[on8st]

Summary

On a Hermes-Lite 2, two receive controls had no route to the DSP that runs on this host:

• The AGC-T slider with AGC Off moved SliceModel::agcOffLevel, and the HL2 backend ignored the request, so every HL2 ran AGC Off at WDSP's channel default of 10 dB (WdspChannel::Config::agcFixedGainDb) whatever the slider said.
• The CW audio peaking filter (APF) was Flex wire text only (apf= / apf_level=). That is why HL2 declared hasAudioPeakingFilter = false and the APF row stayed hidden, although WDSP builds a peaking-filter stage into every receive channel.

Seam

• AGC-off level: main's typed route. SliceModel::setAgcOffLevel emits receiveAgcRequested(SliceAgcRequest::Field::OffLevel) → IRadioBackend::requestSliceAgc (aetherd: migrate receive controls behind the backend seam #5907). Hl2Backend now overrides requestSliceAgc: it handles Field::OffLevel itself and passes every other field to IRadioBackend::requestSliceAgc (the paired setSliceAgc). No new verb.
• APF: one new seam verb, interim. IRadioBackend::setSliceApf(sliceId, on, level) is a defaulted no-op. SliceModel emits apfCommandIssued(on, level) from the operator setters only, never from status application (Principle II), and RadioModel::wireBackendReceiverState() routes it next to the squelch route. It folds into aetherd: consolidate slice receive-control routing #5919's SliceDspRequest::Feature::Apf when aetherd: consolidate slice receive-control routing #5919 lands.
• Flex is unchanged. It overrides neither verb, its apf=/apf_level= text from SliceModel is byte-identical, and its agc_off_level= text comes from FlexBackend::requestSliceAgc as on main. Icom, RTL, Sim and ANAN keep the base no-op and still declare hasAudioPeakingFilter = false.
• The agcOffLevel branch in SliceModel::applyChanges() gets the same unchanged-value guard as the threshold beside it. HL2 now echoes the off level on every slice-state publish, and AgcCalibrationDialog wires agcOffLevelChanged to the calibrator.

Maintainer decisions (KK7GWY, 2026-10-02): (1) the AGC-off level uses main's requestSliceAgc route, not a parallel verb; (2) setSliceApf is accepted as an interim verb until #5919's request type replaces it.

HL2

• AGC-off level → SetRXAAGCFixed, via a new WdspChannel::setAgcFixedGain(). The mapping is gain_dB = 10 + 0.6·(level − 10), so the range is 4..64 dB. The default level 10 gives exactly 10.0 dB, so AGC Off sounds the same until the operator moves the slider. The slope matches the AGC-T threshold's 0.6 dB/unit. WDSP applies the fixed gain only in AGC mode 0, so the value is pushed in every mode and takes effect when AGC is switched off.
• APF → WDSP SPCW (apfshadow: SetRXASPCWSelection/Run/Freq/Bandwidth/Gain, declared in aether_wdsp.h), via WdspChannel::setApf():
  • The filter is centred on the CW pitch in audio Hz, and it follows a pitch change on every receiver, whether or not it is in CW at the time.
  • Level 0..100 sets the bandwidth, because the control is labelled "APF bandwidth" / "Higher values narrow the peak". The map is 200 Hz → 50 Hz, geometric. The default of 50 gives 100 Hz at gain 2.0, which is RXA.c's own construction design.
  • The stage runs only in CWL/CWU. The request is kept through a trip to SSB and comes back with CW.
  • open() selects the double-pole explicitly. An unchanged setApf() returns before the control handshake and the FFTW planner lock.
  • While it runs, both audio outputs derive from I (WDSP's mode-2 double-pole copies I into Q).
  • hasAudioPeakingFilter is now true for HL2. The caps map row says the host runs it.
• Both settings are held per receiver outside Config and applied again on channel rebuild, reconnect and new panadapter, the same way the noise blanker is.
• Both appear in dsp.chains as what WDSP last accepted (apfRun, apfCenterHz, apfBandwidthHz, apfGain, agcFixedGainDb).
• Not persisted, on purpose: like the host noise blanker. After a reconnect the off level is 10 again and the APF is off.

Tests

Both new tests are socket-free and registered in tests/tests.cmake.

• slice_apf_agc_off_seam_test (a recording backend installed through RadioModel's production receiver bindings):
  • Flex's APF wire text is byte-identical.
  • The operator setters emit the APF intent with enable and level together.
  • RadioModel routes the APF to setSliceApf and the off level to requestSliceAgc(Field::OffLevel) for the right slice, including through the AGC-T knob and the AGC-T calibrator's sweep and restore.
  • Status application never re-emits a command, and an echo at an unchanged off level notifies nothing.
  • The base verbs emit no signal, and the base drops OffLevel rather than calling setSliceAgc.
• hl2_apf_agc_off_test (a real Hl2RxDsp and WDSP channel, measured audio):
  1. the level maps;
  2. refusals, and an unchanged setApf() that does not wait on the planner lock;
  3. AGC Off: 50 units is 30 dB of audio, and with AGC on the off level changes nothing;
  4. APF in CWU lifts the pitch over a tone 200 Hz away by more than 10 dB, outside CW it is held and inert, and CWL is centred on the pitch;
  5. both survive a rebuild;
  6. through Hl2Backend: setSliceApf / requestSliceAgc(OffLevel) → WDSP → published slice state and dsp.chains. A pitch set outside CW is the centre on entering CW, and a Mode request still reaches the chain's AGC and leaves the off level alone.
• Mutation-checked (each goes red, and green again when restored):
  • sending the APF only to receivers already in CW;
  • Hl2Backend::requestSliceAgc not handling OffLevel;
  • not delegating other fields to the base.

Results at the head (Linux, RelWithDebInfo):

• Full build. Non-socket ctest 549/550 (including fix(radio): reroute controls a non-Flex radio dropped. Principle VIII. #6049's reroute_dead_controls_test); the one failure is connection_panel_size_test, which fails locally (Four tests fail deterministically on clean main locally: the a11y announcement fixture never receives focus (offscreen) #5811).
• Static gates: engine boundary, touchpoint manifest, capability records (frozen count unchanged), command plane, test registration and the CI test gate all pass, with findings identical to main's.
• on8st's earlier run: 556/556 on macOS.

Hardware evidence (receive only)

Hermes-Lite 2 (gateware 74) with the antenna switch on the dummy load, read through the automation bridge with TX blocked. This run was at an earlier head, when the AGC-off level still used its own verb. The HL2/WDSP half is the same code, and the route change is covered by the tests above.

#	Condition	Expected	Observed	Result
A1	AGC Off, off level 10	fixed gain 10 dB	10 dB	PASS
A2	AGC Off, off level 40	28 dB	28 dB	PASS
A3	AGC Off, off level 100	64 dB	64 dB	PASS
P1–P3	APF not requested (USB / CW / CWL)	not running	apfRun false	PASS
P4	APF on, CW	running at CW pitch	apfRun true, 600 Hz (pitch 600), 100 Hz, gain 2	PASS
P5	APF on, CWL	running at CW pitch	apfRun true, 600 Hz, 100 Hz, gain 2	PASS
P6	APF requested, USB	held, not running	apfRun false, model still on	PASS
P7	back to CW	running again	apfRun true, 600 Hz	PASS

Not covered

• No listening check of the APF on hardware. The audible effect is covered by the offline measurements (sections 3 and 4).
• The APF level slider was not swept on hardware.
• No transmit path is touched, and nothing was transmitted.

Update against main

Merged main into this branch (no rebase). The conflicts were adjacent additions next to #5982's squelch, #5980's pan averaging and #5981's filter length in Hl2Backend, Hl2RxDsp and WdspChannel, plus test registrations; both sides were kept. SliceModel::setAgcOffLevel took main's version, and the following commit moved the HL2 route onto it (decision 1 above). Two later merges brought in #6013 and #6049 cleanly; #6049's band-stack AGC recall reaches requestSliceAgc, which Hl2Backend passes to the base for every field but OffLevel.

Refs #5262, #5917.

73, ON8ST

🤖 Generated with Claude Code

[aethersdr-agent]

1. Issue fit

No linked issue. The PR title and the tests.cmake comment ("G2 of the silent HL2 control map") point at an internal work item I could not find with search_issues; the body does not carry a fixes/closes line. Reviewed against the PR's own stated intent, which is coherent and matches the diff: two controls that were Flex wire text only (apf=/apf_level=, agc_off_level=) now reach WDSP on a host-DSP backend. Per GOVERNANCE.md this is a bug fix with a clear root cause, so no RFC is owed; a one-line issue reference in the body would still help the next reader.

I checked that the fix actually reaches an operator, since that is the whole claim. It does: MainWindow.cpp:7909 gates the P/CW APF row on hasAudioPeakingFilter (now true for HL2), and VfoWidget::relayoutDspGrid/m_apfBtn->setVisible(isCw) shows the DSP-grid APF button on mode alone — applyRadioSideDspVisibility() does not AND m_hasRadioSideDsp into m_apfBtn, so that button was already visible on an HL2 in CW and simply did nothing. Both surfaces now work.

2. Scope

File / group	What it changes	Claimed by the PR?	Verdict
IRadioBackend.h	two new virtual seam verbs, default no-op	yes	In scope. New public surface (a backend interface third parties/other families inherit) — but it is the established setSliceNoiseBlanker shape, and slice_apf_agc_off_seam_test §5 pins the no-op. Worth a maintainer nod, not a blocker
SliceModel.{h,cpp}	two typed intents; guards d.agcOffLevel in applyChanges	yes (the guard is explained in-diff, not in the title)	In scope. The guard is load-bearing, not cosmetic — see §5
RadioModel.cpp	routes the two intents to the seam	yes	In scope
Hl2Backend.{h,cpp}	per-receiver state, push sites, capability flip, dspChains() readback, emitSliceState	yes	In scope
Hl2RxDsp.{h,cpp}	the two level maps, CW gating, re-apply across rebuild	yes	In scope
WdspChannel.{h,cpp}	setApf/setAgcFixedGain, Config fields, validateConfig, open() re-push	yes	In scope
third_party/wdsp/include/aether_wdsp.h	five SPCW declarations	yes	In scope; one is unused (nit 4)
tests/* + tests.cmake	two new registered targets, both joined to AETHER_SETTINGS_CONSUMERS	yes	In scope

Everything in the diff is explained by the stated intent. No CHANGELOG.md edit (correct). No new RadioCapabilities boolean — check_capability_records.py --strict reports 71 at the frozen 71. check_command_plane.py reports 413 across 32 files, unchanged, so no new raw Flex text lands above the seam. No new AppSettings keys. No UI restyle, no default flipped for an existing radio (the 37bee1c2 commit exists specifically to keep AGC Off at the 10 dB every HL2 already ran) — nothing here reads as smuggled preference.

No socket tests added or removed. Neither new test opens, binds, listens or connects; hl2_apf_agc_off_test links Qt6::Network but only pulls MetisProtocol.h for kEp6BlockSamples.

3. Blockers

1. The APF centre goes stale when the CW pitch moves while the slice is out of CW — inline on Hl2Backend.cpp:5398.

setCwPitch() skips a receiver entirely when cwBfoHz(r.mode) == 0.0, and cwBfoOffsetHz (Hl2Backend.cpp:288) returns 0.0 for every mode that is not CW/CWU/CWL. The new pushApf(r) sits inside that guard. setSliceMode() re-pushes mode, filter and shift when entering CW, but never the APF — so the centre Hl2RxDsp still holds is whatever the pitch was the last time this receiver happened to be in CW.

Sequence:

1. Connect. createPanadapter → pushApf(r) with the default 600 Hz; receiver 0 is in USB.
2. Operator sets the CW pitch to 800 Hz. setCwPitch(800) skips receiver 0 (not CW). Hl2RxDsp::m_apfCenterHz stays 600.
3. Operator switches the slice to CW with APF on. setSliceMode pushes shift and filter built from 800; Hl2RxDsp::setMode → applyApf() runs the peak at 600 Hz while the tone arrives at 800.

At the default level 50 the bandwidth is 100 Hz, so a 200 Hz error puts the wanted signal on the skirt: APF-on makes it quieter. That is verbatim the failure the same hunk's comment says it exists to prevent ("the peak stays on the old tone while the signal slides off it — the filter would be attenuating the very signal it exists to lift"). Nothing heals it: no radio echoes the APF, and installChannel() re-applies from the same stale m_apfCenterHz.

It also contradicts pushApf's own contract two hunks earlier — "this is safe to send in any mode and is sent unconditionally, like the blanker" — which is true at createPanadapter and pushInitialState but not here.

Either fix works; the second is the narrower change:

• add pushApf(*r); beside the setShift in setSliceMode's if (r->dsp) block, or
• move pushApf(r) above the cwBfoHz(...) == 0.0 → continue in setCwPitch, which is what the comment already claims.

The test would not have caught this. hl2_apf_agc_off_test §6 does setCwPitch(700) then setSliceMode(0,"CW"), which walks straight into the stale path — but it does not read apfCenterHz at that point, and the next setSliceApf(0, true, 80) pushes the pitch explicitly and masks it. A check of apfCenterHz immediately after the mode change into CW, with no intervening setSliceApf, is the assertion that fails today.

4. Nits (non-blocking)

• WdspChannel::setApf has no unchanged-parameters early-out (inline on WdspChannel.cpp:787). SetRXASPCWFreq/Bandwidth/Gain each land on SetRXADoublepoleFreqs/Gain, and doublepole.c has no guard of its own — each one runs CalcDoublepoleFilter, which rebuilds a ≥2048-tap FIR through build_doublepole_2sided's fftw_plan_dft_1d(..., FFTW_PATIENT). So one setApf() call is three full redesigns, under g_setupMutex (which is fftwPlannerMutex() — I checked, there is no planner race here), on the I/O thread that paces EP2. applyApf() now fires on every setMode, every installChannel, pushInitialState, createPanadapter, setCwPitch and every slider step, and open() adds three more per channel open. The case that bothers me is not the drag the header defends — it is that an operator who never enables the APF now pays three redesigns on every mode change, to keep a stage that is off, off. setMinimumPhase in the same class already has exactly this early-out and cites exactly this planner cost. return true when enabled/centerHz/bandwidthHz/gain all match m_config would remove essentially all of it.
• docs/architecture/radio-capabilities-map.md:150 now contradicts the code (inline on Hl2Backend.cpp:2409). That row still reads hasAudioPeakingFilter | ✅ | ❌ | ❌ with HL2 as the second column, and its Effect text says the row's only effect is Flex slice set <n> apf= — both untrue after this PR. AGENTS.md's capability rule asks for the caps-map doc alongside the declaration; the freeze checker passes because no boolean was added, so nothing catches the drift.
• check(true, ...) at slice_apf_agc_off_seam_test.cpp:224 asserts nothing and prints [PASS]. It proves the base verbs compile and don't crash, which is worth keeping — but MinimalBackend could record and the test could assert nothing was recorded, or the line could be a comment rather than a green check.
• SetRXASPCWSelection is declared in aether_wdsp.h:211 and never called. The design depends on selection staying at RXA.c's construction default of 0 (double-pole) — which I verified in RXA.c:507 — so the declaration is correct but dead. Either call it explicitly (cheap, and makes the dependency enforced rather than assumed) or drop the declaration.
• Neither value is persisted or restored, so a reconnect returns the off-level to 10 and the APF to off while the AGC-T threshold beside it is restored from m_restoredState.agcThreshold. Consistent with the noise blanker's existing precedent, so I read it as a decision — worth one line in the body so it reads that way.
• Enabling the APF forces the audio to mono. xdoublepole in mode 2 (which is how RXA.c builds it, RXA.c:443) does in[2i+1] = in[2i+0] — I is copied into Q before filtering. Harmless for ordinary CW, but if anything on this path ever produces genuinely different left/right audio in CW, APF-on collapses it. WDSP's own design, not this PR's, but undocumented at the seam.

5. What I tried to break, and could not

• The APF/AGC-off echo feedback loop. setSliceApf → emitSliceState → sliceChanged → applyChanges → apfChanged → VfoWidget::syncDsp. applyChanges guards all three (SliceModel.cpp:1524/1528/1602), so an echo at an unchanged value emits nothing and never re-enters a setter. The PR's claim here is accurate — and the new d.agcOffLevel guard is not cosmetic: AgcCalibrationDialog.cpp:280 wires agcOffLevelChanged straight into AgcTCalibrator::onValueChanged, and HL2 now publishes that field on every emitSliceState. Without the guard the calibrator would re-record a sweep point on every VFO step. The seam test pins exactly that.
• The AGC-off dB map against WDSP. SetRXAAGCFixed does fixed_gain = pow(10, dB/20) (wcpAGC.c:549) and xwcpagc multiplies by it only under if (a->mode == 0) (wcpAGC.c:167), so "applies exactly when AGC is Off" and "50 units = 30 dB of audio" both hold, and the default level 10 → exactly 10.0 dB with no rounding.
• The "higher level = narrower peak" direction. PhoneCwApplet.cpp:988/990 and VfoWidget.cpp:2116/2121 both label the slider "APF bandwidth" with "Higher values narrow the peak", so the decreasing geometric map matches what the UI promises. The default 50 → 100 Hz is RXA.c's own construction bandwidth (RXA.c:441). The FIR-length comment is right too: calc_dpole_nc divides by 1.7 then thresholds at 80/40, i.e. 2048 taps down to ~68 Hz bandwidth and 4096 below, at 48 kHz.
• The CWL claim. create_doublepole is built with mode 2 and xdoublepole mode 2 copies I into Q, so one positive centre genuinely serves both sidebands — the "CWL runs the APF on the positive pitch" assertion is not a guess.
• An FFTW planner race. This was my main hypothesis: setApf takes g_setupMutex but not fftwPlannerLock(), and the doublepole design creates plans. WdspChannel.cpp:59 binds g_setupMutex = AetherSDR::fftwPlannerMutex() — same mutex. Refuted.
• Vacuous refusal assertions. WdspChannel::Config bad; bad.apfBandwidthHz = 0.0; only proves something if the default Config is otherwise valid; it is (Receive, 48 k/48 k/48 k, USB, 150–3000), so create() fails on the new APF predicate and not incidentally. Non-vacuous.
• Rebuild and install ordering. installChannel applies the AGC-off level and the APF after setMode, which is required because apfInCircuit() reads m_config.mode; setMode assigns m_config.mode before calling applyApf(); setCwPitch assigns m_cwPitchHz before its loop. The deferred-push path (canPushToChannel() false during a background rebuild) holds both requests and re-applies at the swap. All correct.
• Flex. capabilities().hasAudioPeakingFilter was already true for Flex, the wire text is byte-identical (pinned), and both new seam verbs are base no-ops for every other family. setApf/setAgcFixedGain refuse a transmit channel. I found no Flex-visible change beyond the agcOffLevel echo now being edge-triggered, which is the intended fix.

Findings are reasoned from the code and the vendored WDSP sources in /tmp/aetherclaude/pr-6050, not reproduced at runtime — I have no build and cannot run the measured assertions. Worth noting alongside the green badge: neither new test is in any ctest -R step in ci.yml, so the five passing checks prove this compiles on three platforms and clears the static gates; the measured 30 dB and >10 dB assertions are the author's local evidence only. That is normal for this repo, not a criticism of the PR.

6. Recommendation

Request changes — one blocker, and it is a one-line fix. The engineering here is strong: the level maps are anchored on WDSP's own construction values rather than invented, the refusals are checked and logged instead of swallowed, the applyChanges guard anticipates a real calibrator hazard, and the tests measure audio rather than mirroring fields back at themselves — which is why the one path they don't cover is the one that breaks. Fix the stale APF centre (push the APF from setSliceMode, or drop the CW guard in setCwPitch so pushApf behaves the way its own comment says), add the assertion that reads apfCenterHz right after a mode change into CW with no intervening setSliceApf, and update the caps-map row. The setApf early-out and the maintainer call on the two new seam verbs can follow either way.

Thanks for the unusually legible comments on the WDSP side — the aether_wdsp.h block on apfshadow saved me a lot of time, and every claim in it checked out against upstream/.

---

_{🤖 aethersdr-agent · cost: $15.0086 · model: claude-opus-5}

[on8st]

Thanks for the review. The blocker was real, and so were four of the five nits. Each is handled in its own commit on top of 463386b0. No rebase and no force-push.

Bad news first.

• The blocker was a real defect, and our test walked straight past it. setCwPitch() sent the APF only to receivers already in CW. So a pitch moved while the slice was in USB never reached the receiver, and re-entering CW ran the peak on the old pitch. That contradicts pushApf()'s own comment, which says the APF is "sent unconditionally". Section 6 of hl2_apf_agc_off_test went through exactly that sequence. It read the centre only after a setSliceApf() that re-pushed the pitch, and that hid the fault. Your diagnosis is correct in every detail.
• One premise in the setApf nit does not hold for the vendored WDSP. I added the early-out anyway, for a different reason. CalcDoublepoleFilter has its own unchanged-parameters guard (third_party/wdsp/upstream/doublepole.c:263). It builds the FIR analytically (build_doublepole_1eff, doublepole.c:142, called at :272), and it plans an FFT only when the length changes (setNc_fircore, :274). build_doublepole_2sided is not on this path. A no-op setApf() therefore never caused "three full redesigns". A pitch or level change that really moves the design rebuilds it once per parameter that moved. What a no-op call did cost is below.
• The claim in our own aether_wdsp.h comment that "every call rebuilds the impulse response" was wrong for the same reason. It is corrected in 4aac7f1f.

Hl2Backend.cpp:5398, stale APF centre (38e3760a). I took your second option. pushApf(r) now runs for every receiver before the not-CW continue in setCwPitch() (Hl2Backend.cpp:5400). That is what its contract already said. Hl2RxDsp therefore never holds a stale centre, whichever path runs it later: a mode change, a rebuild or a reconnect. setSliceMode() alone would have covered only the mode change. The test now reads apfCenterHz right after entering CW, with nothing in between (tests/hl2_apf_agc_off_test.cpp:444). It also runs your sequence with the APF on: pitch moved in LSB, then back to CW, centre checked (:475).

WdspChannel.cpp:787, early-out (0947d1dc). setApf() now returns true before the control handshake and the lock when the request matches what WDSP holds (WdspChannel.cpp:800). This is the setMinimumPhase() shape. With WDSP's own guard in place, the saving is the lock, not a redesign. g_setupMutex is the process-global FFTW planner lock, and it was taken on the I/O thread that paces EP2. A no-op call made during another channel's rebuild waited out that rebuild's planning and changed nothing. A no-op call could also be refused while a callback was in flight. Exact equality is correct here, because m_config is written only by an accepted setApf() or open(), and nothing else writes SPCW. The test is section 2b (tests/hl2_apf_agc_off_test.cpp:245). It holds the planner lock itself and makes an unchanged call, which must return within 2 s. Without the early-out, that call blocks until the lock is released. The check then fails instead of hanging.

docs/architecture/radio-capabilities-map.md:150 (d04f80cc). The row now shows HL2 ✅ and says who runs the filter on each family. It also records that on HL2 the flag is an exception to hasRadioSideDsp, in the same shape as hasHostNoiseBlanker, and not a tier under it: the HL2 has no firmware DSP, and the P/CW gate reads this flag alone. The tier table's "Means" cell at :312 is widened to match.

slice_apf_agc_off_seam_test.cpp:224, check(true, …) (67645086). A base verb has no state to read back. So "no-op" can only mean what an observer sees: it emits nothing. The test now spies every signal in the backend's meta-object, calls both verbs, and asserts that no signal fired. It also asserts that more than one signal was being watched, so an empty spy list cannot pass it (tests/slice_apf_agc_off_seam_test.cpp:241, :243).

aether_wdsp.h:211, SetRXASPCWSelection (4aac7f1f). I chose to call it: open() now sets selection 0 explicitly, before the other SPCW calls (WdspChannel.cpp:1360). I did not drop the declaration. Everything this PR reasons about is the double-pole: apfParametersValid()'s divide by the centre, the level map against calc_dpole_nc, and one positive centre serving CWL and CWU through the mode-2 I-into-Q copy. If a WDSP update started create_apfshadow elsewhere, every SPCW call would land silently on the matched or gaussian filter, and both scale the gain by √2. Dropping the declaration would leave that assumption in place. Calling it turns the assumption into an invariant. To be clear about the test coverage: on today's WDSP the call is a no-op (apfshadow.c:48 returns early when the selection is unchanged), so no test can fail when it is removed. I then checked whether the existing measured tests would notice a wrong selection. They do not. With open() forced to selection 1 (matched), every measured APF check still passed, including the >10 dB lift and the CWL centre. So the explicit call enforces the invariant, and no test pins it. I didn't add one, because the only differences I could measure would be tied to a particular filter's exact response.

Review body, the other points.

• Issue fit. No issue describes this defect, so there is no Fixes line. I searched the tracker for the APF, the peaking filter, agc_off_level and the AGC-off level. The work is G2 of our per-radio inventory of silent controls, summarised in our #5262 comment, which names APF among the non-Flex slice controls that reach nothing. The nearest open issue is aetherd: consolidate slice receive DSP and audio control routing #5917, "consolidate slice receive DSP and audio control routing". Its scope includes moving APF encoding behind IRadioBackend, and aetherd: consolidate slice receive-control routing #5919 implements it. That is the routing half of this PR, not the HL2/WDSP half. Level AGC-OFF don't working when use from an external midi interface #5384 (AGC-OFF level from a MIDI controller, closed) is a different defect. I'll add Refs #5262, #5917 to the body.
• New seam verbs as public surface. Agreed that they need a maintainer's nod. The offer in the body stands: I can split feat(seam) into its own PR, or fold the two verbs into aetherd: consolidate slice receive-control routing #5919's request types when that lands.
• Neither value is persisted or restored. That is deliberate and matches the noise blanker's precedent. After a reconnect the off level returns to 10 (the 10 dB AGC Off every HL2 ran before this PR) and the APF returns to off. I'll add one line to the body saying so.
• The APF discards Q. Confirmed: xdoublepole mode 2 overwrites Q with I before filtering (doublepole.c:208-211). WDSP's panel stage is built with no copy (RXA.c:542), so without the APF the two output channels are I and Q. With it running, both are derived from I. We have not measured whether that difference is audible on this path in CW. It is now documented at the seam, in the setApf() comment in WdspChannel.h (4e95237a). There is no behaviour change.
• Neither new test is in a ctest -R step in ci.yml. Correct. The per-PR gate is frozen (check_ci_test_gate.py), so the measured assertions are local evidence. The results are below.

Verification. Each new check fails against a throwaway local commit that undoes its fix: pushApf back below the continue, the early-out disabled, and the base setSliceApf emitting sliceChanged. In that build, hl2_apf_agc_off_test fails on exactly its three new checks (2b and both centre checks), and slice_apf_agc_off_seam_test fails on exactly its new check. The other 554 tests pass. Full ctest at 4e95237a: 556/556 passed (Ninja, Qt 6.8.3, RelWithDebInfo, macOS arm64). The static checks pass at 4e95237a: engine boundary, touchpoint manifest, seam probe table and its scanner, test registration, capability records (71, frozen at 71) and their parser, command plane (413/413), CI test gate, network timeouts, theme seed and shader dialects. No src/gui file changed. Nothing was transmitted.

73, ON8ST

[ten9876]

We're updating this branch against main and landing it (maintainer pass); please don't push to it meanwhile. New commits go on top, with no rebase or force-push.

[ten9876]

Approving at 0d734cb5 (maintainer landing pass).

Maintainer decisions (KK7GWY, 2026-10-02):

1. The AGC-off level uses main's typed route (#5907), with no parallel verb. Hl2Backend::requestSliceAgc handles Field::OffLevel and passes every other field to IRadioBackend::requestSliceAgc. agcOffLevelCommandIssued and setSliceAgcOffLevel are gone, and SliceModel::setAgcOffLevel is exactly main's. Done in a new commit, 9fa82a48, with no history rewrite.
2. IRadioBackend::setSliceApf is accepted as an interim verb that folds into #5919's SliceDspRequest::Feature::Apf when #5919 lands. This is stated at the verb, in the body, and in a note on #5919.

Branch update (Wave 3): merged main three times, signed, with no rebase or force-push.

• bf052e5b: the conflicts were adjacent additions next to #5982 squelch, #5980 pan averaging and #5981 filter length, plus test registrations; both sides were kept. SliceModel::setAgcOffLevel took main's version.
• 4eb5da20 brought in #6013 and 0d734cb5 brought in #6049, both cleanly.

Review findings (aethersdr-agent, 2026-09-30):

• Blocker, stale APF centre: fixed by on8st (38e3760a). Mutation-checked here: sending the APF only to CW receivers turns both §6 centre checks red.
• setApf early-out: 0947d1dc. §2b passes.
• Caps-map row: d04f80cc.
• check(true): 67645086. Now rewritten for the requestSliceAgc route; it also asserts the base drops OffLevel.
• SetRXASPCWSelection called in open(): 4aac7f1f.
• Persistence and mono notes: in the body and the WdspChannel.h comment.

Landing-pass changes (9fa82a48):

• The route change and test rewrite above.
• Comments condensed to AGENTS.md § Comments, including a duplicated ADC-pairing block the merge had left.
• Mutation-checked: Hl2Backend::requestSliceAgc not handling OffLevel fails three §6 checks, and not delegating other fields to the base fails the new Mode check.

Verified empirically (Arch Linux, RelWithDebInfo, at 0d734cb5):

• Full build. Non-socket ctest 549/550. The one failure is connection_panel_size_test, which fails locally (#5811). Socket-owning tests were excluded; both new tests are socket-free.
• Static gates pass with findings identical to main's. The capability-record count is unchanged, and Flex is untouched (it overrides neither verb).

Not verified here: hardware. on8st's receive-only HL2 readings predate the route change. The HL2/WDSP half is the same code, and the new route is covered by the tests. No radio was used in this pass, and nothing was transmitted.

Green freshness: CI and Static Checks on 0d734cb5 were created at 2026-10-03T03:22:14Z. Main's tip is 5ef09b0f4 (#6049, 03:09:45Z), which is older, so the green covers current main.