Small sats, custom microelectronics, and the end of Moore’s lawStory
June 15, 2020
Small satellites and their reduced size, weight, power, and cost (SWaP-C) requirements are challenging microelectronics suppliers to deliver the performance of commercial technology while also maintaining reliability. I discussed these trends in a McHale Report podcast with Tom Smelker, vice president and general manager for Mercury Systems Custom Microelectronic Solutions in Phoenix, Arizona. We also covered sensor processing tends and how artificial intelligence will impact future space applications. Smelker also discussed what he calls the end of Moore’s law and its potential impact on the military electronics market. Edited excerpts follow.
MIL-EMBEDDED: Can you please provide our audience with a description of your responsibilities, what your group does at Mercury Systems, and how the company participates in the space industry as well as the military industry?
SMELKER: At Mercury, I’m the vice president and general manager of custom microelectronic solutions. What that really is, is two-and-a-half and 3D heterogeneous computing, focused for defense and space. We really see ourselves as being that channel between the semiconductor industry and the defense industry for these applications. We’re focusing on enabling our customers to move processing to the sensor edge.
MIL-EMBEDDED: Do you define sensor edge as where the sensors are before the information is even downloaded or downlinked?
SMELKER: Yes, exactly. We can package the microelectronics in smaller packages and put them right there where the sensor is and operate on the data [at that point].
MIL-EMBEDDED: What are the factors driving the development of microelectronics for military space applications these days?
SMELKER: [It is] really being driven by the low-Earth orbit [LEO] being the new frontier. And it’s a really a drastic environmental difference from what previously was our focus on the mid-Earth orbit (MEO) and [geosynchronous Earth orbit] GEO or deep space. So, it brings a whole new capability with being able to use commercial silicon. Small sats are really designated, in that LEO orbit, for short-mission durations. So, the mission duration also changes that equation as well. When you look at it, the risk equation for the GEO, deep space, and even MEO, is completely different than the LEO small-sat opportunities of today.
MIL-EMBEDDED: Small sats have created some headaches for the more traditional payloads of suppliers of microelectronics, especially in terms of radiation-hardening, because their cost constraints are so low, not just about size reduction, but cost. You have a payload that in the past would’ve cost a million dollars, but now they have to deliver the same performance for under $300,000. So, how do you see those changes impacting spacecraft payloads? Is there more COTS [commercial off-the-shelf] instead of custom or is it more just figuring what you need to test versus what you don’t need to test?
SMELKER: It’s both really, I would say; as I mentioned just before, the cost equation for LEO is completely different than the other orbits. And so, that equation has changed. I like to say for the LEO orbit that you have to leverage COTS, astutely, and it’s really focused on custom assembly and packaging that enables reliability. The SWaP-C [size, weight, power, and cost] side of it is a lot different. The small sats also drive the weight performance there as well. So, getting that processing right, that sensor edge that we were talking about earlier, is key. [As is] doing that with commercial silicon in an astute manner that keeps the reliability high enough for the shorter duration that these systems spend in that reduced-radiation environment.
MIL-EMBEDDED: Speaking of the radiation environment, is there anything you have to do to deliver that same performance that you would get in the higher-priced systems to get the necessary rad tolerance, or is it more an 80% solution sort of thing?
SMELKER: I wouldn’t say 80% solution because you have to have more reliability than that, but there’s a lot of upfront testing, early simulations, upfront testing, and analysis to understand where the susceptibilities are. Once that is understood, we’re able to collaborate with our semiconductor partners as well. Then we can design the assembly and packaging to reduce those susceptibilities to that right risk level. It can’t be 100% like the MEO and GEO orbits, where cost isn’t the risk. It’s how long the system can stay up there. It’s really about selecting the right components, doing the analysis, and designing the solution to that environment. And cost is critical in all that. So easy, right?
MIL-EMBEDDED: Going back to Earth, but still talking about sensor processing at the edge, in the last DoD budget requests, the hottest funding areas were for radar upgrades, electronic warfare (EW), and the like. How are those requirements affecting microelectronics designs? Are there similarities to space or is it a different customer need?
SMELKER: That’s a great question. This is an exciting time to be in defense and working on sensor systems. I’d say we’re in a renaissance era for microelectronics. And if you look at it and at commercial and at defense sensor systems, we’re really in what I call the data age, and we would like to be in the information age. People think we’re in the information age, but until we have the processing at the sensor and can operate on all the information that those sensors are creating, in the right amount of time, we’re going to be stuck in the data age.
So that’s where heterogeneous computing really starts to change things. And that’s why we’re seeing a growth in what we call two-and-a-half, 3D heterogeneous computing, where we’re really designing the processing capability to have the right performance be there at the sensor.
I’d say another key part of it, when you look at the radars and EW, is it solves the latency side of it. [Such as] how much time it takes to see the signal react and process it. So being able to process that information much earlier in the sensor chain than before increases the performance, but it also reduces the complexity of these systems as well. You’re going to see where we’re actually reducing overall costs and increasing performance at the same time and increasing reliability.
MIL-EMBEDDED: The space and military microelectronics market is a low-volume market. A lot of companies are able to leverage higher-volume commercial markets with their production lines to maybe drop the prices or component costs down for the military market. Mercury Systems is opening a new facility in Arizona to focus on this technology. Is that something you will leverage in the new facility or is the focus still going to be mostly military as with other parts of Mercury Systems?
SMELKER: Defense is the focus for Mercury, and we’ll maintain our focus on defense, but we are very focused on partnerships with the semiconductor industry. I would say the semiconductor industry is great at the low-mix, high-volume markets. And we’re really good at the high-mix, low-volume markets and being able to take in unique requirements, unique use-case requirements from the radars and EW systems, that we [previously mentioned], and being able to tweak that for [those] specific systems. So, we’re going to continue to stay focused on the defense market, but if there’s opportunities that are low-volume, high-mix, that Mercury can support a customer in, obviously we’ll support that.
MIL-EMBEDDED: Is it safe to say, too, that Mercury Systems is sourcing from some of those higher-volume commercial semiconductor markets? Would this also help the price point?
SMELKER: It will. It definitely helps the price point, and so we will see the price point for semiconductor electronics go down. But at the same time, you’re going to see the requirements and demand for more performance and more innovation in those semiconductor components increase as we move those components to the sensor edge. Unique requirements are going to come out of that which will drive a lot of innovation into the microelectronics [arena]. I’d say the microelectronics are not going to get simpler for moving to the sensor edge, but they’re going to get smaller and be very focused on size, weight, and power. But they will also have increased capabilities to process the information.
MIL-EMBEDDED: Speaking of complexity, at the Embedded Tech Trends Conference this winter, you gave a presentation talking about the end of Moore’s law. Now that might be good news for some people, but bad news for other people. Can you talk a little bit about that concept and how it impacts the military market?
SMELKER: Yes, absolutely. It’s funny because of Moore’s law; the potential end of Moore’s law is almost like talking politics to some people: You don’t do it in mixed company sometimes.
There’s a buzz that Moore’s law is dead. You see the slowing of innovation in the semiconductor industry and that’s really not true. If you look at what Gordon Moore mentioned, he said at one point that it will not be cost-effective to continue to shrink the transistor and have the doubling of capabilities every 18 months, [which is] the trajectory we’re on. But if you actually read his white paper on this, the last page of it talks about where we’re going today. It’s all about heterogeneous computing, rightsizing the capability for the system itself. For instance, for the last 20 years, we really were trying to solve performance increases by putting more and more capabilities within the same silicon, adding processors, adding more memories, adding more interfaces, bringing in functionality that used to be on the board into the silicon. So what it did, was it grew the size of the silicon. It grew the design cycle costs, extended the design cycles, increased the cost of the design cycles, and increased the cost of testing and verification. But then it also reduced yields, because now you have much more complex, larger devices.
[As for] where we’re going now, you don’t need the memories to be at the same technology node as the processor. You don’t need the interfaces to be at the same technology node as the processor as well. When you start designing chiplets that can handle, say, the memory, at a much older technology node or some analog to digital interface at a different technology node than the processor or FPGA [field-programmable gate array], you start to be able to really create all these chiplets and chips that programs can design unique solutions for and rightsize the processing capability. It’s actually a very exciting time to be in the microelectronics business and see how this is revolutionizing how our everyday life is changing, both militarily and commercially.
MIL-EMBEDDED: We talked about innovation. We talked about what’s going on in space and in the military. Don’t limit yourself to that market for this question: What would be a game changer in the microelectronics world, let’s say, five to 10 years out? Predict the future.
SMELKER: I would say what’s going to really change the market and drive the market is artificial intelligence (AI). And if you look at it, AI is going to leverage the heterogeneous capabilities that were designed today, because now we can rightsize those AI algorithms to process that data at the sensor edge that those algorithms will be tuned for. So, you’re going to see a growth in heterogeneous computing over the next five years, but where that will also lead you to is, today, we are processing architectures based on Von Neumann’s processing architecture. And where AI wants to go and where AI will push it is neuromorphic processing, and there’s a lot of investment going on in neuromorphic processing. Heterogeneous computing, by the way, is how neuromorphics are going to be integrated. AI is going to be that push.
So you’re going to see over the next five to 10 years capabilities in processing sensor information, both in military [applications] and in your home, your car, you name it, just take off. That’s going to drive the new processing architectures and change everything. You’re going to see smart sensors absolutely everywhere. If you look beyond 10 years, where’s that going? There are innovations being made today in quantum. Obviously, we’re not there. Are we going to be there in 10 years? Probably not, but we’ll have made some great progress. But 20 years from now, 30 years from now, it will be a completely different world.
Tom Smelker is vice president and general manager of custom microelectronic solutions at Mercury Systems Custom Microelectronic Solutions in Phoenix, Arizona. Prior to joining Mercury, Smelker spent nearly 20 years as a Senior Engineering Fellow and systems design program manager at Raytheon Missile Systems. Smelker began his career as an Undergraduate and Graduate Fellow at the U.S. Army Research Laboratory.