What the analysis does, in four minutes — full transcript

This is the written version of the walkthrough linked from /video. It is the complete explanation, not a summary: the video adds the screen recording, and nothing else. Numbers below come from the same published log as /demo.

1. Start with the log, not the tune

Open the report and look at the log-quality score before any chart. It answers one question: can this flight support a tuning decision at all? A hover-only log, a log that is mostly noise, or a log with no usable step cannot measure tracking, and no amount of analysis makes it able to.

If the score says the log is not good enough, stop. The guide on log quality explains exactly which manoeuvre is missing and how to fly it. Uploading a second bad log costs more time than flying one good one.

2. Read the spectrum before the step response

The spectrum lists the frequencies present in the gyro signal. A motor-noise peak and a frame resonance are different faults, and the frequency, the sharpness of the peak and whether it tracks throttle tell them apart. What the spectrum cannot do is name the loose screw.

This matters for trust: a filter recommendation is traceable to a measured peak. If you cannot see the peak in the chart, do not apply the filter change.

3. Read the step response one axis at a time

For the published flight, roll measured 8.63 % overshoot with a 54.85 ms rise and a 86.65 ms settle. Pitch measured 28.85 %, 47.45 ms and 69.79 ms. Yaw measured 22.42 %, 100.62 ms and 157.02 ms. All three were graded good.

Overshoot and droop look similar on a moving chart and have opposite fixes. Overshoot is the peak above the command: usually too much D, or a filter that lets a resonance through. Droop is the response settling below the command: an I or feedforward problem, and raising D makes it worse. The confidence grade beside each axis says whether the chart is allowed to change gain at all — on a weak step, master, P and feedforward are capped at 1.0×.

4. Read the before/after table, then the provenance

Every line states the old value, the new value, the rule that fired and the measurement that triggered it. The method page defines each of those measurements. If a line does not make sense to you, do not paste it: a tune you cannot explain is a tune you cannot debug at the field.

The table is also where the tool is allowed to say "no change". A chart-grade response inside the targets produces almost nothing, and that is a correct result, not a failed analysis.

5. Copy the CLI and change one profile at a time

Each report ships a CLI block for Cine, Freestyle and Race. Pick the one that matches how you fly, paste it, and change nothing else in the same session. Then fly the verification flight: a hover, some hard stops, a couple of fast turns, and a check of motor temperature after landing.

The verification is the actual test. The comparison between the original log and the re-log is the only evidence that the change helped, and it is why the report asks for a second upload rather than declaring success on its own.

What this walkthrough does not prove

That the tune will feel right to you. A chart-grade response is tracking, not preference.
That the aircraft is mechanically sound. A cracked arm can produce good-looking gyro data and still fly badly.
That the same values transfer to another build. The CLI is derived from this log and this header.

Questions about a specific step belong in the Discord, or by mail to easyquadtune@outlook.com.

Easyquadtune is not affiliated with Betaflight. Published 2026-10-21. Measurements come from the maintainer's own flight logs; an example is illustrative of the method, not a tune for your aircraft.