So I have spent far too many hours on this problem and gave up.
In terms of Kicad, the problems start with the file format. KiCAD uses a redo of some old 1970s lisp S-Expression format. It's a delimiter swamp with positional/sequencial dependencies.
The second is the hardest one to get around. A schematic is a VISUAL communication medium. The LLM has no eyes and very little concept of "spatial relationships". It doesn't understand what "clean layout" looks like. It' s never seen one and has no "attention awareness" of it.
It does better with "web layouts" as the markup and languages being used there are far more explicit and far more commonly "out there". Those markups and languages have elvoled to be human editable from the beginning. It still doesn't see your webpage, if it looks good it's likely just that it picked a very basic "tabular" layout and grids like 90% of websites.
You can get it to do a netlist. You can get it to read/write/edit the netlist and even make an attempt at a BOM. You will need to review it though.
As an example as to why you need to review it, or take other techniques.... try one of these techniques yourself.
Open an empty chat and ask claude about decoupling capacitors on ICs. Tell it you were thinking that "3 decades in paralell" sounds like your best approach. Read the response.
Open a new empty chat context and ask claude exactly the same thing, but suggest your best approach is a single 1uF on the ICs. Read the response.
Both will be confirmatory, affirming your strategy either one. It would seem that both are valid answer. That is because they are both valid answers with almost equal weighting out there on formuns and in schematic/netlist examples.
It highlights something though. Your prompt can be seen as an "observer". The LLMs response will be from that view point. If there is someone obvious on the "back side" of the idea it might be completely blind to it. So, reprompting it with a flip on something key, like "undo a previous design decision" and ask it to present something different, or you approach the original problem from a new direction, maybe if you were very "pro" the idea, as it's yours, try being negative about the idea and see if it gives you a different insight.
So applied to electronics, if you want something that "you" would have produced yourself you have two options. 1. Let it decide and pick which every style was closest in it's context or 2. Go through every single detail with it and ask / chat with it to direct it to what you want.
As someone said... at a point it becomes faster to just do it yourself.
Where I did get some use out of it for PCB design, was converting a verilog file to 74HCxx logic gates. There I was driving the schematic. I had claude grepping the yosys synthesiser output and helping me go through the verilog bit by bit converting different sections to schematic sections. The yosys idea was mine. I was looking at a sequence with multiple inverters and could see it might have few chips if it was completely "inverted". After looking at a few Claude outputs and not being convinced it was seeing the same thing, I asked it "What did yosys do?" and off it went. Brought back the LUTs and whatnot yosys has placed for the circuit, translated those back to 74HC discretes. This helped.
Schematics. My approach to fix this is a bit too much for me to do myself.
OpenSchematic format. Split into several layers. "functional" - the netlist basically. What the PCB editor and electrical constraints see. It may include the BOM so that a simulation can run with just the "functional" input. Then the visual layer. This is where components get "symbols" and "nets" get "wires" etc. Things get "places" and orientations etc. Finally is the documentation layer. From the normal stuff, to datasheets for components to labels, to tooltips to pop outs.
The functional layer is intented to be easily used by human and machines. It is technically all you need to make the circuit.... just a bit of visual layout , "just" LOL.
To make this even slightly viable... it would need to be created and trans-parsers written for at very least KICAD. However even there changes would be required to KiCAD beyond just import/export. Not a one man band project, unless I only want one project for the next 5 years.
The other problem is acceptance. There have been other such endeavours and they just failed into obscurity. Most people just pick one EDA application and assume they are stuck with it.
I have spent quite a bit of time myself on trying to solve this.
I took on another approach, start with specifying what to build, if you specify it well, Claude is pretty good at building a behavioural Spice model for it, which can be simulated. This is pretty useful, but is mostly mathematical/function model and can not be turned into a schematics yet. Next step is to break it down to blocks, Claude is also pretty good at this and can separate them into behavioural models per block. Next step is to build a schematic/netlist per block. Claude can also do this pretty well. Now you pretty much have a schematics of your whole circuit and you can simulate the parts and the whole.
What you have is a schematics in text format, pretty well designed.
What AI is really bad at is the visual, both in Spice and eg Kicad. You can actually get a netlist with all components and nets, all in a big mess, but correctly connected, and you will have to organize it visually.
And human will have to know what to check and be involved so it it far from automated in the way coding is.
What I realized building this, is that the input is really important, so instead I built a tool to go from idea to design, focusing on specifying requirements, v&v and then down to architecture and the output is what an EE would need to start building. This ended up being a very useful, especially when working with a customer specifying a project, which may change along the way, keeping track and traceability.
I really like what I built, and would like feedback from users. The tool is web based, completely free, I would be thankful for any feedback.
Link is in my profile.
Home
Help
Search
Login
Register
