Business models, roadmaps and a little black magic: An interview with Bob Hilton
by Amy Knutson-Strack
There are many who would be envious of a 30-year career in the same industry – an industry where people took the time to train you well and you, likewise, trained others with the same kind of zeal. And yet, it isn't often that we take the time to recognize these individuals for their efforts and contributions to the packaging industry. In early October, however, we were able to do just that by awarding the Second Annual Microelectronics Packaging Technologist Award, sponsored by the Microelectronics Packaging and Test Engineering Council (MEPTEC) and Advanced Packaging magazine.
Developed to recognize individuals who have played an integral role in the development of technologies that have impacted the back-end of the semiconductor industry, this year's award was presented to Bob Hilton, chief technical officer of MTBSolutions (San Jose, CA) during a special MEPTEC luncheon in Santa Clara, CA.
After graduating from the University of Adelaide in Australia, Hilton worked for Phillips for the better part of four years, where he learned semiconductor package development and its associated business, including test and general engineering. Later in his career, he was the vice president of packaging engineering at National Semiconductor, where he pioneered the industry's first commercially available CSP. He is the founder of C-Pak Pte. Ltd. and maintained a consultant business before joining MTBSolutions, which specializes in advanced packaging, assembly, technology and business solutions for the semiconductor industry.
In his more than 30 years of experience in the semiconductor assembly and packaging industry, Hilton has invested his time and energy in advancing packaging and assembly – and has helped others succeed in this industry.
What kind of training did you receive early on in your career? How did this shape your current path?
I received quite an advanced education while at Phillips I was involved in a lot of technical details – you might call it minutia, but because I was in a laboratory environment, it was the equivalent of an advanced study program. We worked on customer application issues, device test parameters and fault tolerant design, so I really gained a tremendous amount of basic instruction, especially relevant to working with big companies.
When we started MTBSolutions about two years ago, we started off with the premise that we were going to work mainly with very small companies and act as a surrogate packaging department. But, in fact, that is not where we ended up because I am back to working with bigger programs and bigger companies.
Not unlike many companies, it sounds like your business model changed.
It's a humbling experience to recognize that no matter how successful you are, your success isn't exactly what you predicted. You've got to define success more loosely than simply measuring, “Did I meet my business goals?” We were reasonably successful in meeting business goals, but we were way off from the intended one.
So we had to refine our business model. We adapted quite well, but I think those first couple of years caused us a little bit of heartburn because when you think you understand the market you are going after, you have a concept of how you are going to structure it. We thought that we were going to work on a lot of very low technology issues, because we thought that we were going to be selling experience, but it seems the subcontractor market covers a lot of that. Those companies try to make it easy for people to do business because they cover quite a few of the basic issues.
Is this provision of high support levels impacting the large subcontractors?
Based on the strong market during the past few years, I think big subcontractors are focusing business on fewer customers with bigger business – rather than more complex issues – because engineering resources are rather limited.
With engineering resources at a premium in the industry, people try to avoid reengineering things; they try to stick to one brand of toothpaste. Consequently, when there is a shift in complexity, these companies can struggle in their need to get to the next plateau. avoid reengineering things; they try to stick to one brand of toothpaste. Consequently, when there is a shift in complexity, these companies can struggle in their need to get to the next plateau. We enable people to take a saturated labor force, set them up with a new technology program, and graft it into their current structure. For instance, Asian factories are very good at detail management, but usually aren't quite so strong at finding the resources to do R&D.
So how has subcontracting changed to absorb this?
The original model for subcontracting was that when you had spillover capacity, you sent the subcontractor a kit and they gave it back to you six weeks later. Nowadays, it doesn't work that way. Companies expect subcontractors to develop the package, so subcontractors have to have their own technology roadmap, a capacity roadmap, space, people and program management.
If you look at the subcontracting model in semiconductor terms, it focuses on a specific core technology, such as generating wafers in very large quantities. But what enables key customers and their business plans are subcontractors with a roadmap. This way, the customer knows when they are going to have a new capacity available, and they also know that they are going to have to drop certain technologies because they are going to become obsolete. This means there's a business relationship structure [between OEMs or FSA companies and subcontractors], and even with competitive cross-pressures, I think they are in balance with a business model based on specializing in core competencies.
This is also happening in assembly, although not to the same degree. They are not as sophisticated and the margin that the assembly companies enjoy is not as high as the margin for wafer manufacture, and I always think that is a sort of paradox. This industry tends to keep assembly subcontractors very closely controlled to the bottom line.
How can you encourage companies to manage this effect of margins?
That's a difficult thing to address. Few companies can fully understand it…but it seems like the people who work in wafer manufacturing have an easier time of it because they work with a sector of a company that is market-oriented. I think assembly people are more task-oriented. Because people are on very short-term, focused programs, they can be stretched a little bit [in terms of workload], but you can't stretch them ridiculously.
So it's critical to have a clear roadmap and roadmap interim achievement locations. By offering bite-sized sequences to a saturated labor force, it's possible to achieve long-term goals in short increments.
Are companies adapting well to developing roadmaps?
There's a broad distribution of companies, but you can look at companies that have outperformed the index [of their counterparts]. Some of the companies that have spun off from National Semiconductor in the linear analog area have outstanding business models in terms of margin, ability to grow and market relations. They very clearly made a great decision in how they wanted to focus their businesses.
Other companies have tended to “defocus,” even though I don't think they think of it this way – maybe they think of it as “risk spreading.” Companies that tend to specialize in a technology, like memory, ASIC or programmable logic, are focusing heavily on specific technologies. This seems to be a much more successful model than the old model of doing 20, 30 different technologies.
You also can't afford to saturate a company's management with too many tasks. To be a billion-dollar semiconductor corporation today is not such a novelty because there are a lot of them. It doesn't take a lot of analysis to see that companies that have focused around one or two key products tend to do very well. That's simplifying the business model.
What do you see as some of the challenges on the horizon that will impact engineers in the future?
One trend is about what I call “capability bandwidth,” which is more power, more speed, more I/Os…more of everything. This area is not as cost-sensitive as time-to-market; time-to-market is everything in this market because, even though it's in very small quantities, the dollars are enormous and a company's whole future is on the line.
Let's put this into context: Most people working in the high-volume assembly area talk in terms of hundreds of millions of circuits, and in the kind of business we're talking about, it's only tens of hundreds of units or maybe just a few thousands of units. It's not the same kind of marketplace. Sure, if a company is application-specific and sells into a computer application, the company sells a several million chips. That's still a relatively low number.
So, I think engineering folks need to relate to the fact that there are mixed tasks. In other words, if we just look at cost and forget cost-effectiveness, price performance or time-to-market, then we'll miss the mark. This is a big challenge for subcontractors because the model is price, price, price. And, to be honest, for the products subcontractors are doing in big volume, that model is correct – but for the products that are in the future, price may not be so important. I don't truly know anyone in this market who just buys for price, even though they say they do. For most people, your technology has to be the best – then they want it cheap.
Also, things that are manufactured cheaply today went through a learning curve, and we can't expect new products to start at the same place on the curve as mature products. When there's a new technology, it's important to back up the curve. Engineers need to ask the question of price, and if they want ultimate speed, then the [cost] numbers will change. I'm not lobbying for people to pay more money, but price cannot be the only determinant.
What technologies play into this?
One of them is speed, or bandwidth. Everybody is saying that I/O is going up, and it is, but clock speeds and address times are really shrinking. Address lines in the gigahertz range are quite typical in higher I/O applications, and this is a challenge.
I recently spoke with a company that is concerned about running a package at 40 GHz. The technological challenge of 40 GHz might not interest a lot of engineers, but it's a question that wouldn't have been asked five years ago. It's an application where they've got to make high volumes without manual tuning because they can't afford it. Traditionally, this type of product was made at tremendous costs by qualified people who knew how to take wire bonds and shape them. Then add a bit of black magic. This has its limitations because, for instance, a little gold fleck can change the whole nature of a circuit when it's running at 40 GHz. Just by breathing on it, it can change.
It's important to recognize that we are now approaching the kind of time demand where materials are becoming extremely limiting. A customer recently wanted to talk about flip chip vs. lead frames and for what the company was trying to do, the lead frame couldn't do it because it has too much inductance and capacitance. The customer thought that this response was quite discouraging because we were trying to push them toward spending money in packaging. That was true, but they failed to grasp that we were providing a technology roadmap. Once you've defined a roadmap, I think you've got to recognize the law of diminishing return.
Most companies ignore the fact that packaging engineering is becoming a limiting part of the equation. If they continue to misuse this resource by designing totally for cost, then companies will eventually optimize around an obsolete technology. This could replicate what happened to many Japanese companies, which clearly beat their U.S. competition in equipment and semiconductors in the 70s and 80s because they optimized everything, but then they ran out of technology and research fell behind development. Companies have to be careful about what they focus on – and if it isn't your technology roadmap and it's just cost, then you might end up with the cheapest obsolete product in the market.
Some technologies do become obsolete, but what are some good mainstays that have advanced the packaging industry?
The single most important thing we did was bring in auto wire bonders with image processing and pattern recognition because it made such a difference for quality, people management, training, turnover and even quality of life. Being an old operations person, it used to be torturous to hire operators to do wire bond because it was a very highly skilled operation. It would take six weeks to train people to use tweezers to pull wire tails and wire bond by aiming and dragging. You had to use a lot of human skills to keep the operators motivated because it was a dead-boring job.
Incidentally, of the people who trained these operators to do this, many learned to manage people in all their different phases and are now executives in the human resources groups of multinational corporations. What they acquired were human skills, and for me to have taken a part in training this human resources pool is surprising because I'm an engineer. I treated our human resource problem like an engineering problem. We used to say, “Together, we are successful.” People react to you talking to them decently. If you say, “Look, I'm really relying on you; you are a key part of my team and what can I do so you and I work better together?” who won't respond to that? What you are saying is, “You're important,” and everyone's a player for it. It tells you just how significant the problem was: engineers had to become experts in managing detailed people issues.
Then the technology moved, and bond pads started to shrink so we had problems where we just couldn't build the product. So we had yield and quality problems, and we had trouble making output along with a labor turnover problem. And our capacities would go up and down tremendously – one year of boom and another year of bust – so one day we'd need 3,000 operators and two weeks later we wouldn't need 1,000. It doesn't matter what type of manager you are, you don't enjoy letting people go, particularly if you have personally invested all of the time it took to train them. Sometimes we'd lay 1,000 people off and a week later we'd be out trying to bring them back in — that's how crazy it was in the 1970s.
But the real solution was to put in automation. And what went with automatic wire bonders was epoxy die attach because it got away from high-temperature gold die attach. Hot machines used to jam and damage the frames, so with cold die attach, suddenly the machines worked all the time. It was a major breakthrough to get those two technologies. It took a superhuman problem and turned it into a minor issue – now it's about another machine. You don't have to worry about it when they don't need it; I can pull the plug.
By taking the tremendous amount of human skill factor out of the semiconductor equation from the beginning of the line to the end, we've automated all of the key programs and we've changed the assembly structure from a human problem to one of a cost of capital. In only a 50-year period it has become a 200 billion dollar industry by wisely investing money to enable this growth. You can't grow at that rate if you don't basically automate successfully. The way the market was growing in the 1970s, if we didn't reduce [the real estate] for a product by an order of magnitude, there wouldn't be enough land in Asia to build factories and there wouldn't be enough people to put in them. That's a pretty fair assessment, as even today there are people limitations in all of the principal operating regions. I wouldn't say that people aren't available, but I think that good people are scarce in key areas. Productivity has been simply astronomical; I think when I started in the industry, one person could make about 1,000 seals a week of a 14 lead DIP. Today, that number would be 150 times higher – the productivity impact of automation has been amazing.
What would you like to see happen in the industry in the near future?
I've had a chance to grow up with this industry because I started in the very early days when it was in primitive form. And I look at people entering the industry today, and I've realized that they've only learned it through college, but not necessarily through practical experience. The disadvantage that I see is that we need to really protect the so-called dinosaurs – I think that practical experience seems to be missing from a lot of new engineers and I don't think they are given the chance to get it. People might think that veterans are obsolete because they don't really keep up with the technology, but I do. We all have limitations in terms of how much technology we are going to learn, and I think that we really need to think over the next 10 years of how we can provide a database of that expertise, and not just let it dissipate and disappear.
You commented that you had excellent training. Are companies foregoing that now?
I don't know whether companies are. It's not totally obvious to me that it's not happening – of course new engineers are getting experience – but the experience they are getting is more on an automation level so it's not so visceral. It took us years to ship the first 20 billion dollars and now we do it in a month. Think about that. In the first few years of shipping 20 billion, we trained a lot of people. And we can't train anybody in a month. You can only keep up if you train more people and really drive productivity. And that means that things have to be more cost-effective, time-effective and we can't waste engineering resources optimizing a low-cost issue. We need to start looking at time- to-market.