Free tool · Betaflight 4.x
Betaflight Blackbox analyzer
Upload your FPV quad’s Blackbox log and get what noise it has, where it comes from and what to change in PIDs and filters, with the CLI commands ready to paste into Betaflight. The log is analysed in your browser and never uploaded.
The log is read entirely in your browser: it is never uploaded to a server.
Diagnosis
Suggested CLI
Before pasting, save a diff all so you can roll back. Apply the changes in small groups, fly 1–2 minutes and touch the motors: if they come back hot, undo the last change.
Gyro noise
Noise peaks
Noise versus throttle
A diagonal band rising with throttle is motor noise (the RPM filter removes it). A horizontal band at a fixed frequency is a frame resonance. The white dots mark the dominant peak at each throttle level.
Step response
How the quad follows a sharp stick input. Ideally it rises quickly to 1.0 and stays there. Above 1.0 means overshoot (not enough D or too much P); a slow rise means not enough P. Dashed lines have low confidence.
Settings read from the log
What it analyses
Motor and frame noise
Gyro spectrum before and after the filters, and a noise map versus throttle. It separates motor noise, which rises with RPM, from frame resonances, which stay at a fixed frequency.
Filters
It checks whether the peaks get through the filters and suggests the fix: RPM filter with bidirectional DShot, dynamic notch range and Q, a static notch centred on a resonance, and gyro and D-term filtering.
PIDs
Roll, pitch and yaw step response from the setpoint and the gyro, cross-checked with the sharp moves in the flight. It measures overshoot and delay and suggests raising D, lowering P or raising P, with a confidence level.
CLI ready to paste
Commands with the parameter names of your Betaflight version (4.2, 4.3, 4.4 or 4.5) and, if you use the sliders, moving the sliders. With a table of current and suggested PIDs.
Indicative recommendations calculated with rules from your log. Every quad is different: apply changes gradually and check motor temperature. AIRLAB FPV is not affiliated with Betaflight.
Guide: how to tune PIDs and filters with the Blackbox
How to tune Betaflight with the Blackbox
- Set up logging: in the Blackbox tab, at 2 kHz or more. At 1 kHz you only see noise up to 500 Hz and motor noise folds back to false frequencies.
- Record the unfiltered gyro: on Betaflight 4.4 and 4.5 the
gyroUnfiltfield is logged by default; on 4.3 or earlier useset debug_mode = GYRO_SCALED. Without it you only see the noise that has already passed the filters. - Fly a test log: 1–2 minutes with good props, slow throttle sweeps from 0 to 100 % and, at the end, 15–20 s of sharp roll and pitch moves (full stick and release).
- Filters first, then PIDs: remove motor noise and resonances; once the signal is clean, adjust P and D from the step response.
- Change a little and check: apply each group of changes, fly 1–2 minutes, touch the motors and record a new log.
Motor noise and the RPM filter
Motor noise rises in frequency with RPM: in the noise versus throttle map it shows up as a diagonal band, sometimes with its harmonics. The RPM filter removes it by tracking the real RPM of each motor, sent by the ESCs through bidirectional DShot telemetry. It needs ESCs running Bluejay, BLHeli_32 or AM32, and motor_poles must match the motor magnets (14 on most 5" motors). On Betaflight 4.3 and later it is enabled with dshot_bidir and rpm_filter_harmonics.
Frame resonances and notch filters
A horizontal band, at the same frequency at any throttle, is a mechanical vibration: arms, stack, camera or loose screws. Before filtering it, check the build. If it is still there, the dynamic notch tracks it when it falls inside its range (dyn_notch_min_hz and dyn_notch_max_hz); otherwise a static notch centred on its frequency (gyro_notch1_hz and gyro_notch1_cutoff) removes it with little delay.
Step response: P and D
The step response shows how the gyro follows a sharp change in setpoint. If it goes above 1.0 and oscillates, it lacks damping: raise D or lower P. If it arrives late without overshooting, it needs more P. With a noisy D term, raising D heats the motors, so the analyzer suggests lowering P instead. Feedforward also speeds up the response: if the overshoot appears right at the start of the move, try lowering it first.
Delay versus noise
Every filter adds delay. With a clean signal (RPM filter on, good props and motors) you can filter less and the quad handles propwash better; with a lot of noise you need more filtering so the motors do not overheat. The analyzer compares the noise before and after the filters to tell you which way to move the filter sliders.
If you are building the quad, start with the FPV build workshop and the FPV drone configurator.
Other free tools
Questions
Which logs does it accept?
Betaflight Blackbox logs (.BBL and .BFL), from the SD card or downloaded from the onboard flash with Betaflight Configurator, and CSV files exported with Blackbox Explorer or blackbox_decode. It also reads EmuFlight and INAV logs in the same format, although the recommendations are designed for Betaflight 4.x.
Is my log uploaded to a server?
No. The log is read and analysed entirely in your browser. It never leaves your computer or phone.
How is it different from Blackbox Explorer?
Betaflight’s Blackbox Explorer shows the graphs and the spectrum, but you draw the conclusions. This analyzer reads the same log, detects the noise peaks and the PID response, and tells you what to change, with the CLI commands.
How should I record the log?
At 2 kHz or more (set blackbox_sample_rate = 1/2 with an 8 kHz loop), with the unfiltered gyro (recorded by default on 4.4 and later; on 4.3 or earlier, set debug_mode = GYRO_SCALED), 1–2 minutes of flight with throttle changes and 15–20 s of sharp roll and pitch moves.
Can I trust the recommendations?
They are a starting point calculated with rules from your log, with a confidence level on each PID change. Apply the changes in small groups, fly, touch the motors and record a new log to check the result.
What is the step response?
It is how the gyro follows a sharp stick input, calculated from the setpoint and the gyro in the log. A good response rises quickly to the requested rate and stays there without overshooting or oscillating.
What if it does not suggest any PID changes?
If the log does not have enough sharp moves, the analyzer does not risk suggesting PIDs. Record 15–20 s of snap rolls and flips on roll and pitch, on top of normal flying, and analyse it again.