Troubleshooting
Common NAP problems, their causes, and how to fix them — organized by symptom.
Updated June 11, 2026
Device shows unknown car or wrong Tesla model
Cause: NAP isn't running. Standard openpilot can't match the pre-AP car and will either fail or misidentify it as the HW1 AP variant.
Fix: Go to Settings → Software → Custom Fork and confirm the URL ends with NotAutopilot/nap-release. If it shows a different fork or plain commaai/openpilot, re-enter the NAP installer URL: https://installer.comma.ai/NotAutopilot/nap-release. See fingerprinting for more on how car identification works.
Steering won't engage — EPAS-related alerts
Symptoms: The device shows an EPAS error, steering is disabled, or engagement immediately drops.
Cause (most common): The EPAS ECU hasn't been patched, or a Tesla software update re-locked the steering firmware.
Fix: Go to Settings → NAP → Actions and run Backup EPAS first (saves a copy of current firmware), then Flash EPAS. Car must be on with key fob inside, foot on brake, parked. Read the full warning screen — power loss during the flash can brick the EPAS module.
Some EPAS firmware variants are not yet supported. If flashing fails or alerts persist after a successful flash, check EPAS firmware for the list of known variants and their status. Post your EPAS firmware version in Discord if yours isn't listed.
Controls mismatch alert
Symptoms: "Controls mismatch" or "CAN error" on screen; device disengages shortly after engaging.
Cause: What NAP sends on CAN and what the car reports back don't agree. Common causes:
- The panda's software is out of date. After a NAP branch update, panda firmware sometimes needs a re-flash.
- A CAN bus is noisy or has a failing termination.
- In pedal mode: the pedal is generating unexpected traffic that the safety model is rejecting.
Fix: Go to Settings → Software → Flash Panda and let the panda firmware update. If the alert continues, check that the adapter and cable connections are fully seated. For pedal-mode mismatch errors specifically, verify the pedal CAN bus setting matches your wiring (Bus 2 is standard through the OBD-C adapter's RJ45 port; Bus 0 for older direct-wired installs).
Pedal not detected or "pedal uncalibrated" blocking engagement
Symptoms: In pedal mode, NAP won't engage longitudinal control; pedal calibration status shows "Not Calibrated."
Cause: The pedal either isn't calibrated yet (new install), calibration was cleared (Emergency Disable was triggered), or the pedal is on the wrong CAN bus.
Fix:
- Confirm Pedal CAN Bus in the NAP settings matches your actual wiring.
- Run Calibrate Pedal — car on, in Neutral, brake held, don't touch the accelerator. The routine takes a minute or two and saves the result.
- If calibration fails repeatedly, check that the Comma Pedal firmware is current (see the pedal install guide) and that the pedal's USB-C or connector is seated.
Engagement will silently refuse pedal-mode longitudinal control if calibration isn't complete. The calibration status field on the NAP settings page will say "Not Calibrated."
Radar alerts or radar not tracking leads
Symptoms: Radar errors on screen; ACC doesn't respond to a car ahead; false lead-car alerts.
Cause (winter / dirty radar): The Bosch radar mounts at the front of the car, and its signal passes through the nose or grille area. Snow, ice, road spray, or a bug-covered nose will attenuate the signal significantly. This doesn't need a software fix — clean the front of the car.
Cause (nosecone mount attenuation): If the radar is mounted behind the nosecone and the Radar Behind Nosecone toggle is off, the expected signal strength won't match and the radar will behave poorly.
Cause (lateral offset): If the radar isn't centered on the car's axis, it will consistently report lead cars as offset. Set the Radar Lateral Offset in the NAP settings to correct this.
Cause (VIN learn not done or done incorrectly): The Bosch radar needs a VIN learn procedure during install to know the vehicle and mount position. If this was skipped or done with the wrong nosecone setting, the radar calibration state may be invalid.
Fix: Check the nose for obstruction first. Then verify the Radar Behind Nosecone toggle and Radar Lateral Offset are correct for your mount. Run Test Radar in the NAP settings to see what the radar is detecting right now. If you suspect a VIN learn problem, see Bosch radar.
Note: Tinkla had an "Ignore Radar Errors" toggle for riding out winter radar noise. NAP doesn't have one — persistent radar errors need an actual fix (clean nose, correct settings, or redo the VIN learn).
Stock cruise won't engage (no-pedal mode)
Symptom: Double-pulling the stalk doesn't set a speed; NAP steers but the car coasts.
Cause: The cruise master switch isn't armed. The orange light on the end of the cruise stalk must be on before NAP can spoof a CC engagement. NAP will never arm the master switch for you.
Fix: Press the stalk end in to arm stock cruise (orange light on), then double-pull to engage NAP. If the orange light won't come on, stock cruise itself may have a fault — check the car's instrument cluster for cruise-related messages.
Also confirm you're above approximately 18 mph (29 km/h). Stock CC won't engage below that speed regardless.
Device won't power on or restarts randomly
Symptoms: Comma device shows no picture, shuts off while driving, or loops through the boot screen.
Cause: Usually a 12 V power issue. The comma device runs off the accessory power connection in the harness, and pre-AP cars are known to have marginal 12 V rail voltage as the main battery degrades. Loose harness connectors are also common after vibration.
Fix: Check the OBD-II power connector and harness connections at the frunk passthrough. Check 12 V battery health. If the device boots fine on a bench but not in the car, the supply voltage may be dropping under load — measure it at the OBD port while the car is on.
See power for accessories if you've added other draws on the same rail.
NAP engages but disengages within a few seconds
Symptoms: Engagement succeeds (blue cruise icon appears) then drops 2–5 seconds later, often with no alert.
Cause: The panda safety model is seeing something it doesn't like — a steering angle rate exceeded, EPAS error codes, or a hands-on override. Or the device is detecting a hands-on torque level that's above the configured threshold.
Check: Look at Settings → Developer → Logs (or pull a drive from comma connect) and look for the disengagement reason. The most common causes in order:
- Hands-on force above threshold (happens when correcting a lane that's not straight)
- EPAS rejecting commands (firmware variant issue — see EPAS note above)
- CAN message timeout (bad connector or noisy bus)