Detangling the Space-COTS Dilemma

by Mont Taylor, Vice President, Business Development, Teledyne e2v HiRel Electronics

Smartphones are by any standard an amazing feat. Their processing power matches that of gaming computers, they combine transceivers that cover frequencies from 600 MHz to the millimeter-wave region, have imaging abilities so good they have eliminated the point-and-shoot camera market, run all day on a battery charge, and fit in a shirt pocket. In short, the amount of leading-edge componentry inside these products is astonishing, and even though high-end smartphones top out at more than $1,000, they’re arguably a bargain.

So why can’t satellite designers use the same parts available to phone engineers, especially when Low Earth Orbit (LEO) satellite constellations are being launched by the thousands? The answer isn’t simple, and how close to this ideal is possible depends upon the degree of risk acceptable to a commercial or defense satellite project.

While traditional satellites cost a billion dollars or more, the current vogue is to compete with these “exquisite” assets with large groups of lower flying, lower cost, small satellites. The question is how to get the cost down while still being able to get the job done.

To answer this, we’ll first look at the spectrum of needs across different industries shown in Figure 1. On the left, commercial semiconductors are produced in huge volumes and are relatively inexpensive. They’re leading edge, with a wide variety of sophisticated functionality readily available. On the downside, they don’t have to last very long, they’re fragile, and models come and go, so they aren’t available for an extended period. Crucially, such parts don’t have to deal with a lot of radiation exposure. At the other extreme, one-off billion-dollar satellites are the opposite in almost every way. Careers end if satellites like these fail, so the parts used are tested extensively and expensively.

In between these extremes are a range of applications with differing needs and methods of testing. Moving to the right in Figure 1, the degree of resilience increases, as does the cost, but the number of component types available decreases substantially, and the parts that are qualified for use tend to be older designs. It is often said that the highest-level space parts often lag commercial ones by up to 20 years.

How Much Risk is Tolerable?

Makers of satellite constellations have some advantages over the designers of traditional satellites, including the ability to take more risk of failure because constellations of a thousand satellites can tolerate a few of them dying, and their relatively low cost allows replacements to be launched.

In contrast, a thousand satellites can’t be allowed to cost a billion dollars each, so costs must be several orders of magnitude lower for the economics to make sense. Working against this is the part quantities: A typical space platform might need 10 to 100 of a particular part, and constellations might need a thousand, but compared with semiconductors in smartphones, it’s a puny number. So, most chip manufacturers don’t want the hassle accompanying devices designed for operation in space.

Why Space is a Hard Place

While the answer to this question has filled many Ph.D. theses over decades, there are some examples of why space is hard on semiconductors, as shown in Figure 2. For example, while satellites float in orbit, the first few minutes of their journey to space are violent, exposing all their components to stresses beyond even those encountered in the automotive environment. And while space may be a near-void, orbiting satellites typically undergo step-function changes in temperature as they go between full sun and the earth’s shadow, suffering frequent and regular thermal shocks from -65° C to +125°C. Large, expensive satellites sometimes have heating and cooling systems that deal with these issues, but lower-cost spacecraft do not.

Another major factor is radiation, which encompasses electromagnetic radiation through subatomic particles to ions of different sizes and energies that are responsible for an extensive number of issues. The amount and types of radiation a satellite will have to tolerate will vary depending upon the altitude of orbit, with Low Earth Orbit (LEO) constellations experiencing less of some types than Geostationary Orbit (GEO) satellites.

In fact, even with shielding from radiation, most GEO satellites receive enough radiation to kill a human in just a few months. Some semiconductors are more prone to these effects than others. Logic, switch drivers, and processors are typically CMOS-based and can thus suffer badly, whereas RF circuits based on GaAs and more recently power circuits fabricated in GaN, can be less affected. A considerable number of space failures result from Electrostatic Discharge (ESD) that is frequently produced by device package lids not being properly grounded, which allows charges to build up from showers of cosmic electrons.

All this being said, the packaging of semiconductors presents its own issues. Plastic-packaged parts were long ago replaced by hermetically-sealed parts in military and space because plastic can sometimes outgas, coating nearby components in chemicals that cause corrosion. Non-hermetic sealing can also allow humidity into packages, particularly on the launchpad in humid locations, ultimately leading to corrosion of the semiconductor inside.

The move to RoHS or lead-free solder in commercial electronics has been good for the planet, but less helpful to space electronics. A phenomenon known as “tin whiskers” can occur when not enough lead is present, where tiny hairs of tin grow and can eventually short between critical parts of circuits. Growth rates from 0.03 to 9 mm per year have been reported and for this reason, space parts typically have leads dipped in tin-lead solder.

For these and other reasons, up to 50% of smaller satellites fail before their mission is supposed to end. While smallsats are significantly less expensive than the large GEO satellites, they are still expensive to design, assemble, test, and launch. They are complex systems with multiple subsystems all designed to work together to accomplish a given objective. In these complex systems, all it takes is one IC to fail and the satellite malfunctions.

Dying Before Their Time

Everything must die eventually, including semiconductors. An old but good mental model of general reliability of electronics is the bathtub curve (Figure 3)—the observation that parts frequently die in the largest numbers early in their operational life (infant mortality), often due to process and workmanship variations. If the design is well understood, the failure rate settles down to a relatively modest and constant level, and the causes of failures tend to be random.

Eventually, wear-out mechanisms dominate, and the number of failures rises at the end of the useful life. A sketch of this is shown in Figure 3 (a). Obviously, something as complicated as a satellite is a system made up of thousands of parts and is only as good as the weakest link: Failure of individual parts can bring the entire system down.

For space, there are additional issues affecting reliability, including radiation, as shown in Figure 3 (b). Broadly speaking, these effects can be cumulative, and as this happens they affect some kinds of semiconductors more than others. For example, the leakage in the many transistors that make up integrated circuits can rise over time, or the threshold switching voltage can change, eventually affecting behavior. Alternatively, single events can be damaging depending on their energy, mass, and where they hit; they can cause transitory events like a logic “one” being read as a “zero,” or more serious latch-up events in which a satellite’s entire system must be rebooted if it can be salvaged at all.

Balancing Risk Versus Cost

It’s impossible to be certain that a part won’t fail at a given time, so space-on-a-budget is all about balancing risk against expense. Decades of experience by engineers in traditional space have led to methods that reduce risk extensively, but the costs and time involved are very high. As an example, a company could design a new part from scratch for space or pay a foundry to take an existing part and redesign it accordingly, if a company can be found to do it. The typical result is an outlay of $4 million to $6 million and a risk that the part will become export-controlled.

An alternative is to characterize the radiation tolerance of individual wafers and keep track of where on each wafer a particular device came from. Process control of wafer manufacture is typically most accurate in the center of a wafer, so for space applications, at a minimum, a percentage of devices around the edge will be thrown away, reducing yield and increasing expense.

Lot traceability means tracking individual devices from when they are diced up all the way through to packaging and test, with extensive documentation to prove it. This is difficult and expensive and doesn’t fit with the typical process flows of high-volume semiconductor suppliers. These concepts are shown in Figure 4.

In addition, devices can be housed in rugged ceramic packages that provide hermetic sealing. Testing is extensive at all stages from radiation testing with a variety of different radiation types to establish tolerance, through extensive shake, rattle and roll, burn-in, and electrical stress testing to weed out failures from process and workmanship issues.

Other possibilities include providing as much radiation shielding as is practical on the spacecraft, and its effectiveness depends on the material and thickness. Even then it can only provide protection from some types of radiation. Temperature-controlled environments on the spacecraft are also an option, but this results in greater power consumption and weight and, of course, cost.

Many of these options are simply not available to designers of smallsat constellations for reasons of cost, project timeline, size, or weight constraints. While the topic of semiconductor part qualification and screening for space will never be exciting reading, it can be pretty exciting for the devices undergoing screening themselves. It is worth taking a minute to look at how this is typically done. The goal is to ensure that the underlying design is acceptable after which the following is performed:

Test to make sure the parts are radiation tolerant up to an acceptable level of the types of expected radiation. As radiation tolerance of components often varies by wafer and by position on the wafer, individual devices must be tracked from before they are diced up through all subsequent stages and have documentation to verify it.

Weed out devices that would form the infant mortality part of the bathtub curve because of workmanship issues such as poor wire bonds, packaging issues, or process problems such as slight mask misalignments during manufacture that can cause failures later. Testing is accomplished through electrical and temperature stress testing and burn-in. These tend to be a combination of steps applied to the whole batch, then more extreme steps applied to a smaller number of parts. The latter are tested and discarded but demonstrate that the remaining devices in the batch are likely to be robust.

By the end of this process, there should be reasonable confidence that the devices are well made with enough miles on the clock that they have passed through the infant mortality part of the bathtub curve, and individually traced and documented to be tolerant enough of radiation for the mission.

A Commercial Part on a Rocket? What Could Possibly Go Wrong?

Going back to the questions raised earlier, the aim is to see whether variants of commercial parts can be used in space while keeping risk within acceptable bounds for the mission parameters and architecture, and whether this can be done at lower cost and faster than would traditionally have been done for a space program. So, can space reliability be realized on the cheap?

One possibility for achieving this is buying standard commercial plastic devices and employing the services of a third-party test facility to screen components to various additional criteria. However, once the test company exercises the device beyond commercial limits there is no recourse: The user loses the cost of the devices plus testing.

Another option is to use devices that have been qualified for automotive use per AEC-Q100. These devices have successfully passed very stringent automotive qualification and test criteria and are a “cut above” standard COTS with one glaring deficiency: the automotive market is not concerned with radiation effects.

So, satellite designers must research the technology in which the automotive device is fabricated to get an indication of potential radiation tolerance. The commercial organizations supporting these devices are not always forthcoming with internal design and process information, especially for inquiries related to relatively small volumes required for space designs.

The next option is to evaluate devices packaged in ceramic and qualified for military and defense applications. However, as with the automotive option, radiation tolerance is not part of the qualification of these devices. Consequently, the satellite designer must research the underlying technology to estimate the potential radiation tolerance. The number of device types available in ceramic packaging in this segment is also much lower and they often lag in performance factors such as speed, CPU performance, and power efficiency.

Taking parts from non-space categories is likely to have a higher probability of success the further to the right you go in Table 1. In each case, the radiation performance is a missing factor that must be established early. For many traditional space applications, a minimum TID (Total Ionizing Dose) tolerance of 100 or 200 krad(Si) might be required, whereas for LEO constellations 20 to 30 krad(Si) might be acceptable given the lower radiation exposure at these orbits and shorter mission life. An example of some of the tradeoffs of using a plastic part for a short low orbit mission is shown in Table 2.

Teledyne e2v HiRel Electronics is a leader in this area, with a long history of working with customers who have tried all of these options. We supply semiconductor parts qualified for the GEO market but also tailored for the lower cost constellation and military markets. One way of keeping costs down is to offer a standard range of parts with the type and degree of testing adjusted to customer requirements (“Source Control Documents” or SCDs).

Typically, Teledyne works with selected semiconductor vendors who have no interest in servicing the military or space markets, but design and manufacture leading-edge devices with a high probability of passing radiation tolerance. The company will buy, characterize, and store entire wafers of devices, either for internal use or at the direction of customers. After this, Teledyne will then characterize the customer’s SCD and provide surety of supply and a detailed documentation trail when requested.

Conclusion

As should be obvious having read this far, a long list of trade-offs must be considered at every stage of building any satellite, whether LEO or GEO, with the latter the most challenging. Nevertheless, builders of the latest smallsat constellations have spent the time and effort required, evidence of which is the growth of this market, that by most forecasts should reach more than $7 billion by 2026.

(121)