Part of the  

Solid State Technology


The Confab


About  |  Contact

Posts Tagged ‘Qualcomm’

Next Page »

Deep Learning Could Boost Yields, Increase Revenues

Thursday, March 23rd, 2017


By Dave Lammers, Contributing Editor

While it is still early days for deep-learning techniques, the semiconductor industry may benefit from the advances in neural networks, according to analysts and industry executives.

First, the design and manufacturing of advanced ICs can become more efficient by deploying neural networks trained to analyze data, though labelling and classifying that data remains a major challenge. Also, demand will be spurred by the inference engines used in smartphones, autos, drones, robots and other systems, while the processors needed to train neural networks will re-energize demand for high-performance systems.

Abel Brown, senior systems architect at Nvidia, said until the 2010-2012 time frame, neural networks “didn’t have enough data.” Then, a “big bang” occurred when computing power multiplied and very large labelled data sets grew at Amazon, Google, and elsewhere. The trifecta was complete with advances in neural network techniques for image, video, and real-time voice recognition, among others.

During the training process, Brown noted, neural networks “figure out the important parts of the data” and then “converge to a set of significant features and parameters.”

Chris Rowen, who recently started Cognite Ventures to advise deep-learning startups, said he is “becoming aware of a lot more interest from the EDA industry” in deep learning techniques, adding that “problems in manufacturing also are very suitable” to the approach.

Chris Rowen, Cognite Ventures

For the semiconductor industry, Rowen said, deep-learning techniques are akin to “a shiny new hammer” that companies are still trying to figure out how to put to good use. But since yield questions are so important, and the causes of defects are often so hard to pinpoint, deep learning is an attractive approach to semiconductor companies.

“When you have masses of data, and you know what the outcome is but have no clear idea of what the causality is, (deep learning) can bring a complex model of causality that is very hard to do with manual methods,” said Rowen, an IEEE fellow who earlier was the CEO of Tensilica Inc.

The magic of deep learning, Rowen said, is that the learning process is highly automated and “doesn’t require a fab expert to look at the particular defect patterns.”

“It really is a rather brute force, naïve method. You don’t really know what the constituent patterns are that lead to these particular failures. But if you have enough examples that relate inputs to outputs, to defects or to failures, then you can use deep learning.”

Juan Rey, senior director of engineering at Mentor Graphics, said Mentor engineers have started investigating deep-learning techniques which could improve models of the lithography process steps, a complex issue that Rey said “is an area where deep neural networks and machine learning seem to be able to help.”

Juan Rey, Mentor Graphics

In the lithography process “we need to create an approximate model of what needs to be analyzed. For example, for photolithography specifically, there is the transition between dark and clear areas, where the slope of intensity for that transition zone plays a very clear role in the physics of the problem being solved. The problem tends to be that the design, the exact formulation, cannot be used in every space, and we are limited by the computational resources. We need to rely on a few discrete measurements, perhaps a few tens of thousands, maybe more, but it still is a discrete data set, and we don’t know if that is enough to cover all the cases when we model the full chip,” he said.

“Where we see an opportunity for deep learning is to try to do an interpretation for that problem, given that an exhaustive analysis is impossible. Using these new types of algorithms, we may be able to move from a problem that is continuous to a problem with a discrete data set.”

Mentor seeks to cooperate with academia and with research consortia such as IMEC. “We want to find the right research projects to sponsor between our research teams and academic teams. We hope that we can get better results with these new types of algorithms, and in the longer term with the new hardware that is being developed,” Rey said.

Many companies are developing specialized processors to run machine-learning algorithms, including non-Von Neumann, asynchronous architectures, which could offer several orders of magnitude less power consumption. “We are paying a lot of attention to the research, and would like to use some of these chips to solve some of the problems that the industry has, problems that are not very well served right now,” Rey said.

While power savings can still be gained with synchronous architectures, Rey said brain-inspired projects such as Qualcomm’s Zeroth processor, or the use of memristors being developed at H-P Labs, may be able to deliver significant power savings. “These are all worth paying attention to. It is my feeling that different architectures may be needed to deal with unstructured data. Otherwise, total power consumption is going through the roof. For unstructured data, these types of problem can be dealt with much better with neuromorphic computers.”

The use of deep learning techniques is moving beyond the biggest players, such as Google, Amazon, and the like. Just as various system integrators package the open source modules of the Hadoop data base technology into a more-secure offering, several system integrators are offering workstations packaged with the appropriate deep-learning tools.

Deep learning has evolved to play a role in speech recognition used in Amazon’s Echo. Source: Amazon

Robert Stober, director of systems engineering at Bright Computing, bundles AI software and tools with hardware based on Nvidia or Intel processors. “Our mission statement is to deploy deep learning packages, infrastructure, and clusters, so there is no more digging around for weeks and weeks by your expensive data scientists,” Stober said.

Deep learning is driving new the need for new types of processors as well as high-speed interconnects. Tim Miller, senior vice president at One Stop Systems, said that training the neural networks used in deep learning is an ideal task for GPUs because they can perform parallel calculations, sharply reducing the training time. However, GPUs often are large and require cooling, which most systems are not equipped to handle.

David Kanter, principal consultant at Real World Technologies, said “as I look at what’s driving the industry, it’s about convolutional neural networks, and using general-purpose hardware to do this is not the most efficient thing.”

However, research efforts focused on using new materials or futuristic architectures may over-complicate the situation for data scientists outside of the research arena. At the International Electron Devices Meeting (IEDM 2017), several research managers discussed using spin torque magnetic (STT-MRAM) technology, or resistive RAMs (ReRAM), to create dense, power-efficient networks of artificial neurons.

While those efforts are worthwhile from a research standpoint, Kanter said “when proving a new technology, you want to minimize the situation, and if you change the software architecture of neural networks, that is asking a lot of programmers, to adopt a different programming method.”

While Nvidia, Intel, and others battle it out at the high end for the processors used in training the neural network, the inference engines which use the results of that training must be less expensive and consume far less power.

Kanter said “today, most inference processing is done on general-purpose CPUs. It does not require a GPU. Most people I know at Google do not use a GPU. Since the (inference processing) workload load looks like the processing of DSP algorithms, it can be done with special-purpose cores from Tensilica (now part of Cadence) or ARC (now part of Synopsys). That is way better than any GPU,” Kanter said.

Rowen was asked if the end-node inference engine will blossom into large volumes. “I would emphatically say, yes, powerful inference engines will be widely deployed” in markets such as imaging, voice processing, language recognition, and modeling.

“There will be some opportunity for stand-alone inference engines, but most IEs will be part of a larger system. Inference doesn’t necessarily need hundreds of square millimeters of silicon. But it will be a major sub-system, widely deployed in a range of SoC platforms,” Rowen said.

Kanter noted that Nvidia has a powerful inference engine processor that has gained traction in the early self-driving cars, and Google has developed an ASIC to process its Tensor deep learning software language.

In many other markets, what is needed are very low power consumption IEs that can be used in security cameras, voice processors, drones, and many other markets. Nvidia CEO Jen Hsung Huang, in a blog post early this year, said that deep learning will spur demand for billions of devices deployed in drones, portable instruments, intelligent cameras, and autonomous vehicles.

“Someday, billions of intelligent devices will take advantage of deep learning to perform seemingly intelligent tasks,” Huang wrote. He envisions a future in which drones will autonomously find an item in a warehouse, for example, while portable medical instruments will use artificial intelligence to diagnose blood samples on-site.

In the long run, that “billions” vision may be correct, Kanter said, adding that the Nvidia CEO, an adept promoter as well as an astute company leader, may be wearing his salesman hat a bit.

“Ten years from now, inference processing will be widespread, and many SoCs will have an inference accelerator on board,” Kanter said.

MRAM Takes Center Stage at IEDM 2016

Monday, December 12th, 2016


By Dave Lammers, Contributing Editor

The IEDM 2016 conference, held in early December in San Francisco, was somewhat of a coming-out party for magneto-resistive memory (MRAM). The MRAM presentations at IEDM were complemented by a special MRAM-focused poster session – organized by the IEEE Magnetics Society in cooperation with the IEEE Electron Devices Society (EDS) – with 33 posters and a lively crowd.

And in the opening keynote speech of the 62nd International Electron Devices Meeting, Seok-hee Lee, executive vice president at SK Hynix (Seoul), set the stage by saying that the race is on between DRAM and emerging memories such as MRAM. “Originally, people thought that DRAM scaling would stop. Then engineers in the DRAM and NAND worlds worked hard and pushed out the end further in the future,” he said.

While cautioning that MRAM bit cells are larger than in DRAM and thus more more costly, Lee said MRAM has “very strong potential in embedded memory.”

SK Hynix is not the only company with a full-blown MRAM development effort underway. Samsung, which earlier bought MRAM startup Grandis and which has a materials-related research relationship with IBM, attracted a standing-room-only crowd to its MRAM paper at IEDM. TSMC is working with TDK on its program, and Sony is using 300mm wafers to build high-performance MRAMs for startup Avalanche Technology.

And one knowledgeable source said “the biggest processor company also has purchased a lot of equipment” for its MRAM development effort.

Dave Eggleston, vice president of emerging memory at GlobalFoundries, said he believes GlobalFoundries is the furthest along on the MRAM optimization curve, partly due to its technology and manufacturing partnership with Everspin Technologies (Chandler, Ariz.). Everspin has been working on MRAM for more than 20 years, and has shipped nearly 60 million discrete MRAMs, largely to the cache buffering and industrial markets.

GlobalFoundries has announced plans to use embedded STT-MRAM in its 22FDX platform, which uses fully-depleted SOI technology, as early as 2018.

Future versions of MRAM– such as spin orbit torque (SOT) MRAM and Voltage Controlled MRAM — could compete with SRAM and DRAM. Analysts said today’s spin-transfer torque STT-MRAM – referring to the torque that arises from the transfer of electron spins to the free magnetic layer — is vying for commercial adoption as ever-faster processors need higher performance memory subsystems.

STT-MRAM is fast enough to fit in as a new memory layer below the processor and the SRAM-based L1/L2 cache layers, and above DRAM and storage-level NAND flash layers, said Gary Bronner, vice president of research at Rambus Inc.

With good data retention and speed, and medium density, MRAM “may have advantages in the lower-level caches” of systems which have large amounts of on-chip SRAM, Bronner said, due in part to MRAM’s smaller cell size than six-transistor SRAM. While DRAM in the sub-20nm nodes faces cost issues as its moves to more complex capacitor structures, Bronner said that “thus far STT-MRAM) is not cheaper than DRAM.”

IBM researchers, which pioneered the spin-transfer torque approach to MRAM, are working on a high-performance MRAM technology which could be used in servers.

As of now, MRAM density is limited largely by the size of the transistors required to drive sufficient current to the magnetic tunnel junction (MTJ) to flip its magnetic orientation. Dan Edelstein, an IBM fellow working on MRAM development at IBM Research, said “it is a tall order for MRAM to replace DRAM. But MRAM could be used in system-level memory architectures and as an embedded memory technology.”

PVD and etch challenges

Edelstein, who was a key figure in developing copper interconnects at IBM some twenty years ago, said MRAM only requires a few extra mask layers to be integrated into the BEOL in logic. But there remain major challenges in improving the throughput of the PVD deposition steps required to deposit the complex material stack and to control the interfacial layers.

The PVD steps must deposit approximately 30 layers and control them to Angstrom-level precision. Deposition must occur under very low base pressure, and in oxygen- and water-vapor free environments. While tool vendors are working on productization of 300mm MRAM deposition tools, Edelstein said keeping particles under control and minimizing the maintenance and chamber cleaning are all challenging.

Etching the complex materials stack is even harder. Chemical RIE is not practical for MRAMs at this point, and using ion beam etching (IBE) presents challenges in terms of avoiding re-deposition of material sputtered off during the IBE etch steps for the high-aspect-ratio MTJs.

For the tool vendors, MRAMs present challenges as companies go from R&D to high-volume manufacturing, Edelstein said.

A Samsung MRAM researcher, Y.J. Song, briefly described IBE challenges during an IEDM presentation describing an embedded STT-MRAM with a respectable 8-Mbit density and a cell size of .0364 sq. micron. “We worked to optimize the contact etching,” using IBE etch during the patterning steps, he said. The short fail rate was reduced, while keeping the processing temperature at less than 350°C, Song said.

Samsung embedded an STT-MRAM module in the copper back end of the line (BEOL) of a 28nm logic process. (Source: Samsung presentation at IEDM 2016).

Many of the presentations at IEDM described improvements in key parameters, such as the tunnel magnetic resistance (TMR), cell size, data retention, and read error rates at high temperatures or low operating voltages.

An SK Hynix presentation described a 4-Gbit STT-MRAM optimized as a stand-alone, high-density memory. “There still are reliability issues for high-density MRAM memory,” said SK Hynix’s S.-W. Chung. The industry needs to boost the TMR “as high as possible” and work on improving the “not sufficiently long” retention times.

At high temperatures, error rates tend to rise, a concern in certain applications. And since devices are subjected to brief periods of high temperatures during reflow soldering, that issue must be dealt with as well, detailed by a Bosch presentation at IEDM.

Cleans and encapsulation important

Gouri Sankar Kar, who is coordinating the MRAM research program at the Imec consortium (Leuven, Belgium), said one challenge is to reduce the cell size and pitch without damaging the magnetic properties of the magnetic tunnel junction. For the 28nm logic node, embedded MRAM would be in the range of a 200nm pitch and 45nm critical dimensions (CDs). At the IEDM poster session, Imec presented an 8nm cell size STT-MRAM that could intersect the 10nm logic node, with the MRAM pitch in the 100nm range. GlobalFoundries, Micron, Qualcomm, Sony and TSMC are among the participants in the Imec MRAM effort.

Kar said in addition to the etch challenges, post-patterning treatment and the encapsulation liner can have a major impact on MTJ materials selection. “Some metals can be cleaned immediately, and some not. For the materials stack, patterning (litho and etch) and clean optimization are crucial.”

“Chemical etch (RIE) is not really possible at this stage. All the tool vendors are working on physical sputter etch (IBE) where they can limit damage. But I would say all the major tool vendors at this point have good tools,” Kar said.

To reach volume manufacturing, tool vendors need to improve the tool up-time and reduce the maintenance cycles. There is a “tail bits” relationship between the rate of bit failures and the health of the chambers that still needs improvement. “The cleanup steps after etching are very, very critical” to the overall effort to improving the cost effectiveness of MRAM, Kar said, adding that he is “very positive” about the future of MRAM technology.

A complete flow at AMAT

Applied Materials is among the equipment companies participating in the Imec program, with TEL and Canon-Anelva also heavily involved. Beyond that, Applied has developed a complete MRAM manufacturing flow at the company’s Dan Maydan Center in Santa Clara, and presented its cooperative work with Qualcomm on MRAM development at IEDM.

In an interview, Er-Xuan Ping, the Applied Materials managing director in charge of memory and materials technologies, said about 20 different layers, including about ten different materials, must be deposited to create the magnetic tunnel junctions. As recently as a few years ago, throughput of this materials stack was “extremely slow,” he said. But now Applied’s multi-cathode PVD tool, specially developed for MRAM deposition, can deposit 5 Angstrom films in just a few seconds. Throughput is approaching 20 wafers per hour.

Applied Materials “basically created a brand-new PVD chamber” for STT-MRAM, and he said the tool has a new e-chuck, optimized chamber walls and a multi-cathode design.

The MRAM-optimized PVD tool does not have an official name yet, and Ping said he refers to it as multi-cathode PVD. With MRAM requiring deposition of so many different metals and other materials, the Applied tool does not require the wafer to be moved in and out, increasing efficiency. The shape and structure of the chamber wall, Ping said, allow absorption of downstream plasma material so that it doesn’t come back as particles.

For etch, Applied has worked to create etching processes that result in very low bit failure rates, but at relatively relaxed pitches in the 130-200nm range. “We have developed new etch technologies so we don’t think etch will be a limiting factor. But etch is still challenging, especially for cells with 50nm and smaller cell sizes. We are still in unknown territory there,” said Ping.

Jürgen Langer, R&D manager at Singulus Technology (Frankfurt, Germany), said Singulus has developed a production-optimized PVD tool which can deposit “30 material layers in the Angstrom range. We can get 20 wafers per hour throughputs, so I would say this is not a beta tool, it is for production.”

Jürgen Langer, R&D manager, presented a poster on MRAM deposition from Singulus Technology (Frankfurt, Germany).

Where does it fit?

Once the production challenges of making MRAM are ironed out, the question remains: Where will MRAM fit in the systems of tomorrow?

Tom Coughlin, a data storage consultant based in Atascadero, Calif., said embedded MRAM “could have a very important effect for industrial and consumer devices. MRAM could be part of the memory cache layers, providing power advantages over other non-volatile devices.” And with its ability to power on and power off without expending energy, MRAM could reduce overall power consumption in smart phones, cutting in to the SRAM and NOR sectors.

“MRAM definitely has a niche, replacing some DRAM and SRAM. It may replace NOR. Flash will continue for mass storage, and then there is the 3D Crosspoint from Intel. I do believe MRAM has a solid basis for being part of that menagerie. We are almost in a Cambrian explosion in memory these days,” Coughlin said.

Leti’s CoolCube 3D Transistor Stacking Improves with Qualcomm Help

Wednesday, April 27th, 2016

By Ed Korczynski, Sr. Technical Editor

As previously covered by Solid State Technology CEA-Leti in France has been developing monolithic transistor stacking based on laser re-crystallization of active silicon in upper layers called “CoolCube” (TM). Leading mobile chip supplier Qualcomm has been working with Leti on CoolCube R&D since late 2013, and based on preliminary results have opted to continue collaborating with the goal of building a complete ecosystem that takes the technology from design to fabrication.

“The Qualcomm Technologies and Leti teams have demonstrated the potential of this technology for designing and fabricating high-density and high-performance chips for mobile devices,” said Karim Arabi, vice president of engineering, Qualcomm Technologies, Inc. “We are optimistic that this technology could address some of the technology scaling issues and this is why we are extending our collaboration with Leti.” As part of the collaboration, Qualcomm Technologies and Leti are sharing the technology through flexible, multi-party collaboration programs to accelerate adoption.

Olivier Faynot, micro-electronic component section manager of CEA-Leti, in an exclusive interview with Solid State Technology and SemiMD explained, “Today we have a strong focus on CMOS over CMOS integration, and this is the primary integration that we are pushing. What we see today is the integration of NMOS over PMOS is interesting and suitable for new material incorporation such as III-V and germanium.”

Table: Critical thermal budget steps summary in a planar FDSOI integration and CoolCube process for top FET in 3DVLSI. (Source: VLSI Symposium 2015)

The Table shows that CMOS over CMOS integration has met transistor performance goals with low-temperature processes, such that the top transistors have at least 90% of the performance compared to the bottom. Faynot says that recent results for transistors are meeting specification, while there is still work to be done on inter-tier metal connections. For advanced ICs there is a lot of interconnect routing congestion around the contacts and the metal-1 level, so inter-tier connection (formerly termed the more generic “local interconnect”) levels are needed to route some gates at the bottom level for connection to the top level.

“The main focus now is on the thermal budget for the integration of the inter-tier level,” explained Faynot. “To do this, we are not just working on the processing but also working closely with the designers. For example, depending on the material chosen for the metal inter-tier there will be different limits on the metal link lengths.” Tungsten is relatively more stable than copper, but with higher electrical resistance for inherently lower limits on line lengths. Additional details on such process-design co-dependencies will be disclosed during the 2016 VLSI Technology Symposium, chaired by Raj Jammy.

When the industry decides to integrate III-V and Ge alternate-channel materials in CMOS, the different processing conditions for each should make NMOS over PMOS CoolCube a relatively easy performance extension. “Three-fives and germanium are basically materials with low thermal budgets, so they would be most compatible with CoolCube processing,” reminded Faynot. “To me, this kind of technology would be very interesting for mobile applications, because it would achieve a circuit where the length of the wires would be shortened. We would expect to save in area, and have less of a trade-off between power-consumption and speed.”

“This is a new wave that CoolCube is creating and it has been possible thanks to the interest and support of Qualcomm Technologies, which is pushing the technological development in a good direction and sending a strong signal to the microelectronics community,” said Leti CEO Marie Semeria. “Together, we aim to build a complete ecosystem with foundries, equipment suppliers, and EDA and design houses to assemble all the pieces of the puzzle and move the technology into the product-qualification phase.”


Mentor Graphics U2U Meeting April 26 in Santa Clara

Monday, April 11th, 2016


Mentor Graphics’ User2User meeting will be held in Santa Clara on April 26, 2016. The meeting is a highly interactive, in-depth technical conference focused on real world experiences using Mentor tools to design leading-edge products.

Admission and parking for User2User is free and includes all technical sessions, lunch and a networking reception at the end of the day. Interested parties can register on-line in advance.

Wally Rhines, Chairman and CEO of Mentor Graphics, will kick things off at 9:00am with a keynote talk on “Merger Mania.“ Wally notes that in 2015, the transaction value of semiconductor mergers was at an all-time historic high.  What is much more remarkable is that the average size of the merging companies is five times as large as in the past five years, he said. This major change in the structure of the semiconductor industry suggests that there will be changes that affect everything from how we define and design products to how efficiently we develop and manufacture them. Dr. Rhines will examine the data and provide conclusions and predictions.

He will be followed by another keynote talk at 10:00 by Zach Shelby, VP of Marketing for the Internet of Things at ARM. Zach was co-founder of Sensinode, where he was CEO, CTO and Chief Nerd for the ground-breaking company before its acquisition by ARM. Before starting Sensinode, Zach led wireless networking research at the Centre for Wireless Communications and at the Technical Research Center of Finland.

After user sessions and lunch, a panel will convene at 1:00pm to address the topic “Ripple or Tidal Wave: What’s driving the next wave of innovation and semiconductor growth?” Technology innovation was once fueled by the personal computer, communications, and mobile devices. Large capital investment and startup funding was rewarded with market growth and increased silicon shipments. Things are certainly consolidating, perhaps slowing down in the semiconductor market, so what’s going to drive the next wave of growth?  What types of designs will be staffed and funded? Is it IoT?  Wearables?  Automotive?  Experts will address these and other questions and examine what is driving growth and what innovation is yet to come.

Attendees can pick from nine technical tracks focused on AMS Verification, Calibre I and II, Emulation, Functional Verification, High Speed, IC Digital Implementation, PCB Flow, and Silicon Test & Yield Solutions. You’ll hear cases studies directly from users and also updates from Mentor Graphics experts.

These user sessions will be held at 11:10-12:00am, 2:00-2:50pm and 3:10-5:00pm.

A few of the highlights:

  • Oracle’s use of advanced fill techniques for improving manufacturing yield
  • How Xilinx built a custom ESD verification methodology on the Calibre platform
  • Qualcomm used emulation for better RTL design exploration for power, leading to more accurate power analysis and sign-off at the gate level
  • Micron’s experience with emulation, a full environment for debug of SSD controller designs, plus future plans for emulation
  • Microsoft use of portable stimulus to increase productivity, automate the creation of high-quality stimulus, and increase design quality
  • Formal verification at MicroSemi to create a rigorous, pre-code check-in review process that prevents bugs from infecting the master RTL
  • A methodology for modeling, simulation of highly integrated multi-die package designs at SanDisk
  • How Samsung and nVidia use new Automatic RTL Floorplanning capabilities on their advanced SoC designs
  • Structure test at AMD: traditional ATPG and Cell-Aware ATPG flows, as well as verification flows and enhancements

Other users presenting include experts from Towerjazz, Broadcom, GLOBALFOUNDRIES, Silicon Creations, MaxLinear, Silicon Labs, Marvell, HiSilicon, Qualcomm, Soft Machines, Agilent, Samtec, Honewell, ST Microelectronics, SHLC, ViaSat, Optimum, NXP, ON Semiconductor and MCD.

The day winds up with a closing session and networking reception from 5:00-6:00pm.

Registration is from 8:00-9:00am in the morning.

Qualcomm Posts Lower Profit, Revenue in Quarter

Thursday, January 28th, 2016


By Jeff Dorsch, Contributing Editor

Qualcomm reported net income of $1.5 billion for the fiscal first quarter ended December 27, down 24 percent from $2.0 billion a year ago. Revenue fell 19 percent to $5.8 billion, from $7.1 billion in the first quarter of fiscal 2015.

The chip design and licensing company forecast revenue in the fiscal second quarter would fall within the range of $4.9 billion to $5.7 billion.

The Qualcomm CDMA Technologies segment produced $4.1 billion in revenue for the first quarter, down 22 percent from a year earlier and up 13 percent from the prior quarter. The Qualcomm Technology Licensing segment posted $1.6 billion in Q1 revenue, down 12 percent from a year ago and down 10 percent from the previous quarter.

“We delivered a stronger than expected quarter with earnings per share above the high end of our initial estimates, driven by better than expected 3G/4G reported device sales and benefits realized from cost actions across the company,” Qualcomm CEO Steve Mollenkopf said in a statement. “We signed several new license agreements in China and are on track with our cost reduction initiatives. Design traction for our new Snapdragon 820 processor continues to be strong, and we expect improving trends in our chipset business in the second half of fiscal 2016.”

Qualcomm reported shipping 242 million MSM devices in the first quarter, down 10 percent from 270 million in Q1 of the previous fiscal year. George Davis, the company’s executive vice president and chief financial officer, forecast second-quarter shipments of MSM devices would be in the range of 175 million to 195 million units.

Device sales totaled $60.6 billion in the quarter, compared with $56.4 billion a year earlier.

Qualcomm estimated Q1 shipments of 3G/4G devices were 307 million to 311 million units, up from the year-earlier figures of 284 million to 288 million units. Average selling prices for those devices in the quarter were $193 to $199, while they were $194 to $200 a year ago.

On a conference call with analysts, Qualcomm executives disclosed a contract dispute with LG Electronics, indicating it may have an effect on Qualcomm’s results during this fiscal year.

Mollenkopf said the company would initiate production of gallium arsenide power amplifiers in 2017.

He also announced that Cristiano Amon, executive vice president of Qualcomm Technologies, has been promoted to president of Qualcomm CDMA Technologies.

Chris Laudani wrote on, “With Qualcomm stock trading at just 3 points off its 52-week low, investors have little to celebrate so far this new year. Over the past year, the stock is down 33%. Last month, management outlined its strategic realignment plan; it will hold an analyst meeting next month in an effort to get the stock moving higher.

“Qualcomm shareholders have to be very disappointed. The stock is back to where it started 10 years ago, despite the fact that Qualcomm stock has returned $14 billion in fiscal 2015 to shareholders. The company has repurchased $11.2 billion worth of common stock and paid out $2.9 billion in dividends, all for nothing.”

He added, “Management is very excited by the possibilities of 5G and believes Qualcomm can lead mobile innovation. In addition, the company sees a $25 billion opportunity in adjacent markets, such as automotive, Internet of Things, small-cell networking, mobile computing and data centers.”

Mentor Graphics Presents at TSMC Forum

Monday, September 21st, 2015

By Jeff Dorsch, Contributing Editor

Mentor Graphics had a hand in presenting two of the 30 papers offered Thursday at Taiwan Semiconductor Manufacturing’s Open Innovation Platform Ecosystem Forum in Santa Clara, Calif.

The 30 presentations were provided in three tracks: electronic design automation, intellectual property, and EDA/IP/services. Mentor’s participation was in the EDA track.

On Tuesday afternoon, HiSilicon and Mentor presented on the subject of “Fill-As-You-Go: Leveraging Calibre SmartFill to cut your 16nm design verification runtime.” The presenters were Zhe Liu, a senior design engineer at HiSilicon, and Bill Graupp, DFM application technologist at Mentor.

The presentation described the work of HiSilicon, Mentor, and TSMC to implement methodology changes.

Later in the afternoon, Qualcomm and Mentor Graphics made a presentation titled “2-5x Productivity Improvement in Converging to a DRC-clean Cell Design – Qualcomm’s Experience with Calibre RealTime.” Tom Williams of Qualcomm was the presenter.

Cadence Design Systems and Synopsys also made multiple presentations at the TSMC OIP Ecosystem Forum.

Monolithic 3D processing using non-equilibrium RTP

Friday, April 17th, 2015


By Ed Korczynski, Senior Technical Editor, Solid State Technology

Slightly more than one year after Qualcomm Technologies announced that it was assessing CEA-Leti’s monolithic 3D (M3D) transistor stacking technology, Qualcomm has now announced that M3D will be used instead of through-silicon vias (TSV) in the company’s next generation of cellphone handset chips. Since Qualcomm had also been a leading industrial proponent of TSV over the last few years while participating in the imec R&D consortium, this endorsement of M3D is particularly relevant.

Leti’s approach to 3D stacking of transistors starts with a conventionally built and locally-interconnected bottom layer of transistors, which are then covered with a top layer of transistors built using relatively low-temperature processes branded as “CoolCube.” Figure 1 shows a simplified cross-sectional schematic of a CoolCube stack of transistors and interconnects. CoolCube M3D does not transfer a layer of built devices as in the approach using TSV, but instead transfers just a nm-thin layer of homogenous semiconducting material for subsequent device processing.

Fig. 1: Simplified cross-sectional rendering of Monolithic 3D (M3D) transistor stacks, with critical process integration challenges indicated. (Source: CEA-Leti)

The reason that completed transistors are not transferred in the first place is because of intrinsic alignment issues, which are eliminated when transistors are instead fabricated on the same wafer. “We have lots of data to prove that alignment precision is as good as can be seen in 2D lithography, typically 3nm,” explained Maud Vinet, Leti’s advanced CMOS laboratory manager in an exclusive interview with SST.

As discussed in a blog post online at Semiconductor Manufacturing and Design ( last year by Leti researchers, the M3D approach consists of sequentially processing:

  • processing a bottom MOS transistor layer with local interconnects,
  • bonding a wafer substrate to the bottom transistor layer,
  • chemical-mechanical planarization (CMP) and SPE of the top layer,
  • processing the top device layer,
  • forming metal vias between the two device layers as interconnects, and
  • standard copper/low-k multi-level interconnect formation.

To transfer a layer of silicon for the top layer of transistors, a cleave-layer is needed within the bulk silicon or else time and money would be wasted in grinding away >95% of the silicon bulk from the backside. For CMOS:CMOS M3D thin silicon-on-insulator (SOI) is the transferred top layer, a logical extension of work done by Leti for decades. The heavy dose ion-implantation that creates the cleave-layer leaves defects in crystalline silicon which require excessively high temperatures to anneal away. Leti’s trick to overcome this thermal-budget issue is to use pre-amorphizing implants (PAI) to completely dis-order the silicon before transfer and then solid-phase epitaxy (SPE) post-transfer to grow device-grade single-crystal silicon at ~500°C.

Since neither aluminum nor copper interconnects can withstand this temperature range, the interconnects for the bottom layer of transistors need to be tungsten wires with the highest melting point of any metal but somewhat worse electrical resistance (R). Protection for the lower wires cannot use low-k dielectrics, but must use relatively higher capacitance (C) oxides. However, the increased RC delay in the lower interconnects is more than offset by the orders-of-magnitude reduction in interconnect lengths due to vertical stacking.

M3D Roadmaps

Leti shows data that M3D transistor stacking can provide immediate benefit to industry by combining two 28nm-node CMOS layers instead of trying to design and manufacture a single 14nm-node CMOS layer:  area gain 55%, performance gain 23%, and power gain 12%. With cost/transistor now expected to increase with sequential nodes, M3D thus provides a way to reduce cost and risk when developing new ICs.

For the industry to use M3D, there are some unique new unit-processes that will need to ramp into high-volume manufacturing (HVM) to ensure profitable line yield. As presented by C. Fenouillet-Beranger et al. from Leti and ST (paper 27.5) at IEDM2014 in San Francisco, “New Insights on Bottom Layer Thermal Stability and Laser Annealing Promises for High Performance 3D Monolithic Integration,” due to stability improvement in bottom transistors found through the use of doping nickel-silicide with a noble metal such as platinum, the top MOSFET processing temperature could be relaxed up to 500°C. Laser RTP annealing then allows for the activation of top MOSFETs junctions, which have been characterized morphologically and electrically as promising for high performance ICs.

Figure 2 shows the new unit-processes at <=500°C that need to be developed for top transistor formation:

*   Gate-oxide formation,

*   Dopant activation,

*   Epitaxy, and

*   Spacer deposition.

Fig. 2: Thermal processing ranges for process modules need to be below ~500°C for the top devices in M3D stacks to prevent degradation of the bottom layer. (Source: CEA-Leti)

After the above unit-processes have been integrated into high-yielding process modules for CMOS:CMOS stacking, heterogeneous integration of different types of devices are on the roadmap for M3D. Leti has already shown proof-of-concept for processes that integrate new IC functionalities into future M3D stacks:

1)       CMOS:CMOS,

2)       PMOS:NMOS,

3)       III-V:Ge, and

4)       MEMS/NEMS:CMOS.

Thomas Ernst, senior scientist, Electron Nanodevice Architectures, Leti, commented to SST, “Any application that will need a ‘pixelated’ device architecture would likely use M3D. In addition, this approach will work well for integrating new channel materials such as III-V’s and germanium, and any materials that can be deposited at relatively low temperatures such as the active layers in gas-sensors or resistive-memory cells.”

Non-Equilibrium Thermal Processing

Though the use of an oxide barrier between the active device layers provides significant thermal protection to the bottom layer of devices during top-layer fabrication, the thermal processes of the latter  cannot be run at equilibrium. “One way of controlling the thermal budget is to use what we sometimes call the crème brûlée approach to only heat the very top surface while keeping the inside cool,” explained Vinet. “Everyone knows that you want a nice crispy top surface with cool custard beneath.” Using a laser with a short wavelength prevents penetration into lower layers such that essentially all of the energy is absorbed in the surface layer in a manner that can be considered as adiabatic.

Applied Materials has been a supplier-partner with Leti in developing M3D, and the company provided responses from executive technologists to queries from SST about the general industry trend to controlling short pulses of light for thermal processing. “Laser non-equilibrium heating is enabling technology for 3D devices,” affirmed Steve Moffatt, chief technology officer, Front End Products, Applied Materials. “The idea is to heat the top layer and not the layers below. To achieve very shallow adiabatic heating the toolset needs to ramp up in less than 100 nsec. In order to get strong absorption in the top surface, shorter wavelengths are useful, less than 800 nm. Laser non-equilibrium heating in this regime can be a critical process for building monolithic 3D structures for SOC and logic devices.”

Of course, with ultra-shallow junctions (USJ) and atomic-scale gate-stacks already in use for CMOS transistors at the 22nm-node, non-equilibrium thermal processing has already been used in leading fabs. “Gate dielectric, gate metal, and contact treatments are areas where we have seen non-equilibrium anneals slowly taking the place of conventional RTP,” clarified Abhilash Mayur, senior director, Front End Products, Applied Materials. “For approximate percentages, I would say about 25 percent of thermal processing for logic at the 22nm-node is non-equilibrium, and seen to be heading toward 50 percent at the 10nm-node or lower.”

Mayur further explained some of the trade-offs in working on the leading-edge of thermal processing for demanding HVM customers. Pulse-times are in the tens of nsec, with longer pulses tending to allow the heat to diffuse deeper and adversely alter the lower layers, and with shorter pulses tending to induce surface damage or ablation. “Our roadmap is to ensure flexibility in the pulse shape to tailor the heat flow to the specific application,” said Mayur.

Now that Qualcomm has endorsed CoolCube M3D as a preferred approach to CMOS:CMOS transistor stacking in the near-term, we may assume that R&D in novel unit-processes has mostly concluded. Presumably there are pilot lots of wafers now being run through commercial foundries to fine-tune M3D integration. With a roadmap for long-term heterogeneous integration that seems both low-cost and low-risk, M3D using non-equilibrium RTP will likely be an important way to integrate new functionalities into future ICs.

5nm Node Needs EUV for Economics

Thursday, January 29th, 2015


By Ed Korczynski, Sr. Technical Editor


At IEDM 2014 last month in San Francisco, Applied Materials sponsored an evening panel discussion on the theme of “How do we continue past 7nm?” Given that leading fabs are now ramping 14nm node processes, and exploring manufacturing options for the 10nm node, “past 7nm” means 5nm node processing. There are many device options possible, but cost-effective manufacturing at this scale will require Extreme Ultra-Violet (EUV) lithography to avoid the costs of quadruple-patterning.

Fig. 1: Panelists discuss future IC manufacturing and design possibilities in San Francisco on December 16, 2014. (Source: Pete Singer)

Figure 1 shows the panel being moderated by Professor Mark Rodwell of the University of California Santa Barbara, composed of the following industry experts:

  • Karim Arabi, Ph.D. – vice president, engineering, Qualcomm,
  • Michael Guillorn, Ph.D. – research staff member, IBM,
  • Witek Maszara, Ph.D. – distinguished member of technical staff, GLOBALFOUNDRIES,
  • Aaron Thean, Ph.D. – vice president, logic process technologies, imec, and
  • Satheesh Kuppurao, Ph.D. – vice president, front end products group, Applied Materials.

Arabi said that from the design perspective the overarching concern is to keep “innovating at the edge” of instantaneous and mobile processing. At the transistor level, the 10nm node process will be similar to that at the 14nm node, though perhaps with alternate channels. The 7nm node will be an inflection point with more innovation needed such as gate-all-around (GAA) nanowires in a horizontal array. By the 5nm node there’s no way to avoid tunnel FETs and III-V channels and possibly vertical nanowires, though self-heating issues could become very challenging. There’s no shortage of good ideas in the front end and lots of optimism that we’ll be able to make the transistors somehow, but the situation in the backend of on-chip metal interconnect is looking like it could become a bottleneck.

Guillorn extolled the virtues of embedded-memory to accelerate logic functions, as a great example of co-optimization at the chip level providing a real boost in performance at the system level. The infection at 7nm and beyond could lead to GAA Carbon Nano-Tube (CNT) as the minimum functional device. It’s limited to think about future devices only in terms of dimensional shrinks, since much of the performance improvement will come from new materials and new device and technology integration. In addition to concerns with interconnects, maintaining acceptable resistance in transistor contacts will be very difficult with reduced contact areas.

Maszara provided target numbers for a 5nm node technology to provide a 50% area shrink over 7nm:  gate pitch of 30nm, and interconnect level Metal 1 (M1) pitch of 20nm. To reach those targets, GLOBALFOUNDRIES’ cost models show that EUV with ~0.5 N.A. would be needed. Even if much of the lithography could use some manner of Directed Self-Assembly (DSA), EUV would still be needed for cut-masks and contacts. In terms of device performance, either finFET or nanowires could provide desired off current but the challenge then becomes how to get the on current for intended mobile applications? Alternative channels with high mobility materials could work but it remains to be seen how they will be integrated. A rough calculation of cost is the number of mask layers, and for 5nm node processing the cost/transistor could still go down if the industry has ideal EUV. Otherwise, the only affordable way to go may be stay at 7nm node specs but do transistor stacking.

Thein detailed why electrostatic scaling is a key factor. Parasitics will be extraordinary for any 5nm node devices due to the intrinsically higher number of surfaces and junctions within the same volume. Just the parasitic capacitances at 7nm are modeled as being 75% of the total capacitance of the chip. The device trend from planar to finFET to nanowires means proportionally increasing relative surface areas, which results in inherently greater sensitivity to surface-defects and interface-traps. Scaling to smaller structures may not help you if you loose most of the current and voltage in non-useful traps and defects, and that has already been seen in comparisons of III-V finFETs and nanowires. Also, 2D scaling of CMOS gates is not sustainable, and so one motivation for considering vertical transistors for logic at 5nm would be to allow for 20nm gates at 30nm pitch.

Kappurao reminded attendees that while there is still uncertainty regarding the device structures beyond 7nm, there is certainty in 4 trends for equipment processes the industry will need:

  1. everything is an interface requiring precision materials engineering,
  2. film depositions are either atomic-layer or selective films or even lattice-matched,
  3. pattern definition using dry selective-removal and directed self-assembly, and
  4. architecture in 3D means high aspect-ratio processing and non-equilibrium processing.

An example of non-equilibrium processing is single-wafer rapid-thermal-annealers (RTA) that today run for nanoseconds—providing the same or even better performance than equilibrium. Figure 2 shows that a cobalt-liner for copper lines along with a selective-cobalt cap provides a 10x improvement in electromigration compared to the previous process-of-record, which is an example of precision materials engineering solving scaling performance issues.

Fig. 2: ElectroMigration (EM) lifetimes for on-chip interconnects made with either conventional Cu or Cu lined and capped with Co, showing 10 times improvement with the latter. (Source: Applied Materials)

“We have to figure out how to control these materials,” reminded Kappurao. “At 5nm we’re talking about atomic precision, and we have to invent technologies that can control these things reliably in a manufacturable manner.” Whether it’s channel or contact or gate or interconnect, all the materials are going to change as we keep adding more functionality at smaller device sizes.

There is tremendous momentum in the industry behind density scaling, but when economic limits of 2D scaling are reached then designers will have to start working on 3D monolithic. It is likely that the industry will need even more integration of design and manufacturing, because it will be very challenging to keep the cost-per-function decreasing. After CMOS there are still many options for new devices to arrive in the form of spintronics or tunnel-FETs or quantum-dots.

However, Arabi reminded attendees as to why the industry has stayed with CMOS digital synchronous technology leading to design tools and a manufacturing roadmap in an ecosystem. “The industry hit a jackpot with CMOS digital. Let’s face it, we have not even been able to do asynchronous logic…even though people tried it for many years. My prediction is we’ll go as far as we can until we hit atomic limits.”

Blog review October 20, 2014

Monday, October 20th, 2014

Matthew Hogan of Mentor Graphics blogs about how automotive opportunities are presenting new challenges for IC verification. A common theme for safety systems involves increasingly complex ICs and the need for exceptional reliability.

Anish Tolia of Linde blogs that technology changes in semiconductor processing and demands for higher-purity and better-characterized electronic materials have driven the need for advanced analytical metrology. Apart from focusing on major assay components, which are the impurities detailed in a Certificate of Analysis (CoA), some customers are also asking that minor assay components or other trace impurities must be controlled for critical materials used in advanced device manufacturing.

Karey Holland of Techcet provides an excellent review of SEMI’s Strategic Materials Conference. The keynote presentation, “Materials Innovation for the Digital 6th Sense Era,” was by Matt Nowak of Qualcomm. He discussed both the vision of the Internet of Things (IoT), the required IC devices (including analog & sensors) and implications to materials (and cost to manufacture) from these new IC devices.

The age of the Internet of Things is upon us, blogs Pete Singer. There are, of course, two aspects of IoT. One is at what you might call the sensor level, where small, low power devices are gathering data and communicating with one another and the “cloud.” The other is the cloud itself. One key aspect will be security, even for low-level devices such as the web-connected light bulb. Don’t hack my light bulb, bro!

Linde Electronics has developed the TLIMS/SQC System. Anish Tolia writes that this system includes an information management database plus SQC/SPC software and delivers connectivity with SAP, electronically pulling order information from SAP to TLIMS and pushing CoA data from TLIMS to SAP.

Ed Korczynski blogs about how IBM researchers showed the ability to grow sheets of graphene on the surface of 100mm-diameter SiC wafers, the further abilitity to grow epitaxial single-crystalline films such as 2.5-μm-thick GaN on the graphene, the even greater ability to then transfer the grown GaN film to any arbitrary substrate, and the complete proof-of-manufacturing-concept of using this to make blue LEDs.

Phil Garrou says it’s been awhile since we looked at what is new in the polymer dielectric market so he checked with a number of dielectric suppliers – specifically Dow Corning, HD Micro and Zeon — and asked what was new in their product lines.

Karen Lightman, Executive Director, MEMS Industry Group, had the pleasure to learn more about the challenges and opportunities affecting MEMS packaging at a recent International Microelectronics Assembly and Packaging Society (IMAPS) workshop held in her hometown of Pittsburgh and at her alma mater, Carnegie Mellon University (CMU).

Ed Korczynski blogs that The Nobel Prize in Physics 2014 was awarded jointly to Isamu Akasaki, Hiroshi Amano, and Shuji Nakamura “for the invention of efficient blue light-emitting diodes which has enabled bright and energy-saving white light sources.”

Yes, GlobalFoundries is hot on FD-SOI. Yes, Qualcomm’s interested in it for IoT. Yes, ST’s got more amazing low-power FD-SOI results. These are just some of the highlights that came out of the Low Power Conference during Semicon Europa in Grenoble, France (7-9 October 2014) blogs Adele Hars.

The Week in Review: October 17, 2014

Friday, October 17th, 2014

Driven by rising demand for fitness and health monitoring features as well as by improved user interfaces, shipments of sensors used in wearable electronic devices will rise by a factor of seven from 2013 through 2019, according to IHS Technology.

Intermolecular, Inc. announced this week that Dr. Bruce McWilliams has been appointed president and chief executive officer. David Lazovsky has resigned as president and chief executive officer and from the Board of Directors to pursue other interests.

Qualcomm announced that it has reached agreement with CSR regarding the terms of a recommended cash acquisition of CSR will be acquired by Qualcomm Global Trading Pte. Ltd.

Texas Instruments this week announced it has shipped more than 22 billion units of copper wire bonding technology from its internal assembly sites and is now in production for major high reliability applications including automotive and industrial.

Element Six this week announced the development of a new thermal grade of diamond grown by chemical vapor deposition (CVD), DIAFILM TM130.

Next Page »