Float vs Fix at a glance
Float — Accuracy: 10 cm–1 m horizontal
- Corrections received but not fully resolved
- Ambiguities treated as real numbers, not integers
- Position jumps of 10–50 cm are normal
- Never adequate for precision survey
- Often a stepping stone toward Fix
- Can look like Fix on some displays
Fix — Accuracy: 1–3 cm horizontal, 2–5 cm vertical
- Carrier phase ambiguities fully resolved to integers
- Position is stable and repeatable
- Centimetre accuracy maintained at speed
- Required for precision survey and stakeout
- Takes 10–60 seconds in good conditions
- Shown in green on most field software
What makes Fix different — ambiguity resolution
To understand Float and Fix you need to understand one concept: carrier phase ambiguity. It is the reason RTK can achieve centimetre accuracy at all — and the reason Float cannot.
GNSS receivers measure position in two ways. The simpler method is pseudorange — measuring the travel time of a satellite signal to estimate distance. Pseudorange gives accuracy of 1–3 metres. The more precise method is carrier phase — measuring the phase of the satellite's radio wave at the receiver antenna. The carrier wave has a wavelength of about 19 cm (for GPS L1). By tracking how many whole wavelengths fit between the satellite and the receiver, and precisely measuring the fractional part, the receiver can measure distance to millimetre precision.
The problem: the receiver knows the fractional part of the carrier phase precisely, but it does not know how many whole wavelengths there are between it and the satellite. This unknown integer number is called the carrier phase ambiguity — or simply the integer ambiguity.
Float
Fix
Carrier phase ambiguity
The integer cycle count is still unknown and treated as a real-valued estimate.
The integer cycle count has been confirmed as a specific whole number.
Position accuracy
10 cm–1 m. The fractional phase is measured, but the unresolved cycle count limits accuracy.
Centimetre precision. The resolved cycle count unlocks the precise fractional measurement.
Resolving the integer ambiguity is what RTK does. Using corrections from the reference station — which has its own precisely known position — the rover can cross-check its carrier phase measurements against the reference and mathematically determine the correct integer values. When the receiver is confident it has the right integers for all tracked satellites, it declares RTK Fixed.
Float means the receiver is still working on this. It has an estimate of the integers — good enough to give sub-metre accuracy — but not yet confident enough to fix them to specific integers. Float accuracy depends on how good the estimate is: anywhere from 10 cm to 1 m, with occasional larger jumps.
The full solution progression
When you connect to an NTRIP service and power up in the field, your receiver moves through several solution types before reaching Fix.
No fix — Accuracy: none
The receiver has no satellite lock and no valid position. It is still acquiring signals. This usually takes 15–60 seconds after power-on outdoors.
Single — Accuracy: 2–5 m
The receiver has satellite lock and a valid position, but no corrections are applied yet. It uses pseudorange only. This is the type of position reported by standard GPS apps on your phone.
DGPS / SBAS — Accuracy: 0.3–1 m
Differential corrections are applied, but only to pseudorange measurements. It is better than Single, but not accurate enough for precision work. This can appear briefly when NTRIP corrections first arrive.
Float — Accuracy: 10 cm–1 m
Carrier phase corrections are applied, but integer ambiguities are not yet resolved. The position is significantly better than Single but not precise enough for survey work. Float is often a transitional state lasting 10–60 seconds before Fix.
Fixed — Accuracy: 1–3 cm horizontal, 2–5 cm vertical
Integer ambiguities are resolved and full RTK accuracy is achieved. This is the solution type required for precision survey, stakeout, machine guidance and drone mapping with ground control points.
Which solution is good enough for what
Single
Float
Fix
Rough navigation — Finding a plot or general location
Good enough
Good enough
Good enough
Drone mapping — Direct georeferencing without GCPs
No
Marginal
Required
GCP collection for drone mapping
No
No
Required
Precision agriculture — Auto-steer with 2–5 cm row guidance
No
Marginal
Preferred
Survey — Topographic surface mapping
No
No
Required
Survey — Cadastral or legal property boundaries
No
No
Required + verification
Stakeout to 1 cm
No
No
Required
Machine control — Earthworks
No
Sometimes
Required
How to get from Float to Fix faster
Float is a transitional state. In good conditions it lasts 10–30 seconds. In challenging conditions it can persist indefinitely. These measures help most:
Go outside with a clear sky view
Ambiguity resolution requires a strong, stable signal from many satellites simultaneously. Even partial canopy cover significantly slows initialisation.
Stand still during initialisation
The RTK engine converges faster when the antenna is stationary. Once Fixed, you can move normally. Collecting data while still in Float can delay Fix.
Use a VRS mountpoint
Shorter effective baselines make ambiguity resolution faster and more reliable. Beyond 20 km from a physical station, switch to VRS and enable GGA.
Enable all satellite constellations
More satellites provide more measurement redundancy. Adding GLONASS, Galileo and BeiDou can halve Float-to-Fix time compared with GPS-only.
Check the NTRIP connection
Verify that bytes per second is non-zero. A dropped and reconnected correction stream resets initialisation and can leave the receiver stuck on Float.
Check PDOP
Above 4, satellite geometry is poor and ambiguity resolution may not converge. Wait for the geometry window to improve, typically within 15–30 minutes.
False Fix — the hidden danger
The most dangerous scenario in RTK is a false Fix — the receiver declares Fixed but has resolved the ambiguities to the wrong integers. The position looks centimetre-precise and stable, but it is wrong by one or more carrier wavelengths (19 cm per L1 cycle).
False Fix produces errors of exactly one or more multiples of the carrier wavelength — 19 cm, 38 cm, 57 cm and so on. It is more common at long baselines, in high multipath environments and during solar storms when ionospheric noise is high.
How to detect a false Fix
Always verify on a known control point at the start of any precision survey. Set up over a point with published coordinates and compare your measured position with the known values. A discrepancy of exactly around 19 cm, 38 cm or 57 cm in any direction strongly indicates a false Fix. Disconnect, move to open sky, reconnect and re-initialise before continuing.
Fix quality indicator — ratio
Many receivers and field software applications report a Fix quality ratio alongside the solution type. A ratio above 3.0 indicates high confidence. A ratio between 1.5 and 3.0 means the Fix is tentative, so verify it on known points. A ratio below 1.5 may indicate that the receiver should not have declared Fix at all.
