…3241)

Standalone split of Fix #2 from #3194, as requested by @jensenpat.

## Problem

`CwSidetonePortAudioSink` picks audio output devices via PortAudio, but
on Windows:

1. **`defaultPortAudioOutputDevice()`** falls straight through to
`Pa_GetDefaultOutputDevice()`, which returns an MME device (the first
host API PortAudio enumerates). MME carries 50–150 ms OS-level
buffering, producing audible CW timing jitter on fast keying.

2. **`findPortAudioOutputDevice()`** matches devices by name substring.
On Windows, the same physical device is enumerated three times under
MME, DirectSound, and WASAPI. The old code treated this as an ambiguous
match and returned `paNoDevice` — causing the sidetone to fall back to
`CwSidetoneQAudioSink` and lose low-latency timing. When it did resolve,
it picked whichever host API appeared first (MME).

## Fix

**`defaultPortAudioOutputDevice()`** — add a WASAPI preference pass
before the `Pa_GetDefaultOutputDevice()` fallback, mirroring the
existing Linux JACK preference exactly. WASAPI shared mode runs at ~10
ms, vs 50–150 ms for MME.

**`findPortAudioOutputDevice()`** — collect all partial-match
candidates, then resolve to the WASAPI entry when exactly one exists.
Logs a clear `qCInfo` message showing which host API won. Falls back to
the existing "matched multiple" warning only if multiple WASAPI entries
are somehow present (pathological case).

## Scope

`CwSidetonePortAudioSink.cpp` only. No changes to `AudioEngine.cpp`, no
RX path changes, no buffer size changes.

## Related

- Partially addresses #3193 (CW sidetone latency component)
- May also help #2528 (TX→RX delay when stopping CW) and the latency
component of #2694
- Sibling of the held #3194 Fix #1 (RX WASAPI ring cap — separate review
track)

…3397)

## What

First, **pure-addition** step of the consolidated audio sink factory
(#3306): a single shared, OS-parameterized format/rate **negotiation
policy** that every audio sink and source will migrate onto, replacing
the ~9 divergent per-sink fallback ladders and per-OS `#ifdef` branches
that are the root of a recurring class of platform audio bugs.

Zero behaviour change — this is the maintainer-endorsed *"land the
helper first, migrate sinks incrementally"* approach (#3306, and the
#3194 thread's "sensitive audio-stack changes must be separate PRs with
soak time").

## Why this shape

The historical bugs escaped CI because the per-OS rate ladders were
compiled behind `Q_OS_*`, so a Linux runner could only ever exercise the
Linux path. The new policy is a **pure function over an injected
`DeviceCaps` snapshot with `TargetOs` as a parameter, not an `#ifdef`**
— so one headless, hardware-free test binary exercises the Windows,
macOS *and* Linux ladders on any runner.

## Contents

- **`src/core/AudioFormatNegotiator.{h,cpp}`** — dependency-free policy
(links only `Qt6::Core`). One ladder owns:
- Windows force-48k + r8brain (#2120/#2123), macOS 48k/A2DP (#1705),
Linux native-24k — divergences preserved as **data**, not `#ifdef`.
- The **universal 44.1 kHz rung** that RX/Quindar/QSO are missing today
(#3385).
- `Int16`↔`Float32` + `preferredFormat()` catch-all (#2669 / #1090 /
#3231).
- macOS mic preferred-rate-first (#2930) and Bluetooth-HFP native rate
(#2615); Windows probe-at-open (#2929).
- The two stereo resampler strategies — **PreservePan** (RX/QSO) vs
**MonoCollapse** (TCI-TX/RADE) — kept deliberately distinct
(#2403/#2459).
- **`tests/audio_format_negotiation_test.cpp`** — table-driven golden
matrix, registered under CTest. **25/25 pass**, including the
44.1k-only-device regression guard.
- **`docs/audio-sink-factory.md`** — the full three-layer design (pure
policy → live Qt wrapper → device-ownership router), the
device-following **"uncoupling"** fix (CW / RADE / Aetherial-Audio
"Pudu" recorder following the selected output), the regression-guard
invariants, and the one-sink-per-PR migration plan.

## Test

```
$ ./build/audio_format_negotiation_test
...
25/25 checks passed
```

## Follow-ups (per the migration plan in the doc)

Live Qt wrapper → migrate RX speaker → migrate PC mic →
`AudioOutputRouter` (closes the uncoupling class) → TCI-TX client-rate →
PipeWire DAX shared resampler. Each a separate, soakable PR.

Refs #3306

💻 Generated with Claude Code (Opus 4.8) with architecture by @jensenpat

Co-authored-by: Codex <noreply@openai.com>

…ry (#3306) (#3398)

> **Stacks on #3397** (the pure foundation). The first two commits here
are #3397's (foundation + live wrapper); the last two are the RX and
TX-mic migrations. Merge #3397 first and this collapses to just those
deltas. Kept as separate commits per the maintainer's
one-change-at-a-time / soak rule (#3194 thread).

## What

Layer 2 + the first two sink migrations of the consolidated audio sink
factory (#3306):

1. **Live Qt-Multimedia wrapper** `AudioDeviceNegotiator` — the thin,
and only, platform-specific bridge. `probe(QAudioDevice)` → pure
`DeviceCaps`; `negotiate()`/`formatLadder()` → openable `QAudioFormat`s.
Smoke-tested against the real default devices
(`audio_device_negotiator_test`, 8/8).
2. **RX speaker** — `startRxStream()` walks the factory's Float ladder
with real `start()` attempts instead of two forked per-OS `#ifdef`
blocks (**net −87 lines**). `m_resampleTo48k` (bool) → `m_rxOutputRate`
(int).
3. **PC mic (TX), macOS + Linux** — `startTxStream()` drives mic
rate/format from the factory's Int16 Input ladder (stereo-then-mono),
removing the per-OS rate lists and the bespoke
`macTxInputRateCandidates()` helper.

## Behaviour (preserved)

- **RX:** Windows/macOS → 48k (WASAPI 24k-resampler artifacts #2120;
macOS A2DP off HFP), Linux → 24k native. New strictly-additive wins: a
44.1k-only output now works (#3306/#3385); the Quindar local monitor now
opens on Windows (old branch `return`ed before
`startQuindarLocalSink()`).
- **TX mic:** standard macOS mic → 48k; Linux → 24k native;
**Bluetooth-HFP mic → native low rate first (#2615)**, fed in via
`macBluetoothNativeInputRate()` + the wrapper's new
`preferredRateOverride`; preferred-rate-first avoids the silent-48k-open
trap (#2930).

## Scope notes / invariants

- Two stereo resampler strategies kept distinct: **PreservePan** (RX,
dual L/R, VITA-49 pan #2403/#2459) vs **MonoCollapse** (RADE/TCI-TX).
- macOS telephony/HFP **output** substitution (#1705) unchanged (device
selection, before the ladder).
- **Windows mic left as-is** — already matches the factory's force-48k +
probe-at-open policy and carries the mono-only USB-mic channel clamp
(#2929) that needs its own soak. That + the device-following
`AudioOutputRouter`, TCI-TX client-rate, and PipeWire DAX are the
remaining increments (see `docs/audio-sink-factory.md`).

## Test / build

- `audio_format_negotiation_test` 25/25, `audio_device_negotiator_test`
8/8.
- Full macOS app builds & links clean (RelWithDebInfo); deployed to the
macOS test box. Soak on Windows/Linux requested before un-drafting.

Refs #3306

💻 Generated with Claude Code (Opus 4.8) with architecture by @jensenpat

---------

Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Jeremy [KK7GWY] <kk7gwy@aethersdr.com>

…#3193) (aethersdr#3241)

Standalone split of Fix aethersdr#2 from aethersdr#3194, as requested by @jensenpat.

## Problem

`CwSidetonePortAudioSink` picks audio output devices via PortAudio, but
on Windows:

1. **`defaultPortAudioOutputDevice()`** falls straight through to
`Pa_GetDefaultOutputDevice()`, which returns an MME device (the first
host API PortAudio enumerates). MME carries 50–150 ms OS-level
buffering, producing audible CW timing jitter on fast keying.

2. **`findPortAudioOutputDevice()`** matches devices by name substring.
On Windows, the same physical device is enumerated three times under
MME, DirectSound, and WASAPI. The old code treated this as an ambiguous
match and returned `paNoDevice` — causing the sidetone to fall back to
`CwSidetoneQAudioSink` and lose low-latency timing. When it did resolve,
it picked whichever host API appeared first (MME).

## Fix

**`defaultPortAudioOutputDevice()`** — add a WASAPI preference pass
before the `Pa_GetDefaultOutputDevice()` fallback, mirroring the
existing Linux JACK preference exactly. WASAPI shared mode runs at ~10
ms, vs 50–150 ms for MME.

**`findPortAudioOutputDevice()`** — collect all partial-match
candidates, then resolve to the WASAPI entry when exactly one exists.
Logs a clear `qCInfo` message showing which host API won. Falls back to
the existing "matched multiple" warning only if multiple WASAPI entries
are somehow present (pathological case).

## Scope

`CwSidetonePortAudioSink.cpp` only. No changes to `AudioEngine.cpp`, no
RX path changes, no buffer size changes.

## Related

- Partially addresses aethersdr#3193 (CW sidetone latency component)
- May also help aethersdr#2528 (TX→RX delay when stopping CW) and the latency
component of aethersdr#2694
- Sibling of the held aethersdr#3194 Fix aethersdr#1 (RX WASAPI ring cap — separate review
track)

…#3306) (aethersdr#3397)

## What

First, **pure-addition** step of the consolidated audio sink factory
(aethersdr#3306): a single shared, OS-parameterized format/rate **negotiation
policy** that every audio sink and source will migrate onto, replacing
the ~9 divergent per-sink fallback ladders and per-OS `#ifdef` branches
that are the root of a recurring class of platform audio bugs.

Zero behaviour change — this is the maintainer-endorsed *"land the
helper first, migrate sinks incrementally"* approach (aethersdr#3306, and the
aethersdr#3194 thread's "sensitive audio-stack changes must be separate PRs with
soak time").

## Why this shape

The historical bugs escaped CI because the per-OS rate ladders were
compiled behind `Q_OS_*`, so a Linux runner could only ever exercise the
Linux path. The new policy is a **pure function over an injected
`DeviceCaps` snapshot with `TargetOs` as a parameter, not an `#ifdef`**
— so one headless, hardware-free test binary exercises the Windows,
macOS *and* Linux ladders on any runner.

## Contents

- **`src/core/AudioFormatNegotiator.{h,cpp}`** — dependency-free policy
(links only `Qt6::Core`). One ladder owns:
- Windows force-48k + r8brain (aethersdr#2120/aethersdr#2123), macOS 48k/A2DP (aethersdr#1705),
Linux native-24k — divergences preserved as **data**, not `#ifdef`.
- The **universal 44.1 kHz rung** that RX/Quindar/QSO are missing today
(aethersdr#3385).
- `Int16`↔`Float32` + `preferredFormat()` catch-all (aethersdr#2669 / aethersdr#1090 /
aethersdr#3231).
- macOS mic preferred-rate-first (aethersdr#2930) and Bluetooth-HFP native rate
(aethersdr#2615); Windows probe-at-open (aethersdr#2929).
- The two stereo resampler strategies — **PreservePan** (RX/QSO) vs
**MonoCollapse** (TCI-TX/RADE) — kept deliberately distinct
(aethersdr#2403/aethersdr#2459).
- **`tests/audio_format_negotiation_test.cpp`** — table-driven golden
matrix, registered under CTest. **25/25 pass**, including the
44.1k-only-device regression guard.
- **`docs/audio-sink-factory.md`** — the full three-layer design (pure
policy → live Qt wrapper → device-ownership router), the
device-following **"uncoupling"** fix (CW / RADE / Aetherial-Audio
"Pudu" recorder following the selected output), the regression-guard
invariants, and the one-sink-per-PR migration plan.

## Test

```
$ ./build/audio_format_negotiation_test
...
25/25 checks passed
```

## Follow-ups (per the migration plan in the doc)

Live Qt wrapper → migrate RX speaker → migrate PC mic →
`AudioOutputRouter` (closes the uncoupling class) → TCI-TX client-rate →
PipeWire DAX shared resampler. Each a separate, soakable PR.

Refs aethersdr#3306

💻 Generated with Claude Code (Opus 4.8) with architecture by @jensenpat

Co-authored-by: Codex <noreply@openai.com>

…ry (aethersdr#3306) (aethersdr#3398)

> **Stacks on aethersdr#3397** (the pure foundation). The first two commits here
are aethersdr#3397's (foundation + live wrapper); the last two are the RX and
TX-mic migrations. Merge aethersdr#3397 first and this collapses to just those
deltas. Kept as separate commits per the maintainer's
one-change-at-a-time / soak rule (aethersdr#3194 thread).

## What

Layer 2 + the first two sink migrations of the consolidated audio sink
factory (aethersdr#3306):

1. **Live Qt-Multimedia wrapper** `AudioDeviceNegotiator` — the thin,
and only, platform-specific bridge. `probe(QAudioDevice)` → pure
`DeviceCaps`; `negotiate()`/`formatLadder()` → openable `QAudioFormat`s.
Smoke-tested against the real default devices
(`audio_device_negotiator_test`, 8/8).
2. **RX speaker** — `startRxStream()` walks the factory's Float ladder
with real `start()` attempts instead of two forked per-OS `#ifdef`
blocks (**net −87 lines**). `m_resampleTo48k` (bool) → `m_rxOutputRate`
(int).
3. **PC mic (TX), macOS + Linux** — `startTxStream()` drives mic
rate/format from the factory's Int16 Input ladder (stereo-then-mono),
removing the per-OS rate lists and the bespoke
`macTxInputRateCandidates()` helper.

## Behaviour (preserved)

- **RX:** Windows/macOS → 48k (WASAPI 24k-resampler artifacts aethersdr#2120;
macOS A2DP off HFP), Linux → 24k native. New strictly-additive wins: a
44.1k-only output now works (aethersdr#3306/aethersdr#3385); the Quindar local monitor now
opens on Windows (old branch `return`ed before
`startQuindarLocalSink()`).
- **TX mic:** standard macOS mic → 48k; Linux → 24k native;
**Bluetooth-HFP mic → native low rate first (aethersdr#2615)**, fed in via
`macBluetoothNativeInputRate()` + the wrapper's new
`preferredRateOverride`; preferred-rate-first avoids the silent-48k-open
trap (aethersdr#2930).

## Scope notes / invariants

- Two stereo resampler strategies kept distinct: **PreservePan** (RX,
dual L/R, VITA-49 pan aethersdr#2403/aethersdr#2459) vs **MonoCollapse** (RADE/TCI-TX).
- macOS telephony/HFP **output** substitution (aethersdr#1705) unchanged (device
selection, before the ladder).
- **Windows mic left as-is** — already matches the factory's force-48k +
probe-at-open policy and carries the mono-only USB-mic channel clamp
(aethersdr#2929) that needs its own soak. That + the device-following
`AudioOutputRouter`, TCI-TX client-rate, and PipeWire DAX are the
remaining increments (see `docs/audio-sink-factory.md`).

## Test / build

- `audio_format_negotiation_test` 25/25, `audio_device_negotiator_test`
8/8.
- Full macOS app builds & links clean (RelWithDebInfo); deployed to the
macOS test box. Soak on Windows/Linux requested before un-drafting.

Refs aethersdr#3306

💻 Generated with Claude Code (Opus 4.8) with architecture by @jensenpat

---------

Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Jeremy [KK7GWY] <kk7gwy@aethersdr.com>

Two independent contributors to RX speaker latency are addressed:

1. App-side RX buffer cap default lowered 200ms -> 100ms. The cap was
   originally set to 200ms (aethersdr#1505) as a conservative cushion for
   high-jitter links. For the common wired/LAN case that is avoidable
   latency. The setting remains user-adjustable (50-1000ms) for
   VPN/SmartLink users who need a larger cushion.

2. Explicit WASAPI shared-mode ring buffer on the Windows RX path. The
   RX QAudioSink was started with no setBufferSize(), so class-compliant
   USB interfaces inherited their driver's default ring (typically
   100-300ms), which stacked on top of the app-side cap and produced
   300-500ms+ speaker latency. We now request a 50ms device buffer
   before start() on Windows, comfortably fed by the existing 10ms RX
   drain timer and mirroring the explicit buffers already used by the
   sidetone and Quindar sinks.

The Windows device-buffer approach revives the RX half of the original
fix proposed in aethersdr#3194 by Ian M7HNF, which was never merged pending
on-hardware validation.

Defaults updated consistently in the engine, the MainWindow settings
load, and the Radio Setup dialog.

Co-Authored-By: Ian M7HNF <idimmock@gmail.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Two independent contributors to RX speaker latency are addressed:

1. App-side RX buffer cap default lowered 200ms -> 100ms. The cap was
   originally set to 200ms (aethersdr#1505) as a conservative cushion for
   high-jitter links. For the common wired/LAN case that is avoidable
   latency. The setting remains user-adjustable (50-1000ms) for
   VPN/SmartLink users who need a larger cushion.

2. Explicit WASAPI shared-mode ring buffer on the Windows RX path. The
   RX QAudioSink was started with no setBufferSize(), so class-compliant
   USB interfaces inherited their driver's default ring (typically
   100-300ms), which stacked on top of the app-side cap and produced
   300-500ms+ speaker latency. We now request a 50ms device buffer
   before start() on Windows, comfortably fed by the existing 10ms RX
   drain timer and mirroring the explicit buffers already used by the
   sidetone and Quindar sinks.

The Windows device-buffer approach revives the RX half of the original
fix proposed in aethersdr#3194 by Ian M7HNF, which was never merged pending
on-hardware validation.

Defaults updated consistently in the engine, the MainWindow settings
load, and the Radio Setup dialog.

Co-Authored-By: Ian M7HNF <idimmock@gmail.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…3897)

## Problem

RX speaker latency is higher than it needs to be, most visibly for
Windows users on external class-compliant USB audio interfaces (#3193),
where the TX→RX transition and general monitoring lag by 300–500 ms+.
Two independent code paths contribute:

1. **App-side RX buffer cap is conservative.** `m_rxBufferCapMs` has
defaulted to **200 ms** since it was introduced (#1505) as a cushion for
high-jitter links. For the common wired/LAN case that is avoidable
latency.

2. **The Windows RX sink never constrains its device buffer.**
`startRxStream()` constructs the RX `QAudioSink` and calls `start()`
with no prior `setBufferSize()`. WASAPI shared mode then applies the
device's *default* ring buffer, which for USB interfaces (Scarlett Solo,
Focusrite, etc.) is typically **100–300 ms**. That stacks on top of the
app-side cap.

## Changes

**1. Lower the RX buffer cap default 200 ms → 100 ms.**
The setting remains fully user-adjustable (50–1000 ms) for VPN/SmartLink
users who need a larger cushion — only the default changes, and existing
users who have set their own value keep it. Updated consistently in
three places: the engine default (`AudioEngine.h`), the settings load
(`MainWindow.cpp`), and the Radio Setup dialog field
(`RadioSetupDialog.cpp`).

**2. Set an explicit 50 ms device buffer on the Windows RX path.**
Before `start()`, on `Q_OS_WIN`, we call
`setBufferSize(candidate.bytesForDuration(50 ms))`. 50 ms is comfortably
fed by the existing 10 ms RX drain timer and mirrors the explicit
buffers already used by the sidetone and Quindar sinks. macOS/Linux are
unchanged (guarded by `#ifdef`).

Combined, these take the worst-case Windows USB path from ~300–500 ms
toward ~150 ms, without touching the high-jitter escape hatch.

## Credit

The Windows device-buffer approach (change 2) revives the RX half of the
original fix proposed by **@M7HNF-Ian** in #3194, which was never merged
pending on-hardware validation. Ian is credited as a co-author on the
commit. (The CW-sidetone half of #3194 shipped separately as #3241.)

## Testing

- Built clean on macOS (RelWithDebInfo, Qt 6 / Homebrew). The Windows
device-buffer block is `#ifdef Q_OS_WIN` and not exercised by the macOS
build.
- **Needs Windows on-hardware validation** with a USB interface (the
#3193 reporter's Scarlett Solo Gen4 is ideal) to confirm the latency
reduction and that 50 ms does not introduce underruns. Left as **draft**
pending that.

Addresses #3193.

💻 Generated with Claude Code (Opus 4.8) with architecture by @jensenpat

Co-authored-by: Ian M7HNF <idimmock@gmail.com>
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>