Google FeedBurner is phasing out its RSS-to-email subscription service. While we are currently working on the implementation of a new system, you may experience an interruption in your email subscription service.
Please stay tuned for further communications.
Get email delivery of the Cadence blog featured here
GOMACTech is the GOvernment Microcircuit Applications and Critical Technology conference. Basically military electronics, with a sprinkling of other government topics like spectrum allocation and space. This year it was held in Reno, Nevada. Next year will be the 50th year it has been running, so I'm not sure what it covered in 1968. I hope that like IEDM they put all the old papers on a USB stick. It won't be the 50th conference since it used to be biennial, but they will be having some special events anyway. Next year it will be in Miami.
It is certainly a different sort of conference to ones I've been to before. I've never had to show a passport or green card to prove I'm a "US person", and never had to sign an NDA before acknowledging that some of the material presented falls under ITAR and may not be disclosed to foreigners. So I won't be blogging those bits. Furthermore, I've never been to a conference that started with the national anthem. The theme of the conference this year was Technologies for Secure Spectrum from DC to Light.
The conference proper starts on Tuesday, and Cadence was one of many exhibitors. Monday was a day of the Trusted Suppliers Industry Session. The US military and other government organizations need to be able to build semiconductor products where the supply chain is completely under appropriate controls. In particular, it must be on US soil and only accessed by people with adequate security clearance. This isn't a problem for specialized processes that are far off the leading edge of semiconductors, there are over 30 foundries in the US (I was surprised the number was that high) but for advanced digital, most of the capacity is in Taiwan, and even some of the on-shore capacity, such as Samsung's huge fab in Austin, don't tick all the other boxes. In John Cleese's words from one of the Amnesty International stage shows, "We can't have Ruskies in the secret service. It wouldn't be secret."
A trusted supply chain begins with trusted design and continues with trusted mask, foundry, packaging/assembly, and test services. One particular trusted supplier of note is GLOBALFOUNDRIES, who took the baton from IBM when they purchased their semiconductor business. It was clearly regarded as a major accomplishment to transfer all the wafers in the Burlington and Fishkill fabs from IBM to GLOBALFOUNDRIES without breaking trusted status. I hate to think how much paperwork was involved in that, especially given that GLOBALFOUNDRIES is owned by the government of Abu Dhabi.
I've never seen them talk about it, but having fabs in the US has to be an edge for Intel custom foundry (ICF) that the offshore foundries cannot match. Thinking about it as I type, it just occurred to me to wonder whether the Altera acquisition was some sort of shotgun marriage under encouragement from the US government, to ensure a trusted supply of FPGAs. You probably already know that since the military builds a lot of systems in very small volumes (by semiconductor standards), they are big users of FPGAs.
The day actually opened with a fascinating presentation by the Chief Security Officer of MGM Resorts International. I will not cover that here, I'll give it a separate post in a few days' time. As a teaser, I'll tell you that gaming is not one of the things he worries much about, even though they own 40% of the big hotels on the Las Vegas strip and more in Macau. He really worries about something that is firmly on our radar in the semiconductor industry: IoT.
The conference proper opened the following day with a keynote from Paolo Gargini, for many years director of technology strategy at Intel, chairman of the International Technology Roadmap for Semiconductor (ITRS), and now chairman of its successor, the International Roadmap for Devices and Systems (IRDS). I'll cover that in its own post, too.
Following the keynote, there were three presentations that made up the Jack Kilby Lecture Series. You probably recognize his name as one of the main people who invented the integrated circuit while at TI, and for which he won the Nobel prize in physics in 2000. I assumed that the lecture series had been started following his death in 2005, but in fact it seems the lecture series was started in 1980, with an inaugural lecture by Jack Kilby himself. The focus of the lecture series this year was spectrum allocation.
The three lectures were by:
All three people were talking about similar issues, how to share spectrum between commercial and government entities. Fred had the government/military view, Preston had the view from Google as a representative of private industry, and Raytheon sort of straddles the divide as a company that is also a major military contractor to the government. Since all three of them covered the same ground from different angles, instead of doing a sort of chronological report, I'll summarize the arguments.
The basic problem is that there is a limited amount of spectrum. In the past, this was allocated statically in frequency bands (the above picture shows how complex it is). The "received wisdom", that Fred insisted is not true, is that the military has a lot of spectrum that they don't use but won't give up. The military always has the argument that military use is priority one, since there are no other priorities if we don't defend our country. After all, the tagline on the bottom of Fred's slides was "Maximizing Warfighter Operational Effectiveness."
A few things have started to change the perception that the military needs to keep every kilohertz it can, forever, even if it doesn't need it right this moment:
Fred seems to be the top spectrum guy for the DoD, if I understand his job correctly. He handles all spectrum policy including international engagements. The top of his priority list is more spectrum sharing. This requires technology innovation. But if everyone can operate as seamlessly as WiFi usually does, then everyone wins since spectrum is used very efficiently. The problem is not just a US problem, many standards are now international, so the problem is global.
Preston said he felt that he'd been invited for a cage match with Fred over spectrum, but "they have finally seen the light" and everyone knows sharing is the future. When a cellphone had one band, then sharing that band was tricky, but now they have as many as 14 so some can be shared dynamically. Spectrum is a way to spur economic growth and to realize the full potential of government-held spectrum.
Another big challenge is the way spectrum is auctioned. It needs to be venture friendly and innovation friendly. Nobody can innovate if the first thing they need to do is pay $10B to buy a national spectrum footprint. Currently we have too much spectrum with presumption-of-renewal licenses. "It's as if we built the roads for ox-carts but now nobody can ever put a car on them."
The spectrum policy has directly caused a low-risk, monolithic usage model since you can't start out small and grow (like Google, Amazon, and Facebook all did). So we are stuck wth AT&T and Verizon with the same business approach, same business model, same equipment, no experimentation. One example Preston has was a sports venue. You want to use spectrum in a stadium? Bad luck, it is sold on a regional or national basis, so you can't. As a result, you can't start small and build out.
Even the commercial area could share more. If you buy stuff for your house, you can only buy WiFi. If a carrier buys stuff to cover your house, they can only buy LTE. But there's no reason that the technology can't be shared, that your little wireless router couldn't also be using LTE (on much lower power, of course) and still allow you to make phone calls on your smartphone. Spectrum class no longer needs to partition the kind of technology we can use. Plus we can constantly refresh the technology in the band since semiconductors get more capable over time.
This fits the needs, too. Dense indoor bandwidth growth is the most expanding segment, not outdoor. Indoor is not well suited to macro-cells (big basestations by the freeway) due to density of users and loss penetrating walls. For 5G this is even more true. Building penetration losses can be 30-80dB, meaning we will need tens of millions of access points, not tens of thousands.
The shy-about-his-first-name G. Van Andrews from Raytheon said he didn't really need to use most of his slides since everyone had already said it. His perspective was less commercial and more military, but as he said, we "cannot expect worldwide access to any spectral regions...even in time of war." In the short term, relocation costs of moving a capability to a new frequency are large. But a lot of those costs are covered by spectrum auctions, and even more could be if the regulations were tweaked to allow using funds from one auction to clear band that might be unlicensed (like WiFi).
One thing Raytheon are working on, and everyone else it sounds like, is the "nirvana of being able to transmit and receive in the same band" which would double the effective bandwidth of any band. DSP technology is getting powerful and so it has the possibility of removing the transmitted data from the received data. The challenge is that the transmit power is orders of magnitude higher than the receive power, so it is like having someone whisper to you while you are simultaneously shouting at them as loudly as you can. It is hard for your brain to separate the whispered signal from the shouted noise.
The big takeaway, if you want a single sound bite, is that spectrum sharing is how bands are going to be shared, between commercial entities, between government and commercial entities, and even in your home between equipment you pick up on Amazon, and the equipment that the carriers buy from Nokia, Ericsson, and Huawei.