Real-time data processing

Modern embedded systems continuously generate data that must be interpreted before the next action can take place. Real-time data processing is therefore not just about speed, but about delivering the required result within a predictable time window. 

TOPIC develops embedded architectures in which latency, bandwidth and processing requirements are considered from the start.

Distributing processing across the system

When input rates exceed what a processor can handle sequentially, processing tasks need to be distributed. FPGA logic can perform parallel and time-critical operations, while processors handle control, application logic and more dynamic workloads.

This division makes real-time data processing possible in demanding applications such as medical signal processing, industrial ultrasound, vision, lidar and radar.

Real-time data integration from sensor to application

Real-time data integration connects the different processing stages. High-speed interfaces such as PCIe and Ethernet move data between sensors, FPGA logic, processors, memory, storage and other systems.

Bandwidth alone is not enough. Buffering, data reduction and workload distribution determine whether the complete data pipeline can keep pace and deliver useful information within the required time.

Optimizing real-time performance

For demanding applications, TOPIC balances processing performance against bandwidth, power consumption and thermal constraints. In one industrial ultrasound application, over 100 modules perform signal processing in parallel.

This illustrates the principle behind scalable real-time data processing and integration: process data close to the source, reduce unnecessary data movement and assign each task to the hardware best suited to execute it.