ControlPoint Manual — versioned with the code; every worked example below is a REAL corpus job and its numbers are pinned by the test suite.
Tape transfers (dh_tape)
What it does
A tape drop is just another height-difference observation — so it adjusts in the SAME network as the levelling, no side-calculator. Re-type a leg as tape in the legs table (or import it as a DH kind=tape record) and it gets the tape stochastic model σ² = a² + (b·L)² (steel defaults a = 0.5 mm, b = 30 ppm, adjustable) instead of √K. The report then shows per-type error factors, prints every tape correction applied, and labels any loop that crosses a tape leg — a mixed loop misclosing is the direct check on the tape.
Worked example — the 2HPX shaft
The champion job IS a two-loops-plus-drop network: six legs to marks named TAPE1…TAPE8 drop 5–8 m over 5–22 m of plan — impossible as staff shots; they are the shaft tape transfers. Typed as tape, the whole job adjusts as one network: level factor 0.797, tape factor 1.123, and three mixed loops (each through two tape legs) all close under 1 mm — the tapes were good.
Theory in one paragraph
A suspended tape's error grows with its length (calibration, temperature, catenary), not with the square root of a distance walked — hence a² + (b·L)² rather than σ²km·K. Giving each observation type its own honest sigma model is what lets one adjustment carry both and still tell you which type misbehaves.