How to achieve centimeter-level GNSS localization for commercial applications, when cost is a factor.
If you are aiming for high-precision, real-time positioning in Germany, you likely already know that standard GNSS isn’t enough. Navigating the landscape of correction services, especially the free options versus the paid ones, can be complex. In this post, we break down why standard GNSS fails at high precision, the correction methods available, and the specific quirks of the German market for real-time correction services.
The Basics: Why Isn’t GNSS Infinitely Precise?
To understand the solution, we have to look at the problem. GNSS receivers use timed signals from satellites to trilaterate their location. However, in the real world, several factors introduce errors, for example:
- Satellite inaccuracies: Satellite orbits are not known to infinite precision, for example due to varying atmospheric drag. Their atomic clocks also have nanosecond jitter and the signal path is subject to varying delay for example due to thermal cycling.
- Environmental influences: Space weather, and Earth-bound weather distort the signal as it travels to the receiver. Additionally the ground under the receiver moves ever so slightly, for example due to gravitational pull from the moon, tidal loading, especially in coastal regions. The movement of the receiver and the antenna placement also has an influence on signal propagation.
To overcome this, a standard GNSS module isn’t enough. First, you need a module that does not only use the transferred GNSS messages, but also measure the signal’s phase. The phase information has higher precision because the carrier signal is of much higher frequency. But to make sense of the phase measurements, you need a reference. The source of this reference and how it’s used is what we’ll look at next.
The Three Main Correction Methods
There are three primary ways to correct GNSS errors using phase measurements, each with its own trade-offs regarding accuracy, convergence time, and infrastructure dependency.
1. RTK (Real-Time Kinematic)
RTK relies on a reference station nearby with a known, exact location. The reference station measures the signal and its phase, and sends that to the receiver.
- How it works: The reference station’s measurement is “subtracted” from the receiver’s measurement, removing all forms of inaccuracies in one go. But the further the receiver is from the reference station, the more the environmental factors differ, so the precision gets worse.
- Accuracy: Best in class (~1-2 cm).
- Constraint: You must be within ~30 km of a reference station. The closer the better.
- Peculiarities: It’s quite easy to operate your own base station.
2. PPP (Precise Point Positioning)
PPP is model based and tries to identify and eliminate all possible interference factors individually. For satellite inaccuracies, PPP relies on precise measurements from ground stations of satellite orbits and their clock deviations. Additionally, it models ground movements and atmospheric influences. The remaining inaccuracies come mostly from tropospheric influence.
- How it works: The PPP algorithm solves a system of equations involving the measurements of the signal and phase, the correction data (orbit, clock, bias), and the influence of the receiver hardware, while recursively estimating the influence of the troposphere and the signals’ cycle ambiguity.
- Accuracy: Good (~10 cm).
- Constraint: Convergence time of a few minutes (~15-30 min) to reach that accuracy.
3. PPP-RTK
As the name suggests, this is a hybrid approach.
- How it works: PPP-RTK follows the same approach as PPP, but combines that with data from a regional network of reference stations transmitting model parameters for ionosphere and troposphere. As a bonus, not having to estimate influences from ionosphere and troposphere eliminates the major contributors to the high convergence time of PPP.
- Accuracy: Better than standard PPP, but generally not as good as pure RTK (~3-5 cm).
- Constraint: PPP-RTK is the “newest” evolution of precise positioning. It can be harder to find (or more costly to use) correction data services and GNSS modules supporting it. There also exist several competing correction data formats (RTCM-SSR, SPARTN, SSRZ) which results in incompatibilities between service providers and modules.
The Data Dilemma: Sourcing Corrections in Germany
All three methods above require a stream of correction data. This is where the situation in Germany can get quite complicated and costly.
RTK: The SAPOS Maze
Generally you have the option to deploy your own RTK reference station, which is useful for municipal application, but doesn’t scale well for larger applications.
For Germany-wide application, you need a correction data provider that operates a nationwide grid of reference stations. These providers usually offer data streams for virtual reference stations (VRS), which require you to send an approximate position first, so the server can forward the correction data stream from the nearest reference station.
In Germany, the primary nationwide service is SAPOS HEPS (“Satellitenpositionierungsdienst – Hochpräziser Echtzeit Positionierungs-Service“). However, it is fragmented into two variations:
1. ZSS-SAPOS (National): The “Zentrale Stelle SAPOS” offers a unified service that works Germany-wide.
- Cost: 10 €/month per device.
2. Regional SAPOS (Federal States): Every federal state runs its own SAPOS instance.
- Cost: Most states provide this for free under Open Data regulations.
- Drawback: Not all states are free. Rhineland-Palatinate charges ~120 €/year per device (the same as ZSS-SAPOS), Bavaria charges ~20 €/year per registration, and Mecklenburg-Vorpommern has a 100 € one-time fee per registration.
Note: If you are deploying a large fleet of devices, you can achieve massive cost savings by connecting to the free regional SAPOS instances, if you can go without servicing Rhineland-Palatinate. There are also other providers, but keep in mind that your application will be dependent on their availability, so free services like RTK2GO are most likely no alternative.
PPP: The Global Player
The key difference of PPP is that it relies on orbit corrections which are globally usable, rather than local reference stations. PPP correction data is also often delivered directly via satellite (L-band), eliminating the need for a separate cellular data stream. This combination makes it indispensable for niche environments, such as the open ocean, deserts, or polar regions, where local infrastructure is non-existent. PPP used to be one of the more expensive approaches, but in 2023 the EUSPA (European Union Agency for the Space Programme) launched the free correction data service GALILEO HAS (High Accuracy Service):
- Cost: Free.
- Benefit: It comes with comparatively low convergence time (~5 min) and is delivered via satellite.
- Drawback: With officially <20 cm, though often better in practice, it’s not as precise as other PPP solutions. It currently “delivers Service Level 1 only with a reduced performance” with “Full Service Declaration targeted in Q1-Q2 of 2027”. It needs a compatible module.
There are other services that provide precise worldwide PPP support, but for general commercial application within Germany, they offer no advantage at low cost compared to the free PPP and PPP-RTK options available. They also are more expensive than RTK with ZSS-SAPOS. Consequently, we excluded them from further analysis.
PPP-RTK: The GEPOS Alternative (Internet or DAB+)
The German BKG (Bundesamt für Kartographie und Geodäsie) and the AdV (Amtliche deutsche Vermessung) provide a free PPP-RTK service called GEPOS (Gemeinsamer amtlicher PPP-RTK-Positionierungsdienst von Bund und Ländern).
- Cost: Free.
- Accessibility: No registration required.
- Benefit: GEPOS is not only available via internet, it’s also broadcast via DAB+, which has good coverage in Germany even in rural areas.
- Drawback: This service uses the niche SSRZ format which is not compatible with any available GNSS module. Instead a closed source converter tool is provided, though only for Android and Windows.
Use-Cases:
Which Path Should You Choose?
Deciding on the right architecture depends heavily on your budget, fleet size, area of operation, and precision requirements. Here is our strategic recommendation. Note that we’re only considering real-time, centimeter-level localization in Germany.
Scenario A:
You need the absolute best precision (1–2 cm)
If you have an internet connection and accuracy is paramount, use RTK.
- Small Area: If you operate in a small area, consider using federal SAPOS or operate your own reference station.
- Large Area + Small Fleet: If you have very few devices, save yourself the engineering headache and buy the ZSS-SAPOS service (10 €/month/device).
- Large Area + Large Fleet: If you have many devices and you are willing to skip coverage in Rhineland-Palatinate, use the federal SAPOS instances. For this to work you will need to write some logic to switch instances based on location, because the federal VRSs will only work with their respective reference stations. (Note: We did not go this path and there are potential unknowns in the federal SAPOS instances and their administration, so test and evaluate properly before going this route.)
Scenario B:
You don’t need the highest precision (~20 cm)
If you only need decimeter level precision, consider using PPP with GALILEO HAS. This is easy to setup and works even without network connectivity. Be aware though that the convergence time is not suitable for all applications.
Scenario C:
You operate in rural areas with poor internet
If you are working in agriculture, forestry, or remote infrastructure you are quite free to choose, because all options work without internet:
- RTK with your own reference station, sending corrections via radio.
- PPP-RTK via DAB+, if you can work around its format conversion issues by using Android or Windows.
- PPP with GALILEO HAS via satellite.
The Verdict
High-precision GNSS in Germany is accessible and affordable, but it isn’t always simple.
While the technology for centimeter-level accuracy is readily available, the most cost-effective paths require navigating a fragmented landscape of regional services or adopting newer hybrid standards like PPP-RTK.
Ultimately, as always, the choice comes down to a trade-off between operational cost and development cost. You can pay for precision and simplicity with a national subscription, or you can invest in development time or hardware to tailor your solution closer to what your business case can live with.
Bonus: Example RTK setup
using federal SAPOS (Berlin) and NEO-F9P module
- Registering on a federal instance: On sapos.de you can find a list of all federal instances. The registration process differs and in some states it’s not even needed. We registered for federal SAPOS in Berlin via the [online form]. It took one business day to receive our login data and another day for the login to be activated.
- Setup the hardware: For testing we used a Holybro H-RTK NEO-F9P Rover. We connected the antenna via a USB-UART-adapter to a development machine running Linux Mint.
- Start the GPS client: We want to use
gpsd, so we checked its man page. Usually it suffices to pass the NTRIP caster URLsapos-be-ntrip, port2101, and VRS mount pointVRS_3_4G_BE; together with login information (USERNAMEandPASSWORD) togpsd, like so:ntrip://USERNAME:PASSWORD@sapos-be-ntrip.de:2101/VRS_3_4G_BE; The full command looks like this:
$ sudo gpsd -n -N -D 4 -F /var/run/gpsd.sock /dev/serial/by-id/USB-UART-adapter ntrip://USERNAME:PASSWORD@sapos-be-ntrip.de:2101/VRS_3_4G_BE 2>&1 | tee gpsd.log
This failed on the first try, becausegpsddoesn’t support the RTCM 3.4 format yet, which this SAPOS instance uses. Luckily RTCM is backward compatible andgpsd; already supports RTCM 3.3, so it was a simple fix that we submitted upstream. Now runninggpsdand checking withgpspipewe got:$ gpspipe -w | grep –line-buffered TPV | jq –unbuffered -c '{status, lat, lon, alt, ecefpAcc, ecefvAcc}'
{"status":2,"lat":52.520843100,"lon":13.409317600,"alt":240.8120,"ecefpAcc":4.5000,"ecefvAcc":0.3100}
{"status":2,"lat":52.520843100,"lon":13.409317500,"alt":240.8080,"ecefpAcc":4.5400,"ecefvAcc":0.3000}
{"status":2,"lat":52.520843100,"lon":13.409317500,"alt":240.8080,"ecefpAcc":4.5400,"ecefvAcc":0.3000}
{"status":2,"lat":52.520843100,"lon":13.409317500,"alt":240.8400,"ecefpAcc":4.5800,"ecefvAcc":0.3100}
{"status":4,"lat":52.520817200,"lon":13.409312700,"alt":239.6610,"ecefpAcc":1.1900,"ecefvAcc":0.3100}
{"status":4,"lat":52.520795700,"lon":13.409304900,"alt":236.4270,"ecefpAcc":0.7700,"ecefvAcc":0.3100}
{"status":4,"lat":52.520788400,"lon":13.409300300,"alt":235.1760,"ecefpAcc":0.6900,"ecefvAcc":0.3000}
{"status":4,"lat":52.520788400,"lon":13.409300300,"alt":235.1760,"ecefpAcc":0.6900,"ecefvAcc":0.3000}
{"status":4,"lat":52.520784300,"lon":13.409298700,"alt":234.6180,"ecefpAcc":0.6600,"ecefvAcc":0.3200}
{"status":4,"lat":52.520783700,"lon":13.409298200,"alt":234.3460,"ecefpAcc":0.6600,"ecefvAcc":0.2600}
{"status":4,"lat":52.520784100,"lon":13.409296900,"alt":234.2430,"ecefpAcc":0.6200,"ecefvAcc":0.1100}
{"status":4,"lat":52.520785200,"lon":13.409297100,"alt":234.1680,"ecefpAcc":0.5500,"ecefvAcc":0.1600}
{"status":4,"lat":52.520785200,"lon":13.409297100,"alt":234.1680,"ecefpAcc":0.5500,"ecefvAcc":0.1600}
{"status":3,"lat":52.520826800,"lon":13.409302700,"alt":239.4330,"ecefpAcc":0.0200,"ecefvAcc":0.1000}
{"status":3,"lat":52.520826700,"lon":13.409302800,"alt":239.4290,"ecefpAcc":0.0200,"ecefvAcc":0.1900}
{"status":3,"lat":52.520826600,"lon":13.409302800,"alt":239.4250,"ecefpAcc":0.0200,"ecefvAcc":0.1900}
{"status":3,"lat":52.520826700,"lon":13.409303000,"alt":239.4270,"ecefpAcc":0.0200,"ecefvAcc":0.1300}
{"status":3,"lat":52.520826700,"lon":13.409303000,"alt":239.4270,"ecefpAcc":0.0200,"ecefvAcc":0.1300}
Interpretation:status(codes aregpsdspecific): 2 = DGNSS, 4 = RTK float, 3 = RTK fix;lat/lon/alt: latitude, longitude, altitude;ecefpAcc: position accuracy estimate in meter;ecefvAcc: speed accuracy estimate in meters per second.4. Verify positioning accuracy: We didn’t do this, but if you really want to know the real accuracy of your solution, you will have to find a public reference station with known coordinates that are more precise than what your solution can provide. In Germany there are public reference points, so called “Geodätische Referenzpunkte” or “Geodätische Grundnetzpunkte“. The location of these is known to the centimeter and they can be used to measure the RTK positioning accuracy. [Wikipedia has a list of all public reference points.][/vc_column_text][/vc_column][/vc_row]
