fix(hl2): the ALC gauge is described as what it measures. Principle VIII.

[on8st]

Fixes #6235

Option 1 from #6235, in the form @NF0T and I settled on in the thread. No ruling was given between options 1 and 2; if the envelope peak (option 2) is wanted, close this.

Limits, first

• Text only. No reading changes. The gauge still shows the same number: on an HL2, the peak of the audio block after the ALC, before the modulator.
• The gap this describes was measured offline, on synthesized speech through the transmit DSP (figures in HL2: the ALC gauge shows the audio peak before the modulator, and is described as the SSB peak #6235): with limited audio the envelope stood 5 to 6 dB above the gauge. Nothing was keyed. Real microphone audio was not measured.
• Flex operators read one word less. The shared accessible description loses "SSB" and becomes "post-software-ALC peak (dBFS)". It stays true for the Flex firmware meter of the same name; docs/architecture/flex-meter-learnings.md and tx-audio-signal-path.md, which describe that Flex meter, are not touched.
• Since feat(hl2): transmit AM, DSB and FM through TXA. Principle II. #5977 the same gauge also runs in AM, DSB and FM on an HL2 (one audio path, Hl2TxDsp::processAudioBlock), so dropping "SSB" is correct there too. Not keyed in any mode.
• No test pins these strings, before or after.

Change

• PhoneCwApplet.cpp (three sites), PhoneCwApplet.h, MeterModel.h: "post-software-ALC SSB peak (dBFS)" becomes "post-software-ALC peak (dBFS)".
• Hl2TxDsp.cpp: the comment that said the value is "the signal reaching the wire" now says it is the audio peak before the modulator, at the ALC target while the ALC reduces.
• Hl2Backend.cpp: two comments, and the HL2 meter's description "Post-ALC transmit peak" becomes "Post-ALC audio peak, before the modulator".
• docs/user/docs/hermes-lite-2.md: one line saying what the ALC gauge shows on an HL2, and that it is not the transmitted envelope.

Full ctest at c5886fa0 (08859e04 merged with main 7864ef5; macOS, Qt 6.12.0, -j8): 801 tests, 799 passed.

• vkamp_connection_test failed and passed on the serial rerun (Replace retained socket fixtures with injected and opt-in coverage #5254).
• kpa500_advanced_widget_test timed out. It does the same on clean main 94e3cd4 here, and only in our launchd test runner: it hangs in backupSelectionRechecksReadAvailability. Started from a terminal it passes 9 of 9 in 0.3 s. This PR touches no KPA500 file.

At c5886fa0, check_engine_boundary, check_test_registration and check_qsettings (--strict) and gen_touchpoint_manifest.py --check pass. check_a11y on the two PhoneCwApplet files at 08859e04: 2 findings, the same 2 as on the base.

73, ON8ST

🤖 Generated with Claude Code

[jensenpat]

Reviewed exact rebased head 73ceed328eefe5d867b11b17cbc186d64b1583b2 against main 3fc45ebe7340863b257c6acd3d81274ec20627da. No blocking findings. Rebase was conflict-free; the contributor's authorship is preserved and the rebased commit has a verified signature.

RadioModel/seam isolation: RadioModel, IRadioBackend, RadioCapabilities, Flex/Icom backend directories, touchpoint manifest and tags are byte-identical to the base. No new includes, family branches, signals, timers, routing, commands, TX gates or persistence changes. Manual adversarial review completed; a separate automated code-review skill was unavailable.

The C++ diff contains only comments and two description substitutions. After removing comments and accounting for those exact substitutions, executable tokens match the base in all five changed C++ files. Shared Phone/CW accessibility text drops “SSB” from “post-software-ALC SSB peak (dBFS)”; Flex readings, units, scales and TX gating are unchanged. Icom retains its unchanged Percent branch and “radio-reported percent” description. HL2's meter description, comments and guide correctly identify the existing post-ALC/pre-modulator audio measurement; no DSP or measurement changes.

Validation on this head: all four boundary/manifest/capability/command-plane checks pass with identical base/head findings; git diff --check passes; generated docs reference check passes (8 pages); log-rule check passes (707 checks, 45 rules). No local build, CTest, bridge or live-radio test was run for this wording-only patch. Fresh hosted CI and Docs checks are pending and must pass before merge. Operator explicitly authorized the wording and merge after the isolation explanation.

Generated with OpenAI Codex (GPT-6 Astra)