Introduction
"Sometimes GNSS signals do not arrive directly - they bounce off surfaces first."
Multipath interference is one of the most persistent and difficult-to-eliminate sources of GNSS error. Unlike ionospheric delays or clock offsets, multipath cannot be corrected by a ground network or predicted from a model. It is site-specific, time-varying, and highly dependent on the local geometry of reflective surfaces. For GNSS professionals, understanding the physical mechanisms of multipath and its distinct effects on code and carrier measurements is essential for diagnosing performance problems and selecting appropriate mitigation strategies.
The Reflection Mechanism
A GNSS signal transmitted from a satellite travels toward the receiver antenna. In ideal conditions, it arrives via a single direct path - the line-of-sight (LOS). In practice, the signal also reflects off nearby surfaces, and these reflected copies arrive at the antenna slightly later than the direct signal because they have travelled a longer path. The receiver's tracking loops, designed to align with the direct signal, are confused by the superposition of the direct and reflected waveforms.
The severity of the resulting error depends on several factors: the amplitude of the reflected signal relative to the direct signal, the additional path length (delay) of the reflected signal, and the phase relationship between the direct and reflected signals when they combine at the antenna. Reflected signals that are in-phase with the direct signal cause constructive interference, distorting the correlation peak in one direction. Out-of-phase reflections cause destructive interference, shifting it in the other direction.
Reflective Surfaces
Not all surfaces reflect GNSS signals equally. The smoothness, conductivity, and geometry of a surface determine the strength and character of the reflected signal:
- Metal structures (aluminium, steel): The strongest reflectors. Metal produces high-amplitude specular reflections capable of introducing large multipath errors. Controlled experiments confirm that aluminium causes the highest multipath levels among common building materials.
- Glass facades: Modern buildings with large glass surfaces are nearly as reflective as metal at GNSS signal wavelengths. Glass-curtain-wall buildings create consistent, strong reflections that persist as a receiver moves along an adjacent street.
- Water surfaces: Calm water is an excellent specular reflector. Near rivers, lakes, harbours, and coastal areas, water-surface reflections produce strong, stable multipath signals. This effect is well-documented in marine applications where docking at low speed makes multipath errors particularly persistent.
- Concrete and masonry: Less reflective than metal or glass, but still capable of producing significant multipath in close proximity. Diffuse rather than specular reflections are more common.
- Vegetation: Trees and shrubs produce diffuse, irregular reflections. The effect changes dynamically as wind moves branches, producing time-varying multipath that resembles noise rather than a systematic bias.
Code Phase (Pseudorange) Impact
Pseudorange measurements are derived from code-phase tracking - the alignment of the receiver's locally-generated code replica with the received signal. The receiver uses the early-minus-late discriminator in a Delay Lock Loop (DLL) to maintain this alignment. When a reflected signal is superimposed on the direct signal, the composite correlation function is distorted, and the DLL locks to a shifted version of the true peak. The result is a pseudorange error - a systematic bias in the measured range to the satellite.
Code multipath errors can range from a few centimetres to several metres depending on the reflection delay and amplitude. In urban environments, multiple simultaneous reflections from different surfaces compound these errors, producing position errors of several metres or more that fluctuate as the receiver moves or as satellite geometry changes.
Carrier Phase Impact
Carrier-phase tracking uses a Phase Lock Loop (PLL) to measure the fractional and integer cycle count of the carrier wave from each satellite. The maximum carrier-phase error due to a single reflected signal is limited to a quarter of the carrier wavelength - approximately 4.8 cm on GPS L1. This is substantially smaller than the corresponding code multipath error, which explains why carrier-phase measurements form the foundation of high-precision techniques such as RTK and PPP.
However, carrier-phase multipath is harder to mitigate. When the reflected signal is strong enough, it can cause the PLL to slip cycles - an event known as a cycle slip - which immediately invalidates ambiguity resolution and forces re-initialisation of RTK solutions. Even without cycle slips, persistent carrier-phase biases introduce sub-centimetre to centimetre-level systematic errors that are difficult to separate from genuine signal variation.
Phase Distortion and Polarisation
GNSS satellites transmit right-hand circularly polarised (RHCP) signals. When a signal reflects off a surface, the polarisation is reversed - the reflected signal becomes left-hand circularly polarised (LHCP). High-quality GNSS antennas are designed to have significantly greater sensitivity to RHCP signals than LHCP signals, providing an inherent degree of multipath rejection. Choke ring antennas and ground planes extend this rejection to signals arriving from low elevation angles, which are the most susceptible to ground-bounce multipath. However, no antenna design can eliminate all multipath in a dense urban or industrial environment.
Detection and Mitigation
Multipath is notoriously difficult to detect from within the receiver, because a distorted correlation function looks identical to a valid one to standard tracking algorithms. Indicators that suggest multipath is active include elevated code-minus-carrier divergence, unusual C/Nâ‚€ fluctuation patterns, and residuals in the navigation solution that are inconsistent with expected noise levels. Mitigation strategies include:
- Elevation masking to exclude low-angle signals most prone to ground reflection
- Narrow correlator spacing in the DLL to reduce the correlation function peak distortion
- Multipath estimating delay lock loops (MEDLL) and similar advanced correlator designs
- Sidereal filtering for static applications, exploiting the near-repetitive nature of multipath from fixed reflectors
- Antenna selection - choke rings, ground planes, and controlled radiation pattern antennas (CRPAs)
Despite decades of research, multipath remains the dominant unmitigated error source in urban and semi-urban GNSS deployments. It is the primary reason why raw GNSS accuracy in cities falls well short of open-sky specifications.