Estimated reading time: 6 minutes

Beyond Design: The Metamorphosis of the PCB Router

The traditional PCB design process is often time-consuming and labor-intensive. Routing a complex PCB layout can consume up to 30% of a designer’s time, and addressing this issue is not straightforward. We have all encountered this scenario: You spend hours setting the constraints and finally hit the Go button, only to be surprised by the lack of visual appeal and the obvious flaws in the result.

History shows a prevalent skepticism among PCB designers, many of whom prefer interactive routing—the setup process often takes longer than manual routing the design. However, a modern autorouter offers a blend of automation and interactive control, allowing users to engage with the system when they choose, thus providing a balance between independence and user intervention. Unfortunately, many designers don’t use their router’s full capabilities.

Over the past 30 years, PCB routers have advanced significantly, evolving from basic dot-to-dot connectors to complex routing systems. Nevertheless, there is still ample opportunity for enhancement. In this month's column, I will delve into the latest advancements in artificial intelligence technology, particularly focusing on its implications for routing technology.

Modern autorouters depend significantly on configuration and setup, utilizing Boolean algorithms, which means their effectiveness is directly tied to the skill of the person configuring them. Fortunately, this configuration can be saved and applied to future designs, saving some time. PCB routers have gone through many different stages of development over the years, from third-party applications that were difficult to learn, use, and interface with, to a cohesive, cloud-based, layout/router environment. IC and PCB routing applications have used many of the same algorithms, with shape-based, push and shove, and rip-up and retry being the most effective.

However, with the advent of reinforcement machine learning, AI can be trained on an extensive library of professional PCB layouts, allowing it to bypass traditional setup processes. This enables the AI to leverage its training to produce high-quality products, effectively reusing constraints and topologies from previous designs.

PCB designers typically invest considerable time in manual tasks, such as component placement and signal routing. However, with the advent of generative AI, these processes can be automated, allowing for faster iterations and greater design exploration during the early phases. For instance, Allegro X AI claims to adhere to constraints related to wire lengths, signal integrity, and power distribution. However, one might question the necessity of analyzing wire lengths at all, considering that propagation delay is a more critical issue rather than length itself.

The system automates component placement and the routing of critical nets, ensuring that signal integrity is maintained while reducing the time required for routing. The process of copper pouring is also automated, facilitating the creation of ground and power pours and planes, which has traditionally been a time-consuming task. Cadence has been developing place-and-route (P&R) tools for IC synthesis for decades and has now adapted the technology for PCB P&R. Shorter interconnects and reduced crossovers are essential for both chip and PCB layout, but critical routing incorporating signal integrity and flight time requirements is of greater importance for the PCB.

Zuken has also introduced AI technology known as AIPR (Intelligent Place and Route), which enhances the CR-8000 platform through a three-stage rollout. The initial phase, termed the Basic Brain, significantly improves user experience by employing the Smart Autorouter to optimize design routing based on learned methodologies and strategies. In the subsequent stage, Zuken’s Dynamic Brain will leverage insights from newly developed PCB designs, incorporating historical design examples into its AI algorithms. This fusion of customer best practices and AI-driven insights promises to accelerate design iterations and substantially boost overall productivity, all within the CR-8000 framework. The final stage, the Autonomous Brain, will possess the ability to self-improve with each project, ushering in a new era of AI-driven innovation.

Siemens has made significant advancements with the introduction of Process Prediction, a key component of their Modern UX Solution. This innovative feature is integrated into Xpedition PCB, Constraints Manager, and HyperLynx Analysis tools. As users design their products, the prediction model is continuously trained in real time, enabling it to make instant predictions. With each new command, the model learns the sequence of a specific user, adapting to their unique behavior. It can accommodate multiple learned sub-processes and seamlessly switch between these paths (Figure 1).

The process begins with Xpedition PCB, where the layout is created, and then constraints are established, including stackup definitions and design rules. Subsequently, it returns to the layout for editing. As the layout is routed, critical constraints are fed into HyperLynx for signal integrity analysis to ensure quality. This process utilizes the DDR batch wizard to conduct timing simulations. In the event of failures, the layout is revisited for fine-tuning. This cycle continues until all requirements are satisfied, completing the entire design flow from start to finish (Figure 2).

The target demographic encompasses inexperienced designers that can leverage pre-trained seed models. Mid-level designers have the option to either train their own models or utilize those created by experienced designers, to enhance process efficiency. Furthermore, expert designers can impart their knowledge of intricate workflows and standardized design processes through these seed models.

At the lower end of the market, we find emerging platforms such as DeepPCB and Flux Copilot. DeepPCB is an end-to-end, fully autonomous pay-as-you-go PCB AI tool. DeepPCB supports all PCB EDA tools compatible with Specctra, including OrCAD, Allegro, PADS, Zuken CR-3000/CR-5000, Altium, EAGLE, KiCAD, EasyEDA, etc. For best results with DeepPCB, manually route sensitive parts and unsupported nets, then let DeepPCB handle the rest while protecting existing routes. The standard version of DeepPCB allows you to design layouts of up to eight layers and 1,200 connections and takes a few hours to complete a design. This may be suitable for KICAD and Eagle users but appears to represent a regression in technology since the ability to autoroute non-critical nets has been around for decades and completes in a matter of seconds.

Flux Copilot, another entry into the AI PCB design realm, is suited to hobbyists who design simple microcontroller projects. It can help with component selection, debug when you are stuck and suggest and create schematic connections for you. Flux Copilot is a direct competitor to KiCAD and EasyEDA. Also, KiCAD Guider provides AI-driven support for KiCAD, delivering design recommendations and detecting errors to enhance the designer’s experience.

The evolution of PCB design has progressed from manual routing to highly advanced AI-driven autorouting solutions, significantly reducing the time and labor required in the process. While traditional autorouters depend heavily on setup and user expertise, modern AI-based tools leverage machine learning to streamline and optimize routing while preserving signal integrity.

As AI integration progresses, PCB designers will have more opportunities to enhance workflow, reduce repetitive tasks, and focus on innovation, ultimately transforming PCB design into a more intuitive, intelligent, and seamless experience.

Key Points

• History shows a prevalent skepticism among PCB designers, many of whom prefer interactive routing.
• Modern autorouters depend significantly on configuration and setup, utilizing Boolean algorithms. This means their effectiveness is directly tied to the skill of the person configuring them.
• With the advent of reinforcement machine learning, AI can be trained on an extensive library of professional PCB layouts, allowing it to bypass traditional setup processes.
• Allegro X AI adheres to constraints related to wire lengths, signal integrity, and power distribution.
• Zuken has also introduced AI technology known as AIPR (Autonomous Intelligent Place and Route).
• Siemens has made significant advancements with the introduction of Process Prediction, a key component of their Modern UX Solution.

Resources

• Beyond Design: Integrating AI into the PCB Design Flow, by Barry Olney and Charles Pfeil
• Cadence, Zuken, Siemens EDA, DeepPCB, Flux Copilot and KiCAD literature

This column originally appeared in the June 2025 issue of Design007 Magazine.