The "shift left" philosophy, which moves critical verification and analysis to earlier in the design process, has already transformed digital design and is now making significant inroads into the AMS domain.
This approach represents a fundamental rethinking of the development process, challenging traditional workflows that have dominated semiconductor design for decades.
In my experience, shift left isn't just about changing when verification happens, but how we reimagine the entire design journey, allowing organisations to identify potential issues at the earliest possible stage. For AMS designers who have historically relied on detailed circuit simulations late in the development cycle, this evolution brings both challenges and unprecedented opportunities for efficiency and innovation.
The Imperative for Change
The key reason organisations are looking for alternative methodologies is because the stakes for getting AMS designs right have never been higher. With advanced nodes below 7nm, even a single design iteration can cost millions of pounds, not just in resources, but in lost market opportunity.
Market windows continue to narrow, with delayed products potentially missing entire commercial cycles. Meanwhile, increasing interdependencies between analogue and digital domains amplify the impact of analogue issues discovered late in the cycle.
What's particularly challenging in modern AMS design is the growing complexity of multiphysics interactions. As circuit speeds increase and silicon manufacturing advances, novel physical effects, including thermal, electromagnetics, and layout-dependent effects, all come into play that were previously negligible.
These multiphysics interactions create new design challenges that traditional methodologies struggle to address efficiently.
Consider a high-performance AMS design where thermal effects from power-hungry digital blocks affect sensitive analogue circuits, or where electromagnetic interference between components creates unexpected performance degradation. Without early multiphysics analysis, these issues remain hidden until late-stage verification, often requiring costly redesigns or performance compromises.
In a global semiconductor market projected to reach £550 billion ($697 billion) this year, driven by advancements in AI, cloud computing, and automotive electronics, such delays represent substantial lost revenue. The UK semiconductor industry, employing approximately 15,000 professionals, faces these same pressures.
Traditional approaches that defer detailed analysis until layouts are nearly complete cannot meet these challenges. The old methods of designing first, verifying later have become unsustainable as complexity grows and margins shrink. Early performance prediction, particularly for multiphysics interactions, is no longer an optional addition to have in strategies, but is now essential for managing risk and ensuring competitiveness.
Enabling Technologies for Early Prediction
The core of this shift-left evolution lies in new methodologies that provide performance insights much earlier in the design process. Engineers now implement behavioural modelling techniques to capture circuit characteristics before detailed implementation. These models provide functional representations that enable early validation of design concepts.
Additionally, advanced simulation techniques can deliver performance estimates with manageable computational loads, making it possible to run thousands of simulations in the time previously required for dozens. Design teams can leverage this expanded coverage to identify corner cases and potential failure modes early, when changes are least expensive.
Comprehensive multiphysics analysis capabilities are particularly transformative in this space. Modern AMS designs require the ability to simultaneously model thermal, electromagnetic, electrostatic discharge, and other physical interactions that affect circuit performance. Early-stage multiphysics modelling enables designers to anticipate these complex interactions before they become problematic, leading to more robust architectures from the outset.
What's truly shifting expectations in this space is the role of AI and machine learning, tools that, not long ago, felt like science fiction. AI models trained on previous designs can now predict behaviour from high-level specifications, enabling quick evaluation of multiple architecture options. This capability transforms the exploration phase of design, allowing teams to evaluate far more alternatives than traditional approaches would permit.
The inherent tension in early-stage analysis is between speed and accuracy. Analysis performed earlier has less detailed information to work with, potentially reducing accuracy. However, successful shift-left strategies navigate this trade-off through progressive refinement approaches. They combine fast approximate methods with selective high accuracy techniques and employ statistical analysis that quantifies uncertainty rather than ignoring it.
Real-World Impact
The benefits of shift-left methodologies extend beyond just finding bugs earlier. Designs developed with early performance prediction tend to be more robust, more power-efficient, and more manufacturable, qualities that ultimately determine product success in demanding applications like autonomous driving or next-generation communications.
Another critical driver for organisations adopting this approach is the financial aspect. In today's competitive environment, being first to market can mean the difference between category leadership and playing catch-up. When redesigns cost millions and delay product introductions by months, early verification delivers a clear return on investment. Companies implementing these methodologies consistently report reduced design cycles and fewer late-stage surprises that require expensive fixes.
Implementation Strategies
Organisations looking to implement shift-left methodologies for AMS design should consider several key strategies that work together to create a comprehensive approach. Investing in behavioural modelling capabilities forms the foundation, encompassing both tool infrastructure and team training. Engineers accustomed to detailed circuit design often need to develop new skills in abstractions and modelling techniques to fully leverage these approaches. This investment pairs naturally with developing a progressive verification methodology that defines appropriate analyses at each design stage, including clear guidelines for when to use fast approximations versus detailed simulations.
Establishing a knowledge base of reusable verification components and models accelerates this process while ensuring consistency across projects.
Critically, fostering collaboration between analogue and digital teams creates the cultural change necessary for success. The shift-left evolution isn't solely about technology but requires cross-functional teams with shared goals who can leverage each other's skillsets to identify and resolve issues at the earliest possible stage.
Building multiphysics expertise across design teams is increasingly essential. As circuit speeds and manufacturing technologies advance, design teams require access to a broader array of physical modelling tools to capture complex interactions. Having this expertise in-house enables organisations to address multiphysics challenges proactively rather than reactively.
Looking Ahead
The increasingly competitive landscape of chip design is forcing companies to adopt more proactive approaches to verification and validation. Getting designs right the first time isn't just preferable but is essential for survival in a market driven by constant innovation and compressed timelines.
As machine learning and AI techniques mature, we can expect to see even more powerful early prediction capabilities emerge. This will enable virtual prototyping of complete systems long before physical implementation, further compressing development timelines.
AI is proving particularly valuable for tackling the non-linear behaviours often seen in AMS circuits, which were previously intractable and required extensive manual intervention.
With innovation cycles expected to continue to accelerate, the ability to predict and prevent problems before they occur becomes a critical aspect of semiconductor development that organisations must master.
Those who successfully implement shift-left strategies transform verification from a bottleneck into a strategic advantage, positioning themselves at the forefront of an industry where anticipating challenges is as valuable as solving them.
Author details: Marc Swinnen, Senior Principal Product Marketing Manager, Ansys
