The receiver is tracking about thirty satellites on two bands, and the integrity prediction still refuses the vertically guided approach planned for the arrival. The contradiction isn’t one: adding frequencies and constellations acts on specific terms of the error budget, and on those alone. The rest — jamming, spoofing, scintillation, the single chain from antenna to computer — comes out of the operation untouched.
An error budget, and three properties that sort its terms
The measured pseudorange differs from the true geometric distance by a sum of known terms: satellite clock and orbit, ionospheric delay, tropospheric delay, multipath near the antenna, receiver noise, receiver-chain bias. Memorizing this budget in order is useless: its terms must be sorted by three properties, and three only.
- Dispersive or not. Does the term depend on the carrier frequency? If so, two frequencies are enough to isolate it and then eliminate it. If not, no combination of frequencies will touch it.
- Correlated or independent between sources. Is it common to several satellites, or even to an entire constellation, or drawn independently for each? Redundancy only dilutes the independent.
- Space-segment or local. Does it originate in the space segment, in the medium traversed, or a few metres from the antenna? Adding more satellites changes nothing about what happens on board.
This sort predicts the outcome without calculation: dual frequency acts on the dispersive, constellation multiplicity on the independent and on the single-source fault. What is neither dispersive, nor independent, nor space-segment resists both — and that is precisely what causes operations to fail.
Dual-frequency systems remove ionospheric errors by using two different signals, while more satellites improve accuracy by giving better positioning data. But they can’t fix everything, like jamming or equipment failures.
An Airbus A350 using dual-frequency GPS and Galileo over Europe might see 24 GPS and 24 Galileo satellites, but if the ionosphere is calm, a single-frequency system could actually be more precise in that moment.
What dual frequency actually removes
The ionosphere is a dispersive medium: the group delay it imposes is proportional to the integrated electron content along the line of sight, and inversely proportional to the square of the frequency. The same line of sight observed on the upper band, around 1575 MHz, and on the lower band, around 1176 MHz, thus provides two equations for one unknown. The term is eliminated instead of modelled.
The reach of this move goes beyond precision. In a single-frequency augmentation service, the ionospheric delay arrives as a grid: interpolated values, an age, and above all a bound on what remains after correction. That bound must cover rare, poorly modelled gradients that vary sharply with magnetic latitude and the solar cycle. It is this bound, not average precision, that most often limits the availability of a vertical minimum: the receiver computes a vertical protection level incorporating this uncertainty, and refuses the operation when it exceeds the alert limit. Removing the term at the source takes out of the calculation the contributor that was inflating the bound. The gain shows up in availability well before it shows up in metres.
« First order » doesn’t mean « in full »
The combination eliminates the first-order term, which accounts for the overwhelming majority of the delay. Higher-order contributions and a ray-bending effect remain, negligible against aeronautical requirements: residual doesn’t mean zero, but the distinction here has no practical consequence. It does have one, however, on noise, and that is where the operation is paid for.
The terms two frequencies cannot reach
What dual frequency does not do follows from the same criterion. Tropospheric delay is not dispersive in these bands: dry air and water vapour delay both carriers equally, so it remains modelled, as before. Multipath is no more eliminated: the geometry of the reflection differs from one band to the other, so the combination does not cancel it out, it mixes it. As for clock and orbit error, it is common to both bands by construction: it still has to be corrected or bounded.
What constellation multiplicity buys: geometry
Here the gain is arithmetic before it is physical. More numerous and better distributed sources lower the dilution of precision, particularly on the vertical axis, which is the constraining one. Above all, they make routine a redundancy that used to be marginal: detecting a faulty measurement requires at least one redundant measurement, excluding it requires two, plus a geometry that lets the test point to the culprit. In the classic single-constellation formulation, where the solution already consumes four unknowns, one extra usable measurement allows detection and a second allows exclusion — if their distribution allows it, which the count alone does not guarantee. With a single system, terrain masking, a prolonged bank angle or fuselage shadowing can drop the count below the threshold at the worst possible moment. With two or three, the count survives.
What remains to be named is what has just been bought: availability and continuity, not integrity. Integrity is a statement of probability — that the position error exceeds a bound without the crew being alerted within the specified time — and it rests on a characterisation of each system: nominal error, probability that a satellite is faulty, probability that an entire constellation is. Without an up-to-date characterisation, satellites are stacked up without the right to tighten the bound.
Hence a counter-intuitive reversal: the value of a second constellation is not the number of satellites it brings, it is its independence. A fault at the scale of a constellation — a bad parameter upload from its ground segment — is not diluted at all by the other satellites of that same constellation; it is diluted by those of another system. That is why an onboard algorithm claiming to support vertical guidance must survive the complete loss of one system.
The arithmetic cost of each added system
Each constellation runs on its own time reference and its own terrestrial frame. Mixing measurements requires either using a broadcast inter-system offset, or estimating it — and estimating it consumes a degree of freedom. Four satellites from a second system do not bring four redundant measurements: the first one pays for the offset. Two constellations of four are therefore not worth one constellation of eight, and the gap shows up exactly where it hurts, in the measurements available for exclusion.
The price we forget to weigh against the gain
The combination that frees the solution from the ionosphere amplifies the noise of the measurements that compose it, by a factor of roughly two to three depending on the bands used. Counter-intuitive but real consequence: under calm ionospheric conditions and with a correct model, a single-frequency receiver can be more precise than a dual-frequency one. So it is not nominal precision being bought, it is the disappearance of a rare, poorly bounded, seasonal term.
The lower band also has its own environment, and it is not a comfortable one: an aeronautical radionavigation band shared with pulsed emitters — telemetry, tactical air navigation, military data links. Receivers protect themselves with synchronized blanking on the pulses, which discards a fraction of the observation time. Where these emitters are densest and most visible, at high altitude above a well-equipped region, the effective carrier-to-noise ratio degrades. The lower band is not a spare upper band.
Finally, the upgrade is not a software one: an antenna and wiring qualified for the upper band alone do not restore the lower band under usable conditions, and the group delays between bands must be known for the combination to make sense. A capable receiver is not a capable installation.
The edge case deserves to be anticipated rather than discovered. Loss of a band in flight — jamming on the lower band, prolonged blanking above a dense zone, failure of a receiver stage — does not degrade performance in gentle steps: it flips the calculation into another regime. The ionospheric term reverts to being modelled, hence bounded, and the vertical protection level jumps at once, sometimes past the alert limit of the segment underway. Depending on the installation, this downgrade is announced, silent, or expressed as a plain refusal of the operation. Knowing which of the three applies to the aircraft being flown is read in the onboard documentation, not discovered in flight: on a final segment, the gap between an announced degradation and a silent one decides the conduct to follow.
The trap. Capable, approved and provided are three distinct things, and the satellite count on the position page tells you about none of them. A dual-frequency multi-constellation receiver flying over a region where the augmentation only broadcasts the single-frequency service gets the performance of the single-frequency service. And the satellites shown as tracked are not necessarily the ones the position solution uses, still less the ones the integrity calculation admits: reading thirty and believing you’re covered is mistaking a reception status for a performance claim.
The right reflex. Ask three separate questions instead of one. What does the installation actually do, meaning which bands and which constellations are genuinely admitted into the solution and into the integrity calculation? What does the service above the destination provide, for the intended window? What does the approval on file authorise? Then check availability with the prediction tool configured as the aircraft that will actually depart, never from the number of satellites in view.
What comes out of both operations untouched
Jamming is a power problem, not a modelling one. Two bands far apart in the spectrum provide real diversity against a narrowband emitter — an onboard repeater, a passenger’s faulty device, a poorly filtered ground installation — but a wideband emitter, or simply a powerful, nearby one, takes both.

The most widely misunderstood point fits in one sentence: constellation multiplicity does not protect against jamming, because constellations share the bands. Adding systems does not add frequencies. An emitter around 1575 MHz simultaneously removes the upper band from every system present, and the effect is by zone, not by satellite.
Spoofing is addressed by neither one. Multi-constellation raises the cost of a coherent falsification and enables plausibility cross-checks: a surveillance-level defence, not a signal-level one. Navigation-message authentication offered by a constellation is a third mechanism, distinct from the previous two, and it authenticates content, not time of arrival: a delayed replay of authentic signals stays authentic.
Ionospheric scintillation escapes the correction logic altogether: amplitude fades and rapid phase variations do not distort the measurement, they drop the tracking loops, and an algorithm has nothing to correct once it no longer has a measurement. Both bands are affected, not always at the same instant: partial diversity, zero immunity. The medium whose delay dual frequency removes can, under another guise, remove the signal itself.
What remains untouched by the question: barometric altimetry, on which altimeter setting, decision and most procedure altitudes depend; database integrity, where a coding error in the final segment does not show; procedure design; and the single chain on board. One antenna, one cable, one receiver: no diversity in the space segment makes up for a single point of failure in the aircraft.
| Term or threat | Dual frequency | Multiple constellations | What still needs treating otherwise |
|---|---|---|---|
| Ionospheric delay, first order | Eliminated by the combination | No effect | Higher-order residuals, no practical reach |
| Bounding of rare ionospheric gradients | The term leaves the calculation, so does the bound | No effect | Nothing for this term; scintillation remains untouched |
| Scintillation and tracking-loop dropout | Partial diversity between bands | No effect: bands shared by all | Known zones and periods, conventional fallback |
| Tropospheric delay | No effect: non-dispersive medium | No effect | Onboard model, as before |
| Satellite clock and orbit | No effect: error common to both bands | Detection and exclusion through redundancy | Augmentation corrections, or characterised bounds |
| Fault affecting an entire constellation | No effect | Handled by the independence of another system | Algorithm surviving the loss of one system |
| Insufficient geometry, masking, banking | No effect | This is its very purpose | Pre-flight availability prediction |
| Multipath near the antenna | No effect: the combination mixes it | No effect | Antenna siting, receiver processing |
| Measurement noise | Amplified, factor of roughly two to three | Attenuated by redundancy | Smoothing and filtering, no miracle |
| Narrowband jamming | The other band may survive | No effect: shared bands | Onboard detection, loss procedure, fallback |
| Wideband or high-power jamming | No effect | No effect | Conventional aid, alternate field, fuel |
| Spoofing | No effect | Plausibility cross-checks only | Message authentication, consistency with inertial reference |
| Altimeter setting, database, flying technique | No effect | No effect | Wherever they have always been handled |
From receiver to operational credit: a chain that must be complete
Between a capable receiver and a lower minimum, there is a chain of conditions, none of which follows from the others.
- The constellation’s service, with its in-space performance commitment.
- The augmentation service level actually broadcast over the region and during the window, when the operation depends on it.
- The performance standard the equipment was built to, and the functions it covers.
- The installation’s approval, as it appears in the flight manual supplement.
- The operator’s authorisation for the intended operation, within its operations specifications.
- The published procedure, and the minima it carries.
- The availability check at flight planning, when required.
Miss one link and the credit does not exist, whatever the quality of the rest. The least intuitive point is the sixth: a procedure published for a given service level does not become more permissive because the aircraft carries better equipment.
Where the change is tangible, it is markedly so: availability of vertical-guidance minima in regions and periods where the ionosphere was the limiting factor, fewer prediction refusals for geometry, tolerance to masking during manoeuvring, and the possibility for a properly bounded onboard algorithm to support vertical guidance without regional augmentation. This last point shifts the geography of what’s possible: it opens three-dimensional operations where no ground infrastructure is planned.
| Question | Single-frequency with regional augmentation | Dual-frequency, multiple constellations |
|---|---|---|
| Handling of the ionosphere | Broadcast grid, interpolated values, bounded residual | Term eliminated by the measurement itself |
| Limiting factor for the vertical minimum | Bounding of the ionospheric residual | Geometry and system characterisation |
| Source of clock and orbit corrections | The augmentation service | Broadcast parameters and characterised bounds |
| Where the integrity claim is formed | Largely on the ground, applied on board | On board, on bounds supplied and kept current |
| Measurements available for exclusion | Often just sufficient | Comfortable, minus one degree of freedom per system |
| Narrowband jammer | Loss of function | The other band may survive |
| Wideband or very powerful jammer | Loss of function | Loss of function |
| Dependence on regional infrastructure | Total, for vertical guidance | Reduced: the onboard algorithm carries the claim |
| What conditions the operational credit | Approval, broadcast service, published procedure | The same, unchanged |
What to go and check at home
Five questions are settled in the onboard documentation, and nowhere else.
- Which bands and which constellations the installation uses, as distinct from those it displays.
- Which of them the integrity calculation admits, which is not the same list.
- What the flight manual supplement authorises, and under which equipment-serviceability conditions.
- Whether the prediction tool used at flight planning models the configuration of the actual departing aircraft, or a generic configuration.
- What the procedure prescribes when a notice declares a system unusable, or reports interference over the area.
Three grey areas remain, better named than endured. System characterisation must be produced, distributed and kept current: the mechanism for its distribution and the conduct to follow when it is out of date are not described with the same clarity everywhere. Handling of a whole-constellation fault varies from one installation to another. Finally, in a fleet with mixed configurations, the same procedure will be available for one aircraft and refused for another with nothing in the cockpit explaining why: comparing across fleets is the only way to notice it, and that is what the exchanges between professionals are for.
For anyone preparing a qualification, the shift in what’s being examined is clear. The question is no longer to explain how a satellite system works, but what today’s integrity claim rests on, which term of the error budget the configuration treats, which it does not, and what to do when the claim vanishes mid-final-segment. Answering through the mechanism rather than the system’s name is what the FL330 training programmes on performance-based navigation work towards.
This article describes mechanisms and sets no limits. In the event of any discrepancy, applicable regulations, current charts and notices, and the operator’s approved documentation prevail over what is written here.
Questions fréquentes
Does a dual-frequency multi-constellation receiver do away with the availability prediction before an RNP approach?
No. The prediction is still required whenever the procedure or the operator’s documentation calls for it, and it covers the aircraft’s actual configuration, not what the receiver can do in general. What dual frequency changes is how often it refuses: with the ionospheric residual no longer inflating the bound, the prediction succeeds more often. It can still refuse for geometry, for an announced satellite outage, or for reported interference over the area.
Does constellation multiplicity protect against jamming?
No, and it is the most widespread misconception. Constellations share the same frequency bands: adding systems does not add frequencies. A jamming emitter in the upper band simultaneously removes that band for every system present, and the effect is by zone. Only dual frequency provides diversity, and only partial diversity, against a narrowband emitter that reaches only one of the two bands.
Does dual frequency correct tropospheric delay the way it corrects ionospheric delay?
No, because the troposphere is not dispersive in these bands: dry air and water vapour delay both carriers identically, so combining the two measurements reveals nothing. The term stays handled by an onboard model dependent on satellite elevation and assumed conditions. This is the general rule to keep: only what depends on frequency can be eliminated by two frequencies.
Are two constellations of four satellites worth one constellation of eight?
No, and the gap cuts both ways. On the cost side, each added system runs on its own time reference: if the inter-system offset is estimated rather than broadcast, it consumes a degree of freedom, so one of the second system’s four measurements does not become redundancy. On the benefit side, two independent systems protect against a fault affecting an entire constellation, which no number of satellites from a single system can do.
Is a dual-frequency receiver always more precise than a single-frequency one?
No. The combination that eliminates the ionosphere amplifies the noise of the measurements that compose it, by a factor of roughly two to three depending on the bands used. Under calm ionospheric conditions and with a correct model, a single-frequency receiver can therefore show a smaller nominal error. The gain from dual frequency is not average precision, it is the disappearance of a rare, poorly bounded, strongly variable term that was the limiting factor for vertical availability.
What does dual frequency change for ionospheric scintillation?
Very little, because scintillation does not distort the measurement: it removes it. Amplitude fades and rapid phase variations drop the tracking loops, and a correction algorithm has nothing to correct once it no longer has a measurement. Both bands are affected, not necessarily at the same instant, which gives partial diversity but no immunity. The right conduct remains knowing the exposed zones and periods, and having a conventional fallback available.

Leave a Reply
You must be logged in to post a comment.