack engineer” that was put forward simply describes an excellent talent with overall thinking, comprehensive technology, and quick problem solving abilities.

A full-stack MCU is also called a full-stack self-developed MCU, which requires hard technology and soft skills. Hard technology means that the core IP and professional IP that are technically difficult are independently developed and integrated by the original MCU vendor. Here, core IP refers to processor IP, such as RISC-V core, 8051 core, etc. Professional IP refers to high-speed interface IP, wireless communication IP and protocol stack, etc. Soft skills refer to the development resources related to this MCU, including data files, source codes, development tools, debugging tools, etc.

Motivation for design A classic MCU generally adopts the industrial division of labor similar to that in the PC field, that is, an SOC model based on IP integrated assembly. The main components include core IP, basic digital and analog peripherals, and professional peripheral IP. Among them: Core IP refers to the processor. Generally, microprocessor core IP licenses are purchased from third parties such as ARM or MIPS. There are multiple types and levels that are related to overall performance. Basic digital peripherals include timers, serial ports, SPI, PWM, GPIO, full-speed USB, etc., and the technology is simple and generally self-developed. Basic analog peripherals include oscillators, ADC, DAC, operational amplifier, power management, etc., and the technology is simple and generally self-developed. Professional peripheral IP refers to peripherals that have relatively professional technology, include controllers and high-speed analog transceivers, involve radio frequency and even baseband algorithms, and require communication protocol stacks and even drivers, such as high-speed USB, Ethernet, WIFI, Bluetooth, etc. Especially for peripherals with high-speed transceiver PHY. Generally, MCUs have no built-in controller or only have built-in controller, and the customer purchases an external PHY chip to connect, or the original MCU manufacturer purchases the third-party IP to be built in, and the manufacturer may pay Million-level licensing fee and IP commission fee.

Now is a more intelligent and digitized era, and everyone is pursuing the Internet of Things with interconnection of all things and interoperability between upper and lower level systems. Networking, high-speed communication, large data storage, wireless and other applications have continuously improved requirements for MCU performance and functions, and the demand for MCUs with comprehensive functions, high integration and high cost performance has increased significantly. Gradually, the traditional MCU assembly method has shown its drawbacks:

Original manufacturers of MCU core IP have already developed processors that support clock frequency over one hundred, floating point, high-speed computing, etc. And the performance is comparable to that of 486 CPU, and it is more and more like a PC. But its powerful functions and versatility may be wasted in many applications. For example, because the peripherals of other manufacturers do not meet the same requirements and it is needed to be connected through frequency reduction, or it cannot serve dedicated peripherals as a general IP. The market demands more professional interfaces, such as Ethernet and wireless interfaces. Most MCU original manufacturers need to purchase third-party IP or require users to add external chips, which virtually increases the cost of end products.

In this case, full-stack self-developed MCUs can be applied in the market more flexibly, realizing integrated chips, which is conducive to global optimization for applications, reducing intermediate links and EMI, and there is no need to pay authorization fees and commission fees to third parties. The cost of the whole device is reduced, and it is completely autonomous and controllable. It can be used for products in special fields as well as for the development of long-term products. The technical cost of a full-stack MCU is undoubtedly very large, but technology and resources are a turning point today.

1. As the core instruction set of MCU, the existing open-source and general simplified instruction set RISC-V can become the mainstream instruction set. On this basis, the original manufacturer can customize development for unused applications, and achieve the best cost and dedicated control. 2. Iterative technology updates, and electronic manufacturing technology upgrades. It gives integrated circuits advantages in size and power consumption. And more complex IPs can be integrated. 3. With unified communication specification and form, the powerful standard interfaces replace the inefficient and partial interfaces, and the integration of a small number of peripherals can realize more applications.

Development Process To develop a full-stack MCU, it is necessary to develop both hard technology and soft skills. 1. A certain cost investment. Original manufacturers should increase technical capabilities and accumulate experience, and the early R&D cost is relatively high. 2. The integration of load resources by excellent architects is required. The connections of various modules, clocks, and power sources inside the MCU are globally monitored. 3. A large number of customer groups can involve more fields.