1 From six Cartesian coordinates to six geometric coordinates

The inertial Cartesian state is

(1)

It contains six independent scalars. For ideal two-body motion, the same physical state may be represented by

(2)

The two descriptions answer different questions. Cartesian coordinates say where the satellite is and how fast it is moving in a chosen frame. The classical orbital elements separate the same information into orbit size, orbit shape, plane orientation, orientation within the plane, and current phase.

A useful interpretation is

GPSORB02 established the need for six constants. GPSORB03 derived angular momentum. GPSORB04 derived mechanical energy and eccentricity. These quantities now yield the classical orbital elements.

2 Required auxiliary quantities

Given r and v, begin with their magnitudes

(3)

The specific angular-momentum vector is

(4)

with magnitude

(5)

The specific mechanical energy is

(6)

The eccentricity vector is

(7)

and its magnitude is

(8)

The inertial reference plane is assumed to be the x-y plane, with unit normal

(9)

The node vector is defined by

(10)

Because n lies in both the inertial reference plane and the orbital plane, it points along their intersection: the line of nodes. For the usual prograde geometry, n points toward the ascending node.

In components,

(11)

3 Semimajor axis from the specific mechanical energy

GPSORB04 derived

(12)

for a nonparabolic Kepler orbit. Therefore

(13)

For an ellipse, ℰ < 0 and therefore a > 0. For a hyperbola, ℰ > 0 and the conventional semimajor axis is negative. At the parabolic limit,

(14)

so |a|→∞. The semilatus rectum

(15)

remains finite and is often a better size parameter near the parabolic limit.

The first element therefore comes directly from energy: no angle extraction is required.

4 Eccentricity from the eccentricity vector

The second element is simply the magnitude of Eq. (7):

(16)

The magnitude determines the conic type,

(17)

(18)

(19)

The direction of e is equally important: for e≠0, it points toward periapsis. That direction will be used to define ω and ν.

5 Why atan2 is preferable to inverse cosine

Many orbital-mechanics derivations introduce angles with formulas such as

(20)

This recovers an angle only in the interval 0 ≤ α ≤ π. The cosine does not tell us on which side of the reference direction the target vector lies. A second sign test must therefore be added by hand.

A more systematic method computes both a cosine-like and sine-like quantity and uses

(21)

where X is proportional to cos α and Y is proportional to sin α. The signs of both arguments preserve the quadrant.

Most numerical libraries use the argument order

(22)

MATLAB, Python, C, C++, and many other environments follow this convention. The returned interval is commonly (−π,π]. Orbital elements are often reported on [0, 2π), so a negative result may be wrapped by adding 2π.

6 Inclination from the angular-momentum vector

The orbital plane is perpendicular to h. The reference x-y plane is perpendicular to k. Therefore the inclination is the angle between h and k.

The usual cosine relation is

(23)

Instead of using only arccos, form the sine magnitude from

(24)

Then

(25)

This naturally returns

(26)

Thus prograde orbits satisfy 0 ≤ i < π∕2, polar orbits have i = π∕2, and retrograde orbits satisfy π∕2 < i ≤ π.

7 Right ascension of the ascending node

The node vector

(27)

lies in the reference plane and points toward the ascending-node direction. Since

(28)

the right ascension of the ascending node is simply its polar angle measured from the positive inertial x axis:

(29)

If the result is negative, wrap it to [0, 2π).

Using Eq. (11), one could equivalently write

(30)

but computing n explicitly makes the geometry clearer and is less error-prone.

The formula requires

(31)

When n = 0, the orbit is equatorial and there is no unique ascending node. That singular case is treated later.

8 Argument of periapsis

The argument of periapsis is the oriented angle within the orbital plane from the ascending-node direction n to the eccentricity vector e.

The cosine is

(32)

To obtain the sign of the sine, note that h is the positive normal to the orbital plane. Therefore

(33)

Combining the two with atan2 gives

(34)

The common positive scale factor ne cancels between the two arguments, so normalization is unnecessary inside the atan2 call.

This formula requires both

(35)

If the orbit is equatorial, the node direction disappears. If the orbit is circular, the periapsis direction disappears.

9 True anomaly

The true anomaly is the oriented angle from the eccentricity vector to the instantaneous radius vector. Its cosine is

(36)

The corresponding signed sine is

(37)

Therefore

(38)

Again, the common scale factor er does not need to be included explicitly.

A useful cross-check follows from the radial-velocity result derived in GPSORB04,

(39)

Hence, for e≠0,

(40)

This makes the physical sign intuitive: before apoapsis the satellite is moving outward and ṙ > 0; after apoapsis it is moving inward and ṙ < 0.

10 The complete nonsingular state-to-elements algorithm

For a noncircular, nonequatorial orbit, the transformation can be written compactly as follows.

First compute

(41)

(42)

(43)

and

(44)

Then

(45)

for a nonparabolic orbit,

(46)

(47)

(48)

(49)

and

(50)

Wrap Ω, ω, and ν into the desired 2π interval after the call to atan2.

11 A GPS-like numerical example

To make the full transformation concrete, consider the inertial state

(51)

(52)

with

(53)

The state was chosen to be representative of a medium-Earth orbit, but the conversion procedure is completely general for a nondegenerate Kepler orbit.

The magnitudes are

(54)

The angular-momentum vector is

(55)

with

(56)

The energy is

(57)

Therefore

(58)

The eccentricity vector is

(59)

so

(60)

The node vector is

(61)

Applying the quadrant-safe formulas gives

(62)

(63)

(64)

and

(65)

Thus the recovered classical element set is

(66)

As a useful numerical check,

(67)

while

(68)

These are cos 30^∘ and sin 30^∘, respectively. Likewise, the true-anomaly formulas recover approximately cos 70^∘ and sin 70^∘.

12 Why the classical elements become singular

The Cartesian state remains perfectly well defined for circular and equatorial orbits, but some of the geometric reference directions used by the classical elements disappear. These are coordinate singularities, not physical singularities.

12.1 Circular inclined orbit

For a circular orbit,

(69)

There is no unique periapsis direction, so ω and ν are individually undefined. Their sum, however, remains meaningful. Define the argument of latitude

(70)

where the symbol u is conventionally used. Directly from the state,

(71)

(72)

so

(73)

12.2 Equatorial eccentric orbit

For a prograde equatorial orbit,

(74)

so

(75)

There is no unique ascending node, and therefore Ω and ω are individually undefined. Their sum remains meaningful as the longitude of periapsis,

(76)

For a prograde equatorial orbit it can be obtained directly from the eccentricity vector:

(77)

Retrograde equatorial motion requires careful sign conventions because the positive orbital normal is opposite the reference z direction; modified or equinoctial elements are generally preferable near that singular geometry.

12.3 Circular equatorial orbit

Now both the periapsis direction and node direction disappear. The four classical angles Ω, ω, and ν do not remain individually meaningful, but the satellite’s azimuth in the reference plane does. Define the true longitude

(78)

For a prograde circular equatorial orbit,

(79)

Again, the physical state is not singular. Only the chosen coordinates are.

13 Practical singularity thresholds

In floating-point software one should not test exact equality such as

(80)

Instead choose tolerances appropriate to the application. A generic implementation might classify the orbit using thresholds

(81)

where 𝜀_e and 𝜀_i are selected based on numerical precision and mission requirements.

This matters because the classical angles become poorly conditioned before the exact singularity is reached. If e is extremely small, tiny state-vector perturbations can rotate the computed periapsis direction dramatically even though the physical orbit has barely changed. Similarly, when i is extremely small, tiny perturbations can move the computed node direction by a large angle.

For estimation and orbit propagation, nonsingular element sets such as equinoctial or modified equinoctial elements are often preferred near these geometries. The classical elements remain extremely useful for interpretation because their geometry is intuitive.

14 Common implementation errors

Several mistakes recur in state-to-element software.

14.1 Reversing the cross product in the node vector

The adopted definition is

(82)

Using h ×k reverses the vector and shifts the node angle by π.

14.2 Using inverse cosine without a quadrant test

A formula such as

(83)

does not distinguish positive and negative n_y. The atan2 form does.

14.3 Swapping the arguments of atan2

The convention used here is

(84)

Swapping those arguments changes the angle definition.

14.4 Forgetting angle wrapping

An atan2 result of −10^∘ may represent the desired orbital angle 350^∘. The mathematical direction is the same, but software interfaces often require one specified interval.

14.5 Applying nonsingular formulas to singular geometry

If e or n is numerically near zero, formulas for ω, ν, or Ω may divide by tiny quantities or return essentially arbitrary directions. The orbit must first be classified and the appropriate replacement angle selected.

15 Connection to GPS broadcast ephemeris parameters

The classical element set derived here is not yet the full GPS broadcast model, but it explains the geometry behind several broadcast parameters. The legacy GPS ephemeris contains quantities corresponding closely to

(85)

plus rates and harmonic correction terms.

There is one important difference from the element set in Eq. (2): the navigation message uses mean anomaly at the reference epoch rather than true anomaly as its phase variable. True anomaly is geometrically natural for locating the satellite on the conic, while mean anomaly is much more convenient for time propagation because it advances linearly in the ideal Kepler problem.

The next lessons therefore move from the geometric angle ν to eccentric anomaly E and mean anomaly M, derive Kepler’s equation, and show why the broadcast ephemeris begins its time update with mean motion and mean anomaly.

16 Summary

One inertial Cartesian state

(86)

contains exactly the information required to reconstruct the six classical orbital elements.

Energy gives the orbit size,

(87)

The eccentricity vector gives the shape,

(88)

The angular-momentum vector gives the plane tilt,

(89)

The node vector gives the orientation of the orbital plane in inertial space,

(90)

The node and eccentricity vectors determine the periapsis orientation,

(91)

and the eccentricity and radius vectors determine the instantaneous position on the conic,

(92)

The classical elements are therefore not an unrelated set of geometric labels. They are a nonlinear change of coordinates on the same six-dimensional orbital state. Their circular and equatorial singularities arise because some geometric reference directions cease to exist, not because the satellite dynamics themselves become singular.

References