Ionosphere Research Performance Report

Dual-Channel VLF/ULF Phenomena Detection Analysis

Introduction

This report analyzes GPU FFT performance for ionosphere phenomena detection, focusing on dual-channel configurations optimized for Very Low Frequency (VLF, 3-30 kHz) and Ultra Low Frequency (ULF, <3 Hz) electromagnetic signal processing.

Ionosphere Phenomena Requirements

Ionosphere research requires specialized signal processing with conflicting trade-offs:

Phenomenon	Frequency Range	Time Resolution	Frequency Resolution	Typical NFFT
Lightning/Sprites	3-100 kHz	<10 ms	~100 Hz	1024-2048
Whistlers	1-30 kHz	10-50 ms	10-50 Hz	2048-4096
Schumann Resonances	7.8, 14, 20 Hz	>1 s	<0.5 Hz	65536-131072
ULF Pulsations	0.001-5 Hz	>10 s	<0.1 Hz	131072+

Key Challenge: No single configuration optimally captures all phenomena simultaneously. Real-time systems must adapt NFFT and overlap based on target phenomena.

Dataset Coverage

This analysis covers 13 dual-channel measurements across 13 configurations.

• NFFT Range: 256 - 131072
• Channel Count: 2 (dual-channel ionosphere configuration)
• Benchmark Types: Throughput, Latency, Real-time streaming

Scientific Metrics Overview

Real-Time Factor Analysis

Real-Time Factor (RTF) measures processing speed relative to signal acquisition rate. RTF ≥ 1.0 is required for real-time operation (processing faster than data arrives).

Time/Frequency Resolution Trade-offs

Interpretation: - Lower-left corner (low freq res, low time res): Impossible - violates uncertainty principle - Upper-left corner (low freq res, high time res): Schumann resonances, ULF pulsations - Lower-right corner (high freq res, low time res): Lightning, sprites, fast transients - Center: Whistlers, moderate-speed phenomena

Phenomena Detection Suitability

Dual-Channel Streaming Performance

Real-Time Compliance & Reliability

No real-time benchmark metrics available. This section requires data from run_realtime.py benchmarks.

Configuration Recommendations

Appendix

Data Sources

• Location: artifacts/data/
• Format: Enriched CSV with scientific metrics
• Filter: Dual-channel configurations only (engine_channels=2)
• Date Generated: 2025-11-03 17:57:05

Ionosphere Research Methodology

All benchmarks use dual-channel GPU FFT processing optimized for ionosphere signal characteristics:

• Channel count: 2 (E/W or N/S antenna pairs)
• Sample rate: 48 kHz (captures VLF band up to 24 kHz Nyquist)
• NFFT range: 1024 - 131072 (covers lightning to Schumann phenomena)
• Overlap range: 0.0 - 0.9375 (balances time resolution vs computational load)

Key Scientific Constraints: 1. Lightning/Sprites: Require NFFT ≤ 2048 for <10ms time resolution 2. Whistlers: Optimal at NFFT = 4096-8192 (10-50ms, 10-50Hz resolution) 3. Schumann Resonances: Require NFFT ≥ 65536 for <0.5Hz frequency resolution 4. Real-Time Factor: RTF ≥ 1.0 mandatory for streaming acquisition systems

System Configuration

• Project: ionosense-hpc-lib
• GPU: CUDA-accelerated FFT processing
• Analysis Tools: experiments/analysis/ (shared with Streamlit dashboard)
• Report Generator: Quarto with Python backend