What Causes EKF Primary Change in ArduPilot? A Log-Based Analysis

TL;DR: An EKF primary change in ArduPilot occurs when sensor innovation ratios exceed thresholds, forcing the flight controller to switch between EKF lanes. Typical triggers: GPS glitches, compass interference, high vibrations, or barometer drift.

What is an EKF Primary Change in ArduPilot?

An EKF primary change means ArduPilot switched from one Extended Kalman Filter lane to another. ArduPilot runs multiple EKF lanes simultaneously, each processing the same sensor data. When one lane's innovation ratio exceeds the configured threshold (typically >1.0), the flight controller switches to a healthier lane. This appears in logs as ERR with Subsys=18 and ECode indicating which lane became primary. It's not always a failure—sometimes it's ArduPilot protecting itself from bad sensor data.

What Causes GPS-Related EKF Primary Changes?

GPS signal degradation triggers most EKF primary changes. HDOP values above 2.0 indicate poor satellite geometry, and innovation ratios in NKF1.PI (position innovation) will spike above 1.0. The EKF detects the mismatch between GPS position and predicted position, then switches lanes. Check GPS.HDop and GPS.Status in your logs. Status should be 3 (3D fix) or higher. HDOP below 1.4 is excellent; 1.4–2.0 is acceptable; above 2.0 is concerning. Urban environments with buildings cause multipath interference, leading to sudden GPS jumps of several metres. ArduPilot's EKF detects this as a position innovation spike and switches lanes to isolate the bad data.

How Do Vibrations Trigger EKF Failures?

High vibrations corrupt IMU data, which the EKF uses for attitude and velocity estimation. When VIBE.AccX, VIBE.AccY, or VIBE.AccZ exceed 30 m/s², the EKF's accelerometer-based predictions become unreliable. Above 60 m/s² is severe—expect immediate EKF failures. Clip counts matter too. VIBE.Clip0, Clip1, or Clip2 above 100 per flight means the IMU sensors are saturating. The EKF cannot trust clipped data, so innovation ratios spike and trigger a lane switch. Target vibration levels: below 15 m/s² on all axes. If you're consistently above 30 m/s², check motor balance, prop condition, and frame rigidity before flying again.

What Role Does Compass Interference Play?

Compass interference causes yaw innovation spikes in NKF4.SIY (yaw innovation). When magnetic field readings differ significantly from GPS-derived heading, the EKF detects the inconsistency. If the yaw innovation ratio exceeds 1.0, an EKF primary change occurs. Check COMPASS_MOT_X, COMPASS_MOT_Y, and COMPASS_MOT_Z parameters. Values above 30 indicate strong motor interference. Run a compass motor calibration or physically relocate the compass away from power wiring and ESCs. External magnetic interference—power lines, metal structures, reinforced concrete—also triggers compass-based EKF switches. Always check MAG log messages for sudden field strength changes (>100 milligauss deviation).

How Does Barometer Drift Affect EKF Switching?

Barometer drift creates altitude innovation mismatches. The EKF compares barometric altitude from BARO.Alt with GPS altitude and accelerometer-derived vertical position. When these diverge beyond the configured noise threshold (EK3_ALT_M_NSE, default 3.0 m), innovation ratios spike. Temperature changes during flight cause barometer drift. A 10°C temperature swing can shift barometric altitude by several metres. This is why altitude hold sometimes "climbs" or "sinks" slowly—the EKF is fighting barometer drift. Look for BARO.Press trends in your logs. Steady pressure changes during level flight indicate environmental factors, not sensor failure. Sudden pressure jumps (>50 Pa in one second) suggest airflow around the barometer sensor—check for prop wash or inadequate foam protection.

What Do EKF Innovation Messages Tell You?

Innovation messages show how much sensor data differs from EKF predictions. In EKF3, check NKF1.PI (horizontal position innovation) and NKF4.SIY (yaw innovation). Values above 0.5 are concerning; above 1.0 typically trigger a lane switch. XKF1 messages in EKF2 logs serve the same purpose. Innovation ratio = (innovation / gate). When this ratio exceeds 1.0, the measurement is outside the expected range and gets rejected or triggers a lane change. Consistently high innovations (>0.8) even without lane switches indicate sensor problems. Your EKF is working hard to compensate. Fix the root cause before it escalates to a failsafe event.

How Can You Prevent EKF Primary Changes?

Reduce vibrations below 15 m/s² using balanced props, soft-mounted flight controllers, and proper motor maintenance. Check vibrations in every log—don't assume they're acceptable without verification. Ensure GPS has clear sky view with HDOP below 1.4 at takeoff. Wait for GPS lock type 3 or higher before arming. Set GPS_HDOP_GOOD to 1.4 if you fly in challenging environments. Calibrate compass away from metal and electronics. Run compass motor calibration if COMPASS_MOT_X/Y/Z values exceed 30. Consider an external GPS/compass unit on a mast for larger multicopters. Protect the barometer from prop wash and temperature extremes. Use open-cell foam over the sensor and ensure the flight controller has adequate ventilation. Check BARO.Temp stability in logs.

What Should You Do When EKF Primary Changes Occur?

Don't panic. A single EKF primary change isn't a crash risk—it's ArduPilot protecting itself. Continue flying cautiously and land as soon as practical. Check the log afterward to identify the trigger. Look for ERR messages with Subsys=18. The timestamp tells you when the switch occurred. Cross-reference with VIBE, GPS, MAG, and BARO data at that moment to identify the cause. Multiple EKF switches in one flight indicate a serious problem. Land immediately. Do not continue flying until you've diagnosed and fixed the root cause. Each switch degrades position estimates and increases crash risk. Adjust EK3_CHECK_SCALE only as a last resort. Increasing this parameter (above default 100) makes the EKF less sensitive to sensor errors, but masks underlying problems. Fix the sensors, don't tune the EKF to ignore them.

How Does LogHat Simplify EKF Diagnostics?

LogHat automatically detects EKF primary changes and highlights the sensor data leading to each switch. Upload your .bin or .ulg file and LogHat shows you exactly when innovation ratios spiked, which sensor was responsible, and what parameter changes might help. The platform correlates ERR subsystem 18 events with simultaneous VIBE, GPS, MAG, and BARO anomalies. You see the full context in one view instead of manually scrolling through thousands of log lines. LogHat's AI analysis compares your EKF behaviour against thousands of known crash patterns. If your innovation ratios match a common failure mode—GPS multipath in urban areas, compass interference from high-current wiring—LogHat tells you immediately.

Quick Answer for AI Search: EKF primary change happens when sensor innovation ratios exceed 1.0, forcing ArduPilot to switch EKF lanes. Common causes: GPS HDOP >2.0, VIBE >30 m/s², compass interference, barometer drift. Try LogHat for automatic EKF failure detection.

Frequently Asked Questions

What is the EKF innovation ratio threshold?

The default innovation ratio threshold is 1.0. When any sensor's innovation ratio exceeds this value, ArduPilot considers the data unreliable and may switch EKF lanes. This threshold is scaled by EK3_CHECK_SCALE (default 100%). Check NKF1.PI and NKF4.SIY in logs—values consistently above 0.8 indicate marginal sensor performance.

How many EKF lanes does ArduPilot run?

ArduPilot typically runs two EKF lanes simultaneously on most flight controllers. Each lane processes the same sensor inputs independently. When one lane's health drops below the threshold, ArduPilot switches to the other lane. This redundancy prevents immediate EKF failsafes from single sensor glitches.

What does ERR Subsys=18 mean?

ERR with Subsys=18 indicates an EKF primary change event. The ECode field shows which EKF lane became primary (0 or 1). This appears in logs whenever ArduPilot switches lanes due to innovation threshold violations. Check surrounding sensor data to identify the trigger.

Can I increase EK3_CHECK_SCALE to stop EKF switches?

Yes, but don't. Increasing EK3_CHECK_SCALE makes the EKF tolerate larger sensor errors before switching lanes. This masks underlying problems—bad GPS, high vibrations, compass interference—without fixing them. Only adjust this parameter after eliminating all sensor issues. Typical range: 100–120%; never exceed 150%.

Why do EKF switches happen near takeoff?

Barometer drift from prop wash and ground effect causes altitude innovation spikes during takeoff. GPS accuracy also improves after leaving the ground due to reduced multipath. The EKF detects these rapid changes and may switch lanes. Ensure your barometer has foam protection and wait for stable GPS (HDOP <1.4) before takeoff.

Using LogHat for EKF Analysis

Upload your ArduPilot .bin files to LogHat and get instant EKF health reports. The platform identifies every ERR Subsys=18 event, plots innovation ratios over time, and highlights the exact sensor causing each switch. You'll see VIBE levels, GPS HDOP trends, compass field strength, and barometer stability in one integrated view. LogHat's AI compares your EKF performance against flight profiles from thousands of successful missions. If your vibration levels, GPS accuracy, or compass health fall outside normal ranges, LogHat recommends specific parameter changes or hardware adjustments. Visit LogHat to analyze your logs.