Bluebud platform aims to standardise DSP-enabled Bluetooth Audio IP
Intended to standardise DSP-enabled Bluetooth audio IP to meet the demands of the Bluetooth audio market, CEVA has introduced Bluebud. The integrated wireless audio platform targets True Wireless Stereo (TWS) earbuds, hearables, wireless speakers, gaming headsets, smart watches and other wearable devices.
Bluebud serves to provide a standardised and self-contained IP that companies targeting the Bluetooth audio market can effectively drop in to an SoC design, dramatically reducing the high entry barriers and time-to-market, said CEVA. It is self-contained with everything from Bluetooth link control, host protocol stack, profiles and audio codecs up to the application able to run on the embedded CEVA-BX1 while Bluebud provides the SoC integrator the option to add a host CPU to run user applications.
Bluebud enables end-product differentiation via support for host-level application software which can run on the CEVA-BX1 DSP or an optional host CPU.
The platform combines CEVA’s RivieraWaves Bluetooth 5.2 IP, with both Classic Audio and LE Audio, and the CEVA-BX1 audio processor, with all the required peripherals for wireless audio. The compact low power design has a footprint of less than 0.5mm2 in 22nm. The single core architecture reduces the overall latency from RF to audio output down to less than 20ms, which will benefit audio/video synchronisation for movies and gaming. The design also incorporates a power management unit supporting idle and sleep power-down modes through clock and voltage scaling.
Users can take advantage of the onboard CEVA-BX1 DSP and the SenslinQ DSP software framework to seamlessly integrate differentiated, value-add software from CEVA, its ecosystem partners or custom software, onto the Bluebud platform. CEVA’s optional software for Bluebud includes ClearVox (multi-microphone noise reduction and acoustic echo cancellation), WhisPro (voice trigger and commands), and MotionEngine Hear, for user control via tap and double tap, head tracking, in-ear detection and activity classification.
For light AI inferencing workloads, Bluebud allows users to run TinyML models, such as a wake word engine, to summon a voice assistant or detect sound events for contextual awareness, using native support for the TensorFlow Lite Micro neural network framework.
Bluetooth 5.2 and its integral LE Audio enhance audio quality and link robustness as well as enable audio sharing. The audio quality improvements are courtesy of the LC3 codec, optimised for the CEVA-BX1, which also enables lower power operation to extend the battery life of TWS earbuds and other audio streaming devices, explained CEVA. The CEVA-BX1 processor combines DSP compute capability with programming and compact code size. Using an 11-stage pipeline and four-way VLIW micro-architecture, it offers parallel processing with a single instruction multiple data (SIMD) ISA, which is widely used in neural networks inference, noise reduction and echo cancellation, as well as high accuracy sensor fusion algorithms.
The software development tool chain includes an LLVM compiler, Eclipse-based debugger, DSP and neural network compute libraries, neural network frameworks support in the form of TensorFlow Lite Micro, and choice of industry leading real time operating systems (RTOS).
The Bluebud platform is available for licensing by end of Q1 2021.