A neckband low-latency TX/RX is a closed-loop, paired system: it exists to keep end-to-end latency and latency jitter below the “lip-sync / motion-sync” threshold while surviving crowded 2.4 GHz conditions without audible dropouts. The engineering boundary is defined by four coupled budgets: latency, RF robustness, clock drift, and power integrity.

Rule of writing depth for this page: every claim must be backed by a measurable signal (timing marker, histogram, rail waveform, or RF counters), and must stay inside the TX/RX paired hardware boundary.

• End-to-end latency (E2E): target < 20–30 ms (define whether the measurement includes host input buffering; lock the definition in H2-1 evidence).
• Latency jitter: control p95 / p99 latency; avoid spikes that “feel like stutter” even if the mean is low.
• RF robustness: keep audio stable under interference by bounding retry depth and enabling rapid channel recovery (use PER/retry counters as evidence).
• Audio output: stable THD+N/noise floor at target output power; prevent pop/click during link recovery and power events.
• Voice uplink (if present): intelligibility under wind/friction noise; avoid uplink scheduling that steals downlink latency budget.
• Power/thermal: rails must survive burst TX current and output power steps without brownout or PLL unlock; verify with synchronized rail captures.

Depth cue: targets are not marketing specs; they are testable guardrails that drive the later “budget table” (H2-2) and “SOP” chapter (H2-11, later).

1) TX-side (input boundary only): The TX ingests audio from USB Audio or analog line-in (or an external extractor feeding line-level). The page focuses on what the interface does to the budgets: sampling clock source, format changes, headroom/noise, and the timing marker location.

2) Codec + packetization: The TX must convert continuous audio into small frames, apply a low-latency codec (or a fixed PCM mode), then packetize for RF with a bounded retry/ACK strategy. This is where fixed latency is created (codec frame) and where jitter risk begins (RF scheduling).

3) 2.4 GHz RF link: The link is optimized for low latency: short packets, fast ACK, limited retransmissions, and rapid channel hop/recovery. Success is quantified by PER / retry-rate / channel-switch counts aligned to audio artifacts.

4) RX-side output: The RX reconstructs audio using a bounded jitter buffer, decodes, then outputs via DAC + headphone amp (or open-speaker driver). The RX must prevent audible artifacts during buffer underflow, link recovery, and power transients.

5) Optional uplink (multi-mic talkback): If the neckband includes a mic array, uplink audio can share the RF schedule or use an independent channel. Either way, uplink must not cause unbounded downlink jitter; this page treats it as a resource scheduling and AFE evidence problem—not an app/OS tutorial.

6) Power tree as a first-class budget: The PMIC partitions rails for RF/SoC/AFE/HP-amp. In practice, many “RF problems” are power droop or ground return problems; therefore power and timing markers must be measured together.

Boundary lock: No LE Audio/Auracast stack details, no UHF system planning, no cloud/app walkthrough. Only the paired TX/RX hardware budgets and evidence chain.

A. Define your latency “start” and “end” points (must be stated explicitly):

• Start (TX input time): where the audio enters the system boundary (USB sample capture point or ADC input). Add a GPIO marker at the exact frame boundary.
• End (RX output time): at the transducer drive node (HP-amp output) or DAC output (if HP-amp adds additional processing).

B. Minimum instrumentation that makes debugging deterministic:

• 2-channel scope (one on audio output, one on GPIO marker) + optional current probe on battery/rail.
• RF counters (PER, RSSI, retry-rate, channel-hop count) exposed via debug UART or log.
• Audio test tone/impulse file + correlation method (for statistically robust latency and jitter histogram).

C. “Two measurements first” rule (fast discriminator):

• Measurement #1: E2E latency + jitter histogram (p50/p95/p99). If p99 spikes correlate with retries, the cause is in H2-3/H2-4 later; if spikes correlate with rail droop, it’s power/ground (H2-8/H2-9 later).
• Measurement #2: RX buffer level (or underrun flags) aligned with RF retry bursts. Buffer underrun without RF degradation indicates clock drift/PLL issues (H2-7 later).

Depth cue: the goal is not a single “average latency” number; it is a repeatable measurement definition that lets every later chapter point back to one of these evidence streams.

“Low latency” is only meaningful when it is decomposed into a budget worksheet. In neckband TX/RX systems, the user notices jitter spikes (p99) more than the average delay. Therefore this chapter defines: (1) fixed vs variable latency segments, (2) the tuning knobs, and (3) the evidence signals that prove each segment.

Depth rule: each line item below must map to a measurement (timing marker, counter, buffer level, or waveform). If it cannot be measured, it cannot be controlled.

• E2E mean latency: within the target band (e.g., < 20–30 ms by your definition in H2-1).
• Jitter limits: control p95 / p99; define a maximum acceptable spike window (stutter threshold).
• Audio stability under loss: audible artifacts must be minimized at a defined stress level (PER sweep), with bounded recovery time.

Practical note: a system can have “fast mean” and still feel bad if retries or clock drift cause buffer depth oscillation. The budget must include those control loops.

Use this worksheet to avoid “blind tuning.” For each row: identify whether it is fixed latency, variable latency (jitter), or both; then decide which knob you can safely turn.

Key insight: you cannot “solve jitter” by codec alone. In practice, p99 spikes are driven by retry bursts, buffer policy, and power/clock excursions. Those must be visible in evidence streams.

• Frame length ↓ → latency ↓, but compression efficiency ↓ and loss sensitivity ↑ (PLC becomes more audible).
• Bitrate ↓ → airtime ↓ (often helps in crowded RF), but audio artifacts ↑; be careful not to “mask” RF issues with quality loss.
• Retry limit ↑ → fewer drops, but p99 spikes ↑; the fix is bounded retry + fast recovery, not unlimited retry.
• Jitter buffer depth ↑ → stability ↑, but constant lag ↑; the correct knob is policy (when to expand/shrink), not only “more buffer.”
• Uplink scheduling → talkback quality ↑, but downlink jitter risk ↑ if slots are not isolated; enforce a maximum jitter budget for uplink contention.

Depth cue: every trade-off must be verified by a stress test: PER sweep + coexistence scenario + synchronized histogram. Otherwise, tuning is just guesswork.

Method 1 — Impulse + correlation (best for statistics):

• Inject a known impulse (or MLS/chirp marker) at the TX boundary and record the RX output.
• Compute cross-correlation to extract E2E latency per trial; repeat to build a histogram (p50/p95/p99).
• During the test, log RF counters (PER/retry/hop) and RX buffer level; correlate spikes to retries vs buffer policy vs clock drift.

Method 2 — GPIO marker + scope (best for root cause):

• Toggle a GPIO exactly when a frame enters packetization (TX) and capture it alongside the RX analog output waveform.
• Trigger on audio artifact events (click/pop) and inspect whether a retry burst or a rail droop occurred at the same moment.

Method 3 — Stress matrix (must include at least these axes):

• RF crowding: channel occupancy / interferer on/off / human-body detune posture.
• Power events: charging on/off; volume step; TX burst duty cycle ramp.
• Clock stress: temperature ramp; long-run drift observation to reveal buffer sawtooth.

Output artifacts without evidence alignment are non-actionable. Always align: latency histogram spikes ↔ retry bursts ↔ buffer level ↔ rail waveform.

In low-latency audio, the limiting factor is rarely raw PHY data-rate; it is unbounded waiting (contention) and runaway recovery (unlimited retries). A proprietary 2.4 GHz link with a custom MAC is chosen to enforce deterministic timing: short packets, fast ACK, bounded retransmissions, and rapid channel recovery—so that p99 latency remains controlled in crowded spectrum.

Boundary lock: this chapter explains PHY/MAC mechanisms only at the level needed to control latency and robustness. It does not teach Wi-Fi/Bluetooth stacks.

• PER / retry-rate guardrails: define an acceptable PER band under a specified stress scene, with retry-rate bounded so p99 latency does not explode.
• Maximum retransmissions: set a hard retry limit per audio frame (or per packet group). Anything beyond the limit is intentionally dropped and handled by PLC (H2-4).
• Fast recovery time: bound “bad channel” recovery (channel hop / redundant channel) to a short window so buffer policy can remain tight.
• Uplink coexistence rule: downlink audio must retain guaranteed airtime; uplink voice is scheduled in bounded windows and cannot create unbounded downlink jitter.
• Airtime budget: keep the link load low enough that retry bursts do not cascade into queueing delay (verify with duty-cycle/airtime estimates).

Depth cue: targets must be stated as limits (bounded retry count, bounded recovery time), not as vague “robust link” claims.

1) Slot/Polling vs contention: to control p99 latency, a low-latency link prefers a scheduled access pattern (time slots or polling) so that each audio frame has a predictable transmit opportunity. Contention-based access can increase throughput but often creates unpredictable waiting time.

2) Short packets + fast ACK: short packets reduce the “damage radius” of interference bursts. Fast ACK confirms delivery quickly so the sender can either proceed or trigger a bounded retry before the audio frame’s deadline expires.

3) Bounded retransmissions (deadline-aware): retries are limited not by a generic link-layer rule, but by the audio playout deadline. Once a packet group exceeds its retry budget, the system intentionally drops it and relies on PLC/jitter-buffer strategy (H2-4) to maintain listenability.

4) Channel plan (hop / redundancy): recovery is driven by short-term indicators (retry bursts, sudden PER increase), not by long averaging windows. The goal is fast return to a “good” channel before buffer expansion becomes audible as delay.

5) Downlink/uplink coexistence: when downlink audio and uplink voice share the schedule, enforce a guaranteed downlink slot. Uplink is allocated bounded windows (or a separate lane) so uplink activity cannot cause unlimited downlink jitter growth.

Engineering handshake with H2-4: the loss shape created by bounded retry (isolated losses vs bursts) must match PLC assumptions and buffer-state transitions.

A. Must-log counters (per second or per frame window):

• PER (packet error rate), RSSI, retry-rate (or retry burst length distribution)
• Channel hop count and hop reason flags (retry burst trigger vs periodic pattern)
• Airtime / duty-cycle estimate (payload size × packets/sec × retries)
• RX buffer level / underrun flags (from H2-4) to connect link behavior to playout stability

B. Correlation workflow (turn counters into root cause):

• Step 1: generate latency histogram (p50/p95/p99) while logging counters.
• Step 2: locate p99 spike windows, then overlay retry bursts and hop events.
• Step 3: if spikes align with retries → MAC/radio contention; if spikes align with buffer underflow but PER is normal → clock/power/scheduling (later chapters).

Depth cue: success is defined by bounded behavior. Even in bad RF, the system should choose “bounded loss handled by PLC” over “unbounded waiting.”

The codec largely sets the fixed latency floor (frame + decode), but “stable listening” in a low-latency system is dominated by jitter-buffer policy and how it cooperates with bounded retries (H2-3) and PLC (packet loss concealment). The goal is to keep p99 latency controlled while shaping loss artifacts into forms that PLC can mask without sounding like stutter.

Boundary lock: focus on codec frame/bitrate/PLC/buffer-control only. Do not expand into app DSP tutorials or LE Audio/Auracast topics.

• Audio baseline quality (clean RF): set a reference THD+N/noise floor at the target output level. This prevents “codec blame” for analog issues.
• Maximum jitter-buffer depth: cap buffer growth so recovery cannot turn into “hidden added delay.” The cap must respect the H2-2 budget.
• PLC audibility threshold: under defined loss injection, PLC artifacts should remain below the annoyance threshold (use consistent test content).
• p99 latency stability: downlink p99 must remain bounded under stress; a system that “feels slow sometimes” is often buffer-policy oscillation.

Depth cue: “max buffer depth” is a safety rail. If it is missing, the system can silently trade user experience for stability and still pass simple “mean latency” checks.

1) Frame length sets the fixed floor: smaller frames reduce base delay but are more sensitive to loss and scheduling jitter. Larger frames improve efficiency and mask some loss but increase fixed delay. The selection must be aligned to the H2-2 latency worksheet.

2) Bitrate/bitpool affects airtime and loss shape: lower payload can reduce airtime (helpful in crowded RF), but also increases compression artifacts and can change PLC audibility. This is why loss injection must be run on the final codec settings.

3) PLC is the “bounded loss safety net”: with bounded retries (H2-3), some packets will be intentionally dropped. PLC must be tuned to the observed loss distribution (isolated vs burst) so artifacts remain less annoying than stutter.

4) Jitter-buffer policy dominates perceived stability: the buffer is not just a fixed number of milliseconds; it is a control loop. Correct behavior is a three-mode policy: Normal (minimum delay) → Stressed (absorbs retry bursts) → Recovery (shrinks slowly with hysteresis).

5) Avoid policy hunting: if buffer expands quickly and shrinks immediately, listeners feel “fast/slow” oscillation. Recovery must require a stable window (low retry bursts) before reducing depth, and depth must never exceed a maximum cap.

Engineering handshake with H2-3: retry bound creates the “loss shape,” buffer policy decides whether it becomes stutter or a tolerable PLC texture.

A. Loss injection is mandatory (PER sweep or controlled interferer):

• Run a defined loss sweep (e.g., PER ladder) and log p50/p95/p99 latency, buffer level, underrun flags, PLC state.
• Use the same audio content each time (speech + transient-rich music) to make PLC artifacts comparable.

B. What “good” looks like under stress:

• p99 bounded: spikes align to bounded retries, not runaway queueing.
• Buffer stable: level rises in stressed mode but stays under cap; in recovery it shrinks slowly after a stable window.
• PLC consistent: artifacts remain stable and non-spiky; abrupt crackles indicate underrun or discontinuity.

C. Minimum outputs to keep for design reviews:

• Latency histogram plots (p50/p95/p99) per stress level.
• Buffer level vs time with PLC state overlay.
• Event-triggered waveform captures for clicks/crackles (align with retry bursts and buffer drops).

Depth cue: do not tune by “listening only.” Every subjective artifact must map to one of: PLC mode switch, buffer underrun, or recovery policy hunting.

A neckband talkback path is dominated by real-world coupling: wind buffeting, collar rub, and structure-borne vibration. The multi-mic AFE must protect headroom (so DSP is not forced to “hide clipping”), keep EIN low for intelligibility, and expose clean hooks for NR/VAD and optional beamforming—without adding avoidable processing latency.

Boundary lock: local talkback chain only. Avoid conference AEC tutorials and platform/stack expansion.

• Speech intelligibility SNR: define target SNR at the AFE output in at least three scenes: quiet indoor, outdoor wind, and walking collar rub.
• Wind/rub robustness: prevent low-frequency bursts from saturating preamp/ADC; define a “no-clip under wind burst” criterion.
• VAD consistency: stable trigger behavior under wind/rub; avoid rapid on/off gating that creates word chopping.
• AFE noise floor (EIN-driven): maintain a low noise spectrum in quiet scenes so “hiss” does not dominate far-end perception.
• Processing latency guardrail: if NR/AGC uses look-ahead, quantify the fixed delay added to uplink and keep it bounded.

Neckband emphasis: wind/rub are typically low-frequency, high-amplitude and non-stationary. Analog headroom protection must come before DSP polish.

1) Mic placement shapes noise: neckband geometry creates asymmetric pickup and exposes microphones to wind and fabric contact. Multi-mic processing is only effective if each channel stays within linear range.

2) Bias + ESD + input protection: stable bias and a correct ESD return path prevent touch/ESD events from turning into long recovery or “scratch” bursts.

3) Preamp/PGA + ADC headroom: allocate gain so quiet speech is above noise floor while wind/rub bursts do not drive clipping. Use an analog HPF where needed to protect the ADC and avoid AGC pumping.

4) NR/AGC/VAD hooks: DSP should classify “wind/rub vs voice” and maintain VAD stability. If a module uses look-ahead, it adds fixed delay; quantify it.

5) Local sidetone (monitor) only: sidetone stability is governed by local loop latency and coupling. Keep the sidetone path short and bounded to avoid squeal.

Lane 1 — AFE noise and headroom evidence:

• Record AFE output spectra in quiet, wind, and collar-rub scenes; verify no persistent tonal artifacts and no low-frequency overload.
• Capture time-domain bursts to confirm “no-clip under wind burst” and to separate saturation from NR side-effects.

Lane 2 — Uplink stability evidence:

• Log uplink latency distribution (p50/p95/p99) and packet loss counters during the same acoustic scenes.
• Align “speech cut” windows with VAD gating flags vs RF loss vs NR/AGC state changes to avoid misdiagnosis.

Practical discriminator: if PER is normal but speech chops, suspect VAD/NR gating or headroom/AGC pumping; if PER spikes align to chops, suspect uplink scheduling or RF stress (H2-3).

On the neckband RX side, the output chain is a frequent failure point: pop/click during power or link events, limiter behavior that trades safety for distortion, and stability problems triggered by cable capacitance or load variation. The design must control mute sequencing, headroom under low battery, and the output return path so user-visible artifacts are bounded and repeatable in tests.

Boundary lock: neckband output chain only. Avoid full pro-IEM system architecture and compliance walkthroughs.

• Max output power and headroom: define output target into intended loads across nominal and low battery conditions.
• THD+N baseline: measure distortion/noise at reference levels in clean conditions to prevent “RF/codec blame” for analog issues.
• Pop/click threshold: set a measurable transient limit at the load during power-on/off, mode switches, and link resume.
• Stability margin: remain stable across cable capacitance and load variation; avoid oscillation that appears only with specific earphones.

1) DAC and reconstruction: the DAC output and its filter path set the baseline noise and can interact with mute timing.

2) HP amp stability under real loads: cable capacitance and varying transducer impedance can reduce phase margin. Stability measures (Zobel / series R / compensation) must be verified with representative cables and loads.

3) Limiter and hearing-safe caps: limiter/clipping strategy is a trade between safety and distortion. The behavior must be deterministic: bounded peak level, controlled release, and predictable recovery after link resume.

4) Pop/click is sequencing + discharge: pops occur when output nodes move abruptly due to rail ramps, charge-pump transitions, or DC offset steps. A robust design aligns mute control with rail readiness and includes a controlled discharge path.

Practical emphasis: “works on bench” is insufficient. Load and cable variation, plus power/link events, are the dominant field triggers.

1) Event-triggered transient capture:

• Capture output waveform during power on, brownout, link resume, and mode switches.
• Overlay mute control and rail status (or a GPIO marker) to prove sequencing, not guesswork.

2) Load and cable stress:

• Repeat transient tests with representative cables and transducers to expose capacitive-load instability.
• Use step response or burst tones to detect ringing/oscillation.

3) THD+N and noise sweeps:

• Measure THD+N vs output level at nominal and low battery conditions to isolate headroom/limit interactions.
• Confirm limiter attack/release determinism by logging limiter flags alongside audio measurements.

Practical discriminator: if artifacts appear only with certain earphones/cables, suspect stability/return path. If artifacts align with rail events, suspect mute/charge-pump/DC offset management.

Low buffer depth reduces tolerance to clock error. With two independent timebases (TX and RX), small ppm offsets and temperature drift can push the jitter buffer toward floor/ceiling, showing up as click/tick, dropouts, or pitch drift/wobble. A stable system defines a ppm budget, selects a master timebase policy, and implements drift compensation that keeps buffer level bounded and artifacts inaudible.

Boundary lock: device-local clock domains + compensation only. Avoid protocol-stack and OS scheduling tutorials.

• ppm budget: allocate allowable frequency error across TX reference, RX local oscillator, and temperature drift so the jitter buffer stays within bounds.
• Lock time: define PLL lock time on cold start and after RF recovery/channel switch; bound “unlock windows” that can trigger ticks.
• Allowed drift rate: limit how fast compensation must react (especially during thermal ramp) to avoid pitch wobble.
• Compensation smoothness: keep the drift loop slow enough to be inaudible while fast enough to prevent buffer floor/ceiling hits.

Practical rule: low latency demands controlling p99 buffer excursions, not only average ppm.

1) Two clock domains: TX packetization is timed by a TX reference, while RX DAC playback is timed by a local RX clock. Even small ppm mismatch accumulates quickly when buffer depth is low.

2) Master policy: choose whether TX pacing is the master (RX adapts playback) or RX playback is the master (TX adapts pacing). The decision affects recovery behavior after RF stress and how timestamps are interpreted.

3) Drift estimator: estimate frequency error from timestamp deltas and buffer level trend. Noise on the estimate drives wobble if the loop is too fast.

4) Compensation mechanism: apply ASRC or controlled playback-rate adjustment to keep buffer level centered without audible pitch modulation.

5) PLL lock and holdover: a clean holdover strategy prevents brief unlock windows from turning into buffer discontinuity during RF recovery.

Two synchronized evidence lanes:

• Timing lane: log PLL lock/unlock and holdover status, plus drift estimates (ppm, drift rate) over temperature and during RF recovery events.
• Audio lane: log buffer level, underrun/overrun flags, and a click/tick detector (or marked audible events) on the same timeline.

Correlation rules (fast discriminators):

• Click aligned to floor/ceiling hits → buffer discontinuity; check timestamp gaps and PLL unlock windows.
• Slow pitch drift with monotonic buffer trend → estimator bias or master policy mismatch.
• Wobble with frequent coefficient changes → loop gain too high or estimate noise; add smoothing and hysteresis.

Ensure event markers (RF recover, channel switch, mode transitions) are timestamped; timing bugs often masquerade as codec or analog issues when markers are missing.

The most common field failures in a neckband TX/RX are power-induced: RF TX bursts droop rails, volume steps pull HP-amp current, and codec/MCU events spike transient demand. If rail partitioning and sequencing are not engineered, the system shows link drops, resets, or pop/click. A robust design separates sensitive rails (AFE/PLL) from bursty rails (RF/HP), chooses LDO vs buck by noise and transient needs, and validates with event-aligned oscilloscope captures.

Boundary lock: internal power tree and events only. Avoid USB-PD protocol walkthrough and charger topology deep derivations.

• Transient droop limit: define max rail sag during RF TX burst, volume step, and link recovery events.
• PSRR and noise for sensitive rails: ensure AFE/PLL/refs remain quiet under RF and HP load transients.
• Battery life: treat average power as an outcome of rail choices, sleep modes, and burst handling—not as the primary knob.
• Charge thermal rise: set allowable temperature rise during charge + play; define current limits and priority behavior.
• Seamless power-path behavior: USB-in insert/removal must not inject pops or trigger brownout resets.

1) Li-ion power-path: a power-path block selects USB input or battery and can support “charge-while-play”. Priority must keep system rails stable even when charge current is limited or USB is inserted/removed.

2) Rail partitioning: isolate sensitive rails (AFE/PLL/refs) from bursty rails (RF TX and HP amp). Partitioning is not only about regulator choice; it also controls return currents and where transients flow.

3) Sequencing and event hooks: align mute and recovery behaviors with rail readiness. RF recover, volume steps, and codec/MCU wake-ups should be treated as explicit events with markers for debug.

4) LDO vs buck placement: choose LDO where noise/jitter sensitivity dominates; choose buck where burst current dominates; consider buck→LDO filtering when both matter.

Lane 1 — Rail integrity (oscilloscope):

• Probe key rails: RF rail, HP-amp rail, SoC core rail, and a clean AFE/PLL rail.
• Measure transient droop and noise injection during RF burst, volume steps, and link recovery.

Lane 2 — Event markers (debug GPIO + logs):

• Toggle GPIO markers for RF TX on, volume change, recover start/end, and wake events.
• Align droop/noise to user-visible failures: link drops, resets, and pop/click.

Fast discriminator: if only AFE/PLL rail noise rises (no droop), expect timing artifacts (H2-7). If RF/SoC rails droop during events, expect drops/resets and recovery pops (H2-6).

In a low-latency 2.4 GHz neckband link, failures in crowded spectrum are often driven by near-field detune (wearing posture), cable-as-antenna (charging cable and ground return), and power/ground injection (buck ripple and HP-amp current loops). With low buffer, brief PER/retry bursts quickly become audible drops or pops. Robustness comes from controlling coupling paths, and proving root cause with event-aligned spectrum/near-field scans plus link/audio counters.

Boundary lock: practical coexistence/EMC coupling-paths only. No certification walkthroughs.

• ESD robustness goal: maintain stable operation without persistent lockups, false drops, or loud pops under expected contact/air discharge environments.
• Crowded-band survival: bound acceptable PER/retry burst duration and define recovery time so low buffer does not underrun.
• Charge-while-use stability: insertion/removal of a charging cable must not meaningfully worsen PER, buffer stress rate, or audio noise floor.
• Noise containment: RF bursts and HP-amp current loops must not inject audible artifacts into AFE/codec reference or create ground-reference discontinuities.

1) Body near-field detune: wearing posture changes the antenna’s effective environment, shifting match and radiation patterns. Small shifts can create short PER bursts in crowded channels.

2) Charging cable as an antenna: a cable provides a strong common-mode path and can radiate/receive interference; it also changes the system return path and can inject noise into sensitive domains.

3) Power & ground injection: DC-DC ripple and bursty load currents (RF TX and HP amp) can create ground bounce and rail noise that maps to timing (click/drift) or link instability (retries).

4) ESD/EFT upset chain: transient coupling can trip reset/interrupt logic, causing reconnect storms and recovery pops if sequencing/mute policy is not aligned.

Lane 1 — RF & EMI observation:

• Near-field probe scans around antenna zone and charging cable entry while repeating posture A/B/C and plug/unplug.
• Spectrum snapshots before/after cable attach; note energy changes near switching tones or strong interferers.

Lane 2 — System & audio evidence:

• Log PER, retry rate, channel switches, and buffer stressed entries; time-align with posture and charging events.
• Capture audio noise spectrum and pop/click timestamps; correlate to rail noise markers and retry bursts.

Fast discriminator: posture/cable-driven PER bursts indicate RF/EMC coupling; fixed-tone noise points to ripple; volume-step-correlated drops point to HP rail/return loops.

Neckband constraints are dominated by mechanical-electrical coupling: mic ports and mesh change acoustic response and wind noise, sweat and corrosion create leakage/noise, cable pulling adds intermittent contact stress, and simultaneous charge + RF + HP load drives thermal rise and derating. Robust design focuses on measurable hotspots and evidence that separates false touch, brownout, and packet loss.

Boundary lock: hardware-verifiable items only. No app UX essays.

• Surface temperature rise: define maximum allowable surface ΔT during charge + RF activity + high output playback, plus derating thresholds.
• False trigger rate: bound unintended button/touch events under motion, collar rub, and sweat conditions using a measurable event-rate spec.
• Ingress-related design points: specify mic port mesh and seals to reduce clogging risk while keeping acoustic response stable (design points only, no certification flow).
• Mechanical stress tolerance: cable pull and neckband flex must not create intermittent contacts that mimic drops or brownouts.

1) Mic port & mesh: mesh and water-resistant layers change the acoustic path; clogging or wetting shifts response and can destabilize VAD/NR behavior.

2) Sweat & corrosion: moisture can create leakage paths and contact instability, raising noise or triggering intermittent faults that look like RF drops.

3) Buttons/touch under motion: collar rub and flex can create false triggers if debounce and sensor placement do not match wearable dynamics.

4) Thermal stacking: charging + RF activity + HP output can create localized hotspots; derating must be predictable and evidenced.

Thermal evidence:

• Log multi-point temperature (pod surfaces and internal sensor) during charge + RF activity + sustained high output.
• Correlate derating events to output power reduction, RF retry bursts, or limiter engagement.

False-trigger evidence:

• Record touch/key event rate under motion, collar rub, and sweat simulation; verify debounce rejects vs accepted events.

Reproduction checklist:

• Posture state, clothing contact, cable attach, ambient temperature/humidity, and activity pattern must be logged to make field issues reproducible.

Fast discriminator: input-event bursts without rail/PER anomalies point to false triggers; resets point to brownout; PER/retry bursts point to coexistence/coupling (H2-9).