Deep Space Communications

Deep space communications is the engineering of a data link across interplanetary distance: the budget, the aperture, the wavelength, the detector, the code and the protocol. The framing under test on this page is that deep space communications scale with demand.

Established That framing fails for an unusually clean reason: the physics scales the wrong way and the supply scales slowly. Received power falls as the inverse square of range, so a mission twice as far away needs four times the aperture-power product for the same rate. The two things that could keep up — antenna hours and flight-qualified optical terminals — are both production rates rather than inventions. This is the one page in this cluster where the binding constraint is genuinely physical and genuinely industrial at the same time.

Established The seam with the neighbouring brief is drawn before anything else, because the two pages used to overlap badly. Deep Space Infrastructure owns communications capacity — how much exists, who is competing for it, and what happens when it runs out. This page owns the link itself — optical link budgets, photon-counting detector arrays, coding, delay-tolerant networking, and relay architectures as engineering. The previous version of this page named Deep Space Network capacity as its own bottleneck; that clause moves out, and the terminal-production bottleneck moves in, because that one is genuinely this page's.

Frontier The single most useful correction here is to read the whole demonstration curve rather than its best point. DSOC returned 267 Mbit/s at 31 million km, 25 Mbit/s at 226 million km, and more than 6.25 Mbit/s out to 2.7 AU — roughly a factor of 44 in rate across roughly a factor of 13 in distance, close to what the inverse square law predicts once ground conditions and pointing are allowed for. A brief quoting only the peak has reported the best point on a curve as if it were a capability.

Established And the demonstration is finished. DSOC flew as a technology demonstration attached to Psyche, launched 13 October 2023, and ran through 2025 with extended-mission opportunities; the neighbouring brief records its final pass in September 2025. Every sentence about DSOC on this page is in the past tense, deliberately, because an earlier version of a page on this site implied the demonstration was still running and that error is not being reintroduced.

Established Start with the equation, because everything else in this subject is a consequence of it. Received power in decibel-milliwatts equals transmitted power plus gains minus losses, and the dominant loss term for deep space is free-space path loss, which in standard form is 32.45 dB plus 20 times the base-ten logarithm of frequency in megahertz plus 20 times the base-ten logarithm of distance in kilometres.

Established Two consequences follow directly and both are counter-intuitive the first time. First, distance enters as 20 log(d), so every doubling of range costs 6 dB — a factor of four in received power. Going from Mars to Jupiter is not “further”; it is roughly an order of magnitude of link budget. Established Second, frequency also enters with a plus sign, so path loss rises with frequency — and yet every generation of deep space communications has moved higher, from S-band to X-band to Ka-band to optical. The reason is that antenna gain rises as the square of aperture divided by wavelength, and for a dish of fixed physical size the gain term more than repays the loss term. That single trade is the whole explanation of optical communications to a reader who has not met a link budget.

Established The reference case for how brutal the distance term is remains Voyager. Voyager 2 was 143.05 AU — 21.4 billion km — from Earth as of February 2026 and returns science data at approximately 160 bits per second. Its radioisotope generators delivered 470 W at launch and decay on an 87.7-year characteristic. The standard link-budget reference gives Voyager as the highest known path loss case, 308 dB as of 2002, requiring high transmit power and high-gain antennas at both ends.

Frontier The comparison worth making explicitly, with its arithmetic shown. DSOC returned 267 Mbit/s at 31 million km. Voyager 2 returns 160 bit/s at 21.4 billion km — about 690 times further. On the inverse-square term alone that is a factor of roughly 480,000 in received power, and the observed rate ratio is about 1.7 million. Both inputs are sourced and the comparison is illustrative arithmetic, not a measured scaling law — but it is the most vivid way to show that this subject is dominated by one exponent.

Established Aperture is the lever, and apertures are civil engineering. Gain scales with collecting area, so a 70-metre dish has roughly four times the area of a 34-metre one. That is why the Deep Space Network's 70-metre antennas exist and why they are irreplaceable in the near term — and it is why the aperture side of the link budget is a construction schedule rather than a design choice.

Established Now the demonstration, in order, because the order is the finding. DSOC's flight hardware was a 22 cm aperture telescope with a spacecraft terminal mass under 29 kg, downlinking at 1550 nm with a 4 W transmitter and receiving an uplink at 1064 nm. The ground segment was the uplink transmitter at JPL's Table Mountain Facility and the downlink receiver at the 5-metre Hale Telescope at Palomar Observatory. Detection at both ends used high-efficiency photon-counting detector arrays, which makes this as much a detector-physics achievement as a laser one.

Established The results with their distances and dates. First light 14 November 2023. On 11 December 2023, 267 Mbit/s at 31 million km, with 62.5 and 100 Mbit/s also demonstrated; 1.3 terabits were downloaded across the test period. On 8 April 2024, 25 Mbit/s at 226 million km — 1.5 AU — against a project goal of 1 Mbit/s at that range. In July 2024 the link was maintained at 467 million km. In early 2025 the downlink exceeded 6.25 Mbit/s out to 2.7 AU, with an uplink of 1.8 kbit/s across 0.2 to 3.3 AU. Established The design expectation at 0.4 AU had been 292 kbit/s uplink and up to 100 Mbit/s downlink; the demonstration beat its own 1-Mbit/s-at-1.5-AU goal by a factor of twenty-five.

Frontier Read the shape rather than the peak and the conclusion inverts without diminishing the achievement. 267 Mbit/s at 0.2 AU becomes 25 Mbit/s at 1.5 AU becomes a little over 6 Mbit/s at 2.7 AU: a factor of 44 across a factor of 13 in distance. DSOC did not repeal the link budget. It moved the whole curve up by one to two orders of magnitude and left its slope intact. Frontier Interest to mark: NASA describes the technology as improving communications performance and efficiency by ten to a hundred times over conventional means. That is the agency on its own demonstration; the rate-and-distance table above is the nearest thing to an independent check, and it broadly supports the claim at the distances tested.

Established What optical buys, stated once and precisely. Two to three orders of magnitude more gain for the same physical aperture, because gain scales as the square of aperture over wavelength and the wavelength drops from centimetres to microns. That is the entire reason a 22 cm telescope with a 4 W laser can compete with a multi-metre dish and tens of watts.

Frontier What optical costs, and it is four things rather than one. Pointing: beam divergence scales with wavelength over aperture, so an optical beam is arcseconds wide where a radio beam is degrees — the spacecraft must point at where Earth will be when the photons arrive, and hold it, from a platform with reaction wheels and thermal drift. Weather: radio at Ka-band degrades in rain; optical is stopped by cloud, so the ground segment has to be sited and duplicated for cloud statistics rather than only for longitude coverage — fundamentally different station economics from a three-complex geometry. Solar angle: a receiver looking at a spacecraft near the Sun is looking at the sky next to the Sun. Photon starvation: at 6 Mbit/s over 2.7 AU the receiver is counting individual photons, so the system is a photon-efficiency problem rather than a classical signal-to-noise problem, which is why the detector arrays matter as much as the laser.

Established And radio does not go away, for a reason with a name. A 160 bit/s X-band link from 143 AU works through cloud, needs no fine pointing, and has been running for forty-nine years. Optical is a throughput technology for the inner and middle solar system; radio remains the technology of extreme range and of command. Frontier This page therefore refuses the “upgrade from radio to light” framing the previous version used and replaces it with a two-regime picture.

Established Delay-tolerant networking is the other half of the engineering answer and it is badly under-covered. The internet protocol suite assumes a seamless end-to-end path, round-trip times in milliseconds and continuous connectivity. Interplanetary links have none of those: one-way light time to Mars is minutes and to Pluto hours, links are intermittent by orbital geometry, and communication suspends entirely during solar conjunction, when the Sun sits between the endpoints. Congestion control, retransmission timers and the three-way handshake all rest on assumptions that fail simultaneously.

Established The answer is store-and-forward with custody transfer. The Bundle Protocol operates as an overlay: data is packaged into bundles which are stored at each hop until a transmission window opens, and custody of the bundle transfers hop by hop, so responsibility for eventual delivery is held by an intermediate node rather than by the origin. The bundle layer supplies custody transfer, segmentation and reassembly, and an end-to-end reliability that the internet protocol cannot provide across disconnected regions.

Established The standards lineage is longer than most readers expect. RFC 4838 and RFC 5050 (2007) gave the architecture and Bundle Protocol version 6. RFC 9171 with companions 9172, 9173 and 9174 (January 2022) gave version 7. RFC 9713 (January 2025) updates 9171. The Consultative Committee for Space Data Systems maintains a profile of the earlier protocol for space mission use. Frontier None of those documents was opened for this rewrite — the standards hosts were unreachable — so they are cited by number, which is how standards are cited anyway, and the exact publication months should be checked before they are quoted.

Established And the deployment record, which is real rather than aspirational. The Bundle Protocol was tested on the UK-DMC satellite in 2008. DINET, the Deep Impact Networking experiment, flew on the Deep Impact spacecraft in October 2008 — the deep-space demonstration, and the one the previous version of this page never named. Delay-tolerant networking was deployed to an ISS payload in May 2009; in October 2012 an ISS commander remotely operated a robot in Germany over it; and it is used in ISS operations today, supporting both protocol versions. Established Two implementations exist: ION, written in C by NASA and JPL under spaceflight software constraints, and HDTN, a performance-oriented C++ implementation running on ordinary operating systems. One flight-constrained and one throughput-oriented implementation is the normal shape of a maturing protocol stack.

Frontier The honest status is a split verdict. Delay-tolerant networking is settled as a specification and as an ISS operational technology. It is a frontier claim as an interplanetary network, because there is not yet a network — there are point-to-point links and one deep-space experiment from 2008. The thing that would make it a network is relay infrastructure, and relay infrastructure belongs to the neighbouring brief.

Established The ground segment is named here only where it is a link fact, and one case is the best in the subject. DSS-43 at Canberra is the only antenna on Earth that can send commands to Voyager 2. Its 2020 refurbishment took the link down for eight months; communications resumed on 2 November 2020 and full communications were restored on 12 February 2021. Frontier The reason DSS-43 is unique is a link budget: Voyager 2's southern declination, combined with 160 bit/s at 143 AU, means only one southern-hemisphere 70-metre aperture closes the uplink. That is the physics of a link expressing itself as an institutional fragility, which is exactly the register this page works in.

Established Two further ground facts are cited once each and handed across. DSS-14 at Goldstone became operational in 1966 as a 64-metre dish and was upgraded to 70 metres in 1988 specifically to support Voyager; the 70-metre antennas are ageing, harder to maintain than modern beam-waveguide antennas, and since 2021 are being completely refurbished rather than replaced by arrays, which takes them offline for months at a time. DSS-23 at Goldstone has an expected operational date of 2026. Frontier What those dates mean for capacity — the construction rate, the contention, the audit finding that the network is oversubscribed — is Deep Space Infrastructure's argument, and this page cites and points rather than re-deriving it.

Established The optical demonstrations beyond DSOC are the evidence for the supply-side constraint, and they were not marginal. The Lunar Laser Communication Demonstration returned 622 Mbit/s down and 20 Mbit/s up, error-free, across 385,000 km in October 2013. The Laser Communications Relay Demonstration launched 7 December 2021 into geostationary orbit. TBIRD on PTD-3, launched 25 May 2022, reached 100 Gbit/s by December 2022 and 200 Gbit/s on 28 April 2023 from low Earth orbit. And the European Data Relay System, first deployed in November 2014, carries inter-satellite links at up to 7.2 Gbit/s, with more than a million minutes of communications and over fifty thousand links by May 2023.

Frontier The European system is the entry worth dwelling on, because it is the only operational service in that list rather than a demonstration. Optical relay at operational scale exists — in near-Earth space, run by Europe, sold as a service. That is a far stronger statement of maturity than “the physics was proved in deep space,” and it sets the right baseline for asking why deep space has no equivalent.

Frontier The supply side is where the evidence runs out, in both directions, and this page says so. Named suppliers of space optical terminals include Tesat-Spacecom, Mynaric, Honeywell, Ball and Sony. No production-capacity figure for any of them was obtainable for this rewrite — which is a negative result worth recording, because it means the public evidence for terminals at fleet scale is thin in both directions rather than adverse.

Established The one hard data point on the supply side is a consolidation. Rocket Lab acquired Mynaric in April 2026, with approval from the German federal economics ministry conditioned on intellectual property, production and research remaining in Germany and Europe, on guaranteed access for German and European customers, and on existing export restrictions to China being maintained; the headquarters stayed in Munich. Frontier That acquisition has two live readings and this page cannot adjudicate between them: that optical terminals are becoming a strategic component to be vertically integrated, or that a standalone terminal business could not sustain itself. No financial or delivery data was obtainable here, so the second reading is carried as a hypothesis and not asserted.

Established The demonstration is over, and its closing report is the agency’s own. NASA records the deep-space optical experiment as having exceeded its project expectations, which is consistent with the rate-and-distance table above — a 1-Mbit/s-at-1.5-AU goal beaten by a factor of twenty-five is an exceeded expectation by any reading. What a completed demonstration does not settle is the thing an operational service is built from, and the distinction is worth stating in one line: the experiment proved that the link closes, once, between one terminal and one borrowed telescope, in weather that happened to cooperate. Frontier The most consequential ground-segment result of the demonstration was therefore not a data rate but a retrofit. A 34-metre radio antenna at Goldstone, fitted with mirror segments on its existing dish surface, was reported to have received the same optical downlink — a hybrid radio-and-optical aperture rather than a purpose-built observatory. If that generalises, the capital question for an optical ground segment changes from “how many new observatories” to “how many existing dishes can be re-surfaced”, and the second question has a far better answer, because the sites already have power, fibre, staff, scheduling and a security perimeter. It is one experiment, reported by the organisation that ran it, and the figures that would establish it — collecting efficiency against a clean aperture, and what the retrofit does to the antenna’s radio performance — were not obtained here.

Established The open questions divide into the link, the network and the supply chain, and only the first has a demonstrated curve to argue about.

Frontier Question one: how far does the ten-to-a-hundred-times improvement claim generalise? NASA states it for optical over conventional means; the fetched rate-and-distance curve broadly supports it at the distances tested, from 0.2 to 2.7 AU. It is not established as a general figure, and the interested party is the agency that flew the demonstration.

Frontier Question two: does optical become the primary deep-space downlink with radio as backup? Near-Earth evidence is strong — 200 Gbit/s from low orbit, an operational European relay service with over fifty thousand links. Deep space has exactly one completed demonstration. Established The counter-position is stronger than it looks: radio remains irreplaceable at extreme range and for command, which is a description of current practice rather than a prediction, and Voyager plus DSS-43 is the whole argument.

Frontier Question three: is delay-tolerant networking the correct architecture for an interplanetary network? Cerf and the delay-tolerant networking research group, the internet standards process, the space data standards committee and NASA's own implementation all say yes. It is settled as a specification and as ISS practice and unsettled as an interplanetary network, because no interplanetary network exists to test it.

Frontier Question four, and it is the adjudicated constraint: can the existing supplier base produce optical terminals at fleet scale? Five named suppliers, one recent acquisition, and no production-rate evidence in either direction obtainable for this rewrite. Frontier The absence of evidence is the finding here, and anyone arguing either side should be asked for a delivery number.

Frontier Question five: will ground-station cloud statistics rather than spacecraft hardware turn out to be the binding constraint on an optical deep-space service? The physics is clear — cloud stops an optical link outright, so availability requires cloud-decorrelated sites — and no availability figure or site-diversity study was obtained here. The argument is sound and unquantified.

Frontier Question six: is photon-efficiency coding as enabling as the laser? At the far end of the demonstrated curve the receiver counts photons, and the figure of merit becomes bits per received photon rather than signal-to-noise ratio. The coding families that maximise it — serially concatenated pulse-position modulation and relatives — are a distinct discipline from the turbo and low-density parity-check codes that dominate radio links. Frontier The claim that photon-counting arrays are the enabling component is sourced; specific code rates and bits-per-photon figures are not, and none appear on this page.

Speculative Question seven, kept short deliberately: interstellar links. Signalling across light-years demands enormous aperture-power products and imposes years to millennia of latency, and is feasible in principle only at very small data rates. No interstellar link budget was obtained for this rewrite and no data rate is asserted. Speculative The one adjacent artefact worth a sentence is the solar gravitational lens, whose focal region begins around 542 to 550 AU with light amplification up to about a hundred billion; it is an imaging concept, and the suggestion that the same focal physics could amplify a link is an extrapolation from an imaging proposal. Getting a spacecraft there belongs to Interstellar Probes and the instrument case to Mega-Telescopes.

Handwave And the framing itself. “Deep space communications scale with demand” requires that supply grow at least as fast as the manifest. The physics scales as the inverse square, the apertures scale on a construction schedule, and the terminals have no demonstrated production rate at all. The claim does its work at the word “scale,” which is doing duty for three different rates, none of them measured against the manifest.

Frontier Question eight: link availability is a joint probability, and no joint statistic for a candidate network has been published. The single-site version of the question — how often is the sky clear at Palomar — is the wrong one. Availability for a service is the probability that at least one station in the network has a cloud-free line of sight, acceptable turbulence and daytime sky background, and a free receiver, at the hour the spacecraft is transmitting. That product is dominated by the correlation between sites, not by the goodness of any one of them: two excellent sites inside the same synoptic weather regime are close to one site, and two mediocre sites on different continents may be worth more than either. What would settle it is a published joint cloud-free-line-of-sight statistic for a named three- or four-station network, computed from satellite cloud climatology, set against a mission’s required availability. The method for that calculation is standard in the optical ground-station literature; the answer for a deep-space network is not on the public record, which is why this page argues the shape of the constraint and quotes no availability number.

Frontier Question nine: acquisition, which is a different problem from pointing and fails earlier. Pointing is holding a beam once you have the target; acquisition is finding it. The demonstrated system acquired against an uplink beacon transmitted from the ground, and the beacon obeys the same inverse-square law as everything else — so as range grows the beacon, not the downlink, becomes the marginal link, and beyond some distance the terminal must acquire open-loop from star-tracker attitude and ephemeris alone. The demonstrated uplink was 1.8 kbit/s across 0.2 to 3.3 AU, which is a link operating close to where beacon-referenced acquisition stops being comfortable. Frontier The number that would settle the question is an acquisition-time distribution against range — how long from window open to locked, at 1 AU, at 5 AU, and how often acquisition fails outright — and no such distribution has been located. Until it exists, “optical works to Jupiter” is an extrapolation of a downlink curve across a different link that has not been extrapolated with it.

Established The first bottleneck is the exponent and nothing removes it. Six decibels per doubling of range, a factor of four in received power. The best demonstration in the subject's history obeys it: 267 Mbit/s at 0.2 AU, 25 at 1.5, a little over 6 at 2.7. Every future mission that goes further pays the same rate Voyager pays.

Frontier The second is pointing, and it is a spacecraft-attitude problem masquerading as a communications problem. An optical beam is arcseconds wide. The transmitter must lead the target, pointing at where Earth will be when the photons arrive, and hold that lead from a platform with reaction wheels, thermal gradients and a science instrument doing its own slewing. Nothing about this is unsolved; everything about it is a per-spacecraft engineering cost that a radio link does not pay.

Frontier The third is cloud, which changes the shape of the ground segment. A three-complex network at 120-degree longitude spacing gives continuous coverage for radio. For optical, availability is a weather statistic, so the requirement becomes multiple cloud-decorrelated sites per longitude — more stations, sited by climatology rather than by geometry. No availability figure for an optical deep-space ground network was obtained for this rewrite, and the economics of it are unquantified here.

Established The fourth is photon starvation at the far end of the curve, which is a detector and coding problem. At a few megabits per second across 2.7 AU the receiver counts individual photons, and performance is set by detector efficiency, timing jitter and code choice rather than by transmitter power. The photon-counting arrays at both ends of DSOC were the enabling component, and they are the part of the system with the least public engineering literature.

Frontier The fifth is terminal production, and this is the bottleneck that genuinely belongs to this page. DSOC was one terminal. A deep-space optical network needs terminals on every outbound spacecraft, ground receivers at multiple cloud-decorrelated sites, and a production line that delivers on mission schedules. The physics is proven at the scale of one, and the industry is proven at the scale of one product line inside a larger company. Those are different maturities, and the first is routinely allowed to stand in for the second.

Established And the sixth is aperture, which this page names and hands across. A 70-metre dish has four times the area of a 34-metre one, the 70-metre antennas date from the 1960s and are being refurbished months at a time rather than replaced by arrays, and DSS-23 is expected in 2026. What that construction rate means for a growing manifest is Deep Space Infrastructure's argument, and the audit finding that the network is oversubscribed belongs there too.

Frontier The seventh is that the receiving apertures are somebody else’s scientific instruments. The demonstration’s downlink receiver was a 5-metre astronomical telescope whose time is allocated to astronomy by a committee, and an operational service cannot be scheduled against an observing programme. So the optical ground segment has exactly three capital routes and they are not equivalent: build dedicated optical stations, which is a new observatory per site; borrow astronomical apertures, which is what the demonstration did and does not scale past a demonstration; or re-surface existing deep-space radio antennas into hybrid apertures, which is the cheapest route, reuses sites that already exist, and rests on a single reported experiment. Frontier The seasonal term makes it worse in a way radio does not experience: an optical receiver looking at a spacecraft near the Sun is looking into a bright daytime sky, so availability has a solar-elongation cycle on top of a weather statistic, and the two are not independent of the mission’s own geometry.

Frontier The eighth is interoperability, and it is the quiet reason radio deep-space operations work at all. Cross-support — one agency’s ground station tracking another agency’s spacecraft during a critical event — is routine at radio frequencies because the space-data standards committee long ago fixed the modulation, coding, framing and ranging so that the terminals are compatible by construction. Optical has the standards but not the practice. The same committee has issued recommendations for optical links, including the high-photon-efficiency family that standardises the pulse-position modulation and concatenated coding this page names in its engineering row, so a common physical layer exists on paper. Frontier No cross-agency deep-space optical cross-support has been demonstrated, and until one is, an optical link is a bilateral arrangement between one spacecraft and one ground segment rather than a service — which matters most in exactly the case cross-support exists for, a spacecraft in trouble over the wrong hemisphere. The standards are cited by number here, as standards are, and the document numbers and issue dates should be checked before they are quoted.

Established This brief records no dependency on another brief. No physics result is missing. The link budget is closed-form, the optical demonstration is complete, the protocol is standardised and running on the ISS, and the detector arrays exist. What is missing is quantity: antenna hours and terminals.

Established The first standing requirement is deep space network capacity ahead of the mission manifest. This page states it as a requirement and does not argue it — the argument, with the forecast history and the contention record, is Deep Space Infrastructure's. What belongs here is the reason capacity is a link problem too: aperture is a term in the equation, and the only way to buy back the inverse-square loss on an existing spacecraft is to point a bigger dish at it.

Frontier The second is space-qualified optical terminals produced at fleet scale, and it is this page's own. Five named suppliers, one demonstration terminal flown to deep space, an operational European relay service in near-Earth space, and one acquisition of a pure-play manufacturer by a launch and spacecraft prime in April 2026. No production rate is published by anyone, which is why the requirement is stated as a rate and not as a technology.

Established The seam with the infrastructure brief is quoted rather than renegotiated, because that page drew it first. It owns the six over-capacity forecasts, the Artemis contention that degraded a flagship robotic mission, the decade-per-antenna construction rate, the commercial relay procurement and the depots. This page owns the link budget, the demonstration curve, the optical-versus-radio trade, delay-tolerant networking, DSS-43's uniqueness as a link-budget consequence, and terminal supply. The shared facts — DSS-23's 2026 date and the oversubscription audit — are cited once here and pointed across.

Frontier Two further hand-offs. Whether to transmit deliberately at interstellar range is a decision rather than an engineering question, and belongs with Interstellar Civilization Models; this page owns only the engineering statement that such a link is power- and latency-bound and feasible at tiny rates. Frontier And directed-energy hardware is genuinely shared: large phased apertures, precision pointing and atmospheric propagation are common to Beam-Powered Propulsion, and the shared heritage is noted rather than claimed.

Established The decisive experiment in this subject has run and is complete. DSOC flew on Psyche from 13 October 2023, achieved first light on 14 November 2023, and produced a rate-and-distance curve from 0.2 to 2.7 AU before finishing in 2025. Its result is a curve, and the curve is the finding. Established It is written in the past tense throughout this page, because the neighbouring brief records its final pass in September 2025 and explicitly corrects an earlier page that implied otherwise.

Established The near-Earth optical experiments are a ladder and they all succeeded. 622 Mbit/s down and 20 Mbit/s up, error-free, across the Earth–Moon distance in October 2013. A geostationary relay demonstration launched in December 2021. 200 Gbit/s from low orbit on 28 April 2023. Frontier And the one that is not an experiment at all: a European inter-satellite relay service running at up to 7.2 Gbit/s with over a million minutes of communications and more than fifty thousand links by May 2023. That is the maturity baseline deep space does not yet have.

Established The protocol experiments are older and more numerous than the optical ones. The Bundle Protocol on the UK-DMC satellite in 2008; DINET on the Deep Impact spacecraft in October 2008, which remains the deep-space demonstration of delay-tolerant networking; deployment to an ISS payload in May 2009; remote operation of a robot in Germany from the ISS in October 2012; and current ISS operational use supporting both protocol versions. Frontier What has never been run is the experiment that matters most: a multi-hop store-and-forward network across interplanetary distance with real relays. That requires relay spacecraft, and the relays are proposals.

Established Solar conjunction is a scheduled experiment that runs itself every synodic period. Communication suspends when the Sun sits between endpoints, and every Mars mission plans a blackout around it. Frontier It is also the cleanest argument that delay-tolerant networking is engineering rather than metaphor: the network must survive a scheduled, multi-week, astronomically caused outage that no terrestrial protocol contemplates.

Frontier The experiment nobody has run on the supply side is a production run. One deep-space terminal has flown. An operational near-Earth service exists. No supplier publishes a delivery rate, and the only recent structural datum is an acquisition. The test that would settle the constraint is a mission set that needs a dozen terminals on a schedule, and no such set has yet placed the order.

Established The flight terminal is now specified by something that flew. A 22 cm aperture telescope, under 29 kg of spacecraft terminal, a 4 W transmitter at 1550 nm, an uplink receiver at 1064 nm, and pointing good enough to lead the Earth across light-minutes of travel time. That is the reference design any deep-space optical terminal is measured against, and it returned 267 Mbit/s at 31 million km.

Established The ground terminal is the harder half and it is astronomical hardware. DSOC's downlink receiver was the 5-metre Hale Telescope at Palomar and its uplink transmitter was at Table Mountain. Photon-counting detector arrays at both ends are the enabling component, and the ground requirement that follows from cloud sensitivity is multiple sites chosen for decorrelated weather rather than for longitude alone.

Established The radio side's engineering requirement is aperture and it is measured in decades. Three complexes at roughly 120 degrees of longitude, each with one 70-metre antenna and three or more 34-metre beam-waveguide antennas. DSS-14 dates from 1966 and was enlarged to 70 metres in 1988 for Voyager; the 70-metre antennas are being completely refurbished rather than arrayed, taking them offline for months; DSS-23 is expected in 2026. This page states those as link facts and sends the capacity argument to Deep Space Infrastructure.

Established The protocol requirement is a stack, and it exists in two implementations for two purposes. ION, in C, constrained to spaceflight software restrictions. HDTN, in modern C++, performance-optimised for ordinary operating systems. Custody transfer, segmentation and reassembly, and end-to-end reliability across disconnected regions are the functions the bundle layer must supply, and the standards that define them are the 2007 architecture pair, the January 2022 version-7 family and its January 2025 update.

Frontier The coding requirement changes regime along the curve. Near the Sun the link is a classical signal-to-noise problem. At the far end it is a photon-efficiency problem, and the relevant figure of merit is bits per received photon. Serially concatenated pulse-position modulation and related families are the discipline; no code rate or bits-per-photon figure is quoted on this page, because no coding primary was opened for this rewrite.

Frontier And the requirement that no single mission can meet: a production line. Terminals on every outbound spacecraft, receivers at multiple weather-decorrelated ground sites, spares, and a delivery cadence that matches a launch manifest. The canonical engineering reference for the whole subject is JPL's deep space link design handbook, which this rewrite could not open and therefore names rather than cites.

Frontier The ground requirement, stated as a capital choice rather than a wish. An optical deep-space service needs receiving aperture at several sites whose weather is decorrelated, and the three ways to get it — new dedicated stations, borrowed astronomical telescopes, hybrid re-surfacing of existing radio antennas — differ by an order of magnitude in cost and by years in schedule. Only the third inherits a site, and the hybrid result is one experiment old. Frontier The interoperability requirement is a standards conformance rather than a construction: a terminal built to the published optical recommendations can in principle be received by any conforming station, which is the property that makes radio cross-support possible and which no deep-space optical pair has yet exercised across agencies. And the requirement nobody has costed is the scheduling layer — a network of weather-dependent receivers implies handover between stations mid-pass and a scheduler that can re-plan on a forecast, which is a different operational instrument from the fixed allocation grid a radio network runs on.

Established The tight seam is with the infrastructure brief and it is honoured in that page's own words. Deep Space Infrastructure owns communications capacity — how much exists, who is competing for it, and what happens when it runs out. This page owns the technology itself: optical link budgets, photon-counting detector arrays, coding, delay-tolerant networking, and relay architectures as engineering. Concretely, the six over-capacity forecasts, the Artemis contention that degraded a flagship robotic mission and the decade-per-antenna construction rate stay there; the link budget, the demonstration curve, the optical-versus-radio trade, the protocol stack and terminal supply stay here.

Established The audit finding on oversubscription is cited once and pointed across, deliberately. NASA's Office of Inspector General reported in 2023 that the network is oversubscribed, with mission impacts and scheduling challenges expected to increase as crewed lunar missions begin. That is a real finding from an auditor whose interest runs against its own agency, and it is the neighbouring brief's argument to make, which fetched the capacity literature directly. This page did not open the report and does not build on it.

Frontier The astronomical adjacency is concrete rather than thematic. Mega-Telescopes owns large apertures as instruments — and DSOC's downlink receiver was the 5-metre Hale Telescope at Palomar, which is an astronomical aperture doing communications work. If optical deep-space service scales, the ground segment competes with astronomy for exactly the same class of asset and the same class of site.

Frontier The directed-energy adjacency is shared hardware. Beam-Powered Propulsion needs large phased apertures, precision pointing and atmospheric propagation modelling — the same three problems an optical uplink has, at higher power. The heritage is genuinely shared; the propulsion application is not this page's.

Speculative And two long-range adjacencies kept deliberately short. The solar gravitational lens is shared property: Interstellar Probes owns getting to 550 AU, Mega-Telescopes owns it as an imaging instrument, and this page owns at most the observation that the same focal physics has been proposed as a link amplifier. Whether to transmit at all across interstellar distance is a decision for Interstellar Civilization Models, not an engineering question for this one.

Established The institutional structure of this subject is unusually healthy on the standards side and unusually thin on the supply side. The protocol lineage runs through the internet standards process and the space data standards committee, with two open implementations and a decade of ISS operational use. Nothing about the network architecture is blocked institutionally.

Established The demonstration institution works too. Optical links have been demonstrated from lunar distance in 2013, from geostationary orbit from 2021, at 200 Gbit/s from low orbit in 2023, and from deep space across 0.2 to 2.7 AU through 2025. That is a coherent programme executed across a decade by a single agency plus partners, and it delivered.

Frontier The supply institution is where the record thins to nothing. Five named terminal suppliers, no published production rates, and one acquisition in April 2026 in which a launch and spacecraft prime absorbed a pure-play manufacturer under ministry conditions on keeping production and research in Europe. Frontier Two readings of that acquisition are live and this page carries both, because no financial or delivery data was obtainable to choose between them.

Established The operational institution has a single point of failure with a name and a date. One antenna on Earth can command Voyager 2, and its 2020 refurbishment left the spacecraft uncommandable for eight months. That is not a management failure; it is a link budget with an institutional shadow — the aperture required to close the uplink at 143 AU exists in one southern-hemisphere location.

Frontier And the coordination institution barely exists for the thing that would matter next. An operational optical deep-space service needs ground sites chosen for cloud statistics, which means international siting, cross-agency scheduling and shared receivers. The European relay service shows the commercial model works in near-Earth space; nothing comparable is organised for deep space.

Established The cluster-level observation applies here too and is stated in full in Space Law and Governance. This brief carries constraints-only adjudications and no brief-to-brief edge. Deep space communications is not waiting on a discovery; it is waiting on antenna hours and a production line.

Frontier The distributive question here is access to a shared, scarce, publicly funded asset. Deep space links are allocated by scheduling, and a scheduling system is a distribution mechanism with no public process. The neighbouring brief documents a case where a flagship observatory's operations were degraded by contention, which is the concrete form this question takes.

Established The spectrum and optical-band question is real and quiet. Radio deep-space allocations are internationally coordinated and protected; optical bands are not allocated in the same way, and an optical ground station is a telescope looking at a sky that astronomers also use. The two communities share apertures, sites and sky, and DSOC used one of the world's historic research telescopes as its receiver.

Frontier Equity of access has a second dimension in the ground segment. A radio network needs three longitudes; an optical network needs many more sites chosen by climatology, which means more host countries and a wider set of participants — or a narrower one, if the sites with the best cloud statistics happen to be concentrated. Nothing obtained for this rewrite quantifies that, and the observation is offered as a structural point rather than a finding.

Speculative And the one genuinely irreversible ethical decision in the subject is not this page's to make. Deliberate transmission toward another star is an act with no recall and no consent process, and the engineering here — enormous aperture-power products, tiny data rates, years to millennia of latency — only describes what such an act would cost. Interstellar Civilization Models owns the decision.

Established A final point about this page's own claims. An earlier version described record data rates from tens of millions of kilometres and treated the demonstration as ongoing. The rates are now given as a curve with its endpoints, the demonstration is written in the past tense, and the Deep Space Network capacity clause has been removed from the bottleneck list because it belongs to the neighbouring brief. Correcting a page's own overstatement in public is part of the job.

Established The civilizational claim usually made here — that communications is what makes exploration usable — is true and is not what the evidence is about. What the evidence is about is that the link budget sets a floor no demonstration has raised, and that the two things which could raise the ceiling are a construction schedule and a production line.

Established The achievement deserves its own sentence before the qualification. A 22 cm telescope with a 4 W laser returned a quarter of a gigabit per second across 31 million km and beat its own goal at Mars distance by a factor of twenty-five. That is one of the cleanest engineering results in this entire category.

Frontier And the qualification is the slope. 267 Mbit/s at 0.2 AU, 25 at 1.5, a little over 6 at 2.7 — the same six decibels per doubling that has governed every link since the first one. Every mission that goes further pays what Voyager pays, which is why Voyager returns 160 bit/s from 143 AU and why one antenna in the southern hemisphere is the only thing on Earth that can tell it what to do.

Frontier The network question is the one with the largest gap between architecture and reality. Delay-tolerant networking is a mature specification with two implementations and years of ISS operations behind it, and its deep-space demonstration was in 2008. An interplanetary internet requires relays, and the relays are proposals. That is a shorter gap than most in this category and it is still a gap.

Established The sentence to leave a reader with. The physics sets a floor that no demonstration has raised, and the two things that could raise the ceiling are a construction schedule and a production line, neither of which has published a rate. The demand side is not in doubt — which is exactly why the capacity argument next door and the link argument here are the same problem seen from two ends.

These horizons track three separate rates — the physics, the aperture schedule and the terminal supply — because the framing under test assumes they move together and they do not:

• 10 yr: Established The link budget does not change; expect every rate quoted for a new mission to sit on the same slope, one to two orders of magnitude above where radio put it if the mission carries an optical terminal. Frontier Expect DSS-23 in service and the 70-metre antennas cycling offline for refurbishment months at a time. Frontier The item to watch is the first operational — rather than demonstration — optical link on a deep-space science mission, and the number to ask for is how many terminals were ordered and delivered on schedule. Frontier Expect delay-tolerant networking to remain what it is today: standardised, running on the ISS, and without an interplanetary relay to be a network on.
• 25 yr: Frontier This is the window in which a deep-space optical ground segment either exists as a set of cloud-decorrelated sites or does not, and in which the terminal supplier base either publishes delivery rates or continues not to. Frontier It is also the window in which the first genuine multi-hop store-and-forward network could operate, if relay spacecraft are flown; that is the neighbouring brief's proposed column, not this one's. Established Voyager will be silent long before the end of this horizon, which retires the single-antenna commanding constraint by attrition rather than by solving it.
• 50 yr: Frontier A solar-system-wide network with optical trunks in the inner system, radio at the edges, and custody-transfer routing between them is the coherent long-horizon architecture, and every component of it exists today at the scale of one. Speculative Whether it is built depends on production rates nobody currently publishes, which is a weaker basis for forecasting than a physical limit would be.
• 100 / 250+ yr: Speculative At this range the interesting question stops being throughput and becomes latency: a network whose round-trip times are measured in hours changes what autonomy means for everything on it, and that is an architecture question rather than a link one. Handwave Interstellar links remain describable only qualitatively — enormous aperture-power products, tiny rates, latencies in years to millennia — and this page asserts no number for them, because no link budget for the interstellar case was obtained.

Frontier One near-term item belongs on this list and is cheap enough that its absence would itself be informative: the first occasion on which two separated ground apertures receive the same deep-space optical downlink in the same pass, or hand one over between them. That single event converts site diversity from an argument into a measurement, and it requires no new physics, no new spacecraft and no new standard — only a second receiver and a scheduler willing to try.

• Depends on This brief records no dependency on another brief. Nothing here waits on a physics result: the link budget is closed-form and has been since the 1960s, the optical demonstration is complete with a published rate-and-distance curve, the protocol stack is standardised through the internet and space-data standards processes with two working implementations, and the photon-counting detector arrays that made the demonstration work exist at both ends. What is missing is quantity — antenna hours and terminals — which is why the adjudication is constraints-only.
• Requires (not on this map) Both constraints are production rates rather than inventions, which is why neither is an institutional token. The first is aperture: gain scales with collecting area, a 70-metre dish has four times the area of a 34-metre one, DSS-14 dates from 1966 and was enlarged in 1988 for Voyager, the 70-metre antennas are now refurbished months at a time because replacing them with arrays was abandoned in 2021, and DSS-23 is expected in 2026 — against a manifest that a 2023 audit by NASA's own inspector general described as already oversubscribing the network. The capacity argument itself belongs to Deep Space Infrastructure, which fetched that literature directly; this page names the requirement because aperture is a term in its equation. The second is genuinely this page's: one deep-space optical terminal has ever flown. A network needs terminals on every outbound spacecraft, receivers at multiple cloud-decorrelated ground sites, and a line that delivers on mission schedules. Five suppliers are named in the public record, one of them was acquired by a launch prime in April 2026, and no production-rate evidence was obtainable in either direction — which is the honest status of the constraint and is stated as such rather than assumed adverse. The third constraint is the ground side of the same quantity problem and it is new to this row: receiving aperture at enough sites with decorrelated weather to give a service-grade availability, since cloud stops an optical link outright and two stations in one weather regime count as roughly one. That is a siting and construction rate rather than a discovery — dedicated stations, borrowed astronomical telescopes, or existing deep-space radio antennas re-surfaced into hybrid apertures — and no joint availability statistic for any candidate site set has been located.
• Enables No typed enabling edge is claimed, because an edge to every outbound mission would be an edge to most of the category. The concrete relationship worth naming runs the other way: this brief's subject is a service that Deep Space Infrastructure shows is already rationed, and the astronomical apertures that Mega-Telescopes owns are the same class of asset an optical ground segment would need — the 5-metre Hale Telescope served as the deep-space demonstration's downlink receiver.
• Adjacent Deep Space Infrastructure owns communications capacity and this page owns the link, in that brief's own words. Beam-Powered Propulsion shares the directed-energy hardware: large phased apertures, precision pointing, atmospheric propagation. Interstellar Probes owns getting to the solar gravitational lens focal region and Mega-Telescopes owns the lens as an instrument. Interstellar Civilization Models owns whether to transmit deliberately at all. Outside this map: detector physics, coding theory, the internet standards process, and the small set of firms that build space optical terminals.

Established “Laser communications delivers 267 Mbit/s from deep space.” That was the best point on a curve, at 31 million km — about 0.2 AU. At 1.5 AU the same system delivered 25 Mbit/s and at 2.7 AU a little over 6.25 Mbit/s. Frontier Quote the curve, not the peak: a factor of 44 in rate across a factor of 13 in distance is what the inverse square law looks like when a real system obeys it.

Established “DSOC is ongoing.” It is complete. The demonstration flew attached to Psyche from October 2023 and finished in 2025, with the final pass recorded in the neighbouring brief in September of that year. An earlier page on this site left the impression that it was still running, and this one writes it in the past tense on purpose.

Established “Deep space communications is upgrading from radio to light.” It is acquiring a second regime, not replacing the first. Optical is a throughput technology for the inner and middle solar system; radio works through cloud, needs no fine pointing, and has been returning 160 bit/s from 143 AU for forty-nine years. Frontier The two-regime picture is the accurate one and the upgrade framing is not.

Frontier “The interplanetary internet exists.” Delay-tolerant networking is a standardised protocol family with two implementations, years of ISS operational use, and a deep-space demonstration flown in 2008. A network requires relays, and the relays are proposals. Established The opposite error is worse: it is not a metaphor either — custody transfer is a real mechanism designed for a real problem, and solar conjunction is a scheduled multi-week outage no terrestrial protocol contemplates.

Frontier “Optical terminals can be produced at fleet scale.” Nobody knows, and this page will not pretend otherwise in either direction. The fetched supplier list carries no capacity figures and the one recent structural event — Rocket Lab's April 2026 acquisition of Mynaric — came with no delivery or financial data. Frontier The honest statement is that the evidence is absent, which is different from evidence of absence.

Frontier “That acquisition shows the terminal business was failing.” It might. It might equally show that terminals have become a strategic component worth vertically integrating, which is the reading the ministry conditions on keeping production and research in Europe would fit. Both readings are live and this rewrite obtained nothing that adjudicates between them.

Established “The Deep Space Network capacity crisis is this page's finding.” It is not, and the previous version of this page wrongly claimed it as a bottleneck. The oversubscription audit is real and the argument belongs to Deep Space Infrastructure, which fetched the capacity literature directly. This page cites it once and points. What replaces it here is the terminal-production constraint, which is genuinely this page's.

Established “DSS-43 is a management problem.” It is a link budget. Voyager 2's southern declination plus 160 bit/s from 143 AU means one southern-hemisphere 70-metre aperture closes the uplink, so one antenna on Earth can command the spacecraft, and its 2020 refurbishment left it uncommandable for eight months. Frontier That is the physics of a link expressing itself as an institutional fragility, and no amount of scheduling fixes it.

Speculative “Interstellar communication is just a bigger version of this.” It is bigger by an amount that changes the character of the problem: enormous aperture-power products, latencies of years to millennia, and feasibility in principle only at very small data rates. No interstellar link budget was obtained for this rewrite, so no rate appears on this page. Speculative The solar gravitational lens is an imaging proposal with a focal region beginning around 542 to 550 AU; the suggestion that its focal physics could amplify a communications link is an extrapolation and is flagged as one.

Frontier And the framing verdict. Deep space communications does not scale with demand. The physics scales as the inverse square, the apertures scale on a construction schedule, and the terminals have no published production rate at all. What DSOC proved is that the curve can be moved up by one to two orders of magnitude, which is a genuine and large achievement, and which leaves the slope exactly where it was.