propwash

Drone flight log analyzer. Parses Betaflight blackbox (.bbl), ArduPilot DataFlash (.bin), and PX4 ULog (.ulg) logs to detect vibration issues, PID tuning problems, motor saturation, gyro spikes, and mechanical faults - with frequency spectrum analysis and actionable diagnostics.

Try it now: propwash.deltave.org - drop a log file in your browser. No install, no uploads. Runs locally via WebAssembly.

$ propwash analyze flight.bbl

── Session 1 ──
  Firmware:    Betaflight 4.2.9
  Craft:       DIATONE ROMA F5
  Duration:    89.0s
  Sample rate: 249 Hz
  Motors:      4

  Timeline (68 episodes):
      10.486s  GYRO SPIKE     pitch peak 677°/s (47 frames) (185ms)
      10.647s  OVERSHOOT      pitch peak 205% (14 frames) (52ms)
      25.429s  MOTOR SAT      motor[0] for 8 frames
      88.255s  GYRO SPIKE     pitch peak 2032°/s (103 frames) (414ms)

  Vibration Analysis (full flight):
    roll axis — noise floor 22.3 dB:
      #1: 205 Hz at 45.2 dB
      #2: 51 Hz at 32.0 dB
    pitch axis — noise floor 21.3 dB:
      #1: 218 Hz at 41.4 dB

  Vibration by Throttle:
    0-25%:   roll dominant 34 Hz    ← propwash
    25-50%:  roll dominant 205 Hz   ← frame resonance
    75-100%: roll dominant 21 Hz    ← PID oscillation

  Diagnostics:
    !!  [motors] Motor[0] saturating significantly more than others
    !!  [pid] Severe pitch overshoot (peak 372%, avg 78%)
    !!  [mechanical] Extreme pitch gyro spike (2032°/s)

Web UI

Features: tabbed interface (Overview, Timeline, Spectrum, Raw Data), interactive ECharts plots with zoom/pan, spectrogram heatmaps, filter overlay, throttle-band analysis, side-by-side flight comparison.

Install (CLI)

Download a prebuilt binary from Releases, or build from source:

cargo install --git https://github.com/Iteratrix/propwash propwash

Usage

# Full analysis with event timeline, vibration FFT, and diagnostics
propwash analyze flight.bbl

# JSON output for programmatic consumption
propwash analyze flight.bbl --output json

# Log metadata and field inventory
propwash info flight.bbl

# Batch triage of multiple files
propwash scan *.bbl *.bin *.ulg

# Side-by-side comparison of two flights
propwash compare before.bbl after.bbl

# Raw frame dump (for debugging / agentic analysis)
propwash dump flight.bbl --session 2 --frames 100-200 --fields gyroADC,motor

What it detects

Events:

• Throttle chops and punches (rapid throttle changes)
• Motor saturation (motor hitting max output)
• Gyro spikes (extreme rotation rates)
• Setpoint overshoot (PID tracking errors)
• ESC desync (single motor spike)
• Firmware messages (PX4 log messages, ArduPilot error codes)

Vibration:

• Frequency spectrum per axis (FFT with Hann windowing)
• Top 5 frequency peaks with noise classification (motor noise vs frame resonance)
• Throttle-windowed analysis (how vibration changes with motor speed)
• Spectrogram (frequency vs time heatmap)
• Accelerometer vibration (FC mounting quality)

Diagnostics:

• Asymmetric motor saturation → mechanical issue on one side
• Severe overshoot → P/D imbalance with axis-specific recommendation
• Extreme gyro spikes → crash/damage detection
• Frame resonance → notch filter recommendation
• Throttle-dependent frequency shift → motor noise (RPM filtering)
• FC mounting issues → accelerometer RMS analysis

Supported formats

Firmware	Format	Status
Betaflight	.bbl blackbox	Full support
EmuFlight	.bbl blackbox	Full support
Rotorflight	.bbl blackbox	Full support
INAV	.bbl blackbox	Full support
Cleanflight	.bbl blackbox	Full support
ArduPilot (Copter/Plane/Rover)	.bin DataFlash	Full support
PX4	.ulg ULog	Full support

Architecture

propwash-core/          Parser library (can be used independently)
  format/bf/            Betaflight-family blackbox decoder
  format/ap/            ArduPilot DataFlash decoder
  format/px4/           PX4 ULog decoder
  analysis/             Event detection, FFT, diagnostics
propwash/               CLI binary
propwash-web/           WASM bridge (powers the web UI)
web/                    Browser frontend (vanilla JS + ECharts + uPlot)

Two-layer API for library consumers:

• session.unified() — format-agnostic sensor data via the Unified trait
• session.raw — format-specific parsed data (behind features = ["raw"])

Typed field access:

use propwash_core::types::{SensorField, Axis};

let gyro = session.unified().field(&SensorField::Gyro(Axis::Roll));

Performance

Parses a 15MB Betaflight log (472K frames) in 250ms. 45x faster than Python parsers.

License

MIT