A New Frontier for Embedded Development

Everything that runs on electricity will live on the network.

Think about the implications.

If it can be plugged in or runs on a battery, if it lives in a car, an airplane, a boat, in a pants pocket or factory, it connects to the network. At some point in time, everything we use on a day-to-day basis will communicate on an IP network.

Application convergence in a networked world is truly a new frontier for embedded development engineers. It is an engineering world filled with very different — and often mismatched — network-connected devices, equipment and systems that all need to be adaptable, multi-format, multi-standard and programmable. It is also a world that can only work with a complete blending of control processing and signal processing. This blending is the key enabling technology behind embedded convergence.

What is embedded convergence? It is not a classic real-time control standing alone, nor is it a classic DSP, and it most definitely is not heterogeneously kludging together MCU and DSP cores. Convergent processing combines the attributes of signal processing and control processing into a single, unified architecture designed from the ground up.

Today, time-to-market pressures and tightly constrained development costs simply won't allow complexities to derail the delivery of embedded designs that integrate signal processing capabilities. In the convergent network-connected world, signal processing is a necessity.

Let's take a look at how convergence class applications work. Processors for the convergent domain have to operate fully in both the application and signal processing domains. Neither kind of processing is an adjunct to the other. So when you go processor shopping for your next convergence class embedded design, be forewarned that any embedded processor not intrinsically architected from the ground up for homogeneous, fully-integrated, high-performance signal processing and high-performance application processing is a losing proposition.

To meet the needs of today's converged applications, an embedded processor should match the demands of a classic ASIC implementation — particularly in regards to processing performance and power efficiency — and the low latency, optimized data flow, memory utilization and network upgradeability of a general purpose processor implementation.

Convergence-class processors should integrate signal processing facilitators, such as DMA controllers and peripherals; include MCU-like peripheral sets such as I/O flags, to facilitate embedded control; and strong RTOS platforms. Finally, look for processor architectures that are blended so you can avoid the traditional schizophrenic development split between embedded microcontroller engineers and DSP engineers.

So what comes next? What happens as technology drives innovation which drives demands for more technology? There is absolutely no indication that consumer, commercial or industrial products will slow their demand for increased embedded processing, data-throughput and intelligent network connectivity. The next generation of convergent applications will go beyond MCU-level control needs and will demand microprocessor-class performance and enterprise-class application processing tasks. High-speed data streams will fly over broadband communications channels, and support for high-end processing applications and high-speed network data streams will include managing all human interfaces, inter-device communications and system-wide control capabilities.

Welcome to the new frontier — your opportunity to innovate.