In short
- USIBD LOA has five bands: LOA10 5–15 cm down to LOA50 0–1 mm, each holding for 95 % of points (2σ).
- Specify measured accuracy (the survey) and represented accuracy (the model) separately.
- Survey accuracy combines instrument, registration and control: 2 × √(σi² + σreg² + σctrl²).
- A 1.9 mm scanner with 2 mm registration and 2 mm control gives about 6.8 mm: LOA30, not LOA40.
The five LOA bands
| Level | Range (95 %) | Typical use |
|---|---|---|
| LOA10 | 5–15 cm | Massing, early feasibility, site context |
| LOA20 | 15 mm–5 cm | Space planning, cable routes, SLAM and drone surveys |
| LOA30 | 5–15 mm | As-built models, MEP coordination, retrofit design |
| LOA40 | 1–5 mm | Prefabrication, busbar and pipe spools, facade brackets |
| LOA50 | 0–1 mm | Industrial metrology, machine alignment |
Ranges from the USIBD LOA specification v3.0; v3.1 is the current release. Tolerances apply in both directions. See the accuracy section.
Measured vs represented accuracy
Measured accuracy is how close the registered points are to their true position. Represented accuracy is how close the modelled elements are to the points. A wall modelled as perfectly flat and plumb can be 20 mm away from a real wall that bows, even if the scan was millimetre-accurate. Specify both, and check the second with model-to-cloud deviation reports.
The error budget
- Instrument: grows with range; quoted values like "1.9 mm at 10 m" assume good targets and angles.
- Registration: how well scans are joined; target-based registration beats cloud-to-cloud on long, repetitive spaces such as data halls.
- Control: how well the cloud ties to the project grid; a total station network keeps it small.
Can this survey claim LOA40?
| Case | Instrument | Registration | Control | 2σ | Band |
|---|---|---|---|---|---|
| Cloud-to-cloud, light control | 1.9 mm | 2 mm | 2 mm | 6.8 mm | LOA30 |
| Targets and total-station control | 1.9 mm | 1 mm | 1 mm | 4.7 mm | LOA40 |
The scanner is the same in both cases; the method decides the band. That is why the datasheet value alone never proves an LOA.
Point spacing: can the scan see the feature?
Accuracy is not resolution. To model a feature reliably, point spacing at that range should be at most about half the feature size. A 6 mm setting at 10 m gives 6 mm spacing there and 12 mm at 20 m, enough for a 25 mm pipe at 20 m but not a 10 mm bolt. Closer setups or a finer setting fix it, at the cost of time and data.
Choosing an LOA
- Space planning, cable routes, quick as-builts: LOA20.
- Retrofit design and MEP coordination: LOA30.
- Prefabricated busbar, pipe spools, CDU connections, facade brackets: LOA40.
Ask for the tightest band only where it is needed; LOA40 everywhere multiplies setups and control work.
Questions about accuracy
What does LOA30 mean?
Measured or modelled points lie within 5–15 mm of their true position 95 % of the time. LOA40 tightens that to 1–5 mm.
Is the scanner's datasheet accuracy the survey accuracy?
No. Registration and control add their own errors; the combined 2σ decides the LOA.
How is accuracy checked?
By independent check points measured with a total station and compared with the cloud, and by model-to-cloud deviation reports for the model.
Does SLAM reach LOA30?
Typically not on its own; 1–3 cm is common, which is LOA20. Control points help, but prefabrication tolerances call for a terrestrial scanner.
Sources
- USIBD Level of Accuracy specification v3.0 (v3.1 current).
- Error budget as used by the accuracy model on this site; example instrument value 1.9 mm at 10 m.