The Role Of ISU Petasys, A Multilayer Manufacturing Company in Realizing a Successful PCB Implementation
John Stephens, Senior Vice President of Sales for ISU Petasys (pronounced “E-soo”), built his first circuit board in 1974. Like many others of us in the industry, he got a first-hand education in PCB design, and manufacturing and assembly by working in the aerospace industry.
He explains, “I started in aerospace at Litton Guidance and Control Systems, where I worked in a process development lab that was basically a prototyping shop where we developed new technology and prototypes. We would teach our suppliers the process that we developed and, if it was a new process, we would build the boards. Aerospace taught me a lot. The lab was great as we were a small team and we did multiple processes and job functions.”
“People will say that no one was building complex boards back then, but in 1976 we built a 16-layer board that went into an F-16.
Over his 40-year career, John has been steeped in the PCB manufacturing process. He states, “From the lab, I went into materials planning where I did scheduling and a lot of ordering. Then, I had the job of interfacing with the fabricators and our engineers. We didn’t work with Gerber data back then, we gave our fabricators film, but there were a lot of issues with it. If there were design errors, the film was fixed but the files weren’t.”
“Eventually, we told our suppliers that we wanted to use electronic data. We used to build boards on little twelve inch square panels and then we showed management that we could put four of those on an 18x24 panel and reduce costs.”
To this day, the standard panel size across the industry is 18x24.
John moved into the quality and reliability group and eventually managed the entire supplier quality engineering group which included both boards and components.
“After Litton, I went to Ambitech, initially as director of quality, and then as director of technical marketing and business development. From there, I went to Merix, and then to my current position with ISU as Senior VP of Sales for North America. My position really exceeds North America because everybody is so global now. Our Field Application Engineers report to me, and we provide input to the company’s R&D and technical staff. And, all of our FAEs have hands-on fabrication or electrical experience.”
“Everything I have done has led me to what I am doing today. Every position I held, I learned something new. I just kept building on my experience and expertise.”
ISU Petasys was founded in 1972 and is headquartered in Dalsung-goon Daegu, South Korea. John explains, “From 1987 to 1997, we doubled the size of the Korean factory by cloning the first factory to build the second. From the standpoint of a business continuity plan, the two buildings are mirror images of each other, so if one building is damaged in some manner, we can still continue our operations in the other. In 2,000 we opened our California factory, and in 2013, we acquired Hunan (MFS Technology based in Changsha, China). Hunan does mainstream technology, but not at the level of complexity that we do in Korea and California. In 2015, we built a third building in Korea to house all of our state-of-the art plating facilities because there was no physical room in Factory 1 or Factory 2.”
As noted in other articles, PCB fabrication companies can be mega-corporations with hundreds of thousand employees who build the consumer devices upon which we have become so reliant.
John says, “We have about $500 million in revenue and that puts us around the 30th in the industry. There are gigantic companies that build high volume products such as automotive products, package substrates, cell phone technology and hand-set technology.”
“What we build represents the most complicated multilayer PCBs fabricated in the world today. These are the products that are used in the core infrastructures for the Telcos, service providers, and the cloud titan data centers. Our core strength is the high-end switching and routing space. Figure 1 is a picture of one of our router switch boards. I believe today that we are the only supplier that builds for all three of the major Telco routing hardware companies. We are one of the largest high-end multilayer fabricators in the world. Prismark (Prismark Partners LLC, an electronics industry consulting firm based in Cold Spring Harbor, New York), did a study a few years ago, and I believe that at the time we were rated number two in ultra high-end PCB revenue. For the purposes of that study, high-end was categorized as anything over 20 layers. Figure 2 is a picture of one of our 36-layer high performance computing boards.”
“80% of our business comes out of North America and Europe,” he continues. “From an engineering and in-customer perspective, North America and North American OEMs represent most of our business. We are also doing a lot of business with the ODMs (original device manufacturers) in Taiwan.”
“The ODM industry in Taiwan is very interesting when it comes to the cloud. There are a lot of ODM products that are still lower tech, but the market is really starting to evolve rapidly with the likes of Google, Microsoft, Amazon, Facebook and even IBM. Now, many of the cloud companies are building their own teams to design the switches for their data centers.”
John notes, “On the foregoing types of products, collaboration has become much more difficult because there are many more players involved. The expertise of the ODMs to build a server is one thing when Intel hands you a reference design. It’s quite different to build a 100 Gb/s or 400 Gb/s switch with little to no high-speed equipment expertise.”
He continues, “When you look at our product and services model, we offer services from quick-turn prototyping all the way through high volume production.”
“Regardless of the application of a particular board that we fabricate, reliability is always a crucial factor. You can’t have a board that works once in a while or even fairly often. It has to work the first time and every time thereafter. This is particularly true of mission-critical parts such as the boards that are built
for the aerospace industry. Figure 3 is a picture of one of our boards that we built for commercial avionics.”
As noted in several articles as well as our two books and our one-to-three day courses, a successful PCB product development strategy is one where the collaborative approach is a critical aspect of that strategy. From his perspective, John cites, “The cloud guys have gone out and built big engineering teams by hiring people from the classic hardware OEMs. Then, they build their own product. But it’s still a collaborative approach. Our involvement now is a little more complicated and more important than ever. There are a lot of different levels of experience with the people designing some really complicated product these days.”
“You create a lot of opportunity if you can show value,” he adds. “Our approach is to provide available engineering resources that really add that value. It’s one thing for a fabricator to say, ‘hey, I will build a stackup, or I will recommend a material’. That’s not really adding much value. Because of the work we have done over the years with the North American OEMs, we have the ability to say, ‘this is what we can do.’ We look at what the engineer needs, where we have to push, and where we need to warn people where they are pushing too hard. This is not an easy skillset to develop.”
John continues, “There are a lot of fabricators, us included to some point, who don’t have the internal resources and expertise to do the foregoing. We continue to train our teams in Korea and China to be able to do a stackup, do a DFM or hand somebody a book that adds that value. We educate designers by giving them trade-offs, showing the pros and cons, defining what’s doable and what’s not. Our team in the U.S. drives our technology roadmap by listening to our customers and then saying ‘this is what we need to work on.’ Our team in Korea is quite capable of developing the PCB processes and working on that side of the equation. The complexity comes in having that closeness to a customer where you can see the need. You can’t always predict it, but you have got to be in front of it. You can’t develop it when someone hands you a set of Gerber data. Our customers prefer to not have us do R&D on their products.”
Design and the Product Fabrication Process
One of the mantras that we continually repeat at Speeding Edge is that the greater number of downstream aspects that you can factor into the design process, the better. This involves engaging, in a timely manner, with your targeted fabrication and assembly house.
As John sees it, “The number one mistake that product developers make is waiting too long before they engage with us. Ideally, a customer will reach out to us when they have the idea that they need to build a PCB. It’s never too soon for people to come talk to us. If they don’t have a schematic that is good. If I ask them if they know how big the box will be and they say yes, then that might be too late, as we may have missed an opportunity to add value. They have likely already decided the size of the PCB. That’s the number one cost driver in developing the PCB—raw material. When you look in an expensive cabinet full of electronics, the ASICs are the number one item on the bill of material and costs, and the PCB or the optics are number two.”
He continues, “Everybody looks at us to save cost. From our perspective, the number one cost on our bill is the laminate. The higher the layer count and the higher the performance of the circuit, the higher the percentage of the cost the raw material becomes. As noted earlier, the standard panel that we use is 18”x24”. That’s 3 square feet. You can design a circuit board 12”x12” which is 1 square foot, but you can only put one of those circuit boards on that 3 square feet. We consider this poor material utilization. I always tell product developers, ‘you are going to pay for 100% of the boards I make whether I ship them to you or not.’ “If my yield is 50%, guess what? You are paying much more than you should be paying.”
Over the years, ISU has made numerous custom-sized panels to accommodate products. But not everything can be custom sized because of the laminate manufacturers’ master sheet size. John states, “We can cut a master sheet any way a customer wants, but if there is a bunch that falls on the floor, it’s wasted and it still has to be paid for.”
Engaging with a fabricator early on in the design process does not mean that once that communication is established, it requires ongoing conversations. John explains, “The problem people have is that they think we need or want to engage a lot. We just need to have a conversation when they have the product idea. Then they may not talk to me for three months. Then they call me back when they have the schematic capture done.”
“It’s at this point that we talk about how the connectors are going to be used, the type of BGA that’s been selected, the pad stacks that will be required and the overall loss budget. We have developed checklists that people use that identify the times when we need to touch base. I ask them to let us know when they are going to be at different stages in the project.”
“This is how we build a relationship with the designer. And, our FAE becomes the answer guy. Customers use the FAE as a resource to ask about different kinds of board questions. We become integrated into the product development process as a knowledge source, but we aren’t intrusive.”
“Bottom line, customers can never engage too soon. This fits into our business model of prototype through production to the end-of-life support for the whole product life cycle. When we look at a prototype, if we give design guidance or perform DFM it’s because we have to live with it forever. Not everyone supports that model.”
Woven into the foregoing is the need to make cost tradeoffs. A PCB is rarely built that has all the “wants” in it. It’s a balance between the cost of the design, the time-to-market requirements for the product, and the overall product development cost.
John explains, “When you’re in the development stage of a product, you can point to things that can be changed that help the process but there are definitely limits in terms of the flexibility of a given product development cycle. One of the issues with the late-to engage-issue is that the schedule is always king and it can’t be impacted. There can be many things that impact the schedule and they can be crunched but the end gate never moves. If we were to say that we need to re-lay out a PCB, the answer from the customer almost always is ‘no we don’t.’”
There comes a point in the process when the driving need is just to press on. John notes, “We make these types of compromises all the time. My reply is ‘I get it. We have to do this, but we have to try and change it before the production design is released. Two spins from now we need to make sure that we are documenting the things that we need to amend to improve the manufacturability, the yields, and the quality, but the impedance can’t come up.’ Whatever the thing is that needs to be addressed has to be put on the table before it’s too late. We have gone through full system qualification, and I can tell you at that point in time, nothing is going to change on that product, except when it goes end of life and doesn’t have to be built anymore.”
“This is where some of the problems come in the lower technologies and even sometimes in the higher technologies. Customers have done things a certain way for a long time. And, their constraints are their constraints, and we just have to figure it out. But, this type of approach is expensive.”
“I think that this is really part of the thing [early engagement] that has recently gotten a little more lost. With the cloud and the ODMs getting into technology that they have never seen before, it’s becoming a bigger issue again. Early engagement is critical. If a customer makes it hard to build we have to charge more to build it,” he says, “not all 28-layer boards are created equal”.
“This is why it’s incumbent upon us to make sure that we can make a good business arrangement. That we can have a good partnership and it’s going to work. If you were to solicit input from our top customers, they would tell you that our ability to work with the design team along with the quality of the product that comes out at the back end, is our number one value to them. We’re not always going to be the lowest cost but we are competitive.”
Challenges, Trends and New Innovations
Similar to what Tarun Amla of ITEQ noted as challenges with the board stackup from the laminate side of the business, ISU faces similar things on the fabrication side.
John explains, “Product developers will claim that they designed a product to an IPC standard. Maybe they did, but it’s also possible they didn’t. We also run into instances where they will say ‘oh we tried to do [it] that way but we ran into this problem so we did it this way.’”
He continues, “We are facing the same problem that we faced 40 years ago. People look at things on a computer and it looks perfect. But that stuff moves around in the real world. It’s not fixed. We’re still building boards with cloth that moves. We’ve just added more layers and more complexity. Stackups have gotten better but people are still faced with conflicting challenges. For example, component manufacturers will make a reference design that everyone is supposed to follow. Sometimes, those reference designs make the chips work better but they don’t make the circuit boards easier [to design or fabricate]. That’s a competing interest.”
“We also build test boards for the semiconductor manufacturers. Figure 4 is the picture of one of these boards. This helps on both sides of the equation. We learn from the semiconductor vendors and they learn from us.”
The same holds true for those companies that design connectors. “We build test boards for a connector company,” John notes. “Then our customers go to use those same connectors, and the connector
manufacturer will say ‘well you guys built the test boards so there shouldn’t be any problems.’ We try to explain to these customers as well as the connector company, that the test board is 80 mils thick, but the board you are going to use it on is 2X that thickness. There are a lot of factors, such as plating anomalies, that people don’t recognize.”
As has been cited in numerous articles, the complexity and functionality of today’s electronic products, particularly those on the high-end of the performance spectrum, continually add to the challenges that we encounter in many products being designed and built these days.
As John notes, “What you could do with a 12-layer 062 board, you can’t do with a 160-mil thick, 36-layer board. What was happening in previous product generations still occurs, but not everything does. This is where the expertise comes in. For instance, stackups are still a problem. Product developers want to know how small of a hole we can drill, or how high an aspect ratio we can plate. They will say they want an 8-mil hole. When we ask if that is the drilled hole or the finished hole size, they don’t know what we mean by the inside hole diameter vs. the outside hole diameter. In high speed products, this along with the capacitances and the clearances becomes super critical.”
He continues, “We are also getting boards with extremely tight registration budgets. The stackup is one of the integral things in helping us achieve better registration. In fact, it drives registration.”
“In addition, people want to know what our capabilities are, and they want to see our technology roadmap. I explain, ‘I prefer not to show you that because there is a chance you will misuse it.’ We have our advanced capabilities, and if they pick every element from the column that says ‘advanced’, we probably can’t build it. To do so would be like standing on one foot and then doing a backflip. I can do a backflip, I just can’t do every combination with the backflip.”
“You can get competing design elements wherein people think that they have used an element before so they should be able to use it all the time. They don’t necessarily understand the interaction or complexity of those elements. As things evolve with complexity, people struggle when they try to incorporate more fabrication element feature sets. In that situation, we like to be able to have discussions on what is possible, and what we need to do with a design to allow the customer to incorporate that feature they need to minimize the risk.”
As we have noted in previous blogs that have been posted, today, complex, high-speed, high-frequency, low-loss designs are the major industry product trends that impact the design process.
John explains, “The biggest challenge is transmission loss associated with the higher frequencies we are operating at now. And, we have to address everything that impacts transmission loss including higher performance materials and the glass that is used within those materials. In some cases, people are getting away from glass and going to some other medium which shows promise. In addition, we have to look at copper roughness. This is not just what the laminate manufacturer may put on the board, but what we, the fabricator, do with that copper. Oxide doesn’t work at high frequencies because it makes things too rough. Even ISU’s oxide—that is really smooth—is not smooth enough so we have to look at other chemistries that are non-etching in nature. Over the years, we’ve been making copper rough to make things stay together when they are put in the reflow ovens. You have this giant piece of smooth copper, and in order to bond the laminate material to it, you have to roughen it up.”
“Because of the return loss on the signals and the plane, the copper cannot be rough anymore. So now, we are using chemistries that are very similar to what a laminate manufacturer uses to coat the glass, or is used to treat foil so that the resin sticks to it. The same thing happens with copper. You have to put a treatment on top of it to make it compatible. This is very different from what we have done in the past and it is a current trend.”
John continues, “The boards are getting thicker, the vias are getting longer, and they don’t work because there are just horrific impingements to the eye so they have to be backdrilled. On a product that we made a year ago, you could have a 12-mil stub. On the product that we are making for next gen 56 Gb/s and 112 Gb/s products, 6 mils is the maximum stub. To be able to drill to that level of accuracy on a really thick board consistently with conventional technology is impossible.”
Because of the foregoing, ISU had to develop a new process so they could drill to a 6-mil stub and do it repeatedly. John explains, “It’s a 4 +/- 2 stub, it’s deep and it’s amazing what the process engineers have achieved. This, along with a high aspect ratio, [is] driving trends. 20:1 with a .2 mm drill is commonplace in today’s high-end networking architectures.”
ISU also worked with drill equipment manufacturers to improve their fabrication processes. John says, “The manufacturers had some really good ideas and good techniques, but the software was limited, so it took some time. Our own engineers had to get involved. We had to improve our flatness control, but at the same time we had to develop an inner layer sensing capability. There’s equipment now that can do this but we still have more work to do on our side.”
ISU also works with laminate suppliers and its own customers to develop new materials and test them through SI and thermal mechanical reliability. “We work with chemical manufacturers on smooth copper,” John notes. “Ideally that chemistry would be compatible with all of the resin systems but that’s not possible. No one makes a universal non-etching solution that is compatible with every manufacturer’s laminate materials.”
The trends of board thickness and package size drives most of the development that ISU does these days. “We always look to enhance our process capability around the evolution of the PCB technology,” he adds. “There are things that are more innovative, and we need to be able to see these things coming. Some of these things have taken us four years to develop. If someone had told me before that I needed a 6-mil stub, I would have told them they were nuts.”
Over the next two to five years, a number of evolutions will take place. John states, “In multilayer PCBs, we are dealing with transmission lines now, but density is going to be our next big challenge, in particular, the density of the BGAs. Think of all of the little parts that go into the build-up products—such as cell phones.”
On the high-end networking product side, the parts keep getting bigger. “On the switching chips that are being built, every part is bigger in size,” John notes. “I think we are approaching a point where we are going to see a density shift in the large format to a much tighter pitch, which is going to increase the I/O density. The way that we build a PCB today is not going to work. I don’t know if it’s two or five years but it’s in that window. We are going to see a shift on some high-end products that are really going to require some new fabrication processes—techniques and approaches that we don’t utilize today. They are used in some regard on some products such as cell phones. But the things you do on a cell phone, you can’t do on a networking card. We are going to have to develop a combination of technologies that are used in a variety of industries, and figure out how to make a 40-layer board out of it. This is why I think density is going to drive the next set of requirements.”
The Impact of COVID-19
As with many other manufacturing venues, the full, long-term effect of COVID-19 on the PCB industry remains to be seen.
John explains, “I think the industry and the people in it have done a really remarkable job in being able to support one another. Because ISU is based in South Korea, we experienced a potential impact. There were visitors from Wuhan who traveled to Daegu for an event, and people got infected, and it spread really quickly within that group. Our factory is located just outside of Daegu. But Korea took a really proactive approach in terms of testing. They were able to test and categorize people very quickly. Initially we were worried that when the outbreak happened in Korea, that we would have to close for a while. But we implemented the protocols similar to what is recommended by the CDC and then some, and we were successful.”
“For us, the impact has been more from what we have to do to maintain our operations. But we haven’t impacted our customers. Our supply chain has seen interruptions, so our productivity was diminished for a time while we went through having to clean things every day and learning how to live in a COVID-19 world. The biggest impact has been on logistics. People aren’t getting on planes to fly around the world. These same planes would be carrying all the stuff that we need to build things. Being in Korea, most of our supplies are flown in. As a result, we had to change to ocean shipments, which is extending our lead times. Each country that we get product from has gone through a phase of having one shut down or shortage to the next.”
“Our customers want us to ship air freight. But we can’t do that because there are no planes available and the wait time has gone up to three days. With that kind of delay, we might as well put the product on a boat. In addition, the air freight charges have gone up by 3 or 4X.”
The aspect of hoarding also plays a role in today’s product development environment. “Some really big OEMs have gone out and brought up capacity and accelerated the delivery of supplies to try and mitigate the impact,” John cites. “This creates a ripple effect. I worry about the backlash of COVID-19 because they have stockpiled stuff and created some supply chain irregularities.”
“We have been expediting a lot of product as a result of COVID to fill the gap from some of the Chinese suppliers. At the same time, our customers are trying to get out in front of the problem because they have seen the impact of COVID as it moved initially from China to Italy and now to Malaysia, Thailand and other countries. Two months after COVID broke out in China, Malaysia got hit and, for a while, we couldn’t get the copper foil that used to come from Malaysia. The government shut everything down.”
“We’re having difficulty getting different products from different parts of the world. These are the kinds of issues resulting from COVID that go beyond direct factory operations. We’re fortunate because we are listed as an essential infrastructure business. In our California plant, we implemented the same thing that was implemented in South Korea—thermally scanning people for their temperature; having everyone wear masks and gloves, or washing their hands and maintaining social distancing within the workplace. People who have office jobs work from their homes. Our big thing has been maintaining our customer support efforts. We have constant status meetings to let our customers know what is going on.”
“The key thing is that no one is taking their foot off the accelerator. Think of the impact of the infrastructure to the communications network with everybody working remote. We’re not wired to have everyone working from home. Think of all the meetings taking place over the Internet. Cisco has been reporting that Webex has never been bigger.”
And, as noted by all the news agencies, the impact of COVID goes beyond the workplace. John states, “You have close to 50 million students not going to class. It’s been estimated that a third of the students literally don’t have the wherewithal for home learning. It can’t be offered to 2/3 of the people and disenfranchise the other third.”
As a result of the foregoing, there are some aspects of the industry that are seeing a high level of growth. “Google classroom is a really big product, as is the need for Chromebooks,” John explains. “And, all of our big networking customers are moving forward, as are the cloud companies. They are going at a faster rate/pace, because a new reality is being created as a result of COVID. And, with some people already being less than enthusiastic about public education, we may end up with more distance learning and kids being home schooled. The products associated with this are going to get more complex, and we are going to become more dependent on them. The paradigm is shifting, and it will be interesting to see what is going to happen next.”
The interface between the design and fabrication process is multi-tiered, and touches on a number of product development techniques from pre-schematic all the way through the final delivery of the product. Leveraging this interface at all phases of the product development process ensures that a product will work as a prototype as well as a finished product to end of life. With the complexity of today’s designs, and the demands put on the PCB, it’s imperative that the relationship between design and fabrication be implemented from the start.