Incize: How to Get the Best Switch Performance with RF-SOI (CEO interview)
Posted date : May 12, 2020

With the CAGR for switches in cell-phone RF front-end modules running at >9% for the next five years, new players want to get in on the action, and established players want to up the ante. The specialists at Incize help wafer suppliers, foundries and fabless companies maximize switch performance starting at the substrate level. CEO Mostafa Emam explains how.

SOI News (SN): Can you tell us about the role Incize plays in the RF-SOI ecosystem?

Incize CEO Mostafa Emam. (Photo courtesy: Incize)

Mostafa Emam (ME): Our clients are wafer suppliers, foundries and fabless companies. The services we offer are testing and modeling of substrates, with the vision of what will happen in the value chain. Based on what the customer will do with the substrates, we do testing and modeling to improve the technology and tune their processes. Although the big players have teams devoted to this, we can add a layer of characterization that they have no expertise in.

Some of our customers are foundries that have been using bulk silicon, but now see opportunities in RF-SOI. But they’re starting from scratch and we help them to adopt the technology. We help them understand the physics behind the technology so they can migrate from bulk to SOI. We help them develop test structures and evaluate their technology.

Then we create models for both fully-depleted and partially-depleted SOI with PD-SOI 130 nm or 60 nm technology dominating the RF front-end module market. We believe RF is an art and you need to see the whole picture.

SN: Can you tell us a bit about the history of Incize?

ME: The RF-SOI story started in the 1990s. Then came trap-rich RF-SOI wafers from Jean-Pierre Raskin’s team at UC Louvain, industrialized by Soitec. Our lab at UC Louvain became known for our expertise in RF-SOI, and in 2011 we created Incize as a spin-off. [Editor’s note: for more background, see this SOI Consortium article about the birth of trap-rich substrates and the company’s founding.] At first our characterization services were very diverse, but by 2014 we focused mostly on RF-SOI because of the big demand. In 2015 we started doing radiation hardness tests for space applications and a new business unit was created. In 2016 we started our modeling and PDK activity, followed the next year by work on GaN on Si.

In 2018, we started offering full support to RF-SOI newcomers, who were starting from scratch, usually smaller players in the RF market. It takes about two years to fully train the engineers, support the technology enhancements, design test vehicles, measure them and finally do the modeling and PDK. So some of these players are now fully established in the RF-SOI market and have contracts in place with big customers.

SN: In your presentations, you often say there is room for all. What do you mean by that?

ME: There is a big market for RF-SOI in the coming years. It can offer the low-power, the low-cost and the high performance. RF-SOI is the only mature technology that combines all of this today. It successfully competes with traditional III-V technology. More foundries want to employ RF-SOI. We show to them that it’s not black magic – you just need to know how it works.

There is a big market for RF-SOI in the coming years. More foundries want to employ #RFSOI. We show them that it’s not black magic – you just need to know how it works. - Incize CEO Mostafa Emam #5G #semiconductors Click To Tweet

On our side, we have the knowledge and the infrastructure. Our added value is that we can do advanced tests the customers can’t do.

So the foundry says there’s opportunities in switches, we’ll do this and develop it all with optimization for specific Ron and Coff [the figure of merit for RF switches]. The foundry develops an RF switch and aiming at certain performance (RonCoff). We help our customers during this development phase. Once the performance target is reached we start developing a model and a PDK.

There is enough demand for RF-SOI, as even entry-level cell phones have SOI chips. Some opt for a fast and low-cost solution. Many target “good enough”, although some target to compete against the big players – it’s a question of their business strategy. And this is where our added value comes in.

SN: Can you provide some more insight into how you see the RF market?

ME: The Front End Module (FEM) is a fast growing market, with increasing demand in terms of volume and performance. This includes antenna switches, LNAs, tuners, filters, etc. Historically, III-V materials have been used for their high performance and high power handling.

However, RF-SOI has become the material of choice, and the biggest driver is integration of the RF switch and LNAs in one chip. It’s not easy to integrate the power amplifiers (PAs) on the same chip (still being on III-V substrates). But as it decreases footprint and cost, there are those who’ll do it. There is no viable competition for SOI – nothing will replace it in the short term. There are other technologies, but they are long term. It’s a stable market with high demand.

SN: For those of us who are not RF experts, can you help us understand the technology?

ME: The switch is sort of the traffic light of the FEM, receiving and transmitting. The simplest RF switch can be composed of only four transistors. Transistors leak power so you need to determine your Ron Coff performance. [Editor’s note: Resistance on vs. Capacitance off, the RF switch figure of merit, is measured in femtoseconds and should be as low as possible. Psemi has a good video explaining it.] When the Coff capacitance is small, the switch is really off. When the On resistance Ron is small, it means low losses, and the switch is turned On. Ron and Coff is a compromise. And as there are many frequency bands and antennas, the FEM becomes very complex.

Another issue is power handling, since the switch is the first stage behind the antenna.

And finally, there is the question of switch linearity. Trap rich SOI wafers suppress harmonics so you have less distortion originating in the substrate. You have to model this – the designer needs to know. In addition to single tone harmonics you also get intermodulation, where, two or more high power signals at two different frequencies create distortion at other frequencies. The danger is that these parasitic signals can be so close that the filter can’t reject them and the useful signals get distorted. This was a killer for the switch created on the bulk substrate. Trap-rich RF-SOI fixed this. So now 100% of switches are on trap-rich SOI substrates. While it’s still a niche market, there is demand from customers for increasing the number of bands – that’s driving this market.

SN: And what happens as we move to 5G?

ME: There’s more and more pressure on the specs. It’s an art when anything changes. Moving from 3G to 4G required complete upgrade of the [foundry’s] models. With 4G, the specs are severe and the FEM must be built on trap-rich substrates. But 5G is not well defined, so the RF industry is taking their best shot. What is clear: the performance requirements get more challenging. Once the technology is understood, it can be implemented. But the foundries and the fabless need our help to do it fast and do it well. We create more value working together.

SN: How do you see things evolving?

ME: The number of foundries today doing switches on RF-SOI is increasing, and this will continue for the next few years. We saw this opportunity and invested in it. Our company is just ten people, and we are self-funded. We are swimming in business, but we have fun. And we’re getting recognition.

Any company with CMOS in place could adopt RF-SOI. But it’s a different mindset. We help with the transition.

~ ~ ~

Click here to read more feature articles in SOI News.

About the author

Adele Hars editor