Like almost everything these days, the art of PCB design has its underpinnings in data. We’re data driven – and swamped in a sea of information. Some of my favorite info is found on the literature offered up by the printed circuit fabricators. We get what’s called a technology roadmap. While maps usually refer to geography, these roadmaps show the way into the future.
Figure 1. Image Credit: MEI - This is a current technology snapshot broken down by relative cost. A roadmap would include what to expect in the future.
Open one up it looks like a spreadsheet with physical attributes on the y axis and calendar years filling up the rows. The numbers are sorted into columns with the values decreasing over time. The left column of data provides today’s mainstream values for important characteristics of the circuit board.
Board designers are going to take on more and more of the load once carried by the chip team and the substrate.
We focus on minimum air gaps and line widths for inner and outer layers. Machining tolerances and dielectric thickness numbers and so on are at their highest in the first column. Mass production boards should have technology that is aligned with standard processes.
When Google launched the first Chromecast dongle, we had Flextronics as the ODM. They owned Multek back then and Multek owned five fab shops. Each one had different equipment and cost structure. In order to use their bottom tier factory, we couldn’t stack micro-vias so I had to spin the board to stair step the 1-2 and 2-3 vias. The more price conscious you are, the more you go for the older factories.
Beyond the standard service, the next column lays out the shop’s best practices as they lean into fine lines and tight pitches. Tooling charges and lead times increase while the number of qualified vendors diminishes. This represents their current prototyping capabilities.
Of course, the envelope is all relative as each fabricator has a unique factory or collection of factories. This space is the home of the smartphone. The players in this market are already designing the phones for 2025 and beyond with the expectation that the technology in their prototypes will be mainstream by the time of the roll-out.
The low-volume, quick-turn fabricators are the ones pushing the boundaries between the emerging technology and the next generation. The third column of the technology roadmap is the wish list of values for the future. Marketing hype replaces actual results in this forward looking vision.
Figure 2. Image Credit: Author - Laser cut board edges with plating closer to the edge than thought possible could change the way PCBs are connected together.
The leading edge features MSAP as in Modified Semi-Additive Process where the copper is sputtered onto certain layers of the stack-up. Let’s say you have an HDI (high density interconnect) board with a 2-6-2 stack-up. The board starts with a 6 layer core and has two more lamination stages to wind up with a 10-layer board. The layers where MSAP can be applied are 4,5,6 and 7. The inner layers of the core are the ones where this is doable. Other layers have a plate-up operation for the micro-vias and that’s not an option with MSAP.
Of course, it is rare to find a fab shop that is up to this level of technology right now but give it time and it should become available wherever you build your little streaming devices. All along, the driver for new PCB technology has been our partners in the component world. The pitch of the first BGA device was measured in miles, or at least fractions of an inch. So 1.27 mm became 1.0 and then, 0.8 to 0.65, 0.50, 0.40, 0.35, and now 0.30 mm pitch BGAs in the market.
Figure 3. Image credit: Author - An insane Rubik's cube may be the forebearer of tomorrow’s chip packaging technology.
By condensing the circuit from perimeter pins of a quad flat pack (QFP) to an ever decreasing geometry on a grid, the chip companies are handing off more of the workload to the printed circuit board. The marketing team calls it Wafer Level Chip Scale Packaging or WLCSP for short. In practice, the chip and the substrate are essentially the same size. There are flip-chip pads on the die with matching pads on the substrate. The substrate has a via-in pad that passes the signal right down to the other side just as though it was a chip on board situation.
In this way, the PCB designer is doing the traditional task of the substrate where there were traces that spread out from the die to a more reasonable pin pitch that supported through-hole via technology. The chip companies could still be building them that way but all of our gadgets would be larger and more power hungry. The race to the bottom isn’t going to relax as long as the Marketing team has any say in the matter. Board designers are going to take on more and more of the load once carried by the chip team and the substrate.
The trends will continue with multiple passive components in the spirit of the old resistor packs. I can imagine capacitors ganged up around integrated circuits while ESD diodes are clustered around connectors in monolithic packaging. This would reduce the bill of material and process time in assembly. The trade-off being that a discrete package can be replaced more economically than a gang package.
On a software front, I think we continue to see incremental improvements in design verification being handed down to the PCB community from the chip development ecosystem. The ability to lock in an IPC class or even a specific vendor’s mainstream capability should help us tailor the data for high confidence in the initial iteration. This is happening one vendor at a time. (Dear Cadence, It sure would be nice to press a key and get a heat map of the board.)
The existential question for the layout engineer is “Will I be replaced by a machine?” My guess is that the simplest layouts could be thrown into the machine with usable results. The vast majority would get a more streamlined user interface but still craft the board. Out on the cutting edge, designers will be granted greater assistance from the machine learning component of AI.
Maybe that line between the human and the machine creeps down but for now, most R&D is done in secret. Machine learning counts on existing models while being part of something like ChatGPT has the potential to expose intellectual property. With this in mind, I can imagine walled gardens of data held close to the vest by companies and institutions that can accumulate and leverage that much relevant information. Even with all of that, there’s still going to be someone in the driver’s seat.
The job pool that seems to be shrinking is that of the librarian which has become more of a service as time passes. The growth as far as the deluge of offers I’m getting is in the start-up sector doing on-site or hybrid work. The jobs range from simple flexes to career boards that will consume you for months. The board designer’s job will evolve over time but it’s here to stay.