Per FlightAware's engineering blog, receivers within 5 miles of airports provide enhanced surface-level tracking. Our location provides:

• Direct line-of-sight to all KBOS runways
• Surface movement detection (taxiing aircraft)
• Low-altitude approach/departure coverage
• Harbor traffic visibility

1. RTL-SDR receives 1090 MHz signals
2. dump1090 decodes Mode S / ADS-B messages
3. adsb-processor aggregates from multiple receivers
4. SQLite database stores aircraft + metadata
5. adsb-server provides HTTP API
6. Web UI displays real-time map

• REST API for aircraft metadata enrichment
• binCraft binary format for area queries
• Trino access for historical data (pending application)

# Test OpenSky metadata
curl -s "http://localhost:8081/aircraft-metadata/A0AF06" | jq

• PiAware feeding from pi.lan (Site 264028)
• AeroAPI available for commercial queries

• Crowd-sourced unfiltered data
• Globe visualization: https://globe.adsbexchange.com/

Status: Stay with SQLite (ADR-001)

SQLite advantages for this use case:

• Zero administration
• Single file deployment
• Python built-in support
• WAL mode for concurrent reads

Track when flight numbers switch aircraft (e.g., AAL2763 from A321 to B738).

SELECT * FROM flight_aircraft_history WHERE callsign = 'AAL2763';

• Metadata caching (permanent, no TTL)
• USE_OPENSKY flag for offline operation
• Rate limiting (2 sec between calls)
• Retry failed lookups after 24 hours

• Leaflet map with aircraft icons
• Selection circles and tracking
• Altitude color coding
• External links (ADSBx, FlightAware)
• Copy aircraft info to clipboard

• Falsehoods Programmers Believe About Aviation
• FlightAware's ADS-B Flight Tracking Network
• Stories from the Cache Crimes Division

• OpenSky REST API
• The 1090 Megahertz Riddle (Mode S decoding)
• Radar Tutorial: SSR/Mode S

• hydra.toml - Platform configuration (IaC)
• VERSION - Platform version
• docs/adr/ADR-001-sqlite-vs-postgresql.md - Database decision
• docs/CHANGELOG-0.2.0.md - Release notes

• Playwright UI testing (hydra-setup-2nd)
• FlightAware AeroAPI integration (hydra-setup-bn1)
• Apache Baremaps custom aviation maps (hydra-setup-hdc)
• Route frequency analysis

• Callsign anomaly detection
• Flight path prediction
• Aircraft type classification from signatures
• Delay pattern analysis

Quick baseline for message rate and aircraft count:

# Message rate (10-second sample)
timeout 10 nc localhost 30003 | wc -l

# Aircraft count from dump1090
curl -s http://localhost:8080/data.json | jq '. | length'

# Pi receiver check
ssh pi.lan 'curl -s http://localhost:8504/data/aircraft.json | \
  jq "{aircraft: (.aircraft | length), with_position: ([.aircraft[] | select(.lat)] | length)}"'

Capture from both receivers simultaneously and compare coverage:

# Capture 15 seconds from both
timeout 15 nc localhost 30003 > /tmp/hydra_msgs.txt &
timeout 15 nc pi.lan 30003 > /tmp/pi_msgs.txt &
wait

# Compare message counts
echo "Hydra: $(wc -l < /tmp/hydra_msgs.txt) messages"
echo "Pi:    $(wc -l < /tmp/pi_msgs.txt) messages"

# Extract unique ICAOs and diff
cut -d, -f5 /tmp/hydra_msgs.txt | sort -u > /tmp/hydra_icao.txt
cut -d, -f5 /tmp/pi_msgs.txt | sort -u > /tmp/pi_icao.txt

# What each receiver sees exclusively
comm -23 /tmp/hydra_icao.txt /tmp/pi_icao.txt  # Hydra only
comm -13 /tmp/hydra_icao.txt /tmp/pi_icao.txt  # Pi only
comm -12 /tmp/hydra_icao.txt /tmp/pi_icao.txt  # Both

When swapping antennas:

1. Run quick baseline (pre-swap)
2. Swap antenna (USB may disconnect)
3. Check service status: systemctl status dump1090-mutability
4. Restart if needed: systemctl restart dump1090-mutability
5. Wait 30 seconds for stabilization
6. Run new baseline (post-swap)

Critical: USB disconnect causes dump1090 to lose device handle even though service stays running. Always restart after physical antenna changes.

For extended A/B testing during antenna swaps:

# Start hourly test (samples every 60 seconds for 1 hour)
cat > /tmp/receiver-comparison.sh << 'EOF'
#!/usr/bin/env bash
LOG=logs/receiver-comparison-hourly.log
echo "Time           H-msgs     H-ac   P-msgs     P-ac" > $LOG
for i in $(seq 1 60); do
  timeout 10 nc localhost 30003 > /tmp/h.txt &
  timeout 10 nc pi.lan 30003 > /tmp/p.txt &
  wait
  printf "%-12s %8s %8s %8s %8s\n" "$(date +%H:%M:%S)" \
    "$(wc -l < /tmp/h.txt)" "$(cut -d, -f5 /tmp/h.txt | sort -u | wc -l)" \
    "$(wc -l < /tmp/p.txt)" "$(cut -d, -f5 /tmp/p.txt | sort -u | wc -l)" >> $LOG
  sleep 50
done
EOF
nohup bash /tmp/receiver-comparison.sh &

Observed results from 2025-12-28 testing:

Key findings:

• Elevation matters more than antenna tuning
• Proper 1090MHz tuning provides more consistent reception
• Bookshelf placement (pi) outperforms window placement (hydra) by ~3x