TORONTO—Emerging memories could find their way to space, but established memories will continue to be launched into orbit because of their proven reliability and predictable behavior in what is a harsh, airless, and unpredictable environment.
With SpaceX launches now a regular occurrence, the recent successful landing of the Mars Perseverance Rover, and even a proposed hotel in space, there’s a strong demand for extra-terrestrial computing and memory capacity. But even since the earliest of NASA missions, protecting memory from radiation and extreme temperatures has been a key challenge — and an expensive one. Even today, however, it’s not always the most advanced technology that solves the problem. Older DRAM and SRAM are still commonly used, although emerging memories such as magnetoresistive random-access memory (MRAM) and resistive random-access memory (ReRAM) may offer solutions.
But regardless of the type of memory, price is just as important factor as the very specific performance, reliability, and radiation requirements in aerospace applications. Spin Memory believes MRAM may be able to meet these demands (where DRAM and SRAM can’t) because of its radiation immunity, operational temperature range, performance and security, as well as being capable of supporting the artificial intelligence platforms that are finding their way into space, said Chuck Bohac, the company’s senior director of business development. “They’ll need a lot of non-volatile memory for those architectures, and in space, the environment is a little more challenging because of the high energy particles that disrupt standard flash and SRAM and DRAM.”
For space-based applications, designers must go to extremes to reduce the number of single event upsets and influence from these high energy particles on the CMOS, he said. MRAM becomes more valuable because the bit cell is inherently immune to radiation. Since it’s magnetically based rather than charge based, the bit cell doesn’t flip when it’s struck by a high energy particle.
“There’s no charge actually fundamentally involved with the storage of the memory itself,” added Tom Boone, Spin Memory’s director of MRAM-Defense and Aerospace. “It’s simply the magnetic orientation of the metals.”
Memory has presented one of the biggest challenges in the realm of radiation-hardened (rad hard) electronics, he said. “Logic is pretty well understood. There’s both design tricks and process tricks that designers can use to actually develop the CMOS to be very robust from a logic perspective, from a power perspective, and all the logistics of the digital processing.” However, because almost all CMOS memories are charge-based, they can be disrupted by high energy particles and radiation, whereas MRAM is strictly based on magnetic materials.
In addition to its immunity, said Bohac, MRAM the benefit of being relatively fast when compared to a compared to a lot of other non-volatile memory technologies. “We’re approaching DRAM speeds in terms of programmability.” MRAM also includes error correction techniques and a high level of endurance that makes it extremely valuable for space-based applications and has already demonstrated the ability to handle extreme temperatures.
Pros and cons
Functional safety is also a big part of Spin Memory’s intellectual property, said Boone, including a CMOS design that helps the company monitor the health and performance of the memory to improve its longevity. While some missions are relatively short, he said, customers want a clear understanding of what the life of the part is. “Even when things do start to fail, they want to be able to utilize it as long as possible for these missions.”
The original MRAM that’s been used in space was based on toggle MRAM, said Boone, and it’s been quite successful in part because it’s very rad hard. However, it has size limitations and tops out at 32 megabits; otherwise, it’s a great technology for space-bound applications. Third generation MRAM technology is Spin-Transfer Torque (STT) MRAM, which has more density, while still being relatively low power — another precious commodity in orbit.
MRAM is not the only emerging memory that has the characteristics that make it viable for space. ReRAM, specifically in the form of conductive bridging RAM (CBRAM), can also handle exposure to radiation. It’s one the reasons that Adesto Technologies, acquired by Dialog Seminconductor in early 2020, has positioned it as an option for small medical devices that are sterilized using radiation, which is something you can’t do with a floating gate technology, said Gideon Intrater, now Dialog’s vice president of strategy and technology. As soon as the device is bombarded with gamma rays or other types of radiation, the device is erased. Typically, the requirement is to be able to program the device before it’s sterilized.
He said the company has shifted focus from building discrete CBRAM devices to embedded applications as there are few additions that need to be made to standard CMOS processes. And because the technology is resilient against all sorts of radiation, including gamma rays, it could be used for aerospace applications. In fact, levels of radiation fore medical applications significantly higher that what might be encountered in space, said Intrater. “You’re trying to do your worst in terms of flooding the system with radiation.”
In space, however, there are other factors to be considered in the design process, including extreme temperatures and other robustness requirements, which is why Adesto hasn’t looked skyward—there’s a lot of investment that has to be made for relatively low volumes of parts sold. That doesn’t mean its CBRAM technology couldn’t end up in space-bound applications in systems built by companies that specialize in building electronics for spacecraft. “There’s really no reason why they wouldn’t take a CBRAM technology and apply it to their devices.”
Even though the business is there, the kind of design thinking that goes into building spacecraft systems that require memory and make money off them is a very different business model, said Intrater. “It’s just not a typical business for a semiconductor company.”
Protecting memory is a pricey exercise, said Jim Handy, principal analyst with Objective Analysis, and in some cases, it means surrounding them with additional materials that can add weight to a spacecraft. The problem is most conventional memories including DRAM, NAND flash, and EE PROM do not respond favorably to radiation such as alpha particles. Without adequate protection, these alpha particles penetrate a DRAM die and generate a high density of holes and electrons in its substrate. The result is an imbalance in the device’s electrical potential distribution that causes stored data to be corrupted.
Error correction can address some of the challenges that memories encounter in space, said Handy, but it adds cost, in part because good error correction requires a whole lot of extra bits of memory as well as an error correction engine. “All that stuff is going to cost more power and power, and even before weight, power is a huge factor in satellites.”
Even if you can reduce the amount of memory requiring a protection or correction with an emerging memory such as an MRAM, it’s not the most favorable business case because of the relatively low volumes in extra-terrestrial memory market. “Space is not going to be a good application for semiconductor because it’s not going to cause high unit volumes to ship,” said Handy. Emerging memory companies that are offering ReRAM and MRAM are all aiming the same target, “and the target just can’t consume very much memory at all.”
One company that’s very much focused on space is BAE Systems, and it’s put a lot of memory in space over the years, said Jason Ross, chief engineer for Memory Product Development at BAE Systems, and even today, space-bound computer systems still rely on older technologies, including SRAM and DDR3 DRAM. The company has more than a hundred of its Rad750 single board computers flying in space, and it has contributed to many high-profile extra-terrestrial systems over the years, including the Perseverance rover and several generations of Mars rovers, as well as many GPS satellites. He said one of the unique aspects of BAE’s space-based computers that is they would generally be considered old technology.
The old standbys
Even as SRAM is being eschewed here on Earth for less expensive memory technologies where feasible, it’s been the been the preferred volatile memory in space, said Ross, and the company has several generations of SRAM products going back to the late 1980s. “Even computers that are built today to go into space systems still rely very heavily on a SRAM. What you might consider as very old memory technology in the commercial field still has a great use in a space.”
Some of BAE’s newer products have computer architectures that use DRAM, too. But regardless of memory type, as a company building systems for use in orbit, there are two key areas for consideration — the environment and the radiation. “You get outside of earth atmosphere and it’s not nearly as compatible with everyday electronics,” said Ross.
The extreme temperature differences are the obvious environment factor, and systems need to operate across a full temperature range that meets military specifications. And although less of a concern for memory specifically, lack of oxygen is a problem for electronics overall, he said. “Some of the cooling techniques that you might see used on the ground don’t actually work because we don’t have any air. Cooling can become an issue, and when you’re talking about a volatile memory such as DRAM, which is sensitive to temperature, that that’s another problem we ended up having to deal with.”
The radiation challenges are due in part to the fact that systems lose some of the natural protection they get from the Earth’s atmosphere from particles that interact with BAE’s systems, said Ross. “Those particles can basically flip the state of a memory bit.” That can lead to the equivalent of the “blue screen of death” because the radiation exposure changes the data stored in memory, which he said is considered to be “non-destructive” because the system can be reset.
But radiation exposure can also lead to physical degradation of the part. “The radiation is harmful to a lot of the commercial technologies that are used in integrated circuits,” said Ross. “Over time, it slowly degrades the part.” That leads to a permanently non-functioning system, and the consequences can be catastrophic.
But unlike SRAM or DRAM, MRAM’s magnetic nature means memory relies on a change of resistance rather than charge. However, even though SRAM and DRAM can be affected by radiation exposure, it can be countered through engineering that is radiation hardened by design, said Ross. “We use a lot of the techniques.” One of them is error correction code (ECC), which is typically used in ground-based systems for applications that require high reliability, as well as adding redundancy. In addition, he said, there are other circuit techniques that can be employed to mitigate how radiation affects the part, right down the transistor level.
Some techniques and testing to qualify a part overlap with those that are employed for automotive applications, which also require a high degree of reliability and inflict a wide range of temperatures onto a device in a harsh environment. “We do have to significantly improve the reliability of the part and how long the part operates,” said Ross. “Space is more stringent than automotive.”
And that adds costs. “Parts that are made to fly in space are definitely more expensive than a commercial part,” he said. “The initial development tends to be the biggest piece of the cost. The other thing that drives up the cost is that we tend to be relatively low volume. A commercial memory vendor might be selling millions of parts a year; we might have cases where we only sell thousands of parts.”
For a company like BAE, selling thousands of a product type a year is a good year, Ross said, and while it’s possible a device containing memory aimed at space applications might have broader applications on the ground, it’s not a priority for the company. In addition, it tends to move slower from node to node than the commercial industry does. “We tend to be one or two generations behind.”
Spin Memory, however, sees other markets such as industrial and automotive as a longer-term target for products initially developed for RAD-HARD space and defense use because the prices will come down and volumes will go up, said Boone. “We’ll be able to leverage the technology is as we continue down the evolution of the learning curve.”
Gary Hilson is a general contributing editor with a focus on memory and flash technologies for EE Times.