20261020-1000-air-hub-commercial-enclosure Dual power input for Air Hub: USB-C PD with priority, M8 4-pin harness, IP54 without sealed connectors

status: approved
createdAt: 2026-10-20 10:30
approvedAt: 2026-10-20 (owner: "可以，按这个思路来做")
relatedTask: 20261020-1000-air-hub-commercial-enclosure

Context

Owner decisions of 2026-10-20:

commercial-grade product;
IP54 and the conformal coating stay;
no sealed M12 / DT connectors;
connectors may sit on the housing side;
USB-C takes power through a PD trigger and also serves flashing;
the harness connector is M8 4-pin (V+, GND, CAN-H, CAN-L) at 9–36 V;
both inputs are fitted. USB-C has priority: while a valid USB-C source is plugged in, the M8 V+ does not supply the board.

Today one 4-wire harness carries power and CAN on a 5.08 mm screw terminal, through a 2 A fuse, a 5.0SMDJ33CA TVS, an SQD50P06-15L reverse P-FET and an ACM7060 common-mode choke. The 5 V rail is a K7805-2000R3L module (8–36 V in) on the Linux carrier and an LMR36520 buck on the Mini. The RK3506 core board needs 4.75–5.25 V. The Linux carrier USB-C is the OTG0 service port (VBUSDET 68k / 100k).

Proposal

1. Supply rules

Board	USB-C valid range	M8 range	5 V-only USB
Air Hub / Air Hub Pro	9–12 V (PD request 12 V)	9–36 V	treated as no USB; M8 powers the board
Air Hub Mini	5–12 V (PD request 12 V)	9–36 V	powers the board

2. USB-C path (both boards)

PD sink: HUSB238 (_002DD variant), VSET for 12 V, ISET 1.25 A. It replaces the 5.1k CC pull-downs. It requests the highest PDO up to 12 V, so a 9 V-only charger gives 9 V.
Protection: SMF15A TVS on VBUS. The USBLC6 VBUS pin moves to +5V, because its 6 V breakdown is below 12 V.
Switch: an LM74800-Q1 ideal-diode controller with two DMN6040SSDQ N-FETs back to back (common drain).
- It blocks reverse current from the board rail into VBUS.
- Its EN/UVLO divider sets the USB on-threshold: 6.5 V on the carrier, 3.8 V on the Mini.
VBUSDET (carrier): the 68k / 100k divider stays, with a BAT54WS clamp to +3V3 for 12 V VBUS.

3. M8 path (both boards)

Connector: a panel-mount M8 4-pin socket with flying leads to a JST XH 4-pin board header. It replaces the screw terminal.
Protection and switch: the 2 A fuse and 5.0SMDJ33CA TVS stay. An LM74502 with two DMN6040SSDQ N-FETs (common source) replaces the P-FET reverse stage.
Priority: the LM74502 OV pin is driven from a divider on USB VBUS. It turns M8 off once VBUS passes 7.9 V on the carrier or 4.4 V on the Mini.
Handover: the USB on-threshold is below the M8 off-threshold, so on both edges the two paths overlap briefly and the ideal diode stops back-feed. There is no gap.
After the switch: the ACM7060 choke and the π filter feed VIN as today.

4. Vanxoak review fixes (Linux carrier)

PHY LED0 drives an LED to GND (anode on ETH_LED0), so the ANEG_Dis strap reads 0.
4.7k pull-downs on the RX_D0, RX_D1 and CRS_DV straps.
Y4, C47 and C48 next to the PHY XI/XO pins.
USB D+ on one layer, without vias.

5. Layout

Both boards are rerouted. USB-C and the M8 header sit on a side edge so the cables exit through the housing side.

6. IP54

Each connector face seals against the housing wall through a gasket.
The M8 socket is panel-mounted and sealed by its own O-ring.
The test (V-08) runs with all cables plugged. Unplugged ports get a rubber cap.
Conformal coating stays as in the family rule.

7. Documents

Spec P05-air-hub.md: §2.4, GW-05, the environment row and AIR-O-04 say "USB-C first; Mini 5–12 V, Pro 9–12 V (PD); M8 9–36 V".

Design appendix and reference-board page: power rows, diagrams and the §6 cost.

Evaluations: platform-combinations.md §8.1, industrial-linux-and-radio.md, emc-safety-certification.md.

Decision records:

the 9–36 V record stays, because M8 is still 9–36 V;
a new family draft decides "YBO bus connector = M8 4-pin".

Handover: refresh the package.

Verification

ERC clean and verify_net 0 mismatches on all three boards.
DRC: 0 unconnected and 0 clearance.
SPICE: VBUSDET at 5 V and 12 V.
Cost CSVs regenerated.
Docs gate: bun run docs:lint && bun run docs:build && git diff --check.