Texas Instruments has introduced the MSPM0G5187 and AM13Ex, two new microcontroller (MCU) families featuring the TinyEngine neural processing unit (NPU) for efficient Edge AI/Machine Learning inference. The TinyEngine NPU provides up to 90 times lower latency and uses over 120 times less energy per inference compared to similar MCUs without an accelerator. The MSPM0G5187 is a low-power Arm Cortex-M0+ MCU, while the AM13Ex Arm Cortex-M33 microcontroller is designed for real-time motor control, starting with the AM13E23019 SKU.
**TI MSPM0G5187 Cortex-M0+ MCU**
– **CPU:** Arm Cortex-M0+ at 80 MHz
– **Memory:** 32 KB RAM with ECC
– **Storage:** 128 KB flash with ECC, 8 KB data flash with ECC
– **Accelerators:**
– TinyEngine NPU delivering up to 2.56GOPS at 80MHz
– MATHACL math accelerator
– **Peripherals:**
– **USB:** 1x USB 2.0 (12 Mbps)
– **Audio Interface:** I2S, PCM, TDM, etc.
– Up to 59x GPIOs, including 2x 5V tolerant
– 3x UART, 2x SPI, 1x I2C
– Timers: 4x 16-bit timers with up to 14 PWM, 2x watchdog timers, RTC with alarm and calendar mode
– **Analog:** 12-bit SAR ADC; 26 channels; up to 1.6 Mbps sample rate, comparator with 8-pin DAC, temperature sensor
– **Security:** AES encryption, cryptographic acceleration, secure boot, secure communication, etc.
– **Debugging:** 2-pin SWD
– **Supply Voltage:** 1.62V to 3.6V
– **Power Modes:**
– RUN – 103µA/MHz (CoreMark)
– SLEEP – 34µA/MHz
– STOP – 199µA at 4MHz
– STANDBY – 1.5µA at 32kHz with RTC and full SRAM retention
– SHUTDOWN – 88nA with IO wake-up
– **Packages:** Various pin and package options
– **Temperature Range:** -40 to 125°C
Supported by the MSPM0 SDK with FreeRTOS and Zephyr RTOS, these microcontrollers are aimed at battery-powered devices handling AI/ML edge workloads. The LP-MSPM0G5187 Launchpad development kit is available for $22. The MSPM0G5187 family is available in production quantities starting at under $1 per unit in lots of 1,000.
**TI AM13Ex Arm Cortex-M33 MCU**
– **CPU:** Arm Cortex-M33 core at up to 200 MHz with FPU, MPU, DSP, TMU
– **Memory:** 128 KB RAM with ECC, EPI for external interfaces
– **Storage:** Up to 512 KB flash with ECC
– **Accelerator:** TinyEngine NPU for AI/ML
– **Peripherals:**
– DMA controller, NVIC, up to 107 GPIO with XBAR, watchdog timer, timers
– **Analog:** 3x 12-bit SAR ADCs, comparators, programmable gain amplifiers, etc.
– **Real-time Control:** MCPWM modules, eCAP, eQEP modules, Device Crossbars
– **Enhanced Interfaces:** Configurable serial interfaces supporting UART, I2C, SPI, CAN FD
– **Clock System:** Internal and external oscillators and PLLs
– **Safety and Security:** Enables IEC61508 SIL-2 and SIL-3 systems, various secure features
– **Debugging:** JTAG and SWD, Trace Macrocell
– **Supply Voltage:** 3.3V
– **Power Modes:** Efficient power modes for different operational states
– **Packages:** Various pin packages available
– **Temperature Range:** -40°C to +105°C
The AM13Ex family is suitable for motor control applications leveraging machine learning for adaptive control and predictive maintenance. The AM13E23019 is available in preproduction quantities starting at $2.45 per unit in 1,000-unit orders.
**TinyEngine Software Support**
Both microcontrollers are supported by TI’s CCStudio IDE, which uses generative AI for accelerated development. The IDE contains over 60 models and examples in CCStudio Edge AI Studio. It supports capture, training, and optimization of custom models and integrates frameworks like PyTorch, TensorFlow, and ONNX. More details are available in the developer documentation.