Every thickness reading is only as good as the velocity behind it. The velocity of the actual material — not the 5900 m/s in the textbook — is what converts time into millimetres, and it changes with alloy, product form, grain structure and temperature. Cast material, austenitic stainless, clad plate and weld metal can all sit several per cent away from wrought carbon steel.

A single timed echo cannot separate velocity from probe zero: both are unknowns in TOF = 2t/c + t₀. Timing two different known thicknesses removes the zero, because the offset is common to both. Subtracting the two equations gives c = 2(t₂ − t₁) / (TOF₂ − TOF₁), and substituting back recovers the zero. This is the two-point calibration every thickness gauge asks for, and it is why the two steps must be well separated — a small thickness difference divides a small time difference and amplifies the timing error.

Use the same probe, cable, couplant and surface condition as the inspection, and take the timings between corresponding points on the waveform — flank to flank or peak to peak, never a mix. If the material and the calibration block are not the same, the velocity derived here belongs to the block, not the component.

Compression wave

Worked example

Calibration method	two_point
Thin step thickness	10 mm
Time of flight on thin step	3.59 µs
Thick step thickness	50 mm
Time of flight on thick step	17.15 µs
Known probe zero (one-point mode)	0 µs
Nominal velocity for comparison	5900 m/s
Measured compression velocity	5899.7 m/s
Measured velocity	5.8997 mm/µs
Probe zero offset	0.2 µs
Deviation from nominal velocity	-0.005 %

40 mm of extra steel (50 − 10) is crossed twice in 13.56 µs (17.15 − 3.59), so c = 2 × 40 / 13.56 = 5.8997 mm/µs = 5899.7 m/s. Back-substituting on the 10 mm step: the metal path takes 2 × 10 / 5.8997 = 3.390 µs, so the zero is 3.59 − 3.390 = 0.200 µs. The velocity is 0.005 % below the 5900 m/s nominal.