Semiwiki EDA Webinar 800x100
Semiconductor Expert Forum

Posted on

What is the impact of missing the 7NM node with EUV?

What is the impact of missing the 7NM node with EUV?
by Robert Maire on 10-23-2016 at 12:00 pm
Categories: Foundries, Lithography, Semiconductor Advisors, Semiconductor Services

ASML reported a quarter that was slightly below expectations coming in at Euro 1.815B in revenues and Euro 0.93 EPS. Orders were a bit soft at Euro 1.4B but well within the normal quarterly variation of a lumpy business. Euro 28M was lost in a currency adjustment associated with the Hermes acquisition.

The guidance for Q4 was between Euro 1.7B and 1.8B with 47-48% gross margin. The company will ship 4 EUV systems for the year, rather than the 6 it had targeted, one unit slipped due to customer issues and one unit slipped due to ASML supplier issues.

Logic was an overwhelming 84% of business in the quarter but is expected to normalize to a better balance between memory and logic going forward.

A logic heavy quarter at 84% of business….
As we have mentioned in previous notes, we think part of the strength was due to both TSMC and Intel spending for 10NM. We had also mentioned our concern that there may have been a temporary hiccup in capex spending by Samsung due to near term cash issues related to the smartphone battery debacle. There is the possibility that some business may have shifted out of the quarter……

EUV remains slow-only 4 systems to ship rather than 6….
EUV continues to move slowly along without a significant change in the relatively slow pace. Concerns continue to be centered primarily on system uptime and availability. However there are still many ecosystems issues that have yet to be finalized as well including pellicles, resist, inspection, etc; etc;.

ASML continues to diligently plug away at the “moonshot” like project but customers are still not willing to bet their multi billion dollar production on a less than firm solution when they have a working , albeit more expensive, solution in multi patterning.

5NM seems to be the insertion point…
Given the ongoing EUV issues, it seems virtually everyone is pushing off the decision which means that 7NM will be multipatterning. The only exception may be Samsung which has made more noise about EUV but has slipped a bit in progress towards 10NM and 7NM (could the slippage be because they are trying to use EUV??).

Given where we stand right now it feels like 5NM is a better bet for EUV.

On the positive side, ASML has gone from 3 EUV customers to 6. Obviously the second tier chip makers are sensing that we are getting closer to EUV so they have to start getting ready as the learning takes several years to come up to speed.

One of the new customers is likely GloFo who is skipping 10NM to go straight to 7NM (which we think is a great decision). However, it seems as if GloFo is doing multipatterning for 7NM not EUV.

Is ASML missing 7NM with EUV a problem???
One of our growing concerns is that ASML EUV may miss a large node opportunity at 7NM. We have heard from many in the industry that there is a strong expectation that 7NM will be a very popular node much like 28NM, which is “the gift that keeps on giving” to the industry.

TSMC and others seem very focused on 7NM being a “big” node as compared to “Lite” nodes such as 22NM. GloFo seems to be betting the farm on it.

If 7NM goes as big as expected and it is only multipatterning then ASML will have missed a significant spending window as 5NM will likely be more of a “Lite” node.

7NM could be the last big node before the industry sees some titanic shift in the core process technology as we approach physical limits of Moore’s Law.

How does an 84% logic quarter impact other equipment makers?
If the rest of the industry follows the pattern of spending that ASML just saw (and it usually does)then we would expect other tool makers to see a similarly logic heavy quarter.

This is obviously a very strong benefit to KLAC which has lived through some lean years as memory ruled the roost. The fact that Gen 5 is now out when logic is ramping makes the possibility of strong potential upside performance. Yield management tools are in strong demand during the ramp period which we are in right now. TSMC needs to get its act together to have strong yields for next years Iphone and 10NM processor.

AMAT has historically had a very strong relationship with TSMC (TSMC is the house that AMAT built) and as such should benefit from increased logic spend we have been hearing about. AMAT has had a more equal mix of memory/logic business.

Lam has historically been much more memory centric and Samsung remains its biggest customer. Given that Samsung feels a bit weaker right now and Samsung logic has been slow , we would not be surprised to see near term weakness associated with this mix in logic heavy spending. Additionally , Lam has historically not had as much success at Intel as other competitors, so heavy spending in Q4 by Intel (up 40% Q/Q) will not benefit Lam as much as it does other players such as KLAC and AMAT.

The stock…
We think ASML remains fully valued in the near term and we have heard no breakthroughs in EUV that make us want to rush out and buy the share. The Hermes acquisition is on track and will obviously cause a bit of a dislocation as it is absorbed. We think ASML is overdone above $100 and would be more likely to look at the shares below $100.

About Semiconductor Advisors
Semiconductor Advisors provides this subscription based research newsletter, Semiwatch, about the semiconductor and semiconductor equipment industries. We also provide custom research and expert consulting services for both investors and industry participants on a wide range of topics from financial to technology and tactical to strategic projects.

Posted on

What’s the Intel Capex Outlook?

What’s the Intel Capex Outlook?
by Robert Maire on 10-23-2016 at 7:00 am
Categories: Foundries, Intel Foundry, Semiconductor Advisors, Semiconductor Services

Intel has terrific QTR & slightly light guide Intel is recovering & transforming at the same time. Whats the Capex outlook? Impact on ASML KLAC LRCX?

Intel reported revenues of $15.78B and earnings of $0.80 for the quarter beating expectations and previous upward guidance. CCG (PCs) were up 21% Q/Q and 5% Y/Y. Data center was up 13% Q/Q and 10% Y/Y. IOT was up 20% Q/Q and 19% Y/Y. NVM was up 17% Q/Q and down 1% Y/Y. Capex for 2016 will come in more or less as expected at $9.5B +-$500M.

All in all, a very good quarter….outlook is for flattish revenues with slightly lower gross margin….slightly below street expectations.

Our quick takeaway is that Intel has been beating numbers handily and we think that there could be a bit of sandbagging especially with the introduction of a new CFO who isn’t going to want to disappoint his first quarter out of the starting gate, so we will see what the numbers finally come in at but we would bet that they will look better when Q4 is done.

Three quarters of above average Capex spend for three reasons….
The company talked about capex ticking upward for Q4 through Q2 of next year due to above average spending on three projects; 10NM rollout in Israel, 3D NAND in Dalian and 3D XPoint. Math suggests that capex could be $3.5B in Q4 as compared to an average run rate of $2.5B or so ( 40% above normalized run rate).

This perfect storm of spending will obviously help tool companies over the next several quarters. Given that Intels spend has been somewhat reduced, this blip, even though temporary, is a good shot in the arm to get these projects up and running

What wasn’t said is more interesting than what was said
It is interesting to note that Intel management did not even mention the word Israel or 10NM development in their prepared remarks. Also gone are comments on 10NM yield ramp and the associated “technology leadership” that Intel has historically touted. Could it be because TSMC has caught up to or passed Intel, with Samsung not far behind?

Part of the transformation of Intel is also transforming it from a technology and advanced production leader to more of an applications driven company. XPoint is an example of an application driven product rather than a Moore’s law driven product.

The de-focus away from Moore’s law leadership is quite clear as Moore’s law has also been missing from Intel’s recent vocabulary.

Results count towards a better stock price
Its hard to argue with success as Intel put up record numbers in what could have otherwise been a downward spiral of PCs. The restructuring though difficult will likely bear fruit going forward as well. It is not unreasonable to expect the stock hitting and breaking through $40 in the near future given our expectations of performance. The dividend doesn’t hurt either…

ASML- Waiting for 5NM…..maybe
As Intel backs away from pushing the limits of Moore’s Law, its clear that the need for EUV is also backing down (along with the alleged 15 unit order that drove ASML stock up huge…). Samsung seems the most committed to EUV but then again they are furthest behind in the race. We see EUV nowhere in sight at TSMC.

However, ASML will still do quite well with nice high margin immersion steppers.
We would expect them to also see brighter capex spending going forward.

KLAC – Gen 5 will be driven by 10NM and 7NM ramps
We expect KLAC to do very, very well with the current ramp of 10NM and 7NM and will be very well rewarded for its efforts on its new Gen 5 tools. Obviously Intel will buy a lot of tools, and we have heard that TSMC is ordering huge…

We think some of the share loss to E Beam will slow as it has run much of its course so the negative news is somewhat behind them.

KLAC will have some catch up to do as they were prepping to be acquired so they will have to step it up again and we could see increased spend after restructuring over a year ago followed by a fallow period of waiting to be bought.

LRCX- Forward guidance will make investors forget KLAM Kaput
We think Lam will put up pretty good Q3 numbers but will guide a very large (much better than expected) uptick in business over the next several quarters.

We would not be surprised to hear of either some pushouts or some business falling out of Q3 into Q4 as lam’s biggest customer, Samsung, may have shifted some spending due to near term cash issues related to exploding batteries. This would make the upside to future guidance look even better.

3D NAND remains the big driver. We would watch the gross margin line as competition may heat up in parts of the etch market. We have heard of some recent success in oxide etch by Semes ( a Samsung company in Korea) as well as an ongoing push by Applied in etch that may eat away at the edges. Rumor also has it that AMEC in China may have etch tools at Intel.

Lam still has its analyst meeting scheduled in November but now rather than a wedding celebration it will likely be a bit more subdued.

The stocks are feeling a bit “toppy”
The stocks have been doing very well, perhaps too well as they are feeling a little more like they are at the end of a strong run. Even though business is clearly very good and we think guidance will be great as well we wonder if that can keep the stocks moving or if they will take a bit of a breather after earnings.

About Semiconductor Advisors
Semiconductor Advisors provides this subscription based research newsletter, Semiwatch, about the semiconductor and semiconductor equipment industries. We also provide custom research and expert consulting services for both investors and industry participants on a wide range of topics from financial to technology and tactical to strategic projects.

Posted on

Webinar on Revolutionary Changes in SOC IP Access

Webinar on Revolutionary Changes in SOC IP Access
by Tom Simon on 10-22-2016 at 7:00 am
Categories: eSilicon, Events, IP, Semiconductor Services

Knowledge is power, and I’ve seen the trend over time of people getting more and deeper access to knowledge as each year goes by. I remember, as a student in high school back the in 70’s, the first time I wanted to buy stock in a company. You could only get a quote by calling a broker or visiting the broker’s office. Today you can get real time quotes on your computer – or phone even. The same goes for researching investments. Before, you needed to have personal connections or pay hefty research fees and commissions. Now you can use google to get everything you need, and then some.

This is just one example. We have seen the same trend in everything you can buy. However, one of the last areas to provide online research and the ability to purchase online is semiconductor IP. Granted this is a niche market, but access to this information is life or death for fabless semiconductor companies.

Looking at the changes in the semiconductor business, it is easy to see how access to each component of the chip design process has expanded. Before the fabless movement, only designers at companies that had a wafer fab could even consider getting access to chip design technology – tools and IP. Then came fabless, but a lot of key information about available IP was shrouded in secrecy. Or, in many cases there simply were no choices available – you took what you could get.

Right now, we are crossing a Rubicon in chip design. The equivalent of the Back to the Future self-tying shoes for IP selection is available to fabless chip designers. eSilicon has rolled out its STAR navigation platform that puts real live data on PPA into the designer’s hands, before money has been committed, contracts have been signed and hard decisions have been cast in stone.

Designers can go online with eSilicon’s STAR navigation system to get detailed information on many types of IP, including memory blocks. Because of the wide range of configurations and options, the correct selection of memory blocks can have a huge effect on every aspect of an SOC design.

To demystify the process and show the level of accessibility they provide, eSilicon is hosting a webinar that will walk designers through the complete process. The webinar will be on Wednesday October 26 at both 9AM and 9PM Pacific Daylight Time. They will cover examining PPA data for memory alternatives and exploring different architectures. With this system it is even possible to download front end models to verify the design choices. Lastly they will demonstrate how a quote is generated so you can purchase the IP online.

More blogs by Tom Simon

Posted on

Fabless Photonic Design Flow Takes Shape as Cadence teams up with Lumerical and PhoeniX

Fabless Photonic Design Flow Takes Shape as Cadence teams up with Lumerical and PhoeniX
by Mitch Heins on 10-21-2016 at 4:00 pm
Categories: Cadence, EDA, Silicon Photonics

This week Cadence Design, Lumerical Solutions and PhoeniX Software hosted a two-day photonic summit and workshop. The first day had nearly 100 registered participants and featured industry leaders from Global Foundries, UCSB, MIT, Hewlett Packard Enterprise, General Electric, Boeing, Rockley Photonics, and Juniper Networks speaking about their current efforts with integrated photonics and the emerging photonics ecosystem. The second day of the summit was comprised of a full-day hands-on session where more than 70 participants were introduced to a new top-down, schematic-driven design, verification and implementation flow for electro-optical integrated circuits.

The new flow features Cadence’s extensive suite of custom, analog and mixed-signal design functionality combined with a new set of capabilities that promises to enable the co-design of systems comprised of electrical and optical components in one design automation environment. One of the things that stood out for me while watching the second-day session was that this wasn’t just another set of disparate tools that were being strung together to make a flow. Indeed, Cadence has thoughtfully made extensions to their design database and infrastructure to enable photonic design. The devil is in the details of course, but we learned that Cadence, PhoeniX and Lumerical have been working together for over two years now to get to this point. It was not an easy endeavor.

There is no way to describe everything that was discussed and shown in the all-day session in this short article, as evidenced by the fact that the participants all went home with an almost 200 page document describing the flow. None the less I will endeavor to give you the 50,000 foot birds eye view of what I saw. First and foremost the flow is an extension of Cadence’s front-to-back SDL (schematic driven layout) flow for custom, analog and mixed signal design. Everything you loved and hated about that flow is still there. Key additions to the flow for photonics included :

  • new layer types in the technology file to enable connectivity checks,
  • new pin types for optical connections
  • new photonics-based schematic checks
  • frequency and time-domain optical circuit simulation support in ADE ( via the Lumerical engines)
  • configuration management of optical and electrical views for circuit simulation partitioning and netlisting
  • a fully parameterized set of photonic building blocks that can be mapped onto different technology processes ( via the PhoeniX engines)
  • the ability to create and characterize photonic modules including the creation of compact models (via PhoeniX layout engines + Lumerical FDTD, electrical and thermal solvers)
  • full curvilinear shape generation including all angle rotations and phase-aware auto-waveguide routing ( via the PhoeniX engines)
  • forward annotation of design parameters and constraints from the schematic to the layout through CDF including the use of parameterized pcells that interact with the PhoeniX curvilinear engines for module and waveguide creation

  • real-time checking of schematic vs layout parameter mismatch
  • back annotation of changes to parameters and physical modules and waveguide routing from from layout to schematic to simulation
  • an extensive wrapping of PhoeniX and Lumerical commands in skill, enabling designers to use the full power of both Phoenix and Lumerical tools directly from within the Cadence environment
  • beginnings of an electrical / optical co-simulation methodology and infrastructure including the use of characterized and parameterized photonic building blocks stored in foundry specific process design kits (PDK).


This set of capabilities is a huge step forward for integrated photonic designers and especially those companies who are struggling to integrate and co-design their electrical and photonic systems. The new flow also fills some significant gaps in Cadence’s previous capabilities for photonic design, namely curvilinear shape and all angle rotations along with photonic simulators and solvers. A couple of remaining areas still to be worked are a more efficient DRC process to reduce the amount of spurious false error reporting caused by all of the curvilinear shapes and more work on the integration and analysis of IC, PIC and packaging. More to come on these items at a later date.

It’s still early days for the flow but the fact that Cadence, PhoeniX and Lumerical were able to take more than 70 participants through a hands-on exercise of the flow in a single day was a testament in and of itself that the flow is already robust enough for the challenge. I was impressed and anyone that knows me, knows that means something. I’m not from Missouri, but I grew up right next door to it. Show me – and that they did.

Also read: The Fabless Empire Strikes Back, Global Foundries and Cadence make moves into Integrated Photonics!

Posted on

Webinar Offers View into TSMC IP Design Methodology

Webinar Offers View into TSMC IP Design Methodology
by Tom Simon on 10-21-2016 at 12:00 pm
Categories: EDA, FinFET, Foundries, Solido, TSMC

Standard cell and memory IP are key enablers for new process node availability. These two items must be in place early and be completely ready for a process node to scale to volume. Development of both leaves no room for error and they require the highest performance possible. Foundries are extremely focused on this and spend a lot of time and energy in delivering this IP in a timely fashion. Variation aware analysis is a critical part of this IP development. However, with the introduction of FinFET devices at 16nm and below, some assumptions about the nature of variation are changing.

Usually with Gaussian distributions, when we talk about sigma, we can draw an equivalence between the number of standard deviations needed to reach a given sigma. Sigma is really a way of specifying the percentage of cases under the curve: thus 3 sigma is 99.865%. This comes about because with a true Gaussian distribution, 3 standard deviations covers this percentage of the cases.

With variation analysis, designers seek to put to rest concerns that outliers can cause a greater percentage of chips to fail due to process variation, voltage fluctuation, high temperature, missed timing margins, etc. Additionally, designers can use statistical design analysis methods to determine optimal design parameters for the highest performance. However, these methodologies rely on being able to simulate the design cases farthest from the mean.

With a Gaussian distribution, getting to 6 sigma, or 99.9999999013% of the cases, would require analyzing out to 6 standard deviations from the mean. With a brute force simulation approach this would mean running billions of samples. To make things worse, circuit design behavior does not have Gaussian distributions. Instead they have long tails, drastically increasing the number of standard deviations that are needed to reach a given sigma.

This point was driven home by Jacob Ou from TSMC and Kris Breen from Solido during their webinar late last September. They point out that a single standard cell, sense amp, or bit cell/slice is really only one small part of a larger design that depends on all the elements working. A tolerable error in a single cell or circuit, becomes intolerable when the likelihood of chip failure depends on thousands, or millions of instances of the same in a larger chip.


Jacob from TSMC talks about how the non-planar nature of FinFET devices leads to new parasitic elements coming into play that can no longer be considered negligible. These include additional capacitive couplings within the FinFET device. The result is a long tail on the performance histogram. In some cases, the distribution curves can even become bi-modal. In the webinar TSMC discusses their use of Solido tools to tackle tough issues in standard cell library development. In many cases they were able to get results more quickly, or in some cases even perform analysis that would have otherwise been impossible.

Solido’s High Sigma Monte Carlo uses a self-validating approach to quickly find cases that are above of the desired sigma and simulate them. An ordering of samples around the sigma threshold is generated and simulated, which also provides algorithmic feedback on the effectiveness of the sample selection process. Because SPICE is used for simulation, there is no doubt in the final outcomes.

TSMC usually keeps their cards close to their chest, but in this webinar they go into details about the results achieved when Solido tools are used in their internal flow. The good news is that the webinar is available for replay in case you missed the live session.

More articles by Tom Simon

Posted on

Apple Car Crumble

Apple Car Crumble
by Roger C. Lanctot on 10-21-2016 at 7:00 am
Categories: Automotive

Software, mechanical and electrical engineers working for auto makers received a huge self-esteem injection this week as events unfolding at Apple suggested that the company had abandoned long-rumored plans for building a car. Considering the fact that Apple still hasn’t delivered a decent navigation app with traffic services, it’s hardly a shock that the company would consider making a car to be a bridge too far.

The bottom line is that making a car IS a pretty tough task. In many respects cross-town rival Tesla Motors has made the whole process look – at least outwardly – entirely too easy, if expensive. The hiring binge at Apple that suggested interest in the automotive industry and the firing binge reflecting its denouement reveal a newly realistic Apple.

It appears that Apple retains an abiding interest in developing automated driving technology, at least according to “observers” and “insiders.” But a car developed in house or in partnership with an existing car maker now appears unlikely.

Observers and analysts were quick to point out that Apple only focuses on high margin opportunities. The theory here is that the low margins in the auto industry created a sour grapes scenario for Apple thereby leading to a bowing out of the race to build a car.

The margin argument is a quaint and convenient one and it ignores the fact that Apple changes the economics of the markets it enters. The whole argument for Apple entering the automotive industry was that it would change the game and rewrite the rules.

By definition, if Apple found certain conditions in the market to be unfavorable, Apple would alter those conditions. Apple fans were looking to Apple to bring innovative transportation solutions to the market that would help reduce traffic, congestion, highway fatalities and fossil fuel consumption.

It is more likely that Apple came to the conclusion that it not only lacked any novel new solutions to these transportation challenges, but adding even MORE cars to the equation would be unhelpful and therefore both financially and spiritually unrewarding. There were bigger barriers for Apple to overcome than low margins. Low margins (negative margins?) haven’t stopped Tesla Motors.

The reality is that Apple is not built or positioned to take on the automotive industry and nothing in its various product offerings suggest a grasp of what it takes to make a car. The nature of creating a car requires an extraordinary level of cooperation and coordination between teams working on different systems.

The internal operational security within Apple, which limits communication between teams, is a big barrier to this kind of collaboration. Additionally, the regulatory oversight and liability exposure characteristic of the auto industry are enough to scare away even the bravest and most robust legal department.

Finally, it’s pretty clear that Apple has always regarded the car as an accessory to the Apple eco-system, which remains true. Apple continues to deliver devices produced under highly controlled circumstances encompassing hardware and software and which work universally and globally in a predictable manner.
Apple’s product development and design discipline have served the company well in its run up to dominance in the smartphone market. Apple’s automotive integrations, such as CarPlay, are universally more predictable and easier to use – even if Apple’s resistance to adopting industry standards periodically creates nightmares for car companies.

This resistance to industry standards is yet another limitation for Apple’s ambitions. When it comes to connectors and interfaces it’s Apple’s way or the highway. It looks like Apple will have to hitch a ride on the automotive highway for the foreseeable future. And that’s very reassuring for all those who design and build cars for a living. It’s a tough job. Thank you for seeing and accepting that, Apple.

Roger C. Lanctot is Associate Director in the Global Automotive Practice at Strategy Analytics. More details about Strategy Analytics can be found here: https://www.strategyanalytics.com/access-services/automotive#.VuGdXfkrKUk

Posted on

The Fabless Empire Strikes Back, Global Foundries and Cadence make moves into Integrated Photonics!

The Fabless Empire Strikes Back, Global Foundries and Cadence make moves into Integrated Photonics!
by Mitch Heins on 10-20-2016 at 4:00 pm
Categories: Cadence, EDA, Foundries, GlobalFoundries, Silicon Photonics

In August I wrote an article proclaiming Score 1 for IDMs vs Fabless and discussedIntel’sannouncement of volume production of their 100G PSM4 and 100G CWDM4 transceiver products.

This week the Fabless Empire strikes back.Daniel Nenni and I attended a two-day Photonic Summit and workshop hosted by Cadence Design, PhoeniX Software andLumerical Solutions. The keynote speaker for the summit was Ted Letavic, Senior Fellow of Global Foundries. Finally, one of the big fabless guys spoke out about integrated photonics and his presentation should strike fear into the hearts of every large IDM currently vested in integrated photonics. Letavic rolled out a succinct and powerful argument for why integrated photonics was a technology that was going main stream and he made it abundantly clear that Global Foundries was entering into the integrated photonics foundry business.

This week should be marked as a water shed day in photonics as it was a day when a major pure-playfoundryproclaimed thatintegrated photonics was real and here to stay and thenwent on toclaimin no uncertain terms that they had demonstrated that they can make high yielding integrated photonics on a 300mm CMOS platform. In fact, Global Foundries isusing a300mm SiGe on SOI process to enable active photo detectors and modulators. And, in case there were any nay-sayers in the audience, Ted’s talk was immediately followed by John Bowers of UCSB who gave a presentation about their accomplishments of bonding III-V materials (most notably on-chiplasers) on to silicon. All of the pieces of the puzzle are now coming together, light sources, high speed modulators, high index silicon waveguides for small, low cost small photonic devices and a growing infrastructure for photonic test and packaging.

The icing on the fabless photonic attack came in two forms. The first was a speaker Aaron Zilkie from Rockley Photonics, a fabless PIC (photonic integrated circuit) chipset startup. Rockley is promising a disruptive change to data center network architectures using an integrated high-speed switching solution comprising a digital packet-switching ASIC with an Optical I/O PIC all integrated into one single low-cost package without the need for power hungry high-speed RF signals traces between the ASIC and the optical components. The solution promises to reduce multiple levels of switches in the data center network effectively flattening the network and providing for a more flexible and higher performing data center.

The fact that a small fabless company could in theory team up with Global Foundries to challenges the likes of Intel in the data center was not lost on those in the audience who came from the IDM side.

And if the startups didn’t both the incumbents then they should have at least been worried by the like of Hewlett Packard Enterprisewho showed a complete reticle field of integrated photonic structures with 1000’s of resonator rings being used to characterize the effects of process variance on their photonic devices.

The second prong of the Fabless attack came in the form of Cadence Design stepping up to the photonics plate.They had made some noise at the Optical Fiber Conference in the Spring of this year but the event they co-hosted with PhoeniX Software and Lumerical Solutions, two well-known photonic design automation companies, clearly showed that Cadence is jumping into the photonics fray with both feet. For those of you who have been living under a digital rock for the last 25 years, Cadence literally owns the analog and mixed signal implementation market with their Virtuoso franchise and they have a very sizeable share of the mixed-signal verification market to boot. Virtuoso is ubiquitous for custom and analog design and they have now combined forces with PhoeniX Software and Lumerical Solutions to produce a state-of-the-art, top-down, electro-optical design automation suite. Lumerical is known for their photonic simulation engines and solvers while PhoeniX is known for their native curvilinear shape engines. Both companies have participated in literally hundreds of photonic tape-outs over the last decade. The combination of Lumerical and PhoeniX with Cadence, plus the entry of a high-volume pure-play CMOS-based foundry spells real trouble for IDMs who have been ruling the high-end integrated photonics markets.

The Fabless vs IDM photonics battle is on. Stay tuned as this story continues to develop!

Also read: Fabless Photonic Design Flow Takes Shape as Cadence teams up with Lumerical and PhoeniX

Posted on

Why Integrate Bluetooth LE IP in a Single Wearable SoC?

Why Integrate Bluetooth LE IP in a Single Wearable SoC?
by Eric Esteve on 10-20-2016 at 12:00 pm
Categories: IP, Synopsys

Did you know that, in over 800 teardowns of mobile and wearable products from 2012 to 2015, wireless chips outnumbered the actual number of products, indicating multiple wireless ICs in some designs ([SUP]1[/SUP])? It could be interesting to look at the advantages of integrating wireless technology such as Bluetooth low energy in a single SoC. Especially for systems where bill of material (BOM) cost and power consumption can be real issues, like Internet of Things (IoT) and wearable applications.

Bluetooth has been initially defined for short range usage, typically headset and smartphone, with paired device broadcasting approach. According with Bluetooth SIG, the future launch of Bluetooth 5 at the end of 2016, beginning of 2017, will suppress these limitations in term of range and broadcasting capability, and also double the speed and by consequence half the power consumption. Let’s have a look at the various wireless architectures integrating Bluetooth low energy, and the preferred process technology associated with each option.

  • Standalone RF transceiver: The controller and PHY are integrated in the transceiver chip that connects to the main SoC which houses the software stack and application code. RF transceiver is implemented in legacy nodes, like 180 nm.
  • Wireless network processor: Several wireless protocols are integrated in a dedicated processor housing the wireless protocol stack, but the application code is in the application processor SoC. This network processor option target mature 90 nm node.
  • Fully integrated wireless SoC: This monolithic, single die implementation is ideal for Bluetooth with the low energy technology for IoT applications. The Link Layer and PHY are integrated into the SoC that runs all software stacks and application codes. The 40nm and 55nm technology nodes are becoming popular for this monolithic option.
  • Combo wireless chipset solution: Several wireless technologies such as WiFi and Bluetooth are integrated into a single transceiver connected to the SoC that includes the digital modem. All software, wireless stacks, and application codes reside in an external non-volatile memory. This solution is prevalent for mobile application processor and leverage aggressive process nodes like 28 nm, or below.

The combo wireless chipset architecture is offering unlimited off-chip memory, giving programmers more resources, but the number of chips (application processor + transceiver + flash) is too large if the goal is to develop a very low cost application, such as wearable or IoT device. Current wearable designs, like fitness wristbands, implementing Bluetooth LE for low bandwidth wireless connectivity, integrate only two chips: one SoC connected to a Bluetooth LE IC via a UART or I2C bus. It is now possible to push the integration for this type of application, by integrating the complete Bluetooth IP (Link Layer + PHY) into a single SoC. It’s technically possible, but what are the benefits?

A monolithic solution will offer the expected benefits like lower power, lower BOM cost and smaller footprint, all of these being extremely valuable for the target application. And it will also offer lower latency, as the data sent over the AMBA AHB bus can reduce latency by 5 to 10 cycles versus the SPI bus.

If we look more carefully at the power consumption, the two chip solution is made of a SoC, say in 40 nm or 55 nm, and one RF transceiver in 180 nm. Integrating the Bluetooth LE as an IP in the SoC will allow to seriously decreasing the power consumption part of the Bluetooth function. This function, implemented in a 180 nm process node (RF transceiver architecture), is now running in the SoC targeting 40 nm (55 nm) process node at much lower (Vdd) voltage, 0.9V for these ultra-low-power processes. This benefit is coming on top of the expected power saving going with the integration of two chips into a mono-chip. For battery powered devices like fitness wristbands, lower power consumption translate into longer usage, and the time between charges could make or break the product.

As mentioned earlier, the price point of a wearable or IoT device solution could be decisive. If it’s too high, the product will stay in a niche: cool, but too expensive to allow for wide market adoption. The wireless integration enables the removal of a complete chip sets, reducing packaging and test cost and removing duplication of power management. This can save over $0.15 in packaging costs and 20-30 extra pads that are required to support the additional wireless network processor. These savings, in conjunction with a reduced PCB footprint, make the total system cost savings very attractive.

The Bluetooth PHY is available on 180-nm, 55-nm and 40-nm process nodes allowing designers to also take advantage of the advanced processes’ power, area and performance benefits, especially on 55-nm and 40-nm. Moreover, the SoC development costs are still reasonable on these process nodes when compared with nodes below 28 nm. Unlike for very high volume applications like smartphone, or performance demanding like servers, targeting 55 nm or 40 nm process nodes allow enough power and cost savings to build successful systems for such low computing, low bandwidth systems.

Last but not least, the DesignWare Bluetooth Low Energy IP solution is qualified by the Bluetooth Special Interest Group (SIG) which is critical for designer success.

Eric Esteve from IPNEST

([SUP]1[/SUP]) According with a survey made by Teardown.com

Posted on

Disarming Trolls

Disarming Trolls
by Bernard Murphy on 10-20-2016 at 7:00 am
Categories: General

An unintended consequence of the ubiquity of the Internet, particularly in social media, is the rise of the troll. Trolls post comments of unbelievable vitriol in some cases, comments that if issued in person and in public might lead to arrest and psych evaluations. Then vitriol turns into viral vitriol and the helpless target is bombarded with hate speech. But you can’t just suspend the rights of trolls. Speech is protected, at least up to a point, in many countries and few of us could honestly claim that we have never indulged in a heated response to a post or email. We may not be as vile as the worst offenders, but we share some of their traits.

In fact, theories of what makes for trollish behavior seem to be in flux. Accepted wisdom is that many of these people are socially awkward misfits (particularly teens and young adults) working out aggression through the anonymity of the Internet. But recent research suggests that many trolls are proud of their opinions, which they feel reflect social norms they want to defend. They are quick to anger, in that state perhaps less aware of crossing lines in self-expression, but mostly they are happy to be identified, to garner credit among like-minded thinkers for their vigorous support of those norms. Clickbait and echo chambers certainly play on this all-too-human weakness.

So “outing” trolls won’t necessarily help, and since none of us are perfect we ought to recognize that we too might be tempted to indulge in trollish behavior. Perhaps it would be preferable to try to block bad posts rather than bad posters, which requires some level of recognition and determination of how to respond. Social media providers are working on varying types of system in this line. Twitter, one of the most visible platforms for troll attacks, has an interesting approach called Periscope, which depends on users rather than machine learning to decide whether a tweet is abusive or offensive. As soon as one reader reports a tweet in this context, Periscope polls a randomly selected jury of other users reading the same tweet, to comment on whether they also find it offensive or abusive. If found guilty, the commenter is put in a 1-minute timeout and comments on the tweet are disabled. Repeat offenders are permanently muted. Nice approach, depending on human rather than artificial intelligence and difficult to game (I would think) given random jury selection.

Then again, Twitter may not have moved fast enough. According to Jim Cramer, Salesforce.com may have walked away from an acquisition in part because of the negative reputation around public perception of hatred associated with Twitter traffic. Which should be a reminder to other social platforms. It’s not just about being morally righteous – it’s also about company valuation.

In Google, a group called Jigsaw has developed (and no doubt continues to develop) a capability called Conversation AI. This is a machine-learning-based approach for which they used 17 million comments on New York Time stories, with moderator flags on offensive/abusive comments, plus data from Wikipedia discussion logs where they used a crowd-sourced service to flag reactions. Google claims this now can match judgments against a human panel with ~90% certainty and a ~10% false positive rate. Not bad, but I’m pretty sure these rates need to improve quite a bit to reach reasonable 1[SUP]st[/SUP] Amendment standards. Meantime Google is planning continued trials with NYT and Wikipedia.

An interesting sidebar here is that Conversation AI was inspired in part by work done by Riot Games on moderating player behavior in their massive multiplayer League of Legends world. Riot Games use machine learning to analyze conversations that has led to players being banned. From this they are able to show players in real-time where aspects of their comments are offensive or abusive. According to that company, providing this feedback has led to a 92% drop in offending behavior, which to me is an indicator that nipping the problem in the bud may be more effective that post-facto censorship.

Facebook doesn’t seem to be (at least publicly) as active in this area, perhaps because you connect only (mostly) to friends and you can unfollow or unfriend anyone who offends you. They do have some capabilities to detect a related problem – someone impersonating your account with the same name and profile. They’re also testing methods to detect intimate images as instances of revenge porn. In both cases, the potential victim is notified but must choose to have action taken (to avoid problems in purely automated responses).

This seems like an area where collaboration between providers is needed, perhaps even more than in domains like general AI. We could even dream that similar methods might encourage a general rise in the level of civility in on-line debate, out of which all kinds of wonderful things might happen (perhaps sane and effective government, to pick just one random example). Details on Google Jigsaw and the Riot Games work come from this Wired article. A Twitter Periscope article can be found HERE and can find an article on characterization of trolls HERE. What I could find on Facebook work in this area is HERE.

More articles by Bernard…