Show HN: Stillwind – PCB part selection as constraint solving | Hacker News Reader

Show HN: Stillwind – PCB part selection as constraint solving

stillwind.ai

7 pointshannesfur5 comments

Hi HN,

We run an electronic parts search engine, and Stillwind Builder is our attempt to automate the selection step itself: it turns a project description into a PCB bill of materials.

At its core, it frames part selection as a constraint satisfaction problem, since your picks depend on each other — a motor driver can't draw more current than the power source is rated for, and every part's supply pin has to be within the voltage range of the rail feeding it. Once the constraints are set up, we evaluate combinations of parts from our database against them. If no combination works, the system adjusts the constraints and tries again. At the end it spits out a BOM.

You can try it directly at the link — the purple examples run with one click, or you can describe any project you like. We wrote up how it works in more detail here: https://stillwind.ai/blog/stillwind-system-builder

We'll be in the comments for the next few hours and happy to answer questions. If you'd want to use this on real projects, we're open to design partnerships: contact[@]stillwind.ai

Open articleView on HN

I assume the system of equations is generated semantically based on the request. Do you have a mechanism to ensure that there is no crucial equation missing or do you have tests on how reliable the constraints are in real applications? I don't know much about electrical components and would depend solely on the output accuracy/compatibility guarantee.

A BOM is cool, but then I still have to design the PCB, source the parts, and make it. Could you do this end-to-end, so when I'm satisfied with the result, it is manufactured and delivered straight to my door?

Basically, this is the end goal :) TBH, there are still some major hurdles to overcome to get to this point though. We are working on some parts of this pipeline already (e.g. schematic creation or distributor/EDA integration), but other pieces are still quite far away. Especially component placement and routing are still largely unsolved problems and are also beyond the intelligence of today's LLMs.

How big is the coverage of electrical components across all categories? Just to estimate the likelyhood of finding the right thing for unusual usecases

We have a focus on more complex ICs as we see them most relevant for this task. In that area we have ~95% of all components that are active (not EOL & recommended for new designs). But we try to have a little bit of every category to make most projects possible. While we expand our database, you can use the external search agent which uses the internet to track down parts that are missing.

Reply on news.ycombinator.com