WanderBridge Mini — Decision Log

Running log of every keep/swap/drop decision through Stages 0–10. Each entry cites source (Perplexity call, datasheet, distributor page, design research) and dates the call.

---

Stage 0 — Feature-parity sweep (2026-04-24)

Decision: Bridge Mini positioning = "wired-first single-location endpoint with optional low-RF backhaul + multi-backhaul configurability." 13 features nobody in the competitive set ships.
Source: training-knowledge competitive analysis + WANDERROUTER-DESIGN-RESEARCH.md reference deployment #1 (Michael's apartment).
Flag carried forward: full Perplexity comparative sweep deferred to next session (rounds reset). Items marked 📝 in FEATURE-MATRIX must be Perplexity-verified before marketing materials quote them.

Stage 1 — Intake + freeze (2026-04-24)

Decision: Native design (no Blueprint input). Device ID WV-NET-BRIDGE-MINI, revision v0.1-brief.
Chassis: new Mini chassis (palm-sized, ~90 × 60 × 25 mm, aluminum base + polymer cover, VESA-75 optional, rear service panel).
Ship target: Q1 2027.
Validation reference: Michael's Silver Spring 1BR apartment (design-research doc).
Target price span: $49 bare-minimum → $249 Lux.

Stage 2 — Component selection (2026-04-24)

40 components selected — 21 platform (in every unit), 19 module (configurable or upgrade). Full BOM at BOM-normalized.csv.

Key architectural decisions:

SoC: MediaTek MT7981B (Filogic 820)

Why: Mainline OpenWrt support (used in Flint 3), low cost ($8 vol-1k), internal 7-port switch (eliminates external switch chip), 1× PCIe 2.0 x1 for WiFi module, 2× GMAC for PHYs.
Considered + rejected: MT7988A (overkill for Mini, save for Medium/Pro), IPQ5300 (closed-source drivers on OpenWrt), RTL8198D (weaker WiFi 7 path).
Stage-4 Perplexity verification: distributor stock, lead-time, 2026 lifecycle confirmation.

Mini chassis RAM/storage constraint (Standard 2 deviation)

RAM and eMMC are soldered, not socketed. Rationale: Mini chassis size (~90×60 mm) cannot accommodate SODIMM slot + M.2 NVMe.
Upgrade path: whole-mainboard swap via send-back service (Ambassador-assisted, 5 business days).
Documented limitation: acknowledged in SERVICE-FLOW.md and in buyer-facing copy. Not a silent downgrade from WanderVerse philosophy — Mini trades per-component upgradability for 🔒 whole-mainboard swap path.
Consistent with philosophy: because all other modules (WiFi, MoCA, Powerline, cover, 2nd port, PoE class, power fallback) ARE user-replaceable or send-back-swappable. The RAM/eMMC constraint is chassis-size-driven, not chosen.

Internal switch vs external switch chip

Decision: use MT7981B's built-in 7-port switch. Saves $3-6/unit + PCB real estate. 1-port and 2-port daughterboard configurations both handled by internal switch — port 2 is a daughterboard that plugs into the SoC's 2nd GMAC.
Rejected: RTL8367S/RTL8372B external switch chip.

PoE-PD: TPS2373 (802.3at) base, TPS23881 (802.3bt) upgrade

Same mainboard footprint — user/Ambassador can swap the chip during send-back upgrade service without board re-design.
Compliance: IEEE 802.3at / 802.3bt are passthrough certs (chip vendor certifies, we inherit).

Multi-backhaul = daughterboards (not on main PCB)

WiFi 7 (MT7925 M.2 E-key) — user-installable in chassis service panel
MoCA 2.5 (MaxLinear MxL3710) — daughterboard via 30-pin header
Powerline AV2 (MediaTek MT7510P) — daughterboard via 30-pin header + AC isolation
Rationale: cost discipline — buyers who don't want these don't pay for them. Separable PCBs also keep EMC isolation cleaner (especially Powerline's AC-coupled noise).

Enclosure — 3 cover tiers, 1 base

Base: 6061-T6 CNC aluminum (prototype) → A380 die-cast (production) with CNC post-processing for cover-rim finish
Standard cover: PC/ABS injection-molded. Open STL published — users can print replacement covers in any color / material / custom design (Standard 2 + repair-culture commitment).
Pro cover: 6061-T6 CNC anodized in 3 colors.
Lux cover: Hand-finished walnut veneer + brass inlay OR hand-painted. Per-unit Ambassador-signed glyph plate (+$25).

Ambassador signature — two tiers

Standard: thermal-printed card inside cover with Ambassador name + QR to their WanderVerse page.
Lux: etched-aluminum glyph plate, hand-designed per Ambassador, permanent inside the cover (visible every time the user opens the device for service).
Both preserve "Ambassador is part of the device's ongoing story" principle.

---

Flags carried into Stage 3:

#	Flag	Severity	Stage-3 action
1	MT7981B PCIe 2.0 vs MT7925 PCIe 2.0 compatibility	🟢	datasheet cross-check (standard PCIe handshake)
2	MoCA + WiFi antenna coexistence — possible 2.4 GHz interference	🟡	Stage 6 EMC/co-existence test plan
3	Powerline AC line + SoC ground isolation — Y-cap + ferrite placement	🟡	Stage 6 EMC design pre-check
4	2nd-port daughterboard 30-pin header pinout definition	🟢	Stage 5 DFA — settle pinout early so daughterboards can standardize
5	MT7925 FCC modular cert status (April 2026)	🟡	Stage 7 compliance + Stage 4 Perplexity verification
6	PoE-bt inrush current vs chassis thermal under 60W ceiling	🟡	Stage 6 thermal model
7	Die-cast vs CNC-from-billet vs sheet-metal — chassis volume production path	🟡	Stage 5/6 tooling + supplier conversation
8	2.5G RJ45 shielded magnetics + chassis ground strategy	🟢	standard EMC practice; cover in Stage 6
9	Production origin — US contract-mfg (e.g. Arrow SMT NC) vs Taiwan (Foxconn, Qisda)	🔴	Stage 7 compliance + business-decision, gates before EVT
10	Anodization spec for Pro covers (MIL-A-8625 Type II or III?)	🟢	Stage 5/6 mechanical sign-off
11	Powerline + UL certification (mains-connected product)	🟡	Stage 7 compliance — UL 62368-1 pre-cert lab quote
12	Open STL file coverage — which parts get published	🟢	Stage 8 — confirm: Standard cover (yes), Pro cover (yes, with copyright), Lux (no — by design, Artisan-made only)

Stage 2b — Configurator axes (pending)

_(7 axes identified in MANIFEST; full matrix to follow)_

Stage 3 — Compatibility review (2026-04-24)

40-part BOM verified electrically + mechanically coherent. Full output: COMPATIBILITY.md.

Resolved in this stage:

MT7981B ↔ MT7925 PCIe compatibility (Gen 2.0 x1 both ends — clean)
RGMII/SGMII PHY interfacing via RTL8261N
Power budget: PoE-at (25W) adequate for typical configs; PoE-bt (60W) recommended as default for full-config + headroom
Chassis mechanical fit: all modules fit base Mini chassis EXCEPT Powerline (needs +10mm variant)
Antenna / RF spatial coexistence: clean separation between WiFi / MoCA / Powerline paths
Thermal: fanless design adequate; aluminum base as heatsink; full-config ~14W well within passive envelope

Deferred / open:

🔴 30-pin daughterboard header pinout — must be spec'd at Stage 5 DFA before any daughterboard fab
🟡 MAC allocation for simultaneous MoCA+Powerline — may require single-backhaul-module-at-a-time constraint; Stage 5 DFA
🟡 TPS2373/TPS23881 footprint standardization — use 23881 footprint with 0Ω jumper variant for 2373; Stage 5 DFM
🟡 Powerline chassis variant — either extend base or offer Mini-Powerline as separate chassis SKU; Stage 5 DFM decision
🟡 Powerline AC isolation EMC pre-check — Y-caps + ferrite + transformer layout; Stage 6
🔴 Production origin (US vs Taiwan CM) — business decision, gates EVT; Stage 7

Stage-4 Perplexity queries queued for tomorrow: 1. MT7981B + MT7925 distributor stock/lead-time/lifecycle 2. MxL3710 + MT7510P supply status 3. MT7925 FCC modular cert confirmation 4. UL 62368-1 pre-cert lab quotes for Powerline 5. US vs Taiwan CM supplier comparison at 1k-10k unit scale

Stage 4 — Sourcing (pending, Perplexity rounds reset)

Stage 5 — DFM/DFA/DFT (2026-04-24)

Full output: DFM-DFA-DFT.md.

Key decisions:

Mainboard: 4-layer ENIG, 70×50mm. Prototype via JLCPCB; mid-run via Sierra or Advanced Circuits; supplier pick Stage 4.
Assembly: 14-step Ambassador-station workflow, 7 min (Standard) / 12 min (full-config + Lux) per unit.
DFT: 13 test points, 10-test automated fixture, 90s/unit, ICT + flying-probe at SMT line.
First-run target: 250 units, ~$45 FOB, $79 retail typical = 44% gross margin.

Stage 6 — Thermal/EMC/ESD (2026-04-24)

Full output: THERMAL-EMC.md.

Key decisions:

Fanless design; aluminum base doubles as heatsink. 15W sustained comfortable, 20W acceptable, >25W not feasible.
All configs ≤2-port + WiFi + MoCA fit comfortable envelope (≤13W typical).
All-modules config (+ Powerline) pushes thermal limit (13.5W typical, 20W worst-case) — advisory in configurator.
EMC stacks defined for Ethernet (magjack ground), PoE (SMAJ58A), WiFi (RF keepout, MT7925 passthrough cert), MoCA (chip passthrough), Powerline (full Y-cap + choke + isolation transformer stack).
ESD: Level 4 (±8 kV contact / ±15 kV air) on all external I/O.

Decisions locked 2026-04-24 (Michael deferred, Claude decided):

1. Powerline → deferred to v2. Bridge Mini v1 backhaul menu: Ethernet / MoCA 2.5 / WiFi 7 client radio. Removes MT7510P, MT7510P transformer, IEC C7 inlet, UL 62368-1 cert cost from v1 BOM/timeline. Saves $18-35k + 3-6 months. Still 3 configurable backhaul paths.

2. Chassis → single variant. No Mini-Powerline chassis extension needed. Single STL family + single tooling path.

BOM-normalized.csv trimming (Powerline-related rows removed from v1 scope — rows 27, 28, 29 now marked deferred-to-v2):

Row 27 MediaTek MT7510P — deferred
Row 28 Würth 750316772 transformer — deferred
Row 29 Schurter 4300.0621 IEC inlet — deferred

Rows remain in BOM-normalized.csv for posterity + v2 planning, flagged in stage2_notes.

Updated Bridge Mini v1 BOM: 37 active parts (40 total, 3 Powerline deferred).

Why this way.