Will we ever be replaced by AI and our own hardware?
Some signs are obvious like the twitchy pointer finger able to click through byzantine menus at a rate that would shame that teenage gamer that occupies the sleeping quarters halfway down the hall. The command and layer names fall into our conversation making half of what we say indecipherable to the layperson. (...or, should I say, the non-layout-person?)
Yes, that would be our name on the high score list on Tetris. Making things fit in a hurry is our line of work. Obviously, we’re pros at connecting the dots, staying in our lane while routing outside the box.
But this is no game. If enough of the little decisions we make in a day go the wrong way, product enablement may be delayed. Enough delay on the engineering samples can sway the customer to shop elsewhere. Taping out a board is small potatoes next to a chip but the consequences for failure are not so different. If the setting is a start-up, it could be an existential threat to the company. A large org will not hesitate to re-org when the seas get rough. Nobody wants to throw away money.
Image credit: Author
No pressure because there are no mistakes, right? A single fatal flaw at a critical juncture in a product development cycle can sink the whole effort. The last-second changes increase the odds of something irregular getting through to the product. Most of the time, a board can be fixed or at least used in a limited way. Still, the internal pressure on the PCB Designer is bound to manifest in these crucial moments. Rework eats profit; no profit, no business.
More and more, we have to rely on the software to retain and manage knowledge of the design requirements. Months go into the design work in some cases. Then, several months may go by before revisiting the project. Embedded knowledge is the best way forward. I find myself taking over designs where much is lost in translation from one CAD platform to another.
The first order of business is understanding what is there and how we can capture the particulars in the constraint manager. The three elements of a trace are width for impedance, length for timing and air-gap for isolation. Components come in minimum, nominal and maximum footprint sizes. All of the sanity checks are intended to one, take ownership of the design and two, pass that legacy on to a future me or someone else.
What if “someone else” is a machine? We love to deride our auto-routers and it doesn’t seem like anyone will come up with a plausible component placement algorithm. With the technology advancing as it is, can we still sit here and believe that a netlist in/board out black box couldn’t happen? Note that my frame of reference is the auto industry and the upheaval going through it with the advent of autonomy. Some serious computation and graphics machines are coming along that should have the horsepower to knock out its own motherboard.
The reason I don’t think it will happen is money. We’re just not a big enough pool of users to justify the investment in replacing us with Artificial Intelligence. CAD is big business but the global motor pool is something else entirely. The small-time makers and entrepreneurs will be doing CAD PCB Layout for the foreseeable future. I’m speculating that this would also hold true for medium-sized enterprises.
Image credit: Amazon
The data-driven FAANG conglomerates may set themselves apart by leveraging their data collection techniques to build an electronic PCB Designer but it would specialize in data centers or similar objectives that benefit the whole but are not for the individual or typical corporations.
No doubt that the one evergreen trend in this business is that things get smaller. We Earthlings will be able to adapt to these changes with course corrections on the fly. One of my younger connections recently asked me if there was a good book about PCB Layout. There might be but if the ink is dry, it’s out of date. The future is being invented in factories around the globe. The PCB vendors are chasing the chip vendors down in scale. We’re going along for the ride.
The devices on my boards right now go down to 0.35 mm pin-pitch. I’m used to that but another designer was asking for tips on 0.30 mm devices. Turn the page, here comes another node. As we continue to shrink the geometry, we continue to rewrite the code. The adaptability of the human element gives us a leg up on our machines despite all of our organic shortcomings.
In my younger days, I was fortunate to stumble into an IPO bound telecom firm. Our market was multi-carrier RF amplifiers for 3G cellular base stations; part of the gear that goes into the little hut at the bottom of the cellular tower.
Image credit: Everything RF
Our distinction was that we were going to automate the tuning of the amplifier by dealing with amplifier control in the digital realm. It used some complex technology to figure out how the amplifier was responding then compensate for drift over temperature to keep the amplifier locked on to the signals and working at peak efficiency. Yeah, science.
Right on the heels of our IPO came another company with the same market in mind but a different approach. They planned on hiring lots and lots of technicians to tune lots and lots of amplifiers and brute force their way to wealth. The one thing that they could do that we could not do was to respond to changes in the market landscape. They took a lot of market share from us in that way when the FCC added a new regulation. That sort of adaptability will keep us PCB Designers in business - at least until We all retire!