Siemens Intros AI-Powered Library Characterizer to Speed IC Design

Not too long ago Siemens announced Solido Characterizer, the newest addition to its Solido Characterization Suite, at its U2U EU occasion in Munich. Developed for foundries and in-house chip design groups, the software program targets library characterization workflows throughout all course of nodes, providing what Siemens says is a seven instances enchancment in throughput for Liberty file technology.

Library characterization is the method of making the timing, energy, and noise fashions that digital design instruments depend on to judge circuit conduct. As course of applied sciences shrink, nook counts develop, and rising codecs just like the Liberty Variation Format (LVF) add statistical variation knowledge to each library, producing correct Liberty information has develop into a bottleneck that may delay tapeout schedules.

On this YouTube video, Sathish Balasubramanian, Head of Merchandise for Solido Customized IC and EDA AI at Siemens EDA, describes Solido Characterizer.

A single characterization run at a complicated node could require simulations throughout lots of of course of, voltage, and temperature corners, and Solido Characterizer makes use of predictive AI to compress that timeline from weeks to days.

Two Engines, One Seven Occasions Speedup

The seven instances throughput is achieved by means of two complementary applied sciences. The primary is a quicker AI engine that drives Liberty technology for multi-PVT creation and applies superior LVF methods, delivering a 5 instances speedup by itself. LVF captures statistical variation knowledge on the library stage, and its rising adoption has added substantial complexity to each characterization run. The AI engine accelerates the technology of the big datasets LVF requires, dealing with each normal timing and energy knowledge and the newer variation-aware codecs.

The second contributor is Solido LibSPICE, which Siemens describes because the trade’s first purpose-built, AI-accelerated simulator for characterization workloads. Basic-purpose SPICE simulators deal with a variety of circuit evaluation, however characterization includes operating giant numbers of structurally comparable simulations throughout many corners with slight parameter variations. LibSPICE exploits that regularity, offering a further two instances or larger efficiency enhance.

Solido Characterizer gives high-performance, high-throughput, trendy normal cell library characterization.

Solido Characterizer additionally integrates with Solido Analytics, giving engineers real-time high quality assurance of outcomes, useful resource monitoring for in-progress runs, and automatic rerun capabilities. Engineers can catch points throughout characterization fairly than discovering them downstream throughout signoff, lowering the pricey cycle of producing a library, discovering errors late, and re-running all the circulation. Extra data is accessible within the Solido Characterization Suite fact sheet.

Generative AI Enters the Characterization Circulation

Past the speedup, Siemens is layering generative and agentic AI capabilities on prime of the workflow by means of its Fuse EDA AI system. A GUI assistant gives tool-specific steering and error decision on the level of want, lowering the training curve for much less skilled engineers. A log analyzer pinpoints errors and recommends fixes by cross-referencing documentation, and an automatic rule technology engine creates customized validation checks throughout Liberty metadata, lookup tables, and corners.

Solido Generator builds an AI mannequin of the library to supply production-accurate .lib fashions at new PVT corners 100x+ quicker than SPICE.

Solido Characterizer additionally works in tandem with Solido Generator, one other part of the Characterization Suite. Generator makes use of baseline Liberty information produced by Characterizer to coach AI fashions that produce further library views with out operating SPICE simulation, producing new PVT libraries in minutes fairly than the hours or days conventional strategies require.

Collectively, Characterizer and Generator type an end-to-end library creation pipeline that addresses each the preliminary characterization bottleneck and the scaling problem of manufacturing libraries throughout giant PVT matrices.

Early Adopters Report Manufacturing-Prepared Outcomes

GlobalFoundries has reportedly been utilizing the software in its design enablement workflows. James Culp, fellow in Design Enablement at GlobalFoundries, mentioned the staff validated and created design margins and libraries of wonderful high quality, sustaining manufacturing accuracy correlated to SPICE fashions whereas reaching a 20% to 30% speedup inside inner flows. Culp mentioned the adoption course of was fast and simple, and that the staff delivered signoff-ready margins and Liberty information for a number of high-profile prospects throughout a number of GlobalFoundries applied sciences.

Anatrix, a supplier of radiation-hardened digital gate libraries, has additionally deployed Solido Characterizer for its single-event latch-up radiation-hardened IP. Greg Pauls, president of Anatrix, mentioned that silicon testing verified post-layout parasitic simulations utilizing the Siemens EDA circulation precisely predicted IC conduct, reinforcing belief within the outcomes and supporting first-pass silicon success.

Solido Characterizer is accessible now and helps normal cell, I/O, reminiscence, and customized cell characterization and handles all Liberty knowledge sorts, together with timing, energy, noise, and variation throughout NLDM, CCS, LVF, and Moments knowledge buildings. The software program is a part of Siemens’ broader Solido Customized IC portfolio, which spans SPICE simulation, variation-aware verification, and IP validation.

All photographs used courtesy of Siemens.