Brin2012Mar17

A response to comments to David Brin's Startram post

The Startram paper by Powell and Maise makes some "interesting" claims. The JHAPL folks assume repulsive magnetic levitation between a floating conductor and a surface conductor. If we help the authors by closing the loops to make coils (paging Dr. Kirchoff!), the result is a huge inductance, on the order of Henrys, and energy storage on the order of gigawatt-years because you are filling nearly a million cubic kilometers with high magnetic field energy. That's assuming that they don't quench their magic superconductors first. The conductors must be very large diameter to keep the B fields below critical. If there is a quench, there will be an electric arc with the power of a multimegaton nuclear bomb.

See also:

• http://launchloop.com/MassDriver

• http://launchloop.com/Superconductor

• http://launchloop.com/ZeroToEighty

Startram is not interesting. It won't work, because it is based on misunderstandings of launch, orbits, coilguns, and superconductors.

Most of this page is to respond to some thought-provoking questions raised about Server Sky around mid-April in the comments. You will find a lot more detail about Server Sky in the pages on this site, and in published papers it refers to. Use Text search, and Recent Changes, to look at the state of things. The site is not complete. This is a work in progress. It will become a book, and more papers, over time. I am sharing all this prior to completion to show my work as I go along, and get feedback.

Here's some excerpts from comments, with links to fuller discussion of them:

• ... a straight road to the Kessler syndrome ...

• ... Active collision warning and avoidance ...

• ... Timing coordination and accuracy ...

• ... digital beamforming ... grating lobes ...

• ... much lower bandwidth and huge antennas ...

For one of the inspirations for Server Sky (there are many), read Ivan Bekey's 2003 book Advanced Space System Concepts and Technologies. Also look for Bekey's book on Worldcat . Bekey, former head of NASA's Advanced Concepts office, tells us to use gossamer structures and connect them with information and maneuvering.

Another inspiration is the electrochromic windows on the 787 Dreamliner. electrochromic mirrors are appearing in buildings, they are cheap, durable, reliable, and thin.

Other inspirations include laptop screen thin films, Linear Signal's phased array TVRO antennas, rad hard hafnium oxide gate stacks, noise-tolerant computing, rad hard indium phosphide photovoltaics, and many other developments in solid state electronics. Moore's Law halves transistor prices every two years, and has transformed almost every industry. Relatively speaking, big iron satellites and space technology move at vacuum tube rates, and that industry is ripe for a Moore's Law makeover.

... a straight road to the Kessler syndrome ...

The Kessler syndrome occurs when derelict satellites collide, and the shrapnel disables other satellites, adding to the density of colliders. For a collision to occur, two quite different orbits must intersect, and the objects arrive at the intersection at the same time. Kessler syndrome is the potential result of:

• Crowded altitudes
• Fragile, large, easily disabled satellites
• Inaccurate tracking
• Inclined orbits
• Loss of manuevering fuel
• Satellites left in orbit after end of mission
• No systems for collecting derelict objects
• No economic incentives (positive or negative) to do so

So, don't do that!

Crowding: Server sky puts many more objects in orbit, but they travel as large arrays - essentially, large satellites connected by maneuverability and active cooperation rather than by aluminum struts. The same satellite mass is distributed more widely and more productively. Don't think of them as a bunch of independent satellites - think of them as a multi-ton satellite divided into cooperating components in nearly identical orbits. Ant colonies, not elephants. The arrays are deployed at 1 radii out, in the van Allen belt, where the non-constellation collider density is far smaller, and few other satellites can survive.

Fragility: The thinsats themselves are arrays of smaller subcomponents; except for memories and processor cores, the radio and maneuvering and optical thrusters are redundant, and can function independently even if debris punches a hole.

Tracking: See Active collision warning and avoidance below.

Inclined Orbits: All arrays are in near equatorial, shallow inclination orbits (less than 0.2 degrees) designed to never intersect with each other. Relative closing velocities will be centimeters per second. See ToroidalOrbits.

Fuel: Thinsats maneuver with light pressure, not expendable fuel. Rapid maneuver is not possible, but they can move tens of meters in an hour. With accurate tracking, that is sufficient to avoid colliders if they are all accurately tracked.

End of mission: Server sky thinsats will go obsolete at Moore's Law rates, at which time they could maneuver to reenter. Given their tiny ballistic parameter, they will come down in days if perigee drops below 1000 km altitude. They can also be nudged back into larger trackable stacks, or get cut apart and used as ballast for thinner next-generation thinsats.

Collecting derelicts: Solar powered electric thrusters take longer to maneuver, but with patience can create far more delta V with less fuel. Jerome Pearson ( author of first technical space elevator paper in Acta Astronautica, 1975 ) says it can be done with tethers. It may also be done with laser ablative thrusters, or VASIMR electric engines, or ... Which is to say, I don't know which option is best, but there are many. Small objects can be "caught and caged" through controlled collision in a capture structure, without matching delta V.

Economic incentives: Given a growing demand for ballast for ever thinner thinsats, the value of collected debris to be reused as ballast will approach that of earth-launched mass. Rather than trash, debris will become valuable. A kilogram in orbit is far easier and cheaper to deliver to the m288 server sky orbit than a kilogram on the ground. Debris passing near the server sky orbit (mostly spent upper stages for GEO satellites) will be the cheapest to maneuver, while removing the biggest collision threats).

So, thinsat arrays can be the solution to Kessler syndrome - a high paying market for the debris that is up there, and an element of the collection process.

... Active collision warning and avoidance ...

Thinsats will be functionally agile. When they are not sending intra-array or ground transmissions, the transmitters are reconfigured to transmit narrow beam, lookdown, time-coded radar chirps. At 6400km (1 earth radius out), they can see far more of the sky than ground radar, at shorter wavelengths, unhindered by the atmosphere. Thinsats are poor radar receivers, but other satellites in different orbits can pick up the signals. It should be possible to track much smaller objects for a much longer time, permitting orbit determination to meters, not kilometers. That makes avoidance much easier.

Lageos demonstrates that we can measure satellite position within fractions of a micrometer over tens of thousands of kilometers.

Thinsats maneuver at about 10 micrometers per second squared - a thinsat can move its 10 centimeter radius out of the way of a sub-centimeter collider in 2.5 minutes, a 7 meter object in 20 minutes. Assuming precision radar and a correspondingly large ephemeris, a thinsat is unlikely to encounter anything it can't avoid and can't survive.

If a thinsat does get irreparably damaged, it may be collected for ballast as mentioned above. If it is completely derelict, another thinsat can be sacrificed to nudge it towards a collector, or lower the perigee for reentry.

... Timing coordination and accuracy ...

This discussion assumes 2020 technology and continued Moore's Law performance improvements. Scale by 2X/2years to whenever you are reading this.

Thinsats do not perform well as autonomous units, just as single ants do not perform well without a colony. The spectacular system performance relies on thinsats working together in an array, trading information and checking each other's measurements in a distributed and non-centralized way. Timing is a discussion of internal thinsat timing.

Internal to a thinsat, up to 100GHz I and Q timing signals will be generated from characterized and tuned SAW oscillator on tuned resonant paths (to reduce timing jitter) and distributed to the transmit chips, for mixing with GHz bandwidth I and Q IF signals from lower speed digital packet modulators. This is for intra-array communication, ground transmissions will use a separate 38GHz timing system.

Assume the thinsats are in a 32x32x32 array with 10 meter spacing ( < 1% shading ). They are in the same inertial frame (very nearly), with relative shear velocities under 20 centimeters per second and shear accelerations under 100 μm/s², array edge to array edge, so there are no appreciable relativistic effects between them. There will be variable distance and timing skew over the 4 hour orbit, but that can be characterized.

Assume 100 millisecond thrust adjustment intervals, and 1μm/s² thrust control steps. The velocity jitter for a thinsat will be ±25 nm/s, and the position jitter will be ±0.6nm. The doppler shift will be 8e-17, and the path time jitter will be less than 2e-28 seconds. The position of the thinsats will be locked in space to within measurement error.

A larger source of timing error could be flexing in the thinsats, with associated position and velocity errors for each antenna element on the thinsat. The primary vibration mode for a 50 μm thick, 20 cm wide thinsat has a many second period. That must be actively damped with the thrusters. Stated more accurately, thrusters (the source of flex energy) must be operated smoothly, to minimize flexing and associated timing error. This kind of error (at MUCH higher speeds) is encountered and conquered when moving wafer platforms in semiconductor photolithography tools; in those, the stage error at every point on the platform must resonate less than a nanometer or so. This will be pre-characterized with CAD during thinsat design, characterized per thinsat in orbit, and acceleration profiles stored in tables.

Over the course of a 4 hour orbit, we will change grid configuration a few times. Arrays will have velocity shear, and the nearest neighbor in the Z direction (earthwards/outwards in the orbital plane) will shift between 5 inline candidates. Maximum velocity along the hypotenuse will be about 3 millimeters per second between nearest neighbors.

Assuming GHz bit rates and megabit packets, distances between neighboring thinsats will change by 3 microns during a packet, a timing delay change of 10 femtoseconds. This will certainly affect accurate clock calibration, but in an easily characterized and predictable way; we don't need to involve the CPU at all, and probably not the calibration DSP, for dealing with this.

These and many other timing errors can be discovered and modeled during design, before manufacturing and launch. No doubt many others will be discovered in orbit, limiting initial performance but contributing to better design of subsequent generations of thinsats. The most important thing is to produce, launch, and gather data from many thinsats. The billionth thinsat will be much better than the hundredth, and the trillionth will be much better than the billionth.

This is the key insight to the rapid development of semiconductor technology - make vast quantities of identical units, and learn everything you can from the failures. That is how the Russians managed to build good rockets in spite of the systemic flaws of communism; this is how Krafft Ehricke told Americans to build theirs (we were too rich to pay attention to him, we built a few big Saturn Vs and Space Shuttles instead). Ants are far more successful than elephants, and right now my kitchen is evidence of that.

... digital beamforming ... grating lobes ...

Ground antennas

At 38 GHz, a meter-squared dipole array with elements at λ/2 spacing (4 mm) would contain 60,000 elements. This seems enormous, but consider the computer screen you are looking at now, with millions of pixel elements for less than $100; it is quite possible that grids of wires, phased planes of distribution conductors underneath, and nonlinear mixer materials could be made to drive these antenna elements. It is likely that this would be quite inefficient and have lousy noise figure, but we can put many kilowatts into a packet at the server sky end; efficiency and good noise figure may not be necessary. So this hypothetical-technology array may reduce the cost of a phased array antenna just as LCD arrays have reduced the cost of solid state light sources.

The angular resolution of the panel will be approximately λ/2W radians, or about 2 milliradians, or about 0.1 degrees. For ground stations directly on the equator, the beams will travel through the server sky orbit and impinge on geosynchronous satellites; not good. However, for ground stations above or below a degree north or south, the beams, aimed at the M288 6411 km altitude, will sweep well below or above the geosynchronous arc. This is an advantage of using a MEO orbit.

In the beginning, we will probably just use traditional parabolic dishes on equatorial mounts pointed at individual arrays; we won't need much azimuth control, and the tracking speed can be less than 3 degrees per minute.

Phased Array Transmitters in Orbit

Traditional phased arrays use two dimensional grids of dipole elements spaced evenly at less than λ/2 CHECKME spacing. If they are spaced more widely, there will "grating lobes", spatial aliasing, with strong signals at other angles besides the primary beam. This happens because the dipole elements also constructively add at large angles.

Server sky arrays are three dimensional, and the spacings are intentionally not regular. The thinsats are spaced far apart so that they do not completely shade each other as the array turns around itself and turns around the earth, changing its orientation towards the sun. That necessarily makes the density of dipoles low enough to splatter significant power to other angles off the primary beam. The depth of the array does provide more spatial selectivity than a single plane of thinsats.

MoreLater - add a PhasedArray page with the phased array simulation outputs (already in the slide shows) and the code. Animations will not help for the book, sigh

However, the thinsats themselves are about 20 cm across, and with a λ/2 spacing of 4mm, each one can produce a beamwidth of 0.02 radians, or about one degree. The spacing between thinsats in an array will not be uniform; small fractional wavelength dithering in position (which we can control to submicron accuracy) smears out the grating lobes. Random dithering is good, though I am working on better dither functions. The off-axis energy is still there, and raises the noise floor for ground receivers off the main beam, but there are no peaks. Spread spectrum correlators can extract signals well below the noise floor; ground receivers can tune out most of the uncorrelated "noise" power, and their 0.1 degree angular selectivity can tune out more. A 1% filled array will scatter 99% of its power off-axis, within the one degree thinsat spread (200 km ground spot). But 1% of the array power from a 32-cubed 10 meter array will be focused on a 12.5 microradian ground spot, 125 meters across, 25,000 times the power level of a the off-beam scattered power of an interfering array.

• BTW, it may be possible to increase array fill without shadowing by clever array design and per-orbit active maneuvering. An interesting geometry problem. A more closely packed array will put more power into a wider main beam, and less into the thinsat footprint. The power ratio will stay the same, a function of thinsat spread and array count.

125 meters is similar to WiMAX selectivity in an urban area. However, urban areas are much better served by optical fiber. Server sky is much better suited to serving suburban and rural areas, or serving antenna farms surrounding urban areas and feeding into them over fiber. Server sky is designed to reach billions of unserved people in the developing world, and grow as they become wealthy. It is insane to get locked in deadly competition with entrenched companies in saturated markets.

MoreLater

Phased Array Receivers in Orbit

MoreLater

Intra Array communication

It will be relatively easy to send signals from satellite to satellite, since the distances are short. Inter-satellite links can be very high bandwidth (many GHz) and high signal-to-noise. We can use frequencies that do not penetrate atmosphere, and The main source of interference

MoreLater

... much lower bandwidth and huge antennas ...

Server sky will operate in the computational and internet paradigm. Limited information (such as search requests, or initialization data) up, large data sets down. We need huge bandwidth down, less bandwidth up. Until we get good enough with the radio, initial applications will be very compute intensive relative to bandwidth, and there are many applications for that. Protein modeling, for example; scientists in developing countries with unreliable power grids and lacking money for big compute farms may be able tor rent time to perform such calculations.

In the longer term, radio will improve to the point that we can take on bandwidth-intensive tasks. One that I am fond of is speech translation; doing that correctly can involve immense computation during the conversation. International calls from remote villages in India to world commodity exchanges, remote classrooms for global education, etc. could help villagers thrive as citizens of the global community. They could uplink and downlink through meter-sized phased arrays attached to existing cell towers.

Protocols will differ from standard TCP/IP, but we will borrow a lot of IPV6 .

MoreLater

This is a work in progress. I am filling in the holes, it will take about two weeks to finish.