Eric Creviston is president of RFMD's Cellular Products Group. He previously served as divisional vice president of RFMD's communication semiconductor division from May 2001 until June 2002, and prior to that served as product line director of RFMD's Nokia product line. Before joining RFMD in 1994, Creviston was employed by Teledyne Electronic Technologies in Mountain View, CA, from October 1987 to November 1994, most recently in the role of commercial products manager. He received his MSEE degree from the University of Illinois in 1987.
MWJ: First I thought we could discuss the various phone sectors as I understand them. There is the emerging phone, which is cost-sensitive and feature-limited, then the feature phone, the smart phone and the new personal media and M2M devices that are cropping up. Does RFMD view the phone market sectors along these lines?
Creviston: Those are the classical tiers that people have used. We’ve begun to look at it slightly differently based on technical requirements. You mentioned connected devices, and M2M, which is not really a new segment but one that’s been growing very rapidly and becoming quite meaningful now at roughly 100 million units a year, and it would include a lot of the USB data modems, for example.
From an RF perspective, they are similar to a smartphone or a high-end multimode phone, with the exception being that a smartphone typically will have a wide spread of power ranges whereas the connected devices and the M2M might more typically run at max power. Therefore, you would design the RF front end to optimize performance and current consumption at higher power levels.
MWJ: I guess I thought that the M2M would replace 2G; that is, I thought 2G technology and frequency bands would be re-allocated for M2M applications as more people abandoned 2G phones for 3G. Is that a misconception on my part?
Creviston: Well, not necessarily. With machine to machine, currently the majority of the applications are low data rate and are using 2G today. For example, both CDMA and GSM are used for asset tracking and remote vending locations, where there is very little data exchanged. But more and more, people are finding ways to use high amounts of data in M2M applications. For example, video surveillance is an obvious application for that. So you will see machine to machine as a category—also include 3G and I’m sure eventually 4G as well.
If we back up, what’s driving everything is the need to provide higher and higher throughput of data for various mobile applications, and the economics of it—to the point of your title—defines how technology is applied. In other words, how cheaply can you provide as much data as possible to as many people as possible. That is driving all of these finer segmentations, and the 3G/4G smartphone category represents very expensive R&D, so OEMs and chip manufacturers want to leverage them over as many regions and as many carriers and applications as possible. And that’s why they tend to drive many different frequency bands as well as multimode because they have to support 3G as well as be backward compatible to 2G. So multi-band, multimode drives a tremendous amount of RF technology into that 3G/4G category.
We also see a 3G entry category now. It’s sort of an upgrade from the 2G emerging category, where people are finding ways to build 3G-capable devices at lower price points. We actually think this is the fastest-growing segment today. We believe this segment will replace the feature phone in a way, which was generally 2.5G previously, or EDGE. The 3G Entry segment will be built around 1-2 bands of UMTS and then in some cases, just GPRS for the 2G backward compatibility and in other cases, EDGE.
MWJ: And the 3G entry phone is going to be targeting specific areas because it’s band-limited, right?
Creviston: Exactly right. So what you’re trying to do there is create something that can sell in the $80-$100 price point per phone and be very, very targeted and high-volume for a given region, say a part of India, for example.
MWJ: What kind of factors go into your decision about which products to pursue?
Creviston: One of the strategies we have at RFMD is to really leverage our scale and our breadth, not just in manufacturing but also in R&D. So the way that manifests itself is we try to be the complete broad-based supplier for each of our customers. We don’t necessarily choose what to go after; we look at our customers’ R&D plans and we cover all of their bases, for the most part. We also have the strategy of doing the entire RF section—so we have the power amplifier, the switch and filtering, and the power management that’s associated with it.
MWJ: Could you walk me and our readers through PowerSmart™, which you have announced?
Creviston: Sure, I’d love to. PowerSmart™ is a dramatic step forward for the industry in terms of what it’s actually doing. PowerSmart™ is the result of over five years of research and development that started when we were previously doing our own complete 3G RF system.
Historically, a typical 3G handset has consisted of a quad-band EDGE Tx Module (PA plus switch), together with discrete power amplifiers and duplexers for each band. And that made perfect sense when there was only one band of UMTS as a bolt-on. Dual-band is still manageable with this architecture. However, if you look inside some of today’s smartphones, there are 4-5 UMTS bands, together with 2G—it’s not a bolt-on any more, it’s becoming a majority of the content. And so it calls for a different way of thinking about it. A different partitioning is now emerging in the RF area—we are seeing a shift to multi-band, multimode Power Cores (power amplifier + RF power management), together with a switch duplexer module.
The big step with PowerSmart™ is what we bring in—quadrature power amplifiers that are very broadband, very VSWR tolerant, and then we add a sophisticated power management system that is able to go inside the power amplifier and control all the bias points very specifically to make the power amplifier optimized for anything from 2G through 4G. So now you really have two amplifier chains inside PowerSmart™—one for the group of frequency bands that are in the low frequency range centered around 900 MHz, and you have a higher band for ones up around 1.8 to 2 GHz. Those two broadband amplifiers can process anything from 2G through 4G. And not just process it but process it competitively—be as good as any 2G amplifier when it’s in 2G, and be as good as any 4G or LTE or TD amplifier when it’s in that mode.
This is because we basically have firmware control over each of these bands which will tell the band, real-time, what it’s going to be processing. So that gives you an idea of what PowerSmart™ is at first blush. It’s firmware controllable, what we call a Power Core, because you really have to have this intimate design of the power management with the power amplifier itself to make this happen--to really get a world-class performance in each mode. Then once you have this Power Core, you start looking at what you can do with it and that’s what’s really powerful because as it turns out, today in typical hardware designs you have to go in and optimize the handsets for many different applications.
In some regions and with some carriers, for example, the most important thing is to have a lot of output power even though you might sacrifice either linearity or current consumption. You really have to have the output power. In other markets or carriers, it’s potentially better to have less output power because you need to keep the linearity down and keep the current consumption as low as possible. And typically today those are two different hardware recipes where you’re actually fine-tuning the hardware for that. With PowerSmart™, it’s the exact same hardware. In fact, within a very short amount of time within our engineering labs, we can generate the lookup tables that would correspond with either of those two setups, for example. And so a phone manufacturer can determine, in real-time as the phones are coming off the production line, how he wants to optimize it.
Even more importantly, over the air you could actually send in the data and tune the RF front end specific to which region, for example, you’re operating in. To go even further, you can change it per burst, in conjunction with the baseband. In this scenario, the baseband is the one that’s actually sending these lookup tables to the power amplifier. It can actually do so between bursts. So OEMs can move way from making a trade-off in a smartphone design that may be optimized in data mode and have pretty poor performance in voice mode because it’s all done in hardware. Since ours can be configured with firmware, between bursts you can tell it, “Hey, get ready, I’m going to send you a 3G data burst,” and the power amplifier becomes best-in-class at 3G data, and the next burst you can say, “Okay, I’m sending you 2G voice,” and the power amplifier changes, and it’s best-in-class in 2G voice. That kind of flexibility and tunability on the fly within the phone, I think, is going to be what really puts this platform over the edge.
It’s obviously also much smaller. Instead of having all of these individual amplifiers, you’ve got them all combined, so it’s much smaller as well. It’s going to make it much more efficient for handset manufacturers to do their R&D because you have one hardware recipe, essentially. So we think it’s going to be very, very differentiating.
MWJ: So you’ve got an advantage clearly on performance and size and reduced bill of materials. How about cost? I’ve got to imagine the market’s pretty driven by cost as well. Maybe cost isn’t an issue if they make it up with the performance and that’s important, but how does that play?
Creviston: We think essentially what we’re enabling is for people to continue to add more and more bands without adding a lot more cost. We think there’s a practical economic ceiling for the RF in the $5-$7 range. If you’re going to be getting up into the hundreds of millions of phones a year, the economics just don’t play out. You can’t afford $10, or $12, or $15 worth of RF content because you need to keep adding bands and to add LTE, and add all this—you can have cellular front ends that have $10 worth of RF content in them, but you’re not going to be making $100 million a year at that. So, what we’re enabling is for the market to continue to grow in terms of RF content, but for the customers to add more and more bands without as much of a cost penalty.
MWJ: I was asking about the number of devices you sold per year before because I was trying to gauge where we are and where you think we might be in terms of the number of mobile devices sold per year in about five years from now.
Creviston: We’re selling roughly 2 million components a day on average and of course there’s more than one per handset on average with all of these multimode systems and so forth. Our model for the handset industry is a little higher than what most people are publicly talking about now because we do have such a strong business in China, and we fully reflect the gray market channel, which is a channel for these devices. We have the handset industry at about 1.4-1.5 billion units in 2009, and then growing this year to on the order of 1.7 billion units. And that also, by the way, is not just handsets. That is all cellular connection, so that does include the machine to machine and connected devices segments as well. In terms of growth (you ask where we will be in 5 years), I certainly don’t have a number for that. I think there’s a lot of different ways to project it going forward. You have to have your own ideas of what’s going to happen in GDP and what kind of saturation there is. I think it’s clear that every few years there’s a major reset to the expectations for saturation. I recall when we were ramping this business in the late 90s, people were looking at Finland and Sweden and some of the advanced countries with communications, thinking that a penetration rate of 50-60% of the population of the country might be reasonable. And now of course we’re seeing many countries where the penetration is over 100%. And I’m sure you know lots of people who probably have 2 or 3 or 4 cellular modems themselves in various places. The market certainly has the potential for getting much, much bigger than it is today.
MWJ: Are there any particular technologies you have your eye on—CMOS—we talked about the importance of the switch—silicon on sapphire? Peregrine just had this partnership with IBM. Are there things that you’re looking at yourself getting into or that you’re wondering how that might mature and change? We’re largely a GaAs industry; that is, much of the devices focus is on GaAs, do you see a transition, an inflection point, or a crossover point rather where CMOS takes the leading role?
Creviston: Well, we’re known as a gallium arsenide company because we’re the largest manufacturer of gallium arsenide, but we actually ship more square millimeters of CMOS than GaAs today. I’m not sure people realize how much CMOS content there is in these power amplifiers, but I think in terms of the switch function, and the actual power amplifier transistor itself, today you’re right, they’re dominated by GaAs because of the cost-performance tradeoff. CMOS itself is actually more expensive for the power amplification and for that final output stage. We are always continuing to do work on CMOS as well as silicon germanium for that part, and for the power amplifier output transistor as well. In terms of the switch, though, we announced at Barcelona this year a family of SOI switches. We definitely see that technology working its way in. Interestingly, though, it’s not for cost as much as it is for performance in these high band multimode, multiband switches where you have high throw count and need a lot of linearity and isolation. We have some intellectual property in the way we build those SOI switches that we think are going to be very helpful in these smartphones going forward.