Radio communication exhibits the highest energy consumption in wireless sensor nodes. Given their limited energy supply from batteries or scavenging, these nodes must trade data communication for on-the-node computation. Currently, they are designed around off-the-shelf low-power microcontrollers. But by employing a more appropriate processing element, the energy consumption can be significantly reduced. This paper describes the design and implementation of the newly proposed folded-tree architecture for on-the-node data processing in wireless sensor networks, using parallel prefix operations and data locality in hardware. Measurements of the silicon implementation show an improvement of 10-20× in terms of energy as compared to traditional modern micro-controllers found in sensor nodes.

Wireless sensor nodes require low-energy components given their limited energy supply from batteries or scavenging. Currently, they are designed around off-the-shelf low-power microcontrollers for on-the-node processing. However, by employing more appropriate hardware, the energy consumption can be significantly reduced. This paper identifies that many WSN applications employ algorithms which can be solved by using parallel prefix-sums. Therefore, an alternative architecture is proposed to calculated them energy-efficiently. It consists of several parallel processing elements (PEs) structured as a folded tree. Profiling SystemC models of the design with ActivaSC helps to improve data-locality. Measurements of the fabricated chip confirm an improvement of 10-20x in terms of energy as compared with traditional MCUs found in sensor nodes.

Today's highly integrated System-on-Chips (SoC) demand innovative architectural modeling means to cope with their energy constraints. In this context, SystemC has become a well-established simulation tool for Transaction Level Modeling (TLM) in the integrated electronics industry but it lacks the support for power modeling. This paper introduces ActivaSC, a flexible, fast, transparent and non-intrusive extension to the SystemC class library which allows capturing the activity information of a digital system being modeled. This information can then be post-processed to estimate power consumption. ActivaSC avoids time consuming design iterations as designers can assess architectural design trade-offs based on system activity early in the design flow. ActivaSC does not require any code alterations nor a specific API, so that it can be used with any modeling style. Finally, several benchmark tests illustrate the superior efficiency of ActivaSC. A speed-up of up to 75% in terms of elaboration overhead and 20% in terms of simulation overhead is realized with respect to prior art.