The 2019 International RF-SOI Workshop in Shanghai was packed to overflowing, with over 500 attendees, noted SOI Consortium Chairman & Executive Director, Carlos Mazure. There were 16 presentations over the course of the day – every one of them excellent – so it will take two posts to cover them all. We covered Danni Song’s compelling keynote in a previous post – see it here. In this post, we’ll cover the remaining keynotes and the morning session, which was dedicated to 5G deployment. In the next post, we’ll cover the afternoon sessions, which were dedicated to the China RF-SOI ecosystem, and the RF value chain.
PDFs of the presentations are not yet posted, and then they will only be available to those whose companies belong to the SOI Consortium. But we’ve summarized them all for you, so read on!
The keynote by RF-SOI legend Jim Cable chronicled his always-innovating journey from digital to RF via sapphire then SOI. (Cable’s work was recognized in an award that evening.) His original vision back in the early days was for a RF front-end module (FEM) + CMOS transceiver. At the time, doing it on sapphire (an insulator) rather than bulk made it much easier, as sapphire eliminated the non-linear capacitances. That was the beginning of their UltraCMOS technology, and though it did very well, sapphire was only available in 6” wafers. So pSemi (or Peregrine, as it was known at the time), engaged with Soitec on bonded-SOS. “It was a killer technology, and the marketshare we won was staggering,” he recalled, and helped convince Soitec RF-SOI was worth looking at. The goal is to handle high RF power levels: you can use SOI to handle higher voltages than you’d think were possible. They added an invention they called HaRP, that dealt with accumulated charges and enabled them to hit the linearity specs on silicon. With that, he explained, they came to completely dominate the switch industry. UltraCMOS evolved, getting 60% smaller with 20x better linearity – but now of course you have 50 switches, not six. He heralded the great partnership they have with GlobalFoundries, noting, “We were pioneers in this field.” In fact, in 2017 they were in the top 10 for IP generation in semiconductor manufacturing. Now comes mmWave, where he says, “We see everything we believed and more.” They’re currently sampling an 8-channel mmWave RFFE (RF Front End).
Michael Reiha’s talk centered on how SOI wafer-leader Soitec is positioning itself on 5G, which, he explained, demands a wider portfolio. Soitec looked at what they could do to make 5G ready for sub-6GHz. Massive MIMO (mMIMO) is an efficient technique to improve throughput. With SOI, you can reduce the power it takes, making it a good choice for urban environments. RF-SOI is a candidate for power amplifiers, and FD-SOI is enabling more users to be added. The concept of network sharing is an opportunity for compact, low-cost filters that can meet the requirements with simpler, lower-cost, higher-efficiency filters. That’s why they’ve just announced a new substrate called piezo-on-insulator (POI). However, total cost-of-ownership is not just how much a product costs, but how much it costs to run it.. Currently, RF and mechanics dominate the bill-of-materials, so you need to decrease the number of RF FEM components and get savings scaled with the array size. The main challenge of SOI is in efficiency, but the advantage is that it can be used in integrating digital with analog sensing and RF. Then you can use AI sensing for tracking temperature, for example, and control for 5G optimization. In short, with RF-SOI, you apply AI to the radio head, especially for things like mMIMO. And btw, he added, Soitec currently has capacity of two million wafers per year.
There are over a thousand 5G networks available today worldwide, said Cederic Malaquin. Adoption is accelerating, driven by 5G cloud gaming, AR/VR/XR, 5G multi-video calling and stadiums. However, carriers need better ROI. 5G should address this so that customers are better served. MIMO and carrier aggregation (CA) are the main techniques supporting network capacity and coverage improvements. 5G NR will bring more spectrum. With each generation putting more content in phones, new spectrum is happening in sub6 and mmWave. The impact of 5G on mobile phones is huge in terms of both content and complexity. Some phonemakers (like Apple and Samsung) are moving towards increased integration. Others (like Huawei) are going more for discretes. Yole sees tuners, switches / LNA as addressable by RF-SOI, but they are less convinced about power amplifiers. They also see SiP (system in package) as prevailing over integration. The 5G mobile and base station markets will really build up in 2022-25. RF-SOI will remain the mainstream technology for switches and antenna tuners through at least 2025: they don’t see anything else replacing it. There is still increasing demand for 8” wafers. 12” wafers growth comes from integrated switch/LNAs, which comes from the Tier 1’s. In the front-end space, Murata leads in dollars by far, followed by Skyworks and others. Though mmWave is not yet clear, there are opportunities for RF-SOI.
STMicroelectronics has a long history in RF-SOI, noted Laura Formenti, dating back to 2000 when they started collaborating with Soitec and Leti. An IDM, ST also offers foundry services. For 5G, their foundry offering includes H9SOIFEM, C65SOIFEM and SOI mmW for high-performance analog, dedicated RF processes for RF switches, LNA, PA plus RFFEM. Then they have FD-SOI for RF, mixed signal and digital integration. From antennas to transceivers there’s an opportunity for full integration. For infrastructure, it depends on the customer preferences. 12” C65SOIFEM was introduced in 2019, and 12” SOIMMW will be introduced in 2020. Both their fabs at Crolles and Rousset, France, are in production. H9SOIFEM is for 4G and 5G sub-6GHz RF FEMs, enabling monolithic integration of the PA, LNA and switches, which is especially good for wifi. The C65SOIFEM is high-performance.
Danni Song said its a question of available spectrum. In sub-6, you get the same level of coverage with fewer base stations; also, sub-6 is much more mature. When will mmWave be ready? It depends. In the US, yes, but in China the spectrum allocation for mmWave has not yet been done. So it’s a wait and see for the industry to be ready. Peter Rabbeni agreed, adding that in the US sub-6 is crowded and conflicts with military bands. Paul Hurwitz added that mmWave is for fixed wireless access. Michael Reiha added that mmWave has advanced a lot even in the last few months (Verizon in Washington, DC, for example), so there is momentum.
Q: Does China lead in the sub-6GHz opportunity?
Jim Cable said that at pSemi they have two business models: mobile and infrastructure. He sees massive MIMO in base stations as huge (though in mobile their role is more supporting Murata). Peter Rabbeni added that they’re working with innovative partners in China, and that GF also offers skills in services and packaging. Yangyang Peng sees big opportunites with 5G for SmarterMicro and China. Paul Hurwitz has seen an increase in the capabilities of RF companies in China, and that the market in China moves faster than elsewhere. Michael Reiha sees China as strategic, and because there is central deployment, they can plan with the right partners.
Q: Data usage will be huge – what will it cost to individual users?
Danni Song said 5G phones will be expensive, but consumers want them, so we need to bring down costs and increase performance – but what about power consumption? Power needs to come down, maybe levering things like sleep mode more. For 3G & 4G, she noted, they had lots of time. People are pushing hard for 5G, but there’s a need for patience. Yangyang Peng said he didn’t want to pay more than for 4G.
In summary, Jim Cable noted that mmWave will demand huge amounts of silicon, to which Paul Hurwitz agreed, and Michael Reiha said Soitec will be ready. Everybody agreed that 3D packaging would be very important, especially for mmWave.
And that’s it for our coverage of the morning. Next up we’ll cover the presentations given during the afternoon.