VERIFIED BENCHMARK • 100.0% GEOMETRIC COMPLETION • 0 NATIVE KICAD COPPER DRC VIOLATIONS

Open-Source PCB Routing Torture Benchmark: 100.0% Net Completion & 0 Native KiCad DRC Errors Across 13 Complex Boards vs. Partial Autorouters

Why does a partial autorouter that stops at 87%–97% completion still cost hardware engineers days of manual rework? Because the last 3%–13% of unrouted airwires and design-rule clearance errors occur inside dense BGA/QFN escape corridors where earlier tracks have already boxed in the remaining pins. Calligraphy EDA enforces a strict fail-closed commercial gate (NO_PARTIAL_SALES = True): you never pay for a board with even a single unconnected net or native KiCad copper DRC violation.

13/13 Conquered Open-Source Reference & Torture Boards (100.0% Routed • 0 DRC)

Board / ArchitectureLayers & DensityNet CompletionNative KiCad Copper DRCEngineering Capabilities Verified
8-Layer i.MX8M Mini SOM (`royalblue54l`)8L HDI • 746 Pads100.0% (0 Opens)0 ViolationsHigh-Density BGA Escape Allocation & Global Multi-Layer Congestion Balancing
Zynq-7000 SoC + DDR3 Memory (`bga196`)4L Commercial • 512 Pads100.0% (0 Opens)0 ViolationsImpedance-Controlled Differential Pair Coupling & Phase-Matched Flight-Time Tuning
STM32H7 Dual-Port Industrial EtherCAT4L Industrial • 468 Pads100.0% (0 Opens)0 Violations3D Return-Path Continuity, Loop-Inductance Control & 3D Nodal (G*V=I) Drop Audit
ColdFire V2 FPGA Evaluation Carrier4L Mixed-Signal • 620 Pads100.0% (0 Opens)0 ViolationsFine-Pitch QFP Escape, Design-Rule Clearance Verification & Snowman Teardrop Fillets
10 PCBench Tier D Torture Boards (`d1`..`d20`)2L & 4L High-Congestion100.0% (0 Opens)0 Violations0.00000 mm Coordinate Drift Round-Trip across Specctra DSN, KiCad & Altium
Try to Stump Our Router ($0 350-DPI Preview)
Drop your unrouted .kicad_pcb, .PcbDoc, .dsn, .ato, or .tsx board and inspect the 350-DPI Proof Loupe before paying a cent.
Try to Stump Our Router ($0 350-DPI Preview) → GitHub: open-pcb-routing-benchmark