(This article is based on an in-depth presentation Terry gave at the SOI Consortium’s Fully-Depleted Tech Workshop, held during VLSI-TSA in Taiwan, April 2013.  The complete presentation is freely available on the SOI Consortium website.)

~ ~ ~

FinFET Isolation Considerations and Ramifications — Bulk vs. SOI

By Terence Hook, Senior Technical Staff Member, IBM Semiconductor Research and Development Center

Fully-depleted transistor technologies, both planar and fin-type, are now in the mainstream for product designs. One of the many interesting topics in the new 3D FinFET technology is the approach to isolation. In this article, key elements that differentiate junction-isolated (bulk) and dielectric-isolated (SOI) FinFET transistors are discussed, encompassing aspects of process integration, device design, reliability, and product performance.

BULK VS. SOI BASICS

In bulk wafers, isolation is formed in a manner similar to planar devices, with implanted wells and shallow-trench-isolation oxide separating fins from one another.

With an SOI (silicon-on-insulator) wafer, however, the fins are formed in the silicon layer, the isolating dielectric is already extant, and no well implants are required.

Figure 1: Schematic representation of bulk junction and dielectric-isolated FinFETs

The most important differences in the devices formed in these two manifestations lie in the shape of the fin, the processes that determine the effective fin height, and the presence of doping, which consequently affects the device in many adverse ways such as the variability and the reliability.

The final realization of the full potential of fully-depleted FinFETs is dependent on optimally addressing the issues enumerated herein. Dielectric isolation is shown to provide superior characteristics in all of the above-named aspects. Figure 1 shows a schematic representation of FinFETs for the two isolation architectures, with the various critical points of distinction noted as are discussed below.

FIN SHAPE

Definition of the fins on an SOI wafer is relatively straightforward; vertical fin sidewalls may easily be obtained.

In a bulk-based process, as the spaces between the lower, electrically inactive portions of the fins must be filled with an insulator, some angling of the fin is required to prevent the formation of voids.

Figure 2: Typical bulk junction and dielectric-isolated FinFET fin profiles

Bulk and SOI fin profiles are pictured in Figure 2.  As tapering the fin compromises the subthreshold slope and degrades the effective drive current as well as the output conductance, minimization of the taper is important to the electrical integrity of the device.

BULK: DOPING IN THE FIN

Whereas in an SOI design the transistor-transistor and subfin source-drain current paths are inherently interrupted by the dielectric layer, in a bulk-based process adequate doping for electrical isolation and latchup immunity needs to be established.  This requires additional masking levels and connections for electrical bias.

Conventional design criteria of doping, depth, and overlay tolerances apply to the deep interdevice isolation wells, but suppression of undesired current in the drain-source region has unique features in the FinFET configuration.

Suppression of punchthrough current requires some level of doping at least in the bottom portion of the fin. The adverse effects of doping on mobility and random-dopant-fluctuation have been reported; non-uniform doping is particularly egregious as it increases capacitance without a concomitant increase in drive current.

However, the level of doping required depends on the alignment of the gate and the source junction depth. An optimum choice for the conjunction seeks to minimize the dopant required while respecting physical process window constraints (see Figure 3).

Figure 3: Short-channel effects as a function of doping and gate recess depth relative to the source junction depth in bulk FinFETs

Another adverse effect of doping in the fin is the implication for the gate work function. For junction-isolated FinFETs, the gate metal work function is established so as to provide the desired threshold voltage in the presence of doping; for undoped dielectric-isolated FinFETs the appropriate work function is closer to midgap, which reduces gate leakage and improves reliability.

Figure 4: Voltage operating range as a function of fin doping

Between RDF-driven Vmin and work function-driven Vmax, the operating window of bulk FinFETs is more limited than that of undoped SOI FinFETs (see Figure 4).

PRODUCT AND CIRCUIT DESIGN CONSIDERATIONS

Designing with planar bulk technology has historically differed from planar SOI technology in three aspects: well contacts, self-heating, and floating body effects.

At the expense of area, planar bulk technology has enjoyed the advantages of controlling the threshold voltage through the well potential.  No such benefit exists in bulk FinFET devices, as it is not possible to influence the transistor through the well bias except in the spurious and undesirable region below the active fin.

In fully-depleted devices the concept of a floating body (charge storage in an isolated neutral region) is not applicable, so SOI and bulk FinFETs behave the same way for all switching scenarios.

Self-heating effects, while not important for fast switching operation, can be relevant for DC circuits. While large-area planar structures will continue to enjoy the advantage in thermal conduction relative to SOI traditionally observed, bulk and SOI FinFETs have very similar self-heating characteristics, as the only difference in thermal conductance is a tall, thin sliver of silicon, which provides only a small increase in thermal conductance.

While bulk FinFET technology has lower soft error rates than planar bulk technology, SOI FinFETs are better yet.

VARIATIONS

Fin height variation has a much more serious impact than the planar analog of transistor width variation. Wide transistors (i.e., many fins) have the same variation as narrow (i.e., few fins).

Figure 5: Calculated dependence of SOI and bulk transistors on key process variations, and relative variations in the two architectures

Whereas in the SOI-based version the electrical fin height is determined by the starting silicon thickness, in the bulk-based FinFET process the fin height is determined by several processes, and the distinction between “active” and “inactive” fin is blurred by the conjunction of the gate alignment with the source junction.

The sensitivities to various key variables have been calculated with hardware-calibrated 3D simulations, and the variation of those key parameters determined with respect to state-of-the-art processes (see Figure 5).

The fin variation-driven performance tolerance of a bulk FinFET is larger than that of an SOI FinFET.   That benefit of SOI is not only found in sort yield and worst-case design corners, but smaller variation within a chip enables a faster chip for any given level of leakage.

CONCLUSION

Complete realization of the benefits of fully-depleted transistor architecture is affected by the choice of isolation. Increased range of operating voltage, process simplification, reduced variation, lower soft error rate, and higher circuit density are all features of a dielectric-isolated architecture.

For these reasons the ability of an SOI-based FinFET to reap the full benefits of fully depleted transistors is demonstrably superior to a doped, bulk-based implementation.

~ ~

~  ~

Hearing the news that GlobalFoundries would be offering two flavors of FD-SOI, ASN asked the company to explain the strategy further. Here are the responses provided by Subi Kengeri, Vice President of Advanced Technology Architecture.

Subi Kengeri, VP Advanced Technology Architecture, GlobalFoundries

What do you see as the FD-SOI benefits for chip designers?

• Lower SRAM Vmin for retention and lower operating Vmin for Logic
• Wider range of Voltage operation for performance/power trade-off
• Total dielectric isolation equates to lower capacitances, lower leakage, and latch-up immunity
• Ultra-thin silicon film provides excellent electrostatic control and optimum transistor performance
• Back-bias control gives an additional speed boost
• Simple planar process using same front end and back end as our 28SLP process, which means fewer process steps and fewer masks, helping to absorb the additional substrate cost

What are your plans for making FD-SOI available to your customers?

We are the manufacturing partner for ST’s FD-SOI technology. We also are planning to offer the technology to other customers who may be interested, but we have not announced details yet. We are the only pure-play foundry with deep experience in both bulk and SOI technologies, which allows us to offer a broader range of technologies at advanced nodes.

GlobalFoundries’ Fab 8 in upstate NY

Can you elaborate on the “maximum” version of FD-SOI — tuned for specific applications — what sorts of things would those be?

Examples of features in the Maximum version of FD-SOI:
a. Back-bias capability on logic for higher performance
b. Denser SRAM by taking advantage of lesser variability of Fully depleted device
c. Base Vts tuned for specific applications (performance vs power trade-off)

And the “minimum” version — a simple and “out of the box” FD-SOI technology — who/what is this for?

a. No Back-bias supported
b. All SRAMs are foot-print compatible to 28SLP
c. Fully depleted device offers better Vmin and power advantages: Optimized for Mobile Applications

Are there any special logistics in terms of the PDK, IP, etc?

a. PDKs are similar to bulk CMOS, except the models will support a 4-terminal device for Back-bias
b. In the base version (termed as minimum version above), IP’s Physicals are fully compatible with bulk CMOS, but would require electrical re-characterization to take advantage of improved FD-SOI device characteristics
c. In the extended version (termed maximum version above), IPs will be designed to take advantage of Back-bias for better performance/power trade-offs in specific applications

What is the next node, and when will that roll out?

See slide 8 of [this] presentation:

~  ~  ~

~  ~

In a recent press release, the SOI wafer leader Soitec said that chips built on its SOI wafers were found in over half of the smartphones and tablets in the market worldwide. 50%? That’s a lot! How do they figure that? The answer: RF.

As seen here, RF chips account for a large part of cellphone components. (Source: Soitec & UCL, ESSDERC ’12 RF Workshop)

With all the talk right now about FD-SOI for application processors, the importance of the RF chips might seem to get a little lost. Don’t expect that to last.

Soitec’s wafer shipments for RF apps have increased by 400 percent in the last two years. In their current fiscal year (which ends this month), the company says it will have shipped over 200,000 engineered wafers to customers making chips for mobile comm. Those wafers translate into about 2.5 billion ICs for RF front-end module apps, which covers half of the 600 million smart phones and 100 million tablets expected to be produced this year.

Soitec, of course, does several flavors of SOI (including bonded silicon-on-sapphire aka BSOS, and high-resistivity (HR) SOI, which Soitec markets as their Wave SOI™ product line) as well as epitaxial GaAs wafers. It all adds up.

If app chips are the heart of the smartphone, RF is the soul. But in terms of chips and substrates, the RF side of the mobile world is much more complicated than the app side. Different functions have different needs, and those needs have traditionally been best met by disparate starting substrates. Devices can have eight of more chips and modules, and the chips in any given set can have different starting substrates, depending on the critical parameters.

The advent of LTE – “long-term evolution” aka 4G – will have a phenomenal impact on the RF components market, with analysts looking for RF components to almost double in value over the next five years. Look for an alphabet soup of new chip modules designed to handle the enormous complexity of evermore frequency bands.

Front-end modules (FEMs), which handle the back-and-forth of signals between the transceiver and the antenna, already contain multiple parts, including switches, power amplifiers, antenna tuning, power management and filters. With FEM real-estate reduction tracking at 15%/year and market growth continuing to increase at 15%/year for at least another five years, the quest is on for better, cheaper FEM solutions. Some are targeting SoCs, some will be multi-chip modules.

(Source: Soitec & UCL, ESSDERC ’12 RF Workshop)

A couple years ago, Soitec put together a really useful white paper on substrate technologies for RF. You can see, for example, that in choosing a substrate for switches, linear resistivity is the key parameter. This is something that can be addressed by several substrates, including GaAs, SoS and HR-SOI: the deciding factors are the trade-offs between performance and cost.

(Source: Soitec & UCL, ESSDERC ’12 RF Workshop)

There are huge opportunities in RF for the greater SOI & engineered substrates communities, so in coming issues of ASN, this is a topic we’ll be covering more. Upcoming articles by Professor Jean-Pierre Raskin of UCL (his group is working on a new generation of HR-SOI with enhanced signal integrity), as well as Peregrine and Skyworks, among others, are in the works.

Stay tuned!

~ ~ ~

~  ~

The 2013 Common Platform Technology Forum showcased “the latest technological advances being delivered to the world’s leading electronics companies,” so of course SOI-based topics were well-represented. Happily, those of us who weren’t able to get over to Silicon Valley were able to attend “virtually” via a live stream (which is now reposted – click here to register and watch it yourself).

The Common Platform Alliance is IBM, Samsung and GlobalFoundries, operating, as IBM’s Dr. Gary Patton points out, as a “virtual IDM”.

Here’s a round-up of the SOI-based highlights.

DR. GARY PATTON, VICE PRESIDENT OF SEMICONDUCTOR RESEARCH & DEVELOPMENT CENTER, IBM

In his keynote address, Gary covered the following SOI-based innovations:

Flexible computing with FD-SOI. (Courtesy: IBM, Common Platform Technology Forum 2013)

• FinFETs: As ASN readers know, IBM is driving FinFETs very hard. With ARM & Cadence, they taped out their first 14nm FinFET processor last fall (on SOI). Gary’s talk gave an overview of the evolution of device structures, including PD-SOI (the basis for IBM’s Watson supercomputer), FD-SOI, FinFETs and future structures and materials.

• Wearable electronics & folding displays – IBM has developed a new, low-cost technique that starts with the FD-SOI technology developed with ST and Leti, for manufacturing silicon-based electronics on a flexible, plastic substrate. Gary showed a sample, and said that “research suggests that flexible, affordable electronics can be made with conventional processes at room temperature.”

• Silicon nanophotonics – most all of the industry’s nanophotonics work is on SOI, and IBM is no exception here.  As Gary notes, “…the key innovation isn’t just the technology…it’s the fact that it’s commercial and scalable…”.
• Carbon nanotubes breakthrough – IBM has attained 10,000 working nanotube transistors on a single device using standard semiconductor processes.  As we noted in ASN when this news broke last fall, IBM researchers fabricated trenches made of hafnium oxide onto SOI wafers, which allows the self-assembly by the carbon nanotubes into neat rows rather than a spaghetti-like tangle.

As seen here, carbon nanotubes start on an SOI wafer. (Courtesy:IBM, Common Platform Technology Platform 2013)

MIKE NOONEN, EXECUTIVE VP, GLOBAL SALES, MARKETING, QUALITY & DESIGN, GLOBALFOUNDRIES.

In Mike’s keynote on particularly innovative customers, he covered ST’s FD-SOI technology.  Here are the main points he made about it:

• STMicroelectronics has been a partner in the Common Platform.
• FD-SOI leverages 80% FEOL of the 28nm SLP; the BEOL is identical to 28nm LP.
• “You can really dial-in optimal transistor performance,” he said.  The thin silicon channel introduces “interesting and exciting capabilities”, including:
  - lower leakage, lower capacitance, enhanced latch-up immunity, electrostatic control;
  - speed boost through back biasing;
• This technology is a simpler planar process:
  - reduced masks offsets cost;
  - considerable IP reuse.
• With a nod to Soitec, the world-leader in SOI wafers, he said, “Soitec has been a really enthusiastic evangelist of this technology, and I really want to acknowledge their efforts in making Fully-Depleted over SOI something that the industry has become very excited about.”  He added that they’re joined by MEMC and SEH as SOI substrate suppliers.
• Regarding the roll-out, he concluded, “A PDK of this technology is available this quarter, and GlobalFoundries has partnered with ST for volume manufacturing and will be entering risk production in the 4th quarter of 2013, with volume production in the first half of 2014.”

GlobalFoundries’ keynote highlights FD-SOI. (Courtesy: GlobalFoundries, STMicroelectronics, Common Platform Technology Forum 2013)

HANDEL JONES, OWNER & CEO, INTERNATIONAL BUSINESS STRATEGIES

In a “fireside chat” with Brian Fuller, Silicon Valley Bureau Chief, EETimes, Handel Jones touched on a number of SOI-related topics.  (In case you missed it, Handel recently wrote an excellent article for ASN on FD-SOI vs. Bulk & FinFET economics.) In addition to his general discourse on the impact of design & process issues on cost/gate, the importance of the ecosystem, and general industry outlook, here are some of Handel’s SOI-related observations during the forum chat:

• RF: he is particularly impressed with IBM’s work on RF, which he says is “…doing extremely well.”  As you may have seen previously in ASN, IBM’s CMOS 7RF SOI technology, which the company says offers significant cost advantages to designers of mobile handsets, has been on SOI for over five years.
• FD-SOI: When asked about any single, major disruption on the horizon, he noted that designing with FinFETs for mixed signal is tough, so there may be a delay there.  However, FD-SOI looks very positive, he says. He sees FD-SOI offering lower power, lower cost/gate, re-usable IP and scalability to 14nm.

~~

We said it was a game changer, and ST-E’s put together a really good page on their website that shows how they’re doing it.

By way of reminder, the NovaThor L8580 integrates an eQuad 2.5GHz processor (the mobile industry’s fastest) based on an ARM Cortex-A9, an Imagination PowerVR SGX544 GPU running at 600Mhz and an advanced multimode LTE modem on a single 28nm FD-SOI die.

The ST-E site is well worth looking at yourself – but in the meantime, here are a few of the highlights they’re sharing:

* With FD-SOI, you can do much more with body-biasing (aka back-biasing) than you can in bulk (which suffers from too much leakage). Thanks to the ultra-thin insulator layer in FD-SOI, the biasing creates a buried gate below the channel, so it effectively acts like a vertical double gate transistor. This facilitates the flow of electrons, leading to a higher voltage in the body, and faster switching of the transistor.

(Courtesy: ST-Ericsson)

* With FD-SOI, you can hit higher speeds with lower operating voltages. This is because the buried oxide layer prevents electrons from leaking away as they travel through the channel from the source to the drain (this sort of leakage is a major source of power consumption in 28nm bulk, which depends on doping to handle leakage). Interestingly, this graph shows ST-E going down to 0.5V – which is incredibly impressive.

(Courtesy: ST-Ericsson)

* As the (now award-winning) folks over at ST and Leti described for us a few years ago, designing a good SOC involves using the right blend of low, standard and high-Vt devices according to the target application and how it’s being used at any given time. The ST-E designers use this feature to apply different voltages independently to the top and the buried gates of the FD-SOI transistor, which effectively changes its characteristics. By choosing optimal combinations of the voltages, the transistor characteristics can be transformed from those of a very high-performance transistor to those of a very low-power transistor. A processing core built up of such transistors can operate as if it were in fact two cores – one optimized for high performance and the other for low power. (You can’t do this with FinFETs, btw.)

(Courtesy: ST-Ericsson)

Now are you starting to see why it’s a game changer?

—Adele Hars is the editor in chief of Advanced Substrate News.

~  ~

What a great start to 2013: at CES in Las Vegas, ST-Ericsson announced the NovaThor™ L8580 ModAp, “the world’s fastest and lowest-power integrated LTE smartphone platform.” This is the one that’s on STMicroelectronics’ 28nm FD-SOI, with sampling set for Q1 2013.

And it’s a game changer – for users, for designers, for foundries, and for bean counters.  Here’s why.

The NovaThor L8580 integrates an eQuad 2.5GHz processor based on the ARM Cortex-A9, an Imagination PowerVR™ SGX544 GPU running at 600Mhz and an advanced multimode LTE modem on a single 28nm FD-SOI die.

ST-Ericsson’s NovaThor(TM) L8580 on ST’s 28nm FD-SOI features a 2.5Ghz eQuad(TM) app processor with ultra-low power consumption. (Courtesy: ST-Ericsson)

In the eQuad CPU architecture, each processor core can operate as a high-performance core or a very-low-power core, depending on what’s needed at the moment. Since all the eQuad cores can adapt to the needs of the user at any given time, there’s no need for the dedicated low-power cores found in other multi-core CPU architectures. Remember, the 2.5GHz cores in the L8580 are the mobile industry’s fastest, or conversely, at 0.6V in low-power mode, the industry’s most battery-friendly. With all 2.5GHz cores working together, expect blazing high-performance when you’re doing something like browsing the web. But when phone’s your pocket, those cores will take barely a sip of power.

The NovaThor L8580 is essentially a straight port from 28nm bulk to 28nm FD-SOI of the (very successful) NovaThor L8540, with just a bit of tweaking to fully leverage cool things you can do with FD-SOI, like biasing to increase performance and conserve power.

For the folks designing smartphones and tablets (and ultimately for the end-user), that port to FD-SOI gets the NovaThor L8580:

• CPUs running 35% faster and GPU and multimedia accelerators running 20% faster. In terms of multimedia performance, they’re supporting 1080p video encoding and playback at up to 60 frames per second, 1080p 3D camcorder functionality, displays up to WUXGA (1920×1200) at 60 frames per second and cameras up to 20 megapixels. (Hence their use of the descriptive “extraordinary”.)
• 25% less power consumption than rival architectures when running at high-performance  levels – think Cooler Operation.
• A low-power mode can deliver up to 5000 DMIPS at 0.6V – more than enough computing power for the majority of applications in everyday use. A key point here is that it enables stable SRAM operation at 0.6V – have you heard of anyone matching this? The result is that this low-power mode consumes 50% less power to deliver the same performance compared with alternative solutions in bulk CMOS.

It all adds up to big battery savings – this is the extra day CEO Didier Lamouche promised us in Barcelona last year when they announced this chip.

Click here to view the embedded video.

ST-Ericsson has posted an amazing video, filmed live at CES 13. In the first part of the demo (re: high-perf), on a Samsung Galaxy S3, they’ve got the Sky Castle 3D Graphics Demo launching twice as fast on FD-SOI as the bulk equivalent, and hitting 2.8GHz! And in the second demo (re: low power), they’re hitting 1GHz using just 0.636V, which would take 1.1V on bulk.

Design Highlights

For the ST-E designers, most of the IP blocks were directly re-used from the bulk design, so the porting to FD-SOI was extremely simple and fast.

For the manufacturing folks over at STMicroelectronics (and starting this year, at GloFo), FD-SOI is a planar technology that re-uses 90% of the process steps used in 28nm bulk. The overall manufacturing process in FD-SOI is 12% less complex, so they’ve got lower cycle time and reduced manufacturing costs (bean counters take note, please). They also point out that the manufacturing tools for FD-SOI are much simpler than those required for FinFETs.

Wondering what’s next? The 14nm FD-SOI node is already in development, the ARM Cortex-A15‘s  on the radar, and the FD-SOI roadmap is already defined up the 10nm node.

With FD-SOI, you can do much more with body-biasing (aka back-biasing) than you can in bulk (which suffers from too much leakage). Thanks to the ultra-thin insulator layer in FD-SOI, the biasing creates a buried gate below the channel, so it effectively acts like a vertical double gate transistor. This facilitates the flow of electrons, leading to a higher voltage in the body, and faster switching of the transistor. (Image courtesy ST-Ericsson)

With FD-SOI, you can hit higher speeds with lower operating voltages. This is because the buried oxide layer prevents electrons from leaking away as they travel through the channel from the source to the drain (this sort of leakage is a major source of power consumption in 28nm bulk, which depends on doping to handle leakage). Interestingly, this graph shows ST-E going down to 0.5V – which is incredibly impressive. (Image courtesy of ST-Ericsson)

(Image courtesy ST-Ericsson)

(Image courtesy ST-Ericsson)

As the (now award-winning) folks over at ST and Leti described for us a few years ago, designing a good SOC involves using the right blend of low, standard and high-Vt devices according to the target application and how it’s being used at any given time.  The ST-E designers use this feature to apply different voltages independently to the top and the buried gates of the FD-SOI transistor, which effectively changes its characteristics. By choosing optimal combinations of the voltages, the transistor characteristics can be transformed from those of a very high-performance transistor to those of a very low-power transistor. A processing core built up of such transistors can operate as if it were in fact two cores – one optimized for high performance and the other for low power. (You can’t do this with FinFETs, btw.)

Just Posted: FD-SOI video & white paper

Just as this blog was going online, ST-Ericsson posted an excellent, in-depth white paper; and in partnership with STMicroelectroics, a YouTube video detailing the how’s and why’s of FD-SOI.Here are the links — you really don’t want to miss these:

• Multiprocessing in Mobile Platforms: the Marketing and the Reality
In this white paper, ST-Ericsson’s Marco Cornero and Andreas Anyuru “…illustrate and compare the main technological options available in multiprocessing for mobile platforms, highlighting the synergies between multiprocessing and the disruptive FD-SOI silicon technology used in the upcoming ST-Ericsson products.”

• An Introduction to FD-SOI

STMicroelectronics and ST-Ericsson have teamed up on this excellent video, which garnered 1250 views within the first four days of its posting on YouTube. The animations and comparisons highlight why FD-SOI is so fast, and so cool.

~ ~

~  ~

If you want to cut through the noise surrounding the choices for 28nm and beyond, an excellent place to start is the SOI Consortium’s Fully Depleted Technology Symposium.

As a member of the design and manufacturing communities, this is your chance to see and hear what industry leaders are actually doing. Planar? FinFET? The Consortium’s been doing these symposia during major conferences for going on four years now, and lively debates always ensue.

(Courtesy: Hilton Hotels & Resorts)

This next FD Tech symposium happens the first day of the IEEE’s IEDM conference in San Francisco – Monday, December 10^th at 8:15pm. Conveniently, it’s also taking place in the same building – at the SF Hilton.

Top technologists from STMicroelectronics, ST-Ericsson, IBM, ARM, Altera, LETI, Soitec, MEMC and others will be debating comprehensive Fully-Depleted Technology solutions.

But perhaps most importantly, we’re going to get the first product-level benchmarking results of 28nm FD-planar for mobile SoC and FPGA applications.  That’s silicon proof straight from the companies who are doing it.

If you’ve been following recent ASN postings from STM, ST-Ericsson, IBM and others, you know these folks are really excited about the results they’re seeing.

Here’s a peak at the presentations planned for the symposium:

• Planar Fully-Depleted Technology at 28nm and below for extremely power-efficient SoCs:  SoC level 28nm Planar Fully-Depleted silicon results
  By Joel Hartmann, Executive VP Front-End Manufacturing & Process R&D, STMicroelectronics
• Evaluation and benchmarking of 14nm planar Fully-Depleted Technology for FPGAs
  By Jeff Watt, Ph.D. Fellow, Technology Development, Altera Corporation
• Challenges and comparisons of designing power-efficient SoCs with planar Fully Depleted transistors and FinFETS
  By Rob Aitken, ARM Fellow
• Second-generation FinFETs and Fin-on-Oxide
  By Ed Nowak, IBM Distinguished Engineer and Device Chief Designer, Semiconductor R&D Center, IBM Systems and Technology Group

The presentations will be followed by a Q&A.

Admission is free, but space is limited, so you must reserve in advance – click here to go to the special registration site.

To recap, it’s the:

Fully-Depleted Transistors Technology Symposium
Hilton San Francisco Union Square Hotel (333 O’Farrell St.)
Monday, December 10^th, 2012
8:15pm to 10:30pm

Food & refreshments will be provided.

We won’t all be in San Francisco, so if you can’t get there, the presentations will be posted on the SOI Consortium website (you can also get the presentations from previous events there, too, as well as excellent white papers).

If you do go and want to share your reactions on Twitter, use #FDchipTech and @soiconsortium.

This will be a great event – don’t miss it!

~~

~  ~

What would a port to 28nm FD-SOI do for your design?  A recent announcement by CMP, STMicroelectronics and Soitec invites you to find out.  Specifically, ST’s CMOS 28nm Fully Depleted Silicon-On-Insulator (FD-SOI) process – which uses innovative silicon substrates from Soitec and incorporates robust, compact models from Leti – is now available for prototyping to universities, research labs and design companies through the silicon brokerage services provided by CMP (Circuits Multi Projets®). ST is releasing this process technology to third parties as it nears completion of its first commercial FD-SOI wafers.  What you can get from CMP is the same process technology that will be available to all at GlobalFoundries in high-volume next year.

The CMP multi-project wafer service allows organizations to obtain small quantities of advanced ICs – typically from a few dozen (for a prototype, say) to over a hundred thousand units (for low-volume production). CMP is a non-profit, non-sponsored organization created in 1981, with a long history of offering SOI and other advanced processes. It offers industrial quality process lines – with industrial-level, stable yields. Headquartered in Grenoble, France, CMP has over 1000 clients in 70 countries.

The cost of ST’s 28nm FD-SOI CMOS process at CMP has been fixed at 18,000 €/mm², with a minimum of 1mm².  At this point in scaling, that gets you about two million gates – about eight million transistors.  So the pricing is very aggressive for an advanced technology node – and it comes down if you get more than 3mm², and even more if you get >15mm², Kholdoun Torki, CMP Technical Director explained to ASN.

Dr. Torki was kind enough to elaborate a bit on the particulars for us. Here’s what he says. The ST design kit contains a full-custom part, and standard-cells and I/O libraries with digital design-flows supported under Cadence Encounter and Synopsys Physical Compiler. The design-kit is from ST Front-End Manufacturing and Technology, Crolles. CMP delivers this design-kit under NDA.

Devices are supported for UTSOI (ultra-thin SOI) models, which were developed by and are the property of Leti.

The UTSOI model is available under Eldo from Mentor and Hspice from Synopsys. It is also expected to be available for Spectre (Cadence) and for Golden Gate and ADS (Agilent) within the next few months.

CMP provides the first level support (installation, and general questions on the use of the kit). Multi-Projects Wafer runs are organized at ST Crolles. For low volume production, a quote is issued on a case-by-case basis, on request.

The ST 28nm FD-SOI offering has a true 28nm BEOL metallization with .1µ metal pitch, says Dr. Torki.

CMP also has offered the Leti 20nm FD-SOI R&D process since 2010. (In fact for those looking even further ahead, Leti has predictive model cards down to 11nm.) It is expected the 20nm FD-SOI process from ST, incorporating strategic technology from Leti, will be available from CMP towards the end of next year, although the exact date has not yet been fixed.

How it works

In Multi-Project Wafer runs, costs are shared (and reduced) because the reticle area is shared across customers. CMP offers one-stop shopping, including:

• NDA processing
• the design-kits linking CAD and processes, and related support
• Design submission, checking, and final database to the Fab
• Wafer sawing and Packaging
• Export license processing
• Chip delivery

Because reticles are shared across multiple designs, CMP customers benefit from very attractive pricing. (Courtesy: CMP)

Last year (2011), CMP handled 273 circuits, including prototypes, low-volume production runs and industrial applications.

For organizations like the 77 customers in 23 countries using 28nm bulk CMOS through CMP’s program, migrating from 28nm CMOS bulk to 28nm FD-SOI will be seamless, says Dr. Torki. There are no disruptions in process or design. There are the same layer numbers and names, so they can load a bulk design directly into an FD-SOI design environment. They use the common design-rules platform (ISDA alliance design-rules), and bulk devices can be co-integrated with FD-SOI devices as needed.

These are real, leading edge chips and circuits we’re talking about. Here’s what you get:

• 28nm HK/MG FD-SOI with ultra-thin BOX and ground plane
• 10 Cu metal layers: (6 thin + 2 medium + 2 thick)
• Triple Well (Deep N-Well allows the P-Well to be isolated from the substrate)
• Single IO oxide + Single core oxide.
• Double VT: 1.0V Low Vt transistors (LVT) + 1.0V super Regular Vt transistors (RVT)
• Low Leakage (high density) SRAM using LP core oxide
• IO supply voltage: 1.8 V using the IO oxide.
• Ultra Low k inter-level dielectric
• 0.10µ metal pitch
• Self-aligned silicided drain, source and gate
• Poly and active resistors: Silicide protection over active areas for ESD protection
• CMP for enhanced planarization (on STI, Contacts, Metals and vias).

FD-SOI Transistor (Courtesy: ST)

The 28nm FD-SOI standard-cells, IO cells and related IP are all from ST. The CORE cells Libraries include:

• CORE_LL: Low Power LVT
• CORE_LR: Low Power RVT
• CLOCK (LL and LR): Buffer cells and the same for clock tree synthesis
• PR: Place and route filler cells.

The IO cells Libraries include:

• Digital
• Analog
• Flip-Chip bumps
• ESD

You can find more details at the CMP website, or from the paper Dr. Torki presented at the 2012 SOI Conference.

So this represents a real opportunity.  Universities, often doing important research for industrial partners, have long known the value of using services like CMP’s. But with this latest ST-CMP-Soitec announcement, the fabless world can do more than kick the tires – they can take 28nm FD-SOI for a real test drive.

FD-SOI promises an extremely cost-effective, performance-enhanced, power-miser of a chip.  Wouldn’t you like to give it a try?

~~

~  ~

It’s a bright green light from the world leaders in SOI wafer capacity. Soitec, the world leader in SOI wafer production, and long-time partner Shin-Etsu Handatai (SEH), the world’s biggest producer of silicon wafers, have extended their licensing agreement and expanded their technology cooperation.

SEH is a $12.7 billion company, supplying over 20% of the world’s bulk silicon wafers. SEH’s relationship with Soitec goes way back: they were one of the original corporate investors back in 1997, and the first to license Soitec’s Smart CutTM technology for manufacturing SOI wafers.

With its 300mm SOI wafer production fabs in France and Singapore, Soitec has an expandable installed industrial base of two million wafers per year.

As Horacio Mendez, Executive Direct of the SOI Consortium told ASN, “This is a very significant announcement. The substrate supply chain is fully engaged: we have multiple independent suppliers that can clearly meet the market demands for all key sectors, including mobile devices. As the advanced technology nodes ramp, the wafer production is in place; and very importantly, the capacity is expandable to provide maximum flexibility to customers.”

SEH has been manufacturing standard SOI wafers using Smart Cut technology for years. And last year, the company said it had completed development of its ultra-thin BOX (aka UTB — the wafers used for planar FD-SOI) substrates. Nobuo Katsuoka, director of the SOI program at SEH, recently told Semiconductor Manufacturing & Design, “SEH is delighted to deliver the products on request.”

Wafers for FD-SOI (a “planar” “2D” technology) have Angstrom-level uniformity in their ultra-thin layers – so it’s excellent news that the the industry’s two leaders are both supply sources.

SOI wafers for FinFETs (a “vertical” or “3D” technology, for which the top silicon and insulating BOX layer don’t have to be ultra-ultra-thin) have also long been available from Soitec, SEH and other sources.

With respect to this announcement, SEH’s Katsuoka said, “We are very excited about the business opportunities for SOI products, and we look forward to working with Soitec to extend the global supply chain for new products, such as FD-SOI and SOI for FinFETs, which are showing potential benefits in mobile and embedded applications. Our relationship with Soitec has been a very positive and fruitful one, and we are excited to extend that collaboration. The unique features of Smart Cut will enable our two companies to jointly improve global output for existing and new SOI products.”

As Steve Longoria, SVP of WW Business Development at Soitec, told ASN, “The wafer is the front end of the manufacturing process. This announcement is a proof point of new energy for robust, multi-source supply for impending high-volume demand.”

BEYOND LOGIC

The newly announced Soitec-SEH agreement also extends the companies’ commitment to wafers for a broad-range of areas. For example, there are major market opportunities in SOI for RF devices, power, MEMS/sensors, photonics and more.

The agreement also extends to R&D for technologies of the next wave. We might think of Smart Cut as an SOI technology, but in fact it’s really a manufacturing technology that can be applied to a huge range of wafer materials. As a result of the extended agreement, SEH will continue to use Soitec’s industry-defining Smart Cut technology to manufacture SOI wafers.  What’s more, SEH will now also be able to extend its Smart Cut manufacturing capabilities to other materials, a trend commonly referred to as Silicon on Anything or SOA (any material on top of which there is a thin film of plain silicon), which will allow SEH to further expand its scope of applications.

So with an abundance of opportunities, a robust multi-source supply chain for the front end of the chip manufacturing process, top-quality wafers that enable savings and efficiencies – in short, better end-user value – it’s all systems go for high-volume demand.

This illustration shows how Smart Cut, Soitec’s proprietary engineered wafer technology, works. The industry standard, this revolutionary wafer bonding and layer splitting processes makes it possible to transfer a thin layer of material from a donor substrate to another substrate, overcoming physical limitations and changing the face of the substrate industry. The Smart Cut technology was originally developed by the CEA-Leti. Soitec holds exclusive exploitation of CEA-Leti rights into the Smart Cut technology, including the right to sublicense to SEH. The technology was made viable for SOI high-volume commercial production by Soitec, and is now protected by more than 3,000 patents owned or controlled by Soitec.

~  ~

In the spring of 2012, STMicroelectronics announced the company would be manufacturing ST-Ericsson’s next-generation (and very successful) NovaThor ARM-based smartphone/tablet processors using 28nm FD-SOI process technology. With first samples coming out this fall, ASN talks to Jean-Marc Chery, Executive Vice President, General Manager Digital Sector, Chief Technology & Manufacturing Officer, STMicroelectronics about the manufacturing process and the expected results.

Jean-Marc Chery, Executive Vice President, General Manager Digital Sector, Chief Technology & Manufacturing Officer, STMicroelectronics (Photo credit: Artechnic)

Advanced Substrate News (ASN): You taped out ST-Ericsson’s 28nm FD-SOI NovaThor in the beginning of September. Did that go as you expected?

Jean-Marc Chery, STMicroelectronics (JMC): 28nm FD-SOI is a pretty exciting technology, allowing better design optimization (for higher speed and power efficiency) than traditional bulk technologies, still reusing most of manufacturing bricks of planar 28nm LP technology and the same design flow and methodology.

Adoption of 28nm FD-SOI for ST-Ericsson’s NovaThor has not introduced any major difficulty in its design, and the FD-SOI version has been taped out shortly after the Low-Power bulk version. Of course special care has been dedicated to further optimize power, exploiting FD-SOI exceptional flexibility and low-power capabilities.

On the manufacturing side, FD-SOI does not introduce additional complexity: on the contrary, process steps are reduced and thus cycle time.

ASN: Can you talk about the results you expect to see or have seen in the chip? Is there anything about it, or perhaps about the ARM core in particular, that makes it especially well-suited to FD-SOI? Is there anything about the transistor back-biasing capability (which enables significant performance enhancements and power optimization) in the design that makes it challenging to manufacture?

JMC: The wide supply range (ranging from 1.2V down to 0.6V) with excellent performance, and extended back-biasing capability (allowing dynamic modulation of the transistor threshold voltage) offered by 28nm FD-SOI technology have allowed us to exploit the ARM implementation to offer an improved maximum frequency and reach an overall power reduction for the various operating modes of the SoC.

About back biasing, this is a standard feature of FD-SOI technology with no particular challenges for manufacturing. Of course, its dynamic usage to optimize operating points for power (or speed) requires an appropriate device architecture to fully benefit from it.

ASN: In the press, STMicroelectronics has indicated that the 28nm FD-SOI has better power and performance than the industry’s first-gen bulk 22nm FinFETs. Would you say that your choice of FD-SOI puts you in a position of strength, in that you’ll have the mobile industry’s leading technology for 28nm and a choice of mature technologies at 14nm?

JMC: 28nm FD-SOI technology is a unique offer in the SOC industry, allowing the introduction of a fully-depleted technology with a low-cost solution and in a timely manner.

28nm FD-SOI is a planar technology derived from 28nm LP bulk technology, with the same design rules and allowing direct layout reuse (or simplified porting) of basic building blocks and IPs, benefiting from inheriting their maturity level. Also on the manufacturing side, 28nm FD-SOI technology uses the same equipment as Low Power bulk CMOS in a simplified process flow. In ST/Crolles facility we are reaching yield levels comparable to 28nm LP bulk ones, proving that FD-SOI process does not introduce major yield detractors.

A smooth library and IP migration flow coupled with rapid availability for manufacturing is driving the success of this 28nm technology.

Looking at the technology roadmap, the same incremental step for the 14nm node is under development and is on track.

The STMicroelectronics fab in Crolles, France. (Photo credit: Artechnic)

ASN: The plan was to start production in your fab in Crolles, then shift to GlobalFoundries for high-volume production in 2013 — is this still the schedule? From a manufacturing standpoint, what does it take to get a fab ready for FD-SOI production (does it take much longer than a typical bulk scaling transition)? Are there any special tools or other preparations needed?

JMC: For manufacturing, 28nm FD-SOI technology uses the same toolset as for 28nm LP bulk. Process development is complete, and ST/Crolles fab is now working to bring yield at production levels and complete the qualification of the technology, as done for any other.

Phase-in of the technology at GlobalFoundries is planned to start Q1 2013, with process qualified and with production level yield foreseen for Q4 2013.

The ST Crolles fab is highly automated, and already runs a broad mix of products in addition to the new FD-SOI chips. The accumulated assets the company has invested in this fab will increase capacity to 4500 wafers/week by the end of 2014. (Photo credit: Artechnic)

ASN: Let’s talk about the Crolles fab for a minute. Although it may be considered small compared to the big pure-play foundries, some aspects you share with the big foundries – like a large mix of product and advanced automation, right?

JMC: Crolles’ technology mix encompasses Advanced CMOS 28/40 nm, Imaging Sensors, embedded Non Volatile Memories starting at 55nm for Microcontroller and Analog on CMOS 110nm. This mix optimizes very well the accumulated assets we have invested in this Fab toward 4500 wafers week capacity over the next two years.

ASN: How do you see the impact of STMicroelectronics’s decision on the industry? Do you expect others to follow? Will other companies be able to leverage your technology at your foundry partners?

JMC: We would like very much for others to follow us. Through GlobalFoundries, ST is making its FD-SOI technology available to anyone in the microelectronics industry. The ST wide set of silicon-proven 28nm foundation libraries and IPs, encompassing not only basic libraries (std-cells, srams, I/Os) but also complex AMS IPs, is also available to be licensed to those customers aiming for quick access to the technology.

~~