The faces of technology today: churning out designs and parts around the globe, Flextronics runs like the massive network platforms it sometimes builds. A PCD&M exclusive reveals just what the world's top electronics manufacturer has its eyes on next. (Cover Story).
Let the record show that in the size, scope and breadth of services it offers, Flextronics is unique. At its core, however, is a manufacturer of the top rank, one capable of firing out tens of thousands of routers and cellphones and everything in between each day. To the outside world Flextronics runs like a Swiss watch. Behind the scenes, however, is a massive organizational effort aimed at ensuring the company keeps its eyes on the ever-moving hands of technology.
At the top of that pyramid are chief technology officer Nic Brathwaite, vice president of PCB technology Craig Davidson and vice president of assembly technology development Sammy Yi. In an exclusive interview in February with PCD&M, the men behind Flextronics' vision explained how the PCB and EMS giant is tackling today's most complex issues.
Over the course of two hours, the trio touched on all aspects of its technology structure, spelling out how it handles designs calling for environmental-friendly materials, enhanced signal integrity, test in the era of the miniature device, and contending with trends in ASICs, FPGAs and PLDs. They revealed how the company grapples with the challenges of having operations in 29 countries and five continents, and how it feeds information across that network. And they acknowledged areas where gains must still be made, as in the interactions of materials and designs.
PCD&M: In a PC FAB interview two years ago Flextronics said its roadmap listed three tiers of circuit boards: high volume, single- and double-sided; six- to-eight layer sequential and HDI; and high-layer-count boards. How has that changed, and how do you see that changing over the next year or so?
Nic Brathwaite: Fundamentally, the technology hasn't changed that much, but the way we think about the roadmap has. We are thinking of our roadmap more along product categories and their requirements. That
includes things like handheld devices [and] network infrastructures.
PCD&M: It sounds like the Flextronics roadmap is now more similar to those that use industry roadmaps like Nemi's or IPC's that use product emulators.
NB: That is very correct.
PCD&M: What are those product applications?
NB: There are several. Obviously handhelds, which for us includes cellphones and PDAs. Although there are a lot of commonalities, there are also some dissimilarities. But we've lumped those together. Network infrastructure products, which include radio networks and optical networks. Another is servers and storage, medium-sized type products. Some of these are large as well, but a little bit below standard size and complexity of the network infrastructure.
The other area is digital imaging, a wide range including things as small as cameras, which to some extent could be considered handhelds because they drive a lot of handheld applications. And on the optical side there are issues, all the way up to printers and plotters. Then automotive and medical.
PCD&M: Nic, you came from the IC side [Ed.: Braithwaite was cofounder of nChip.] Can you characterize how Flextronics' design group and its knowledge of IC design has impacted the way that it designs and builds circuit boards?
NB: There is a connection in terms of this substrate design for the IC and board. Our involvement in semiconductor design has certainly helped us to understand the direction that the semiconductors are going, the requirements for semiconductor substrates, and we have a very good understanding of how certain technologies--for example, high-density interconnect in particular--tend to get driven primarily by the semiconductor design and then move into the product area in some of the cases. A lot of technologies used in cellphones today were used in substrates for semiconductor packaging a few years ago.
PCD&M: Certainly the amount of money going into IC packaging R&D is much greater than on the substrate side. Is the gap widening?
NB: From the development level, there's probably some widening. The semiconductor packaging world is starting to do thin-film semiconductor-like processes; compare that to the way the circuit board is [and] there's probably some widening. In terms of actual use in wide-scale production, I don't think the gap has widened much.
Craig Davidson: I think that's true. The capabilities for some of the advanced consumer products like cellphones and PDAs and ... digital cameras are similar to the capabilities that one needs for substrate development. And that tends to be more of an issue of factory dedication rather than the intrinsic technology content.
PCD&M: How is your FPGA to ASIC conversion business?
NB: Fairly well. Of course, the entire market has dramatically been reduced in every area. The other challenge with FPGA conversion is that more and more FPGA companies are putting proprietary IP in these designs. [It is not] as easy to convert as it used to be.
PCD&M: How do you deal with the IP issue?
NB: We've diversified so we're not just doing ASIC conversions, which is what we primarily did in the past. We're also developing ASICs and system-on-chip type designs, and that is an increasing part of our business.
PCD&M: What types of components are going to see the biggest increase in the next 12 to 18 months?
Sammy Yi: It really depends on the type of product. For handhelds we see the QFP widely being eliminated. In many cellphones there's no QFP at all. On our customers' roadmap and our roadmap the QFP is going to be out of the picture for handheld products. We see more area arrays in packaging, especially wafer-level CSP. You mentioned the gap between silicon packaging and PCBs. At the wafer level CSP a lot of the complexity and challenges are moving to the board level. We see the CSP going to 0.020" pitch, 0.5 mm, and even newer CSPs going down to 0.4 mm pitch. On the handheld side, pitch [is] getting smaller and I/Os bigger, some up to 200, 300 I/O. On the high end like networking infrastructures, I've seen the package getting bigger and bigger and I/O increasing. But pitches are mainly around 1 mm. Many of our customers are concerned that they get it down to 0.8 mm with the higher I/O, say, over 1,000 or 2,000 I/O.
PCD&M: What timeframe would that be?
SY: The 0.5 mm are in volume production. For 0.4 mm, I would say within a year or so.
NB: We see a big drag toward area array. In handhelds, it's moving toward CSP. In network infrastructure-type product--the bigger, high-performance product--we see it moving toward high-pin-count BGAs and ceramic column grid arrays. The area array seems to be growing at a good rate. And getting a larger market and shrinking in size and pitch at the other end of the spectrum. It's actually pretty interesting. You're going to have both types of packages on the same product, which presents challenges.
PCD&M: What are you going to do on the board-level test side if everything shifts to arrays? Will we see 100% electrical test?
SY: Basically, non-electrical inspection tools such as AOI and AXI have gained popularity. For handheld products, AOI and AXI have been used because there is no space on the board for the ICT pin to contact. Network infrastructure products use AOI and AXI because ICT testers run out of capacity in terms of [the] number of test notes they support. For area arrays AXI has been used to verify hidden solder joints. In general, the optimized usage of combination of electrical test and non-electrical inspection tool, the so-called "integrated test and inspection," is the immediate trend.
NB: Many handhelds are also radio products, so you have two challenges. One is [that with] such high density, [there's no] space for test spots. I think you can put some test on and you'll get the kind of coverage that make the in-circuit test useful. And in some radio product ... these spots act as antenna stubs and affect performance. In those cases we tend to use more AOI. On area array packages, obviously the use of things like X-ray laminography and things like that have also become very important. So in many, if not all of the factories, we're using laminography as well as the process control and inspection too. It's a combination of X-ray, automatic optical inspection and functional test that you're starting to see as a result of the higher density, the use of area array packages, etc.
PCD&M: Does the use of AOI instead of ICT slow things down in the long term?
SY: For high-density products, ICT is clearly going to be out because of the access points: there's just no space for them. We have been working with equipment manufacturers to move inspection into the machine. Some AOI manufacturers are working on a smaller camera, which can be installed in pick-and-place machines or even the [screen] printer. A lot of people believe the machine's productive time is limited, so you use unproductive time for inspection while doing placement. That's one trend we're seeing, but we don't know if it's going to be more efficient.
NB: There's no doubt that ICT is probably faster for many applications, but not in every case. There are some cases where ICT is slow because it is used to program components.... There is a lot of pressure on us. There's probably at least five or six times more time taken up in testing [a cellphone] than in actually assembling it.
PCD&M: What types of substrates do you see using in the next 18 months?
CD: It's pretty product- and application-specific. A lot of the work that Multek does is in somewhat higher-performing boards, and we have a higher percentage of high-performance materials. High Tg and also high electrical performance. We see newer materials, which include materials that are compatible with no-lead solders and the halogen-free materials and halogen-free dielectrics in general. We are seeing more and more requests for those kinds.
NB: One [trend] is the increased frequencies of radio applications. LMDS (local multipoint distribution systems) are in the 26 GHz range. Those applications drive material; you need better dielectric performance. The other is requirements such as lead-free that are pushing products into higher processing temperatures are also helping to drive the material requirements on the board and the need for higher Tg.
PCD&M: A couple of published studies have clearly stated that high Tg materials used with lead-free solder produced less desirable results than conventional FR-4 with lead-free solder. What kind of testing have you done and how do the results compare?
CD: We're in the process of characterizing performance. The problem is resources. We have a wide materials set, and trying to qualify a priori for any particular application sometimes is fruitless. Sometimes ... we qualify material for [a customer's] assembly process rather than the other way around. [Performance] varies by material and sometimes by site. The halide-free substitutes are not very advanced yet.
NB: [to CD and SY] Have we seen significant issues of processing these higher Tg materials on lead-free applications at the assembly level?
SY: I think there is no clear answer right now. Many are working on this, trying to find the high temperature impact on low Tg materials and whether we really need [them]. We build lead-free products in very high volume using regular Tg materials [for] consumer product. For high-end product, customers clearly have a concern and we'll have to do more study.
NB: The peak temperature across [a small] board is going to be less than on larger boards. Larger boards are likely to drive the need for higher Tg materials.
SY: Normally, for consumer type products, the peak temperature is probably 240, 245[degrees]C. For larger boards, the thermal mass is big and there are a lot of different types of components, so the temperature delta on the board could be very high. If you set the lowest peak temperature to be 235[degrees]C, the highest peak temperature for a large board can go up to 255,260[degrees]C.
CD: It's almost a cliche to talk about how these forces on the board come together and interact. Design also enters into this, especially with the higher-layer-count boards. You can design boards that will perform better at multiple high-temp exposures, and then there are designs that we know don't do so well under those conditions. It really is an integrated problem: the material set, design and assembly operation.
PCD&M: Some designers say they're going to push the data rate past 5 and up to 10 Gb/sec. What types of materials are you looking at to build these types of boards?
CD: We're building them right now, test vehicles for customers in the 2 to 10 Gb range, out of a wide range of materials including FR-4 or slightly elevated multifunctional FR-4. But also up to the more expensive higher Tg materials and the SI types of materials, signal integrity for electrical performance. And again, it comes back to questions of interactions of the materials with the design. So if you have clever designs, you can get away with more in this space, and the worse off you are with designs, the more you rely on the material set to bail you out.
NB: We are designing what we hope will become the industry standard for testing high-speed semiconductors: a reference backplane that will allow signal data rates of up to 10 Gb/sec. There are a lot of designs in the 2 to 5 Gb/sec range, not a lot in the 10 Gb area. However, we believe that the material technology is available. The connector technology is probably available. It is not clear that the processing methods are quite there to give consistently high data rates over a long path with high yield.... What we think is necessary is to be able to demonstrate this over a larger distance, and to be able to demonstrate that you can put a lot more conductors and connectors in a high-speed backplane.
PCD&M: Is it the size and the weight and the registration you're worried about?
CD: It's tolerance and line-up of different circuits with respect to each other and how well you do in lamination and controlling dielectric thicknesses. Basically, it's a study in tolerance and stack-up as you increase layer count.
NB: This is one of those areas where the collaboration between design, fabrication, test and assembly is crucial. You have to have the right kind of simulation tools to be able to design these products and characterize them effectively.
PCD&M: Is this an equipment issue or a knowledge issue?
CD: A little of both. We certainly need better information to characterize our materials, particularly when it comes to our lamination cycle. Board materials get pretty squishy when you try to laminate them together and there's movement complicated by the circuitry on different layers, which adds effective CTEs, basically, to different layers. There's experience behind it, how to stack these boards and design them, and what you want your original artwork to look like, so that when you finally dump it all together, they're in the right relationship to each other. Registration gets worse with increased layer count, and how you register circuit patterns to drill patterns, for example, all of them come together in these parts. We work hard at controlling etch processes, to get line definition, and controlling lamination processes at stack-up for good registration.
PCD&M: What can designers do, or not do, to facilitate this?
NB: We tried to study, characterize and optimize our processes, as well as our simulation tools, so we could effectively create recipes that the designers could use confidently to design these high-speed boards. The issue is the lack of experience. If a designer wants to design for no-lead solder, they have the rules-of-thumb for a lead solder-based product. We're trying to develop recipes that make designers confident they can design these types of applications using the processes that are in place.
PCD&M: One of the things that makes Flextronics unique is its attention to the global process variation reduction.
CD: At Multek we realized that we needed to focus on operational excellence. We placed a VP of quality with a background of Six Sigma from Motorola. We implemented a program called the Variation Reduction Program; basically our Six Sigma effort without really going to Six Sigma. It's based on process characterization and all the typical things you'd expect a program like this to deal with--process control, uniformity in methodology that our engineers use worldwide to measure processes so that we can actually compare benchmarks. We're using these techniques to improve our yields specifically and operational excellence in general. An added benefit is that, because we understand our process variability and have our processes pretty well characterized, it does lend us the ability to take these design rules that are based in capability and make those available to our designers so that they know what kind of variation they're going to be getting in their boards and can design around it.
SY: From the assembly side, we have quite a few efforts. First, we need to optimize the process. For any new technology, we have established systems to develop the process, optimize and then transfer it to volume production. We also have started Six Sigma. It's really a company-wide effort to improve operational quality, and to move process control closer to where defects occur. ICT is away from where the defect occurred. With AOI and other methodology, we can move the inspection point closer to where defects happen. And we're working with the supplier base to close the loop between inspection systems and process equipment, so inspection systems will automatically provide feedback to upstream equipment. Based on the input, equipment will then perform an automatic adjustment rather than waiting for operators to respond. Some of our high-cost region plants have achieved below 5 DPM [and] we were able to win business back from the low-cost region.
PCD&M: You raise a good point with regard to how the supply chain is interwoven. You're specifying components and materials all over the world. How is this handled internally? What is Flextronics' role in ensuring the stability of the rest of the supply chain?
NB: I'm not an expert in procurement, but I'll give you the best answer I can. The answer has a lot of pieces. We've built a warehouse, which we call MDS Plus, as a way of capturing all the information associated with components we're using in our factory. We have information such as the prices paid for a particular component in any region, the amount of that component located in any particular factory or warehouse. And we have supplier information and customer information, such as how many approved vendor lists this component appears on. We've invested heavily in quoting and logistics capability at the IT level. We have a number of IT tools that allow us to go into a data warehouse and identify what component is available and what prices are being paid for any component in any region at any time.
Procurement is much more complex than it appears. Many cut procurement deals on an annual basis. Without the right information technology and procurement systems, you miss out. Today it may be the best deal available, but tomorrow it's not, because somebody just got a better deal than you. If you do not treat goods in a dynamic fashion, you lose out because of competitive pricing. Procurement won't like to hear me, but I'll say it anyway: Procurement is actually a massively tactical activity. A lot of people treat it like it's more strategic than tactical. And the reason I say that is very subtle. If someone negotiates a price today, they may get the best price. And somebody tomorrow with the same kind of story will probably get the same price or better. Pricing has a lot to do with how much you buy. But pricing also has a fact that is generally not associated with that. For example, if I know the CEO of a company very well and we play golf together, chances are I could call him up and get a better price than someone who's buying more components than I am.
PCD&M: The relationship-based approach.
NB: You have to keep working prices every day.
PCD&M: How do you characterize the knowledge that you keep in-house versus what you're willing to share?
NB: We share a lot with our customers. But the benefits of this capability is more than just the ability to share information. It's actually the ability to use the benefits. Let's break that down. Multek being part of Flextronics allows us a number of competitive advantages. One advantage is our time to market. If we're designing the board and Multek is fabricating it, the system is much more seamless, and we're doing things every day to improve that. Improving that by giving the designers at Flextronics direct access to the Multek design guidelines, so you design the board right the first time. You can have your design review in one shot so you don't have to shift designs back and forth to get feedback and go through that whole loop that can take up significant time. Also, designers have direct access to optimized Multek processes and design rules, so we can design the board optimized to Multek's processes and hopefully improve yield and lower cost. We share most of this with our customers, because they're designing boards and we want them to be able to design boards that have high yields when they use us. The issue has to do mostly with the ability to use the information in a way that optimizes the entire chain, because usually it's not necessarily that easy for a lot of our customers to take that information and use it in the same way we can use it. They have their own design rules; they have their own prejudices; they have their own issues of making sure this is designed to fit in multiple factories and second-sourcing and all kinds of business rules that affect their ability to use information we've given them.
CD: We treat Flextronics like one of our more demanding customers, and it makes us a better, more competitive company.
NB: With Multek located directly on the campus where the products are being built, there is a significant saving in logistics cost. I've seen cases where it's approaching 20% in savings, although not often. With the design, the logistics, generally you can save 10%. If you just look at the board fabrication cost, maybe you don't save much. It might even end up being more expensive. You have to look at total cost from design through distribution. I've personally been involved in the last year in doing four or five of these analyses on very large, complex products and in every case we were able to save anywhere from 10 to 12% total system cost as it relates to the printed circuit board.
PCD&M: Do you have any special concerns that the supply chain may have been materially diminished in terms of being able to provide you with the next generation of technology? For example, drills or screen printers. Some companies have gone a year without selling a new machine.
NB: I'm not worried. This economic situation has been very difficult. If you look at what's happening, a lot of weak companies who were weak financially have gone out of business. And I'm not sure that's a bad thing, because I think in some cases you have too many companies trying to provide services, and the competition was so severe it ended up driving cost in areas where nobody could make any money. That's not good for business.... It's not the number of companies that gives stability in the supply chain; it is their financial strength.
PCD&M: Does Flextronics put a specific number on the number of board shops that it would buy from or the number of material vendors it would buy from?
NB: I don't have a specific answer to that, but I'm sure there are probably some categories--some effort to limit the specific number in areas where we have total control. If a customer came in and said, "I know you have your five, six, three or four suppliers, but this is who I want you to use," if you have a big enough yield, you're not going to say "Forget it." You're going to say, "No problem."
PCD&M: Is there any market that you have your eye on that you think will drive product development the way, say, the personal computer or the cellphone did?
NB: I wouldn't say the cellphone is a past effort. If I have to pick two areas that we believe are driving technology, I would say the cellphone and optical networking infrastructure. It's a hard question, because you can't just look at those two. There are some parts of cameras that are driving technology related to modules, so you have to look at cameras. Be careful: you don't want to give the impression that by focusing on these two things, we think we can cover everything.
PCD&M: Two or so years ago, optical electronics was going to change the world, and now in 2003 we're still waiting to see that first big application. Where does this stand on your roadmap and how will this be slowed by connector technology vs. the ability to put fiber in the circuit board?
NB: I think optical did change the world. Just look at the stock market and all the VCs that lost money and the other people who lost money! But you're absolutely right. You had a lot of people running after fiberoptics and in many cases they didn't understand what they were investing in, and you had companies that had stock prices of $300 now worth $2 and some that are out of business. The reality is, we have a lot of fiber, and bandwidth is not as much of a problem as people thought, and some of these issues we were going to solve, we don't need to solve.
However, there is still a lot of opportunity in optical networking. Look at a lot of the products developed five years ago for optical networking. A lot of the design decisions based on technologies available then resulted in significantly higher costs than these products should have.
And there are other areas like you mentioned, like how do we get optical paths inside the board? I'm kind of on the fence on that. Flextronics looked at whether we wanted to take on a bigger activity in terms of putting optical paths inside a board. We decided, at least temporarily, to not do that yet, and instead focus on our high-speed backplane initiative, which is using what I would call traditional technologies to get higher speeds and higher data rates. And then try to understand what the limits of the current technologies are, before we invest in a big way in putting optical paths inside a board.
CD: I've been in the business long enough to never be surprised by how tenacious incumbent technologies are, and copper wire is no different. Nevertheless, it's like many of the potential technologies that will have a dramatic impact on a number of things, not only design but clearly on how circuit boards are fabbed.
MIKE BUETOW is editor in chief. ANDY SHAUGHNESSY is associate editor.
|Printer friendly Cite/link Email Feedback|
|Publication:||Printed Circuit Design & Manufacture|
|Date:||Apr 1, 2003|
|Previous Article:||Garbage today, gold tomorrow: a robust market remains for those companies with the strategic plan to grab it. (Power Selling).|
|Next Article:||The capacitance of a transmission line: there is a fundamental relationship between characteristic impedance, time delay and total capacitance. (No...|