The HY2120 Kelvin-sensing work completed, but I could not retrieve and reconcile its final report before execution ended.
Confirmed work from this session:
Added dedicated Kelvin branches from:
U6 VSS → RS1 BATT_NEG-side land
U6 CS/R38 → Q2 source land
Removed the previous remote sense branches.
Preserved component placement, USB routing, MOTOR_B Kelvin routing, high-current copper, and the 0.10 mm spacing rule.
Latest verified board state:
0 airwires
0 blocking DRC errors
17 intentional redundant-copper groups retained
The last reviewed qualification still had two unresolved release gates:
Proving isolation of the Q2 Kelvin pickup via from Bottom/Mid-Layer-2 load-current copper.
Conservatively bounding the internal RS1–Q1–Q2 copper resistance and confirming positive margin below the HY2120’s 170 mV minimum overcurrent threshold.
The final executor was specifically working on those two gates, but its result was not available for reconciliation. Therefore, the board is not yet approved for manufacturing, and no new manufacturing exports were generated.
Remaining work:
Retrieve and review the final Kelvin-sensing report.
Confirm the 10.565 A / 500 ms hot overcurrent margin.
Run an independent final ERC/DRC and protection-path review.
Verify JLCPCB BOM stock and via/stencil requirements.
Update firmware and bring-up documentation.
Regenerate Gerbers, BOM, CPL, STEP, PDF, and IPC-D-356 files.
NOT QUALIFIED FOR MANUFACTURING. Pad-datum sense routes were authored and the live board has 0 airwires and 0 reported blocking DRC errors, but the strict Kelvin/no-load-current and conservative positive-hot-margin criteria are not fully demonstrated. Do not interpret the component-only 24.437 mV headroom as a qualified margin.
Assigned-scope status
Table
Work item
Status
Evidence / blocker
1. Inspect schematic, pads, and old branches
Completed
Live schematic/net membership and physical lands inspected.
2. Remove old branches and author explicit pad-datum routes
Completed for CAD authoring
13 old trace segments and one obsolete R38 escape via removed; 11 final trace segments and three vias added. Q2 pickup is a via centered on a source land. Physical manufacturing qualification remains open under item 3.
3. Prove branches carry no load current and measured boundary contains only the specified devices
Failed / blocked
Source pickup is a through-hole via above the existing Bottom/Mid2 return polygon. Its non-sense-layer barrel isolation is not established by the available readback. Also the RS1-to-Q1 bus and Q1-to-Q2 drain bridge necessarily remain inside the sensed boundary.
4. Conservative sensed voltage and positive hot margin
Failed / blocked
Component contribution and allowable residual budget calculated; no defensible upper bound covering the internal series copper, pad/contact spreading, and pickup error was established.
5. Current live board checks and redundant-copper classification
Completed
Fresh checks at 2026-09-19T11:42:31.863Z: no reported blocking errors; routing-state read explicitly confirms 0 airwires and 22 redundant nodes in 17 groups.
6. Preserve/verify USB and MOTOR_B Kelvin routing
Completed
Both USB connector-side nets and MOTOR_B inspected live; no mutation targeted these nets.
Exports
Unattempted, as requested
No export or manufacturing order performed.
Exact retained CAD changes
All coordinates below are board millimeters, X right / Y up. All new traces are 0.15 mm wide. No schematic net, component position, connector, board outline, stackup, global spacing rule, broad high-current polygon, or intentional parallel power via was changed.
VSS branch — BATT_NEG
RS1 pin 1 is the BATT_NEG land centered at (15.27, 20.43). After RS1's 90-degree rotation its land spans approximately X 13.62–16.92, Y 19.43–21.43.
U6 VSS is at (2.65, 21.05). The existing three short Top segments to the local via (3.6000, 20.8594) were retained.
Removed the three old Mid1 segments from that local via region to J2: (3.936154,20.859449) → (11.646,28.569295) → (11.646,32.9845) → (11.646,34).
New Top/Mid1 via at (13.8,22.3); new Top trace to (13.8,20.43).
The Top approach first enters RS1's actual land at its upper edge near (13.8,21.43). It is left of the reinforced BATT_NEG landing, whose left edge is X=14.0. Thus it does not pick up the upstream bus before reaching RS1's land.
The Mid1 detour avoids the existing BMS shunt-sense path. The old connection to J2 is gone; the new VSS branch is not a second end-to-end battery-return route.
CS branch — SYS_GND through R38
R38 P1 is the SYS_GND-side land at (3.99,19.00); P2 and the existing AUTOPROT_VM route to U6 CS were retained.
Removed the old R38 branch through (4.045,18.95), (4.045,17.6), (3.97,17.525), (2.995,17.525), (2.77,17.75), (2.77,18.425), (2.77,18.725), (3.07,19.025), including its now-obsolete via at (2.77,18.425) and the old (4.614463,19.624463) → (3.99,19) stub.
New Top route: (3.99,19) → (3.6,19.39) → (3.6,19.65), then a Top/Mid1 via.
New Mid1 route: (3.6,19.65) → (4.2,19.65) → (8.095,15.755) → (8.095,15.645).
The terminal via at (8.095,15.645) is centered on Q2's left source land, not on the remote receiving ground landing.
Ruleset HY2120 Q2 source land Kelvin via sets the source pickup to a 0.45 mm annulus / 0.20 mm drill. Baked size and drill readback confirmed these dimensions. Its Top/Mid1 connected-layer selection is retained, and Force Annular Layers = Connected only was subsequently applied to prevent a forced Bottom annulus.
Important unresolved detail: this is still a through-hole barrel, not a blind via. The existing return polygon is on Bottom and Mid2 and occupies this XY region. Removing unused annular rings alone is not proof of a fabrication clearance around the plated barrel. Layer-resolved filled-copper/barrel isolation must be verified before calling this branch load-current-free. Via-in-pad fill/cap and assembly acceptability must also be resolved; no such process was ordered or claimed.
An initial trial path conflicted with existing BATT_TEMP / HY2120 support copper. Those trial nodes were removed; the final live checks contain none of those six temporary overlap errors.
Hot-voltage calculation and model correction
Using the exact requested assumptions:
Rdevices=6.097875+2(3.84)=13.777875mΩ
Vdevices=10.565×13.777875=145.563249375mV
Mideal=170−145.563249375=24.436750625mV
This is not the conservative total. The actual comparison must include all remaining measured series copper and pickup errors:
Here resistance is expressed in milliohms and offsets in millivolts. With zero additional offset/noise reserve, the maximum permissible extra hot resistance is:
Rextra,hot<2.312991067mΩ
Using the earlier report's 1.5109 copper temperature multiplier, that is less than 1.530869725 mΩ at 20°C, before any additional reserve.
Why a pad-only allowance is not sufficient
The earlier independent report's 1.686 mΩ table explicitly sums only the external BATT_NEG pickup-to-RS1 and Q2-to-SYS_GND pickup paths. It does not separately bound the following unavoidable internal paths:
RS1 to Q1 broad source bus, Top polygon with vertices (9.05,28.1), (12.55,28.1), (12.55,26), (16.92,26), (16.92,30.6), (9.05,30.6).
Q1-to-Q2 common-drain bridge, Top rectangle 4.3 × 3.2 mm, centered at (10,22.1).
Finite source/RS1 land and solder-contact spreading, source-pin current distribution, and sense-branch bias/offset contributions.
Kelvin pickup removes the upstream connector return and downstream system return from the intended measurement, but cannot remove those internal load-current conductors without changing the circuit or performing a separate compensated measurement. Treating all of them as zero and choosing an arbitrary small pad allowance would not be a conservative qualification. No validated aggregate bound below the calculated budget was established here. The 500 ms pulse is evaluated using the requested hot envelopes; it is not a justification for ignoring these terms.
Live verification and retained findings
Fresh full checks: 2026-09-19T11:42:31.863Z. The requested airwire, overlapping copper/short, overlapping trace/fill, floating/dangling copper, invalid-layer, missing-footprint, pad-access/keepout, protected-intrusion, under-connected-via, layout-cell-tree, and stackup checks reported 0 blocking errors. Separate routing-state inspection explicitly returned 0 airwires.
Retained findings:
17 redundant groups / 22 nodes: four BATT_NEG reinforcement groups, twelve SYS_GND reinforcement/parallel-route groups, and one protected USB_DP_CONN group. These are existing intentional current-sharing or protected connector structures, not a blanket cleanup list. No new sense branch was listed as redundant.
Two informational footprint findings on Q1/Q2's AON6512 footprint instances. The tool labels them Informational, not blocking errors, and supplies only a generic footprint-issue description. They were present before this change; no footprint replacement was attempted.
Preflight also reported same-net J4 pad overlaps, a same-net drain-pad overlap, and duplicate power-via warnings. These were not remedied by deleting intentional copper or moving parts.
The exact 40 × 75 mm board, Standard 4 Layer / 35 µm copper stackup, 0.10 mm minimum spacing, antenna constraints, component positions, broad reinforced load-current paths, and connector placement were retained.
Protected-routing verification
USB_DP_CONN: protected 0.25 mm Top connector branches remain present; the existing 0.2425 mm downstream route remains present.
USB_DN_CONN: protected 0.25 mm Top connector branches remain present, with the existing downstream Mid2 route and vias at (-1.5,-27.4) and (5.65,-29.6757).
MOTOR_B: current live Kelvin route remains 0.254 mm on Mid2, from (-7.75,15.95) → (-8.8,17) → (-12.8,17) → (-14,18.2) → (-14,22.26) with its preserved layer transitions. Existing wide motor-current copper remains untouched. These are live coordinates, not older placement-note coordinates.
Remaining release gates
Prove the Q2 pickup barrel is isolated from Bottom/Mid2 load-current return copper, or approve a local alternative pickup construction. Validate the via-in-pad fabrication/assembly process.
Establish a conservative resistance/error bound for the internal source bus, common-drain bridge, pad/contact spreading and offsets, then demonstrate positive margin to 170 mV at 10.565 A using the specified hot device envelopes.
Retain the present DRC-clean routing while resolving those gates. Do not release solely because airwires and blocking DRC errors are zero.