Cadence Design Systems has started bringing artificial intelligence (AI) into the fold on its flagship chip design suite to help designers build smaller, faster processors that consume less power and cut costs.
Now, the electronic design automation (EDA) giant is turning its eyes to the system level. It’s rolling out a new tool that applies AI to optimize the placement and routing of components on a printed circuit board (PCB).
Built into the Santa Clara, California-based company’s Allegro X platform for system-level design, the software —called Allegro X AI—can automatically create physical layouts of small to medium PCBs. Thus, it helps home in on layouts with equal or better quality compared with manually designed devices. Cadence said it enhances engineers’ productivity, using AI to reduce a process that previously took several days to a matter of hours.
“Automation akin to digital IC design is needed to bring about much-needed transformational change in the way PCBs are designed today,” said Saugat Sen, VP of product marketing at Cadence.
Like many other EDA tools currently on the market, the company said Allegro X AI leverages the scalability of compute in the cloud to help designers save time and resources.
Place, Route, Repeat
What’s wrong with the traditional PCB design cycle whereby AI-powered tools are now becoming necessary?
The placement and routing of a PCB is “a time-consuming, manually intensive, and serialized process, and [it] becomes even more of a time-to-market bottleneck as PCB complexity increases,” said Sen. “Current processes typically do not scale either with available compute resources or with complexity of design.”
Every circuit board starts as a high-level concept scrawled on a napkin or scribbled on a whiteboard at the office, where engineers figure out what they want the device to do and the parts it needs to accomplish that task.
One of the first steps in the process is selecting the logic, analog, power, and other chips and components that will populate the PCB, preferably those that are priced competitively and are in stock. Once component selection is done, they start plotting out the schematic, a conceptual diagram of the PCB, and all of the building blocks. They then review the blueprint to pinpoint any mistakes—just one could cause the device to malfunction in the future.
One of the last phases is physical layout, which involves the careful placement of hundreds, or sometimes thousands, of IC packages and other parts in a limited space and in a way that minimizes the amount of copper wire between them. Then Allegro X or other EDA tools are used to simulate the performance of the device and make sure it works when it comes to signal integrity (SI), power integrity (PI), or other metrics. The device must be designed to tolerate heat, cold, vibrations, and other harsh conditions that can make or break it.
It’s also key to keep in mind whether the circuit board can be easily manufactured (also known as “design for manufacturing” or DFM), assembled (DFA), and tested (DFT) at a cost that makes sense, said Cadence.
But everything on the circuit board rarely works the first time through. Mistakes are inevitable. As a result, companies usually must redo the circuit board multiple times before it’s production ready.
Other times, components that were available at the start of the design process are out of stock or delayed when it’s time to take a PCB to be manufactured. That can mean redesigning large portions of a design.
The number of possible configurations for a circuit board is vast. Traditionally, engineers had to rely on their expertise to manually place and route everything together in a way that meets the project’s goals.
AI at the Wheel
However, according to Cadence, business as usual isn’t working for many engineers these days. Modern PCBs are too complex, skilled engineers are too scarce, manufacturing costs are too high, and project deadlines are too tight.
With everything working against human designers nowadays, Cadence and a number of companies, including Celus and U.S. startup Jitx, are exploring the potential of AI software to ease the pain of PCB design.
Cadence said Allegro X AI unites more traditional types of computational functionality used in physical design with a wide range of AI technologies. But it’s unique in its product line with the “Generative” AI at its core.
This is the same underlying AI technology as ChatGPT, a system released by Microsoft-backed OpenAI that can take a text prompt and create content that’s in some cases indistinguishable from a human response.
Instead of placing and routing all components manually, engineers can tell Allegro X AI what they care about. It then generates a layout with everything they don’t care about already addressed. Cadence said Allegro X AI results in a reduction of 10X or more in turnaround time for the physical design of a PCB.
Users can adjust the physical, mechanical, and electrical constraints of the design and create variations of the PCB, opening the door to feasibility analysis in the early phases of a design project.
“For the PCB designer who produces layout, Allegro X AI becomes a new tool that multiplies their productivity by allowing them to complete more boards in a fixed amount of time as well as produce better boards by leveraging layout insights that previously would never have been considered,” said Sen.
Cloud-Powered EDA
One of the main advantages of Allegro X AI is that it can explore many more possible physical designs than engineers have time to these days due to ever-tighter deadlines. Engineering departments also are being pushed to the limit, given the difficulties of hiring hardware engineers lately.
The Allegro X AI tool can automatically take care of placement, metal pouring, and critical net routing. Equally as important, it integrates system design and analysis technologies through the Allegro X platform, which allows it to take into account the electrical and thermal constraints of the PCB. The Allegro X platform also helps make sure the PCB can be manufactured and tested. Furthermore, customers can use the system analysis and other features of Allegro X to further improve the output of the AI, said Sen.
The company plans to train the AI technology in the Allegro X platform internally on what works and what does not when it comes to PCB design over time. Customer designs will not be used in training.
Cadence isn’t trying to take human engineers out of the loop entirely with Allegro X AI. The company said it serves as more of a creative assistant that helps free up engineers to focus on innovation where it counts.
“The AI technology was also able to provide me with placement options that I had not considered [before],” said Allan Norgaard, a PCB designer at Velux, which makes remote-controlled windows, blinds, and shutters. “The technology dramatically decreases design time and will fundamentally change how we do designs.”
Other early customers of Cadence’s Allegro X AI tool include leading industrial equipment manufacturer Schneider Electric, which hopes to use it to significantly cut down its PCB development time.
“The hardware designer can make assessments on density and complexity—and adjust the electrical design to ensure a rapid and efficient completion of the design,” said Jean-Christophe Dejean, VP of product lifecycle management processes at Schneider Electric.