Chapter II: Part 2

The Echo I satellite was launched into a circular orbit inclined at an angle to the plane of the earth’s equator. In _Figure 2_ this equatorial plane intersects the plane of the satellite orbit along the line OPM. The point O represents the center of the earth, the point M is on the satellite orbit, and the Point P is on the equator. At any instant, the satellite may be located in its orbit by the angle θ, which is measured between the line OM and the line OQ, where the point Q is the satellite’s location. If the satellite moves in a circular orbit, as in this case, the angle θ is proportional to time. That is, we can write θ = _nt_. We call _n_ the _angular speed_ of the satellite; one way of measuring this is in degrees per second.

Thus, the satellite is whirling at a constant speed about the earth like a stone tied to a string. Let us examine the physics of this situation a little more closely with the help of _Figure 3_. If the satellite is moving with the velocity _v_, then we know that the centrifugal force acting on it is

(_mv_²)/(_r_),

where _m_ is the mass of the satellite and _r_ is its distance from the center of the earth. Obviously, no string ties the satellite to the earth, but the force of gravitational attraction between the earth and the satellite has the same effect. Newton’s law of mutual attraction tells us that this force is proportional to the product of the two masses divided by the square of the distance between their centers, or

(_km_)/(_r_²),

where _k_ is a constant that essentially represents the mass of the earth.[1] Newton’s law also tells us that this force will be pointing toward the center of the earth if the earth is spherical. When the satellite is in circular motion, the centrifugal force and the gravitational force must balance each other. Hence we have

(_km_)/(_r_²) = (_mv_²)/(_r_)

and from this we can solve to find that the velocity of the satellite must be equal to

_v_ = √((_k_)/(_r_)).

In the case of the Echo I satellite, which was designed to have a radial distance of _r_ = 5000 miles, this velocity amounts to about 4.4 miles per second. The time for one revolution in orbit is obtained with the formula

_T_ = (2_πr_)/(_v_).

For the Echo satellite this time, _T_, turns out to be just about two hours.

Calculating the Orbit of Echo I

These basic physical principles of satellite motion can give us many useful answers. They tell us how fast we must move a precision tracker to follow the satellite through the sky, how much time a satellite will spend above the horizon, and how long will be the time from one chance of seeing it to the next. However, in the Echo project we were not merely concerned with planning our experiments from hour to hour; we also needed to know how the satellite would move for weeks and perhaps months in advance. When you study the motion of a satellite over such a length of time, you discover that its circular orbit will not remain the same as it was at launch. This fact had been observed on other satellites and was to be expected also with Echo.

In everything we have said so far it was assumed that the earth was a perfect sphere, which is the way a geographer’s globe presents it to us. In reality, the earth is somewhat flattened, with its diameter from the north pole to the south pole being somewhat shorter than its diameter at the equator. One way of looking at this is to visualize the earth as a sphere with some material added in the equatorial zone, which we may call _equatorial bulge_. This bulge causes Echo’s orbit to have a slow “wobble” about the earth’s polar axis, somewhat like that of a spinning top.

Another force that makes the satellite’s orbit shift slightly is the faint pressure caused by the light from the sun. Although this pressure is much too small for us to perceive without the help of very delicate instruments, it is enough to affect a satellite, which has nothing to support it in space and is exposed to solar pressure for a very long time. Since the Echo balloon is a plastic sphere, 100 feet in diameter, that weighs only a little more than 100 pounds, the light rays striking its surface are enough to cause a second “wobble” effect. This wobble centers about the line from the earth to the sun. Light pressure also forces the orbit to go slightly out of round from a perfect circle, and other gradual effects on the satellite’s orbit are caused by the gravitational attraction of the moon and the sun.

All these disturbances are ever-present and act simultaneously, and a satellite’s total response to them is very complicated. Fortunately, however, most of the changes take place at a very slow and uniform rate, and we can predict them fairly accurately.

Calculating the Orbit of Telstar I

In Project Telstar we had to calculate the satellite’s orbit from observations made by our precision trackers. This introduced a few problems in addition to the ones we encountered with Project Echo. In the first place, the orbit of the Telstar satellite is a elongated ellipse, as indicated in _Figure 4_, rather than being almost circular, as in the case of Echo I. We mentioned earlier that a precision tracker can furnish data on a satellite’s elevation angle, _E_, and azimuth, _A_ (see _Figure 1_). It can also give us a reading for ρ, the distance from the tracker to the satellite (_Figure 4_). If we know the position of the tracker on the earth, we can reduce the quantities _A_, _E_, and ρ to the angle θ and the distance _r_ (measured from the center of the earth to the satellite). These two quantities locate the satellite in the plane of its orbit, but in order to describe its position completely we must also specify this orbital plane. In _Figure 5_ the orbital plane is shown as a shaded surface, with θ and _r_ being the same as before. You will recall that the line OM represents the intersection between this plane and the equatorial plane; we call the angle _i_ between the two planes the _inclination_ of the orbit. Finally, we have the angle Ω between the line OM and some line OA to the point A, which we can choose as any convenient spot in the equatorial plane. Now we have specified the orbital plane completely. The point A can be found from day to day by fixing its position relative to a certain star in the sky.

Figures 4 and 5 tell us something about the geometry of the satellite’s _position_ in space, but for the complete story we must also give the _time_ at which it can be found there. For this purpose, there are some astronomical laws that relate position on an elliptic orbit to time. Two of these are illustrated in _Figure 6_; in looking at this figure, you should imagine that you are standing off to one side of the orbital plane to get a good view of the entire orbit. The longest dimension of the ellipse, 2_a_, is called the _major axis_; this dimension is related to the satellite’s _period_—the time it takes to go once around the ellipse. More than three hundred years ago the astronomer Johannes Kepler observed that the period _T_, of an ellipse is

_T_ = 2π√((_a_³)/(_k_)),

where _k_ again was (using Newton’s work) essentially the mass of the earth.

Instead of a complete revolution, we may only be concerned with part of one orbit. Let’s say that this part lies between the two positions P₁ and P₂ that the satellite occupies at the two times _t_₁ and _t_₂ (see _Figure 6_). Then another of Kepler’s laws says that the ratio between the time difference _t_₂ - _t_₁ and the period _T_ equals the ratio between the sector of the ellipse OP₁P₂ and the area of the entire ellipse.

Now let us see how we can use the quantities _r_, θ, _i_, and Ω as well as Kepler’s two time laws to determine the motion of the satellite in space. Suppose that we have made observations of the Telstar at two times _t_₁ and _t_₂ and that we have measured its distance along lines ρ₁ and ρ₂ in _Figure 7_. In other words, we know that at these two times the satellite was at the points P₁ and P₂. Since three points determine a plane, we know in this case that P₁, P₂, and O define the satellite orbital plane. Knowing this, we can now calculate the angles θ₁ and θ₂, the distances _r_₁ and _r_₂, and the angles _i_ and Ω. (The detailed formulas for this are derived from analytic geometry.)

However, we still do not know the length and the width of the particular ellipse the satellite is following and how this orbit is oriented within the orbital plane. Let us imagine again that we can stand off to one side of the orbit and take a good look at it; _Figure 8_ shows us what we would see. There are the two points P₂ and P₁ at which we have observed the satellite. We know the positions of these points relative to each other and in relation to the center of the earth, because we have already calculated _r_₂, _r_₁, θ₂, and θ₁, But any number of ellipses could be made to pass through these two points. Some might be very large, others might be so narrow that they would intersect the earth and thus be impossible. However, only one of these ellipses will satisfy the time difference that we observed between P₁ and P₂. In other words, the shape and period of this particular ellipse must be such that it will cause the satellite to pass through P₁ and P₂ in exactly the time interval _t_₂ - _t_₁. If we work out our time formulas, we will convince ourselves that there is only one such ellipse. When we have found it, we have determined the orbit of the Telstar satellite from the two observed positions and times.

In principle then, we could keep track of the Telstar satellite by making a pair of observations P₁ and P₂, and then predicting ahead a short segment of an orbit that is the ellipse we have computed. After a while we must verify this ellipse with two more observations P₃ and P₄, predict ahead over another segment, and verify again with P₅ and P₆ (see _Figure 9_), The reason we have to keep taking new measurements is that the elliptic orbit does not remain the same. As we discussed in connection with Echo I, the orbital plane will “wobble” about the earth because of the equatorial bulge. We also know that the orbit’s major axis will revolve within the orbital plane. As we have seen before, these effects are small and can be represented by appropriate mathematical formulas. If we calculate them, we will see the connection between one pair of observations and a later one, and eventually we can increase the time interval between successive pairs of observations. There are also mathematical formulas that we can use to predict the position of the satellite for many revolutions in its elliptic orbit.

In order to predict orbits successfully, we must also realize that the measurements we obtain from a precision tracker, such as the angles _A_ and _E_ and the distance ρ, are always subject to small inaccuracies. Thus it is not really possible to take just two measurements like P₁ and P₂ and determine a satisfactory orbit from them. In reality, our tracker takes many readings, and these are averaged to give adequate information about the orbit. Therefore, the picture we have in mind is not quite like _Figure 7_, but rather like _Figure 10_. Here the trackers have established a series of points that are somewhat scattered, and by taking averages we can calculate an orbit that passes through them in a smooth fashion.

The trackers we have mentioned so far have given us azimuth and elevation angles and also the distance to the satellite at every instant. Sometimes we must use simpler instruments that do not yield all this information. They might, for instance, only give us the two angles. The mathematics of calculating an orbit from such measurements is somewhat different, but the process is fundamentally the same as we have discussed here.

When you do these calculations for the Telstar satellite from one day to the next—and especially if you have more than one satellite to keep track of—the amount of work will become quite large. Nowadays our calculations are done for us on electronic computers, which both receive information from the tracking instruments automatically through Teletype or DataPhone channels and send back information concerning future positions of the satellite to the ground stations. There are still quite a few problems to be solved, and we are presently working on ways of making all this equipment perform the orbit predictions for the Telstar satellites automatically and efficiently.

Franz T. Geyling _was born in Tientsin, China, and received a B.S. in
1950, an M.S. in 1951, and a Ph.D. in 1954 from Stanford University.
He joined Bell Telephone Laboratories in 1954, and has been engaged in
celestial mechanics studies of rockets and satellites, as well as
stress analysis of submarine cables._

CASE HISTORY NO. 2
What Color Should a Satellite Be?

Peter Hrycak
_Mechanical Engineer—Member of Staff, Electron Device Laboratory_

THE PROBLEM

_It is important for a satellite to stay at the proper temperature while it is orbiting in space. The instruments aboard it must continue to operate properly, and one way of insuring this is to keep them from being exposed to extreme heat or cold. We can, of course, regulate a satellite’s temperature somewhat with various kinds of devices, and we can see that one of its ends does not point towards the sun for too long. But in designing the Telstar satellite we also wanted to control temperature in an easier way: by covering the satellite’s external surface with material with the best properties—including the right color—for maintaining its over-all temperature at the right level._

The Radiation of Heat

A satellite’s temperature is determined by the balance between the heat that enters the satellite and the heat that leaves it. This means that we must be concerned with how heat is transferred. Heat can be transferred in three ways: by _conduction_, when two bodies are in direct contact and their molecules collide; by _convection_, which utilizes the movement of warm currents in a fluid; and by _radiation_, in which heat energy travels as electromagnetic waves at the speed of light. With a satellite, we are concerned only with the last of these, since the only way energy can be gained or lost in space is by radiation.

In the transfer of heat by radiation, the surface of the heated body—such as a satellite—is very important. All energy gained must be absorbed at the surface; all energy leaving must be emitted at the surface. So the physical properties of this surface control how energy is absorbed and how it is emitted. The origin of the radiant energy is vitally important; most surfaces, for instance, will behave differently when exposed to solar radiation from the sun’s temperature of 10,000° Fahrenheit than when exposed to radiation from nearby objects at room temperature.

Absorptivity and Emissivity

The physical property of a material that controls the way it absorbs radiant energy is called its _absorptivity_, and the property that controls its emission of energy is its _emissivity_. For absorptivity we use the symbol α; for emissivity we use the symbol ε.

When radiant energy reaches a surface, only a certain part of it is absorbed; the rest is either reflected, just as light rays are reflected, or else passes right through it. The absorptivity, α, of a substance tells us what percentage of radiant energy it will absorb. A perfect absorber, or _black body_, would absorb all the radiant energy that reached it. If such an ideal substance existed (which it doesn’t) we would say it had an α of 1. The actual absorptivities of real substances are indicated by numbers between 0 and 1: The α of black velvet cloth, for example, is about 0.97; that of a polished silver mirror is about 0.08 for solar radiation (absorptivity for most polished metals for room temperature radiation is even lower).

We measure emissivity, ε, in very much the same way. A hypothetical black body would emit all the energy it possibly could and have an ε of 1; the emissivities of real substances are indicated by numbers between 0 and 1. For any given frequency (or color) of light, a substance’s absorptivity and emissivity are equal; however, the total spectrum of frequencies of the energy absorbed is usually different from that of the energy emitted.

The ratio between emissivity and absorptivity, α/ε, is very important, as we shall see later. If this ratio is greater than 1, it means that a substance absorbs heat faster than it emits it, and thus tends to become warmer. If the ratio is less than 1, the reverse is true—the surface emits radiant energy at a faster rate than it absorbs it, and tends to become cooler.

How We Measure the Radiation of Heat

This is one of the fundamental relationships of modern physics:

_Q__body = ε_A_σ_T_⁴.

It was discovered experimentally by Josef Stefan in 1879, and verified theoretically by Ludwig Boltzmann; it is known as the _Stefan-Boltzmann Law_. This formula tells us the amount of radiant energy, Q_body, that will be emitted by a body having the surface area _A_ when it is at the temperature _T_. Temperature, here, is measured in degrees Rankine (°R), or Fahrenheit temperature above absolute zero (to calculate degrees Rankine, add 460 to the temperature in degrees Fahrenheit). The expression ε_A_ is used to show that only a certain fraction of the energy that would leave a perfect black body of area _A_ will actually leave a real body of the same size; the size of this fraction is determined by the body’s emissivity. The symbol σ is a quantity we call the _Stefan-Boltzmann constant_.

We can also calculate the heat from the sun that will be absorbed by a body. If we let _S_ be the total amount of solar energy that would be absorbed by a perfect black body, α_S_ will be the amount that is actually absorbed by a body with an absorptivity of α for solar radiation. If our body is a spherical satellite, the sun’s rays will only strike it from a single direction. Thus only an area equivalent to the sphere’s cross-section (largest inscribed circle) will receive energy at any one time. Since, _as shown in the sketch_, this area (_a_ = π_r_²) is one-fourth that of the sphere’s total surface area (_A_ = 4π_r_²), we know that the radiant energy from the sun that is absorbed will be

_Q__sun = (_A_)/(4)α_S_.

A man-made satellite’s position relative to the earth is very like that of the earth in relation to the sun; the earth, after all, is itself a satellite of the sun. And during most of its useful life a satellite will be in _thermal equilibrium_—it will be losing just as much heat energy by its own radiation into space as it will be gaining from other sources, primarily the sun. Since the total amount of energy it absorbs is equal to the amount of energy it emits, _Q__body is equal to _Q__sun. This means that we have the equality

ε_A_σ_T_⁴ = (_A_)/(4)α_S_.

Now, if we solve this for temperature, we will get

_T_ = ((α)/(ε) × (_S_)/(4σ))(^¼).

This equation is well known in astronomy, and has been used for more than 80 years to calculate the temperatures of various objects in the sky. Today, we still find it useful for measuring the surface temperatures of man-made satellites such as Telstar. Since both _S_ and σ are known constants (in this case, we use the quantities _S_ = 445 and σ = 0.173 × 10⁻⁸), you can see that temperature is dependent on the α/ε ratio.

Finding the Right Surface for Telstar

shutter (closed)
electrical heat transfer
insulation blanket
heat transfer by radiation
shutter (open)
solar cells
electronic chassis
heat transfer by circulation
thermal control mechanism
nylon lacing
microwave antennas
frame

In designing the Telstar satellite, both its internal and external temperatures had to be controlled. The electronics canister inside the satellite operates best if it stays at approximately room temperature of 65 to 75°F. This much heat is supplied in the canister by dissipation of electrical energy from the solar cells. The container is well insulated to keep its temperature relatively stable, and it has shutters that open automatically if it begins to get overheated (_see above_). The operating characteristics of the solar cells on Telstar’s surface also had to be considered; they work better at rather cool temperatures. So we decided to keep the satellite’s skin at an average temperature of about 0°F, although temperatures actually will range quite a bit above and below the average as the satellite moves from sun to shadow.

Now, using this average temperature of 0°F (converted to 460°R) as _T_ in our formula, we can solve for α/ε. We find that this gives us a ratio of approximately 0.7 for the satellite’s surface. However, this presents a problem. Almost 40 per cent of Telstar’s surface is taken up by its power plant of 3600 sapphire-covered solar cells. These cells, unfortunately, have a relatively high α/ε ratio—their α is 0.8 and their ε is 0.54, for an α/ε of 1.5. This means that the portion of the surface not used by either solar cells or antenna openings must, in order to give us an over-all average of about 0.7, have a very low α/ε ratio—less than 0.3.

To get this sort of ratio, we had to select carefully the material for the outer surface of the Telstar satellite. There were many kinds of surfaces that might have been used; they could have been metal or non-metal, rough or smooth, shiny or dull. And they could have been any color from black to white. However, to get a 0.3 ratio we needed something with a relatively high emissivity for the low-frequency electromagnetic radiation that the satellite emits and a rather low absorptivity for the high-frequency radiation coming from the sun. High emissivity meant that we should use a nonmetal surface rather than polished metal, since the emissivity of nonmetals is quite high at the temperatures in which we were interested, while that of polished metals is relatively low. And, to get low absorptivity, we decided that the color of these surface areas should be very close to a pure white.

There were several substances that met our requirements. After testing a number of them, we decided to use aluminum panels coated with a thin layer of aluminum oxide (Al₂O₃). This coating is very pure, hard, and stable, and we left it rough to minimize changes due to meteoroid abrasion. Its α/ε ratio is 0.24. The aluminum oxide coating can be applied by means of the plasma jet process—particles of aluminum oxide are heated to a partially molten state, mixed with gases, and then sprayed onto the cleaned, pre-coated aluminum panels (_see illustration above_).

Using this carefully selected outer surface has helped solve the temperature-control problem. Since Telstar has been in orbit its internal and skin temperatures have kept well within the ranges we wanted them to. Thus you can see how some basic formulas from classical physics helped us choose the right material for the satellite’s surface—and even what color it should be. The blue-and-white checkered appearance that Telstar I finally took on was no accident—it was the result of carefully combining various colors and materials in just the right amounts to obtain the temperature balance we needed.

Peter Hrycak _was born in Przemysl, Western Ukraine, and received a
B.S. in 1954, an M.S. in 1955, and a Ph.D. in 1960 from the University
of Minnesota. He joined Bell Telephone Laboratories in 1960, and has
worked on low temperature refrigeration problems and thermal design
and thermal testing of the Telstar satellite._

CASE HISTORY NO. 3
How Can We Make Optical Measurements on a Satellite?

Jeofry S. Courtney-Pratt
_Physicist—Head, Mechanics Research Department_

THE PROBLEM

_Since the first Telstar satellite went into orbit, we have tried to trace its path through space precisely. But we also have had to keep a constant check on the position, or attitude, that the satellite takes as it travels. We are particularly interested in the direction of the spin axis about which it revolves, and we also want to know its spin rate, which is the number revolutions the satellite makes each minute. Although these might seem relatively simple jobs, they actually turned out to be rather complicated. And only at virtually the last minute, just before the satellite’s design was finally set, did we think of a new way of using reflected flashes of sunlight to report on its spin axis and spin rate._

Why We Want to Know About the Spin Axis

When the satellite was injected into its orbit, it was spin-stabilized to keep it from tumbling over and over, much as a rifle bullet is stabilized by being spun about its longitudinal axis. The Telstar satellite is roughly spherical, and it was designed to spin with the helical antenna end as its north pole and the antenna bands as its equator. On July 10, 1962, the satellite was given an initial spin of 177.7 revolutions a minute. As we expected, this rate is decreasing gradually; after two years it will only be spinning one tenth as fast.

The most important reason for keeping a close watch on the satellite’s spin axis is to make sure that microwave signals are sent and received steadily. It isn’t possible to build an antenna that radiates at exactly the same power in all directions. Telstar’s antennas work very well, but they operate better in the direction of the satellite’s equator than they do towards its poles. This means that if the spin axis is constantly changing transmission will fade in and out—even at times passing through “null” where no transmission at all is possible. No single fixed orientation is perfect for the spin axis, but we decided that the best average position would be to keep it always perpendicular to the plane of the earth’s orbit. We tried to make sure that the spin axis would not vary by more than five degrees from this direction at any time—although it probably could depart as much as 15 or 20 degrees without doing serious harm.

A second reason for being careful about the satellite’s spin axis is the problem of heat balance. If one end of the satellite points constantly at the sun and the other end does not, the end near the sun will get much too hot and the other will get much too cold. Therefore, we tried to fix the spin axis so that it stayed perpendicular to a line drawn from the satellite to the sun.

We also wanted to get a continuing report on the effects of the magnetic field of the earth at high altitudes. We knew these would cause the spin axis to change with time, or _precess_, but we couldn’t be exactly sure what these changes would be.

Since the orientation of Telstar’s spin axis was so important we installed a _torque coil_ in the satellite. This is a coil of wire in which, upon a signal from the ground, an electric current can be made to flow. The current produces a magnetic field that interacts with the earth’s magnetic field to change the position of the satellite’s spin axis. However, we could not be sure that this device would work properly—and this is another reason why we wanted to keep track of the exact position of the spin axis.

Ways of Measuring the Spin Axis

One of the devices built into Telstar is a set of six _solar aspect cells_ spaced at regular intervals around the satellite. These give a fairly accurate indication of the angle between the spin axis of the satellite and a line joining the satellite and the sun. When sunlight strikes these solar cells, they produce electric currents, and the value of the current from each cell is sent back to the ground via telemetry. Three of the cells are in the satellite’s northern hemisphere; three are in the southern hemisphere. If Telstar’s north pole were pointing to the sun, for example, the three northern cells would record large, equal currents; those in the southern hemisphere would show zero current. But if the spin axis were perpendicular to the satellite-sun line (as we want it to be) all six cells would report equal, average-sized currents, which would fluctuate as the satellite spun around. The solar cells were carefully calibrated before Telstar was launched, and we estimate that they can tell us the angle between the satellite’s spin axis and the satellite-sun line to within one or two degrees.

However, this one angle is not enough to locate the spin axis exactly. As you can see in _Diagram 1_, there are many possible positions for the spin axis OP that have the same angle θ with the satellite-sun line OS. These positions all would lie on the surface of an imaginary cone OPP′ that has OS as its axis and 2θ as its vertex angle. We need to have a second measurement to find a single position for the spin axis. As late as November 1961 we had not found a satisfactory way to make such a second measurement. Then Donald Gibble of Bell Telephone Laboratories suggested that we observe the reflections of sunlight from mirrors fitted onto the satellite[2].

Only when a satellite is in the right position can you see the reflection of sunlight from a plane surface on its body. _Diagram 2_ shows how flashes of reflected light are observed. The light of the sun, S, is reflected from a plane surface, R, on the satellite to our observing station, T, on the earth. If we imagine the line ORB drawn perpendicular to R, we know, from the law of reflection, that the angle of incidence, _i_, made by the sunlight to this line will be equal to the angle of reflection, _i_′, between the reflected light and the same line. The law of reflection also tells us that the sun, the line ORB, and the observing station all must now lie in the same plane. And, since we can calculate where the satellite is in its orbit at this exact moment, we can locate line ORB. But what about the spin axis? We know where on the satellite our reflector R is located, so we know ahead of time what the angle θ′ between ORB and the spin axis, OP, will be. We call it the _flash angle_. Thus we can tell that the spin axis will be somewhere on the surface of an imaginary cone OPP″ that has ORB as its axis and 2θ′ as its vertex angle[3].

In _Diagram 3_ we have combined our two measurements of the satellite’s spin axis. You can see that the two cones will intersect along two straight lines, OP and OQ; these are thus the only possible positions that will satisfy both our measurements. Actually, of course, only one of these lines is the true location of the spin axis. And it is usually obvious which one it is, when we consider all our other data about the satellite’s position.

Using this technique, if we measure the exact times when we see flashes of reflected sunlight from Telstar, we can combine that information with data from our six solar aspect cells and get a good plot of the position of the satellite’s spin axis.

In theory, this looked like a very promising idea. But finding a satisfactory way to put it into practice was something else again. Our first thought was simply to make use of the light reflected from the sapphire covers on the satellite’s solar cells. However, these covers have a low coefficient of reflection and do not form a completely flat surface. This means that the light reflected from them is very much reduced in intensity and spreads out too much to give us the precise readings we want. On the other hand, if we attached a plane mirror with a high reflection coefficient to the satellite, we thought we could pick up the minute flashes of reflected light from a distance of as much as a few thousand miles. So we decided to press ahead with this scheme and install one or more reflectors on the satellite.

By the time we started work on the mirrors, the final design of Telstar I was almost complete; this meant that we had to squeeze our mirrors aboard it as best we could. The most stringent physical requirement in designing them was weight; they could not add more than half a pound to Telstar’s total load. Nor could they project more than one-eighth inch from the satellite’s surface, or they might interfere with the radiation pattern for the main antenna. We also decided to make the mirrors out of highly polished metal, since any other possible material might break too easily. And the mirrors had to be as flat as possible, so the beam of reflected sunlight would not diverge by more than one degree.

Thus we had to design mirrors that would be very thin, very shiny, very flat, very light, and almost unbreakable. After much experimenting, we solved this rather tricky problem. The mirrors we added onto Telstar I, as you can see in _the illustration below_, were machined from aluminum alloy sheet, carefully polished by hand with abrasive papers, and buffed on a cloth wheel. Finally, we evaporated a thin layer of pure aluminum onto their surfaces to improve their reflection coefficients and make them resistant to corrosion. The three mirrors were fastened to the surface of the satellite with small screws, which had to be tightened and shimmed very carefully so that the mirrors stayed as flat as possible.

Locating the Mirrors on the Satellite

As we mentioned above, the flash angle θ′ between the satellite’s spin axis and a line perpendicular to the mirror is very important in our calculations. We made detailed studies of the various flash angles that would be possible during the first 60 days after launch. We plotted the times when the satellite would be above the horizon while our Crawford’s Hill, New Jersey, observing station was in darkness, and we made allowance for satellite orbits that might deviate slightly from the planned one. These calculations told us that the best flash angle for the mirror would be 68 degrees—which is the angle made by the first facets above Telstar’s equatorial antenna band. So we located a flat mirror on one of these facets. Because one of the solar aspect cells was already installed in the center of this facet, we were forced to cut a circular hole out of the center of the mirror.

But we knew that one mirror could not do the whole job. After Telstar I had been in orbit more than 30 days, the 68-degree mirror would only be in position to give infrequent flashes, and one at about 95 degrees would be more useful. This presented two problems. First, no facet on the satellite makes a 95-degree angle with the spin axis. However, we could use one of the facets just below the equatorial antenna, which makes a 112-degree angle, and groove or _facet_ the mirror so that its reflecting faces became narrow strips slanted 17 degrees away from the base at the angle of 95 degrees (112 - 17 = 95). Our second problem was space—since there was not enough room left on any of the 112-degree facets to mount a second large mirror, we substituted two smaller mirrors and mounted them 120 degrees apart. This arrangement lets us know from which mirror we see flashes—the plane mirror gives one flash for each revolution of the satellite; the faceted mirrors give two flashes for each revolution of the satellite.

How We Record Flashes from the Mirrors

Now we had finally found a satisfactory way to reflect a train of tiny flashes—much too faint to be seen by the naked eye—from Telstar as it passed across the sky during the night. But our main aim was to record the exact times when these flash bursts occurred. With this information, we could, using the method we described above, tell very accurately both the satellite’s spin axis and its rate of spin. We do not have space to describe the many problems that had to be solved in setting up the equipment to record the flashes. Let us merely outline the procedure that we finally devised:

1. _To pick up the satellite’s flashes we use a 12-inch-aperture photoelectric telescope mounted on a radar trailer (shown in illustration below). It is pointed by means of prediction drive tapes produced by an electronic computer; these are based on data from previous passes._

2. _On clear, dark nights when the satellite is at relatively short range, we can see it with an auxiliary finder telescope, and then adjust the large telescope precisely. Or, if the satellite’s high-frequency beacon has been turned on, the Holmdel microwave antenna can automatically point our large telescope._

3. _When flashes of light are picked up by the telescope, they fall directly onto the cathode of a photomultiplier tube. They are then filtered out from the random light in the night sky and amplified._

4. _Rather than make a continuous recording of the output—one night this would have produced a record twelve miles long for us to pore over—we use an electronic trigger. This is the time base of an oscilloscope, whose sawtooth output is set in operation only if a signal of four volts or more is received (photo below)._

5. _A pen recorder makes a continuous line on a revolving drum, with a heated stylus connected to a galvanometer marking a permanent record on heat-sensitive paper. Any signal output from the oscilloscope is picked up by the galvanometer and causes the pen to make a sawtoothed mark; when the paper is unrolled from the drum, these marks are clearly visible as notches in a series of otherwise straight lines._

6. _A synchronous timer marks the chart every ten seconds, and we are able to time individual pulses with a precision of one tenth of a second. Because the beginning and end of a train of pulses are not always distinct, we can only determine the center of a burst of flashes—which we use as our most important time indication—to within two seconds. However, this is accurate enough, for a change of only one degree in the orientation of the satellite’s spin axis would change the time of the flash burst center by about half a minute (see below)._

7. _We use a second oscilloscope to check on whether the signals we receive are genuine flashes or just accidental stray light. This oscilloscope has a long-persistence screen, which we use as a temporary memory. The pulses traced on its cathode ray tube are automatically photographed by a 35-mm camera while they persist on the screen. We can then examine the photograph to see if the pulses are genuine, which we ascertain from (a) their shape and size and (b) the intervals between successive pulses. Looking at the photographic record also confirms whether we are observing flashes from the 68° plane mirror or the 95° faceted mirrors. We can calculate the satellite’s spin rate by measuring the intervals between individual flashes._

Results

Telstar I was launched on July 10, 1962. That evening, beginning on the satellite’s seventh pass, we were able to detect trains of flashes from the mirrors. We assumed that, since Telstar had been launched almost exactly according to plan, its spin axis would be perpendicular to the plane of the earth’s orbit, and we calculated when we should see the flashes. And, each time, we actually saw them within two minutes of the times we had predicted—so we knew that the spin axis was almost exactly where it should be.

Our measurements have continued whenever the weather and other conditions permitted. Combining readings from the bursts of flashes and telemetry data from the solar aspect cells, we have accurately plotted Telstar’s spin axis; it has continued to precess very much as we predicted it would. We have also seen what happens to the spin axis when the satellite’s torque coil is turned on. And, by measuring the intervals between flashes, we have made very precise measurements of the spin rate, which is gradually decreasing mostly according to schedule. However, the plot is showing some small unexplained variations of spin decay rate, and a study of them will, we hope, throw light on some of the variations of the earth’s magnetic field.

For future communications work, particularly with satellites at longer ranges, it would seem to be preferable to use stiffer, flatter mirrors and to make them from beryllium rather than aluminum alloy. More accurate tracking means, more observatory sites, and more powerful telescopes will also be needed. But for this first experimental use our little mirrors have worked very well.

Jeofry S. Courtney-Pratt _was born in Hobart, Tasmania, Australia, and
received a Bachelor of Engineering degree from the University of
Tasmania in 1942 and a Ph.D. from Cambridge University in 1949. He was
also awarded an Sc.D. by Cambridge in 1958. He joined Bell Telephone
Laboratories in 1958, and has done research in high-speed photography,
optics, optical masers, the properties of materials, and the physics
of the contact of solids._

CASE HISTORY NO. 4
How Do We Keep Solar Cell Power Plants Working in Space?

Kenneth D. Smith
_Electronics Engineer—Member of Staff, Semiconductor Device Laboratory_

THE PROBLEM

_Before we learned about the Van Allen belts, we expected that the solar cells used to power satellites would last for many years in space. We thought they would be damaged only by cosmic rays, micrometeorites, and occasional bursts of particles from the sun. But when the solar plants of some American satellites went out of action after only a few weeks in orbit, we realized that in the future solar cell power units would need better protection from radiation damage. We had learned that satellites—and particularly medium altitude communications satellites—must spend a lot of time in regions where they will be struck by thousands or even millions of high-speed radiation particles each second._

_This fact forced us to change almost all our thinking about solar power plants for satellites. To make sure they would last for several years, we had to design new types of solar cells and devise new ways of mounting them. We also had to revise our estimates of how much power we could expect to get from our cells._

_If a communications satellite is to go into regular commercial service, it must continue working for several years in space. The Telstar satellite, however, was designed as an experimental project, and we decided that two years would be a reasonable lifetime to plan for. When Project Telstar began, our problem was to develop solar cells that would operate in an environment subject to strong radiation effects—and keep on operating there for two years._

Organizing the Work

Our work on the solar cells for Telstar I began in October, 1960. With just a little more than a year to go before the satellite had to be ready, there was no time to lose. So we decided to break down the over-all problem into three parts:

—Finding out how radiation would affect various kinds of solar cells;

—Making experimental cells and, when the best had been picked,
determining the best ways to make them in the large quantities we
would need; and

—Developing ways to mount the cells on the Telstar satellite so that
they would withstand the stresses of being launched, the effects of
radiation particles, and extreme changes in temperature.

A different group of people began work simultaneously on each of these three parts of the problem, with each of them going ahead under the assumption that the others would be successful. Each group had to find the answers to many very interesting questions, but since our space is limited we can only discuss some of them here. Before doing so, however, we must say something about what a solar cell is and how it works.

Technical Background on Solar Cells

There are two ways of making a silicon solar cell. In one, the body of the cell is what we call _n-type_ silicon—that is, pure silicon that has been doped with a small number of impurity atoms of an element such as phosphorus or arsenic (from group V of the periodic table). This kind of semiconductor[4] conducts electricity by means of a supply of free-to-move electrons (negative charges) caused by the presence of these impurity atoms. To make a workable solar cell from n-type silicon, a thin surface layer of p-type silicon is formed by diffusing atoms of a material from group III of the periodic table—usually boron—into the silicon. Metallic contacts then are made to these two regions. This kind of cell is known as a _p-on-n cell_.

The second type of solar cell is just the reverse. It begins with a body of p-type silicon (with impurity atoms from a group III element) and conducts electricity by means of “holes”—vacant sites where electrons might be but are not. These holes act as free-to-move positive charges. We can make a solar cell from this material by diffusing a layer of n-type impurity, such as phosphorus, into it. We call this an _n-on-p cell_ (see the _figure below_).

titanium-silver evaporated contact with solder dip finish
antireflection coating
contact gridding for lower series resistance
15 mil wafer, p-type
0.4 micron front layer, n-type

The key to the operation of either type of solar cell is the junction between the regions of n-type and p-type material—what we call the _p-n junction_. In an actual n-on-p cell this junction is only about twenty millionths of an inch below the surface, since that is the thickness of the n-layer. At this point, where the hole-rich p-region meets the electron-rich n-region, there is a permanent, built-in electric field. As shown in the figure below, the n-layer has many free electrons (indicated by minus signs) and a few holes (circled pluses), while the p-region has many holes and a few electrons. When the cell is in equilibrium, thermal agitation causes some holes to diffuse into the p-region. We call these stray holes and electrons _minority carriers_ (the circled pluses and minuses in the figure). Thus, the n-layer has a slight positive charge and the p-body has a slight negative charge; this results in a difference in potential across the junction, which in silicon amounts to about seven-tenths of a volt.

Sunlight is made up of individual corpuscles of energy called _photons_. When these photons are absorbed in or near a cell’s p-n junction, they liberate both a free-to-move negative charge and a free-to-move positive charge—this is called generating a _hole-electron pair_. The electric field across the p-n junction causes the holes to flow to the p-side and the electrons to the n-side of the barrier. This flow tends to make the p-side positive and the n-side negative, so that, when a load is connected between them, a useful external voltage (amounting to about six-tenths of a volt) is produced, and electric current will flow. Thus, we have converted light energy into electrical energy.

Only part of the energy in light can be used to generate an electrical output, since a good deal of the light striking a cell is absorbed as heat or is reflected from its surface. The percentage of solar energy that can be converted into usable electric power is called the cell’s _conversion factor_ or _efficiency_. Although this can theoretically be as high as 22%, the best cells we have made in the laboratory have conversion factors of only about 15%, and the better commercial cells have efficiencies of 12% or more.

Although both p-on-n and n-on-p cells were made in early laboratory studies, the p-on-n cells gave a somewhat higher output. As a result, all the American commercial solar cells up to 1960 were of this type, and they were used on all satellites before Telstar I. (Russian satellites, we believe, have used n-on-p cells from the beginning.)

The U.S. Army Signal Corps Research and Development Laboratory, however, decided to make both p-on-n and n-on-p cells and compare their performance. This laboratory work led to a surprising discovery: The n-on-p cells were several times as resistant to energetic particle radiation as were comparable p-on-n cells. These results were announced in 1960, and confirmed by our measurements and those of other laboratories. The timing was very fortunate, since we had just learned of the greatly increased radiation hazards presented by the Van Allen belts.

Finding Out About Radiation Damage

Now, having given you a very brief account of how a solar cell works, let us return to our three-part problem. The first objective was to study all the aspects of radiation damage. To do this, we had to find out how much radiation the Telstar satellite would encounter; we needed to estimate the concentration of high-energy particles—both electrons and protons—at various altitudes and locations. Several government agencies are now carrying on research in this important area, but at the time of the Telstar I launch we did not know exactly how much radiation the satellite would run into. And the high-altitude nuclear explosion of July 9, 1962 (the day before Telstar I went into orbit) may have increased the quantity of high-energy electrons injected into its path.

We also wanted to find out whether electrons and protons would do the same damage to solar cells. Several kinds of cells were exposed at Bell Laboratories and at various university research laboratories to a wide range of radiation dosages. The experiments showed, generally, that the damage effects of electrons and protons should be about the same. Although protons are 1840 times as massive as electrons, there are a great many more electrons in the Van Allen belts, so that an unprotected solar cell would be much more likely to be injured by electrons than by protons.

In fact, we found that the Van Allen belt protons have so much energy that they can go through transparent shielding material as much as several centimeters thick and still damage a solar cell. Thus, to screen our cells from protons we would need very thick transparent cover plates, and this added weight would be intolerable. So we decided to use no proton shielding at all.

With electrons, the situation is different; they are much lighter and have much less energy. Also, if their energy is reduced below a certain level (about 180 thousand electron volts) electrons will not be able to knock silicon atoms out of position, and thus cannot harm a solar cell. We experimented with a number of different kinds and thicknesses of cover plates, and found that transparent material with a mass of 0.3 gram per square centimeter would slow down electrons enough to make them no problem.

Another radiation study helped us take advantage of the fact that solar cells respond differently to light of different wave lengths. If the surface layer of a cell is extremely thin, it will absorb blue, green, and yellow light well, but may be much less sensitive to the deeply penetrating red and infrared waves. We experimented with n-on-p cells having very shallow p-n junctions, exposing them to an extremely strong radiation dosage. The cells still responded very well to blue and green light, even though most of their response to infrared and red light was lost. These findings convinced us that we should work to make our new cells as blue-green sensitive as possible, since they were going to be exposed to heavy radiation.

Designing and Making the Best Solar Cells

After it was discovered that the n-on-p cell was more resistant to radiation, we decided to make an all-out effort to develop an n-on-p cell that could be manufactured in quantity for our new satellite. Since we didn’t know whether we could solve this problem in time to meet the Telstar I launch date, we “hedged” by designing the new n-on-p cells to be the same physical size (one by two centimeters) as conventional p-on-n cells. Thus, if the n-on-p project hit a snag, we probably could use regular p-on-n cells.

As you can imagine, making a solar cell to fit the very high requirements we had set for the Telstar satellite is not an easy job—and making these cells by the thousands is even more of a task. During October, November, and December of 1960 we carried on a crash program in which we made hundreds of experimental cells in our laboratories, using a variety of materials and many different manufacturing techniques.