FD-SOI At The Edge

Experts at the Table, Part 3: Pushing fully depleted SOI to 10/7nm for AI at the edge; what’s missing in the supply chain.


Semiconductor Engineering sat down to discuss changes in the FD-SOI world and what’s behind them, with James Lamb, deputy CTO for advanced semiconductor manufacturing and corporate technical fellow at Brewer Science; Giorgio Cesana, director of technical marketing at STMicroelectronics; Olivier Vatel, senior vice president and CTO at Screen Semiconductor Solutions; and Carlos Mazure, CTO at Soitec and chairman of the SOI Consortium. What follows are excerpts of that conversation. To read part one, click here Part two is here.

L-R: James Lamb, Giorgio Cesana, Olivier Vatel, Carlos Mazure.

SE: What happens as FD-SOI moves down to 12nm? Do we encounter some of the same problems as with finFETs?

Mazure: From a radiation hardness standpoint, FD-SOI is going to be better because that is tied to the amount of silicon volume. From an NBTI standpoint, there is no data showing FD-SOI transistors getting worse as they shrink.

Lamb: Some of that depends on how densely you’re going to package everything, though.

Mazure: There has been work done done at Leti showing FD-SOI down to at least 10nm, which corresponds to 7nm for the foundries. There is no indication that transistors degrade faster or bring up new problems. There are no surprises there.

SE: Where do you see new markets for FD-SOI?

Cesana: AI is one new application. You get out of the traditional MCU market where you have a von Neumann machine. You have something else, as well. But what you do with it it depends on the application.

Mazure: This is particularly important where you have an application at the edge.

Cesana: You can use this for how a big motor is running. So you put several sensors in there, taking vibration, speed, and so forth. At the end, you want more analysis inside. You cannot do that with an MCU. You need something else, which is based on deep learning. It cleans the signal, then sends it out to something else which does the full analysis.

SE: So basically you’re cleaning the data through pre-processing?

Cesana: Yes.

Lamb: We see that, as well. With remote sensors, some of those are outdoors. We don’t want a massive power supply, so they need to be running off a battery or solar or some other type of power generation. The data needs to be cleaned so it can go to where it can be connected up to a live circuit.

SE: When you put multiple sensors together, sometimes there are unexpected interactions between different sensors. Does the insulation layer in FD-SOI help solve that?

Lamb: Not really, because we condition the output of the sensors, so we’re providing exceptions and alarms. That’s a typical WiFi or Bluetooth sensor.

Mazure: But FD-SOI is getting a lot of attention with Industry 4.0. European-wide, ecosystem-wide programs are being put into place by materials and equipment companies, IDMs, foundries, design companies. They go all the way up to the systems houses with automotive and aerospace to co-optimize the application and see what needs to be put in the different nodes. It’s an interesting holistic view of driving the application.

SE: We used to think about data outliers, but tolerances are becoming so tight that more things are now outliers. That requires a much bigger change from across the supply chain.

Mazure: Until now we have been focusing less on the supply chain. We had to get the applications running. But now things are moving. There is a strong pull from different segments. There is a need for more capacity. So now we have to focus more on the supply chain, the equipment, the materials. We don’t want that to limit the growth of this technology. We started a discussion with SEMI this year. We will continue doing this. The theme today is the SOI supply chain.

Vatel: We are at the bottom of the food chain. We are starting to integrate and understand the need of FD-SOI. We are trying to find ways to differentiate in products. We are not yet there today. We are getting more involved in this process and looking at what can trickle down to the design community.

Lamb: On the materials side, it’s very similar. We’ve had to start looking into design and layout very early on, trying to prepare for markets for transistor and device design. A lot of our focus was on lithography in the past, but you’re getting behind if you stay there. We have to work with designers and where things are going and when they need a certain technology in place. We have to integrate across a wider bandwidth just to be effective in the industry.

Mazure: If you look at the substrate, it’s no longer a simple substrate. SOI is the simplest version of an engineered substrate. You’re adding more into the substrate for functionality. We need to understand what will be the applications to make sure we translate that into the proper substrate solution. That will come with specs, with requirements, and this takes us to the equipment companies to make sure we can process these devices and meet the specifications.

Leave a Reply

(Note: This name will be displayed publicly)