Joe Fjelstad Breaks Down His Occam Process
Joe Fjelstad recently met with the I-Connect007 Editorial Team to discuss the potential benefits of his Occam process for solderless assembly. This technique allows assembly of the PCB without the risks associated with traditional surface-mount processes, such as solder joint failure. Has the time come for the industry to embrace Occam?
Nolan Johnson: Joe, we wanted to talk to you about shrinking device sizes and features because you have been doing some work in that area.
Joe Fjelstad: I actually go back to Tessera where we were developing chip-scale packaging in the form of the micro-BGA in the early 1990s. Tom Di Stefano was the company founder. The original name of the company was IST Associates. The micro-BGA addressed solder reliability issues at that time. Tab was the rage at that time using the interconnect on the exterior of an IC and fanning out to a still fairly fine pitch. Tom had the notion of putting a little flex circuit on top of the chip and then fanning it inward to a relatively coarse pitch above the surface of the IC. That was the earliest of the chip-scale packages and the one that really broke the ice for the CSP revolution.
Tom knew from his experience at IBM that there were concerns about the reliability of flip-chip and use of underfill and other things that people were looking at to try and deal with the inevitable coefficient of thermal expansion mismatch between a silicon chip at 3 ppm/C and FR-4 with a CTE of 24 ppm/C. He and his colleagues at the time came up with the idea of putting a buffer between the chip and the PCB substrate. It became a poor man’s flip-chip and something that everybody could use. That was almost 30 years ago, and we’re seeing the benefits of that still today.
Johnson: One of the things you’ve been working on recently is the Occam process. Tell us about that.
Fjelstad: For about a dozen years, I’ve been pounding sand. I think it is an inevitability, but it involves a sea change in terms of how things get done. Happy Holden is familiar with my pursuit of some order in the industry and the idea of locking in on a fundamental grid pitch. Over a quarter-century ago, I put together a little Lego module with Vern Solberg at Tessera and said, “Here’s what the future looks like.” All the component terminations are on a common grid, and the result is the ability to route more predictably and easily.
Quite frankly, I think we can do it, but the reality is that it involves a lot of attention to and embracing of order, rather than the cacophony that we have today. The situation today is a seemingly endless number of IC package types, lead formats, and pitches. These were the result of choosing to transition to surface mount without fully appreciating its potential and power. Sadly, we didn’t immediately jump to area array. Instead, we went to peripherally leaded components first, and then that’s when the wheels of the orderly design fell off, and the 80% rule stepped in. It’s nonsense.
Johnson: Walk us through it, for those who aren’t familiar with your Occam process.
Fjelstad: The idea sprang from my opposition to lead-free because tin-lead is a superior solution, and there was no real risk of harm to the consumer from lead in electronic solder. However, as soon as the EU mandated lead-free, I had an epiphany while sitting down to write an article, and I wrote three words—assembly without solder—and the Occam process came immediately to mind. Reverse the manufacturing process. Rather than building a PCB and soldering components to it, build a “component board” and build the circuits up on it using an additive process.
This was the first way I saw, but there are many ways to do it. Fundamentally, the idea was to invert or reverse the process. The Occam process concept completely bypasses the soldering process and all of the problems associated with it, which we’re all quite familiar with, and we have been dealing with in earnest for the last dozen years or more. The transition to lead-free was not easy, nor was it required in the long run, if you look at the science.
Johnson: I’m trying to imagine your component board and the circuitry on top.
Fjelstad: I’ve created a number of graphics over time, but in cross-section, it becomes fairly clear what the process steps are. There’s actually a significant reduction in the number of steps required for manufacturing and electronics assembly. In fact, with help from my good friend, PCB designer John Goodrich, a demo was cobbled together, showing how one could build something like this at home using 5-minute epoxy, conductive ink, and nail polish. I won’t bother you with the details here, but it worked. While I have a preference for bottom-terminated components (BTCs) and land grid arrays (without solder balls), the concept will also work with legacy components; it is just not optimal. The idealized components from my perspective are only components that exist on a common grid and only components that have bottom termination.
Here’s a simple thought experiment that illustrates the process. Think of putting all the components flat on a surface to which epoxy does not adhere well. Build a little dam around the components, pour epoxy over them to seal them in place, lift off the cured embedded component assembly, and then start doing your buildup on the exposed terminations. There are obviously some cleaning steps to make sure the contacts are not contaminated, and there may be a number of layers of insulation that are going to have to be built onto it, depending on the number of copper layers required. My thinking along those lines has been the possibility of using a photoimageable resist because it’s very easy from a buildup standpoint and building layer on layer.
Barry Matties: How does this fit into miniaturization? Is this something that would fit for all fine-pitch boards, or is there a physical limit to where you can go with this?
Fjelstad: I believe there is a significant opportunity to build very component-dense assemblies, but super fine pitch circuits may not be required. Unfortunately, the industry is accustomed to using solder. They also have a lot of understanding and sunk costs in manufacturing materials and processes. Many IC packaging houses have taken the pursuit of the flip-chip method, which they’ve become pretty skilled at. If you’re very good and have developed your skills for that, you don’t want to give it up, nor do you want to strategically, if you think about it. It keeps your competitors in line. Moreover, once you’ve developed a particular skill, you loathe to give it up, especially if it gives you a proprietary advantage.
I am advocating a simple man’s solution. I am seeking to create a solution that makes it so everybody can build these things. For several years, I’ve talked about disintegrating ICs and allowing designers to design with IP blocks that are fully packaged and tested and burned in, and maybe even built on earlier nodes. I was suggesting this before the current idea of chiplets caught on, with the difference that I see these functional blocks as packages with common grid pitch terminations. In this manner, we could build devices that will have incredible long-term reliability. Potentially, the reliability could run into decades—even centuries, for some of the earlier nodes of IC fabrication—and rather than getting down to the nanometer nodes, where you’re looking at prospectively months before devices start to wear out due to the diffusion of the metals through the insulator.
Johnson: Your idea would have the benefits of what a lot of people see in 3D printed circuits, and the benefits for standard components and they would be vibration-proof and very much coated and protected from environmental extremes.
Fjelstad: Spot-on observation, Nolan! At a design show some years ago, I saw a laser scanner, and the company could take that laser scan and immediately convert it, for my purposes, to a component carrier substrate. I’ve written that up in a number of my disclosures. “You can do this as a direct print of the substrate.” The other thing that’s fun about printing substrates like this is that if you look at the ways those substrates are designed, the reality is to save the material, as much as they can, these things will virtually “print air” because of the filigree nature of printing to save time and material, and then only the outer surfaces will be solid.
From a high-performance material perspective, you could have a very low dielectric constant to the substrate because, again, the carrier substrate is mostly air. It’s like building something on Styrofoam, which is exceptional from Dk and Df perspectives; however, the printed substrate would be stronger. Every time I turn around, I keep on seeing additional advantages to this technology, and 3D printing fits into this scheme, including the 3D printing of the substrate and laying down both the insulator and circuit layers sequentially.
Johnson: Early on, you mentioned fewer manufacturing steps. I can see that being the case.
Fjelstad: Absolutely. A lot less equipment and energy and much denser. The perspective benefits are cost, performance, reliability, and design simplicity. Those are the big broad brush strokes, and then there are all kinds of subsets from the standpoint of thermal management, resistance to shock and vibration, and even hermeticity. The more you unpack the concept, the better it gets. But this remains, for the moment, an idealized approach, but it is ready to become an ideal approach as to how these things might be done when the industry is ready to embrace the change. Then, those modules can, and will, be built. Normally, I only demonstrate or show single-sided assemblies, for the most part, but double-sided assemblies are quite doable. I do not, however, suggest that the Occam process is a panacea. I can still see a role for solder in the future—hopefully, tin-lead solder. I think that represents an interesting opportunity for us going forward.
Johnson: You said it’s not a panacea. What are some of the challenges?
Fjelstad: Mostly, it relates to legacy component types. On a number of occasions, people have said to me, “What about an electrolytic capacitor?” I suggest one use the technique that they need to use for whatever it is they have to put together for their particular assembly. But these things can be modularized to a point where you have essentially put them in play. I had a couple of opportunities to visit Peter Drucker, the guru of American management, at Claremont College some years ago. He sat at a table and talked; there was no agenda, syllabus, or anything of that nature, as I can recall—just a stream of consciousness. Perhaps his most important suggestion for good management is that one needs to put people into positions where their strengths can be fully utilized, and whatever weaknesses they have don’t matter. I have mentally applied that same thing to manufacturing processes. Use processes where you can fully tap into their benefits and then make sure that they have weaknesses that they don’t come to the fore.
Johnson: How large have you gone with your designs at this point?
Fjelstad: Again, I’m still working largely in the laboratory of my mind. That said, I have taken a part-time role with the folks at SAIC. I had interest from the folks at the Navy’s Crane PCB facility back in 2008; I visited and discussed the idea with them, and they had an intern build a simple circuit using the Occam concept, but the wheels fell off the economy in 2008, you will recall, and although we continued to talk, there was never the ability to do much about that due to a lack of budget. Now, they have kind of a state-of-the-art facility, so we’re eager to go through an exercise and do something along this line in the coming months.
At around the same time, I gave a seminar in Timisoara, Romania, in 2008. A professor from Bucharest Polytechnic University was in the audience; we exchanged contact information and have corresponded and remained friends over the years. They have continued to look for ways to collaborate with Occam, and I have been there several times. Now, I’m heading there in the next month to visit with them and give another seminar. One of the major manufacturers of automotive electronics is keen on taking a look at the Occam concept to make better and cheaper electronics. It looks like they’re seeking some funding from the EU to be able to go out and do some proof of concept for their own purposes.
Happy Holden: HP took it up in that a lot of our boards were chip on boards, which there was no solder. Then, we came up with our second generation of calculators that had no solder in them. And then we came up with tab, which also didn’t use solder. HP has been a pioneer with solderless connections.
Fjelstad: Exactly. The “Finstrate” structure that HP came up with was a real landmark. HP was one of the greatest gold users of all time, though.
Holden: That’s because we never had any tin-lead or reflow.
Fjelstad: I wasn’t aware of that.
Holden: We had tin-nickel and put gold on tin-nickel. I started reducing the thickness of gold, and we introduced bright acid tin on top of tin-nickel. Then, we went to SMOBC, an emergent silver, but never any tin-lead.
Johnson: Earlier, you made a point around not seeing solder go away, but it would change how much we use it so that tin-lead comes back into play. Where do you see the trade-off now that you’ve been working with your Occam process? If we go down this path, what are the circuits going to look like?
Fjelstad: I’ve been working with a very talented designer and Athena Tech founder, Darren Smith, off and on for a number of years. Darren picked up on the Occam idea when I first trotted it out a dozen years ago. He’s a very competent designer, and he jumped on it. He said it absolutely made sense, and he did a design comparison. It was pretty dramatic. We complemented each other, and I tended to challenge him. A few years back, I said that I needed to show something to somebody, and he did a redesign of something he had designed for traditional assembly. It was impressive. That reminds me that a good designer can understand pretty quickly. Early on, I was invited to give a talk at one of the PCB design shows early on about it. I gave a little keynote address in a room full of designers; for the most part, the designers listening nodded their heads, grasping the potential. On the other hand, they look at it and think, “Who’s going to make it?”
Again, my thinking goes back to the idea that if we’d picked up on the general Occam concept with one of the major vertically integrated OEMs that built it all, from ICs to packages and PCBs, including all of the raw materials. It would have already been done. IBM was a showcase of vertical integration because they made silicon, ICs, lead frames, glass, copper, and epoxy. They made everything that was required for their circuits, and I always thought they were a bit crazy until I visited their manufacturing facility in Endicott, and discovered that they weren’t crazy; they were genius with how they set and controlled everything. It’s important to note that in the early days of this industry, every circuit was designed using a standard grid. That was because component terminations were all provided on 100-mil centers.
As a quick side note on that, however, when I went to work in the Soviet Union, their dual-inline packages were 2.5 millimeters—not 2.54 mm (100 mils). With a 16-pin DIP components could be interchanged, but if it got big, then the run-out didn’t work. In other words, you could pretty quickly get a run-out on the things they would work for a few pins, but when you got to larger pin counts, it didn’t work anymore. That was serious but also an amusing by-product of using English versus metric units.
I need to go back to Darren to complete a thought. A while back, I gave a paper at IPC on the Occam concept and wanted to show the potential savings, so enlisted Darren’s help. I said to him, “If you don’t mind, take something you’ve designed—non-proprietary, please—and then do a redesign using the Occam concept for reverse manufacturing. Have every component with terminations on a common grid. 0.5 millimeter is preferred because below 0.5-mm assembly yields for solder typically drop off.” That paper included an example of what he did. It went from 12 layers to six layers, and, as I recall, it went from an original 110 mm by 140 mm down to 70 mm by 30 mm. Then, he got creative and made it into a rigid-flex and separated digital from analog so that it all folded up. It weighed a fraction of what the original design was, and he consumed a small amount of the space. Something I forgot to say earlier is that I told him at the time not to go to DigiKey of other component catalog and look for actual components but pretend as if those components existed with terminations on 0.5-mm pitch, including a 442 I/O FPGA that was the centerpiece of that design.
Holden: If you want the figures for the Occam, you have them in our circuit designer’s guide to solderless assemblies. All the Occam figures you need are in high resolution in your free eBook.
Fjelstad: I’ve been trying to keep that comparison image out there in front of people to allow it to sink into their minds. At the same time, most of them are going to look at it and say, “Wow! That’s a neat idea.” Invariably, people are agreeable to the fundamental idea, but on the other hand, they have to do what they have to do to get through the day. I understand that. When I’m asking a semiconductor packaging manufacturer to make a device, make all your packages with 0.5-mm pitch terminations, including depopulations of the fundamental pitch. Again, I chose 0.5-mm pitch because that’s about the point where the wheels fall off for most people in terms of solder assembly. In other words, it’s a potential for being able to use design for both solder and solderless assembly. When you’re getting ready, you can finally embrace Occam.
The only not insignificant problem is that with solder, the designer needs to provide room for rework and repair. One of the things I stress with Occam assembly is that rework and repair are crutches to support a technology that has lots of problems. That’s what you have to do when you don’t do the job right or when you have a technology that’s incapable of being able to deliver the kinds of quality that you need for your product. Solder is a great technology, but it doesn’t need to be used for everything.
Holden: Joe, have you been following our articles about vertical conductive structures (VeCS)?
Fjelstad: I have been looking at that, and I’ve also been an admirer of Joan [Tourné]’s work for a long time.
Holden: I’m also talking with Joe Dixon, who is the VP of R&D for WUS and happened to work for me in Palo Alto at a PCB shop in the latter part of the ‘70s. WUS had been building all these things and supplying them to customers, but I introduced him to your pictures of multiple nets per via. What company was that with?
Fjelstad: I was still at Silicon Pipe then. I had some good responses to that idea of being able to build a split via and get rid of a lot of the inductions and other limitations relative to traditional vias. We had fun in the early days and were completely unfettered by convention.
Holden: That’s part of my lecture for the electrical engineering students at Michigan Tech on future technologies they may see in their careers. Why should a via only have one net connection? Why can’t a via have two, three, or four nets using the same via?
Fjelstad: I’m also keen on the idea of using vias as resistors and putting a dab of resistive material in some of the holes between process steps and then plating it over; they’re not necessarily great for high power, but printing resistors and capacitors, etc. You can establish a fairly precise resistor. It won’t necessarily be great for high power, but most resistors aren’t high power.
Holden: It would be interesting to know that. I licensed an Occam-like process from the University of Helsinki since they never recognized you or gave you any money, but embedded ICs is the Occam process because we do it exactly the way you outlined. You laser drill down onto the pads of the silicon and then plate directly for embedded.
Fjelstad: The one thing that I’ve held in a very stringent way was the idea of putting everything in packages and testing them. If ICs change with each die shrink, and the pad out changes, then you have to rip up and throw away everything it did. If you put it in packages, then what you did is still good, and you don’t have to go back and rethink and redesign because somebody went through a die shrink. The packages provide the standards, which are the glue that will hold them together.
Johnson: What I’m picking up from this conversation is that it’s a combination of how you’re securing the components and then working with standard die sizes or standard pad sizes and placements. Those are two separate but not necessarily connected ideas; when you put the two of them together, that’s where the real benefit to the Occam process comes into play.
Fjelstad: Yes, and there’s an overarching view that needs to be taken into this thing. Viewed in isolation without perspective, it doesn’t make a whole lot of sense. I see this holistic overview of what it all looks like. Again, the Lego module that I put together is the linchpin for the Occam approach. It’s something that anybody can get their heads wrapped around quickly. Legos definitely work. If every one of the chips used has its own unique pad out or it’s on a different pitch, it all falls apart. It doesn’t make any sense, but if you adhere to fundamental precepts and standards, it works. This is part of the lesson we tried to teach at Tessera where we talked about using TLS. Tessera laminated substrate a 0.5-mm substrate that looked like an early breadboard much like IBMs standard boards from the 1970s.
The idea of building a standard substrate like what we proposed at Tessera early on is where I fully grasped the idea of needing to have something that created order and made sense, but it involves doing things differently. Circling back to that idea of the IC package foundries, remember that they have all of these different lead pitches. They are all depopulations of all these different pitches, and then you have all the different sizes of these structures and various finishes for them. I don’t want your nickel and gold or your solder on the termination. Give me bare copper; now, how much more are you going to charge me for making your process simpler and with fewer steps to control and validate? It all costs money.
There’s a great quote I read when I was at Boeing almost 40 years ago, when I first picked up a copy of Wind, Sand and Stars by Antoine de Saint-Exupéry. There was a passage in the book of him talking about the wing of an aircraft that really struck me: “A designer knows he has achieved perfection not when there is nothing left to add, but when there is nothing left to take away.” That has been a cornerstone of my thinking in the four decades since.
Holden: What’s the status of the Occam prize?
Fjelstad: Thanks for mentioning that, Hap. It’s on hold at the moment, but I will resurrect that at some point in time. I was frankly disappointed that I didn’t get enough people to participate earlier. The biggest impediment is the designer will have to start learning something new and a bit different. I was trying to incentivize designers with a monetary prize to expand the horizons of their thinking and consider something both new and ultimately simpler. I am confident that a tipping point will eventually be crossed.
Johnson: Joe, we really appreciate it. Best of luck with Occam. It would revolutionize the industry.
Fjelstad: Thank you, Nolan. I’ll keep at it.