SCuM's new Python RTL generator enables AI-assisted chip design

In a new arXiv paper (2608.17042), Brandon P. Hippe and David C. Burnett present a Python-based RTL generator designed for the Single-Chip Micro Mote (SCuM), a crystal-free radio architecture targeting ultra-low-power IoT devices. The generator is closely integrated with simulation and testing environments, making it easier to optimize digital baseband hardware against the chip's specific architecture. This flexibility allows the same RTL to be deployed across FPGA implementations and ASIC tape-outs, reducing design iteration time.

The work addresses a critical bottleneck: wireless communication hardware consumes a large share of a device's power budget, especially noise-reduction circuits in the analog front end. SCuM deliberately trades noise performance for lower power while staying compatible with IEEE 802.15.4 and Bluetooth Low Energy. By enabling tighter co-optimization of the digital baseband with the analog front end, the new generator promises significant power savings. The authors also highlight its potential in AI-assisted design workflows, where machine learning could automate RTL generation and hardware optimization for future IoT chips.