Restoring CFAR Validity for Single-Channel IoT Sensor Streams: A Monte Carlo Comparison of CFAR-Family and Sequential...

Real-time event detection in IoT mesh sensor networks must balance sensitivity against the false-positive load placed on a constrained mesh radio. We present a Monte Carlo comparison of the Temporal Spectral Noise-Floor Adaptation (TSNFA) detector against classical comparators drawn from the radar Constant False Alarm Rate (CFAR) family and from sequential change detection: the Lipski FFT energy detector, Cell-Averaging CFAR (CA-CFAR), Ordered-Statistic CFAR (OS-CFAR), and state-machine Cumulative Sum (CUSUM), with four further family members (GO, SO, TM, VI) measured in one additional configuration. All detectors fit a Cortex-M0+ class envelope, process a 1-D 100 Hz time series in 128-sample frames, and use temporal reference windows in place of spatial reference cells. The simulation covers nine SNR levels (1.5 to 24 dB) at 10 nodes with scale checks at 50 nodes, each test configuration replicated five times over 24 h. The recommended TSNFA configuration, selected by an estimator ablation, combines a mean trigger statistic, a gated median noise floor, and a three-frame confirmation rule. TSNFA detects 100% of events at 12 dB SNR and above, 99.5% at 5 dB, and 80.3% at 3 dB, with 99.6-100% event precision and 0.0017 false-positive clusters per node hour at a cost of 2.6 s of added detection latency from confirmation. CA-CFAR and OS-CFAR detect every event but produce 42.6 and 36.8 false-positive clusters per node hour; the Lipski detector reaches 1.24 clusters per node hour at 41-45% precision but detects no events below 6 dB; CUSUM detects 16-72% of events across the SNR range. TSNFA is the only detector tested that combines high detection rate, high precision, and low false-positive load within the Cortex-M0+ envelope.