GNSS / GPS Surveys

Every method in this guide — static, RTK, PPK, OPUS — is the same underlying measurement dressed up differently. The receiver times signals from satellites and turns those times into ranges. The difference between a $200 phone chip and a $10,000 rover is which part of the signal they use and what they do about the errors.

• Code positioning (~2–5 m). The receiver matches the coarse code stamped on the signal. This is phone-grade autonomous GPS. Fine for navigation, useless for measurement.
• Carrier-phase positioning (~cm). The receiver counts wave cycles of the carrier itself — a 19 cm wavelength it can resolve to a few millimetres. The catch: it doesn’t know the whole number of cycles between satellite and antenna. Solving that unknown is called fixing the ambiguity.
• Differential correction. Most error sources — satellite clock drift, orbit error, atmosphere — are nearly identical for two receivers tens of kilometres apart. Put one receiver on a known point and the errors cancel out of the other. Every survey-grade method is some flavor of this trick.

That last bullet is the whole game. One receiver = navigation. Two receivers (or one receiver + a correction network) = surveying.

Pick by what the point is for. Control that everything else hangs on deserves hours of observation; a topo shot that gets averaged into a surface deserves seconds.

• Static (mm–cm, hours). Receiver sits on a tripod over the point and logs raw data for 30 minutes to several hours. Post-processed against other stations. The most accurate method there is — this is how control networks, calibration baselines, and CORS stations themselves get coordinates.
• OPUS (cm, 2+ hours + upload). Static’s no-second-receiver shortcut, for the US: log 2+ hours, upload the RINEX file to NGS’s free OPUS service, get back NAD83 coordinates computed against the national CORS network. The standard way to put real coordinates on a project base point.
• RTK — real-time kinematic (cm, seconds). Your base on a known point radios corrections to your rover; the rover fixes ambiguities on the fly and reads centimetre positions in real time. Needs a radio link and a base within ~10–20 km.
• Network RTK / NTRIP (cm, seconds, no base). Same idea, but the "base" is a state or commercial network of CORS stations streaming corrections over the internet. No base to set up, no radio. The full how-to — mountpoints, VRS, your state’s network — is in our NTRIP & RTK guide at /free/ntrip-guide.
• PPK — post-processed kinematic (cm, after the fact). Log raw data on both base and rover (or drone), solve the ambiguities in software afterwards. No radio link to drop, no corrections to stream — the favorite for drone mapping, where the "rover" is moving at 15 m/s and a dropped link would hole the dataset.

Here's the full RTK loop run on real ground — base setup on a known point, radio link, rover initialization, and what the screen tells you while it happens.

The data collector reports a solution status and a precision estimate on every shot. Knowing what they mean is the difference between measuring and guessing.

• FIXED. Ambiguities resolved to whole numbers — the receiver genuinely knows the cycle count. Centimetre confidence. The only status you store a shot under.
• FLOAT. Ambiguities still estimated as decimals. Position can wander a few decimetres while looking stable. Don’t store control or stakeout under float — wait, re-initialize, or move for better sky.
• PDOP. Geometry quality of the satellites in view — lower is better. Under 2 is excellent, over 4–5 means the satellites are bunched and the solution is weak in some direction. Mask angle (typically 10–15°) trims unhealthy low-elevation signals at the cost of geometry.
• RMS / precision estimates. The receiver’s own opinion of shot quality. Treat the on-screen number as an optimistic floor, not a guarantee — it can’t see a multipath reflection or a wrong antenna height.
• Baseline length. Error grows with distance from the correction source — roughly 1 ppm on top of the fixed error (1 mm per km). A 30 km single-base RTK shot is not the same quality as a 3 km one, even if both say FIXED.

• Multipath. The signal bounces off a building, truck, or chain-link fence before reaching the antenna, lengthening its path. The receiver can’t tell. Symptoms: a FIXED solution that disagrees with itself shot to shot near reflective surfaces. Cure: move, or wait for different geometry.
• Canopy and obstruction. Trees attenuate and scatter the carrier. You may keep a float solution that looks tantalizingly close to fixing. Under real canopy the honest tool is a total station traverse from a GNSS pair set in the open.
• Wrong antenna height. The classic. A 2.000 m pole logged as 1.800 m moves every elevation by exactly 20 cm, and nothing on the screen will ever flag it. Measure the pole, photograph the setup, and check a known point before production work.
• Wrong fix. Rare with modern multi-constellation receivers, but a solution can fix to the wrong integer and report centimetre confidence while sitting decimetres off. This is why control gets redundant occupations (and why stakeout checks into existing monuments before trusting the day’s setup).

• 1. Anchor the project. OPUS or static on a base point (or hold published control). Write down what datum and epoch the coordinates are in — future-you will need it.
• 2. Check before you measure. Shoot a known monument first. If the check is off by more than the error budget, stop and find out why — base coordinates, antenna heights, datum mismatch.
• 3. Measure with the right method. Control: static or repeated RTK occupations. Topo: RTK/NTRIP singles. Drone ground control: RTK/PPK with two occupations on each GCP.
• 4. Close the loop. Re-shoot the check point at the end of the session. Bracketing the work between two good checks is what makes the whole day defensible.

For the network-corrections side of this — NTRIP credentials, mountpoints, VRS vs single-base, and a directory of every US state network — continue with the NTRIP & RTK guide at /free/ntrip-guide.