Flux Upgrade Graduates AI Assistant to AI Circuit Co-Designer
I gave Copilot's latest upgrade a trial run to see if artificial intelligence could take a circuit from idea to PCB layout. Here's what I found.
Electronic design is poised to enter a new era of automation with an upgraded tool from Flux. The AI-powered EDA company has announced a significant upgrade to its browser-based EDA CAD tool. The newly-released version of Copilot is the second major upgrade since the platform's introduction in 2023 and gives it AI-driven hardware design capability.
Flux AI-assisted EDA CAD tool in action.
Copilot As a Co-Designer
Prior to this release, Copilot helped designers by delivering AI-driven design advice. With the new incarnation, it can take on the role of co-designer, interpreting a designer’s idea, wiring up the schematic, and producing the PCB layout and bill of materials (BOM).
Flux combines schematic capture, PCB layout, and BOM management in one unified package. The browser- and cloud-based platform can assist with a wide variety of design projects. Flux Copilot takes chat-like conversational prompts and works with a designer to turn those thoughts into a schematic design. It keeps track of the bill of materials as the design progresses and closes the design cycle with a human-like AI PCB routing.
Flux Copilot conversational design flow.
Copilot can also be used to modify or complete hand-built schematics. As an assistant, it is ready when called upon but does not get in the way when not needed. For example, an engineer might select a set of integrated circuits needed to design a wireless sensor. The engineer can place the ICs, as with any CAD tool. Then, the designer instructs Copilot to add bypass capacitors and power rails. Copilot can further follow instructions to connect ports and communications pins.
Making Hardware Not So Hard
In designing the new version of Copilot, Flux studied artificial intelligence tools used for writing software code. They found that AI could write complex software quickly, but they did not find anything comparable for hardware. The result is the new release of Flux Copilot, inspired by the ease with which AI generates software code.
The new AI Copilot, trained in design languages and best practices, can access datasheets and the CAD library. From this knowledge base, it intelligently creates a set of block diagrams for a circuit, chooses or suggests components, and, block by block, designs the circuit. In the video below, Flux explains and describes the functionality of Flux Copilot.
Real-World, AI-Assisted Design Process
As an experienced microcontroller designer, I decided to put Copilot to the test. I started the conversational design process much like I would with other AI tools. Copilot takes an iterative approach to that conversation. For example, I gave Copilot the following prompt:
The initial conversation.
I then suggested an NXP Kinetis Arm M0+ microcontroller that I have used in similar designs. Copilot found the NXP KL27 and three other options for me to consider.
Copilot offered an architecture with three different blocks: a LiPoly power supply and battery management block, a processing block, and a motor driver block. With more Q&A, Flux suggested an MC33932 motor driver, also from NXP, and a BQ24075RGTT battery management chip.
The three-design block approach closely follows how I would create the same circuit, and the driver chip is the exact part I had in mind. Flux Copilot continued the Q&A process to create schematic diagrams for each block and wired the blocks together. The process helped me achieve my design goals without the mundane work that comes between the steps where I add value.
During the process, the tool created the bill of materials (BOM) and prepared for the layout phase. Copilot has a large component library at its disposal and will suggest options based on what it has learned about the project so far. It made component suggestions, such as bypass capacitors or pullup/pulldown resistors, based on specifications it pulled from IC datasheets.
At each step, Copilot asked me, as the designer, to make the final decision, allowing substitutions or modifications along the way. Copilot created the BOM as it went through the process. While engineers must double-check the final BOM, they are relieved of most of the time-consuming aspects of BOM management.
A New Paradigm for Hardware Design
Copilot can significantly lower the bar for beginning product designers. By following best practices, assisting with component design, and ensuring that wires go where wires need to go, Copilot speeds up the development of more reliable products. Beginners can follow along and assist with the design process as they develop their personal design skills. Flux does caution, however, that, like all automated systems, a human does need to double check. While AI is a helpful tool, it is not yet advanced enough to operate without supervision.
Experienced engineers can maintain control over complex designs, such as those with FPGAs, big CPUs, or specialized signal requirements. However, even in such applications, there is often a lot of activity that is more about time than skill. As an experienced designer myself, I felt that Flux Copilot worked with my thought process. It automated difficult, time-consuming tasks while not diminishing my creative contributions.
Flux provides an ample set of tutorials, reference materials, sample design projects, and a Slack community for additional support for beginners and experienced designers alike.
All images used courtesy of Flux.
Still, need help? Contact Us: [email protected]
Need a PCB or PCBA quote? Quote now