Tuya RS485 Sensor PCB


I can build this as a separate proof-of-concept project: 12 VAC input, protected rectifier/power supply, Tuya Wi‑Fi module, isolated/protected RS‑485 interface, one power LED, and one pairing/reset button on a simple long narrow PCB.
The PDFs could not be text-extracted in Chat mode, so I need Agent mode to inspect the protocol/registers, create the new project, select parts, build and verify the schematic, then make the basic PCB. Click Continue in Agent Mode and I’ll proceed; I’ll only stop for clarification if the sensor’s wire pinout or required measurement variant cannot be established from the attachments.
Continue in Agent Mode
I’ll extract the sensor wiring/protocol, create a separate proof-of-concept project, build and verify the schematic, then make a simple long narrow PCB layout.
Created a thread, a project
Tuya Inline Chemical Sensor POC

Tuya Inline Chemical Sensor POC

Proof-of-concept long narrow Tuya Wi-Fi interface board for a 4-wire chemical sensor using MODBUS-RTU over RS-485. Powered from external Class 2 12 VAC, with...

a month ago

·

Build chemical sensor POC schematic
Starting...
View Thread
The sensor interface is confirmed: 9–24 VDC probe power, yellow/green RS‑485, and MODBUS‑RTU at 9600‑8‑N‑1. The new project is created, and I’m finishing the power and communications schematic verification now.
The verified 39-component schematic is complete in Tuya Inline Chemical Sensor POC. It uses 12 VAC input, rectified probe power, a 3.3 V buck, Tuya CBU, protected RS‑485, one power LED, and one pairing/reset button.
For the first PCB, I propose a simple 2-layer rectangular board around 100 × 25 mm, components on both sides, cable connections at opposite ends, and the CBU antenna at the outer edge. I’ll also correct the CBU footprint’s undocumented center pad before placement—confirm this layout plan and I’ll proceed.

Bench PCB Routing Gate
Resumed surgical pass — 2026-09-03
STOPPED at the authorized hard limit. The U4 VDD connection was completed with a 0.12 mm local trace, but U4:SDA ↔ R6:P2 returned blocked-no-path in both directions at 0.12 mm. Per the instruction to stop when any connection still fails, the reset and backlight interventions were not resumed.
Exact SDA blocker
  • U4:SDA pad center: (5.65, -38.5) mm; its left escape faces C12:P2 (GND).
  • C12:P2 center: (4.28, -38.5) mm. The copper-edge gap between C12:P2 and the U4:SDA pad is approximately 0.27 mm.
  • A 0.12 mm trace with the existing 3.8 mil (0.09652 mm) keepout needs approximately 0.31304 mm, exceeding the available corridor by approximately 0.04304 mm.
  • The adjacent U4 VDD and SCL pad rows leave only approximately 0.20 mm vertical copper-edge gaps, also too small for that trace plus unchanged clearance.
  • A through-via escape below the pad is blocked by the existing 5 V trunk centered at y = -40 mm with 2.0329 mm width; its upper copper edge is approximately y = -38.98355 mm.
The minimum physical correction is move C12 left by at least 0.05 mm or authorize an U4-SDA-only clearance of no more than 0.075 mm. Both are outside the resumed authorization: only R19 could move, and no clearance relaxation was authorized.
Fresh final counts after stop

Table


CheckCount
Airwires4
Dangling traces0
Floating copper0
Foreign-net overlaps/shorts0
Missing footprints0
Invalid layers0
ERC findings0
The four live airwires are: GND stitching via ↔ C12:P2, DISP_RESET_N Trace 8c15 ↔ R18:P1, I2C_SDA R6:P2 ↔ U4:SDA, and DISP_BL_SW R19:P1 ↔ U5:OUT. The prior 3V3 Trace a09c ↔ U4:VDD airwire is resolved.
Result
FAILED — manufacturing gate not passed. Four airwires remain after the allowed two targeted routing attempts per blocked connection. No files were exported.
Final live checks
Final routing-state read and DRC were run on 2026-09-03 after all mutations.

Table


CheckFinal countStatus
Airwires4Blocking
Dangling traces0Pass
Floating copper0Pass
Foreign-net overlapping copper / shorts0Pass
Missing footprints0Pass
Invalid layers0Pass
Stackup validation errors0Pass
Overlapping traces/fills, protected intrusions, under-connected vias, pad-access/keepout errors0Pass
ERC findings0Pass; the ERC query returned no findings
Edge-placement warnings2Non-blocking: J1 and U3
Routing preflight reports Board valid: yes, 57 PCB components, 39 nets, and no routing-feasibility anomalies. The board outline remains the 50 × 100 mm rounded rectangle. The stackup remains the valid 1.30 mm custom four-layer stackup: Top signal, Mid-Layer 1 ground plane, Mid-Layer 2 power plane, Bottom signal.
Non-blocking edge warnings
  • J1: through-hole 12 VAC engineering-sample connector at (-21.5, 24.2) mm is inset from the left edge rather than satisfying the library edge-placement expectation.
  • U3: Tuya CBU antenna section intentionally projects beyond the top edge; the warning is consistent with the antenna edge-placement intent.
Remaining blocking airwires

Table


NetExact endpointsAttemptsObserved blockerMinimum next intervention
3V3Trace a09c ↔ U4:VDD2Deterministic router returned blocked-no-path in the 0.5 mm-pitch U4 left-side escape, where VDD/SDA/SCL and existing top copper tightly constrain pad access.A local U4-only 0.12 mm escape/neckdown rule would be the minimum automated-routing rule change, but it conflicts with the instruction that 0.10–0.12 mm neckdowns are allowed only at J4.
I2C_SDAR6:P2 ↔ U4:SDA2Same 0.5 mm-pitch U4 escape congestion; both routing directions returned blocked-no-path.Same single U4-only 0.12 mm escape/neckdown exception; otherwise a manual via-assisted route is required without changing global clearances.
DISP_RESET_NJ4 pin-3 neckdown Trace 8c15 ↔ R18:P12The J4 pin-3 stub is boxed by the already-routed adjacent J4 escape lanes and local ground stitching; both routing directions returned blocked-no-path.No global rule relaxation is justified. Use one manual via-assisted J4-local escape while retaining the existing 0.10 mm neckdown, or approve a surgical reroute of one neighboring J4 escape.
DISP_BL_SWR19:P1 ↔ U5:OUT2R19 is below the wide PRESS_RAW corridor and inside the now-dense J4/R19 escape area; both routing directions returned blocked-no-path.No global rule relaxation is justified. Use a manual via-assisted escape below R19; if automated routing is mandatory, a small R19-only placement adjustment inside the approved LCD zone is the minimum placement change to evaluate.
The final board check still says the board is routing-feasible, so the four blockers are deterministic targeted-router/path-access failures rather than stackup or board-definition failures.
Mutation summary
  • Removed 6 traces called out by fresh overlapping-copper DRC; this cleared all 7 initial foreign-net overlap/short findings. The deleted nodes were on I2C_SDA and 3V3.
  • Removed 4 subsequently exposed 3V3 floating/dangling trace segments.
  • Regenerated 12 deterministic neckdowns after cleanup.
  • Successfully routed 12 targeted connections, including LCD SCK, DC, LEDA, MOSI, most reset routing, J4 3V3 and local decoupling, restored 3V3 branches, and LED 3V3.
  • Preserved component placement, board outline, stackup, and unrelated routed copper except for the DRC-identified conflicting I2C_SDA/3V3 traces that had to be removed; their unresolved endpoints are listed above.
  • Resumed surgical pass — 2026-09-03

  • Exact SDA blocker

  • Fresh final counts after stop

  • Result

  • Final live checks

  • Non-blocking edge warnings

  • Remaining blocking airwires

  • Mutation summary