Reading Blackbox without drowning

Blackbox is a mirror, not a magic wand. It shows what the flight controller thought happened — not whether your prop was bent. Start with three traces and one question: is this mechanical or software? Most pilots drown because they open a full-session log, zoom to 100%, and try to fix twelve problems at once. One maneuver, one change, one comparison.

When NOT to log (or not to trust logs)

Situation	Why logs mislead
Bent prop, loose motor	Gyro noise dominates — filters and PID look "wrong"
Motors heat-soaked	Noise floor up; afternoon log ≠ morning tune
Mixed changes last session	You cannot attribute anything
Logging rate too low for props	Peaks smeared — you guess wrong
First pack on fresh build	Voltage highest; tune always feels best

Fix props, tighten stack, cool motors. Logging a bent quad just prints noise you will try to fix with D-term — and then you will buy motors.

Setup: logging that helps

Before analysis:

Logging rate — 2K minimum for most 5"; 4K if you chase fine propwash and have SD space
Debug modes — start minimal: gyro, PID, setpoint; add FFT later when basics make sense
One bad move per log — fly deliberately: one hard corner, one flip, one punch-out
Note metadata — prop type, pack number, ambient temp; future you will forget

Betaflight Blackbox Explorer, Plotsplosion, or your preferred viewer — same ideas. The tool matters less than discipline.

Trace 1: Gyro (filtered vs raw)

Gyro is ground truth for what the FC sensed. Compare raw vs filtered gyro on roll or pitch.

Pattern	Likely meaning	Next step
Spiky raw, smooth filtered	Filters working hard	Check if stick feel is mushy — may be over-filtered
Spikes on throttle punches	Often prop/motor/mechanical	Inspect props, bearings, bell play
Constant fuzz all throttle	ESC noise, loose mount, bad bearing	Mechanical; see tuning after a crash
Spikes only one axis	Frame resonance or loose camera	Check that arm, mount, standoff
Raw and filtered both wild	Severe mechanical or desync	Land; bench test motors

Common mistake: cranking D because the gyro line looks "busy" without checking whether the quad felt bad in that moment. Correlate with stick input and throttle.

Trace 2: Setpoint vs actual (roll/pitch)

Setpoint is where the PID wants to be; gyro (or PID sum) shows response. This is the core tracking view.

Pattern	Likely meaning	Knob direction
Actual lags setpoint on snaps	P low, FF low, or over-filtered	Small P or FF bump; check filters
Actual overshoots then settles	P high or D low	Reduce P slightly or add D carefully
Oscillation on settle	P/D balance or mechanical	Do not only add D — check buzz source
Tracking good in hover, bad in turns	Propwash / throttle region	See propwash tuning
Slow wobble (1–3 Hz)	I windup or P too high on heavy quad	I down or P down — context matters

Zoom to one flip or one 180 — not the whole pack. Align setpoint step with the moment you felt washout.

Trace 3: Throttle

Throttle correlates problems to power regions:

Punch-outs — prop/motor load, ESC noise, filter stress
Mid-throttle cruise — resonance, bearing issues, ESC timing
Decaying throttle in turns — classic propwash zone

Draw a mental line: if ugly traces only appear when throttle drops in a hard turn, you are in propwash territory — not always "more P everywhere."

Motor RPM trace (with bidir DShot) helps confirm ESC telemetry is sane — erratic RPM alongside gyro fuzz means fix telemetry before RPM filter tuning. See RPM filter and bi-directional DShot.

Trace 4: PID sum and motors (when ready)

When basics are clear, look at PID sum and per-motor outputs:

Motor saturation — one motor pinned while others are not → mechanical twist, bad mix, or damaged motor
Oscillation in PID sum — tune issue if mechanicals clean
Asymmetric motors — crash damage, reversed motor, or ESC channel fault

Do not start here on day one — gyro + setpoint + throttle solve most field questions.

Learn traces in order: gyro → setpoint vs gyro → throttle → PID sum last. Note prop type, pack number, and maneuver name in a phone note — future you will not remember "the weird split-S" file.

Workflow

1. One bad maneuver per log — do not fix everything at once
2. Compare before/after ONE change (save profiles!)
3. Land when motors hot — logs lie when hardware is tired
4. Ask: did this trace look bad when I felt bad?
5. If no → look at rates, vtx, or pilot input; if yes → one knob

A/B comparison tips:

Same battery class and similar voltage
Same maneuver name — "left hand power loop" not "sort of a loop"
Overlay logs in Explorer if your tool supports it
Write the change on paper: "FF roll +5" not "tweaked stuff"

Symptom → trace → action map

In the air you feel	Look at	Often fix
Sloppy snaps	Setpoint vs gyro on roll	FF, P, or filter reduction
Buzz after flip	Gyro settle region	D + filter, not more P
Nose drop in dive	Throttle + pitch tracking	Mix habit, I, or tune — also camera angle
Mid-throttle hum	Gyro raw at cruise throttle	Mechanical, ESC timing, RPM filter
Gets worse each pack	Compare pack 1 vs pack 4 logs	Heat — hot climate filters

Common mistakes

Analyzing a 6-minute log without zooming to events
Changing PID because gyro looks noisy when feel is fine
Ignoring throttle column — propwash is throttle-dependent
No baseline log before changes
Trusting logs from a quad with a known bent prop "just to see"

When to stop staring at logs

If traces look mediocre but the quad feels great in goggles, stop tuning. Good enough beats perfect on a Sunday field.

If traces look fine but feel is bad, check rates, camera angle, vtx antenna, and radio link before another PID pass.

If traces and feel are both bad and mechanicals are verified, follow Betaflight tuning basics order: filters → P → I → D → FF.

Mechanical check before PID: noise on one motor channel = inspect corner; spikes at same throttle = prop/ESC; ugly only when stick moves = more likely tune. Swap props, re-log, then open configurator.

Blackbox rewards patience. One trace, one question, one change — then fly again. Your logs should get shorter over time because the quad stops misbehaving, not because you stopped looking.