[bnsgeyer]

This PR provides a pretty extensive update to analytic tune.
Adds dynamic page build to accommodate both attitude and position controller tuning
Adds SID Axes for Plane
Update params to include changed params in 4.7

[AP-Review]

Deprecated — see below for the updated review.

Previous review (2026-10-07)

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: REQUEST CHANGES

Reviewed at head 8b1a0ced78.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/webtools/337/1.html#prWebTools-337

Thanks for the big overhaul. I reviewed head 8b1a0ce and reproduced the problems below with synthetic SysID logs.

The main one: the notch parameters from the log no longer reach the form. load_log sets INS_HNTCH_/INS_HNTC2_ before the form exists, and setup_FFT_data doesn't reload them (https://github.com/ardupilot/webtools/pull/337/files#diff-3d5b6156863de28ac11e95577f9808c1e7f02f0cf9fb8865666d25dc7892880cR1213). Roll/Pitch/Yaw predictions therefore always use the default 150 Hz / 75 Hz enabled notch. Moving the notch loop, including set_bitmask_size, into setup_FFT_data should fix it.

Other things to look at:

• Calculate rebuilds the form, so edited or loaded gains are reset to the log values (https://github.com/ardupilot/webtools/pull/337/files#diff-3d5b6156863de28ac11e95577f9808c1e7f02f0cf9fb8865666d25dc7892880cR1142). Load Parameters before the first Calculate throws.
• The change handler returns for SELECT elements, so changing notch enable or mode has no effect (https://github.com/ardupilot/webtools/pull/337/files#diff-3d5b6156863de28ac11e95577f9808c1e7f02f0cf9fb8865666d25dc7892880cR1122).
• On Vertical-Accel, 'Position ctrl only' also applies der_H to the measured acceleration curves (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1188).
• That branch passes H_pilot rather than H_pilot_tf (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1215). The new k < H_acft[0].length+1 loops also produce one extra point.
• intrate divides by trecord instead of multiplying by the sample period (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1539). That is 40x too large for 10 s at 400 Hz.
• Save Parameters on the vertical pages omits PSC_VELZ_* or PSC_ACCZ_* and the FILT notch selected by NEF/NTF (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1810).
• Minor: a Plane log with axis 24 is treated as fixed-wing (the axis > 19 check), a double Calculate click can decorate the form twice, and some debug console.log calls are left in.

Which ArduPilot PR adds SID axis 24 and the Plane position axes? I couldn't find them in master.

Not covered: accuracy against a real flight log, and real-browser checks of the fixed-wing and quadplane paths.

[AP-Review]

Deprecated — see below for the updated review.

Previous review (2026-10-08)

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: REQUEST CHANGES

Reviewed at head 807a548f15.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/webtools/337/2.html#prWebTools-337

Thanks for the notch fix. I re-checked at 807a548, and the logged INS_HNTCH_/INS_HNTC2_ values now reach the form on the first Calculate, so that earlier problem is resolved.

The main remaining issue: Calculate rebuilds the whole parameter form and rewrites it from the log (https://github.com/ardupilot/webtools/pull/337/files#diff-3d5b6156863de28ac11e95577f9808c1e7f02f0cf9fb8865666d25dc7892880cR1142). Gains that were edited or loaded with Load Parameters are silently reset when Calculate is pressed again, for example for another SID run or time window. Load Parameters before the first Calculate throws. Base kept the edits. Building the form only when the vehicle or axis layout changes, or carrying values across the rebuild, would fix it.

Still open from last time, unchanged at this head:

• The change handler returns for SELECT elements, so notch enable and mode changes do nothing (https://github.com/ardupilot/webtools/pull/337/files#diff-3d5b6156863de28ac11e95577f9808c1e7f02f0cf9fb8865666d25dc7892880cR1122).
• Vertical-Accel 'Position ctrl only' differentiates the measured acceleration curves (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1197), and passes H_pilot rather than H_pilot_tf (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1224). The new length+1 loops also add points.
• intrate divides by trecord instead of multiplying by dt, about 40x too large for 10 s at 400 Hz (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1548).
• Save Parameters on the vertical pages omits the other PSC_VELZ_/PSC_ACCZ_ group and the NEF/NTF FILT notch (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1819).

New this round: the Lateral/Longitudinal loader reads DVN/VN and DVE/VE into the position variables (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R1497), so the position response is really a velocity response. Should those be TPN/PN and TPE/PE? If velocity was intended, a comment would help.

Minor: the new console.log("PARM", PARM) (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R704) and other debug logs; the set_bitmask_size loop left in load_log (https://github.com/ardupilot/webtools/pull/337/files#diff-0dad09add812fbabafe507751f37abfbed999d85b065b6ded70ea1c01a0bc027R714) never takes effect; a double Calculate click can decorate the form twice.

The question about which ArduPilot PR adds SID axis 24 and the Plane position axes is still open.

Not covered: accuracy against a real flight log, and real-browser checks of the fixed-wing and quadplane paths.

[AP-Review]

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: REQUEST CHANGES

Reviewed at head 05c47d4c4c.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/webtools/337/3.html#prWebTools-337

Re-reviewed at 05c47d4: one blocking regression remains, plus several should-fix problems in the new position and vertical pages. Since 807a548, the DRB velocity integration has been fixed and now scales correctly.

Blocking

1. AnalyticTune/index.html:1142 Calculate resets edits (still open): every click rebuilds the form and reloads log values, so typed or imported gains are lost, and Load Parameters before the first Calculate throws. Keep the form when the layout is unchanged, or carry the values across the rebuild.

Should fix

• AnalyticTune/AnalyticTune.js:1497 Position reads velocity (still open): PSCN/PSCE position arrays read DVN/VN and DVE/VE; use TPN/PN and TPE/PE.
• AnalyticTune/AnalyticTune.js:1426 Heading wrap: the arithmetic mean of DesYaw (logged 0..360) gives about 180 for a north-facing vehicle and flips the body-frame signals; use atan2(mean sin, mean cos).
• AnalyticTune/index.html:1122 SELECT changes ignored (still open): notch enable and mode changes do nothing.
• AnalyticTune/AnalyticTune.js:1818 Vertical export incomplete (still open): Save Parameters drops the other PSC_VELZ_/PSC_ACCZ_ group and the NEF/NTF FILT notch.
• ESLint on current main: main now runs ESLint on PRs, and main plus this PR gives 150 errors, mostly undeclared variables. Rebase, declare them, then refresh the suppressions.

Minor (5)

• AnalyticTune/AnalyticTune.js:1197 Vertical-Accel der_H (still open): 'Position ctrl only' also differentiates the measured acceleration curves.
• AnalyticTune/AnalyticTune.js:1224 H_pilot vs H_pilot_tf (still open): passing the DC bin shifts the prediction by one, and the new position-predictor length+1 loops (AnalyticTune.js:437) add a trailing non-finite point.
• AnalyticTune/AnalyticTune.js:704 Debug logs and bitmask sizing (still open): console.log("PARM", PARM) and others; the set_bitmask_size loop at AnalyticTune.js:714 is undone by the rebuild.
• AnalyticTune/index.html:1208 Double-Calculate race (still open): two quick clicks decorate the form twice, and the next log load throws.
• AnalyticTune/AnalyticTune.js:2332 Axis 24 (still open): a Plane log with axis 24 is classed as fixed-wing; which firmware PR emits axis 24 and the Plane position axes?

Not fully reviewed: accuracy against a real SysID flight log, real-browser checks of the fixed-wing and quadplane paths, and the live CI result.