Flux Adds AI Design Assistant to Its Browser-based PCB Design Tool
Flux is launching an integrated AI-powered “chatbot-like” design assistant that’s integrated into its web-based online PCB design tool.
With ChatGPT in the forefront of discussion these days, it’s no surprise that the engineering world is finding ways to leverage generative AI technology. Along just those lines, today Flux launched what it’s calling the “industry’s first AI-powered design assistant integrated into a PCB design tool.”
Flux Copilot is an AI-powered design assistant integrated into the Flux browser-based PCB design tool.
In this article, we explain the need that Flux Copilot addresses, we explore how it works, and we share thoughts from our interview with Matthias Wagner, Flux CEO and co-founder and Kerry Chayka, product expert and hardware engineer at Flux.
A Design Assistant with EE Knowledge
As Wagner explains, Flux Copilot is a custom-trained large language model (LLM) AI that basically “lives” inside the “Comments” panel of the Flux tool. It uses its deep wealth of knowledge on electrical engineering principles to do a variety of tasks.
This includes selecting parts and verifying their compatibility, giving feedback on schematic designs, and doing design analysis. Flux Copilot can also assist you in evaluating component alternates, do “cost-down” analysis of designs, and perform electrical rule checks (ERCs). While the initial release of Flux Copilot focuses on these circuit design related tasks, Flux plans to include PCB layout kinds of tasks in future upgrades of Copilot.
You can ask Copilot engineering questions right in Flux brower’s chat, the same way you might ask an engineer that was on-line collaborating with you. You can also place multiple chat threads in Flux on different topics.
You can ask Flux Copilot questions about your schematic design, and it answers right in the Flux’s chat panel. (Click image to enlarge)
“We’ve brought Copilot into these chats,” says Wagner. “That means that you can talk to it like a human and interact with it like a human.”
“It [Copilot] has the full awareness of the schematic and where things are and how they're connected, what the part numbers are, and what the pins are—all that kind of stuff.”
This also has benefits for tracking the development process. “If somebody else were to open this project they too can see all that conversation,” says Wagner. “They too can engage here with the AI chatbot and ask follow up questions, you know, or Iterate on the solution with the chatbot—maybe add constraints, remove constraints, and so on.”
Like an Always-available Senior Engineer
As Wagner points out, Copilot in many ways fills the role of an experienced engineer. “In some organizations you have the luxury of having very senior engineers,” he says. “But the ratio of them is typically one to a hundred or so. That means it’s difficult to get their time. Even then, it’s hard to provide the whole context of your project to them and there can be some latency there.”
“What we can do here with Co-pilot is enable you to help yourself now,” says Wagner. “Sure, somebody senior can definitely review the design later, but you can leverage the Co-pilot’s knowledge in the moment. Also, what’s great here is that this lets you document the full thought process. Nobody has to just reverse engineer that from looking at the schematic.”
Overcoming the Blank Sheet of Paper Problem
In our interview with Flux, Chayka provided a demo of Flux Copilot in action. In particular, he showed an example of using it to help get a project started. “One thing that can be the most difficult early in a design is staring at this blank canvas,” he says. “How do I turn my dream for what a project is into an actual project?”
In his demo, Chayka told Copilot he wants a board design with a microcontroller, and then he wants it to have an analog input that works up to 600 kHz, and he wants to know what support circuitry it needs.
Starting from scratch, in the demo Chayka asked Flux Copilot to create a schematic of a board with an MCU with an analog input that works up to 600 kHz. (Click image to enlarge)
Copilot came back with a recommendation of a specific STMicroelectronics (ST) 32-bit MCU that embeds a 5 Msample/s, 12-bit ADC. Copilot drew the schematic of the circuit. “Copilot then told me I need a 3.3 V voltage regulator and maybe a crystal oscillator,” says Chayka. “It then tells me about the ‘reset’ and ‘boot’ modes.”
A lot of this kind of information, he says, is something you’d need to look up and go learn about, which all takes time. Copilot gets all that information for you.
“It even gives me the values I need for the different schematic components like the reset pull-up. That’s one of the kickers for me.”
More Capabilities Plan for Copilot in Future
As mentioned earlier, this initial release of Copilot focuses on the circuit design side of PCB design. We asked Wagner what future capabilities they plan to add in later upgrades. “We want to expand to PCB layout support,” he says. “Copilot could perhaps make change proposals itself, like a revised bill-of-materials (BOM), for instance. Then, we want to be able to have Copilot analyze those bill-of-materials (BOM) replacements, for example, and see if those result in PCB netlist changes.”
Being able to tweak a PCB design is another task that Copilot could help with. “Let’s say you wanted to move the two capacitors both 20 millimeters to the left,” says Wagner. “Can you do that without retracing PCB connections? Those kinds of dialogue-driven conversations about your PCB design is something Copilot would be good at.”
All About Circuits readers should mark their calendars for next Tuesday, May 2, because we will be airing our next Moore's Lobby podcast. The guest next Tuesday will be Matthias Wagner, Flux CEO and co-founder who we spoke to for this article.
All images used courtesy of Flux