What you'll learn:
- What is Copilot and how does it use AI?
- The capabilities of Copilot.
- Applications and future functionality.
PCB design has always been a tricky endeavor, especially when the designs become more complex. The number of layers, traces, planes, and electronic components can easily lead to a host of problems during the design and testing phase, many of which can be easily overlooked.
Wrong trace geometry, misplacement of decoupling capacitors, starved thermals, and defective components are some common issues encountered during PCB design. Flux, the company behind the innovative browser-based design tool of the same name, recently released an AI-powered design assistant that can help prevent those PCB design issues and more.
The company's Flux design tool helps solve some of the challenges engineers encounter during the PCB design phase via cloud-based and community-shared design resources. These include templates, component resources, layout, and section reuse sub-layouts that can be dropped into certain sections without the need to start from scratch. Flux's Copilot raises the design bar further by integrating an AI-powered electrical-engineer design assistant to augment the browser design platform.
How Does Copilot Work?
Copilot, a custom-trained AI assistant that uses the large language model (LLM) residing inside user projects, can provide feedback, advice, and analysis of designs via a comment section on the design panel (Fig. 1). The tool uses a wealth of knowledge garnered from EE principles and other sources to provide feedback and analysis on users' schematic designs.
Moreover, it can provide part information taken from datasheets and even offer component alternates, helping to save on costs and part availability. Copilot leverages conversational AI to help users with part selection and compatibility via feedback and data analysis using those schematics and even performs electrical rule checks (ERCs), ensuring circuits will function as intended.
The initial release of Copilot focuses on hardware and circuit design-related tasks—trace routes, component placement, etc. It will include PCB layout tasks in future upgrades.
To that end, users can ask Copilot engineering questions in much the same fashion as using online collaboration software to communicate with other engineers through text messaging. Users can even place multiple chat threads in Flux that concern different topics related to their designs. So, in essence, users are able to “talk” and interact with Copilot like you would a human, which also could be used to track project development, as anyone can see the conversation when opening the schematic.
Benefits of Copilot
Copilot is like having access to an always-available senior EE that can answer questions and help users through the project design phase. This saves time and money and allows engineers to produce higher-quality hardware in a shorter time than using traditional CAD software. But this poses an interesting question: How to overcome the blank paper dilemma that can sometimes be encountered with producing an initial design?
In the above video, the engineer demonstrates how Copilot can generate new ideas and explore new design options by asking the AI to generate a list of components for a general use case or a minimum set of components needed to create an IC. AI then generates a list of the components needed and even provides a description of why those components were chosen.
Subsequently, the component list is transferred to another section called “design optimizations,” where it offers suggestions for cheaper components and design optimizations to make the IC more efficient. Once the components have been selected, Copilot can then analyze the schematic and help root out any potential errors that may be encountered with the design.
Still a Work in Progress
While the Flux Copilot AI assistant can help users through the design phase with virtually any piece of EE knowledge and BOMs for efficient design, it's still in its infancy and isn't 100% accurate in its initial launch (Fig. 2). As such, Flux states that "Copilot is to be viewed as a guide for the designer and not a substitute for the designer's judgment or expertise." It also should be noted that it only works for the IC side of PCB design, meaning users can't use it to create a PCB layout. However, future updates may include that feature.
Beyond its IC development/design capabilities, Flux states that Copilot also can be employed as an educational tool for both students and operational engineers. For example, a student may use the AI tool as a learning aid by asking it questions, such as what is the function of a component in the circuit. OEs can tap it for insight on productivity improvements by asking it for optimum replacement parts.
While Copilot's capabilities may seem light at launch, it also can be used for other applications unrelated to circuit design, such as food preparation. During an interview with Flux CEO and Co-founder Matthias Wagner, I asked if Copilot could perform other tasks outside the scope of IC design, including providing instructions on how to make brownies.
Amazingly enough, Wagner asked Copilot if it could do just that, and after a few seconds, managed to produce a great recipe, stating, "While my primary focus is on helping with electronic circuit design, I can provide a simple brownie recipe for you!"