BannerforSemiWiki 800x100 (2)

Top 10 Reasons to Use Vivado Design Suite

Top 10 Reasons to Use Vivado Design Suite
by Paul McLellan on 03-23-2014 at 7:05 am

Here are the top 10 reasons to use the Xilinx Vivado Design Suite to design your All Programmable Devices:

Reason number 10: Accelerate verification by over 100XThe Vivado Design Suite System Edition lets you do design at the C, C++ or systemC level. But a side-benefit is that you can use these languages for verification at performances much higher than using raw RTL. Maybe 10,000X faster.


Reason number 9: Comprehensive hardware debugVivado Design Suite’s probing methodologies are intuitive and flexible. Designers can choose a strategy that best suits their design flow using RTL design files, synthesized design and XDC constraint file. Or netlist insertion. Or interactive Tcl or scripts to automate probing.

Reason number 8: Vivado IDE for design and simulationThere is a complete, fully-integrated set of tools for design entry, timing analysis, hardware debug and simulation encapsulated in a state of the art integrated design environment (IDE).

Reason number 7: Block based IP integrationVivado IP integrator is the industry’s first plug-and-play IP integration design environment. This provides a graphical and Tcl based correct-by-construction design flow.

Reason number 6: Model-based DSP design integrationUsing System Generator for DSP, the industry’s leading high-level design tool for converting DSP algorithms into production quality hardware. It interfaces to the industry’s quasi-standard modeling tools, MATLAB and Simulink (from Mathworks).

Reason number 5: High level synthesisVivado High Level Synthesis takes C, C++ and SystemC and automatically converts it into Verilog or VHDL for further synthesis. This previously somewhat esoteric technology is now completely mainstream with hundreds of users.

Reason number 4: Performance, performance, performanceMuch faster run-time than the competition. Multiple compilations per day even on huge designs. Much smaller memory footprint than the competition.


Reason number 3: Robust performance and low powerInterconnect is now the performance bottleneck in modern programmable-logic devices. The place & route algorithms in Vivado break this performance bottleneck to deliver high performance results at the push of a button. Higher performance than the competition. It also delivers lower system power.

Reason number 2: Fit more into the device, fasterA tool’s ability to fit more functions into an All Programmable device directly translates into savings in system-level cost and power by letting you select the smallest device for your design. The advanced fitting algorithms combined with the architecture of the Xilinx 7 devices allow nearly 100% utilization to be achieved whereas competing devices on the same benchmark top out at 67%, meaning a larger and more expensive device would be required.


And the number 1 reason for using Vivado Design Suite:That is how you get your hands on Xilinx’s All Programmable devices. Because like in those old Visa ads, they don’t take American Express.

A full white-paper that dives deeper into each of these reasons is available here.


More articles by Paul McLellan…


Mentor Acquires BDA!

Mentor Acquires BDA!
by Daniel Nenni on 03-23-2014 at 7:00 am

Mentor Graphics acquired Berkeley Design Automation this morning. The details of the deal were unannounced. This is a strong move by Mentor to challenge Cadence and Synopsys in the nanometer analog/mixed-signal market and nanometer memory characterization market, respectively. Mentor not only acquires the technology and team, but BDA’s list of blue-chip leading-edge customer engagements including the likes of Qualcomm, Samsung, Sony, Fujitsu, Broadcom, TSMC, and NXP.

BDA established itself in the mid-2000’s by solving circuit verification problems that no other EDA company could solve—problems like silicon-accurate device noise analysis on an entire post-layout closed-loop integer-N and fractional-N PLLs, high-speed I/O, and ADCs—while co-simulating with a standard Verilog simulator. (It hurts to even think of problems like this). (See Silicon Correlation)

BDA built its business by literally asking leading-edge analog/mixed-signal (A/MS) design teams for the problems that no other simulator can handle and providing the solution. BDA would then move “downstream” to run circuit simulations that other simulators could run, but BDA’s Analog FastSPICE simulator would run them 5x-10x faster than any other foundry-certified simulator. They take the same approach to this day. (See BDA History)

Customers and industry pundits alike give BDA credit for sending a loud-and-clear wakeup call to Cadence and Synopsys. Cadence responded with APS, which AFS continues to beats by over 2x in runtime and 10x in capacity. Synopsys tried XA with little success, and later bought Magma for FineSim. While FineSim is fast, it’s not nearly accurate, robust, or feature-rich enough for advanced A/MS circuits. Moreover, FineSim is not well integrated into Cadence’s Analog Design Environment, which is a must-have to compete with Cadence for “real” A/MS design team. Meanwhile BDA recently announced AFS Mega which brings SPICE accuracy memory application, and announced TSMC has adopted it for all of their 16nm FinFET SRAM characterization. (See TSMC and BDA)

In apparent effort to slow or stop BDA, last spring Cadence hit them with a lawsuit regarding AFS integration into ADE. (See Cadence vs. BDA) There was immediate overwhelming customer support for BDA, and the case was dismissed with prejudice earlier this year. The undisclosed settlement agreement gives BDA a multi-year ADE integration with ADE, where standard Cadence Connections members have to renew annually.

With this move Mentor is suddenly not just on the map, but arguably takes a substantial technical lead nanometer-scale circuit verification. Mentor removes the small-EDA company barriers that BDA continuously faced and get them immediate access to the broad semiconductor market. Look for Mentor to couple AFS with Questa in the front end to take on AMS Designer and to couple AFS with Calibre in the back-end to create highly-differentiated flow to silicon.

Congratulations to everybody at BDA on the great exit. It was a pleasure working with you!

More Articles by Daniel Nenni…..

lang: en_US


DAC: Automotive, IP and Security

DAC: Automotive, IP and Security
by Paul McLellan on 03-21-2014 at 5:18 pm

DAC is in the first week of June in San Francisco as I’m sure you already know if you are reading this. Historically DAC has focused on electronic design automation (EDA) and embedded software and systems (ESS). This year there are three new areas: automotive, Intellectual Property (IP) and security.

Automotive
Ever increasing feature content enabled by electronics and software and the associated complexity has made development of automotive electronics and software amongst the largest challenges for the automotive industry. Today, the functionality, efficiency, time-to-market, cost, quality, safety, and security of a new vehicle are determined more by embedded systems and software than by any other factor. The Automotive Track is a new, unique forum addressing design automation and design methodologies to enable automotive designers and integrators to meet their unprecedented challenges.

There is a dual keynote Automating the Automobileby Henry Buczkowski, the Henry Ford Technical Fellow at (surprise) Ford Motor Company and Jim Tung, Mathworks fellow at (surprise again) Mathworks.

IP
This year DAC introduces the IP Track on Monday, focused on semiconductor IP. This track includes six sessions running in two rooms adjoining the exhibit floor, and provides creators and users of IP with an open forum to exchange information on state-of-art IP products and the tools and methodology to create, incorporate and validate IP in SoCs. Supporting the IP technical track is the exhibition floor which includes over 20 leading IP suppliers suppliers and providers.

The associated keynote is The Great SoC Challenge (IP to the Rescue!)is by Sir Hossein Yassaie, CEO of Imagination Technologies.

Security
As design of integrated circuits (ICs) and embedded systems is increasingly global, designers and users of ICs, intellectual property (IP) and embedded systems are increasingly facing trust issues. These systems are vulnerable to a variety of hardware-centric attacks, such as side channel analysis, reverse engineering, IP piracy, hardware Trojans and counterfeiting. The Security Track at DAC highlights the emergence of security and trust as important dimensions of hardware and embedded systems design, dimensions that must be considered side-by-side with power, performance, and reliability.

The security keynote The Intel Security Architecture Visionis by Ernie Brickell, Chief Security Architect at Intel.

The technical program for DAC will be announced on March 27th, next Thursday. Details about DAC are on the (all new) DAC website at dac.com.


More articles by Paul McLellan…


SEMI Breakfast Forums: the Internet of Things

SEMI Breakfast Forums: the Internet of Things
by Paul McLellan on 03-21-2014 at 4:29 pm

Coming up on April 10th is the SEMI Silicon Valley Breakfast Forum Internet of Things—Driving the Microelectronics Revolution. It runs from 7am to 10.45am and will be held at SEMI Headquarters which is at 3081 Zanker Road in San Jose.

Widespread adoption of the Internet of Things will take time, but the movement is advancing thanks to improvements in underlying technologies. New processor architectures, interfaces, and memory structure are improving efficiency and density while enabling smaller and lower-power applications.

With the potential to impact all aspects of society and our lives, now is the time for executives across all industries to think strategically about the opportunities likely to emerge from an intelligent and connected world.

There is no such thing as a free lunch at the forum but at least there is free continental breakfast from 7-8am. I love that phrase since we are on the continent of North America but continent referred to in the breakfast is Europe.

Denny McGuirk, President of SEMI will welcome everyone at 8am.

The main speakers are:

  • At 8.05 Karen Bartleson who, in addition to working at Synopsys (and beating my EDAgraffitti blog in 2009 in Denali’s America’s Next Top Blogger grrr), is also President of the IEEE Standards Association
  • At 8.40 Alfonso Velosa, a research director at Gartner where he focuses on ecosystem relationships for IoT, smart cities and electronics.At 8.05
  • At 9.00 Mike Rosa, who works within the 200mm systems group at Applied Materials. He focuses on new and emerging technologies, especially MEMS (which is an important component of IoT)
  • At 9.20 there will be a 20 minute networking break
  • At 9.40 Kerry McGuire Balanze, who is vice-president of strategy for the IoT business unit of ARM
  • At 10.00 there will be a panel discussion until 10.45 when the meeting will wrap-up

Details, including links for registration, are here. Early bird registration ends on March 27th so hurry.

The Arizona Breakfast Forum, also Internet of Things, is on April 17th at Intel’s Ocotillo Campus building OC2 at 4500 South Dobson Road. It starts at 7.30am. Speakers are from TIRIAS research, Semico Research, Medtronic, Freescale and Intel. Full details and registration (early bird for this one ends on April 4th) is here.

Then the Texas Spring Forum How the Internet of Things Will Impact New Manufacturing Solutions is on April 22nd at Intel Austin, 1300 South Mopac Expressway. It also starts at 7.30am. Speakers are from Freescale, Cisco, ARM, Brewer Science, Applied Materials and Intel. Full details (early bird ends April 15th) are here.


More articles by Paul McLellan…


Evaluate MEMS Devices out-of-fab Before Fabrication

Evaluate MEMS Devices out-of-fab Before Fabrication
by Pawan Fangaria on 03-21-2014 at 10:30 am

MEMS design and fabrication is highly complex in the sense that the fabrication process heavily depends on the design, unlike IC fabrication which has a standard set of processes. A slight change in MEMS design can alter its fabrication steps to a large extent. For example, setting device parameters such as capacitance or linear displacement can affect the choice of the film thickness, etch rate, sidewall profile and so on. The design and process are so much tied together that many iterations through the fab are required (which consume costly resources and time) in order to get a perfect build. While an IDM has to keep its fab resources deployed for such a build-and-test experimentation in-house, a fabless design house has to additionally incur time for its design to take several tours through an external fab. This all has significant impact, first on cost of design and manufacturing and then turn-around-time, thus squeezing the window of opportunity which is already small in today’s competitive semiconductor market.

In such a scenario, there is nothing like having a software tool at the designer’s desk which can provide a 3D model of the complete MEMS device based on its layout and process description. The iterations between the design and the process can take place virtually at the terminal itself until the final model of the device is perfect to enter the fab for manufacturing.

I am impressed with the SEMulator3D Virtual Fabrication Platform from Coventorwhich simplifies this build-and-test cycle for MEMS in the most economical and fastest way through its Voxel Modeling Engine and virtual metrology operations for measurement of critical technology parameters. Above is an example of RF Tunable Capacitor model derived from its layout and process behavior, which has its actuation part in silicon wafer (black block on top) bonded on the RF part in glass wafer (yellow and white block at bottom).

How do you describe the process and metrology steps through the process editor? Above is the edited process file for the RF Tunable Capacitor. It has steps defined similar to a real fabrication flow. Several measurement types can be defined such as CD (critical dimension), difference between max and min elevation, thickness of films at particular locations, angle of a sloped sidewall and so on.

SEMulator3D can show a cross section of the device model at any location. In the above cross section of a RF Tunable Capacitor, CD (finger width of the comb drive), Film Thickness (structure thickness) and Line to Line Spacing (width of the perforation plate) are shown. The virtual metrology measurement options provide ‘3D DRC’ for validating these dimensions. Particular regions (e.g. Comb drive fingers stator and rotor Line-to-Line) can be selected to view in 2D layout editor for virtual metrology and analysis.

DRIE (Deep Reactive Ion Etching) has scalloping issues which need to be analyzed and controlled. To characterize the scalloping effect, comb fingers of the Tunable Capacitor were etched using SEMulator3D’s DRIE Custom Python module which provides anisotropic etching through a time-domain multiplexed processing scheme. In this scheme, etching and polymer deposition are alternated to pattern high aspect ratio of structures.

Above images show the SEM (Scanning Electron Microscope) view of fabricated and SEMulator3D view of simulated DRIE profile of inter-digitated comb fingers. The average values of parameters such as height, width and rate are highly dependent on design parameters and etching equipment. There is good agreement between the simulation and experimental results.

The image above shows a scallop dimension, simulated on SEMulator3D, of 501nm height and 176nm width. Longer etch cycle times are responsible for larger scallop heights that in turn are responsible for deeper scallops. A SEM view of similar fabricated etched parts compare well with the modeled view by SEMulator3D.

The Virtual Fabrication Platform of SEMulator3D accelerates the development time for MEMS devices by a large extent at significantly lesser cost and higher accuracy by eliminating in-fab set up and process variations during unit process interaction studies and quantitative analysis.

Alexandre Mehdaoui of Coventor has described in great detail the overall Virtual Fabrication process for MEMS with nice examples, pictures and references in his new whitepaperposted at Coventor website. It provides an interesting learning about what goes in fabrication and how that has been automated in a software tool like SEMulator3D.

More Articles by Pawan Fangaria…..

lang: en_US


ARM, Cadence and the Internet of Things

ARM, Cadence and the Internet of Things
by Paul McLellan on 03-20-2014 at 6:30 pm

There is clearly a lot of hype about the Internet of Things (IoT) right now, but also it is clear that it will be a real market. In fact, it already is with various medical, fitness and home-appliance products already available. At CES in January, wearables was probably the biggest trend. That doesn’t always pan out (3D TV was the biggest trend a couple of years ago and how’s your 3D TV working out?) and in terms of acres of silicon it will not compete with smartphones for some time.


I wrote here about a report on the Internet of Things that ARM had commissioned from the Economist Intelligence Unit. That was 6 months ago but the basic information in the report is very much still current. With acquisitions like Google buying Nest, if anything the conclusions of the report that IoT is real and inevitable seem even stronger now.

In the meantime, the Maker movement isn’t waiting for true products to appear in the space they are making them themselves. Some of these will turn out to be early prototypes of ideas that will become widespread. For example, hereis a little circuit that will send you a text message when your dryer is done. If your dryer is in the garage or the basement where you don’t always hear it, wouldn’t you want your dryer to have that capability. Or hereis an open source-thermostat, basically a Nest you can build yourself. Just Google “internet of things maker” to see some of the other interesting ideas that are floating around: the future is closer than you think.


IoT devices are characterized by needing a combination of technologies:

  • A microprocessor to run all the software including the communications stack
  • Flash and other memories
  • Analog (or analogue as ARM likes to call it, you have to read it with a British accent) for sensors
  • RF for networking
  • Perhaps MEMS devices such as accelerometers

There is a perception that creating such a mixed signal design is hard. ARM and Cadence have an upcoming webinar MCU Mixed Signal Design Challengesat 11am Pacific on April 9th on the topic. I doubt that they are going to claim that such designs are easy, but the combination of Cadence’s Virtuoso environment with ARM’s processor and physical IP makes it a lot smoother than it used to be. Indeed, design tools have advanced so much in the last few years that IoT chips with all their various blocks can be designed so that they are production quality from the start, and work first time.

The webinar will be moderated by Curt Schwaderer of OpenSystems Media and will feature:

  • Diya Soubra, CPU Product Marketing Manager from ARM
  • Joel Rosenberg, Platform Marketing Director from ARM
  • Mladen Nizic, Engineering Director from Cadence.

More details including a link for registration are here.


TLM Modeling Environment Goes Commercial

TLM Modeling Environment Goes Commercial
by Daniel Payne on 03-20-2014 at 6:00 pm

The most successful EDA companies typically choose a domain where they have deep knowledge, then serve a few leading-edge customers that are willing to work with a start-up in exchange for early access to that new technology. The theory is that if you can satisfy the leading-edge customer then you can also satisfy the rest of the market segment. Magillemis one such EDA company that has worked closely with STMicroelectronics over the years on ESL tools that use IP-XACTand help manage complexity, interoperability and design re-use for SoC designers. This week Magillem announced something a little bit different, because they have signed an OEM agreement with STMicroelectronics to offer ST’s TLM (Transaction Level Modeling) modeling environment and methodology.

The EDA industry has provided tools to model at a variety of levels, and each for a different purpose:

  • Transistor-level, for SPICE circuit simulation and layout versus schematic checking
  • Gate-level, for functional simulation and fault modeling
  • Cell-level, for collections of transistors and gates
  • Register Transfer Level (RTL), a language based approach using VHDL, Verilog or SystemVerilog. RTL can be automatically synthesized into gate or cell levels.
  • Transaction Level Modeling (TLM), a higher-lever approach to modeling a digital system, where details of communication between modules are separated from the details of the implementation of functional units or of the communication architecture. (Source: Wikipedia)

Magillem already offered five tools for SoC design:

  • Magillem IP-XACT Packager
  • Magillem Platform Assembly
  • Magillem Register View
  • Magillem Generator Studio
  • Magillem Flow Control

This new TLM modeling fills a product gap for Magillem, and will allow SoC designers to build and analyze virtual platforms using SystemC, one of the faster growing segments in EDA today. SystemC is a set of C++ classes and macros which provide an event-driven simulation interface in C++ (Source: Wikipedia). SystemC is a standard defined by Accellera, and Magillem is a board member of Accellera.

Expect more product details and availability to be announced as we get closer to DAC in June. There is a press release at the Magillem site with quotes from Philippe Magarshack, Executive Vice-President and General Manager, Design Enablement and Services STMicroelectronics and Cyril Spasevski, Chair and CTO of Magillem. You can also visit with Magillem at CHIPEX in Tel Aviv, Israel at the end of April.

lang: en_US


Handel Jones on FD-SOI vs FinFET

Handel Jones on FD-SOI vs FinFET
by Paul McLellan on 03-20-2014 at 1:27 am

Handel Jones has a new white-paper out titled Why Migration to FD-SOI is a Better Approach Than Bulk CMOS and FinFETs at 20nm and 14/16nm for Price-Sensitive Markets. Handel has done an in-depth analysis of the wafer and die costs of the various approaches, bulk planar (what we have been doing up to now), FD-SOI and FinFET. The analysis takes into account depreciation, equipment maintenance, direct/indirect labor, facilities, wafer cost, consumables, monitor wafers and line yield.


It turns out that at 28nm, FD-SOI is already marginally cheaper than bulk planar CMOS. It has a smaller number of masks and fewer processing steps. The difference is a bit bigger at 20nm. At 16nm the comparison with FinFET shows that FD-SOI is a lot cheaper. See the diagram above.


He then goes on to analyze die costs, taking into account gross die per wafer and yield. That results in the graph above. The money quote is that:At 14nm/16nm, the FD-SOI die cost for a 100mm[SUP]2[/SUP] die is 28.2% lower than the bulk FinFET die cost and has higher yield. The leakage of FD-SOI devices is projected to be comparable to that of FinFET devices.

For the very highest performance SoCs, then FinFET is presumably worth the cost. But for anything not on the bleeding edge then FD-SOI might be a more cost-effective solution. FD-SOI, like FinFET, has much lower leakage than bulk, but FinFET can have issues with dynamic power due to the high gate capacitance.

STMicroelectronics has been the trailblazer for FD-SOI and has working products in 28nm. It is not a theoretical alternative to FinFET, it is real. The whole supply chain is starting to fall into place.

At EDPS in Monterey on 17th/18th April I will be talking about FD-SOI in more depth. My working title is Praise FD-SOI, slag FinFETs but maybe I’ll find something a little more politically acceptable. There are also presentations that afternoon on FinFETs (praise FinFETs, slag FD-SOI perhaps) and on 3D-IC (praise TSVs, slag Moore’s Law completely). Come along and watch the wrestling match. Moore’s Law as we used to know it is over: 28nm won. It is not just cheaper than every process that came before it, it is cheaper than every process that is going to come after it. Details on EDPS are here.

The final conclusions of the white-paper:

  • at 28nm and 20nm, the lower power consumption and higher performance of FD-SOI compared to planar bulk CMOS gives major competitive advantages to FD-SOI in high volume portable applications.
  • the lower cost of FD-SOI die compared to 16nm FinFET die provides an overwhelming advantage to utilizing FD-SOI for high volume applications at this technology node.

Handel’s white-paper is here.


More articles by Paul McLellan…


Sewn open: Arduino and soft electronics

Sewn open: Arduino and soft electronics
by Don Dingee on 03-19-2014 at 3:00 pm

As several other recent threads on SemiWiki have pointed out, the term “wearables” is a bit amorphous right now. The most recognizable wearable endeavors so far are the smartwatch and fitness band, but these are far from the only categories of interest.

There is another area of wearable wonder beginning to get attention: clothing, which has drawn the interest of researchers, makers, and moms alike. The endgame as many see it is smart clothing: the weaving of electronics, sensors, and conventional fabrics into something called e-textiles. However, while athletes, soldiers, and other niches may get sensor-impregnated jerseys sooner, affordable clothing based on exotic advanced fabrics for most consumers may still be 20 or 30 years away by some estimates.

Right now, we have these anything-but-soft computing structures – chips, circuit boards, displays, switches – adaptable for some clothing applications. Still missing are some key elements, most notably power in the form of energy harvesting or smaller and denser batteries. The influence of water-based washing machines and their adverse effect on most electronics also looms large.

How do we cross this gap? It’s not all about advanced R&D; these types of challenges are well suited for experimentation and the imagination of makers. Several Arduino-compatible maker modules – all based on Atmel microcontrollers – have jumped in to the fray, showing how “soft electronics” can help create solutions.

photo courtesy Becky Stern

Maybe I’ve built one or two too many harness assemblies using expensive, mil-spec circular connectors, but the fascinating thing to me is what makes all these boards wearable. Small size is nice, but anybody knows a project needs wiring, right? You’ll notice the large plated holes on the first several offerings: these are eyelets for conductive thread, literally intended to sew these boards to other components like fabric pushbuttons. Many projects also use snaps, similar to 9V battery connections, to disconnect boards for conventional washing of the garment.

The other side of this is the software. One of the attractive features of Arduino is the IDE, real live C-style programming simplified for the masses, with functions designed to perform I/O on the Atmel MCU. Code is edited on a PC or Mac, and compiled into a sketch and uploaded to the board. There are so many examples of code for Arduino maker modules out there available in open source, it makes it easy to find and integrate functions quickly.

If that all sounds crazy, consider the pioneer for this is Leah Buechley of the MIT Media Lab, one of the thought leaders of the maker movement and an expert on e-textiles. She is the brain behind the LilyPad, the original 2” diameter Arduino wearable circa 2007 commercialized through SparkFun, with the most recent version featuring the ATmega32u4 and native USB.

Adafruit took the next steps with two wearable boards. FLORA is slightly smaller than the LilyPad and retains the same familiar circular profile and ATmega32u4 MCU. GEMMA goes even smaller, 1.1” in diameter, packing an ATtiny85 on board with a USB connection for easy development.


Not to be outdone by circles, squares and rectangles have recently come back into form. SquareWear 2.0 comes in two versions, the 1.7” square variant with a coin cell socket onboard, both including the ATmega328 MCU with simulated USB, high current MOSFET ports, a light sensor, and a temperature sensor. Seeed grabbed the ATmega32u4 and designed it into the Xadow, a tiny 1” x 0.8” expandable unit with integrated flat cable connectors for daisy chaining.


These aren’t just toys for creating flashing LEDs; there is no shortage of sensors and connectivity, including displays, GPS, Bluetooth, and more compatible with these wearable maker modules. Their popularity is growing: Becky Stern of Adafruit claims there are over 10,000 units of FLORA shipped so far, and they are the darlings of maker faire fashion shows and hackathons.

Besides the upside for makers, maybe this sewing angle will finally allow us to explain electronics to our moms, after all. Until we get to the fulfilled flexible future of e-textiles and more advanced technology, the conductive thread of soft electronics will stitch together creative ideas using somewhat familiar tiny modules with today’s microcontrollers.

lang: en_US


Triple Patterning

Triple Patterning
by Paul McLellan on 03-19-2014 at 1:00 pm

As you can’t have failed to notice by now, 28nm is the last process node that does not require double patterning. At 20nm and below, at least some layers require double patterning. The tightest spacing is typically not the transistors but the local interconnect and, sometimes, metal 1.


In the litho world they call double patterning LELE. This stands for litho-etch-litho-etch which describes the steps taken. Using the first mask, half the polygons are patterned and then etched. Then using the second mask, the other half of the polygons are patterned and then etched. This imposes some restrictions on what can be put on the mask, since it has to be so-called two-colorable, meaning the polygons can be split into two masks such that there each mask has polygons that are sufficiently far apart (there are lots of articles on SemiWiki about the details of this). One problem with LELE is that the second mask is registered onto the first pattern using the fiducial marks (like any other mask) and so there is a level of misalignment that the design must be able to cope with, approximately 10nm, so you can’t actually double the density using double patterning.

So what is beyond double patterning?

Three things:

  • triple patterning called LELELE in the lithography world
  • self-aligned double patterning (SADP) also sometimes called sidewall image transfer (SIT)
  • EUV, 14nm wavelength instead of 193nm, so we can go back to single exposure

EUV isn’t happening for 10nm unless either there is a miracle in improvement of the power of the light source (and some other problems are solved). Or 10nm slips out several years. Both are possible. The economics of 10nm are somewhere between unknown and dubious.


Triple patterning is pretty much the same as double patterning except that the polygons are partitioned onto three masks. The constraints on the design are different, to ensure this can be done. The big advantage of triple patterning over double is that it is a lot denser. Not three times as dense since the 3 masks will still have some misalignment. In fact because with 3 masks the alignment problems are worse (the distances are tighter) it looks like 3-4nm misalignment is the maximum, which is very hard to achieve.


SADP/SIT removes the potential misalignment between the multiple masks. First a mask is used to put down a temporary sacrificial structure called a mandrel. Then sidewalls are constructed on the sides of the mandrel. The mandrel is removed. This is now double patterning but due to the method of construction there was never a time when a mask was critically aligned on the previous one so the misalignment is removed. Sometimes this is also called pitch doubling since the sidewalls are half the pitch of the mandrels.

This SADP/SIT process can be repeated, using the sidewalls as new mandrels and constructing new sidewalls. This gives self aligned quadruple patterning. With enough process steps (and money) you can get down to 7nm like this without EUV.

One of the biggest decisions in process architecture is to decide what the pitch for various layers should be. The most critical pitch is that for local interconnect and metal 1 since these have major impact on the density of memories and standard cells. Too big a pitch and the process is not competitive since the area is to large. Too tight a pitch and the areas are small but the process is very expensive due to all the extra process steps and masks required. This is why there is so much attention on 14/16nm on what the metal 1 pitch is. An additional complication is that with these two layers it is not possible to live with 1D structures (lay down a grating and then use a cut mask to divide it up) since it is essential to be able to run in both the X and Y directions.

For 10nm it looks like SADP/SIT is needed for a few layers and then double patterning above that. Triple patterning seems like it maybe too expensive to be worth it.


More articles by Paul McLellan…