High-precision GNSS is not a single technique. RTK, PPK and conventional static relative positioning use common observations between a reference and a rover. PPP processes undifferenced observations from one station with precise orbit, clock and bias products. Calling PPP a differential technique obscures this difference in observation models.
This comparison focuses on what data each method needs, where its corrections originate, when it produces a result and how dependency loss affects service. Those characteristics provide the basis for project field validation and method selection. For the systems and basic error sources, see GNSS Systems and Positioning Principles.
Two meanings of DGNSS
Differential GNSS (DGNSS) can broadly mean any technique that uses reference stations to improve a user solution. In a narrower equipment or service context, DGNSS often means code differential positioning, where pseudorange corrections improve on standalone positioning.
To avoid ambiguity, this article uses “code differential,” “single-base RTK” and “Network RTK.” PPP remains under precise point positioning, not under differential methods.
Quick comparison
| Method | Main observations and corrections | Result time | Infrastructure | Main trade-off |
|---|---|---|---|---|
| Code differential | Pseudorange plus reference code corrections | Real time | Reference or wide-area service | Simple and quick to recover, but does not resolve carrier integers |
| Single-base RTK | Reference and rover carrier phase | Real time | Nearby reference and data link | Fast high-precision relative position, but sensitive to baseline, atmosphere and communications |
| Network RTK | Regional errors estimated by CORS, delivered as VRS/MAC/FKP or similar | Real time | Reference network, processing center and communications | Wider coverage, with dependence on the network model and service availability |
| PPK | Reference and rover carrier phase | Post-processed | Concurrent reference data and processing software | No live correction link and can be reprocessed, but no result at the time of collection |
| Static relative | Long-duration carrier phase at a stationary point | Post-processed | Reference or network and simultaneous observations | Suitable for control and stability checks, but the point must remain still throughout the observation period |
| PPP | Undifferenced code and phase plus precise orbit/clock products | Real time or post-processed | Global precise products and complete models | No nearby reference, but convergence and product consistency matter |
| PPP-RTK | PPP-like undifferenced observations plus satellite biases and regional atmosphere | Mainly real time | Regional network and state-space correction service | Can accelerate single-receiver integer resolution, but only within a suitable service |
These labels describe observation and correction methods. Project accuracy and availability are established through field validation of geometry, obstruction, atmosphere, baseline, hardware, software and quality control.
How the observation models differ
Code differential
A reference with known coordinates estimates each satellite’s pseudorange error and supplies code corrections. Because it does not resolve carrier-phase integer ambiguities, initialization is straightforward and recovery after interruption is usually easier. Code observations are noisier, so it should not be expected to provide the same resolution as carrier-phase positioning.
It is useful for navigation, asset inventory and general GIS data collection. Centimeter-level relative position or small-deformation time series generally require carrier phase.
Single-base RTK
RTK uses reference and rover carrier phase in real time to estimate the baseline and resolve integer ambiguities. Reference data normally arrives by radio or network in a format such as RTCM. Performance depends on common satellite tracking, continuity of the data link and whether atmospheric errors remain correlated across the baseline.
Fixed status is therefore not sufficient evidence of correctness. Residuals, geometry, correction age, reinitialization and reference stability also require checks. See Baseline Length and GNSS Relative Positioning for the full discussion.
Network RTK
Network RTK is not simply a distant reference. Multiple continuously operating stations estimate regional ionospheric, tropospheric and orbit-related errors and represent corrections through approaches such as VRS, MAC or FKP.
Taiwan’s e-GNSS VBS-RTK service generates virtual reference observations near the user’s approximate position before the receiver performs a short-baseline RTK solution [1]. Users need not build a reference, but remain dependent on service coverage, frame realization, communications, correction format and network-model quality.
PPK
PPK and RTK can use the same reference/rover carrier-phase observations; processing time is the principal difference. PPK obtains the complete data after the mission and can use forward/backward processing, parameter changes or another reference dataset without a continuous field correction link.
Post-processing creates more room for diagnosis and reprocessing, but it cannot automatically repair multipath, bad timestamps, cycle slips or poor reference data. Whether PPK outperforms RTK depends on data completeness and processing strategy.
Static relative positioning
Static surveying assumes that the point remains fixed throughout the observation period and uses changing satellite geometry and redundant carrier observations to estimate one coordinate. NOAA’s OPUS guidance likewise recommends longer observations to improve the opportunity for reliable ambiguity fixing and multipath mitigation [2].
Persistent reflectors can, however, create repeating or systematic multipath. A longer observation period does not make every error average away as white noise. Its duration must be designed around the baseline, equipment, environment and coordinate objective, with antenna height, model and observation files retained for traceability.
A permanent monitoring installation does not automatically use a static solution. Monitoring estimates position as it changes over time; a conventional static solution assumes one constant coordinate during its observation period. The two models answer different questions when real motion occurs.
PPP and PPP-AR
The IGS characterizes PPP as processing undifferenced observations with predetermined precise satellite products [3]. A nearby reference observation is unnecessary, but consistent orbit, clock, antenna, tide, atmosphere and bias models are still required. Real-time PPP can use orbit, clock, code-bias and phase-bias corrections such as those from the IGS Real-Time Service [4].
Float PPP ambiguities contain satellite and receiver hardware biases and generally require convergence. PPP-AR adds consistent code and phase-bias products so that undifferenced ambiguities recover an integer property. It should not be conflated with ordinary float PPP.
PPP-RTK
PPP-RTK retains an undifferenced user model while a reference network supplies more complete state-space information. Wang, Khodabandeh and Teunissen [5] describe network corrections that include satellite clocks, phase biases and ionospheric delays to support fast single-receiver integer ambiguity resolution.
PPP-RTK can operate without a user-deployed base, but it still depends on a reference network. The network, correction generation, communications and user model must be compatible. Convergence and fixing behavior can change outside the service region or when correction quality degrades.
Questions for comparing methods
1. Absolute coordinates or local relative change
Control surveying and cross-region consistency emphasize the reference frame. Local deformation monitoring emphasizes short-term relative stability and whether the reference is genuinely stable. Different methods—or two methods used as cross-checks—may be appropriate.
2. Required latency
Tasks requiring immediate guidance or warning should evaluate the latency and availability of real-time methods such as RTK, Network RTK, real-time PPP or PPP-RTK. PPK, static and post-processed PPP allow more complete reprocessing when results can follow the mission.
3. Available infrastructure
Check nearby reference data, correction coverage, communication reliability, precise-product latency and raw-data retention. Not deploying a local reference does not remove external dependencies.
4. Moving, stationary or deforming
UAV and vehicle trajectories are kinematic. A static control survey assumes no motion during an observation period. Deformation monitoring observes a permanent installation whose coordinates change over time. Choose the model that matches the measurand; a stationary antenna alone does not make a solution static.
5. Evidence of trustworthiness
In addition to coordinate dispersion, define solution availability, convergence and reconvergence, gaps, external checks, reference stability and change records. For safety monitoring, integrity and visible failure signals can matter as much as nominal accuracy.
Comparing and validating with measured behavior
For permanent-monitoring data, RST can reprocess the same raw observations with different observation periods, solution configurations and reference data, then compare them against independent references and long-term time series. Processing is designed and validated from observed failure modes rather than treating one configuration as the answer for every site.
- Fixed is a solution state, not proof of correctness. Measured data can enter a wrong integer solution while familiar quality statistics still look reasonable, so continuity and external evidence remain necessary.
- Longer observation is not unconditionally better. It may improve ambiguity reliability and coordinate stability, but it also increases latency and can make adjacent outputs share more data. The trade-off must be tested for the site and task.
- A field named sigma may not represent the same physical quantity. It may be formal covariance, dispersion within an observation period or a post-processing statistic. Its definition and origin must be established before methods are compared.
- Multi-frequency and multi-constellation processing changes ambiguity dimensionality. More observations usually help, but reliability does not necessarily improve if validation assumptions and fixed thresholds still come from a simpler single-system model.
Evidence limits of quality indicators
| Indicator or evidence | What it tells us | What it cannot prove on its own |
|---|---|---|
| Fixed status | The processor accepted one integer ambiguity set | The integers are correct, or the displacement is real |
| Ambiguity ratio | Separation between the best and second-best candidates under the current model | Immunity to integer-multiple bias, or one fixed threshold for every constellation and site |
| Sigma | Internal dispersion or convergence under that field’s definition | Quality can be compared across sites or methods without tracing the field semantics |
| Longer observation | After a correct fix, it usually adds redundancy and lowers coordinate dispersion | The shortest window fixes at every site, or one window is best for every task |
| Time-series continuity plus an independent reference | Agreement with other times and external evidence | The reference itself is stable, or every systematic error has been removed |
The following de-identified internal example shows a qualitative relationship, not a universal performance curve. This public article does not include the sample size, site mix or complete processing protocol; each project must establish its own values through field validation.
For clients, the comparison framework connects each method description to timeliness, availability, reference traceability and wrong-solution risk. Field validation establishes performance for the intended conditions, and the delivery documents record the validated project parameters and acceptance evidence.
Roles in deformation monitoring
Real-time, post-processed and static control solutions can complement one another, but none should be declared “truth” by default. Real time provides timeliness, post-processing enables diagnosis and reprocessing, and static control can check the frame. Their data, assumptions and time scales must be recorded before they can provide an effective cross-check.
Reference motion enters relative results with the opposite sign. Whether a project uses a single reference or a network service, it must explain the origin and maintenance of the reference coordinates and what happens when the reference is suspect—not only inspect rover fixed status.
RST applies measured-data comparison and validation to its near-real-time differential GNSS monitoring service, including reference stability and long-term time-series quality checks. Project delivery documents record the validated processing configuration, decision thresholds and supporting field evidence.
GNSS monitoring series
This article compares positioning methods; the adjacent articles explain GNSS fundamentals and the effect of baseline length.
Previous: GNSS systems and positioning principles | Series 2 of 7 | Next: Baseline length and data quality
References
- National Land Surveying and Mapping Center, Taiwan. VBS-RTK real-time kinematic positioning technology. https://egnss.nlsc.gov.tw/content.aspx?i=20150625102049287
- NOAA National Geodetic Survey. OPUS: the Online Positioning User Service. https://geodesy.noaa.gov/OPUS/about.jsp
- International GNSS Service. Precise Point Positioning with Ambiguity Resolution Working Group. https://www.igs.org/wg/ppp-ar/
- International GNSS Service. Real-Time Service. https://igs.org/rts/
- Wang, K., Khodabandeh, A., & Teunissen, P. J. G. (2017). A study on predicting network corrections in PPP-RTK processing. Advances in Space Research, 60(7), 1463–1477. DOI: 10.1016/j.asr.2017.06.043
Frequently asked questions
Q: Which is more accurate, RTK or PPK? Neither method is always more accurate. Both can use carrier-phase relative positioning. RTK’s advantage is immediacy; PPK’s is the ability to reprocess after the complete dataset is available. Baseline, atmosphere, completeness, reference quality and processing often matter more than the label.
Q: Is static always the most accurate method? Long-duration static relative positioning can provide high-quality coordinates when its no-motion assumption holds and the observing environment and processing are sound. Systematic multipath, antenna-height errors, an unstable reference or real motion during the observation period do not disappear merely because the period is long.
Q: Why does PPP not require a nearby reference? PPP uses undifferenced observations and precise satellite orbit, clock and other products. Those products still come from global or regional GNSS infrastructure. A user does not need to deploy a nearby base, but the solution still depends on external products and corrections.
Q: How do PPP, PPP-AR and PPP-RTK differ? Ordinary PPP may keep float ambiguities. PPP-AR adds consistent satellite code/phase-bias products to fix undifferenced ambiguities. PPP-RTK also uses regional atmospheric corrections to accelerate convergence and integer resolution within a service area. The service specification determines the actual messages and capabilities.
Q: Does a permanent GNSS monitoring station use static positioning? Not necessarily. A permanent installation describes the hardware. If the solution estimates position changes for each period, it is a time-series or dynamic-state problem. Conventional static positioning assumes one unchanged coordinate for the entire observation period.
Need help assessing reference configuration and data quality for permanent GNSS monitoring? Contact RST Ltd..