Differential GPS (DGPS) Explained

Differential GPS is the parent concept behind every GPS accuracy improvement available to civilian users. The principle: if two receivers track the same satellites at the same time from nearby locations, most of the error sources are common between them (atmospheric delays, satellite clock errors, ephemeris errors). A reference station at a surveyed point can measure those common errors and broadcast corrections to nearby receivers, which apply the corrections and recover sub-3-metre accuracy from a 5-metre baseline GPS.

The /learn/how-gps-works pillar covers the standalone GPS system; this article covers the differential- correction technique that powers most augmentation systems.

The DGPS principle

A standalone GPS receiver computes its position by trilaterating from pseudorange measurements. Each pseudorange has an error budget (covered in /learn/gps-accuracy-explained) with contributions from:

• Ionospheric delay (largest error source)
• Tropospheric delay
• Satellite ephemeris error
• Satellite clock error
• Multipath
• Receiver noise

Many of these errors are spatially correlated. A receiver 10 km from a reference station tracks the same satellites through essentially the same atmosphere, with the same satellite ephemeris and clock errors. The errors at the rover position are nearly identical to the errors at the reference station.

DGPS exploits this:

1. The reference station has a precisely-surveyed position (known to ~1 cm from previous geodetic work).
2. The reference station tracks all visible satellites and computes the pseudoranges it should observe (given the known satellite ephemeris and its own known position).
3. The difference between the expected pseudorange and the observed pseudorange is the error contribution from ionosphere + troposphere + ephemeris + clock for that satellite at that moment.
4. The reference station broadcasts this per-satellite correction to any nearby receivers.
5. A roving receiver applies the corrections to its own pseudorange measurements before computing its position.

The result: typical DGPS accuracy is 1–3 m horizontal, compared to ~5 m for standalone GPS. The improvement comes from removing the spatially-correlated error component.

Classical DGPS: the USCG beacons

From 1990 to 2020, the US Coast Guard operated the NDGPS network — 86 reference stations broadcasting DGPS corrections on marine-band radio frequencies (typically 285–325 kHz). Originally designed to support maritime navigation in US coastal waters and on the Great Lakes, the network later expanded inland to cover most of CONUS and Alaska.

Key facts:

• Coverage area per station: ~150–300 km radius.
• Broadcast standard: RTCM SC-104 messages.
• Receiver hardware: needed a separate beacon receiver in addition to the GPS receiver.
• Operational accuracy: 1–3 m horizontal.

The NDGPS network was decommissioned starting in 2016 and fully shut down by 2020. The reasoning: WAAS (the FAA's satellite-based augmentation system) provided comparable accuracy with national coverage, no separate receiver hardware needed, and no ground-station maintenance burden.

The USCG decommissioning is sometimes characterised as the end of DGPS, but it's only the end of that particular DGPS implementation. The technique itself lives on in every other GPS augmentation system.

Modern DGPS descendants

Today's DGPS lives in many forms:

SBAS (Satellite-Based Augmentation Systems)

Wide Area Augmentation System (WAAS, US), EGNOS (EU), MSAS (Japan), GAGAN (India), SDCM (Russia). These are global-coverage augmentation systems that broadcast corrections via geostationary satellites instead of ground beacons. The underlying technique is identical to classical DGPS — a network of reference stations computes corrections and distributes them — but the distribution is via satellite rather than radio beacon.

Accuracy: 1–2 m horizontal. Coverage: continental in each case (WAAS covers North America, EGNOS covers Europe, etc.).

RTK (Real-Time Kinematic)

A higher-accuracy variant of DGPS using carrier-phase measurements rather than pseudorange measurements. Carrier-phase provides millimetre-level precision over short baselines (under 30 km from the reference station). The /learn/rtk-gps article (when shipped) covers this in depth.

Network RTK / Virtual Reference Stations

A network of CORS stations computes corrections that collectively model the atmospheric and orbital errors as a function of position. The rover's approximate position is sent to a server, which generates a “virtual” reference-station correction tailored to the rover's location. Sub-2-cm accuracy without needing a dedicated base station.

Major commercial network-RTK services: Trimble VRS Now, Leica HxGN Smart Network, Topcon TopNET. Public services: NOAA NGS provides RTK corrections from the US national CORS network for free.

PPP (Precise Point Positioning)

Uses globally-published precise satellite orbit and clock products from IGS without needing a local reference station. Slower convergence (15+ minutes) but global sub-decimetre accuracy.

NTRIP — the modern correction distribution standard

NTRIP (Networked Transport of RTCM via Internet Protocol) is the dominant modern protocol for distributing DGPS / RTK corrections over the internet. Reference stations publish correction streams to NTRIP casters (servers); rovers connect via mobile internet and subscribe to the nearest caster.

NTRIP enabled the consumer rollout of high-precision GNSS: any device with internet access (smartphone, surveyor's rover, RTK-equipped vehicle) can connect to a caster and receive real-time corrections. Hundreds of NTRIP casters operate globally, ranging from free public services (NOAA NGS, EPN, ETRS89 networks) to commercial subscription services.

DGPS accuracy compared to alternatives

| Technique | Typical accuracy | Coverage | | ----------------------- | ---------------- | ----------------------------------- | | Unaugmented GPS | ~5 m | Global, GPS-visible | | Classical beacon DGPS | 1–3 m | Within ~300 km of beacon | | SBAS (WAAS, EGNOS) | 1–2 m | Continental (per system) | | Network RTK (NTRIP) | 1–2 cm | Within ~30 km of nearest CORS | | PPP | 1–10 cm | Global, sub-decimetre after convergence | | Multi-GNSS unaugmented | 2–4 m | Global, multi-constellation visible |

For most consumer applications, SBAS is the practical default because it's free, requires no separate hardware (built into modern smartphones and standalone GPS devices), and covers entire continents. Beacon DGPS was supplanted by SBAS. RTK and PPP are reserved for survey-grade work.

When DGPS still makes sense

A few niches where classical DGPS-style corrections (not RTK, not satellite-augmented) are still used:

• Marine navigation in areas without SBAS coverage. Some countries still operate local DGPS beacons for coastal use.
• Industrial precision applications — robotics, agricultural machinery, harbour autonomous vehicles — that need 1–3 m accuracy with minimal hardware overhead and don't need RTK's centimetre-level precision.
• Legacy infrastructure that hasn't migrated to SBAS or network RTK. Backwards compatibility for embedded systems designed in the 1990s.

For new installations, the answer is almost always SBAS for metre-scale or network RTK for centimetre-scale; classical DGPS is a legacy technology.

A short history

• 1989: USCG begins developing DGPS for maritime navigation.
• 1990–1996: Selective Availability is active; DGPS becomes practical because the SA-induced error is largely spatially-correlated (and therefore removable by differencing).
• 1996: USCG NDGPS reaches full operational capability.
• 2000: SA turned off. DGPS still useful because of remaining atmospheric / ephemeris errors, but the urgency reduces.
• 2003: WAAS begins operational service for civil aviation. Within a decade, WAAS-capable receivers are mainstream.
• 2008: NTRIP becomes the dominant internet-based correction protocol.
• 2016–2020: USCG NDGPS network gradually shut down.
• 2020s: Network RTK via NTRIP becomes mainstream for surveying; SBAS is the default for consumer accuracy.

The trend: corrections moved from broadcast (terrestrial then satellite) to two-way internet distribution. Modern high-accuracy GPS is essentially an internet-connected service with a satellite-tracking front-end.

A worked DGPS example

Suppose a rover at an unknown position receives the GPS signal from satellite 12 and computes a pseudorange of 22,341,200 m. A reference station 5 km away, with a surveyed position, also receives satellite 12 and computes its own pseudorange of 22,338,500 m. From the known reference position and satellite ephemeris, the reference can compute that the true pseudorange should have been 22,338,200 m. The difference — 22,338,500 − 22,338,200 = 300 m — is the error budget for satellite 12 at that moment, attributable mostly to ionospheric and tropospheric delay.

The reference broadcasts this 300 m correction. The rover subtracts it from its own pseudorange: 22,341,200 − 300 = 22,340,900 m. This corrected pseudorange has the spatially-correlated error removed; only the rover-specific multipath and receiver noise remain. With four such corrections (one per satellite), the rover's position fix improves from ~5 m to ~1–2 m.

The actual computation handles range-rate corrections (the time derivative of the static correction, to interpolate between update intervals) and per-satellite weighting based on the satellite's elevation angle. The RTCM SC-104 message format encodes all of this with sub-metre numerical precision.

DGPS in regulatory contexts

A few notes on how regulators have treated DGPS over the years:

• FAA WAAS is certified for Category I aviation precision approaches (down to ~200 ft decision height). Higher-precision Cat II and III approaches use ground-based augmentation systems (GBAS) rather than SBAS.
• Maritime IMO had the USCG NDGPS as the standard navigation- aid system for coastal vessels; the IALA (International Association of Lighthouse Authorities) now treats SBAS as the equivalent.
• Surveying boards in many countries certify network RTK services for specific accuracy claims (e.g., for cadastral surveying, the network must demonstrate < 4 cm horizontal accuracy at 95% confidence).
• Autonomous vehicles use combinations of network RTK, inertial sensors, and visual odometry; the GPS-augmentation component is unregulated but vendor-tested.

Common misconceptions

“DGPS is obsolete.” The classical USCG beacon network is decommissioned, but DGPS as a technique is foundational to every modern augmentation system. SBAS is DGPS via satellite; RTK is DGPS with carrier-phase; network RTK is DGPS via internet. The principle hasn't changed.

“DGPS requires a paid subscription.” SBAS-based DGPS (WAAS, EGNOS, etc.) is free. Network RTK from commercial providers is paid; free network RTK from public CORS networks (NOAA NGS, EUREF EPN) is available for many regions. Classical beacon DGPS, while it existed, was free public service.

“DGPS requires a base station you own.” Not for SBAS or public network RTK. Classical beacon DGPS used a network of fixed government-operated beacons. Modern users typically subscribe to a network rather than operating their own base station; setting up your own is reserved for survey-grade work or specialised applications.

“DGPS only works in the country where it's deployed.” True for some regional systems (classical USCG NDGPS worked only near USCG beacons), false for others. SBAS-style augmentation systems are continental — WAAS covers all of North America including Canada and Mexico. Network RTK services have coverage areas defined by their CORS networks.

“DGPS improves accuracy proportionally to distance from base.” Roughly the opposite — DGPS accuracy degrades as the rover-base distance increases, because the spatially-correlated assumption (same atmosphere, same satellite errors) holds best for nearby pairs. Classical beacon DGPS was good to 1 m at the beacon location and 3+ m at the edge of the coverage area; modern network RTK fights this with VRS (virtual reference station) modelling that interpolates between nearby CORS stations.