> Acoustic Detection Architecture_

Physics Doesn't Lie.
Every Drone Makes Noise.

Modern autonomous drones operate in total radio silence. Billion-dollar radar arrays see nothing. We don't look for radio signals. We listen for the physics that cannot be switched off.

No RF dependency
Zero cloud latency
Edge inference only
Sub-$1k / node
// Why Acoustics
01 // The Physical Law

Rotors Cannot Hide

Every propeller generates a Blade Passing Frequency defined by rotor mechanics. This is Newtonian—no firmware update eliminates it.

Typical quad BPF range
80 – 800 Hz
02 // Harmonic Fingerprint

Every Platform is Unique

Motor, propeller geometry, and airframe resonance combine into a harmonic signature. Our models are trained to exploit this distinctiveness at the class level.

Analysis
Fundamental + harmonic orders
03 // Analog-Digital Hybrid

Fidelity at the Source

High-SNR MEMS microphones feed a discrete analog front-end before the ADC fires. Signal conditioning in the analog domain preserves the frequency content ML depends on.

Data points per analysis window
16,000+
// SIREN Node // Hardware Architecture
Core Compute // PSoC 6
Cortex-M4 ML CORE
TinyML inference & sensor fusion
Cortex-M0+ COMMS CORE
LoRaWAN stack, power management, watchdog
Hardware Security Module CRYPTO
AES-128 acceleration, secure key storage, trusted boot
Asymmetric dual-core design enforces hardware-level isolation between inference and networking. A compromised comms layer cannot alter classification integrity.
Sensor Suite // Acoustic Array
MEMS Microphone Array PRIMARY
High-SNR // wide frequency response // omnidirectional
6-Axis IMU INTEGRITY
Tamper detection, drop-deployment sensing, orientation
Analog Front-End FIDELITY
Discrete pre-amp, anti-aliasing filter, programmable gain
The hybrid analog-digital front-end captures what all-digital architectures discard. Fidelity at the source is the prerequisite for accurate edge classification.
Networking // LoRaWAN Mesh
Range (LoS)
15+ km
Band (US)
915 MHz
TX Power
< 50 mW peak
Topology
Self-Healing Mesh
Sub-GHz spectrum operates below the noise floor of jamming hardware optimized for 2.4GHz/5.8GHz drone control links. The mesh persists when consumer RF bands are denied.
Power // Autonomous Operation
Standby Draw
< 5 mW
Active Inference
~85 mW
Solar Input
Optional MPPT
Battery Runtime
72h+ nominal
Aggressive duty-cycle scheduling keeps the ML core dormant until acoustic thresholds are crossed—reducing energy draw by an order of magnitude vs. continuous-inference designs.
// Competitive Analysis
Traditional Radar / RF
$50,000 – $500,000+
Shields.Systems SIREN
Sub-$1,000 / node
Radio-Silent Drone Detection ✗ Blind ✓ Primary Use Case
GPS-Denied Environments ◐ Degraded ✓ Native
Deployment Fixed / Contractor Throw & Go
Power Requirement Grid Power Battery + Solar
Cloud Dependency ◐ Vendor-Dependent ✓ Zero Cloud
Emissions Signature High / Active Passive / Minimal
False Positive Mitigation ◐ Geometry Only ✓ Acoustic Fingerprint
Per-Unit Cost $50k – $500k+ < $1,000
// Target Deployment Profiles
PROFILE // 01

Private Estates & Infrastructure

Persistent perimeter mesh for high-value sites. Solar-powered, zero-maintenance, no recurring infrastructure costs.

Recommended: 8–20 nodes
Coverage: up to 2km² continuous
PROFILE // 02

Temporary Event Venues

Rapid deployment for stadiums, festivals, and dignitary protection. Single operator. Self-configuring mesh. No tools required.

Recommended: 4–10 nodes
Setup: < 15 min, carry-on packable
PROFILE // 03

Contested & Denied Zones

GPS-denied, RF-jammed environments where legacy systems fail by design. Passive detection. Integrates with HARPY for kinetic response.

Operates under active jamming
Zero RF emission in passive mode
// Next Step

Engineering Collaboration Welcome

Engaging defense primes, system integrators, and private security operators. Technical data packages available under NDA.

Contact Engineering Team
// Status
Hardware Rev PSoC6 // v1.0
ML Model CNN // TFLite Micro
Network Protocol LoRaWAN 1.0.4
Stage Early Prototype
Location Denver, Colorado