A theodolite's angular accuracy spec, expressed in seconds of arc, is the single number that defines its class: 1, 2, 5, 9, and 20-second tiers cover virtually every instrument on the market, and each step roughly halves the angular error at the target [S1][S4].
For traverse work, control surveys, and structural alignment, the published angular rating directly drives the horizontal-distance error contribution, because an angular error of N arc seconds translates into roughly 5 mm per 100 m per N arc second at the target [S3][S6].
What the "seconds of arc" rating actually means
One second of arc equals 1/3600 of a degree, or about 4.848 microradians, and at 100 m sight distance a 1-arc-second instrument resolves a lateral displacement of roughly 0.5 mm against the target [S3]. Modern digital theodolites quote the manufacturer-rated standard deviation of horizontal angle measurement in this unit, commonly displayed as "1''", "2''", "5''", "9''", or "20''" on the spec sheet [S1][S4][S7].
High-quality theodolites reach 1 arc second of resolution, putting geodetic-class instruments into the same angular tier as first-order theodolites of the pre-electronic era, while surveying-grade electronic units routinely cluster between 2 and 5 arc seconds [S7]. Entry-level construction units publish 9-arc-second to 20-arc-second ratings and are intended for rough site layout rather than control [S1][S6].
The 2-arc-second tier is the most common "professional workhorse" rating, sitting between the 1-arc-second geodetic class and the 5-arc-second construction class; a 2-arc-second instrument gives materially tighter measurement tolerance than a 5- or 10-arc-second unit and is the typical spec quoted for engineering and cadastral work [S4][S6].
Accuracy tiers and the projects they match
Most theodolites sold today are digital, with internal absolute encoders replacing the vernier scales used on legacy transits, and the published seconds-of-arc rating comes from those encoder resolutions combined with the compensator and circle-reading electronics [S2][S5]. Old analog vernier theodolites typically read to "less than a minute of arc" (i.e. 30 to 60 arc seconds), which is why modern digital units at 5 arc seconds already represent a 6x to 12x improvement over historical mechanical instruments [S2].
Selection bands: a 1-arc-second theodolite fits first-order geodetic control, deformation monitoring of dams and tunnels, and high-precision alignment of accelerator and turbine shafts; a 2-arc-second unit covers second-order traverse, cadastral boundary, bridge and tower alignment, and most engineering surveys; a 5-arc-second unit handles general construction layout, building column line setting, and topographic detail; a 9-arc-second to 20-arc-second unit is the budget tier used for rough grading, residential layout, and reconnaissance [S1][S4][S5][S6][S7].
Inside a tier, optical plummets, dual-axis compensators, and absolute encoders versus incremental encoders all move the real-world performance, but the spec sheet is still anchored on the seconds-of-arc figure, and that single number is what the surveyor uses to budget angular error in a traverse or alignment network [S5].
How the angular rating propagates into field error

Angular accuracy drives distance-proportional error: a 5-arc-second theodolite sighting a target 200 m away can carry about 5 mm of lateral uncertainty from the angle alone, and that error grows linearly with sight length, so long sights amplify the penalty of a coarse angular rating [S1][S3]. For this reason, deformation-monitoring networks, where sight lines often exceed 300 m and the target displacement to be detected is sub-millimetre, demand 1-arc-second or better instruments, not 5-arc-second units that would bury the signal in angular noise [S7].
For traverse work, the spec sheet's angular rating also governs the allowable angular closure per leg, with typical practice accepting roughly 2 to 5 arc seconds per station times the square root of the number of stations, so a traverse of 10 legs on a 2-arc-second instrument should close to better than about 16 arc seconds of angular misclosure [S1]. Vertical-angle accuracy is published separately but tracks the horizontal rating, often within the same order, because both axes share the encoder and compensator stack [S2][S3].
Setting, levelling, and centring errors are independent of the rated angular accuracy and can swamp it: a 1-arc-second theodolite set up over a point with 5 mm centring error loses its angular advantage on long sights, which is why the optical plummet, the tribrach, and the three-screw levelling sequence are treated as part of the angular accuracy budget in field practice [S5].
Digital theodolite vs total station vs GNSS: where pure angle still wins
GNSS RTK delivers 8 mm to 20 mm horizontal accuracy in real time, which is fine for topographic and cadastral work but cannot replace a 1-arc-second theodolite for tasks that demand sub-arc-second angular control along a fixed reference line, such as plumbing a tower or aligning a turbine shaft [S5]. For pure structural and industrial alignment, where the reference is a known azimuth rather than a coordinate, a digital theodolite set on that azimuth is the right tool, not a total station or RTK receiver, because the angle-only workflow is simpler and the instrument is cheaper than a robotic total station for the same job [S1][S5].
A theodolite covers angle-only work, while a total station adds an electronic distance meter (EDM) and onboard computer for combined angle-plus-distance, which is the right call when the deliverable is a stakeout point or a coordinate file rather than an alignment [S1][S2]. Optical theodolites with internal glass circles and an electronic readout sit between the two: same angular tier as a digital theodolite, but without the EDM and software stack, and that is the natural fit for structural and industrial alignment crews who already have a separate distance workflow [S2][S5].
What "1 arc second" really costs and what the spec does not cover

Pricing tracks the seconds-of-arc rating: a 1-arc-second geodetic digital theodolite costs several times more than a 5-arc-second construction unit, and a 20-arc-second entry-level model is the cheapest tier on the market, so the spec is also a price band [S1][S6][S7].
The published rating is the manufacturer-quoted standard deviation under DIN or ISO test conditions, usually ISO 17123-3, and field accuracy is typically worse because of atmospheric refraction, sight-line obstruction, target centring, and operator-dependent collimation; if a job is on the boundary of two tiers, specify the tighter one rather than the looser, because the spec is a one-sigma number and the 95% tail is roughly twice as large [S3][S4].
Two practical signals to watch: first, whether the spec sheet publishes horizontal and vertical angular accuracy separately (a real spec, not a single combined number), and second, whether the instrument carries a dual-axis compensator with a published operating range (typically ±3' to ±5'); both are quick markers of an instrument that will actually deliver its rated seconds of arc in the field [S2][S3][S5].
For deeper selection criteria on adjacent precision instruments, see Single-Phase vs Three-Phase Power for Stud Welders: Spec-Driven Selection, which uses a similar spec-first selection frame, and the broader theodolite reference page for the underlying instrument geometry.
The underlying component specifications are covered under arc welder, and precision filter.