Tensilica HiFi 1 DSP extends battery life in wearables and always-on devices
Described as ‘breakthrough’ audio / video performance, the Cadence Tensilica HiFi 1 DSP has low energy consumption which enables small form factor, low-cost consumer and mobile devices, as well as automotive and industrial devices, to offer increased functionality with low energy consumption, says Cadence.
The DSP is claimed to extend the duration of voice communication and music playback, which allows always-listening functionality for voice commands, without compromising the battery life.
It is designed for use in small battery wearable devices such as TWS (true wireless stereo) ear buds, hearing aids, Bluetooth headsets and smart watches.
Increasingly, hearables and wearables are adopting the low complexity communications codec (LC3) standardised by the Bluetooth Special Interest Group (SIG) in 2020. Interest in hands-free and touch-free control is also driving demand for always-on, or always-listening, devices that respond to voice wake-up commands. Appliances are adopting these capabilities. All these devices and applications require a small form factor and long battery life.
The LC3 and LC3Plus codecs were co-invented by Cadence and Fraunhofer help minimise energy consumption for battery-constrained Bluetooth devices.
The HiFi 1 DSP delivers low energy encoding and playback of LC3 and other Bluetooth codecs and low energy key word spotting for voice wake-up, in the smallest footprint HiFi DSP. Compared to the HiFi 3 DSP, claimed to be the most popular audio DSP in the industry for the target applications, the HiFi 1 DSP offers 11 to 16 per cent lower area, 60 to 73 per cent greater cycle and energy efficiency for machine learned (ML) -based “OK Google” key word spotting and person detect applications. They also offer greater than 18 per cent cycle efficiency and 14 per cent energy efficiency for LC3 decoding, adds Cadence.
“The HiFi 1 DSP from Cadence significantly lowers the energy required to run the always-on class of AI applications, such as the TensorFlow Lite Micro (TFLM) networks, speech wake word and person detect, enabling battery-constrained devices to run for longer lengths of time,” said Pete Warden, technical lead of TFLM at Google.
Special instructions including arithmetic coding for LC3 encode and decode make the HiFi 1 DSP the most energy-efficient DSP for Bluetooth LE Audio, says Cadence. Instructions for ITU-T3GPP-standardized 2019 BASOPs accelerate speech codecs and run them with high energy efficiency to increase talk time.
Another benefit is the neural network ISA and load/store capability which accelerate key word spotting and other ML workloads to reduce energy consumption. Memory access-optimised ISA reduces improves performance with small cache sizes, confirms Cadence.
The HiFi 1 DSP is available now.