feat: add DRC check for vias too close to pads by lucaferri-dev · Pull Request #113 · tscircuit/checks

lucaferri-dev · 2026-03-08T15:09:01Z

Summary

Adds a new DRC check that detects vias placed too close to SMT pads
Complements existing pad-to-pad spacing checks
Uses configurable minimum clearance (default 0.2mm)

Closes #44

Test plan

Unit tests for via-to-pad distance calculation
Tests for edge cases (overlapping, just-touching, far apart)
All existing tests still pass

Implements issue tscircuit#44 — a new DRC check that validates via placement relative to pads (both SMT pads and plated holes). Uses rect-to-circle and circle-to-circle distance calculations for accurate gap measurement. Closes tscircuit#44 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

graphite-app · 2026-03-08T15:11:55Z

tests/lib/check-via-to-pad-spacing.test.ts

+import { expect, test, describe } from "bun:test"
+import { checkViaToPadSpacing } from "lib/check-via-to-pad-spacing"
+import type { AnyCircuitElement } from "circuit-json"
+
+describe("checkViaToPadSpacing", () => {
+  test("returns error when via is too close to a rectangular SMT pad", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+      {
+        type: "pcb_smtpad",
+        pcb_smtpad_id: "pad1",
+        shape: "rect",
+        x: 0.5,
+        y: 0,
+        width: 0.4,
+        height: 0.3,
+        layer: "top",
+      },
+    ]
+    // Via edge at 0.3, pad left edge at 0.3 => gap = 0mm, well below 0.2mm default
+    const errors = checkViaToPadSpacing(soup)
+    expect(errors).toHaveLength(1)
+    expect(errors[0].message).toContain("too close to pad")
+  })
+
+  test("no error when via is far from pad", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+      {
+        type: "pcb_smtpad",
+        pcb_smtpad_id: "pad1",
+        shape: "rect",
+        x: 2,
+        y: 0,
+        width: 0.4,
+        height: 0.3,
+        layer: "top",
+      },
+    ]
+    const errors = checkViaToPadSpacing(soup)
+    expect(errors).toHaveLength(0)
+  })
+
+  test("returns error when via is too close to a circular SMT pad", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+      {
+        type: "pcb_smtpad",
+        pcb_smtpad_id: "pad1",
+        shape: "circle",
+        x: 0.7,
+        y: 0,
+        radius: 0.2,
+        layer: "top",
+      },
+    ]
+    // center-to-center = 0.7, via radius = 0.3, pad radius = 0.2 => gap = 0.2mm
+    // gap + EPSILON (0.005) >= 0.2mm minSpacing => no error
+    const errors = checkViaToPadSpacing(soup)
+    expect(errors).toHaveLength(0)
+  })
+
+  test("returns error when via is too close to a plated hole", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+      {
+        type: "pcb_plated_hole",
+        pcb_plated_hole_id: "hole1",
+        shape: "circle",
+        x: 0.6,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.5,
+        layers: ["top", "bottom"],
+        pcb_component_id: "comp1",
+        pcb_port_id: "port1",
+      },
+    ]
+    // center-to-center = 0.6, via radius = 0.3, hole radius = 0.25 => gap = 0.05mm < 0.2mm
+    const errors = checkViaToPadSpacing(soup)
+    expect(errors).toHaveLength(1)
+    expect(errors[0].message).toContain("too close to pad")
+  })
+
+  test("respects custom minSpacing parameter", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+      {
+        type: "pcb_smtpad",
+        pcb_smtpad_id: "pad1",
+        shape: "rect",
+        x: 1.0,
+        y: 0,
+        width: 0.4,
+        height: 0.3,
+        layer: "top",
+      },
+    ]
+    // Via edge at 0.3, pad left edge at 0.8 => gap = 0.5mm
+    // With default 0.2mm: no error. With 0.6mm: error
+    expect(checkViaToPadSpacing(soup)).toHaveLength(0)
+    expect(checkViaToPadSpacing(soup, { minSpacing: 0.6 })).toHaveLength(1)
+  })
+
+  test("returns empty array when no vias", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_smtpad",
+        pcb_smtpad_id: "pad1",
+        shape: "rect",
+        x: 0,
+        y: 0,
+        width: 0.4,
+        height: 0.3,
+        layer: "top",
+      },
+    ]
+    expect(checkViaToPadSpacing(soup)).toHaveLength(0)
+  })
+
+  test("returns empty array when no pads", () => {
+    const soup: AnyCircuitElement[] = [
+      {
+        type: "pcb_via",
+        pcb_via_id: "via1",
+        x: 0,
+        y: 0,
+        hole_diameter: 0.3,
+        outer_diameter: 0.6,
+        layers: ["top", "bottom"],
+      },
+    ]
+    expect(checkViaToPadSpacing(soup)).toHaveLength(0)
+  })
+})


This test file violates the rule that states 'A *.test.ts file may have AT MOST one test(...), after that the user should split into multiple, numbered files. e.g. add1.test.ts, add2.test.ts'. The file contains 8 test() calls within a describe block (lines 6, 34, 60, 87, 117, 145, 161). To fix this violation, split the tests into multiple numbered files like check-via-to-pad-spacing1.test.ts, check-via-to-pad-spacing2.test.ts, etc., with each file containing only one test() call.

Spotted by Graphite (based on custom rule: Custom rule)

Is this helpful? React 👍 or 👎 to let us know.

This file was from an earlier commit and has already been removed. The tests are now split into individual files under tests/lib/check-via-to-pad-spacing/ (one test per file).

One test per file per project convention. Moved tests from single file into check-via-to-pad-spacing/ directory. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

seveibar

no snapshots (not yet reviewable) please have one clear test with a pcb snapshot

Address review feedback: add SVG snapshots with error rendering to the error-detecting test cases so reviewers can visually verify the DRC check behavior. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev · 2026-03-08T17:17:54Z

Added PCB snapshots with error rendering to the two error-detecting tests (via-too-close-to-rect-pad and via-close-to-plated-hole). Tests were already split into one-test-per-file in a previous commit. All 7 tests pass.

lucaferri-dev · 2026-03-08T17:22:01Z

Hi @seveibar, the PCB snapshots were actually pushed just after your review (commit 0334297). Both via-too-close-to-rect-pad and via-close-to-plated-hole tests now include SVG snapshots rendered with shouldDrawErrors: true. Each test file has exactly one test. Please take another look when you get a chance!

graphite-app · 2026-03-08T17:22:40Z

lib/check-via-to-pad-spacing.ts

+function getPadRadius(pad: Pad): number {
+  if (pad.type === "pcb_smtpad") {
+    if (pad.shape === "circle") return pad.radius
+    if (
+      pad.shape === "rect" ||
+      pad.shape === "rotated_rect" ||
+      pad.shape === "pill" ||
+      pad.shape === "rotated_pill"
+    ) {
+      return Math.max(pad.width, pad.height) / 2
+    }
+    return 0


Documentation bug: Comment on line 15-16 claims the function "returns the half-diagonal (conservative bounding circle)" for rectangular pads, but the implementation at line 27 returns Math.max(pad.width, pad.height) / 2, not the actual half-diagonal.

The correct half-diagonal would be:

return Math.sqrt(pad.width ** 2 + pad.height ** 2) / 2

For a 1mm × 0.3mm pad:

Current code: 0.5mm

Correct diagonal: 0.523mm

This underestimates the bounding circle by ~4-5% for rectangular pads. While this fallback path is only used for unknown shapes (the main paths use proper rectangle-to-circle distance), it could still cause missed spacing violations for unsupported pad shapes.

Suggested change

function getPadRadius(pad: Pad): number {

if (pad.type === "pcb_smtpad") {

if (pad.shape === "circle") return pad.radius

if (

pad.shape === "rect" ||

pad.shape === "rotated_rect" ||

pad.shape === "pill" ||

pad.shape === "rotated_pill"

) {

return Math.max(pad.width, pad.height) / 2

}

return 0

function getPadRadius(pad: Pad): number {

if (pad.type === "pcb_smtpad") {

if (pad.shape === "circle") return pad.radius

if (

pad.shape === "rect" ||

pad.shape === "rotated_rect" ||

pad.shape === "pill" ||

pad.shape === "rotated_pill"

) {

return Math.sqrt(pad.width ** 2 + pad.height ** 2) / 2

}

return 0

Spotted by Graphite

Is this helpful? React 👍 or 👎 to let us know.

The comment and implementation are now consistent — the current code on line 27 uses Math.sqrt(pad.width ** 2 + pad.height ** 2) / 2 which is the half-diagonal. This was fixed in commit 0c04968. The bot may have been looking at an older version.

- Add circuit JSON asset with board, components, traces and vias for a realistic PCB snapshot showing via-to-pad spacing violations - Fix getPadRadius to use actual half-diagonal (sqrt) instead of max/2 for the fallback bounding circle on rectangular pads Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev

Added a realistic PCB snapshot test with a board, components (R1, C1), traces, and vias — one via placed too close to a pad. The snapshot clearly shows the violation highlighted in the SVG.

Also fixed the getPadRadius fallback to use the actual half-diagonal (Math.sqrt(w² + h²) / 2) instead of Math.max(w, h) / 2 as flagged in the Graphite review.

All 8 tests pass.

Renders SOIC8 chips with a via placed too close to a pad, showing the DRC error highlighted in the SVG snapshot. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev

Added a clear PCB snapshot test (pcb-snapshot.test.tsx) that renders real SOIC8 chips using tscircuit, with a via placed too close to a pad. The SVG snapshot shows the DRC error highlighted on a realistic PCB layout.

Remove pcb-snapshot.test.tsx which produced an SVG with unrelated DRC errors (trace-to-edge, component-outside-board) that obscured the via-to-pad check. Keep via-too-close-to-pad-snapshot.test.ts which uses a clean JSON fixture and produces a focused snapshot showing only the relevant elements. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev · 2026-03-08T18:42:40Z

Snapshots have been added since your review (which was on commit 0783bc3, before any snapshots existed).

Removed the noisy pcb-snapshot.test.tsx which had unrelated DRC errors from tscircuit rendering. Kept via-too-close-to-pad-snapshot.test.ts which uses a clean JSON fixture with a focused PCB snapshot showing just the relevant elements (board, pads, vias, traces).

There are also snapshot tests for specific pad shapes (rect pad, plated hole). All 8 tests pass.

seveibar

don't use the word soup, just one test is all that's needed.

You're not handling most smtpad/plated hole shapes, so this is going to create a lot of bugs. We need to have comprehensive coverage, you could use getBoundsOfPcbElements from circuit-json-util to approximate for now, and we can have a more intelligent computeGapBetweenCopper(elm1, elm2) inside circuit-json-util in the (near) future

Otherwise looks good

Use getBoundsOfPcbElements from circuit-json-util for generic pad shape handling instead of manually dispatching on every shape variant. Reduces to one clear test with a real tscircuit PCB snapshot per review feedback. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev · 2026-03-08T19:55:36Z

Addressed the review feedback:

Replaced manual shape dispatch with getBoundsOfPcbElements from circuit-json-util — this handles all pad shapes generically (including future ones) using bounding boxes instead of manually switching on each shape variant.
Simplified to one clear test with a real tscircuit PCB snapshot (pcb-snapshot.test.tsx), removed all the individual unit tests.

All 67 tests pass.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev

Addressed the feedback:

Replaced manual shape-specific gap calculations with getBoundsOfPcbElements from circuit-json-util for comprehensive pad shape coverage
Consolidated all tests into a single test using the realistic circuit JSON fixture

All 68 tests pass.

graphite-app bot reviewed Mar 8, 2026

View reviewed changes

refactor: split via-to-pad-spacing tests into separate files

0783bc3

One test per file per project convention. Moved tests from single file into check-via-to-pad-spacing/ directory. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

seveibar requested changes Mar 8, 2026

View reviewed changes

fix: add PCB snapshots to via-to-pad spacing tests

0334297

Address review feedback: add SVG snapshots with error rendering to the error-detecting test cases so reviewers can visually verify the DRC check behavior. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

graphite-app bot reviewed Mar 8, 2026

View reviewed changes

lucaferri-dev commented Mar 8, 2026

View reviewed changes

feat: add clear PCB snapshot test with real tscircuit circuit

1bc5814

Renders SOIC8 chips with a via placed too close to a pad, showing the DRC error highlighted in the SVG snapshot. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev commented Mar 8, 2026

View reviewed changes

seveibar requested changes Mar 8, 2026

View reviewed changes

test: add single via-to-pad spacing test with realistic circuit

b525442

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

lucaferri-dev commented Mar 8, 2026

View reviewed changes

Conversation

lucaferri-dev commented Mar 8, 2026

Summary

Test plan

Uh oh!

graphite-app bot Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

seveibar left a comment

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev commented Mar 8, 2026

Uh oh!

lucaferri-dev commented Mar 8, 2026

Uh oh!

graphite-app bot Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev left a comment

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev left a comment

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev commented Mar 8, 2026

Uh oh!

seveibar left a comment

Choose a reason for hiding this comment

Uh oh!

lucaferri-dev commented Mar 8, 2026

Uh oh!

lucaferri-dev left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants