0760-22278043
32-bit MCU, all competing for the smallest

Your next 8-bit MCU is 32-bit. This is a sentence that ST has said before.


In the MCU field, some people roll up performance, while others roll down size and price. After all, MCU selection is a "just right" skill, and if problems can be solved at a lower cost, there is no need for more expensive and larger ones.


The MCU and CPU fields are different, the strongest performance does not necessarily mean the best, and excessive performance is also a shameful behavior. Every embedded engineer should be a Virgo, playing MCU performance to the extreme. So, there are still many veterans who will not give up their 8-bit and 16 bit MCUs.


However, as an MCU manufacturer, we still need to constantly improve our product line. So, manufacturers have been constantly competing on this track.


32-bit MCU, constantly making small and cheap
TI is not only competing for the smallest size, but also for higher cost-effectiveness, so it has already laid out Cortex-M0/M0+cores and launched the MSPSM0 series products. The MSPM0C1104, launched in March last year, was considered the smallest MCU at the time. It was packaged in wafer level packaging (WCSP) and had an area of only 1.38mm ², equivalent to a black pepper.


In terms of configuration, MSPM0C1104 is equipped with a 24MHz Arm Cortex-M0+core, 1KB SRAM, 16 KB memory, 1.62 V~3.6 V power supply voltage range, 3-channel 12 bit ADC, and 6 general-purpose input/output pins. Integrated with high-speed on-chip oscillator, with an accuracy range of -2% to+1.2%, without the need for external crystal oscillators. Compatible with standard communication interfaces such as UART, SPI, I2C, etc., with a working temperature range of -40 ° C~125 ° C.

 

图片

 

TI's MSPM0C1103 is also a cost-effective product, priced at around $0.25, with a minimum package size of 1.6 x 0.86mm, a 24 MHz M0+core, 8 KB Flash, and 1 KB RAM. Excellent simulation performance, suitable for scenarios such as sensors and power monitoring where accuracy is higher than speed.

 

图片

 

ST has always been obsessed with replacing 8-bit MCUs. The STM32C0 Cortex-M0+launched by it starts at $0.21, and some products have a minimum package size of 1.70 × 1.42mm. At the same time, it provides as many I/O resources as possible within a limited size, and only requires one pair of power pins for power supply. STM32C0 is also included in ST's 10-year long-term supply plan and supports IEC 61508 functional safety, providing more assurance for products that require long-term stable supply.
With the continuous expansion of product models, STM32C0 also offers different functional options. For example, STM32C051 increases the Flash capacity to a maximum of 64KB; STM32C071 adds a USB connection; STM32C091 further expands to a maximum of 256KB Flash and 36KB SRAM; STM32C092 supports FDCAN communication.
In terms of hardware design, STM32C0 maintains partial pin compatibility with STM32G0 and integrates an internal clock with an accuracy of 1%, which helps simplify peripheral circuit and PCB design.

 

图片

 

Subsequently, ST launched the STM32C5 of the Cotex-M33 in March this year, based on ST's 40nm process platform, with a maximum clock frequency of 144MHz and CoreMark performance of 593 points. It can integrate up to 1024KB Flash and 256KB SRAM, and is equipped with key peripherals such as Ethernet, OctoSPI, FDCAN, etc. Offering options from 20 pins to 144 pins, supporting a maximum operating temperature of 125 ° C, it can cover a wide range of mainstream application scenarios.

 

图片

 

Of course, when it comes to being cheap and compact, we cannot avoid the RP2040/RP2350. RP2040 dual core Cortex-M0+, 133 MHz, 264 KB SRAM, requires external Flash. RP2350 has stronger performance and its production price is still below $1. The development board Pico/Pico 2 has comprehensive support and a large community. Suitable for low-cost products that require high computing power.

 

图片

 

Renesas' RA0E1 is a 32-bit solution suitable for pursuing cost-effectiveness and long-term stability, with a price of approximately 0.6~0.7 US dollars per thousand pieces, Cortex-M23,32 MHz,64 KB Flash,12 KB RAM。 Supports TrustZone security and industrial grade reliability.

 

图片

 

We are also doing small things domestically. Hangshun announced last year that the packaging area of the 32-bit MCU HK32F005 is 1mm ². In June of this year, Hangshun continued to promote HK32F001, using WLCSP (wafer level chip size packaging) technology to further compress the packaging area to 0.6mm ², setting a new record. In addition, Hangshun has also released a small-sized M4 MCU HK23F403.
WLCSP can be simply understood as a technology of "packaging to chip size". It does not wait for the chip to be cut before subsequent packaging, but directly completes the relevant processes on the entire wafer, and then cuts it into individual chips. This reduces the need for multiple processes and chip handling in traditional packaging, and also helps improve production efficiency. There is basically no obvious packaging space left outside the chip, so the length and width of the final device are very close to the bare chip. Compared to common packaging methods such as SOP and QFP, its advantages in volume and weight are very obvious, making it particularly suitable for products that are sensitive to PCB space.

 

图片

 

In the cheap and compact race track, Pulan has always been a difficult mountain to surpass. PY32F002B is the most typical product, priced at around 8 cents, with a 24MHz M0+core, 24 Kbytes Flash, and 3 Kbytes SRAM.

 

图片

 

The CH32V003 from Qinheng is a very cost-effective MCU on the RISC-V track, with a price of about $0.1 per thousand and an extremely low price. The RISC-V is 32-bit, 48 MHz, 16 KB Flash, and 2 KB RAM.

 

图片

 

Controversy between 8-bit and 32-bit
Compared to 8-bit MCUs, 32-bit MCUs have multiple advantages. For example, it has stronger processing power, wider memory bus, and provides UART, USB, Ethernet, etc CAN、SDMMC、 Integrate more advanced communication interfaces such as LCD controllers, package more instructions in each cycle to achieve faster execution speed and lower power consumption, and access various libraries and well maintained drivers at any time.
However, 32-bit MCUs are not omnipotent. Achieving optimal low-power operation is much more complex than simple 8-bit MCUs, making programming and debugging more difficult. For applications with limited memory, the code may be larger than 8-bit MCUs. Of course, these are not the key points, the most important thing is still the price.
Some engineers have expressed that the 50 cent 32-bit M0/RISC-V from Qinheng and Puran, which supports Arduino, feels that 8-bit has no meaning anymore. The power consumption can now be achieved with 32-bit technology, and the advantage of 8-bit technology is almost non-existent. Domestic M0 models can cost less than 1 yuan, while for consumer products, industrial products can cost less than 2 yuan. And now more and more devices need to be connected to the network, so 8-bit is not enough.
Some people also say that there are many 8-bit MCUs, but there are not many that are still called "MCUs". Even more extreme, it means that 32 bits are already priced like cabbage, and we firmly do not use 8 bits.
Opponents say, even if it's cheaper, do you have a 50 cent 32-bit? They have a light control MCU for 0.5 yuan, and you're using M0? Of course, some people say that there are really large quantities of CM0, but 8-bit machines with very few legs like PIC OTPs can indeed be rolled up. Similarly, there is also a calculator with 4 bits of cow dung. This extreme is not discussed, and there is no need to compare ordinary 8-bit machines with CM0.
Now that 32-bit MCUs are rolled up like this, will you still use 8-bit MCUs? Welcome everyone to discuss together.