Generated: 2026-05-14 09:28:15
Source: compiled_workbook.xlsx
Configurations: 10n / 12dB · 10n / 18dB · 50n / 12dB · 50n / 18dB

1. Run Inventory

Four Monte Carlo configurations were run, factorial across two network sizes (10 and 50 nodes) and two SNR levels (12 dB and 18 dB). Each configuration was simulated for 24 hours of synthetic time and replicated five times with independent random seeds for variance estimation.

Configuration Origin N nodes SNR (dB) Duration MC Runs Total Events Preset
10n / 18dB20260513_102615101824.0h5ACCURATE
10n / 12dB20260513_091959101224.0h5ACCURATE
50n / 18dBcurrent501824.0h5ACCURATE
50n / 12dB20260513_161751501224.0h5ACCURATE

2. Detector Pool

Five detection algorithms were evaluated. TSNFA is the proposed cascade. Lipski FFT, CA-CFAR, OS-CFAR, and CUSUM are classical baselines drawn from radar and statistical signal processing literature, used here at their canonical published settings to allow a fair comparison.

TSNFA (proposed)

Two-stage cascade: spectral consistency (FFT-binned, Defence-1) followed by median-CFAR rank statistic (Defence-2). Designed for low-SNR seismic monitoring with strict bandwidth constraints.

Lipski FFT

Per-bin energy detector firing when bin power exceeds μ + kσ for a minimum number of consecutive bins. Canonical k = 3 (Lipski et al.).

CA-CFAR

Cell-averaging Constant False Alarm Rate detector using a sliding mean of N reference cells. Threshold multiplier α = 7.71 derived from P_fa = 10⁻³ via the Finn-Johnson formula.

OS-CFAR

Order-statistic CFAR using the 75th-percentile (k = 24 of N = 32) of the reference cells. Threshold multiplier α = 37.33 from P_fa = 10⁻³ (Rohling 1983).

CUSUM

Tartakovsky's bounded-memory cumulative-sum change detector with α_fa = 10⁻⁵ and a finite window K_end = 100 to prevent unbounded accumulation in stationary noise.

3. Algorithm Parameters

The canonical settings used across all four runs. These are fixed by-design parameters drawn from the published literature, not tuned to the data. Identical settings ensure detectors are compared at their published operating points rather than at empirically-optimized configurations.

DetectorParameterCanonical Value
TSNFAγ_d (Defence-1 threshold)
γ_a (Defence-2 threshold)
ζ (CFAR multiplier)
Lipski FFTk (sigma multiplier)
N_bins_min (min consecutive bins)3
CA-CFARN_ref (reference cells)32
P_fa (false-alarm prob)
OS-CFARN_ref (reference cells)32
k_rank (rank statistic)
P_fa (false-alarm prob)
CUSUMα_fa (false-alarm rate)

4. Headline Results

The central result table. Each detector evaluated across all four configurations on the four most diagnostic metrics: detection rate, event precision, false-positive cluster count, and per-node network load.

Detector10n / 18dB10n / 12dB50n / 18dB50n / 12dB
Detection Rate (%)
TSNFA100.00%100.00%100.00%99.97%
Lipski99.73%99.73%99.85%99.85%
CA-CFAR100.00%100.00%100.00%100.00%
OS-CFAR100.00%100.00%100.00%100.00%
CUSUM68.14%51.30%70.81%51.08%
Event Precision (%)
TSNFA100.00%100.00%100.00%100.00%
Lipski1.65%1.66%1.72%1.69%
CA-CFAR2.25%2.25%2.29%2.29%
OS-CFAR2.59%2.59%2.65%2.65%
CUSUM1.70%1.28%1.78%1.26%
FP Clusters
TSNFA0000
Lipski12574125196750168603
CA-CFAR919491885030350305
OS-CFAR795279534328643285
CUSUM843784414602947348
Network Load (B/hr)
TSNFA392 B/hr232 B/hr2.0 kB/hr1.2 kB/hr
Lipski241.3 kB/hr240.9 kB/hr1.19 MB/hr1.18 MB/hr
CA-CFAR145.8 kB/hr145.9 kB/hr715.9 kB/hr716.1 kB/hr
OS-CFAR150.0 kB/hr150.1 kB/hr737.0 kB/hr737.2 kB/hr
CUSUM12.2 kB/hr12.2 kB/hr55.1 kB/hr57.0 kB/hr
Key finding: TSNFA achieves 99.9–100% detection rate with 100% precision (zero false-positive clusters) across all four configurations. No classical comparator achieves both. Lipski and CA-CFAR achieve high DR (~99.7–100%) but with precision below 3%. OS-CFAR shows severe SNR-dependent collapse. CUSUM shows moderate brittleness.
Figure 1. Detection rate matrix — heat-map overview of detection rate per detector × per configuration. TSNFA's row is uniformly deep green; OS-CFAR's row shows the SNR-dependent collapse. Open standalone →

5. ROC Analysis

For each detector, the simulator swept a threshold multiplier in post-processing (using the saved per-frame strengths) to trace out the ROC upper envelope. Hollow circles mark the canonical operating point — the threshold value that produces the headline metrics in Section 4.

Figure 2. ROC envelopes per configuration. Toggle detectors via the legend; drag to zoom into the upper-left corner. Open standalone →

6. Network Bandwidth

Network load per detector, expressed as bytes per hour each sensor node transmits. Total mesh load is per-node × (N − 1), where N − 1 is the number of sensor nodes (one node is the sink). The "vs TSNFA" column shows the bandwidth ratio, revealing how much more traffic each comparator generates at the same operating point.

Detector10n / 18dB10n / 12dB50n / 18dB50n / 12dB
Per-nodeTotal meshvs TSNFAPer-nodeTotal meshvs TSNFAPer-nodeTotal meshvs TSNFAPer-nodeTotal meshvs TSNFA
TSNFA392 B/hr3.5 kB/hr1.0×232 B/hr2.1 kB/hr1.0×2.0 kB/hr96.4 kB/hr1.0×1.2 kB/hr57.0 kB/hr1.0×
Lipski241.3 kB/hr2,171.5 kB/hr614.8×240.9 kB/hr2,168.5 kB/hr1,039.4×1.19 MB/hr58,157.7 kB/hr603.3×1.18 MB/hr57,988.5 kB/hr1,017.8×
CA-CFAR145.8 kB/hr1,312.4 kB/hr371.6×145.9 kB/hr1,312.8 kB/hr629.3×715.9 kB/hr35,081.3 kB/hr363.9×716.1 kB/hr35,089.5 kB/hr615.9×
OS-CFAR150.0 kB/hr1,350.3 kB/hr382.3×150.1 kB/hr1,350.7 kB/hr647.4×737.0 kB/hr36,113.7 kB/hr374.6×737.2 kB/hr36,124.1 kB/hr634.0×
CUSUM12.2 kB/hr109.9 kB/hr31.1×12.2 kB/hr109.9 kB/hr52.7×55.1 kB/hr2,699.7 kB/hr28.0×57.0 kB/hr2,790.9 kB/hr49.0×
Key finding: At 50n × 12 dB, Lipski generates 1025× more per-node bandwidth than TSNFA, CA-CFAR generates 617×, and CUSUM generates 49×. OS-CFAR's lower bandwidth (146 B/hr vs TSNFA's 1160 B/hr) is a consequence of its 96.8% miss rate — a silent detector consumes little bandwidth.
Figure 3. Per-node and total mesh bandwidth across configurations. Log y-axis spans three orders of magnitude. Open standalone →

7. SNR Robustness

How much each detector's detection rate degrades when SNR drops from 18 dB to 12 dB. Robust detectors show ≈ 0 percentage-point drop; brittle detectors collapse. Cells highlighted: red for severe brittleness (>30 pp drop), amber for moderate (10–30 pp), green for stability (< 1 pp).

DetectorN = 10N = 50
18 dB12 dBΔ%18 dB12 dBΔ%
TSNFA100.00%100.00%≈0100.00%99.97%+0.03
Lipski99.73%99.73%≈099.85%99.85%≈0
CA-CFAR100.00%100.00%≈0100.00%100.00%≈0
OS-CFAR100.00%100.00%≈0100.00%100.00%≈0
CUSUM68.14%51.30%+16.8470.81%51.08%+19.73
Figure 4. Detection rate vs SNR, paired panels per network size. TSNFA holds a flat 100% line; OS-CFAR shows the steep collapse. Open standalone →

8. Network Scaling

Behaviour of each metric as network size grows from 10 to 50 nodes. Detection-quality metrics (DR, precision) are per-node algorithmic properties and should be invariant in N. Cost metrics (FP cluster count, network load) scale linearly because each node operates independently.

Figure 5. Six-metric scaling grid. Solid lines: high SNR. Dashed lines: low SNR. Open standalone →

9. Bandwidth–Precision Trade-off

Each detector × configuration plotted as a single point in 2D (network load, event precision) space. The upper-left corner is ideal: high precision and low bandwidth. Marker size encodes network size.

Figure 6. Bandwidth-precision trade-off. TSNFA occupies the upper-left "precise + cheap" corner alone. Open standalone →

10. 3D Quality Space

Each detector × configuration as a bubble in three dimensions (FAR × network load × detection rate). Bubble size encodes event precision (big = useful triggers, small = noise). A "good" detector is a big bubble in the front-left-top corner: low FAR, low load, high DR, high precision.

Figure 7. 3D quality space. Drag to rotate. The translucent green ceiling marks DR = 100%. Open standalone →

11. Methodology