A 105 MS/s 4.18 nJ/inference 0.765 mm² Mixed-Signal Streaming ML Engine for Wireless Baseband in 12LP+ FinFET
By Oscar Castañeda, Lavinia Recchioni, Sueda Taner, Thomas Burger, Jérémy Guichemerre, Christoph Studer
ETH Zurich, Switzerland
Abstract
We present SIXPPAC, the first streaming machine-learning engine for real-time integrated sensing and communication (ISAC) in wireless systems. SIXPPAC integrates two SAR ADCs and a digital binary-weight six-layer neural network (NN). The 0.765 mm² 12 nm FinFET engine stores all NN parameters on chip, enabling sustained 105 MS/s processing of 8b I/Q baseband samples at 0.8 V. By deploying bipolar inner products using AND instead of XNOR gates, SIXPPAC consumes 4.18 nJ/inference (0.42 POPS/W) at 0.8 V. We evaluate SIXPPAC on three example applications to demonstrate its versatility.
To read the full article, click here
Related Semiconductor IP
- Verification IP for MIPI SoundWire
- 100M/1G/ TSN Endpoint DO-254 IP Core
- Image Signal Processor
- H.265 / HEVC Video Codec
- H.264 / AVC Video Encoder
Related Articles
- Multimode: How to design a programmable baseband device for multiple wireless standards
- SoC Test and Verification -> Testing mixed-signal Bluetooth designs
- MPEG Standards -> End user reaps benefit of wireless multimedia structure
- MPEG Standards -> Streaming audio needs interoperability
Latest Articles
- A 105 MS/s 4.18 nJ/inference 0.765 mm² Mixed-Signal Streaming ML Engine for Wireless Baseband in 12LP+ FinFET
- Agentic AI for Heterogeneous System Design Across Chip, Scale-Up, and Scale-Out Domains
- Minimum RISC-V: Maximum FPGA Efficiency
- The Era of Vibe Tapeout: Dark Age or Golden Age for Analog Circuit Design?
- Understanding the Power Consumption of Modern High Bandwidth Memory: Experimental Analysis and Modeling Using Real HBM2 DRAM Chips