The question was put forth at Siemens’ EDA Tech Day in May: Which of the following can be replaced by AI?

• Input
• Schematic
• Footprints
• Placement
• Routing
• Deliverables (DfM, validation, etc.).

It was posed by a user who indicated that routing takes up about 30% of the time of a typical design spin. In classical Pareto thinking, that makes it the best target for process improvement.

Certainly, efforts are vigorously being made to develop intelligent automation for all the above and more. I'm calling it intelligent automation because 1) it’s a less scary term than artificial intelligence and 2) isn’t that what we are really after? And for its part, Siemens says that in the electronics design context today, automation without AI is superior to AI itself, while algorithmic automation is state-of-the-art.

But the rub is the amount of data is extraordinary, and few users, if any, want to share them.

Take land patterns and models, for instance: more than a dozen CAD primes and third-party vendors supply these, plus every company has all their internal data going back years or even decades. And fabricators and assemblers have all sorts of historical data on manufacturability and yield using those land patterns as well. But all that data is prized as corporate IP – even when purchased off-the-shelf from the vendor.

The issue, as I see it, is knowing what works with the specific fabricator and assembler, which is less a data availability issue than it is a data application issue.

Software vendors typically train their automation tools on what they perceive are standard design processes. Layout, schematics, analysis: the activity is modeled and then used to predict what an experienced designer would do next. (Insider news flash: Siemens has plans to release constraint prediction, which looks at previous designs and says, in effect, do you want to do the same thing?)

Yet we are in something of a stagnant phase when it comes to software automation. That’s not to say tools haven’t changed or improved: the development of 3-D models makes many CAD suites much more effective for visualizing the populated board inside the enclosure, among other things. But most would agree that if designers continue to hand-route their boards – and the show of hands at the Siemens event and at our own PCB Detroit conference in June indicated this is near-unanimous practice – then the routing functions are widely perceived to be rooted in 1980s technology.

We know that different design rules and models are limited to the design data that are available and shareable. Thus, if each design is fundamentally the same – substrate materials, conductive traces, vias, components – yet different in enough ways to make it unique, AI solutions will remain a long way away. If ever.

Here's the issue: What if AI, implemented into CAD, doesn’t do what we need it to do? What if we spend tens or hundreds of millions of dollars on trying to replicate – not replace, but replicate – the design function, to find that the concepts are effective only for the most basic of boards? Are we, as an industry, going to be OK with that? What could we have been working on instead?

Let’s go back and focus on the goals: faster, more reliable product development. Vendors, let’s focus less on the terminology and more on how to position or upgrade the tools to accomplish that. And users, I get that we are intentionally late adopters in large part because we don’t want to be the proverbial canaries in the coal mines. Let someone else be the lucky one to find the bugs in the software. In the interest of moving forward, however, maybe we raise our hands and provide real-use feedback so we can realize the productivity gains we’ve long sought.

Next-gen. We truly enjoyed our week in the Motor City during our premiere of PCB Detroit last month. We are especially grateful to our exhibitors, especially Cofactr, which sponsored our networking event, and Wayne State University, which hosted the conference.

One of the aspects that I came away with was how eager the local college students were in learning about the PCB industry. They showed up all day, listened intently even though the content was usually aimed at more-experienced designers and engineers, and stuck around and chatted (read: networked) after the classes ended each day.

This runs counter to the perception that college students are software-focused and not necessarily interested in hardware, other than buying the latest iPhone. Several of those present were also enrolled in the PCEA Certified Printed Circuit Designer course we taught at Wayne State last spring, and have now passed the CPCD exam. And there was widespread lobbying – from academia and industry – for PCB Detroit to return next year.

All in all, we feel we learned as much as the attendees. The industry often talks about how we must attract the next generation. It’s great to be part of an organization that is working on doing just that.

Mike Buetow is president of PCEA (pcea.net); This email address is being protected from spambots. You need JavaScript enabled to view it..