Is RF-SOI the same thing as RF on FD-SOI? No, it’s not. However, the runaway success of RF-SOI and the growing list of recent announcements related to FD-SOI with integrated RF has lead to some confusion in the press and social media. The two are different technologies, addressing different markets, and built on two very different types of SOI wafers. The use of one technology or the other depends on the requirements of the targeted RF application.
For the non-technical reader, here is a bit of basic background. At the most simplistic level – RF: radio frequency – is part of the analog family, and as such is all about waves. And when you talk about waves, you talk about losses over distance (attenuation), speed, wavelength and frequency – which is why the RF design has a rep of being something of a black art. The distance to cover, the power envelope and the amount of data to carry over that distance (and of course, the cost) determine the chip solutions. An important part of the RF chip solution is the choice of the wafer substrate itself.
So here’s a quick primer to help sort out what’s what. Please bear in mind, though, that this is a fast-evolving world, so what you’re about to read is not a definitive and forever what’s what – but more of a general (and simplified) “this is how it is currently shaking out”.
RF-SOI – Talk to the Tower
When it comes to using your mobile device for data transmission over a 2G, 3G, 4G/LTE/LTE-A (and next, 5G) network, you still need dedicated RF front-end modules (FEMs). FEMs handle the back-and-forth of signals between the transceiver and the antenna. They contain multiple parts, including switches, power amplifiers, antenna tuning, power management and filters. Traditionally, they were built on gallium arsenide substrates. But more and more, the multiple chips in FEM chipsets are being reduced to single SOCs built on a special class of high-resistivity SOI wafers. This is the realm of RF-SOI. The wafers for RF-SOI are designed specifically to handle the special needs of getting a lot of data transmitted wirelessly, often over relatively long distances.
The latest standards (LTE-A and 5G) raise the stakes ever higher, requiring mobile devices to support more bands, higher frequency bands, and emission and reception on adjacent bands with downlink and uplink carrier aggregation. (Carrier aggregation refers to the simultaneous reception of multiple frequency bands to improve data throughput.)
For RF designers, that means choosing substrates that favor low RF loss and high RF linearity. A couple of years ago, SOI leader Soitec, in partnership with UCL, brought breakthrough RF-SOI wafer technology to the market (read about that here). Now, a few generations later, Soitec estimates that one billion RF devices are produced each quarter using their advanced and enhanced Signal Integrity™(eSI)wafers for RF. In fact it would be nigh near impossible to find a smartphone that doesn’t have an RF FEM based on RF-SOI wafer technology.
Here at ASN, we’ve covered many of the leaders in RF-SOI FEMs over the last few years. Click on any of these names to get an idea of what they’re doing: IBM (now part of GlobalFoundries), Peregrine, SkyWorks, TowerJazz, ST, Qorvo, Sony, Qualcomm, Grace, Toshiba and MagnaChip. To learn more about the latest developments in wafer technology for RF-SOI, click here. With demand soaring, Soitec’s most advanced RF-SOI wafers are now also being produced by Simgui in China – read about that here.
In fact, the cover story and technical features of the October 2015 issue of the prestigious Microwave Journal is dedicated to RF-SOI – click here to read it.
So in terms of terminology, that’s “RF-SOI”. Now let’s look at how RF on FD-SOI is different.
RF in FD-SOI – for digital integration
When we talk about RF in FD-SOI, we’re typically talking about some RF functionality being integrated into SOCs that are essentially digital processors. True, you can integrate RF functionality into an SOC built on planar bulk (it’s generally agreed to be a nightmare in bulk FinFETs, though). But you can integrate RF into your digital SOC much more easily, efficiently and with less power if you do it in FD-SOI.
RF/analog has a (well-deserved) rep of being the most challenging part of chip design. Analog/RF devices are super sensitive to voltage variations. The digital parts of a chip, which have strong, sudden signal switching, can raise havoc with nearby analog/RF blocks. This means that the analog/RF designers have to care acutely about gain, matching, variability, noise, power dissipation, and resistance. They use all kinds of specialized techniques: FD-SOI makes their job a lot easier (good explanation in slide 8 here). What’s more, FD-SOI’s analog performance far exceeds bulk.
What sort of chips are we talking about? For now, we’re talking about processors for mobile devices, for IoT, for automotive, for consumer electronics. When we say “RF in an FD-SOI SOC”, we’re currently talking about chips that are connecting over a relatively short distance to a nearby box or device (<100m for local WiFi, or a few meters for Bluetooth or Zigbee, for example).
ST’s new set-top-box processors on 28nm FD-SOI (read about them here) are a great example. They are the first on the market integrating 4×4 802.11ac Wi-Fi (using IP from Quantenna) and High Dynamic Range support. This means the set-top boxes can reliably serve lots of HD video via WiFi to multiple users throughout the house (hopefully ending the cry: “Who’s hogging all the Wifi?!?”). ST credits their 28nm FD-SOI silicon technology with providing that highly-efficient RF, state-of-the-art WiFi performance and robustness required for reliable video delivery inside the home.
For RF on FD-SOI – as in other FD-SOI apps – designers use SOI wafers with ultra-thin silicon, ultra-thin insulating BOX and phenomenal top silicon thickness uniformity. These wafers are not the special high-resistivity wafers used in RF-SOI. Rather, they are the latest generations of the same (amazing!) FD-SOI wafers that Soitec introduced in 2010. (For an excellent, in-depth interview with the Soitec FD-SOI wafer guru on the supply chain and the most recent developments, click here.)
This is the type of wafers that GloFo, ST, Samsung, Freescale, Sony, several other companies in Japan and many more around the world are using when they say they’re doing RF on FD-SOI. Bear in mind that this level of SOC integration is fairly new (Samsung and TSMC just announced RF integration into SOCs for the first time in 2014 on 28bulk). But using FD-SOI technology and the corresponding ultra-thin SOI wafer substrates makes life much easier for the RF folks on the design teams, gets far better performance and far lower power at a much more attractive cost.
Further ahead, FD-SOI is also a candidate for transceivers and baseband/modem SOCs, which require high-performance digital and analog/RF integration. But even with transceivers on FD-SOI, you’ll still need the FEM on RF-SOI to handle the interface.
So, that’s the current difference between RF-SOI and RF on FD-SOI.
Hope that helps to clear things up?