Summary

The signal-strength-based position estimation algorithm may have inverted SNR weighting, causing nodes with weak signals to be estimated as close rather than far.

Background

In LoRa, SNR scale runs roughly from -20 dB (near noise floor, very weak) to +10 dB (strong). A value of -12 dB is a poor signal. If the estimation code treats a more-negative SNR as higher weight (i.e. closer), the math is backwards.

Suspected Bug

The weight formula 10^(SNR/10) correctly assigns low weight to weak signals (e.g. -12 dB → ~0.06). However, if the code inverts this — using 1 / weight or negating the exponent — a node seen only at -12 dB SNR would be pulled toward the anchor rather than pushed away from it, placing the estimate at the anchor's location regardless of actual distance.

Observed Symptom

A node with -12 dB outgoing SNR (reported in traceroute) is being estimated as inside the reporter's house — the only anchor that observed it. With correct weighting, the uncertainty radius should be large (~5 km for a single low-confidence anchor), signaling the estimate is unreliable. Instead the estimate appears spatially confident and wrong.

What to Check

• src/ or dist/ — position estimation / trilateration logic
• Verify the SNR-to-weight mapping: higher (less negative) SNR should yield higher weight, meaning the node is pulled toward that anchor
• Verify that a single low-SNR anchor produces a large uncertaintyKm, not a small one

Expected Behavior

• Low SNR → low weight → large uncertainty, estimate treated as unreliable
• High SNR → high weight → node pulled toward that anchor confidently

Actual Behavior

Estimate collapses to anchor position with apparent confidence despite a single -12 dB observation.

Authored by NodeZero 0️⃣