What changed, and whether an old report is still valid

Current build: 1.0.5+rMj426Y. Entries are written for pilots, not for a release feed: if a change alters a number, it says so and says whether to regenerate an old report.

1.0.5+rMj426Y — measurement honesty and a crawlable site

Step response rebuilt. Both curves are normalised by the commanded step, rise and settle are in milliseconds, the window stops at stick release, and the chart carries a confidence grade. Reports made before this build can show 0 % overshoot where the new estimator measures real overshoot. Regenerate those reports from the original log.
Bandwidth formula corrected. The old estimate was 1000 / (2π × rise × 1.6), which produced roughly 9 Hz for a 10.7 ms rise and was then clamped to 18 Hz — so the gap to the profile target was always positive and gain was always raised. The estimate is now the 10–90 % rise-time approximation, 350 / riseMs.
Settling compared to a settle target, not to a frequency. A millisecond value was being compared with a hertz target, which added D-term on every log. Each profile now has a settle target in milliseconds.
Weak charts cannot raise gain. Master, P and feedforward are capped at 1.0× unless the chart is chart-grade. See /method.
One support address. Every screen, email and fulfilment reply now uses the same inbox, and the previous address that did not exist is gone.
HTML report export. The report can be downloaded as a single HTML file next to JSON and Print/PDF, so it opens offline without a browser print sheet.
Static pages. /demo, /sample-report, /guides, /compare and the keyword pages now return their own HTML to a crawler instead of the app shell, with their own canonical URLs.
Upload path. The first action on the homepage is the upload. Files over 48 MB are sent in parts, with progress and a retry.
Privacy-friendly funnel events. Twelve events record buckets — file size, firmware family, confidence, outcome — and never a filename, hash, email, key or PID value.

How to read an older report

The JSON export contains the build it was made with. If that build predates the step-response rebuild, compare only the spectrum, filter and diagnostic sections; treat every overshoot, rise and settle value as superseded. The cheapest fix is to re-run the same log. The first log of a fresh install is free, and a previously used free log is still held on the device so it can be analysed without uploading it again.

Reports are not silently updated. Nothing on the server rewrites a report you already downloaded. A new analysis produces a new report; the old file is yours.

Found a number that looks wrong? Send the log's header section and the report JSON to easyquadtune@outlook.com. A reproducible disagreement is the most useful bug report this project gets.

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.