16-Mic Acoustic Radar (FPGA PoC)
Real-time sound source localization using a 16-microphone phased array. FPGA-accelerated beamforming for 360° acoustic awareness in industrial and defense environments. A technological Proof of Concept.
16-Mic Acoustic Radar (FPGA PoC)
A technological Proof of Concept demonstrating real-time acoustic beamforming on FPGA. 16 MEMS microphones phased to pinpoint sound sources with <2° angular resolution.
Technical Specifications
| Parameter | Specification |
|---|---|
| Platform | Intel Cyclone 10 LP / Xilinx Artix-7 |
| Microphone Array | 16x MEMS PDM (Uniform Circular Array) |
| Sample Rate | 48 kHz / 24-bit per channel |
| Beamforming | Delay-and-Sum (DAS) / MVDR (Capon) |
| Response Time | < 10 ms (Real-time Visualization) |
| Angular Resolution | ~2° Azimuth |
| Output Interface | HDMI (Heatmap overlay), UDP Stream |
The Challenge
Visual cameras have blind spots and struggle in low light or smoke. Traditional acoustic sensors are omnidirectional—they know something is loud, but not where it is. The goal was to build a “sound camera” capable of visualizing noise sources in real-time to detect drone swarms, machinery faults, or gas leaks, without the latency of CPU-based processing.
Our Solution
We engineered a Massive Parallel Acoustic Processor directly on the FPGA fabric, bypassing the CPU for the heavy lifting:
- PDM Demodulation: 16 parallel CIC filters decimate Pulse Density Modulation (PDM) streams from MEMS mics into PCM audio.
- Delay-and-Sum Beamforming: The FPGA computes 1024 observation beams simultaneously. By delaying signals based on the array geometry, we constructively interfere sound coming from a specific angle and cancel out the rest.
- Heatmap Generation: The acoustic energy map is overlaid on a visual camera feed via HDMI, allowing operators to “see” sound.
Technology Stack
FPGA Core (VHDL/Verilog)
- Parallel DSP Blocks: 16x simultaneous MAC operations for beamforming.
- CORDIC Algorithm: For real-time phase shift alignment.
- DMA Controller: Zero-copy transfer of heatmap data to the visualization buffer.
Application Areas
- Counter-UAS: detecting drone propeller noise before visual contact.
- Industrial IoT: Pinpointing gas leaks (hissing) or bearing failures in complex machinery.
- Smart City: Gunshot detection and localization triangulation.