A phase-count record measures change over an interval, not an instantaneous distance
A ground beacon sends radio signals to a receiver aboard a satellite. The receiver compares the arriving cycles with its own oscillator. How does that record become useful geodetic information? The first step is to identify the counted quantity and its interval, not to treat a phase value as a ready-made distance or a frequency value as a complete velocity.
DORIS uses ground transmitters and satellite receivers. Its observation chain therefore needs the correct emitting and receiving events, frequency channels, oscillator conventions and antenna endpoints. This guide follows a simplified version of that chain through original examples, then shows why an ionospheric combination can still require a frequency-dependent endpoint correction.
The numerical records and antenna offsets below are fictional. They are not real DORIS data, instrument calibrations or operational processing instructions. Current specialist documentation is required for actual RINEX/DORIS fields, clock corrections, relativistic models and phase-centre conventions. The aim is to understand which quantity each calculation produces and what information it still lacks.
Geodesy subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1 · Chapters 1–2: Read the counted quantity
Route 2 · Chapters 3–5: Convert and time the interval
Route 3 · Chapters 6–8: Keep frequency endpoints consistent
Route 4 · Chapters 9–10: Audit uncertainty and continuity
Route 5 · Chapters 11–12: Transfer and report
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Read the counted quantity
Distinguish the uplink system from the record you interpret
Write a declared count model before converting units
Convert and time the interval
Work a phase difference in both arithmetic directions
Keep oscillator comparison separate from path change
Recognise a finite integration rather than a midpoint measurement
Keep frequency endpoints consistent
Normalize channels before combining their observations
Follow the frequency-dependent antenna endpoint into the combination
Difference the endpoint contribution over the actual count interval
Audit uncertainty and continuity
Propagate covariance through phase differences and channels
Check phase continuity and record changes before interpreting motion
CHAPTER 1 OF 12 · Read the counted quantity
1. Distinguish the uplink system from the record you interpret
Doppler Orbitography and Radiopositioning Integrated by Satellite, or DORIS, observes signals transmitted by ground beacons and received on satellites. AVISO’s technical description identifies nominal beacon frequencies of 2036.25 megahertz and 401.25 megahertz. These are system facts, not the deliberately rounded frequency pair used in some arithmetic examples below.
The CNES document on the GINS Doppler observation equation describes the comparison of received emitter cycles with cycles generated by the receiver oscillator. It also distinguishes proper times at the emitter and receiver from coordinate time. That is why an observation record cannot be interpreted safely from a column labelled phase without its definition.
A phase measurement, a phase difference, a Doppler count and a cycle rate are related but different quantities. An accumulated phase can include an arbitrary reference level. A difference between two compatible phase measurements removes a constant level, while retaining the change across their time interval.
Before calculating, name the satellite, beacon, channel, observation type, endpoint time tags and sign convention. Check the applicable format and processing documentation. A generic GNSS downlink equation is useful comparative background, but should not be copied into an uplink DORIS record merely because both systems use radio carriers and orbiting receivers.
CHAPTER 2 OF 12 · Read the counted quantity
2. Write a declared count model before converting units
For a transparent nonrelativistic teaching model, define P(t) as received carrier phase minus a nominal receiver-generated phase, in cycles. Assume matched nominal frequency f, common time units and no propagation or instrument terms for the moment. Let positive Δρ mean that the signal path becomes longer across the observation interval.
Under our chosen convention, the geometric phase difference is ΔP = −fΔρ/c. A longer path makes the received phase lag more relative to the nominal comparison. Therefore the range-change equivalent is Δρ = −(c/f)ΔP. Reversing the phase-difference convention reverses the corresponding sign; it does not reverse the actual physical motion.
The simplified equation omits oscillator differences, atmospheric effects, moving-endpoint refinements and proper-time transformations. The GINS technical document supplies a much richer model for its stated processing convention. Our equation isolates a conversion so that its dimensions and limits can be checked without pretending it is a complete operational observation model.
Write the units alongside the equation. Multiplying cycles by c/f, a wavelength in metres per cycle, gives metres. Dividing that result by a declared interval gives metres per second. A raw phase number without frequency and interval information does not supply either a range change or a range-change rate on its own.
CHAPTER 3 OF 12 · Convert and time the interval
3. Work a phase difference in both arithmetic directions
Use a fictional nominal frequency of two gigahertz and the exact vacuum speed c = 299,792,458 metres per second. The wavelength is 0.149896229 metres per cycle. If the compatible phase difference is minus 1000 cycles across two seconds, the geometric range-change equivalent in our convention is plus 149.896229 metres.
The interval-average rate is 149.896229/2 = 74.9481145 metres per second. That number is not automatically the full speed of the satellite. It is an interval path-change equivalent under the simplified model. The line-of-sight geometry and other terms must be handled before inferring a state vector.
Check the inverse calculation: −fΔρ/c returns minus 1000 cycles. Now retain the same phase difference but change the nominal frequency to one gigahertz. The wavelength doubles, so the equivalent range change doubles. Retain the original two-gigahertz frequency but double the interval: the range change stays fixed while its average rate halves.
These changes test whether the reader has kept phase, frequency and duration distinct. Do not transfer the fictional two-gigahertz wavelength as an exact DORIS coefficient. Actual channel frequencies, frequency offsets and processing conventions belong to the applicable record. Rounded examples make the mechanism visible; they do not replace the instrument metadata.
CHAPTER 4 OF 12 · Convert and time the interval
4. Keep oscillator comparison separate from path change
Suppose the receiver comparison includes a small fractional frequency mismatch ε under the teaching convention. Over an interval Δt it contributes fεΔt cycles. The count becomes ΔP = −fΔρ/c + fεΔt. Its rate-normalised length equivalent is w = −cΔP/(fΔt) = Δρ/Δt − cε.
Thus a constant fractional mismatch produces a constant apparent range-rate term in this reduced model. For ε = 10⁻⁹, its magnitude is 0.299792458 metres per second. This fictional sensitivity calculation is not a statement about a real beacon or receiver’s oscillator performance.
A constant phase offset behaves differently. If P(t) contains the same added level K at both endpoints, K cancels in the phase difference. A frequency mismatch accumulates with time and does not cancel that way. Confusing the two would lead someone to infer that differencing phase removes all clock-related effects.
In real processing, emitter and receiver oscillator behaviour, time scales and corrections require the documented model. The transferable check is to ask whether a term is constant in phase, proportional to elapsed time, frequency dependent or tied to another event. Similar numerical residuals over a short interval do not establish that their physical causes are the same.
CHAPTER 5 OF 12 · Convert and time the interval
5. Recognise a finite integration rather than a midpoint measurement
The phase difference corresponds to path change over a finite interval. Dividing by elapsed time gives an average, not necessarily the instantaneous range rate at its midpoint. A midpoint label can be convenient, but convenience is not proof that the observed function was linear enough for the intended interpretation.
Choose a fictional path curve ρ(t) = ρ_0 + 100t + 0.1t³ metres, with t in seconds relative to the interval centre. At t = minus one second the offset from ρ_0 is minus 100.1 metres; at plus one it is plus 100.1. The two-second average rate is therefore 100.1 metres per second.
At the midpoint, the derivative is 100 + 0.3t² = 100 metres per second. The difference of 0.1 metres per second is caused by curvature of the rate, not a conversion error. For a quadratic path curve with constant acceleration and symmetric endpoints, midpoint rate and interval-average rate happen to agree; that special case does not settle the cubic example.
To compare observations and predictions fairly, evaluate the model over the same interval and with compatible event definitions. An instantaneous state sampled at one labelled epoch may require integration or an appropriate approximation before comparison with a finite phase count. Increasing printed timestamp precision does not resolve a mismatch between those two quantities.
A shifted interval is different from a changed duration. For a separate teaching curve ρ(t) = ρ_0 + vt + at²/2, moving both interval endpoints later by ε while retaining the same duration changes the average modelled rate by aε. Choose fictional a = two metres per square second and ε = 0.01 seconds: the difference is 0.02 metres per second. The arithmetic does not require a duration error.
This is why time-tag interpretation belongs beside count interpretation. A correction to an epoch label, a correction to oscillator rate and a correction to an accumulated phase are not interchangeable operations. Test each through its own equation and document whether it has already been applied. A residual that looks constant across one short pass does not, by itself, identify which of these operations was missing.
CHAPTER 6 OF 12 · Keep frequency endpoints consistent
6. Normalize channels before combining their observations
A dual-frequency combination must first put its inputs into compatible quantities. Raw cycle counts at different frequencies do not have the same length scale. Let D_H = −cΔP_H/f_H and D_L = −cΔP_L/f_L be compatible interval range-change equivalents under the same phase convention.
For a simplified leading-order dispersive model, write D_H = G + A/f_H² and D_L = G + A/f_L². G represents the common nondispersive interval contribution; A is a signed coefficient under the declared phase convention. This model describes changes over the same interval, not arbitrarily paired observations with different time windows.
Eliminating A gives G = (f_H²D_H − f_L²D_L)/(f_H² − f_L²). The existing GNSS ionosphere-free guide develops the general cancellation algebra. The separate DORIS issue here is keeping the counted intervals, uplink conventions and antenna reference points compatible with the channel combination.
Use a rounded teaching pair f_H = five times f_L. If D_H = 102 metres and D_L = 150 metres, then G = (25×102 − 150)/24 = 100 metres. The leading-order terms were two and fifty metres. This cancellation does not prove that all remaining terms share the same physical endpoint or that oscillator and instrument effects have disappeared.
CHAPTER 7 OF 12 · Keep frequency endpoints consistent
7. Follow the frequency-dependent antenna endpoint into the combination
Two frequency channels can be associated with different effective antenna phase centres. A combination that cancels a dispersive term can therefore inherit a geometrical endpoint contribution. A primary IDS report discusses this issue for DORIS processing; our calculation uses an invented offset and explicitly declared signs rather than adopting a real antenna coefficient.
Let the high-frequency range equivalent refer to a chosen point, and let the low-frequency input contain an extra projected endpoint contribution h(t). Write r_H = R + I_H and r_L = R + I_L + h. If α and β are the ionosphere-free coefficients, α+β = 1 and αI_H+βI_L = 0, then αr_H+βr_L = R + βh.
For our rounded frequency ratio five, α = 25/24 and β = −1/24. If h = d sinE in a declared fictional vertical-offset model, the combined range contains −d sinE/24. The combination has removed the stipulated dispersive term but has not removed this frequency-dependent endpoint contribution.
This is a model-definition result, not a universal DORIS correction sign. A different definition of h or reference point changes how the correction is written. State the actual phase-centre model and whether its correction has already been applied. A label such as ionosphere-free describes one cancellation job, not a certificate that all represented distances terminate at one common physical point.
CHAPTER 8 OF 12 · Keep frequency endpoints consistent
8. Difference the endpoint contribution over the actual count interval
For a phase-count observation, it is the change of the endpoint contribution across the interval that matters. The toy combined range-change term is β[h(t_2) − h(t_1)]. Treating the endpoint correction as one fixed range-rate bias without inspecting its geometry can therefore misrepresent the observation.
Choose fictional d = 0.5 metres, β = −1/24 and elevations E_1 = 30 degrees, E_2 = 90 degrees. Then h_1 = 0.25 metres and h_2 = 0.5 metres. The combined interval contribution is −0.25/24 = −0.010416667 metres. Across ten seconds its average-rate contribution is approximately −0.001041667 metres per second.
A fixed h at both endpoints would cancel from the interval change, even though it would still affect an absolute range-like quantity. A changing projection does not cancel. This distinction connects endpoint handling to the same interval-versus-instant question developed earlier, without assuming that a physical antenna actually changes shape during the pass.
For actual data, use the specified ground and satellite antenna conventions and the documented observation model. Do not substitute the fictional half-metre offset or rounded coefficients. The useful practical question is whether the modelled correction was evaluated at the same events and reference points as the count, not whether one plausible correction value appeared somewhere in a station description.
CHAPTER 9 OF 12 · Audit uncertainty and continuity
9. Propagate covariance through phase differences and channels
Let the two endpoint phase errors each have standard deviation σ cycles. If they are independent, the phase-difference variance is 2σ². If they share covariance k, that variance is 2σ² − 2k. Dividing by interval duration and multiplying by wavelength carries the result into a range-rate uncertainty under the teaching conversion.
Choose σ = 0.01 cycles, zero covariance and the fictional two-gigahertz wavelength. Across two seconds, the rate standard deviation is 0.149896229×√2×0.01/2, approximately 0.001059926 metres per second. The same endpoint errors across a ten-second interval give one-fifth of that formal rate uncertainty, provided the error assumptions remain applicable.
Adjacent count intervals often share an endpoint phase. In a three-epoch teaching sequence with independent phase errors, differences P_1−P_0 and P_2−P_1 have covariance minus Var(P_1). They are not independent merely because the reported intervals have different labels. A later fit should preserve this transformed dependence.
The dual-frequency combination has another covariance transformation: Var(αD_H+βD_L) = α²Var(D_H)+β²Var(D_L)+2αβCov(D_H,D_L). This formula is not an optional refinement when channels share effects. Distinguish conditional random uncertainty from missing endpoint or time-scale models; the covariance calculation does not automatically cover every systematic discrepancy.
CHAPTER 10 OF 12 · Audit uncertainty and continuity
10. Check phase continuity and record changes before interpreting motion
A constant phase reference cancels only when it is consistent across the endpoints. If a phase discontinuity occurs between them, the count includes that step. Under our conversion, a one-cycle step in a fictional two-gigahertz channel produces a 0.149896229-metre equivalent change, with sign determined by the step convention.
Across a ten-second interval, that corresponds to approximately 0.014989623 metres per second. This is not necessarily a physical path change. Flags, station or receiver configuration information and continuity checks belong with the observation rather than being discarded once a numerical difference has been formed.
Changing the count interval may make a discontinuity’s apparent rate contribution change because the same step is divided by a new duration. A true smooth path rate and a discrete phase step therefore need not scale in the same way across the test. This is a diagnostic idea, not a complete algorithm for detecting or repairing actual DORIS records.
Also check whether clock or phase corrections have already been applied. A second application can create a new step or trend that resembles the problem being investigated. Keep supplied data, correction history and derived quantities separate enough that another analyst can reconstruct which operation produced each value.
CHAPTER 11 OF 12 · Transfer and report
11. Transfer the calculation to a changed interval and geometry
Return to the cubic path ρ(t) = ρ_0 + 100t + 0.1t³ and change the symmetric endpoints to minus two and plus two seconds. The offsets are minus 200.8 and plus 200.8 metres. Their difference divided by four gives 100.4 metres per second, while the midpoint derivative is still 100.
The average-versus-midpoint discrepancy became four times larger because the interval squared entered the cubic contribution. More phase accumulation did not make the interpretation automatically more instantaneous. A longer interval may improve some random-noise measures while increasing sensitivity to model variation or discontinuities.
For the endpoint example, keep the fictional elevations unchanged but double the physical offset d from 0.5 to one metre. The combined endpoint range change doubles. Keep d fixed but make E_1 and E_2 equal: it vanishes in this simple difference. These changes test geometry and time differencing separately from the dispersive cancellation.
Finally add the same constant compatible range-change contribution B to both frequency channels. Since α+β = 1, B remains in the combined result. The operation did not remove every common term. This is especially important when deciding what additional models are needed before the count contributes to an orbit or station-coordinate estimate.
CHAPTER 12 OF 12 · Transfer and report
12. Report the counted quantity before the inferred geodetic product
A useful report states the observation type, frequency definition, phase convention, beacon and receiver identifiers, endpoint time tags, integration interval and time-scale treatment. It then describes conversion, channel combination, phase-centre transfer and other applied corrections in their actual order.
Keep the observed interval quantity separate from the inferred orbit or station state. A path-change rate constrains a projection within a model. It does not directly supply an entire velocity vector, long-term station trend or absolute distance. Those conclusions require additional geometry and observations, with uncertainty carried into the estimated parameters.
Preserve reproducible checks: cycle-to-length conversion and its inverse; a constant phase offset that cancels; a frequency mismatch that remains; the cubic-window counterexample; and a changing endpoint projection that survives ionospheric cancellation. These challenge different failure modes rather than merely demonstrating that the software produces numbers.
The gold standard is not treating a dual-frequency radio measurement as a universal corrected ruler. It is following the count through time, frequency, geometry and uncertainty until the represented quantity is explicit. When the operational model adds details beyond our teaching approximation, those details strengthen the inference only if their conventions and evidence remain visible.
Independent worked case and evidence checks
Independent record exercise: use the fictional two-gigahertz channel and compatible phase values P_0 = 12,345 cycles and P_1 = 11,845 cycles at epochs separated by five seconds. The count is minus 500 cycles. With λ = 0.149896229 metres per cycle, the range-change equivalent is plus 74.9481145 metres and its interval-average rate is 14.9896229 metres per second.
Add a constant 200 cycles to both phase values. The count and converted result remain unchanged. Now add one cycle only to the second endpoint. The count becomes minus 499, and the range-change equivalent decreases by one wavelength. This separates a common reference level from a discontinuity across the actual interval.
Use a second rounded channel with frequency one-fifth of the high channel. Convert its count to compatible length before combining. Suppose the converted interval quantities are D_H = 51 metres and D_L = 75 metres. Their leading-order combination is (25×51−75)/24 = 50 metres. The dispersive contributions in this invented example were one and twenty-five metres.
Attach a separate endpoint term to the low channel: let h change by plus 0.12 metres across that interval. Its contribution to the combined interval quantity is βΔh = −0.12/24 = −0.005 metres. If the desired result refers to the high-channel endpoint in this declared model, remove that represented contribution with the corresponding opposite correction. Do not apply a memorised sign without reconstructing the equation.
For an uncertainty check, take independent phase-error standard deviations of 0.02 cycles at the two high-channel epochs. The count standard deviation is √2×0.02 cycles. Multiply by the wavelength and divide by five seconds to obtain the interval-rate uncertainty. Then introduce a positive covariance of 0.0002 square cycles between the phase errors and recompute the difference variance as 0.0004+0.0004−0.0004 = 0.0004 square cycles.
Finally inspect two adjacent five-second intervals from three phase epochs. The shared middle error enters the first difference positively and the second negatively. Retain its negative cross-covariance when fitting a longer sequence. Averaging counts as independent observations would describe a different stochastic model, even if their arithmetic means happened to match.
The exercise has not recovered an actual satellite orbit or a DORIS station coordinate. It establishes the meanings and transformations of fictional interval observations. Actual processing must restore the specialist event, clock, propagation, antenna and relativistic terms that were deliberately simplified here.
Worked comparison
| Operation | What it produces | What still needs checking |
|---|---|---|
| Difference compatible phases | A cycle count over an interval | Continuity, observation type and sign |
| Multiply by −c/f | A declared range-change equivalent | Frequency definition and omitted model terms |
| Divide by interval duration | An interval-average rate | Not automatically the midpoint rate |
| Form a leading-order dual-frequency combination | Cancellation of the stipulated dispersive term | Oscillators, endpoints and residual terms |
| Transfer the frequency-dependent antenna endpoint | A quantity referred to a declared point | Projection at both interval events |
| Propagate covariance | Conditional random uncertainty | Model discrepancy and shared dependence |
Questions and practical checks
Is a DORIS phase count an instantaneous Doppler value?
It represents a phase difference over a defined interval. A rate obtained by dividing by duration is an interval quantity. Comparing it with an instantaneous midpoint model requires a justified approximation or matching integration.
Does phase differencing remove oscillator effects?
A consistent constant phase level cancels. A frequency mismatch accumulates over time and can remain as a range-rate contribution. Actual emitter and receiver time conventions require the specialist model.
Why discuss antenna endpoints after an ionosphere-free combination?
The combination cancels a specified frequency-dependent term. Different effective channel endpoints can contribute a geometrical term that survives. Its reference and sign must be reconstructed from the observation model.
Can I use the rounded frequencies and half-metre offset here for real data?
No. They are fictional teaching inputs. Use the actual DORIS record definitions and applicable instrument models. The nominal system frequencies were identified separately, not substituted for every processing detail.
What should be preserved with a derived count?
Original phase fields, time tags, channel metadata, continuity information, corrections and covariance assumptions. These let another reader distinguish a measured change from a modelling or bookkeeping artefact.
Continue with the connected subjects
Compare the related mechanisms without substituting their conventions
Compare GNSS Doppler geometry and receiver clock drift
Study the general ionosphere-free cancellation and covariance algebra
Separate phase continuity from a physical displacement claim
Follow the reference point and estimated product
Connect instrument reference points with surveyed local ties
Compare antenna reference-point and phase-centre distinctions
Keep the epoch of a geodetic product explicit
Sources and scope
CNES: the Doppler observation equation in GINS, including emitter and receiver cycles
AVISO/CNES: DORIS ground beacons and nominal transmission frequencies
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 1 November 2026.
Continue through observations, reference points and combined information
Use the measurement model to identify what a record represents, the local survey to connect physical endpoints, and the adjustment model to retain the information and uncertainty. These are complementary steps, not interchangeable corrections.
Compare the observation quantities
Radio-source structure: follow complex visibility into a group-delay slope
Connect the reference and the estimate
Surveyed local ties: connect the instruments’ reference points
Normal-equation combination: retain nuisance uncertainty and avoid duplicate information