ZF and SiliconAuto have developed an interface chip with companion microcontroller for automated driving, writes Nick Flaherty.

The specialised I/O chip performs data acquisition and preprocessing in real time for automated driving and is tightly coupled with the XMotiv M3 microcontroller from SiliconAuto.

The system is fully agnostic to the choice of high-performance system on chip for the compute, enabling system developers to integrate any preferred AI inference chip via standardised high-speed interfaces.

The I/O interface chip integrates all required automotive sensor-interface intellectual property along with sensor preprocessing capabilities such as low-latency camera image signal processing and on-chip radar signal processing.

This is coupled to the M3 microcontroller, which is based on a 160 MHz ARM Cortex-M33 core, 128 KB of SRAM and 1 MB of flash memory alongside a range of interfaces including CAN and LIN.

The combination reduces power consumption by limiting data transfers to Double Data Rate memory and by reducing clock speeds, and the I/O interface chip provides the flexibility to connect to any generation of the latest low-power AI inference engines.

The chip is manufactured in a 28 nm process technology and offloads sensor data acquisition tasks and sensor data preprocessing to free up expensive CPU power on the performance SoC. As such, the expensive application cores of the performance of SoC can fully concentrate on perception and driving functions.