| Episodes drawn on | 150 |
| Cited here | 47 |
| Claims extracted | 108 |
| Gathering | concept union |
| Other themes | |
Space electronics is the design and construction of electronic hardware for spacecraft, distinguished less by exotic circuitry than by an environment in which radiation upsets stored state, and by an operating regime in which hardware cannot be touched after launch and is built in ones or handfuls.[518][77] Most small satellite electronics contain nothing unusual in themselves; the effort goes into reliability, because a ten percent unit failure rate that would be acceptable in a consumer product means losing the mission.[518] Since the early 2010s the field has been reshaped by the CubeSat form factor and by commercial rideshare launch, which together moved access to orbit from relationship-based contracting through large prime contractors to a payload slot bookable from a web page with a deposit of a few thousand dollars.[123][705] The associated engineering culture, commonly set against the top-down, waterfall, subcontractor-layered model of the 1960s to 1980s, is defined by tolerance of different part specifications and much faster iteration.[584]
The radiation environment
Upsets and their rate
Memory in spacecraft accumulates single-event bit flips over the duration of a mission, and the Voyager spacecraft suffered such a flip, recovered from and attributed to a cosmic ray strike — an attribution inferred rather than measured, because no cosmic ray detector was aboard to prove it.[77] The upset rate is nonetheless predictable rather than mysterious: irradiating memory in a beam produces a known percentage of bit flips, and multiplying that susceptibility by the measured cosmic ray density in orbit predicts the flip rate actually observed on orbit.[77] Latch-up is the more destructive counterpart, a radiation-induced failure mode that can destroy a spacecraft if power is not cycled quickly, which is why the power distribution stage of a small satellite monitors its loads and protects against it.[518]
Mitigation
The classic mitigation for space memory is to abandon density and use physically large storage cells, so that a bit is held by on the order of fifty thousand electrons rather than fifty and requires a much harder particle strike to upset.[77] The other standard defence is redundancy at the memory level: triplicate the memory and compare, so a single corrupted copy can be outvoted.[77] Triple modular redundancy generalises the idea to computation, running three copies and voting on the result; it works because the probability of two or three independent copies faulting in exactly the same way is vanishingly small, and a divergent three-way result is treated as everyone being wrong rather than a winner being picked.[497]
Material and device properties supply mitigations of a different kind. Gallium nitride is intrinsically close to immune to radiation damage, which allowed a power semiconductor vendor to build radiation-hardened GaN transistors and integrated circuits that are now widely designed into new satellites.[719] The RP2040 and RP2350 microcontrollers turn out to be unusually radiation resistant for an incidental architectural reason: they carry no on-chip flash and boot from external memory, removing the floating-gate storage that is normally the most radiation-sensitive structure on a microcontroller die.[687] Pairing such a flash-less microcontroller with an MRAM or mask ROM boot device yields a radiation-resistant processor system built entirely from commercial parts, because MRAM stores state magnetically and is itself extremely radiation resistant.[687] Early customers took commercial RP2350 parts, ran their own radiation testing and concluded the devices were suitable to fly, an instance of a mainstream consumer microcontroller being qualified for orbit by its users rather than by its vendor.[687]
Qualification
A radiation-tolerant computer is qualified up a ladder of increasingly expensive environments: radiation chambers on the ground first, then high-altitude balloon flights, then sounding rockets that go up and come back down, and only then an orbital demonstration.[497] After a year-long experiment aboard the International Space Station in 2017, one such computer was flown in two loaf-of-bread-sized CubeSats, RadSat-U and RadSat-G, which return operating data on each pass over a ground station so that upset behaviour can be measured on orbit rather than inferred.[497] Some radiation-sensing parts sold for satellite use serve a narrower purpose entirely, existing to detect a nuclear detonation and assert a hardware pin when one occurs so that the spacecraft can immediately drop into a protective low-power mode.[68]
Components and procurement
Space-grade semiconductors are sold on a per-unit basis with individual test and certification records for each device, rather than being screened by lot sampling like commercial parts.[68] Vendors price flight-qualified parts against the value of the mission rather than the cost of the silicon, because a single failure destroys a payload worth billions and the part has to survive on orbit for many years.[632] The resulting prices attach to old silicon as readily as new: a twenty-year-old CPLD listed on a general distributor at three thousand dollars carries that price because it has been radiation hardened and put through pre-testing and characterisation, not because of the die.[584]
Passive component miniaturisation lags in flight hardware, so a satellite builder asking for a 0201 flight-qualified capacitor is ahead of what the industry supplies, and 0402 is the realistic practical minimum size for spacecraft use.[596] Mechanisms carry their own material constraints, since space mechanisms use specialist lubricants formulated for vacuum: any volatile constituent in a conventional lubricant evaporates away in orbit and leaves the actuator dry.[705] Material choice is also constrained at the end of life rather than the beginning, because a CubeSat built of a dense refractory metal such as tungsten would survive reentry and reach the ground, and would not be approved; aluminium, FR-4 and standard solder ablate in the upper atmosphere and disperse as particles.[497]
Supply chains outside the parts themselves shape what is buildable. The board houses and machine shops that survive in high-cost regions such as the San Francisco Bay Area do so on aerospace and medical work, because only those segments carry margins high enough to absorb the local cost base.[268] Such suppliers price and schedule for aerospace programmes, so a hardware startup approaching them typically finds quotes with roughly twice as many digits as expected and timelines measured in multiples of years, making them a poor fit for fast-moving small teams.[268]
The CubeSat standard
Envelopes and mechanics
A CubeSat is a standardised very small satellite measuring ten centimetres on a side with a mass under one kilogram, a form factor designed so that satellites can be handled as a commodity payload.[123] Sizes are expressed in units, where one U is a 10 by 10 by 10 centimetre cube, and larger spacecraft are built as integer multiples such as 3U, giving a family of standard envelopes rather than bespoke geometry.[584] A 3U is therefore a stack of three cubes, 10 by 10 by 30 centimetres stowed, with antennas and solar panels that deploy after release so that the on-orbit envelope is larger than the launch envelope.[518] Because the spacecraft are small, launch pricing and manifesting are driven by mass rather than by the volume constraint of the form factor.[584]
The de facto mechanical and electrical standard inside the envelope has long been the PC/104-derived stack, with mounting holes on an irregular 87.24 millimetre pattern and a large 54-pin through-header that makes an assembled stack effectively impossible to separate again without damage.[518] Modules are pitched at around eleven millimetres per board in that stack, which has to accommodate the inter-board connectors and any taller components, so usable component height is a hard mechanical budget.[518] When a builder migrates off the legacy stack to its own backplane, the practical route is a legacy interface board that bridges the remaining PC/104-format modules onto the new bus, rather than redesigning every subsystem at once.[518]
Subsystems
The power chain of a small satellite comprises maximum power point trackers for the solar panels, a battery charger and battery, and a power distribution stage that also monitors the loads and protects against latch-up.[518] A fully vertically integrated CubeSat builder makes solar panels from bare cells mounted onto PCBs, plus solar regulators and maximum power point trackers, flight computers, UHF radios, and the complete attitude determination and control system of star trackers, reaction wheels, magnetorquers and sensors.[679] Buying a commercial off-the-shelf subsystem such as a power and battery module does not eliminate that work so much as move it, because the effort shifts to integrating the subsystem with the flight computer and the rest of the spacecraft, which is why some builders choose full in-house design instead.[679]
At these production volumes unit cost effectively drops out of the design trade space: one flown payload processor module had a total production run of five boards, so per-unit component and fabrication cost was irrelevant next to reliability and schedule.[349] Power rather than silicon area sets instrument architecture: with a total instrument power budget of perhaps fifty watts, only one to two watts may be available to compute a four to eight thousand point FFT behind a six gigasample per second analog-to-digital converter, which rules out FPGAs and forces custom silicon.[483]
Reaching orbit
CubeSats reach orbit as secondary payloads, because launch vehicles are sized in fixed capacity increments and primary satellites are built light, leaving spare payload capacity on most commercial launches into which small satellites are packed around the main payload.[123] Even at that scale the economics are not hobbyist: around the early 2010s the launch alone ran roughly thirty-five to fifty thousand dollars, and the hardware being flown represented anywhere from ten thousand to a hundred thousand dollars of value going into low Earth orbit.[123] Rideshare access is brokered: a launch service provider aggregates many 6U CubeSats into its own dispensers, ships those dispensers to the launch company for integration, and one Transporter-class mission can carry over a hundred separate space objects manifested by many different companies.[679] Cleanliness requirements are imposed partly by that provider rather than by the payload, since the spacecraft must be visibly clean before the launch provider will accept it onto the vehicle, independent of any contamination sensitivity of the instrument itself.[679]
The change this represents is large. Before commercial rideshare, reaching orbit meant buying a dedicated rocket, an investment on the order of a hundred million dollars once the satellite had to be built large enough to justify the vehicle, which restricted spaceflight to governments and a few extremely well-capitalised private ventures.[728] Amateur satellite organisations historically sidestepped that by obtaining launches free or at token cost as goodwill payloads; once CubeSats standardised the small-payload market and commercial providers began selling rideshare slots, that goodwill capacity was priced, and a slot that had been free came to cost on the order of three hundred thousand dollars.[401] Dedicated small launch vehicles occupy the gap in between, sized around a payload of roughly one hundred and sixty-five to four hundred and fifty pounds and standing around thirteen metres tall against roughly eighty metres for a heavy lifter.[478] Falling launch prices also push operators up the mass curve, so a builder moving from a 25-kilogram CubeSat-class platform to a 250-kilogram spacecraft bus does so because launch cost no longer punishes the extra mass and the larger bus buys larger optics.[679] Small satellites in the current market accordingly span roughly one hundred to five hundred kilograms, sitting between CubeSats and traditional large spacecraft.[679]
Schedule risk has moved with price. Launch latency rather than build time once dominated a small satellite programme: hardware was sealed into its capsule and then waited, first on the ground and then aboard the space station until crew time was available to move it to the deployer pod, each step adding delay outside the builder’s control.[220] A rideshare slip on a commercial launch is now typically measured in a month rather than the years a builder historically had to plan around.[691] The deployment mechanism remains a failure mode independent of the satellite: on one station-based launch the deployer did not respond to repeated deploy commands, leaving a batch of finished CubeSats stuck inside the dispenser with no recourse from the ground.[220]
Propulsion and attitude control
Satellite attitude control is normally done without expending propellant, by spinning reaction wheels up inside the spacecraft so that conservation of angular momentum rotates the satellite in the opposite sense.[410] Reaction wheels are a recognised wear-out item, and a batch of satellites launched together suffered a common-mode failure in which their wheels began to stick and gum up roughly a year into the mission.[410]
Where propellant is carried, its efficiency compounds, because the propellant itself has to be launched: a thruster three times as efficient does not merely save mass at liftoff, it buys three times as much manoeuvring for the same tankage.[701] A common use of onboard thrusters is orbit raising after separation, since a rideshare payload is dropped in whatever orbit suits the primary mission and must manoeuvre itself from there, while station keeping against atmospheric drag in low Earth orbit similarly determines mission lifetime.[701] The customers for electric propulsion are bus integrators, the companies that assemble complete satellites, rather than the end users of the data those satellites produce.[701] Constellation operators split on make versus buy: SpaceX builds its own argon Hall effect thrusters and flies one on every Starlink satellite, Amazon’s Kuiper is building its own thruster system, and most other operators buy from specialist suppliers.[701]
Propellant selection is as much a supply chain decision as a performance one. Xenon is the traditional Hall effect propellant and remains excellent on pure performance, so a switch to zinc is not a performance improvement but a manufacturability and sourcing decision.[701] Zinc is solid, so it packs more densely than any gas, cannot leak, and costs about four dollars a kilogram.[701] Choosing a solid propellant also eliminates a whole class of supply chain risk: no precision orifices, no clean room gas handling, and no valves rated for four thousand PSI noble gases that are made by a single vendor on eighteen-month lead times and are contended for by semiconductor fabs.[701]
Operations
CubeSats are designed to commission themselves without ground intervention: after deployment a healthy satellite detumbles, works through its commissioning sequence, establishes a stable orbit attitude and reports itself ready for commands on its own, so an operator can miss the first contact pass entirely and still find the fleet alive.[220] The characteristic failure is the opposite and is silent and terminal — some spacecraft never establish first contact after deployment, and there is no way to send anything up to diagnose them, so the hardware becomes an expensive object that eventually deorbits and burns up.[728] A satellite that deploys but never transmits telemetry is not necessarily dead, and operators have used a large dish to listen for the harmonics radiated by the microcontroller crystals aboard the spacecraft, which is enough to establish whether the processors are running at all.[518]
Because the flying article cannot be inspected, a standard practice is to build a functional replica and keep it at the operations site, so that code changes and commands can be tested against an exact copy of the flight hardware before being uploaded.[349] Satellites are built in batches for the same reason, with at least one complete flight-identical unit retained on the ground after launch so that anomalies can be reproduced and fixes rehearsed against real hardware.[518] Uploading new firmware to an on-orbit satellite is treated as one of the highest-risk operations in a mission, because a corrupted image or a single wrong bit in the transfer can leave the spacecraft unrecoverable.[323] The European Space Agency has evaluated MicroPython for satellite use precisely to avoid that: with a stable interpreter core resident on the spacecraft, changing behaviour becomes a matter of uploading a new script file and running it rather than rewriting the executable image.[323]
Constellations and services
An Earth-imaging CubeSat constellation can be flown in a sun-synchronous ring so that the planet rotates underneath the satellites, turning the fleet into a continuous line scanner for the whole surface and replacing the conventional model in which a satellite is commanded to point at a specific spot.[220] The resulting imagery is georegistered after the fact by searching an area for known ground control points and then propagating the fit through the rest of the image strip: once one frame is matched, neighbouring frames can be found and shifted until all control points are satisfied, correcting for parallax, terrain and pointing error.[220] The commercial case for such constellations rests on their short life, since the satellites cannot maintain orbit and decay and burn up after roughly five years, so the operator refreshes the whole fleet with current-generation hardware rather than flying twenty-year-old silicon.[394]
Communications constellations changed the density of low Earth orbit within a few years: Starlink was planned at a scale of roughly two and a half to five thousand satellites, and at least two other operators announced systems of comparable size.[478] A laser-linked constellation can beat terrestrial fibre on long-haul latency because light travels appreciably faster through free space than through glass, with simulations suggesting a New York to Seattle link could take roughly half the time of the current fibre route.[478] Satellite connectivity nonetheless splits into distinct market segments by ground segment complexity: a broadband constellation aimed at internet service needs a phased array ground terminal, which suits gateway and premises installations but not a low-power sensor left in a field, so broadband capacity does not automatically serve satellite IoT.[518] A service targeting devices already in the field must therefore close its link budget against whatever antenna those devices already have, typically an omnidirectional PCB trace antenna at around fifty percent efficiency, with any requirement that the customer modify existing hardware treated as fatal to the product and the difficulty deliberately loaded onto the space segment instead.[728]
Ground equipment carries its own history of constraint. Older GPS receivers took minutes to achieve a first fix and asked the user whether the unit had moved more than about three hundred kilometres since last use, because the receiver’s stored rough position and almanac had to be invalidated before it could reacquire; modern receivers lock within seconds when the previous position is still roughly valid.[301]
Design and verification practice
Spacecraft software is verified by exhaustive simulation rather than by field testing, because there is only one article and no opportunity to iterate: every retry protocol, every RF condition and every scenario expected on orbit has to be exercised in a model before flight.[519] Building on portable abstractions such as Zephyr and validating in a simulator such as Renode means a mid-project change of flight hardware does not invalidate the software effort, because the underlying interfaces stay the same.[691] Simulation-first development has compressed schedules to the point where concept to orbit inside one year is achievable for a small spacecraft, rather than the multi-year cycles of traditional aerospace.[691] Where a software defect is unrecoverable, engineers inspect the compiler’s output and check it against expectation for anything that looks wrong, a routine practice on flight software that is not warranted for the overwhelming majority of designs.[626]
Hardware verification is similarly conservative. High-value prototype boards are powered up for the first time under a two-person rule, with a second engineer independently verifying the bench supply settings before the board is connected, because a single slip on a hundred-thousand-dollar assembly is unrecoverable.[494] Aerospace and defence measurement practice sits at the top of the calibration hierarchy, with instruments run against daily calibrations and every variation recorded, because an out-of-tolerance measurement on a satellite or radar is only discovered once the hardware is dead junk in orbit; cellular infrastructure tolerates more, handsets more again, and amateur equipment is calibrated only as far as the user needs.[533] European small satellite builders can use the ESTEC test facilities in Noordwijk, including a very large anechoic chamber, essentially free of charge, and a CubeSat-class satellite was characterised there to measure how its multiple antennas couple to and interfere with one another, a test few small companies could otherwise afford.[427]
Tooling reflects the same institutional split as the rest of the sector. Large aerospace firms sit on Mentor Graphics or Cadence rather than mid-market EDA packages, and the very largest, such as Boeing, run all of them at once because different divisions were acquired as separate companies and kept their own toolchains.[505] Product lifecycle management is a known weak point even at well-resourced hardware companies: SpaceX built its own system because nothing available on the market was adequate, and established enterprise firms with unlimited procurement budgets manage hardware data in spreadsheets because legacy products were built for process rather than for speed during new product introduction.[577]
Old and new space
Old space describes the 1960s to 1980s programme model of top-down design, waterfall development and layered chains of subcontractors beneath a prime, against which new space is defined by tolerating different part specifications and iterating much faster.[584] Commercial space before the 2010s meant a small number of large primes such as Boeing and Maxar building large satellites and launch vehicles, with slow design cycles and comparatively few spacecraft in orbit; the present density of on-orbit hardware is a recent change.[701]
The shift inverted the aerospace career ladder. In traditional large aerospace it can be a decade before a design reaches orbit, and a given seniority may be required before an engineer touches flight hardware at all, whereas at a fast-moving launch company an engineer in their early twenties can have a board of their own design deployed on satellites.[401] The compression is visible in company timelines as well: a satellite startup founded in 2021 put its first three spacecraft in orbit in 2024 on a rideshare, roughly two and a half years and under twenty million dollars from company formation to hardware on orbit.[728] Volume manufacturing has followed the same trajectory, with SpaceX building Starlink user terminals in what is described as the largest PCB assembly plant in the United States.[720]
Public programmes and cost structure
The Apollo moon programme cost roughly eighty billion dollars in inflation-adjusted terms, while NASA’s 1989 ninety-day study for a crewed Mars mission was costed at a minimum of four hundred and fifty billion dollars, an order-of-magnitude jump that ensured Congress never funded it.[137] That architecture became unaffordable through internal consensus, every department insisting its own idea be included until the scheme was so convoluted and expensive that Congress rejected it outright — design by internal stakeholder accommodation being a recognised failure mode for large programmes.[227] The Mars Direct architecture, developed by a small task group at Martin Marietta including Robert Zubrin as a reaction to that study, proposed a long-stay, directly-launched, repeatedly-flown campaign costing fifteen to twenty percent of NASA’s existing budget rather than a new multi-hundred-billion-dollar appropriation.[137]
Launch vehicle economics have been a recurring source of the same pattern. The Space Shuttle was only partially reusable, since the orbiter returned but the boosters and external tank were effectively expended each flight, so the per-launch cost stayed enormous despite the reusable framing.[201] The Space Launch System is fully expendable at a cost on the order of four and a half billion dollars per flight, difficult to justify commercially against a fully reusable competitor, and its architecture is essentially a mid-1970s launch system built decades later.[647] An internal budget audit of the Artemis programme found it so far behind schedule and over budget that further allocations to Boeing and the other contractors were withheld, with the second vehicle the specific trigger.[647]
The structural cause sits upstream of the agency. NASA’s programme structure is set by Congress, which both selects the mission and attaches the constraint that contractors in every funding state must participate, and the resulting geographically distributed contractor architecture is a principal cause of the schedule and cost overruns later attributed to the agency itself.[647] Contrary to common perception, the large majority of NASA’s project portfolio is aeronautical research such as jet engines, wing configurations and aircraft design, even though the majority of the money goes to heavy-lift launch programmes.[647]
Amateur and improvised practice
The AMSAT Fox series began as 1U CubeSats providing a standard reusable bus of battery system, RF chain and computing, into which university groups could drop experimental electronics, the bus supplying the data downlink while the experimenters supplied only the payload.[401] Most amateur radio satellites operate as a bent pipe, retransmitting a signal in real time on a second frequency with no onboard processing, the practical effect being that a handheld transceiver is given an antenna a couple of hundred miles high and its range extends from a town to a large fraction of a continent.[401]
Near-space work admits improvisation that orbital work does not. A high-altitude balloon payload was protected against condensation by coating the assembled boards completely in clear nail polish as an improvised conformal coating, on the reasoning that the boards would be chilled hard at altitude and then descend into humid summer air where water would condense directly onto cold copper.[401] The coating worked as intended: the payload fell through a thunderstorm and was recovered with visible globs of water sitting on the microcontrollers, and the electronics were still running.[401] What aerospace recruiters looked for in a student flight project was not the result but the reasoning — conformal coating the boards, designing for the environment, and being explicit about which requirements were mandatory and which were merely desirable.[401]
Amateur work at the launch vehicle end carries a different hazard. Publishing even the existence of amateur rocket work attracts proliferation interest, and a New Zealand builder who put up a website merely demonstrating that a cruise missile could be built by an individual received offers of large sums from correspondents in Iran and Pakistan for the plans, which he referred to his national security service.[538]
Historical antecedents
Entry into the early United States astronaut corps required an engineering degree: Neil Armstrong and the whole of the original Mercury 7 and the Apollo astronauts were engineers, and many were also test pilots.[110] The materials and reliability research produced alongside those programmes remains a primary reference source for terrestrial electronics as well as spaceflight, NASA having published extensively on lead-free soldering, being one of the world’s leading investigators of the tin whisker problem, and doing substantial work on component reliability and radiation hardening.[25] Much of the primary record has nonetheless been lost to media attrition, the original Apollo 11 moon landing tapes and the Pioneer telemetry tapes both being gone, and even recovered media presents reading it back as a hard engineering problem in itself, so keeping multiple physically separated copies is the only defence.[132]
Ground station design produced techniques still in use. The Holmdel horn antenna, a parabolic reflector integrated into a horn, was built for the passive Echo satellite experiment and paired with a liquid-helium-cooled maser as its low noise amplifier, an early example of cryogenic front-end design for space communications.[328] Its advantage over a conventional dish is very low side lobes, because an ordinary parabolic antenna scatters energy off the edge of the dish and therefore sees the ground, picking up the noise temperature of the Earth, which is far warmer than space and ruins very low noise measurements.[328]
Spacecraft engineering has also produced solutions with no terrestrial analogue, such as the original NASA spacesuits, which carried no closed refrigeration cycle and instead cooled by venting compressed refrigerant into the vacuum of space, using the expansion into the low-pressure environment to chill a reservoir while a circulator distributed the cooling.[618] Other technologies suit space and remain marginal on the ground: concentrator photovoltaic cells are designed to run at very high temperatures under focusing lenses and become more efficient as they get hotter, which suits satellites and space probes but leaves them a niche technology on Earth, where they will not displace conventional rooftop panels.[205] Microgravity aboard the International Space Station has been proposed as a manufacturing environment for ultra-precise parts, because without gravity a molten mass can form a perfectly round object, the practical obstacle being that the finished part still has to be brought back down.[283] The chip-scale atomic clock developed under DARPA funding followed the reverse path, finding its market wherever a stratum-three frequency reference is needed without the volume pricing of a conventional standard, with a unit flown as an experiment on the space station.[459]
Long-dormant hardware occasionally demonstrates how durable the underlying electronics can be: a satellite launched in 1967 that failed to reach its intended orbit and lost contact began transmitting again forty-six years later, when its uncontrolled tumble happened to bring the solar panels into enough sunlight to power the transmitter and simultaneously aligned the antenna toward Earth.[322]