Chip Design Solutions for Deep Sub-micron Effects - Part II

Part II is a continuation of Part I on the descriptions and critique of products with improved capability for better timing closure, signal integrity control, power, and electro-migration analyses. In this segment, the author discusses products from two suppliers with different approaches to solving the problems. Avant!, having a history of acquisitions and proven physical design solutions for deep sub-micron chips, is leveraging and building upon products and technology that they already offer in the market. Magma Design Automation is developing all products with new technology and a singular data model. Magma has as its goal a complete correct-by-construction solution from synthesis through layout.

Deep Sub-micron Tools - Avant!
Avant! refers to its product program to solve the design challenges from deep sub-micron effects as the "SinglePass Initiative." The company has been known for its success in the physical design arena. Some of its newer products offer capabilities at the logical or RTL levels.

Jupiter includes some RTL analysis, design planning technology, and deep sub-micron synthesis technology. The tool is designed to use the proven physical implementation algorithms from Apollo, Saturn, Planet, and Mars, and to present these capabilities to logic designers in a form familiar to them. Logic designers may then choose the architecture using estimates of what the real physical implementation might be.

Apollo, for timing-driven place and route, takes into account capacitive effects, clock skew, and variances due to slew rates. Timing constraints are applied on paths and globally. While doing this, compaction is applied within these constraints.

Saturn synthesizes clock trees with lowest skew, highest speed, and load balancing. It evaluates logic and inserts buffers and sizes cells. For example, it may apply a stronger driving cell. It may re-optimize the logic if particular paths fail by applying different Boolean logic functions. The netlist is then automatically re-generated. For engineering change orders (ECOs), Avant! has incremental capabilities to avoid full re-layout of a chip.

Mars-Rail assesses voltage drop and optimizes the bus implementation. It provides analysis and advice, such as when the bus is loaded too much with too many switching cells. Wire width and length are suggested to control electro-migration, passing this information to Apollo for proper implementation. Mars-Xtalk investigates cross coupling between wires and develops the metrics for the closeness of wires for the specific path.

Avant! believes that the design community needs improved or new design solutions for low-power design, signal integrity control, and the addressing of manufacturing issues for optical proximity, phase shift, and protection diodes.

Avant!'s tools operate on the Milkyway data base, which Avant! believes improves productivity and efficiency.

"SinglePass was pioneered to improve design productivity for very deep sub-micron designers. The solution includes new technologies that optimize weighted characteristics of a complex design such as area, timing, power, noise, and manufacturing issues. By addressing these challenges, RTL, logic, and physical designers achieve timing closure with fewer interations, and as a result bring their products to market more quickly." Bill Alexander, Corporate Product Management, Avant! Corporation.

Blast Fusion - Magma Design Automation
The Blast Fusion system contains built-in physical design and optimization engines and analysis tools that utilize a single, memory-resident data model. This unified data model allows the tools to work concurrently, optimizing for logic, area and signal integrity of the design while maintaining timing during the place-and-route process. A user of Blast Fusion uses a design flow consisting of the following major steps:

To import a design a structural net list is read, flattened, and unmapped. Target libraries are read, and timing constraints are set.

For timing optimization the design is re-optimized and mapped onto SuperCells™. (SuperCells are functional placeholder cells that have a fixed delay but variable size.) Blast Fusion automatically abstracts SuperCells out of the target library. Each functionally equivalent family of cells is represented by one SuperCell that has a fixed delay, but variable area.) All delays in the circuit are determined and frozen.

For area optimization, the cells are placed and simultaneously sized to meet the timing based on the actual loads. Buffers and clock are inserted.

For wire optimization, detailed wiring is performed, maintaining the delay budgets through tuning the wire widths and spacings. A GDSII file is then output.

Magma states that this methodology applied with technology described below enables Blast Fusion to establish the best possible timing for a design prior to detailed layout, and to hold that timing constant throughout the place-and-route process, delivering a final layout that meets the predicted timing without any iterations back through synthesis.

Since its initial market introduction, Magma has enhanced Blast Fusion's initial capabilities for managing signal integrity problems. Blast Fusion now provides a correct-by-construction design flow that automatically addresses antenna effects, crosstalk, and electromigration of signal nets. Techniques used include adding jumpers, inserting buffers, varying wire width and spacing, and re-ordering nets. Magma plans further enhancements to Blast Fusion's signal integrity capabilities that are expected to include IR drop management, power rail design, and electro-migration control.

Their goal is to offer a complete silicon implementation system from RTL specification through logic synthesis and final physical placement and layout. They believe that the key to delivering an RTL-to-GDSII design environment is their "gain-based synthesis" technology. This technology builds upon their SuperCell approach and is to provide the designer fast, high-capacity synthesis with guaranteed timing. Magma expects this solution will offer a faster overall design process.

At this writing, 3Dlabs is currently designing a production chip using Blast Fusion, and Fujitsu and Texas instruments are in the process of deploying the tool to their ASIC design centers.

"The EDA industry is in uncharacteristic agreement about the toughest challenges of deep submicron design. Achieving timing closure and eliminating iterations is the goal of each EDA company. However, customers must look closely at the long-term viability of the varying approaches. Enhancements to existing solutions provide incremental improvements, but are not scalable to 0.15 micron and below process technologies. Magma's unique approach addresses the industry's next-generation process and design technologies." Rochelle Perry Drenan, Director of Product Marketing, Magma Design Automation.

Comparison of Strategies
The product strategies for these companies contrast in that Magma is developing a new chip design system with a new architecture for melding logical and physical design. At the heart of their system is their FixedTiming™ methodology implemented as described above. Avant! uses as its operational tool foundation the Milkyway data base with an evolutionary deep sub-micron tool technology to accomplish their goal of single-pass, or perhaps reduced-iteration, deep sub-micron chip design. While users of the Magma system at this writing depend upon design synthesis from other EDA companies, they expect to provide such capability for tighter coupling of logical and physical design.

Avant! can build upon its established customer relationships to upgrade already operating physical design tools with increased functionality and offer tool additions. They will be challenged by some of their competition on the level of seamless integration that their tools offer. However, the time span for Avant!'s tools to be introduced at companies is expected to be less when compared to Magma's approach. The author expects that Avant! has the opportunity to rapidly realize a good return on their R&D investment. Their physical design system is viewed by the author as a somewhat open system in that a user may be able to substitute a tool from another EDA company, even though the substitution may not take advantage of the common Milkyway data base. With Magma's planned system, it not evident that they will allow for such a mixed environment. However, this opinion should be verified directly with Magma.

The introduction of a tool system such as Magma's typically requires a long period of time. There is the evaluation, the library insertion, the training on the methodology and on the tool set, the system qualification to the 'golden' level, and the launching of the system to multiple design sites. If the realized results are sufficiently favorable, users will make such an investment of time to achieve their goals. From a business perspective, Magma is taking the higher risk to achieve all of its objectives but, if successful, their business reward has the potential to be quite high.

Considering the extensive list of products described in Part I of this series, design users are confronted with many choices. Because of the time it takes for such evaluations, users will likely choose to evaluate only a few solutions. Some suppliers may find it difficult to convince customers to evaluate their products.

Workstations and Design Tools Archive

Guides and Experts Analog Avenue PLD EDA Tools PLD DSP EDA Embedded Systems Power Test


		Click here to get your listing up.