The right event can still be written in the wrong axes
A satellite signal leaves one event and arrives at another. In the interval, Earth-fixed coordinate axes rotate. A range calculation cannot subtract two coordinate triples merely because both are labelled Earth-fixed; it must establish that their axes refer to a compatible orientation. The satellite emission state and receiver reception state need one common coordinate description.
This guide isolates that coordinate problem. It explains why rotating the representation of an earlier satellite event is different from moving the satellite to a later event. Small original examples then show why a sizeable coordinate change can have a much smaller range projection, and why the sign must be checked from declared conventions.
All coordinates, angles and tolerances used below are fictional teaching inputs. They are not operational Earth-orientation parameters or survey corrections. The aim is a transferable habit: make events, frames and transformations explicit before interpreting the difference between two numbers.
Geodesy subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1: Identify events and rotating axes
Separate the physical point from its coordinate representation
Route 2: Work a rotation and range
Route 3: Test signs and approximations
Route 4: Check model interfaces
Route 5: Transfer and report
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Identify events and rotating axes
Separate the physical point from its coordinate representation
Define the passive rotation used in this guide
Work a rotation and range
Rotate a fictional emission state and preserve its length
Calculate a range before and after resolving the mismatch
Derive the first-order range projection
Test signs and approximations
Reverse geometry to test the sign
Examine a zero first-order correction without declaring no effect
Link rotation angle to flight time without mixing clocks
CHAPTER 1 OF 12
1. Separate the physical point from its coordinate representation
A coordinate triple is a description of a point or event relative to specified axes. Rotating those axes changes the numbers without moving the represented event. This is a different operation from asking where a moving object will be later. Confusing the two creates a particularly subtle error because both operations can produce new triples with the same units.
Imagine marking a point on a sheet and laying transparent axes over it. Rotate the axes while keeping the point fixed on the sheet. Its coordinates change because the representation changed. Now keep the axes fixed and move the point instead. Its coordinates also change, but for a different reason. The same numerical pattern does not establish the same physical operation.
For a GNSS range model, name satellite emission event E and receiver reception event R. An orbit model supplies a satellite state at E. The receiver endpoint belongs to R. Their separation needs coordinates expressed in a common reference description. A transformation between axis orientations can be necessary even after the two correct physical events have been selected.
The wider Geodesy root explains reference systems and changing Earth orientation. Here we narrow the task to a single signal flight. This is not the annual movement of a station through a coordinate epoch, nor the construction of a global frame. The timescale and purpose of the transformation must remain attached to the calculation.
CHAPTER 2 OF 12
2. Define the passive rotation used in this guide
Choose a two-dimensional horizontal illustration with z unchanged. Define R(θ) acting on (x,y) by x′ = x cosθ + y sinθ and y′ = −x sinθ + y cosθ. This is the passive rotation convention used throughout our examples. With a small positive θ, a point initially on the positive x-axis receives a negative y-coordinate in the rotated representation.
ESA Navipedia uses this matrix form to express a satellite emission state in Earth-fixed axes associated with reception, with angle related to Earth rotation during travel time. The source distinguishes the emission-state calculation from this coordinate transformation. Our numerical angles are intentionally enlarged to make the geometry visible, not proposed values for a real signal flight.
A different software interface might use an active rotation matrix or the inverse direction. That does not automatically make it incorrect. It means its input and output conventions differ. The full statement is that R converts this named representation into that named representation. A bare matrix with an angle but no direction is incomplete documentation.
A simple sign test uses (1,0) and a positive right angle. Our matrix produces (0,−1). Its inverse R(−θ) returns the original vector. Use that known synthetic test before interpreting a complicated orbital result. If the sign test fails, additional digits from a realistic ephemeris will not repair the convention mismatch.
CHAPTER 3 OF 12
3. Rotate a fictional emission state and preserve its length
Let the satellite emission coordinates in one representation be s = (10,0) distance units. Choose θ = 0.1 radians for the classroom calculation. Using cos0.1 ≈ 0.995004165 and sin0.1 ≈ 0.099833417 gives s′ ≈ (9.95004165,−0.99833417). The event represented is still the earlier emission event.
Its distance from the common origin remains ten units: the square root of x′² + y′² equals ten, up to rounding. This follows from cos²θ + sin²θ = 1. A rotation preserves vector length. If a purported pure rotation substantially changes that length, inspect the matrix, units, numerical implementation or whether another transformation was included.
Length preservation does not mean the receiver-to-satellite range stays unchanged when only the satellite representation is transformed into the receiver’s axes. The receiver coordinates must be interpreted consistently. The operation corrects a mismatch between two descriptions; it is not a physical rotation of one endpoint while the other remains attached to the same original axes.
Do not rotate a position twice because two routines both mention Earth rotation. An orbit-state provider may already return the representation expected by the range model. Document the output reference and the applied transformations. A correct matrix applied to an already transformed vector introduces a new error while passing its own length-preservation test.
CHAPTER 4 OF 12
4. Calculate a range before and after resolving the mismatch
Set a fictional receiver reception coordinate r = (0,2) in the target axes. If the untransformed satellite triple (10,0) is incorrectly treated as though it already used those axes, the computed range is √(10² + (0−2)²) = √104 ≈ 10.198039 distance units. This is the deliberately mismatched calculation.
Using the transformed emission triple gives ρ = √[(9.95004165)² + (−0.99833417−2)²]. Equivalently, because the satellite origin distance remains ten, ρ² = 100 + 4 − 2r·s′ = 104 + 40sin0.1 ≈ 107.9933367. The resulting range is approximately 10.391984. The difference from the mismatched calculation is approximately +0.193945 units.
The equivalence of the component calculation and the dot-product calculation is an independent arithmetic check. One uses coordinate differences directly; the other uses the angle-sensitive cross term. They should agree within rounding. If they do not, inspect which receiver coordinates and transformed satellite values entered each expression.
The transverse satellite coordinate changed by almost one distance unit, yet the range changed by only about 0.194 units. That contrast is not a contradiction. Range responds to the endpoint change along the line of sight, with higher-order geometric terms. A coordinate correction magnitude cannot automatically be reported as an equal observation or receiver-position correction.
CHAPTER 5 OF 12
5. Derive the first-order range projection
For a small coordinate change δs, write the original range as ρ₀ = |s − r| and its unit line-of-sight vector from receiver to satellite as u = (s − r)/ρ₀. To first order, the range change is δρ ≈ u·δs. This derivative follows by differentiating the squared distance, not by assuming that every component of the endpoint change acts along the range.
Under our passive rotation, the first-order satellite change is δs ≈ (θy,−θx). Substitute it into the projection. Since u = (x−rₓ,y−rᵧ)/ρ₀, the numerator becomes θ[(x−rₓ)y − (y−rᵧ)x] = θ(xrᵧ − yrₓ). Thus δρ ≈ θ(xrᵧ − yrₓ)/ρ₀ under these declared conventions.
For s = (10,0), r = (0,2) and θ = 0.1, this gives δρ ≈ 0.1 × 20/√104 ≈ 0.196116 units. The exact value was approximately 0.193945. The difference is about 0.002171 units. The approximation captures the main effect but does not equal the exact answer at this intentionally large teaching angle.
The equation is useful because its assumptions are visible. It uses a small rotation, a chosen sign convention and a first-order expansion about a specified geometry. It does not convert a range correction into a complete position solution. For that, the effects across all observations and estimated parameters must be considered together.
CHAPTER 6 OF 12
6. Reverse geometry to test the sign
Keep s = (10,0) and θ = 0.1, but move the fictional receiver to r = (0,−2). The original mismatched range is still √104. After the same transformation, the squared range becomes 104 − 40sin0.1, not 104 + 40sin0.1. The correction changes sign because the receiver lies on the other side of the x-axis.
The first-order expression confirms this immediately: xrᵧ − yrₓ = 10 × (−2) − 0 = −20, so δρ ≈ −0.196116. The absolute rotation angle and satellite radius are unchanged. Geometry determines whether the range grows or shrinks. A universal instruction to add a positive rotation correction would fail this paired test.
Now reverse θ instead while returning the receiver to (0,2). The first-order correction also reverses. Reversing both geometry and angle restores its original first-order sign. These paired cases help reveal an implementation that takes an absolute value too early or uses an active matrix with a passive convention.
For exact calculations at finite angles, some even-order terms do not simply reverse sign. Therefore distinguish the sign symmetry of the first-order contribution from an assertion that every exact finite-angle range difference is its perfect negative. The educational test should state whether it checks the linearised model or the full trigonometric calculation.
CHAPTER 7 OF 12
7. Examine a zero first-order correction without declaring no effect
Take fictional s = (10,0) and r = (2,0). The untransformed separation is eight units. Our first-order expression gives xrᵧ − yrₓ = 0, so the linear term vanishes. That does not prove the exact finite-angle correction is zero. It says this geometry has no first-order range sensitivity to the small rotation.
At θ = 0.1, the transformed squared range is 100 + 4 − 40cos0.1 ≈ 64.1998334. Its square root is approximately 8.012480. The range increase is about 0.012480 units, arising at higher order. The satellite coordinate changed substantially in the teaching picture, but the first-order range projection is zero because the initial displacement is perpendicular to the original line of sight.
Reduce the angle and the higher-order effect shrinks rapidly. Expanding cosθ ≈ 1 − θ²/2 gives ρ² ≈ 64 + 20θ² for this geometry. Expanding the square root about eight gives δρ ≈ 1.25θ². At θ = 0.1 this predicts 0.0125, close to the exact teaching value.
This is a useful changed-context lesson. A zero derivative means local first-order insensitivity, not necessarily exact invariance. Similar reasoning appears in optimisation, calibration and error propagation. Before concluding that a correction can be ignored, compare an appropriate bound on omitted terms with the actual requirement for the quantity being estimated.
CHAPTER 8 OF 12
8. Link rotation angle to flight time without mixing clocks
In the simplified constant-rotation description, θ = ωτ, where ω is the adopted angular rotation rate and τ is the physical travel interval in a compatible time account. A pseudorange divided by signal speed can contain clock-offset contributions and other effects. It should not silently become a geometric flight interval merely because both quantities are expressed in seconds.
Use fictional ω = 0.01 radians per time unit and τ = 3 time units. Then θ = 0.03 radians. If a clock-label interval of 3.2 were substituted without justification, the angle would be 0.032. The difference of 0.002 radians is a timing-input error propagated into the representation transformation. No real Earth-rotation parameter is implied by these values.
For a small flight-time perturbation δτ, the angle perturbation is ωδτ in this simple model. The induced satellite coordinate perturbation is approximately ωδτ(y,−x), before its range projection is taken. This chain makes the units and sensitivities explicit: time to angle, angle to coordinate, coordinate to range.
The emission-time companion guide develops clock labels and geometric iteration separately. Use it when the uncertain input is which earlier event was recovered. Here, assume that event has been identified and check how its state is represented. Keeping the two interfaces separate prevents an Earth-rotation correction from being asked to repair a wrong satellite epoch.
CHAPTER 9 OF 12
9. Do not propagate the satellite to reception as a shortcut
Suppose a fictional satellite has one state at emission and another at reception. Propagating its orbit forward changes the physical event being represented. Rotating its emission-state coordinates into reception-associated axes changes only the representation of the earlier event. The operations may both involve time, but they answer different questions.
A classroom counterexample makes this visible. Let a point remain fixed in an inertial drawing while the coordinate axes turn. Its coordinates in Earth-like rotating axes change despite zero physical velocity in the drawing. Conversely, let a point move while the axes remain fixed. A propagation accounts for its actual motion even without any coordinate rotation. Neither operation replaces the other.
A complete processing sequence should therefore name both the orbit-state epoch and the axes used for the output. If a function called reception coordinates returns an emission event expressed in reception axes, the name may be reasonable but needs documentation. If it returns the satellite physical state at reception, it cannot be substituted into the same light-time geometry merely because the words sound similar.
The practical audit is to trace one synthetic signal through each function. Label the event before and after every call. A pure representation change should preserve the event identifier. An orbit propagation should change it deliberately. This interface test often exposes conceptual mistakes that ordinary checks of array size, units and numerical convergence cannot detect.
CHAPTER 10 OF 12
10. Keep the GNSS coordinate correction distinct from a gyroscope experiment
The Sagnac effect is a broader rotation-related propagation phenomenon. eduKate’s existing Sagnac guide explains counter-propagating waves around a closed interferometer and the use of phase differences to measure rotation. That is a different learning task from constructing a consistent one-way GNSS endpoint range in rotating coordinates.
The connection is useful, but the experimental arrangements should not be collapsed into one formula. A closed-loop interferometer has its own area, orientation and wave-path relationships. The coordinate examples here use an emission endpoint, a reception endpoint and a stated rotation convention. Importing a loop-area formula into these examples would replace the model rather than complete it.
Likewise, this coordinate transformation alone is not a full relativistic GNSS timing account or a measurement of frame dragging. Real precision modelling includes additional effects under defined conventions. Stating that limit preserves the value of the calculation: it resolves a particular representation mismatch without claiming to explain every timing or gravitational contribution.
A reader moving between the two articles should ask what was observed in each arrangement, which reference was adopted and which physical inference follows. Shared terminology can provide a bridge, but a bridge connects distinct places. It should not erase the differences in observables, equations and evidence that give each explanation its meaning.
CHAPTER 11 OF 12
11. Test invariants, inverses and independent range expressions
A robust synthetic check uses more than one property. First verify norm preservation: |R(θ)s| = |s|. Second verify inverse consistency: R(−θ)R(θ)s = s. Third verify that rotating both endpoint representations by the same orthogonal matrix preserves their separation. These test matrix behaviour independently of a particular satellite geometry.
The third property is especially important. If both s and r already use the same axes, a common rotation gives |Rs − Rr| = |R(s − r)| = |s − r|. A changed range in that test suggests an implementation problem. Our earlier range change occurred because an initial mismatch was being resolved, not because common coordinate rotations alter physical distances.
Then compare the direct component range with the dot-product expansion |s′ − r|² = |s′|² + |r|² − 2s′·r. Use a non-axis-aligned vector as well as the simple examples. Include positive and negative angles and a zero-angle identity case. A routine that passes only one hand-picked geometry may still have a swapped component or sign.
Finally compare exact trigonometric and first-order results as the angle is reduced. The disagreement should behave consistently with the omitted higher-order terms for the tested domain. These are numerical and representation checks. They do not independently validate an orbit product, clock model or receiver observation, which remain separate parts of the evidence chain.
CHAPTER 12 OF 12
12. Report the transformed state and the surviving uncertainty
A useful conclusion specifies that the satellite state at emission was expressed in axes associated with reception using the declared transformation. It states whether range was computed exactly within the model or with a first-order approximation. It does not describe the rotated coordinate difference as physical satellite movement during the signal flight.
For a real processing audit, preserve the time scale, emission and reception labels, orbit-state reference, transformation direction and any already applied correction. Record the input used for flight time and its relationship to clocks and geometry. These details make it possible to discover a duplicate rotation or wrong-event substitution after the final coordinate has been produced.
If a corrected solution agrees better with a suitable independent reference, that supports the combined model under the comparison conditions. It does not make every residual a rotation effect or establish universal future accuracy. Other observation terms and the independent reference have their own uncertainties. Keep the claim proportionate to what was actually tested.
The transferable lesson is that numbers can change because the world moved, because the reference moved or because the representation changed. A trustworthy explanation identifies which operation occurred. GNSS makes this distinction unusually consequential, but the underlying habit applies whenever data cross coordinate systems, clocks or software interfaces.
Independent worked case and evidence checks
Use fictional satellite emission vector s = (6,8), receiver reception vector r = (1,−2) and angle θ = 0.02 radians under the declared passive convention. The original mismatched squared range is 5² + 10² = 125. The first-order numerator xrᵧ − yrₓ is 6 × (−2) − 8 × 1 = −20, so δρ ≈ −0.4/√125 ≈ −0.035777 distance units.
Calculate the transformed coordinates independently: x′ = 6cos0.02 + 8sin0.02 ≈ 6.15878937 and y′ = −6sin0.02 + 8cos0.02 ≈ 7.87840805. The satellite origin distance remains ten. The corrected squared range can also be written 105 + 20cos0.02 − 40sin0.02 ≈ 124.1960535. Its square root is approximately 11.1443283, compared with √125 ≈ 11.1803399.
The exact change is approximately −0.0360116, close to but not equal to the first-order prediction. Explain why the sign is negative before using a calculator. Then halve θ and recompute both forms. The approximation improves as the deliberately small-angle conditions become stronger, though the required tolerance must still be defined.
As a final check, rotate both correctly compatible endpoint vectors by the same additional angle. Their separation must remain unchanged within rounding. If it changes, investigate the implementation rather than claiming that a physical range depends on the arbitrary orientation chosen for a common set of axes.
There is another useful invariant: composition. For the declared planar convention, R(α)R(β) = R(α + β). Transform the emission vector first by 0.012 radians and then by 0.008. Compare the result with a single 0.020-radian transformation. The agreement checks both matrix multiplication order in this planar setting and the way the interface carries its angle units. General three-dimensional rotations about different axes do not share this simple interchangeable order.
Now deliberately enter 0.020 as degrees in one routine and radians in another. Both may preserve vector length perfectly while returning different directions. Norm preservation alone will not detect the unit mismatch. A known-angle synthetic test and the expected range projection provide additional checks. This illustrates why multiple independent invariants are more informative than repeatedly checking the same property.
For a quantitative tolerance exercise, suppose the teaching task allows at most 0.0001 distance units of range approximation error. The first-order result at θ = 0.020 differs from the exact result by about 0.000235, so it does not meet that invented requirement. Calling the angle small is not sufficient. Either retain the exact trigonometric form or justify another approximation whose error fits the stated purpose. This tolerance is fictional and must not be read as a survey standard.
Check units before evaluating the sine and cosine. A matrix can be mathematically correct while its input angle uses a different unit from the function’s expectation. Record radians or degrees explicitly at the interface. This simple check complements, but does not replace, the event and reference checks developed above.
Worked comparison
| Operation | What changes | What must remain clear |
|---|---|---|
| Select emission epoch | The satellite physical event used | Time scale and light-time model |
| Transform emission-state coordinates | The axes describing the same event | Direction and sign convention |
| Rotate both compatible endpoints | Their common representation | Their separation is invariant |
| Use a first-order correction | The approximation to a range change | Neglected terms and tolerance |
| Estimate receiver position | Parameters inferred from many observations | Geometry and complete observation uncertainty |
Questions and practical checks
Why are two Earth-fixed triples not automatically compatible?
Their axis orientations may be associated with different epochs in a signal-flight calculation. Establish the representation convention before subtracting them.
Does rotating a vector mean the satellite moved?
Not in a pure coordinate transformation. The same earlier event receives another numerical representation. Orbit propagation to a later event is a different operation.
Does a metre of coordinate correction mean a metre of range correction?
No. First-order range change is the projection along the relevant line of sight. Geometry and higher-order terms determine the relation.
Can a zero first-order term prove there is no exact effect?
No. The aligned example has a vanishing linear term but a nonzero finite-angle range change. Check omitted terms against the intended tolerance.
Is this the same calculation as a closed-loop Sagnac interferometer?
No. They connect through rotation-related propagation, but the observables and geometrical arrangements differ. Use the existing Sagnac guide for the closed-loop experiment.
Continue with the connected subjects
Establish the wider references
Connect Earth geometry, rotation and reference systems
Build the GPS range and receiver-clock estimation model
Separate coordinate epochs and frame-motion questions
Compare the related but different rotation experiment
Study the Sagnac effect in a closed rotating interferometer
Keep observation epochs and signal identities traceable
Sources and scope
ESA Navipedia: satellite emission states expressed in reception-associated Earth-fixed axes
ESA Navipedia: the separate task of recovering signal emission time
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence. Sources were checked on 31 October 2026.
Continue through the observation model
Follow the next guide that matches the unresolved part of the observation. These companions address different model terms; none replaces an independent accuracy check.
Orientation and phase: distinguish wind-up, ambiguity and observation continuity
Emission timing: recover the transmitting event before evaluating its state
Height and atmosphere: test which elevation signatures separate the fitted parameters