Applying experience to design to build better boards.
Some things change, some stay the same. The Pretenders were almost certainly not thinking about making PCBs, but it certainly applies to this industry.
As my musical musing may suggest, I’ve been here a long time – just over 50 years, in fact. (Chrissie Hynde was still at art school when I started.) I’ve seen many changes. I even caused a few! (More on that later.) But some things stay the same, and one of these is the disconnect between PCB designers and the fabricators who do the making.
The situation has certainly improved, but dialogue between the two groups remains rare, and I still find myself delivering similar presentations on this issue to OEM engineering teams. Panelization is a prime example. Even now, too few engineers understand how to design a PCB to be produced as part of a panel, and not in isolation. Clearances, tolerances and feature sizes must be determined to account for effects such as stretch or movement of the PCB layers during fabrication.
Working with the PCB supplier helps to fully understand what is required, but often this is not done, certainly not with any regularity. Additionally, the PCB manufacturer should work with the designer to fully understand the “functionality” requirements that may override those of achieving the ultimate yield.
Full understanding of CAD and CAM is essential when designing and manufacturing a great PCB, but most people rely on specs instead of knowledge. I remember struggling to get approval to invest in a CAM system for one design office I worked in. I only wanted to be sure aspects like minimum gaps and trace sizes had been handled properly, but it raised a few eyebrows. We can rely too heavily on tools like these. I like to emphasize the “A” in CAD stands for Aided, and the real design work happens in the engineer’s mind. It’s worth remembering, not so much for fear machines might take over, but because using our brains and our experience is the way to get better boards.
What’s changed in 50 years? During my time in the industry, I’ve worked with large OEMs, consulting practices, and PCB manufacturers all over the world, and on numerous projects like early personal computers such as the BBC Micro in the UK and huge telecom OEMs at the other extreme. “Function First” has become one of my slogans. Another is: We can design with tolerances and clearances that permit fantastic yield, or a great price, but it means nothing if the boards don’t perform in the field. I’ve seen catastrophic failures caused by designing to meet a cost target, for example, by increasing pad or trace clearances, without considering the effects on the ground path.
This brings me to one of the changes I’m happy to take responsibility for. I first presented Max Copper, a strategy for ground-plane connection and PCB design to minimize noise, in the late 1970s. Taking advantage of the fact that the “star point” is the quietest point on the board, Max Copper design manages connection distances and loop areas to make the entire board into a star point from a noise perspective. The first commercial product to adopt the concept was the IBM PS2-30 PC of the mid 1980s. Today, so I’m told, more than 85% of boards in the world are designed this way. At the start, we were only dealing with 5V and ground, whereas today there are many more power domains, so the underlying principles of Max Copper are even more important.
I joined Ventec about four years ago as director of OEM technology. In this role in the laminates business, I have become heavily involved with the electro-mechanical aspects of PCB design, with a major focus on z-axis geometry. As I mentioned, markets are demanding miniaturization, which adds pressure to squeeze board sizes in every axis. Thinner laminates hold the key to reducing z-height, which can permit more room inside a rack for extra boards to be fitted, or to add extra layers that potentially permit higher pin-count ICs for greater feature integration or reduce x- and y-axis measurements.
Value analysis. All this speaks to my notion of value analysis, which highlights how an astute PCB design can hold the key to giving customers a bigger return from each cubic centimeter of space – whether this is expensive real-estate in a telecom office or data center, or inside the latest-generation smartphone to boast extra functionality, faster performance, and longer battery runtime.
The laminate is at the heart of this quest to continue increasing performance and energy efficiency, while building thinner and thinner boards. Reducing laminate thickness brings each layer closer to the ground plane, and for a given set of laminate properties the trace widths have to be reduced to maintain dielectric separation. This challenges designers to maintain signaling speeds and control power consumption, bearing in mind skin effects at high signal frequencies. Of course, reducing trace widths causes additional power dissipation, which is unwelcome, to say the least, in any context, whether it’s the tiniest IoT device or smartphone or a large telecom switch or data center server.
New ultra-low dielectric constant (ULDk) materials have changed the game. Conventional design approaches have focused on reducing dissipation factor (Df) to save power, which has tended to produce laminates with Dk’s in the region of 3.2 to 3.4. To maintain dielectric separation with these types of materials, constrained by the impedance triangle, trace widths are now moving below 50µm in the latest smartphone designs. This seriously limits the ability to improve product performance, power consumption and battery life. By reducing the Dk to below 3.0, we can usurp that impedance triangle to maintain trace width and dielectric separation at reduced z-height.
What lies ahead? Despite these great strides taken by resin technologies, remember the PCB is an assembly comprising multiple components. We will probably find the next generation of performance gains by thinking differently about reinforcement materials. Traditional woven glass materials, for example, have inherently non-uniform z-height, and the effect of this becomes more significant in thinner laminates. Other natural variations such as pitch and thickness also have increasing influence on z-height. New construction styles, using single strands of 5µm or 2µm glass can reduce the variation in z-height from tens of microns to just a few microns.