Semi-Methodology
Souletics Software · Semiconductor Engineering
From Disruptive Idea
to Tapeout
End-to-end semiconductor physical design — from VHDL concept through schematic capture, custom layout, full-chip integration, and tapeout signoff. 20 years at Intel. Schematic to silicon. Every step.
AI Accelerators
NPU, TPU, inference engine silicon
Memory
SRAM, HBM-adjacent, embedded memory
CPU & GPU
High-performance compute silicon
Security Chips
Cryptography, secure enclave silicon
RF & Mixed-Signal
Analog, RF, mixed-signal custom blocks
Full-Chip SoC
System-on-chip integration & signoff
Design for Manufacturability — Front End First
Wire length, wire width, spacing rules, and abutment constraints should all be determined at the front end — before physical implementation begins. DFM violations discovered late in the flow cost time, mask money, and yield. Our methodology embeds manufacturability constraints from the first schematic pass, eliminating the costly late-stage iterations that plague most physical design programs.
The Engineering Design Flow
Every chip starts with a disruptive idea. Architecture is defined in hardware description language — VHDL or Verilog — capturing the intended behavior of every block before a single transistor is placed. This is where AI accelerator architecture, security chip algorithms, and memory subsystem behavior are formally specified. DFM constraints are embedded here — wire budget, block sizing, critical path targets.
RTL behavior is translated into circuit-level schematics — transistor-accurate representations of every cell, block, and interconnect. For digital blocks, synthesis maps RTL to standard cell netlists. For analog and custom blocks, hand-crafted schematics define transistor sizing, bias networks, and signal paths. This is where digital meets analog and where custom logic is born.
Physical implementation of every cell and block. Custom digital layout — full-custom transistor-level implementation of critical path cells, memory bitcells, and high-performance logic. Analog layout — precision placement of matched devices, guard rings, shielding, and substrate contacts to minimize noise and parasitics. Wire length and wire width targets established in front end are enforced here. Abutment compatibility verified between adjacent blocks.
The full chip comes together. Floorplanning positions every macro and custom block. Power planning delivers clean supply across the die. Placement distributes standard cells. Clock tree synthesis balances skew across every register. Routing connects every net while honoring spacing rules, via enclosures, and density targets. Abutment issues between custom blocks and APR regions are resolved here — this is the integration challenge that separates experienced physical designers from novices.
DRC — Design Rule Check: Every polygon on every layer verified against the foundry's PDK design rules. Min width, min spacing, enclosure, extension, fin snap grids, density fill. Zero violations required before tapeout. LVS — Layout vs Schematic: Physical layout extracted and compared against schematic netlist. Shorts, opens, and parameter mismatches identified. LVS clean required. ERC — Electrical Rule Check: Antenna violations, floating nodes, ESD compliance, density uniformity for CMP. All three must pass with zero errors.
Timing signoff at all PVT corners — fast corner (FF, low temp, high voltage) and slow corner (SS, high temp, low voltage). Setup and hold verified across every path. Waveform verification — gate-level simulation with extracted parasitics confirms functional correctness at signal level. Dummification — metal and poly dummy fill added across every layer to meet foundry density requirements for CMP uniformity. This is the final preparation before the GDSII is released.
The GDSII file — containing every polygon on every layer of the complete chip — is delivered to the foundry. Masks are cut. Silicon is manufactured. 20+ tapeouts completed at Intel, including a solo cryptography chip tapeout. Every step above has been executed in production, at leading-edge nodes, on real silicon that shipped. That is not a claim. That is a record.
Cadence Virtuoso
CADENCE DESIGN SYSTEMS
Cadence Innovus
CADENCE DESIGN SYSTEMS
Cadence Tempus
CADENCE DESIGN SYSTEMS
Synopsys Design Compiler
SYNOPSYS
Synopsys PrimeTime
SYNOPSYS
Mentor Calibre
SIEMENS EDA
Cadence Spectre
CADENCE DESIGN SYSTEMS
Synopsys IC Validator
SYNOPSYS
180nm – 65nm
Legacy / Mature
Automotive, industrial, analog, power management. Fully depreciated fabs. Cost-effective and robust.
45nm – 22nm
Mainstream
FinFET introduction. IoT, embedded, mixed-signal. Strong IP ecosystem.
14nm – 10nm
Workhorse
Intel's 14nm era. High-performance compute, server CPUs. Multi-patterning expertise required.
7nm – 5nm
Advanced
EUV lithography. AI chips, mobile SoCs. Highest density and performance.
Ready to start?
Schedule a Technical Consultation
20 years. 20+ tapeouts. Schematic to GDSII.
Bring your chip concept and let's talk about what it takes to get to silicon.