Space Exploration Technologies Corp. received exactly two U.S. patents on August 4, 2026. For a company of its filing volume that is a quiet day, but the two records are unusually legible together: each takes a cost constraint rather than a performance ceiling as its subject, and each solves it by exploiting something the constellation already provides.

US12701567B2, titled “Timing half-duplex transmissions,” names Peter J. Worters, Martin S. McCormick and Andrew G. Whitlow as inventors and is classified under H04W 72/121, H04B 7/18513, H04L 5/16 and the H04W 56 timing-synchronisation range. A half-duplex terminal cannot transmit and receive at the same moment. That makes it cheaper and simpler than a full-duplex terminal — one radio chain instead of two, no high-isolation duplexer — and it is the sort of economy that matters when the terminal count runs to millions. The cost is scheduling difficulty, and it gets worse as the satellite's footprint grows, because terminals near the edge of a beam are meaningfully farther from the spacecraft than terminals beneath it.

determining, at a satellite, a maximum round-trip propagation delay associated with one or more user terminals of a plurality of user terminals, determining, based on the maximum round-trip propagation delay, a respective downlink to uplink (DL/UL) offset for each user terminal, wherein the DL/UL offset comprises a time difference between a first downlink frame at the satellite and a first corresponding uplink frame at the satellite— U.S. Patent No. 12,701,567, source

The mechanism is to size the gap between downlink and uplink around the worst case. The satellite determines the maximum round-trip propagation delay across its terminals, derives a downlink-to-uplink offset from it, and transmits that offset inside the downlink frame itself, so terminals learn the schedule from the same signal that carries their data. The result the claims recite is the point of the whole exercise: a terminal at the maximum delay and a terminal at a shorter delay both land in the same uplink frame at the satellite. The nearer terminal simply starts transmitting later. Frame occupancy stays full rather than being padded out to accommodate the farthest user.

Two details of the claim set are worth recording because the abstract does not carry them. Every independent claim requires the user terminals to be half-duplex; the abstract says only “user terminals.” And the claims compute the offset from the maximum delay and one or more additional delays — elsewhere identified as terminal processing time and the transmit-to-receive and receive-to-transmit mode transition times — which the abstract also omits. Those transition times are precisely the overhead a half-duplex radio incurs and a full-duplex one does not, so their presence in the claims is not incidental. The patent carries three independent claims covering the same limitations as a method, as a non-transitory storage medium and as a system. One small artefact survives into the issued text: claim 3 reads “where Nis an integer,” missing a space.

The second grant: using the stack as the fixture

US12698104B1, “Spacecraft antenna restraint system,” names Garrett P. Simard, Stefano A. Bauk, Robert D. Giglio, Andrew C. Thompson and Joshua A. Dunford, and is classified under B64G 1/66 with H01Q 3/08 and H01Q 1/288. It describes a first spacecraft with an antenna movably mounted to its chassis, a second spacecraft adjacent to it in a stacked launch configuration, and a passive restraint system defined between the two that substantially prevents the antenna from moving while in the launch state.

The word doing the work is passive. An antenna that must articulate on orbit has to be immobilised through ascent, and the conventional answer is dedicated hardware — latches, pyrotechnic or burn-wire releases, hold-down brackets — every piece of which is mass that flies to orbit and never does anything again, plus a release event that can fail. Here the restraint is a geometric relationship between two satellites that are already stacked against each other. Separation is the release. In a launch architecture that flies many identical spacecraft per mission, deleting per-satellite hold-down hardware and its actuation is a recurring saving multiplied by the flight rate.

Reading the pair

The two records sit at opposite ends of the same system, and both are about the economics of doing something many times. One reduces the bill of materials on the ground by making the cheaper class of terminal schedulable. The other reduces mass and mechanism count on each spacecraft by borrowing structure from its neighbour. Neither is a capability the constellation lacked; both are ways of paying less for capability it has.

What the records do not disclose should be stated plainly. Neither patent names a product, a constellation, a satellite generation, a terminal model or a launch vehicle. Neither gives unit costs, mass figures, terminal counts, production volumes or deployment dates. The commercial framing above follows from what the claims recite — half-duplex terminals, a stacked launch configuration — and not from any disclosure about programmes or economics, of which there is none. These are granted patents rather than applications, so the claims have been examined and allowed; but a granted claim describes a right to exclude, not a shipped design.

The classification split between the two grants is itself informative. The timing record sits entirely in the H04 communications ranges, sharing space with terrestrial cellular scheduling art; the restraint record sits in B64G 1/66, spacecraft-adaptations territory, with H01Q antenna codes attached. A constellation operator generates patentable subject matter in both fields at once, and a portfolio search confined to either range on its own would return half of this issue day.

One further caution on timing. Issue dates lag filing by years, and nothing in either document indicates when the described approach entered service, if it has. The August 4 issue day is a record of what was granted, not of what is new.