Jonathan Herman

The orbital mechanic who wired the earth

By VastBlue Editorial · 2026-03-26 · 15 min read

Series: The Inventors · Episode 7

Where Cables End

Three billion people do not have reliable internet access. Not because the technology does not exist, but because the economics of physical infrastructure do not support building it where they live. Fibre optic cable costs approximately $20,000 to $40,000 per kilometre to lay in developed areas with existing road and conduit infrastructure. In undeveloped terrain — mountains, deserts, jungles, islands — the cost can exceed $100,000 per kilometre. For a remote village of a hundred people in sub-Saharan Africa or the Amazon basin, the infrastructure investment required to deliver broadband will never generate a return. The cable will never come.

Satellite internet was supposed to solve this problem decades ago. Geostationary satellites, parked at 35,786 kilometres above the equator, can see a third of the earth's surface from a single position. But the physics of geostationary orbit impose a penalty that no engineering can remove: latency. A signal travelling from ground to geostationary orbit and back covers over 71,000 kilometres. Even at the speed of light, the round trip takes approximately 600 milliseconds. Add processing delay at the satellite and ground station, and real-world latency reaches 700-800 milliseconds. This is enough to make video calls awkward, online gaming impossible, and real-time applications — financial trading, remote surgery, interactive cloud computing — impractical.

The latency is not a bug that can be patched. It is a consequence of orbital altitude. The higher the orbit, the larger the coverage footprint but the longer the signal path. There is no geostationary solution to the latency problem, because the problem is the speed of light.

The Graveyard of Satellite Dreams

The idea of blanketing the earth in low-orbiting communication satellites is not new. In the early 1990s, Motorola conceived Iridium — a constellation of 66 satellites in low Earth orbit at 780 kilometres altitude, designed to provide voice telephony anywhere on the planet. The name came from the original plan for 77 satellites, matching the atomic number of iridium. The engineering was breathtaking: cross-linked satellites forming a mesh network in space, each satellite carrying 48 spot beams, the whole system capable of routing a call from a handheld phone in the middle of the Sahara to a landline in Manhattan. Motorola spent over $5 billion building and launching the constellation between 1996 and 1998.

Iridium filed for bankruptcy in March 1999, nine months after commercial service began. The phones cost $3,000 each. Calls cost $3 to $7 per minute. The handsets were bulky and required a clear view of the sky — they did not work indoors, in cars, or in cities with tall buildings. Meanwhile, terrestrial cellular networks had expanded far faster and far cheaper than anyone in 1990 had predicted. By the time Iridium's satellites were operational, GSM coverage already served the vast majority of the world's paying customers. Iridium had built an elegant technical solution to a problem that the market had largely solved by other means.

Teledesic, backed by Bill Gates and Craig McCaw, was even more ambitious. Filed with the FCC in 1994, the plan called for 840 satellites in low Earth orbit providing broadband internet — not just voice — to the entire planet. The projected cost was $9 billion. Teledesic never launched a single operational satellite. The company scaled back its plans repeatedly — from 840 satellites to 288, then to 30 — before quietly shutting down in 2002. The combination of the dot-com crash, uncertain demand projections, and the sheer capital intensity of building a constellation from scratch proved fatal.

O3b Networks, founded in 2007, took a different approach. Rather than LEO, O3b placed satellites in medium Earth orbit at approximately 8,000 kilometres — a compromise altitude that offered lower latency than geostationary orbit (about 150 milliseconds round trip) without requiring the enormous constellation sizes of LEO. The name stood for "Other 3 Billion" — the unconnected population. O3b found a viable niche serving maritime, energy, and telecommunications backhaul customers, and was acquired by SES in 2016. But MEO was still a compromise: latency was better than geostationary but still too high for truly real-time applications, and the constellation could not match terrestrial fibre performance.

The pattern was consistent. Every previous attempt at satellite broadband had failed for one or more of three reasons: the satellites cost too much to build and launch, the constellation required too many satellites to be financially viable, or terrestrial alternatives improved faster than expected. The fundamental problem was economic, not technical. You could build a satellite constellation. You could not build one cheaply enough, and replace failed satellites fast enough, to compete with ground-based alternatives that were getting better every year.

The Low Orbit Trade-off

SpaceX's Starlink constellation chose a fundamentally different architecture: low Earth orbit, between 340 and 614 kilometres above the surface. At this altitude, round-trip latency drops to 20-40 milliseconds — comparable to terrestrial fibre and fast enough for any real-time application. The trade-off is coverage. A single LEO satellite can see only a small patch of Earth — roughly 1,000 kilometres in diameter — and it crosses the sky in about five minutes. To provide continuous coverage to any point on Earth's surface, you need not dozens of satellites but thousands, arranged in carefully coordinated orbital shells and planes.

Jonathan Herman was among the early engineers who worked on this coordination problem at SpaceX. The challenge he contributed to was not putting satellites in orbit — SpaceX had largely solved the launch problem with the reusable Falcon 9. The challenge was operating them once there. A 6,000-satellite constellation is not simply one satellite multiplied by six thousand. It is a dynamic, interacting system where every satellite's position affects every other satellite's coverage, where orbital mechanics and user demand create constantly shifting optimisation problems, and where failure of any component must be absorbed without service degradation.

Shells, Planes, and the Geometry of Coverage

The Starlink constellation is organised into concentric orbital shells — groups of satellites sharing the same altitude and inclination. The first and largest shell operates at 550 kilometres altitude and 53 degrees inclination, meaning the satellites' orbital paths extend from 53 degrees north to 53 degrees south latitude. This shell alone provides coverage to the most densely populated band of the earth's surface, from roughly Patagonia to northern Scotland. Within this shell, satellites are distributed across 72 orbital planes — imagine 72 circles around the earth, each tilted at the same angle but rotated slightly relative to the others, like the segments of an orange peel. Each plane contains approximately 22 satellites, evenly spaced along the orbital path.

Additional shells at 540 kilometres and 70 degrees inclination extend coverage further toward the poles. A shell at 560 kilometres and 97.6 degrees inclination — a sun-synchronous orbit — fills in the polar regions. Each shell adds coverage and capacity, but also adds coordination complexity. Satellites in different shells move at slightly different velocities due to their different altitudes, and cross each other's paths at the points where their orbital planes intersect. The constellation must be designed so that these crossings do not create collision risks or coverage gaps.

The handoff problem is perhaps the most demanding aspect of the coordination challenge. A user terminal on the ground communicates with whichever Starlink satellite is currently overhead. But that satellite is moving at 7.5 kilometres per second relative to the ground. It crosses the user's visible sky in roughly five to seven minutes, then disappears below the horizon. The system must seamlessly transfer the user's connection to the next satellite approaching from the opposite horizon — without dropping packets, without resetting TCP sessions, and without the user noticing anything has happened. This handoff occurs continuously, for every active user, across millions of simultaneous connections.

The beam-steering technology in the user terminal makes this possible. The original Starlink dish — the "Dishy McFlatface" that became an unlikely consumer technology icon — uses a phased array antenna that can electronically steer its beam across the sky without physically moving. The terminal tracks the serving satellite, shifts its beam to the incoming satellite at the optimal handoff moment, and maintains the connection through the transition. The timing must be precise: hand off too early and the signal from the outgoing satellite is still stronger; too late and the outgoing satellite has dropped below the minimum elevation angle, degrading signal quality through atmospheric path length.

The Laser Link Revolution

The inter-satellite laser link system represents perhaps the most technically demanding component of the Starlink constellation — and potentially the most transformative. Each second-generation Starlink satellite carries four optical laser terminals, allowing it to establish high-bandwidth communication links with neighbouring satellites in the constellation. These are not radio links. They are infrared laser beams, operating at approximately 1,550 nanometres wavelength, focused to hit a target the size of a coin from a distance of thousands of kilometres while both the transmitting and receiving satellites move at orbital velocity.

The pointing accuracy required is staggering. To maintain a stable laser link between two satellites separated by 2,000 kilometres, both moving at 7.5 kilometres per second on slightly different trajectories, the optical terminal must aim with precision measured in microradians — thousandths of a degree. The vibration from the satellite's own systems, the thermal expansion and contraction as the satellite moves between sunlight and shadow every 90 minutes, and the gravitational perturbations that subtly alter each satellite's trajectory all conspire to knock the beam off target. Active tracking systems using fine-steering mirrors and position-sensitive detectors must continuously correct the beam direction, hundreds of times per second.

Each laser link can carry approximately 100 gigabits per second — comparable to a major terrestrial fibre trunk. With four terminals per satellite, and thousands of satellites in the constellation, the aggregate bandwidth of the space-based mesh network is enormous. But the real advantage is not bandwidth. It is routing.

In the conventional internet, a packet travelling from London to Tokyo must traverse undersea cables that follow geographic constraints — routing around continents, through chokepoints like the Suez Canal or the Strait of Malacca. The physical path is far longer than the great-circle distance. In the Starlink mesh, a packet can travel in a nearly straight line through space, hopping from satellite to satellite along the shortest geometric path. For distances greater than approximately 3,000 kilometres, the space route can deliver lower latency than the best terrestrial fibre — because light in vacuum travels faster than light in glass, and the path is shorter. This has profound implications for financial trading, where microseconds matter, and for any application sensitive to long-distance latency.

The Collision Problem

Low Earth orbit is not empty. As of 2025, there were approximately 10,000 active satellites and over 30,000 pieces of tracked orbital debris — spent rocket stages, defunct satellites, fragmentation debris from collisions and explosions. Each piece moves at roughly 7.5 kilometres per second. At that velocity, a collision between a Starlink satellite and a ten-centimetre piece of debris would release energy equivalent to several kilograms of TNT — enough to destroy the satellite and create hundreds of additional debris fragments, each potentially capable of destroying another satellite in a cascading chain reaction known as the Kessler syndrome.

Donald Kessler, a NASA scientist, described this cascading failure mode in a 1978 paper. The concept is simple and terrifying: above a certain density of objects in orbit, collisions generate debris faster than atmospheric drag can remove it. Each collision creates fragments that cause further collisions, in a self-sustaining chain reaction that could render entire orbital bands unusable for decades or centuries. The Kessler syndrome is not theoretical — the 2009 collision between the defunct Russian Cosmos 2251 and the active Iridium 33 satellite created over 2,000 trackable debris fragments, many of which remain in orbit.

Starlink satellites carry krypton-powered Hall-effect ion thrusters that allow them to adjust their orbits autonomously. The collision avoidance system evaluates potential conjunction events — moments when two objects' predicted trajectories bring them within a defined threshold of distance — and calculates avoidance manoeuvres when the collision probability exceeds acceptable limits. These calculations must account for uncertainty in both objects' positions (orbital tracking has limited precision), the cascading effects of a manoeuvre on the satellite's future trajectory, and the potential impact on neighbouring Starlink satellites that may need to adjust their own positions in response.

The Factory That Changed the Economics

Every previous satellite constellation had failed, in part, because of the cost of building and launching satellites. A traditional geostationary communications satellite costs $150 million to $400 million to build and $100 million to $150 million to launch. Each satellite is essentially hand-built — custom-designed for its specific mission, assembled in clean rooms by highly specialised technicians over 18 to 36 months. At these economics, losing a single satellite is a catastrophic financial event. Building a constellation of thousands is unthinkable.

SpaceX attacked this cost structure from both directions. The reusable Falcon 9 rocket reduced per-launch cost to approximately $15 million for an internal Starlink mission — roughly one-tenth the cost of a traditional expendable launch. Each Falcon 9 carries 60 flat-packed Starlink satellites in a single launch, bringing the per-satellite launch cost below $300,000. But the larger revolution was in satellite manufacturing itself.

SpaceX builds Starlink satellites in a factory in Redmond, Washington, at a rate that would be unrecognisable to the traditional satellite industry. At peak production, the facility has produced up to 45 satellites per week. The satellites are designed for mass production: standardised components, automated assembly processes, integrated testing, and a design philosophy that prioritises manufacturability over per-unit optimisation. Each satellite weighs approximately 260 kilograms (for the v1.5 generation) and is designed to be flat-packed — stacked like books in the Falcon 9 fairing, deploying their solar arrays only after release in orbit.

The vertical integration is total. SpaceX builds its own phased array antennas, its own ion thrusters, its own laser communication terminals, its own star trackers, its own onboard processors. Where the traditional satellite industry relies on a supply chain of specialist vendors — one company for the bus, another for the payload, another for the propulsion, another for the solar panels — SpaceX manufactures most components in-house. This eliminates vendor margins, reduces integration complexity, and allows rapid design iteration. When an engineer identifies an improvement, it can be incorporated into production satellites within weeks, not years.

Four Million Subscribers and the Equity Question

By 2025, Starlink had over four million subscribers in more than seventy countries. The service was being used by military forces in Ukraine (where it became a critical communications backbone after terrestrial infrastructure was destroyed by Russian attacks), scientific expeditions in Antarctica, maritime shipping companies that previously relied on expensive and slow geostationary satellite links, airlines offering in-flight WiFi, and — most significantly for the original thesis — rural communities across five continents that had never had broadband access of any kind.

In Ukraine, Starlink's role was dramatic and widely documented. When Russia invaded in February 2022, targeted strikes destroyed much of Ukraine's terrestrial communications infrastructure — cell towers, fibre nodes, switching centres. Within days, SpaceX shipped thousands of Starlink terminals to Ukraine. The system became the backbone of military communications, enabling drone operations, artillery coordination, and command-and-control functions that would have been impossible without it. Civilian applications followed: hospitals transmitted medical records, local governments maintained administrative functions, and ordinary citizens stayed connected to the outside world. Starlink did not win the war. But it demonstrated, under the most extreme conditions imaginable, that a satellite constellation could replace destroyed terrestrial infrastructure almost overnight.

In the Amazon basin, Starlink has connected indigenous communities — Marubo villages along the Ituí River, Yanomami health posts in the upper Orinoco — that had no previous internet access of any kind. The nearest fibre connection might be 500 kilometres away through dense rainforest. Cellular coverage does not exist. The only previous communication option was expensive satellite phones with voice-only capability. These communities now have the same 50-200 Mbps broadband available to a suburban household in Ohio. The social consequences — access to telemedicine, educational resources, market information, and communication with the outside world — are still being understood.

The pricing challenge, however, is real. Starlink's standard residential service costs approximately $120 per month in the United States, plus $599 for the terminal hardware. In many developing countries, SpaceX has adjusted pricing — in Nigeria, the monthly cost has been reduced to approximately $35, and the terminal to $299 — but even these reduced prices exceed the monthly income of many potential users in the regions that need connectivity most. The physical layer no longer discriminates by geography. The economic layer still does.

Infrastructure engineering at its most consequential does not build faster tools for people who already have tools. It builds the first tool for people who have nothing. The orbital mechanics problem is, in a very direct sense, an equity problem wearing a physics costume.

Editorial observation

The Regulatory Constellation

Operating a satellite constellation is as much a regulatory challenge as a technical one. The radio spectrum that Starlink uses to communicate between satellites and ground terminals is a shared, finite resource, governed internationally by the International Telecommunication Union (ITU). Every satellite system must coordinate its frequency allocations with other systems to avoid harmful interference — a process that involves negotiations with national regulators in every country where the service operates. SpaceX has filed extensively with the ITU and the FCC, submitting detailed technical descriptions of its constellation design, power levels, beam patterns, and interference mitigation techniques. These filings, while bureaucratic, are the legal foundation on which the entire system operates.

The space debris question has drawn particular scrutiny. SpaceX has committed to deorbiting every Starlink satellite within five years of the end of its operational life — far more aggressive than the 25-year guideline previously recommended by international standards. The satellites are designed to use their ion thrusters to lower their orbit at end of life, increasing atmospheric drag until the satellite re-enters and burns up completely. For satellites that suffer thruster failure, SpaceX has designed them to passively deorbit through atmospheric drag alone within approximately five years from their operational altitude. No Starlink satellite is designed to become permanent debris.

The astronomical observation controversy adds another dimension. Starlink satellites are visible to the naked eye shortly after launch, when they are still in their lower parking orbit before raising to their operational altitude. Astronomers have documented interference with ground-based telescope observations — bright streaks crossing long-exposure images, contaminating data from optical and radio surveys. SpaceX has responded with several mitigation measures: a sunshade visor on early satellites (later replaced by dielectric mirror film on the v2 Mini generation), reduced surface reflectivity, and coordination with astronomical observatories to provide orbital predictions so that observations can be timed to avoid satellite passes. The measures have reduced but not eliminated the impact. The tension between global connectivity and ground-based astronomy remains unresolved.

The Orbital Layer

What Herman and his colleagues built is not a product. It is an infrastructure layer — a new stratum of global connectivity that sits between the terrestrial networks below and the geostationary platforms above. The engineering challenges were immense: coordinating thousands of satellites in multiple orbital shells, solving the handoff problem across millions of simultaneous connections, building laser links that maintain alignment across thousands of kilometres of vacuum, manufacturing satellites at automotive scale rather than aerospace scale, and navigating a regulatory environment designed for a world with hundreds of satellites, not tens of thousands.

The fact that all of this works — that a teacher in rural Montana can video-call a student in Lagos with 40 milliseconds of latency, that a research station in Antarctica can upload data at 100 Mbps, that a fishing vessel in the South Pacific can stream weather radar in real time — is the result of engineering decisions made by people whose names do not appear on product launch stages or magazine covers. The orbital mechanic does not seek applause. The orbital mechanic seeks a lower eigenvalue in the coverage optimisation matrix, a tighter pointing budget on the laser terminal, a cheaper bill of materials on the satellite bus. The work is invisible. The consequence is that the internet, for the first time in its history, has no geographic boundary.

Previous satellite constellations failed because they treated space as a premium environment for premium customers. Starlink succeeded because it treated space as a factory floor and orbit as a supply chain problem. The revolution was not in the physics. It was in the economics.

Editorial observation

Sources

1. SpaceX FCC filings — Starlink constellation modification applications, 2018-2024 — https://fcc.gov/ecfs/
2. SpaceX Starlink technical documentation and press materials — https://www.starlink.com
3. McDowell, J. "Jonathan's Space Report" — independent tracking of Starlink deployments — https://planet4589.org/space/stats/star/starstats.html
4. ESA Space Debris Office — conjunction event analysis reports, 2023-2025 — https://www.esa.int/Space_Safety/Space_Debris
5. del Portillo, I. et al. "A Technical Comparison of Three Low Earth Orbit Satellite Constellation Systems." MIT Lincoln Laboratory, 2019.
6. Kessler, D.J. and Cour-Palais, B.G. "Collision Frequency of Artificial Satellites: The Creation of a Debris Belt." JGR, 1978.
7. Bloom, D.A. et al. "The Iridium System: A Revolutionary Satellite Communications System Developed with Innovative Applications of Technology." Space Technology, 1996.
8. Teledesic FCC filings and corporate history, 1994-2002 — https://fcc.gov/ecfs/
9. O3b Networks / SES mPOWER technical documentation — https://www.ses.com/networks/networks-and-platforms/o3b-mpower
10. Handley, M. "Delay is Not an Option: Low Latency Routing in Space." HotNets 2018. — https://dl.acm.org/doi/10.1145/3286062.3286075
11. International Telecommunication Union (ITU) Radio Regulations and satellite coordination procedures — https://www.itu.int/en/ITU-R/space/Pages/default.aspx
12. Hainaut, O.R. and Williams, A.P. "Impact of satellite constellations on astronomical observations." Astronomy & Astrophysics, 2020.
13. Iridium LLC Chapter 11 bankruptcy filing, U.S. Bankruptcy Court for the District of Delaware, Case No. 99-45005, March 1999.
14. Starlink.com — availability, pricing, and coverage map data, accessed 2025 — https://www.starlink.com/map
15. Pavur, J. and Martinovic, I. "SOK: Building a Launchpad for Impactful Satellite Internet Research." IEEE S&P Workshop, 2022.
16. Humphreys, T. et al. "Signal Structure of Starlink Ku-Band Downlink." IEEE Transactions on Aerospace and Electronic Systems, 2023.