Specifications
Processor
Arm Cortex-M4F microcontroller running at up to 80 MHz
On-chip SRAM
Up to 512 KB
Floating point
Integrated floating-point unit on the Cortex-M4F core
Embedded FPGA
2400 effective logic cells of eFPGA fabric for hardware acceleration
eFPGA RAM
64 Kb dedicated embedded FPGA RAM
Power efficiency
As low as 50 microwatts per MHz, and 12.5 microwatts per DMIPS in always-on operation
External memory
Support for up to 16 Mbit SPI NOR flash
Sensor manager
Dedicated sensor manager and flexible fusion engine for multi-sensor pipelines
Voice path
Low-power sound detector with I2S and PDM-to-PCM voice interfaces
Peripherals
I2C, SPI master and slave, UART and DMA with FIFOs
Clocking
On-chip oscillator plus RTC clock domains for low-power scheduling
Analog
Integrated ADC and LDO
Classification
Ultra-low-power multicore sensor processing platform, not a general-purpose FPGA
Toolchain
Programmable with QuickLogic's eFPGA tool flow alongside standard Cortex-M4F software development
Evaluation platform
SparkFun Thing Plus EOS S3 development kit in a Feather-compatible form factor
Target workloads
Concurrent voice, motion and environmental sensor processing in battery-powered devices
Overview
The QuickLogic EOS S3 occupies a specific niche in the edge computing stack: it is a sensor-processing SoC designed to stay awake permanently on microwatt power budgets rather than a high-throughput inference accelerator. Its two halves are an Arm Cortex-M4F microcontroller running at up to 80 MHz with up to 512 KB of SRAM, and a 2400-cell embedded FPGA fabric with 64 Kb of dedicated RAM.
That combination lets designers partition always-on workloads — keyword spotting, motion classification, sensor fusion — between software on the Cortex-M4F and hardware pipelines in the eFPGA, trading logic cells for power. QuickLogic quotes figures as low as 50 microwatts per MHz and 12.5 microwatts per DMIPS, which is what makes continuous listening and sensing viable on coin-cell or small-battery devices.
The device integrates the support circuitry those designs need: a sensor manager and flexible fusion engine, a low-power sound detector, I2S and PDM-to-PCM voice interfaces, I2C/SPI/UART, DMA, ADC and LDO. QuickLogic ships it in a SparkFun Thing Plus development kit in the Feather form factor for evaluation.
Key Benefits
Microwatt-class always-on operation — from 12.5 microwatts per DMIPS — enables continuous voice and sensor monitoring without draining small batteries. The eFPGA fabric (2400 logic cells, 64 Kb RAM) lets teams offload deterministic low-level processing into hardware while keeping higher-level logic in Cortex-M4F software. Integrated sensor management, voice interfaces and analog reduce external component count, and the Feather-format SparkFun dev kit shortens prototyping.
Applications
Always-on voice triggers and keyword spotting in battery-powered devices; wearables and hearables; industrial sensor nodes running motion, environmental and vibration fusion; smart home and building sensors; asset trackers needing years of battery life; and embedded designs that need deterministic hardware pipelines alongside a microcontroller.
Request a Quote — QUICKLOGIC EOS S3 — ULTRA-LOW-POWER EFPGA SENSOR AND VOICE PROCESSING SOC
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →