CubeSats and small satellites

The CubeSat is not a technology. It is an interface standard, and that distinction explains almost everything about its effect on the industry.

What the standard actually specifies

Proposed in 1999 by Jordi Puig-Suari at Cal Poly and Bob Twiggs at Stanford, the CubeSat specification defines a satellite that fits a standardised dispenser. A single unit, one U, is a 10 cm cube, formally 100 × 100 × 113.5 mm, with a mass limit that began at 1 kg, moved to 1.33 kg and is now 2 kg per unit in the current design specification.

Crucially, the standard defines the dispenser interface: rail dimensions, protrusion limits, the deployment switch arrangement, and the inhibits that keep the satellite inert until it has left the dispenser. The original Poly Picosatellite Orbital Deployer took three units.

What the standard fixes

External envelope, rail geometry, mass per unit, deployment switches, and safety inhibits. Everything the launch provider needs to know.

What it deliberately does not fix

Anything inside. Bus architecture, power, avionics, propulsion, payload. The internals are entirely the builder problem.

Why that split matters

A launch provider can certify a dispenser once and then accept any conforming payload without reassessing it. That is what turned a bespoke negotiation into a catalogue purchase.

The real innovation

Not miniaturisation. Standardising the interface between a satellite and its ride so that the two can be developed independently.

Before this, flying a secondary payload meant a bespoke mechanical and safety negotiation with a specific launch provider for a specific launch, taking years and costing more than most of the hardware. Afterwards it meant building to a published envelope and buying a slot.

Size classes

The CubeSat sits inside a broader taxonomy that is loosely used and worth pinning down.

Standard CubeSat form factors have grown steadily as the concept proved itself:

The trend from 1U toward 6U and 12U reflects a real lesson. A 1U CubeSat is severely constrained in power, and power is what limits payloads: at roughly 2 W of orbit-average generation from body-mounted cells, a 1U satellite cannot run a meaningful instrument and a radio at the same time. A 6U with deployable arrays reaches 20 to 50 W, which is enough for an imaging payload and an X-band downlink.

What commercial parts buy

The economic case rests on using terrestrial commercial components where traditional spacecraft use radiation-hardened ones.

The trade is genuine and should be stated honestly. Commercial parts are not radiation-hardened, have no lot traceability, and change without notice. Reliability is bought back architecturally instead of in silicon: watchdogs, memory scrubbing, current-limited supplies that survive latch-up by power cycling, and acceptance of a five-year rather than fifteen-year life. In low Earth orbit below the belts, where the annual dose is a fraction of a krad, that trade closes. See Space radiation.

The same approach fails outside that envelope. A CubeSat in MEO or in interplanetary space sees a radiation environment that commercial parts do not survive, which is why deep-space CubeSats such as MarCO and CAPSTONE use more conservative electronics and cost far more than their LEO equivalents.

The early-failure problem

The CubeSat community’s honest reckoning with its own reliability is one of the more instructive episodes in modern spaceflight.

Analyses of the early population, most systematically by Michael Swartwout, found that a large fraction of CubeSats launched between 2003 and roughly 2014 never returned usable data. A substantial share never transmitted at all. The failures were concentrated in university-built spacecraft and clustered in the first hours and days after deployment.

Every branch in that diagram represents a real and common failure. The pattern is that CubeSats failed at commissioning, not during the mission, and failed on basic functions rather than on their science payloads.

The causes were consistent across programmes:

The population has improved markedly, and the reasons are unglamorous: commercial bus vendors selling flight-proven subsystems, better testing discipline, and organisations flying many satellites and learning across them. Planet’s iterative approach, deploying successive Dove generations and treating early losses as information, is the clearest example of engineering reliability through volume rather than through analysis.

What the form factor is good at

Constellations of many cheap sensors. Planet operates the largest Earth observation constellation ever flown, built on 3U satellites, achieving daily global coverage at medium resolution. No conventional programme could afford the satellite count. See Earth observation.

Technology demonstration. Flying a new component to raise its technology readiness level is exactly the kind of low-consequence, schedule-driven mission the form factor suits, and it feeds the maturation pipeline described in systems engineering.

Distributed measurement. Some measurements require simultaneous observation from many points, which is impossible with one large satellite regardless of budget. Multipoint magnetospheric and radio occultation measurements are genuinely enabled rather than merely made cheaper.

Access. Universities, small countries and small companies fly missions that would otherwise be impossible. That is a change in who participates, not just in cost.

What it is not good at

The honest limits matter as much as the capabilities, and they are all consequences of physics rather than of engineering effort.

Aperture

Resolution is diffraction-limited by aperture diameter. A 10 cm optic cannot resolve what a 1 m optic resolves, at any budget. This is a hard physical bound, not an engineering shortfall.

Power

Body-mounted cells on a 3U give a few watts. High-power radar, high-rate downlinks and electric propulsion all need tens to hundreds of watts, which needs deployable arrays and therefore volume.

Pointing stability

Small moments of inertia mean small disturbances produce large angular accelerations. Arcsecond stability in a CubeSat is achievable but is a research problem rather than a purchase.

Lifetime

Commercial parts and minimal shielding suit five years in low orbit, not fifteen in GEO.

Thermal

A small body has little thermal mass and a high surface-to-volume ratio, so internal temperatures swing widely over each orbit.

The aperture limit is the one most often wished away. Angular resolution is $1.22\lambda/D$ regardless of how clever the satellite is, so a 10 cm telescope at 500 km altitude cannot resolve better than about 3 m in visible light, and no amount of processing recovers information the optics never collected.

Deep space

CubeSats have left Earth orbit, and the results are encouraging with qualifications.

MarCO, two 6U spacecraft, flew to Mars alongside InSight in 2018 and relayed its entry, descent and landing telemetry in near real time. They demonstrated that the form factor could do interplanetary cruise, communication and navigation, and both exceeded their objectives.

CAPSTONE, a 12U, reached a near-rectilinear halo orbit around the Moon in 2022, the orbit family selected for Gateway. See Lagrange points.

LICIACube imaged the DART impact at Dimorphos.

The qualification is cost. These are not $100 000 satellites. A deep-space CubeSat needs radiation-tolerant electronics, deep-space-capable communications, propulsion and navigation, and the resulting cost is tens of millions of dollars. The form factor carries over; the commercial-parts economics do not.

References

• CubeSat Design Specification, Cal Poly San Luis Obispo, current revision. The primary standard.
• Swartwout, M. “The First One Hundred CubeSats: A Statistical Look”, Journal of Small Satellites, 2:213, 2013, and his subsequent annual surveys of the CubeSat population.
• Klesh, A. T. and Krajewski, J. “MarCO: CubeSats to Mars in 2016”, SSC15-III-3, Small Satellite Conference, 2015.
• Cheetham, B. “Cislunar Autonomous Positioning System Technology Operations and Navigation Experiment (CAPSTONE)”, ASCEND 2021.
• Boshuizen, C. et al. “Results from the Planet Labs Flock Constellation”, SSC14-I-1, Small Satellite Conference, 2014.