QuickLogic Partners with Antmicro on Voice-Enabled Edge AI QuickFeather Development Board
With embedded FPGA, a "Flexible Fusion Engine," and an Arm Cortex-M4F, the open hardware board is due to launch in the second quarter.
QuickLogic has announced a partnership with Antmicro to bring a new open hardware Feather-format development board to market, designed for voice-enabled edge AI projects: the QuickFeather.
"Machine learning applications are being deployed at an amazing rate and the new QuickFeather board will further accelerate that trend," predicts Brian Faith, president and chief executive officer of QuickLogic. "Developers love the fact that it and its associated Renode simulation framework are open source, making it even more attractive for implementing ML algorithms on endpoint IoT applications."
"An open hardware development board for a cost effective, FPGA-enabled SoC platform coupled with useful sensors, supported in a mainstream open source RTOS and the open source Renode simulation framework, QuickFeather is ideally positioned for use in tiny ML applications such as SensiML's AI Software Platform and Google's TensorFlow Lite," claims Michael Gielda, vice president of business development at Antmicro, of QuickLogic's launch. "We are proud to be helping QuickLogic build an open hardware and software ecosystem that can serve as a model for the entire industry."
The QuickFeather is based around QuickLogic's EOS S3, a system-on-chip with embedded field-programmable gate array (eFPGA) and QuickLogic's "μDSP-like" Flexible Fusion Engine (FFE) and an Arm Cortex-M4F core running at up to 80MHz. The SoC includes 512kB of static RAM, while the board design also includes a GigaDevice 16Mb flash memory module.
For on-board sensors, the QuickFeather boasts an mCube MC3635 accelerometer, Infineon DPS310 pressure sensor and IM69D130 PDM digital microphone — the latter for the board's targeted always-on voice-activated edge AI tasks — alongside a user-definable button and RGB LED. Power is provided over a USB connection, or from a lithium-polymer battery which integrates into the built-in charging circuit.
The board is to be supported in Antmicro's Renode simulation framework, the Zephyr real-time operating system (RTOS), and works with popular micro machine learning frameworks including TensorFlow Lite while boasting a fully open source workflow including Symbiflow support on the FPGA side. At the moment, however, it's a paper launch: QuickLogic has promised to get the hardware out by the end of the second quarter, along with its addition to Renode.
In the meantime, more information — and a link to be registered for notification when the boards go on sale — can be found on the QuickLogic website.