In the dynamic realm of system design, ensuring signals arrive unscathed at their destinations is just the tip of the iceberg. The intricacies evolve with escalating package densities, finer PCB traces, and soaring frequencies, requiring a synergy of electrical, mechanical, electromagnetic, and thermal dynamics. The growing complexities and niceties unveil the need for optimal performance, a reliable process where human ingenuity meets computational prowess, all while navigating organizational silos hindering collaboration among experts from diverse disciplines. System-level optimization has become necessary rather than a luxury to meet these soaring demands.
The manual workflow of building, testing, prototyping, refining, and eventually manufacturing has become a significant limitation. The current approach to design optimization heavily relies on the designer's intuition, where prototypes are created, and simulations are run to assess alignment with goals. However, as electronic designs pursue enhanced performance, increasing complexity, and compactness, the optimization challenge surpasses human intuition. Advanced optimization methods are needed to handle the growing complexity of modern designs.
AI-Based Optimization
Cadence has introduced the Optimality Intelligent System Explorer, a new optimization technology that leverages AI to help designers tackle modern design challenges. This technology offers multidisciplinary design analysis optimization (MDAO) capabilities and can perform system-level optimization seamlessly, from the IC to the package and up through the board. Integrating multiphysics analysis tools with the Optimality Explorer ensures that the desired outcomes are achieved. Automation greatly accelerates the optimization process, making it easier for engineers and designers to achieve their goals more efficiently.
The Optimality Explorer workflow specifies input parameters, optimizes system criteria, and employs multiphysics analysis tools for simulations. It automates the optimization process, producing the optimized design and final curves. Users can optimize for parameters like return loss, insertion loss, crosstalk isolation, and system criteria such as eye diagram, Jitter, and BER. For effective optimization, designers must consider geometric variables such as line width, spacing, length, stack-up for traces, pad size, anti-pad geometry, drill size, and stub length for vias. Parameters like PVT corners, on-die termination (ODT), jitter equalization, and more must also be considered when creating models.
Optimality Explorer aims to empower designers for seamless, user-intervention-free design optimization. Its algorithm automates the optimization process, ensuring a smooth and user-friendly experience. It produces significant results in fewer than 500 iterations, achieving faster convergence than traditional methods. It is referred to as in-design AI-driven multi-disciplinary analysis and optimization.
Optimality Explorer simulates and optimizes complex 3D layouts efficiently and accurately, handling traditionally considered challenging to optimize scenarios. It includes field solvers for PC package interconnects and can handle various scenarios usually regarded as challenging to optimize, such as maximizing cross-hatch.
Parameters and Considerations in Optimization
For instance, in a system communication channel, you have transmitters, receivers, interconnects for PCB, packages, and interposes. These components are abstracted as IBIS-AMI models for your transmitter-receiver, and between them are traces and vias.
To ensure optimal channel performance, it is important to consider various geometric variables such as line width, spacing, length, stack-up for traces, pad size, anti-pad geometry, drill size, and stub length for vias. Parameters such as PVT corners, on-die termination (ODT), and jitter equalization should also be considered when creating models.
During the optimization process, it is necessary to specify the design parameters that require optimization and the desired optimization goal. Further, it is essential to create additional surrogate models to optimize these parameters effectively.
Optimality Explorer builds and trains machine learning models based on an initial dataset generated using a random search. It continually analyzes the simulations, updating design variables and calculating objective functions and constraints until it reaches a stopping criterion and convergence.
Optimality Explorer is designed to simplify the optimization process for designers, especially when there are many tunable parameters to consider. Its algorithm automates the optimization process, making it user-intervention-free and easy to use.
Compared to traditional methods that require over 2500 iterations for comparable outcomes, the Optimality Explorer can produce significant results in fewer than 500 iterations, achieving faster convergence.
Realizing Efficiency Gains with Optimality Explorer
In intricate circuit layouts, using individual traces and vias alone is not enough. A combination of these components is required to create interconnected designs where each component influences the behavior of the others.
Differential Pairs Between Two Cross-Hatch Planes
Optimality Explorer simulates and optimizes complex 3D layouts efficiently and accurately, tackling scenarios that are traditionally challenging to optimize. It includes field solvers for PC package interconnects and can handle various scenarios usually regarded as challenging to optimize. For instance, it can maximize cross-hatch patterns in differential pair designs to achieve better results. Optimality Explorer can also reduce the number of simulations required for an exhaustive sweep and reach the target much faster.
Optimality Explorer can optimize both pre-layout and post-layout designs. For example, it optimized an RF power divider, achieving the target with only 46 simulations, compared to an exhaustive sweep that would have taken over 3 million simulations. Optimality Explorer’s versatility extends to handling designs with many parameters, as demonstrated in optimizing a trip patch antenna with 16 parameters in just 71 iterations.
Future Horizons: Expanding the Optimality Platform
The team behind the Optimality Explorer tool is currently working on expanding the platform to cover the thermal and fluid dynamics domains. This involves the incorporation of the Celsius 3D Solver for thermal analysis and CFD for fluid dynamics. Additionally, electrical constraints will be integrated into the existing constraint manager in the Allegro X Design Platform, providing users with a more comprehensive solution. The development team will continuously offer updates on these improvements.
Driving Multiphysics Analysis of Electronic Systems
Deciphering the labyrinth of high-speed signal optimization in modern systems is a multidimensional challenge. Optimality Explorer breaks through the limitations of the conventional human-intensive optimization process by replacing the traditional interactive flow of the design-test-refine loop with AI-driven technology that results in optimal system design solutions. Optimality Explorer stands as a beacon, guiding designers through this intricate landscape, offering automation, efficiency gains, and a path toward the future of comprehensive design solutions.
Watch the CadenceTECHTALK, Optimality Intelligent System Explorer: Applying AI/ML Technology for Rapid Design Optimization to learn more.