Measure a Defect's Length Using the Frequency

This post contains a measurement most inspectors do not know is available to them.
You can measure the axial length of a defect from the frequencies you're using.
Here is where that comes from.
A real defect is not a single sharp step. It has a front edge and a rear edge, and both of them reflect. Sometimes those two reflections arrive together and add. Sometimes they arrive opposite and cancel.

We tested an 80 percent deep notch with 6.44 percent cross-sectional area at 32, 64 and 128 kHz. Three frequencies gave three different reflections. But when the notch length is divided by the wavelength, all three data sets fall on one curve.

So the governing quantity is not the defect length, and it is not the frequency. It is the ratio of the two.
The curve peaks when the axial length of the defect is about a quarter of the wavelength.
Now read that backwards.

Test at several center frequencies. Find the one that gives the maximum reflection. The axial length of the defect is then a quarter wavelength of that frequency.

Quarter wavelengths for the torsional mode in steel: 1 inch at 32 kHz, 0.5 inch at 64 kHz, 0.36 inch at 90 kHz, 0.25 inch at 128 kHz. In the field data in this module, one indication peaked at 90 kHz, so its axial length is about 0.36 inch. Another peaked at 64 kHz, so about half an inch. A third peaked at 128 kHz, so about a quarter inch.

That is a defect length obtained from a frequency, and it cannot be obtained at a single frequency at all.

It is the strongest argument I know for operating at multiple center frequencies — not that detection improves, though it does, but that a quantity becomes available which otherwise does not exist.