As I sat down in the SEMI Arizona Chapter breakfast meeting a few weeks ago, a moment of semiconductor history flew right before my eyes before the IoT sessions started.
We were seated in the cafeteria of Freescale Building 94 on Elliot Road in Tempe, a place I’d been many times before, except this time may have been the last. NXP is consolidating facilities in Arizona early in 2016 and soon vacating Building 94 for good. Glancing around the
cafeteria and adjacent courtyard, I noted a sign on one of the doors leading out from the back of the room:
“Daniel E. Noble Auditorium”
Motorolans and IEEE members might recognize that name. He’s the reason Motorola came to Arizona in the first place, a pristine dry environment ideal for building discrete transistors in 1949. Outstanding technologists at Motorola have been dubbed Dan Noble Fellows for decades, and the IEEE has recognized accomplishments with an award in his name since 2000. Dr. Noble has been gone since 1980. (Photo courtesy ETHW.org.) Seeing that sign, and realizing that it too would probably soon disappear, got me thinking.
I’m not a huge fan of the idea of disruption, especially on the IoT. For those who may not have noticed, our economy, our politics, and much of our way of life have been severely disrupted lately. Disruption and “disruptive innovation” are two very different things.
Disruptive innovation is a long-term outcome in retrospect years after the fact, not a stated objective before the start. Disruption tends to create short-term asymmetric gains for a few, and losses for many. Watching Twitter a few months ago, this Tweet during a chat stopped me in my tracks:
A5. I innovate w/o disrupting. Disrupting disparages the past. #daretobe
— Kate Nasser (@KateNasser) August 12, 2015
Those are interesting thoughts as we create the IoT. I’m all for not accepting the status quo and not doing things the way we’ve always done them – that motivates and shapes innovation, and it may in fact prove to be disruptive innovation as Christensen defines it when history looks back. But, if your primary mission with the IoT today is “disruption”, you’re doing it wrong.
The whole point of the IoT is to get things and people to work together. It’s good to see people talking about real issues, not just smart refrigerators and robots taking our jobs, in meetings like this.
Tony Massimini of Semico set the tone for the SEMI Arizona IoT sessions by saying, “There is no such thing as the ‘IoT market’.” The opportunity is there, but right now, it is hyperfragmented. Much of the focus is on gathering data and trying to make sense out of it. He then discussed sensor fusion and MEMS trends, the same stuff that we have been talking about here for a while. Two of Tony’s slides caught my attention. With smart card applications pulled out, the 32-bit MCU average selling price curve according to Semico looks like this:
Keep in mind, that is an average, but two messages should be clear. The notion that the IoT is going to drive trillions of parts and move everything to two cents per part is complete nonsense. The semiconductor industry is going to fight back against continued MCU ASP declines by adding performance and features, in order to maintain economic viability.
The other Semico slide that was very interesting (and I’ve been asked by SEMI not to show too many slides, so we are cherry picking visuals here and avoiding proprietary stuff) is a contrast between the sensor hub ecosystem in 2014 and 2015. Three things happened: sensor fusion software firms like Movea, Sensor Platforms, and Xsens got acquired; big IoT chip players like Atmel, Broadcom, and Freescale got bought; and a new oval called “integrated hub” shows up with five names in it. Those integrated hub players are Qualcomm, HiSilicon, Intel, Marvell, and MediaTek. Asians were relatively late arrivals on mobile, not so on the IoT, and what Qualcomm does here may well determine its future.
Ian Chen of Freescale whipped out a new term, “Internet of Tomorrow”, and put four mega trends under it. His main point, after a short diversion into 400B diapers of the future with sensors (technology, yes, but do we really need that?), was we are just beginning to tap all this data we are collecting. Some of it isn’t even available for machine learning, trapped in some data lake after capture or worse yet spilled after streaming.
One noteworthy observation is most of the real insight is happening in frequency domain, a powerful argument for DSP capability even at the low end. Ian sees 200 sensors per car, 20 sensors per smartphone, and 8 devices with sensors per person by 2025. He also raised a question to think about: is it time to open source security? He thinks it is baseline value most startups don’t have time or expertise to efficiently provide.
The Intel presentation from David Formisano can be summed up in one acronym: Enhanced Privacy Identification, or EPID. It would have been more interesting with the Wind River perspective overlaid. We know Intel is trying to tell an end-to-end story from edge to infrastructure, and there was little news beyond that.
The next presentation was fascinating, a look at how Medtronic views their opportunity space. One visual jumped out as rather stunning. We talk a lot about how IoT chips and MEMS sensors need to focus on reducing power consumption and smaller package size, but take a look at this Medtronic chart on miniaturization trends in implantable cardioverter defibrillators:
It isn’t chip packaging, or power consumption, but tantalum capacitors and to a lesser extent batteries that are holding further miniaturization progress back – even Medtronic with all their capability is a bit stuck. We have heard the battery laments from mobile and wearables, and cries from energy harvesting attempts that output density is too low to be cost effective for many applications. This is the first I’ve seen someone point a finger at tantalum capacitors as a limiting factor, however; granted, Medtronic is doing more than sensing, actually generating pulses. It is a reminder that the IoT is not all sensors, and there are some actuators and power electronics involved.
Sandeep Gupta from the Impact Lab at Arizona State University echoed a point several presenters made – all these wearables are super cool, but can a healthcare provider actually read the data? Do they even want it or trust it if it was not from a device they issued you? Gupta looks across the entire data lifecycle, everything from sensor scavenging (battery plus harvesting assist, for reliability) to modeling and security.
Here’s an interesting factoid: 80 years of EEG data is 5000 TB. We know the big data lake is growing to zettabytes, but in an IoT context, all that data has to be captured and transmitted. Or, does it? Gupta’s teams are working on generative models which are reducing transmit rates up to 40x in trials. Oversimplified, if samples conform to the model with partial differential equations and context looking for state changes and thresholds, don’t transmit the raw data. This could mean a huge energy savings at the edge as models emerge for various applications. It also may change the notion of postponing analytics until the cloud, or even the gateway. (Do I hear Spark on Android, anyone?)
The last presentation from Eric Miller of PADT was a bit more whimsical, but with a serious point. He defined a fictional product, the “SmartHammerXL” with integrated sensors, and explored design decisions based on use case examination, right down to Joe drives more nails than Bob.
Simulation shouldn’t stop at functional verification, and it also has to reach into multiple domains. PADT happens to use ANSYS in their service offering, and Eric’s comments apply to broader use of simulation. If you toss an IoT product out there and let people figure out how to (mis)use it, bad things will happen. He said simulation used to be on stuff that was broken or likely to fail; now, it drives every phase of product definition and exploration before it ever reaches market.
These sessions provided some deep insights, thanks to SEMI Arizona for allowing me to attend. I’m not sure Dan Noble foresaw billions of transistors on one chip, or billions of IoT chips spread across the globe. He did imagine the impact transistors could have. Today, the tech community stands on those and many other shoulders to see a new horizon, the IoT. Let’s honor that legacy with the right kinds of IoT innovation for the future, not just disruption for quick profit.
Don’t forget to follow SemiWiki on LinkedIn HERE and thank you for your support!