Solutions

Recommended parts

Why these parts

• WCH CH32V203: 144 MHz, 49 µA/MHz, dual USB, CAN; MounRiver Studio and an RT-Thread BSP are both available, making it a good course workhorse.
• GD32VF103: a mature Nuclei N200 ecosystem with plenty of examples and boards, good for side-by-side teaching and porting labs.
• CH32V103: 5 V supply and USB FS, compatible with 5 V sensor modules, which simplifies the power design of a lab platform.
• CMiot CM32M433R: 144 MHz, 144 KB SRAM, dual CAN, and Chinese national cryptographic algorithms, suited to advanced labs and localization topics.

Risks and tradeoffs

• None of these parts has built-in wireless; BLE/Wi-Fi labs need an external module or an ESP32-C3 / CH582M design.
• Prices move with batch and channel; confirm with a distributor before volume purchases for teaching.
• CH32V203 has only 20 KB of SRAM, so memory use must be kept in check when running a GUI or a large protocol stack.

Toolchain and ecosystem

Recommended parts

Why these parts

• CH32L103: still only 49 µA/MHz at 96 MHz, 0.37 µA standby, LPUART/LPTIM, well suited to long-duty-cycle acquisition.
• CH32V006: 48 MHz, 8 KB SRAM, 62 KB Flash in a small QFN package; the peripherals are enough for multi-channel ADC sampling.
• APT APT32F1023: 48 MHz, 4 KB SRAM, 64 KB Flash; the vendor selection guide quotes 5 µA deep sleep, clearly cost-driven.
• CH32V002: the minimal 4 KB SRAM / 16 KB Flash tier, suited to an architecture that takes one sample and wakes a radio module.

Risks and tradeoffs

• APT's toolchain and community material are thinner than WCH's; budget time for porting and debugging before production.
• Vendors test power under different conditions (voltage, temperature, peripheral clocks), so re-measure on your own board.
• RAM retention ranges and the number of wake-up sources in low-power modes constrain the protocol stack you can use.

Toolchain and ecosystem

Recommended parts

Why these parts

• CH582M: BLE 5.0, 0.7 µA sleep, 448 KB Flash, with a complete vendor BLE stack and examples; suited to wearables and HID.
• Telink TLSR9218: 96 MHz, 256 KB SRAM, 2 MB Flash, 0.7 µA sleep; suited to products that need larger buffers and OTA.
• CH592F: BLE 5.4 plus a segment LCD driver, suited to thermostats and panel devices with a display.
• Bouffalo Lab BL702: BLE + USB, 132 KB SRAM, suited to peripherals that need USB and wireless at the same time.

Risks and tradeoffs

• BLE certification (BQB/FCC/CE) is tightly coupled to the stack version; confirm the SDK already supports your target protocol.
• At 20 MHz, some parts leave little headroom to run a display and the wireless stack together.
• OTA partitions and reserved areas in on-chip Flash cut into usable space.

Toolchain and ecosystem

Recommended parts

Why these parts

• ESP32-C3: single-core RISC-V at 160 MHz, 400 KB SRAM, Wi-Fi + BLE 5; the most complete ESP-IDF and Matter ecosystem, with abundant production documentation.
• ESP32-C6: Wi-Fi 6 + BLE 5 + 802.15.4, suited to a new generation of appliances that need Thread/Matter.
• Bouffalo Lab BL616: 320 MHz, 532 KB SRAM, Wi-Fi 6 + BLE; suited to cost-sensitive gateways that still need compute headroom.
• GD32VW553: Nuclei N307 + Wi-Fi 6 + BLE, a smooth migration path for teams already working with GigaDevice MCUs.

Risks and tradeoffs

• RF certification and antenna design are the main hurdles; module designs usually reach production faster than chip-down designs.
• Wi-Fi peak current can reach hundreds of mA; size the supply and decoupling for the peak.
• For the Wi-Fi 6 target use cases of ESP32-C6, confirm router compatibility and the power budget.

Toolchain and ecosystem

Recommended parts

Why these parts

• APT APT32F1730: 105 MHz, FPU/DSP, CAN; aimed at motor and digital-power designs, with a peripheral mix tuned for PWM/ADC.
• CH32V203: 144 MHz, dual USB, CAN, 49 µA/MHz; suited to small drives that also talk to a host.
• CH32V303: 144 MHz, 64 KB SRAM, 256 KB Flash, CAN; enough resources for dual-motor control or complex protection logic.
• GD32VF103: 108 MHz, dual CAN, 12-bit ADC; a mature ecosystem, good as a control-algorithm validation platform.

Risks and tradeoffs

• Motor control is sensitive to ADC sampling timing, PWM dead time, and op-amp bandwidth; validate on a real power stage.
• Some parts do not publish their PWM channel count; confirm with the vendor before committing.
• Isolation, gate drive, and current sensing are analog design problems; the MCU is only one part of the loop.

Toolchain and ecosystem

Recommended parts

Why these parts

• CH32V307: 144 MHz, 64 KB SRAM, 256 KB Flash, dual CAN + 10M Ethernet PHY; suited to device control and protocol conversion.
• CMiot CM32M433R: 144 MHz, 144 KB SRAM, 512 KB Flash, dual CAN, SM3/SM4; suited to localization and security-sensitive designs.
• HPMicro HPM6360: 600 MHz, 512 KB SRAM, CAN + Ethernet; suited to high-real-time multi-axis control and edge algorithms.
• CH32V407: 200 MHz, 200 KB SRAM, 992 KB Flash, Ethernet and USB HS; suited to industrial HMI panels with a display.

Risks and tradeoffs

• The HPM6360 has no on-chip Flash; it needs external NOR/HyperFlash, which affects boot time and BOM.
• Industrial sites demand EMC, isolation, and wide-temperature performance; retest against the full-product standard.
• Whether the Ethernet PHY is integrated and whether external magnetics are needed directly affects layout and cost.

Toolchain and ecosystem