Ultra low power signal processing
Truly autonomous sensor nodes will enable revolutionary applications from ‘go-anywhere’ healthcare monitoring to ‘install-and-forget’ home and industrial automation. To make this a reality, Holst Centre and its partners are developing ultra low power (ULP) signal processing techniques, architectures and devices that will improve the intelligence and overall power consumption of nodes.
Adding a ULP DSP to a sensor node means the node can do more advanced and more energy-efficient data processing. So it can independently react to input changes without needing to contact the basestation. This speeds up response times and reduces the power used for wireless transmissions – and hence the nodes overall power consumption.
The program targets DSPs that will reduce power consumption for algorithms, such as used for biomedical and wireless baseband processing, to below 20 µW, allowing energy scavenging applications. To achieve these levels, the DSP has to be optimized for specific applications. Initially, we are focusing on secure biomedical sensor networks, developing Application-Specific Instruction set Processors (ASIPs) that can be implemented in standard CMOS processes.
The program looks at all aspects of DSP creation from algorithm to IC design, up to demonstrator level and FPGA prototypes. Success relies on involving experts from a wide range of areas throughout the value chain – from concept to silicon. DSP algorithms, hardware architectures, embedded memories, circuit design and manufacturing processes all need to be optimized for the target application.
To do this, the team uses processor architectures, algorithm mapping techniques and deep sub-micron design to integrate partners’ system and algorithm requirements with the lowest energy consumption possible. A deep understanding of the final system and application is essential, allowing good technology transfer.