The first thing you probably asked yourself when learning how to lay out PCBs was “can’t the computer do this?” which inevitably led to the phrase “never trust the autorouter!”. Even if it hooks up a few traces the result will probably be strange to human eyes; not a design you’d want to use.
But what if the autorouter was better? What if it was so far removed from the autorouter you know that it was something else? That’s the technology that JITX provides. JITX is a company that has developed new tools that can translate a coarse textual specification of a board to KiCAD outputs autonomously.
How do you use JITX? At this point the company provides a front end to their tools; you use their website contact form to talk to a human (we assume) about what you want to make and how. But watching their demo videos (see the bottom of this post) gives a hint about how the tooling actually works. In brief; it takes a specification in a domain specific language that describes the components to use, then compiles (synthesizes?) that into KiCAD files that can be sent to fab.
Removing the Human from the Equation
What level of abstraction does JITX work at? If it was an autorouter you’d select components and do some floor planning to place components before hitting the go button and getting a coffee, maybe even lay down some of the more complex tracks. That is to say, there is no abstraction at all. JITX can operate at a much higher level. You can specify which parts go exactly where and use it as an autorouter but where the tool shines is when the human specifies less, not more. The user can plug in specific components, board outlines and the like. But they can also just say “micro USB connector” and “Teensy 3.6” and JITX will figure out what connects where, place components, lay out tracks, and crop to a board outline.
But it can go even higher level than that. [Duncan] the founder and CEO told IEEE that you can “request a board with BLE and a microphone” and the tool can do everything else. It will select parts, figure out support infrastructure like power supplies, place the components, route the board, and emit design files. That’s damn near the entire design from napkin to product, which incidentally is a service JITX offers.
Control Freak v Push Button Designs
So the next question is who is this for? We have electrical engineers who may feel somewhat displaced by a technology that replaces large parts of their workflow. Even if they used it, do they effectively become programmers? Or HDL-authors (nowadays engineers who write HDL would be working on FPGAs or similar)? One could imagine an EE using a tool like JITX to augment their process by designing the totally bone-headed parts of a project which require no creativity. We’re not sure JITX would agree, but the not-groundbreaking-design market seems pretty large; enormous in fact. Especially when you consider that it includes entire products that are unsexy to an engineer but still need to be created. It might even include empowering people and organizations who would not have considered making hardware before.
Anyway, what does that market look like? Let’s consider a few examples:
- There’s often a need to make little bits of connective tissue in the form of simple boards with funky connectors to adapt between arbitrary connectors. A tool that could take, for instance, DigiKey part numbers and excrete design files for production could be a nice streamline. The end game here might be getting the board fabbed and assembled too, so the process would be completely turn key. Though such a product would need to be extremely fast or it might only be useful for nasty fine pitch connectors which are hard to prototype by hand.
- The firmware engineer who needs a test board made. Some firmware engineers do the electrical CAD to make a board themselves, but early prototypes are usually constructed by bolting together development boards. This is a great solution! But if a tool existed that effectively did the same thing but output a PCB instead we could imagine it finding traction.
- The largest and most terrifying market might be the Generic IoT Device. How many times does an EE really need to design something with a modularized nRF52, pick-your-favorite-sensor, and a coin cell? If JITX could generate form-factor designs with common “generic” devices like this it would probably be a huge leg up for a certain class of simple product.
Looking at theJITX website, they actually offer all of these in a pretty interesting fixed pricing model. Interconnect boards are “Adapters” and are a set at $499 a design given certain constraints. They also offer to design entire PCBAs below a certain complexity in 24 hours and at a constant $3,000. It sounds like a napkin sketch is all that’s required to make a board.
We’re pretty excited about what the future holds for more intelligent design tools! Tell us what you think you might use a tool like JITX for (or not!) in the comments.
Check out this demo video which includes a literal napkin sketch. It includes some description of the DSL used to specify the board (an expanse of FR4 to connect a bunch of dev boards together). They move on to specifying the parts instead of breakouts and use the tool to redesign the board at that granularity.