Today’s profound statement: “don’t fall in love with your tools, figure out the biz process change first.” Mixed-signal SoC designers are having ample challenges with their design process and are in need of design management, but don’t want another tool to do it. Continue reading “Webinar Alert – Helping Mixed Signal not be Mixed Up”
Poor Dr. Sanjay Gupta, Emmy award-winning chief medical correspondent for CNN, a neurosurgeon and professor and now an explainer of distracted driving as part of CNN’s weeklong report on Driving While Distracted which concluded last Saturday. He offers a detailed medical explanation of driver distraction as only a neurosurgeon could while more or less obscuring the reality that distracted driving is a lethal proposition and one that the responsible parties have failed to address.
– CNN Special Report: Driving While Distracted
Who is responsible for distracted driving? Obviously it’s the wireless carriers, the car companies and companies like Apple and Alphabet that are feeding the endorphin-drenched proposition of smartphone usage in cars. Dr. Gupta helps us understand the cognitive deficits brought on by various smartphone-related activities. Significantly, he declines to recommend a solution.
Just how serious is the problem? The National Highway Traffic Safety Administration estimates that distracted driving is responsible for approximately 10% of all highway fatalities or about 10 deaths every day. Pretty serious stuff.
One of the daily Driving While Distracted (DWD) segments that unfolded this past week on CNN talks about potential solutions to the problem – but both of the solutions featured involve aftermarket devices that work with smartphone apps to disable or block certain functions while driving.
Like a dozen other distraction mitigation apps and devices introduced or demonstrated during the past 6-7 years in response to the emerging scourge, most of which have fallen by the wayside, there is no indication that these devices or their predecessors will see widespread adoption. There is no path to market described in the CNN segments, so the demonstrations of the two solutions – Cellcontrol and Groove – aren’t much more than feel good pablum.
Approaching the problem from the aftermarket essentially blames the victim – the driver, the user of the technology. These aftermarket devices put the onus on the consumer to curb their own “bad” but quite predictable and, according to Dr. Gupta, understandable behavior.
By the end of the week, the problem remains as intractable as it was at the beginning of the week. We hear from the victims and the experts but we don’t hear from the culprits: the wireless carriers, car companies, Apple and Alphabet.
Notably absent from the various segments is a consideration of Apple’s CarPlay and Alphabet’s Android Auto in-vehicle integration platforms. Apple’s offering replaces the familiar radio dial in the center console of the car with an icon-laden iPhone/iPad-like interface once a phone is connected. The Android Auto system substitutes a GoogleNow screen which all but begs for the driver’s attention to calendar and contact alerts.
As distracting as these two systems are, they are imminently superior solutions to using a smartphone without an in-vehicle connection. In fact, most drivers can’t be bothered with connecting their smartphones in their cars – the process for which can be both annoying and distracting – though it is intended to be neither.
In essence, whether the driver uses the smartphone interface provided by the car maker (some, like GM, go so far as to provide a hidden cubbyhole to remove the smartphone itself from view) or not, the smartphone in the car – which could be a lifesaver in the event of a crash – is, instead, a ticking time bomb of distraction.
The emergence of the smartphone has introduced a powerful location aware device capable of summoning assistance in the case of an emergency as well as communicating contextual alerts and navigation cues. The challenge for phone makers and car makers and application developers is to carefully tread the fine line between guiding and assisting or distracting the driver.
In a post-IoT world, devices ought to be inherently location aware. A car should not only sense the proximity of danger, it should also detect any attention deficits on the part of the driver. These systems ought to be simultaneously built into the car, the smartphone, and the wireless network itself.
The CNN DWD special overlooked a solution provider, Global Mobile Alert, which is working in close cooperation with the U.S. Department of Transportation to see its technology employed as part of the Affiliated Connected Vehicle Test Bed. The GMA technology is made available license-free to the more than 70 companies participating in the Test Bed for the purposes of testing and demonstration.
The GMA application will alert active users of smartphones to the proximity of traffic lights, railroad crossings, school zones and other hazards. The application will also communicate wirelessly the signal phase and timing of any traffic light to the driver during an active wireless call.
While distracted driving accounts for approximately 10% of all highway fatalities in the U.S., intersections are responsible for 33% of those fatalities. Enhanced location awareness, enabled by the smartphone, is the ideal solution to the problem – though the technology is best built into the vehicle as well.
The USDOT Test Bed is only one path to market for GMA. In connection with the Test Bed GMA has been reaching out to the 70 participating companies to ensure they have all obtained appropriate credentials and access to GMA’s code base. Meanwhile, GMA is reaching out directly to wireless carriers and car companies to seek the broadest implementation of its distraction mitigation technology.
Like the stricken families featured in the CNN videos, crash victims and distracted drivers alike, GMA’s chairman and founder, Demetrius Thompson, was himself hit twice by distracted drivers – events that served as the inspiration for the GMA application. The original application, developed before digital maps were widely available, combined the latitude and longitude of key locations with an alert based on the crosswalk chirp associated with assisting blind pedestrians.
The bottom line is clear. We have the technology to quell the crisis of driving while distracted. More details are available from GMA at www.globalmobilealert.com.
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
Continues to Outgrow in a Flat Capex Environment. Is September the 2016 Peak with a softer December? Lam reported June, Q4 , revenues of $1.55B and EPS of $1.80, handily beating estimates and besting relatively high expectations for a positive spin and outlook for H2. Guidance was for a Sept quarter of $1.625B in revs and $1.77 in EPS which is also ahead of current estimates.
Outperformance of the Industry coupled with Leverage…
Lam continues to do a much better job of growing revenues in a flat market as compared to the overall industry. Through a combination of share growth, served market expansion and good customer and product mix, Lam is continuing to do a better job despite the fact that capex for the semiconductor industry has been stuck in neutral for the last three years at $33B overall.
3D NAND & NVM remains the driver with DRAM in doldrums…
We note the change from describing memory as NAND to NVM which is inclusive of Intel/Microns XPoint technology which NVM but not NAND. NVM grew from last quarters 43% of business to 51% percent in the reported quarter.
Logic (read that as Intel) was flat at 7%, reflecting continued slow spending and stretching out of nodes while foundry spending was up from 23% of business to 27% of business largely on increasing spend for 10NM roll out and early 7NM buys. DRAM continued to drop from 27% of business to only 15%
While Sept Q is strong , December may slow due to lumpiness?
In trying to read between the lines of guidance for the second half of the year and Q1 guide it sounds like Q1 (September) will come in like a lion while Q2 (December) may exit a bit more like a Lam. This seemed to be confirmed by Q&A comments talking about “lumpy” business.
We continue to see these yearly cycles and ups and downs that drive the stocks and the companies back and forth but Lam has done a good job of making like a rachet and moving up the bar with every yearly cycle.
Still upbeat about KLAM merger..
Management didn’t comment very much about the merger other than to continue their unwavering support of the transaction.
Management claims that the industry is in favor of the merger as compared to the Applied TEL merger and describe it as “pro competitive”. We would not agree with the “pro-competitive” description and would suggest that though this merger is not as unpopular as the Applied/TEL deal , not everybody is in love with it.
Ruminations about Remedy Ruination…
As we have previously stated in several notes, the longer this deal takes, the more time that regulators have to ask for more “pounds of flesh” to approve the deal or find more reasons not to approve the deal. Longer is certainly not better and obviously worse.
Whereas AMAT could afford to walk away from the TEL deal after finding that the remedies required made the deal unattractive, we don’t think Lam can walk away even in the face of costly remedies as the KLAM deal gets them to the same size as AMAT and without it they are half the company AMAT is.
Management is still looking at the “next few months” and the analyst meeting (although it could always be rescheduled) is set for November so it sounds a lot like an October expectation for approval if it slips again, past that timeframe, we would then reduce the odds to less than 50%.
The stock…..
Lam’s stock has had a very strong run and deservedly so. They have done a great job. However, much of the upside is priced in and not much risk is in the stock. We think that the 90’s and $100 may prove to be a barrier this year especially in light of what could be a soft ending to the year and KLAM merger uncertainty. Even if the merger goes through we may see significant remedies that may reduce the upside.
All in all what we heard was a confirmation of a story that was already reflected in most of the stocks- a better H2 and strong 3D NAND-
Given all this we would take some money off the table and lock in some of the great gains we have gotten….
As Pokemon Go invades the world, let me give you a firsthand player’s description of the game and why the next generation of augmented reality apps will energize the fabless semiconductor ecosystem and greatly benefit TSMC.
While I am not a “gamer” per say, I am a technologist and am always looking for new semiconductor market drivers. I am also a father of four and uncle of many more so I get to pretend I’m a kid again on a regular basis.
My 13 year old nephew came to visit last month and introduced me to Pokemon Go. I was already familiar with Pokemon since I have 4 millennial children and have suffered through years of Pokemon cartoons, video games, cards, and even Pokemon Halloween costumes that we still have in the attic somewhere.
Getting my gamer nephew outside is always a challenge so I jumped at this opportunity to combine three things that I enjoy: Walking for miles and miles, researching semiconductor applications, and beating my nephew at a video game.
Loading the app was easy, free, and from what I have read more than 100 million people have already done it. The app is true to the Pokemon series: You walk around and toss Pokeballs at Pokemon characters that pop up around town. There are designated PokeStops where you can get more balls and other items that help you in the game. There are also PokeGyms where you can train your captured Pokemon by battling others. Funny thing, there is a PokeGym right outside my regular gym so you can see the commercial applications already. In fact, I’m a bit surprised that all Starbucks are not PokeStops.
The PokeStops are a great example of the power of crowdsourcing and have clear commercial applications. Most of the hundred plus stops (Danville is lousy with them) I have visited were created by the crowd and connected with the app via the camera and GPS so you get an augmented reality Google map to follow. If you really want to know what augmented reality is download the Pokemon Go application and see for yourself. You can also turn the Pokemon Go AR off to compare.
You get experience points for doing PokeThings and credit for the miles you have PokeWalked. To be clear this only works while you are walking. I left my app on during a 3 hour car drive and also a 40 mile bike ride and got no walking credit. My nephew did catch a couple of Pokemon in the car when I drove him home so hopefully they will disable that for safety’s sake.
Of course there is a PokeStore where you can buy things which has raked in more than $200M during the games limited release in July. Pokemon Go just went worldwide (with the exception of China, India, and Iran) so expect a PokeBank rush in the coming days. Nintendo stock has also jumped adding billions of dollars in valuation.
Bottom line: Pokemon Go is a work in progress but a great example of augmented reality on your smartphone which is a preview of things to come, absolutely. Pokemon Go and other AR apps will push the sales of smartphones into the double digits again in the not too distant future (my opinion).
In regards to semiconductors, this app absolutely consumed my iPhone 6 and my data plan. Most serious PokePlayers have portable chargers because PokeWalking also dramatically reduces battery life.
This also reminds me of two keynotes I attended many years ago. First it was Andy Grove saying that software was the limitation with PCs not hardware followed by Bill Gates saying the complete opposite. Both men were right of course because as we build smartphones the software will consume them forcing us to build bigger and better smartphones. It is a never ending race that has driven the semiconductor industry since the beginning, right?
So what does this all have to do with TSMC? TSMC dominates the smartphone wafer business and that will not change for the next few years (28nm, 16nm, 10nm, and 7nm). In fact, I am betting the iPhone 7 (which is full of TSMC chips) will exceed sales expectations thanks in part to AR applications like Pokemon Go.
While no one was really paying attention, and the company didn’t have an earnings call anyway, KLAC put up a strong beat. Revenues came in at $919M versus street of $842M and EPS came in at $1.77 easily beating $1.42 street.
However this was at the expense of the September quarter as it appears that business in the September quarter was pulled into June leaving a bit of a hole in September (so far) We will see if business from December gets pulled into September
Kountdown to KLAM….
On the Lam conference call, management was very firm about the deal closing in October, which is when, coincidentally, the clock runs out (on Oct 20th).
We think that KLA’s strong performance likely increases the pressure on Lam to get the deal done. The combination of the two companies looks even better now than when it was first proposed given the strong results of both companies just reported.
Clock unlikely to be extended…
The boards of KLA and Lam could have pushed out the Oct 20th deadline on July 20th but didn’t do so. Pushing out the deadline for the deal would have given some breathing room for Lam management to get the difficult negotiation with regulators for remedies done. Instead we now have less than 90 days to get it wrapped up.
Could KLA get a better deal from Lam???
Maybe if I were on KLA’s board I would let the clock run down or out in order to get leverage to reset the deal to even better terms for KLA shareholders. Lam is very pregnant and needs to get the KLAM deal done more than AMAT needed to get the TEL deal done. This suggests that Lam would likely be willing to pay the price as the deal looks better with every quarter that passes. We don’t see Lam walking away as AMAT did….
Increased willingness to pay higher remedy costs???
With the deal clock running down Lam may be stuck paying higher remedies to regulators to get the deal done quickly before it times out. Because if the deal times out they will not only still have to deal with the remedies but also the potential of resetting the deal value with KLA….which obviously would make the deal less attractive for Lam shareholders.
We are sure that regulators aren’t stupid and know that Lam is up against a deadline and will likely use that as a lever in their favor to get a better deal.
Is a bad deal better than no deal? Probably yes…..
The KLAM deal would make the combined entity equal in size and better than Applied. Without it, Lam’s alternatives are not good as there aren’t that many good partners that can move the needle as much as KLA can.
Whatever the pain or cost in the short term will likely be worth it in the long term…
just keep calm and carry on
The stock…
We wonder if maybe we would want to put on a “straddle/strangle” trade on Lam as we get closer to Oct 20th as the volatility of the stock is likely to be very high, in both directions. Investors will likely ignore the remedy costs of the deal and just focus on wether the deal gets done or not (people will have to figure out the costs later on after the dust settles). We think there is minimally 10-15% of volatility either way.
To be clear, we do think the deal gets done but the costs may be higher than anticipated.
Time for a little fun again. Most of us played this game when we were kids. It fairly quickly degenerates into “infinity plus one” or the even more preemptive “whatever you say next plus one”. But if you’re not allowed to use infinity and you have to name the number and demonstrate how you get to it, is this still interesting? For mathematicians this falls within a very active area of mathematics called combinatorics. Better yet, you can understand interesting aspects of this area with no more than high-school math, which makes it appealing to math geeks like me.
Benchmarks for sizing very large numbers
Start with a hierarchy of operations on how to build big numbers. First you can add (1+1+…+1), then you can multiply (repeated addition: a+a+…+a), then you can exponentiate (repeated multiplication (a*a*…*a). Repeated exponentiation should be next (a^a^…^a); this is called quadration, usually written as [SUP]n[/SUP]a (for “n” a’s in the list). Then you can go to pentation and further, but notation quickly becomes a problem. Knuth’s (yes, that Knuth) up-arrow notation resolves this: a↑n is just a[SUP]n[/SUP], a↑↑n is [SUP]n[/SUP]a, a↑↑↑n is pentation, a↑↑↑↑n is hexation and so on. You can simplify the notation further: a↑[SUP]k[/SUP]n is a followed by k ↑’s followed by n.
These numbers grow reallyfast. 2↑↑2 is 4 but 3↑↑3 (3[SUP]27[/SUP]) is over 7 trillion, 3↑↑↑3 is 3 to the power of over 7 trillion, and 3↑↑↑↑3 is 3 quadratedto over 7 trillion, or 3 with a tower of over 7 trillion exponentiations by 3. The normal question to ask at this point is what these look like as a power of 10. I’m glad you asked. Apart from the very lowest cases, these numbers are so massive that there is no reasonable way to represent them as powers of 10, unless the power is also a number in up-arrow form. By comparison, the number of atoms in the universe (~10[SUP]80[/SUP]), Shannon’s number (yes, that Shannon – this is a lower bound on the game-tree complexity of chess – is ~10^120), even googleplexes (10^10^100) are puny, entirely negligible compared to these numbers.
The Knuth notation is useful but those arrows are awkward, so there’s a different animal called the Ackermann function. One version with two arguments, A(m,n), has a simple recursive definition and extends the Knuth method, in fact A(m,n) = 2↑[SUP]m-2[/SUP](n+3)-3. So A(4,2) is ~ 10[SUP]19,729[/SUP]. But the single argument version is a doozy. A(n) is an abbreviation for A(n,n) – this grows faster than almost all other functions. A(3) is 61, but A(4) is ~ 2^2^2^65,536 and each step beyond grows the number of up-arrows, putting these far beyond any possibility of representing them in repeated exponentiation.
By the way, when you get up to these levels, in a[SUP]n[/SUP] the value of n dominates and the value a (the base) becomes increasingly unimportant. So if you’re thinking that you could give these functions a big boost by changing the base from 2 to 10 – don’t bother. That simply multiplies the value of the exponent by between 3 and 4, a completely negligible change when the exponent is at these sizes and the next number up in the sequence is super-exponentiation levels larger. Put another way, yes, replacing 2 by 10 in the base makes the number bigger, but only in an uninteresting way.
The next step with the Ackermann function is to recursively nest – define A[SUP]2[/SUP](n) as A(A(n)), A[SUP]3[/SUP](n) as A(A(A(n))) and so on. At each step, you evaluate the inner function in the nesting, which gives you an incredibly huge value, which becomes the argument to the next function in the nesting. Graham’s number, once thought to be the largest number ever encountered in a math proof, is thought to be ~ A[SUP]64[/SUP](4).
Things this size can really be found (in math)
Building these gigantic numbers is temporarily entertaining, but their real point is to provide benchmarks against which large numbers found in math problems, like Graham’s number, can be measured. I’ll start with an easy example, Goodstein sequences, for which you need to understand “hereditary base-n” notation. This is very similar to conventional base-n notation (like binary) except that powers also have to be represented in base-n. For example, 16 in base 2 is 2^4 and 4 in base 2 is 2^2, so 16 in hereditary base 2 is 2^2^2. To generate a sequence starting from n, represent n first in hereditary base-2. To get the next value in the sequence, increase all instances of the base (in exponents also) by 1, then subtract 1 from the number. If we start with 4, we get:
[TABLE] align=”center” class=”cms_table_grid” style=”width: 500px”
|- class=”cms_table_grid_tr”
| class=”cms_table_grid_td” | Increase Base
| class=”cms_table_grid_td” | Subtract 1
| class=”cms_table_grid_td” | Value
|- class=”cms_table_grid_tr”
| class=”cms_table_grid_td” |
| class=”cms_table_grid_td” |
| class=”cms_table_grid_td” | 4 (=2[SUP]2[/SUP])
|- class=”cms_table_grid_tr”
| class=”cms_table_grid_td” | 3[SUP]3[/SUP]
| class=”cms_table_grid_td” | 2.3[SUP]2[/SUP]+ 2.3 + 2
| class=”cms_table_grid_td” | 26
|- class=”cms_table_grid_tr”
| class=”cms_table_grid_td” | 2.4[SUP]2[/SUP]+ 2.4 + 2
| class=”cms_table_grid_td” | 2.4[SUP]2[/SUP]+ 2.4 + 1
| class=”cms_table_grid_td” | 41
|- class=”cms_table_grid_tr”
| class=”cms_table_grid_td” | …
| class=”cms_table_grid_td” | …
| class=”cms_table_grid_td” | …
|-
It’s not obvious, but this series always terminates (returns to 0) no matter what starting number you choose. But I suggest you don’t try to complete the table above. The length of the sequence for 4 is 3*2^402,653,211 – 2. For 5 it is >A(4) and for 7 >A(8). For 8 it is ~ A[SUP]3[/SUP](3). From 12 you have to switch to an entirely new notation (and incidentally this number is also bigger than Graham’s number). This is an easily-defined sequence whose length (as a function of the starting value) very quickly becomes incredibly huge and continues to grow even more stupefyingly huge. But to make you the unbeatable champion of the biggest number battle, I want to introduce you to TREE(3).
Explaining the TREE function in detail is not incredibly difficult but would make this blog too long, so I’ll just give an abbreviated version. The idea is to build a tree with a single root according to certain rules. The argument to the function defines the maximum number of branches allowed at any node. And you cannot extend the tree in such a way that an upper part of the tree (nearer the root) can be mapped into an embedding on a lower part (this is where I’ll skip details, but they’re really not too hard). It has been shown that for any argument k, there is a maximum possible size for that tree, called TREE(k).
TREE(1) is trivially 1 (just a single branch between root and one child), TREE(2) is easily shown to be 4, but TREE(3), which doesn’t seem like it should be unusual, rockets to a size which makes everything we’ve seen up to this point completely unnoticeable. A very weaklower bound for TREE(3) is A[SUP]A(187,196)[/SUP](1). A(187,196) is already an inconceivably large number. TREE(3) is probably very much bigger than that many recursive nestings of A on the initial argument. Don’t worry about the starting argument of 1 not letting this get off the ground; A(1)=2 so the final number is far, far beyond any intuitive conception. And yet it’s a very weak lower bound for TREE(3). And TREE(3) is still finite.
To summarize:[INDENT=2]TREE(3) >>> A[SUP]A(187,196)[/SUP](1)
A[SUP]A(187,196)[/SUP](1) >>> A[SUP]3[/SUP](3)
A[SUP]3[/SUP](3) >>> Goodstein(4)
Goodstein(4) ~ 3*2[SUP]402,653,211[/SUP]
…which is ~ 10[SUP]100,000,000[/SUP] times bigger than big numbers in the physical world
So now you know how to win the biggest number game. Just say TREE(3). This number is finite (amazingly) and you can describe how it is constructed. You’ll leave your competitors in the dust. If you want to learn more about ridiculously huge numbers, there’s a great YouTube series HERE.
(Of course some opponents will answer TREE(3)+1, slightly more clever opponents will answer TREE(4) or similar, and really clever opponents will apply the recursive nesting trick and reply TREE[SUP]TREE(3)[/SUP](3). So you might want to add a rule that no-one can build on a previous answer.)
Watching a spirited debate on Twitter this morning between Tom Peters and some of his followers reminded me of the plot of many spy movies: silently killing an opponent with a lethal injection of some exotic, undetectable poison. We are building in enormous risks in more and more big data systems. Continue reading “Lethal data injection a much bigger threat”
I’ve been a daily Facebook user for many years now and it keeps me in touch with family, friends, some business contacts and even a handful of high-tech companies. My first impression is that Facebook is a very successful, cloud-based, social platform staffed with software developers and a few marketing mavens. On closer examination I’ve come to learn that Facebook in fact does have hardware engineers doing a variety of mechanical, 3D and electrical design tasks.
Source: Facebook
The first hardware projects that I recall at Facebook were for custom racks, servers, storage systems and network switches used in their data centers. Now Facebook has hardware design teams in many locations:
Island of engineers can be only so productive, so the latest trend at Facebook is to have these diverse teams meet together or even move into shared facilities in order to share ideas and get prototypes into the field much quicker than before. A new, large facility has been created in Menlo Park with 22,000 square feet of space for collaboration and is fondly called Area 404, kind of a play on words as a missing web page is typically called a 404 page. At Area 404 the engineering teams can now do modeling, prototyping and failure analysis in one spot, decreasing iteration times from weeks to just days. Folks that worked in the Connectivity Lab, Oculus, Building 8 and from Infrastructure can now form teams at Area 404 and attack new designs more efficiently.
Test and debug work are done in the electrical engineering labs, while mechanical work is done in the prototyping workshops. Computer Numerically Controlled (CNC) machines at Area 404 include:
9-axis mill-turn lathe
Electron microscope
Aquila aircraft for Internet connectivity
The Infrastructure group at Facebook designed an open rack, top-of-rack switch called Wedgeusing parts from their own labs. Engineers in the Connectivity Lab built hardware for the flying Aquilaplatform to distribute network services to un-reached regions around the globe. Finally, the VR group has designed their own hardware for the Facebook Surround 360 camera rig, outer shell, and Oculus prototypes.
Facebook really is quite the diverse company spanning both a social platform and hardware that can help to make us smarter, have better user experiences, all in a more connected world. Their new lab space was the next step in growth for the company to realize hardware designs more quickly than ever before.
The MIPI Alliance was founded in 2003 by large IDM to standardize chip-to-chip interfaces in the wireless phone (mobile) segment. The various MIPI specifications (CSI, DSI, DigRF and many more) have been adopted by the application processor chip makers (usually large IDM or fabless, like Intel or Qualcomm initially and many more joining from Asia like MediaTek, Spreadtrum or HiSilicon to only name a few). As important was the adoption from the companies developing peripheral IC like camera sensor, display controller or RF IC as most of these MIPI specifications are functional: CSI stands for Camera specification interface, DSI for Display and so on. Until 2010, not only the MIPI technology was exclusively used in mobile, but there was no demand coming from other, non-mobile, segments and consequently no strategy from the Alliance to push for MIPI usage beyond mobile.
During the last couple of years, MIPI as a low power/low EMI technology has generated high interest in other segments, automotive, wearable, medical and industrial or IoT. That’s why the MIPI Alliance has decided to open and promote the Developers Conference, and that’s good news for the semiconductor industry, for several reasons reviewed in this paper.
If you try to evaluate how many IC including at least one MIPI interface are produced every year to support smartphone manufacturing and you quickly come to several billions. Taking as an example a MIPI powered camera sensor IC (or a display controller IC), every year 100’s of million if not billion of these chips are shipped to the mobile manufacturers. One empirical, but validated economic law states that the higher the production volume, the lower the chip cost. Keeping the sensor IC example, it’s a production proven chip, equipped with a standard based, low power, low EMI interface (MIPI CSI) which can be available for any other application, like for example to interface with an image recognition processor in the automotive segment. Thanks to the huge production volume in the mobile industry, MIPI powered IC have reached an optimized, low production cost, benefiting to any kind of application.
Because MIPI CSI is standardized, the chip maker developing this specific processor can decide to outsource the interface function, buying an IP to a vendor like Synopsys, Cadence or Mixel, avoiding assigning resources to internally develop the function if he is not familiar enough with the technology. If you take a look at the MIPI DevCon agenda, you will notice that all the above listed IP vendors are giving one or more presentation, illustrating the strength of the ecosystem around MIPI. This ecosystem is active for years and includes test equipment manufacturers, Verification IP (VIP) and EDA vendors, foundries, etc. If a chip maker addressing non-mobile segment decides to integrate certain MIPI interface as an IP, he can rely on a dynamic ecosystem.
According with Peter Lefkin, managing director at the MIPI Alliance, this 2016 MIPI DevCon to be held on September 14, 15 in Mountain View, at the hearth of the Silicon Valley, is intended to “learn how MIPI technology is facilitating new capabilities within mobile while at the same time being extended to other markets, such as IoT, automotive, wearables, industrial and augmented/virtual reality”. Let’s take a look at the four tracks:
In the first track, some presentations from Mixel, Synopsys or Intel will directly address applications in automotive or IoT, illustrated by use cases. MIPI adoption is segments beyond mobile is real, especially in automotive. If you try to figure out the electronic architecture of the 2025 car, there will be numerous displays (not only for entertainment but also on the dashboard) and numerous cameras. Some cameras will certainly be used for ADAS, but it’s likely that the rear-view mirror will be replaced by electronic systems. In all cases, it will be possible to integrate a MIPI Serial Interface, like Display Serial Interface (DSI), Camera Serial Interface (CSI) or the I3C sensor dedicated interface, whose specification has just been released at the beginning of August 2016.
Some MIPI members, chip makers and IP vendors, have already developed solutions to support I3C, the second track will be dedicated to I3C and its impact on cameras and other sensors. There will be presentations from Intel, Cadence Design Systems, Qualcomm, Synopsys, Lattice Semiconductor or Microsemi. That is, two of the top three chip makers, the top two interface IP vendors and two FPGA vendors who have decided to invest into MIPI technology and who are able to offer low cost/low power FPGA solutions. Did we say ecosystem in this paper? Here is a good example of the I3C specification ecosystem!
The best standard organization, generating very smart and complexes specifications, would never be successful if these specifications never reach production and consumer adoption. That’s why the last two tracks are so important. The third track addresses verification and debug, with papers dealing from verification of mobile SoC designs to system SW development or UFS card certification test. The presentations given in the fourth track are dealing with product development and prototyping, focused on the two functional specifications enjoying the largest adoption rate (see the above picture), Camera and Display.
That’s the first time that the MIPI Alliance organize a development conference widely opened to the electronic industry (including the non-MIPI members) in the Silicon Valley and this will be the opportunity for system architects, engineers, designers, test engineers, engineering managers, and business and marketing executives working in various segments and not only the mobile segment to learn about the MIPI technology and the various MIPI specifications. No doubt that some of them working in emergent markets will realize that integrating certain MIPI interfaces in their system could be a wise decision.
From Eric Esteve from IPNEST
WHAT:MIPI DevCon: Moving Mobile Forward, the Alliance’s first annual developers conference
WHEN & WHERE:Sept. 14-15, 2016, at the Computer History Museum in Mountain View, Calif.
WHO:The conference agenda is designed for system architects, engineers, designers, test engineers, engineering managers, and business and marketing executives. Members of the media and industry analysts are invited to attend with complimentary registration.
WHY:MIPI Alliance technology is driving new capabilities within mobile and impacting markets, such as the Internet of Things (IoT), automotive, wearables, industrial, and augmented/virtual reality. MIPI DevCon 2016 will provide the latest information on MIPI specifications for implementation in mobile and other emergent markets.
TO REGISTER:Find more details and registration links at mipi.org/devcon including a $49 “early bird” registration fee available until Aug. 19.
PROGRAM DETAILS:The MIPI DevCon 2016 agenda features expert commentary and presentations from MIPI members representing the industry’s top companies working in mobile, IoT, automotive and other fast-growth industries.
The Center for Automotive Research released its own study estimating that the proliferation of car sharing services – expected to grow to 51,000 vehicles and 3.8M users by 2021 – will result in a reduction of 12,700 new and used vehicle sales a year in the U.S. through 2021.
http://tinyurl.com/j298v3b – Study Deems Ride Hailing a Non-threat to U.S. Auto Industry So what is really happening and what is going to happen?
Attendees at the Center for Automotive Research’s Management Briefing Seminar in Traverse City and certainly executives throughout the automotive industry breathed a collective sigh of relief when revised U.S. July sales numbers reported Wednesday hit 1.52M for a 0.5% year-over-year increase and a seasonally adjusted annualized sales rate of 17.86 in the U.S., the highest SAAR of 2016, according to Automotive News.
The news of this latest albeit marginal sales gain was reassuring to many but comes against a backdrop of rising average vehicle prices (up 2.5% this year to $34.264, according to Kelly Blue Book) and increasing vehicle loan and lease terms, though forecasted delinquencies (10%) are well below the peak (13%) during the downturn. Skepticism is growing that the auto industry, at least in the U.S., can extend its growth beyond current record levels and, more likely, is headed for one of its cyclical downturns.
Ford Motor Company CEO Mark Fields pointed to car sharing and ride hailing as contributing to future demand as vehicle utilization increases. AN quoted GM’s Steyn echoing Field’s optimism as she described the growth of Maven’s fleet to 3,500 cars in New York, Chicago, Boston, Ann Arbor, Mich., and Washington, DC. She noted that Maven customers have driven more than five million miles and that GM views the business as complementary.
http://tinyurl.com/zpg9bu6 Why GM is Embracing Car Sharing
Car sharing fundamentally hinges on the creation of a network of vehicles. As such, it is best suited to urban settings for the purpose of convenient access to vehicles for vehicle-less urbanites and visitors.
Before launching Maven earlier this year, GM tested the peer-to-peer car sharing waters with its investment in RelayRides, quickly discovering that sharing your car with neighbors, while a potentially lucrative value proposition, did not scale. GM is continuing to explore the P2P approach in Europe, but P2P has clearly been relegated to niche status.
Multiple car companies along with GM including Ford, Daimler, PSA and BMW are now in the ad hoc car sharing business, some offering one-way rentals (BMW, Ford, Daimler), while others require a round trip (GM). The real challenge will come from managing the maintenance of these small fleets – something rental car companies are better equipped to oversee.
In the short-term, car sharing fleets may represent a way to attract new car buyers. While AvisBudget’s ZipCar and Daimler’s Car2Go started out with low-end vehicles, there are indications that these fleets are diversifying. BMW has begun using its i3 and GM has been making a range of larger utility-type vehicles available.
Car sharing fleets may ultimately serve as a means for introducing new cars to potential buyers. So far, though, the services have not been integrated into vehicle sales operations suggesting that this collateral outcome is not currently part of the plan. (A possible exception is advantageous lease terms offered by car makers to Lyft and Uber ride hailing drivers for new cars.)
Perhaps the biggest challenge for car companies entering the space is managing the process of interacting directly with consumers. Car companies have yet to integrate new car dealers into their car sharing programs – which means they are taking on the potentially expensive logistics of fleet management directly – a function which, long-term, has “outsource” written all over it.
In addition to creating ad hoc urban fleets of shared cars, auto makers are setting the stage for car sharing as a feature. Volvo is leading the way, testing a smartphone-based car sharing proposition capable of allowing any Volvo vehicle to be unlocked, started and shared via a smartphone-delivered code.
The final evolutionary stage of transportation-as-a-service arrives in the form of shared driverless cars. The first step on this path is driverless shuttles in defined environments already introduced or contemplated from Singapore and Copenhagen to the National Harbor in Washington, DC, and downtown Las Vegas.
The remaining question is whether a network of shared vehicles complements or supplements the existing non-networked/owned car park. When asked about the impact of the higher vehicle utilization rates associated with car sharing on vehicle production, Tesla Chairman and CEO Elon Musk said, on yesterday’s earnings call:
“Well, I think the demand for autonomous cars will vastly outweigh the production capability. So it’s more in our mind that the global fleet of vehicles is about 2.5 billion roughly and total new vehicle production per year is only about 100 million. So, the fleet is basically turning over every roughly 20, 25 years. So we would have to make some truly enormous number of autonomous vehicles for there to be any land saturation because it will basically be the only car anyone wants to buy.”
Unless of course, you don’t actually buy a shared autonomous vehicle – you only borrow it. Then, maybe, you don’t need quite so many. That is the question behind Morgan Stanley’s view of peak auto demand arriving in just a few short years.
The Morgan Stanley perspective is disturbing and it does seem to ignore rising demand in emerging markets generally and China in particular. So, the makers of cars will have us all believe that car sharing will stimulate demand. CAR will have us believe that car sharing won’t negatively impact demand very much. Morgan Stanley will have us believe that car sharing will cap demand.
These are radically different views that hinge on consumer behavior and the economics of car ownership. Factor in the even higher cost of cars capable of automated driving and a budding market built on borrowed vs. owned transportation begins to emerge.
All of which means the auto industry is in the midst of a high wire act between an evolutionary advance toward ever-safer and ever-more-fuel-efficient and ever-more-expensive cars and a leap to fully automated and shared operation. At its core, shared and automated vehicle access represents a solution to traffic congestion, vehicle emissions, and rising highway fatalities.
The greatest change of all will be behavioral – ie. convincing consumers to share rather than to own. According to Steyn at Maven, consumer reaction has so far been enthusiastic. If she’s right, we all need to take a closer look at those Morgan Stanley figures.
