15.2 Discontinuity Detection and Characterization

The ASNT NDT Level II ultrasonic general exam lists Discontinuity Detection as official UT topic 5 on the current outline (prior to 5 February 2027). Detection is not "I saw a blip, so I found a crack." Detection is whether the echo is relevant, what family it belongs to, and whether the setup could even see the family you are being asked about. Characterization happens in the same motion of the probe. If you freeze one peak and argue about its height, you have not detected a discontinuity. You have photographed a voltage.

An indication is what the instrument displays. A discontinuity is an interruption in the typical structure of the material. A defect is a discontinuity that failed the written acceptance standard. Topic 5 lives in the first two words. Topic 6 (the next section) applies the standard. Candidates who jump from a bright echo to "reject" skip the work this topic tests.

Relevant versus nonrelevant

Use the same three-box discipline the surface methods teach, with ultrasonic causes.

Nonrelevant does not mean "walk away." A geometric collector can hide a crack at the same address. Stamping "just the root" on every half-skip peak without scanning is how incomplete penetration and root cracks leave the job.

Geometry

Predictable surfaces return predictable echoes. A single-V weld root is a corner reflector at a metal path you can calculate from thickness, refracted angle, and skip. An ID roll, a counterbore, a backing bar, a mismatch shelf, a weld cap corner, and the opposite-surface corner of a plate edge all have loci. If the echo sits on that locus, has the echodynamic of that corner, and does not leave it when you scan, it is nonrelevant geometry until something extra appears.

Mode-converted signals

A shear wave that hits a surface near the first critical geometry can convert to a longitudinal wave, a head wave, or a creeping / lateral-wave family. Those paths arrive at times that do not match the simple shear skip you plotted on the screen. Plate (Lamb) modes in thin product do the same. Wedge reverberations and internal probe noise are not mode conversions, but they share the rule: if the time and probe position do not fit a plotted shear path in the part, do not promote the peak to a mid-wall crack. Change angle, change mode, or use a straight-beam confirmation. Mode-converted peaks can be useful (creeping-wave surface work), but they are nonrelevant to the plotted shear gate until you re-identify the path.

Grain noise

Grain noise (backscatter, "grass") is energy returned by grain boundaries and other small impedance steps, especially in coarse austenitic welds, cast stainless, and some nickel alloys. Grass rises and falls as you scan; it does not hold a single geometric locus the way a slag pocket or a crack tip does. It is nonrelevant as a discontinuity, and it is a detection limit (see the last heading in this section). Do not 6 dB-drop a blade of grass.

Planar versus volumetric

After the echo is a relevant candidate, name the shape family. The general exam and every weld procedure use this split because it changes how you scan and how dangerous amplitude is as a number.

A planar reflector is a surface. Lack of fusion sits on a bevel or an interpass face. A lamination sits in the rolling plane. A crack is a plane that may be vertical, branched, or faceted. If your beam glances along that plane, most energy goes elsewhere. That is why a second angle, an opposite-side scan, or a tandem / pitch-catch pair exists: not to "get more gain," but to change incidence.

A volumetric reflector returns energy over a wider range of approach angles. Clustered porosity is the teaching example. Isolated slag can behave in between — elongated slag along a pass can look almost planar. Characterization is a scan, not a single-word guess from one frozen A-scan.

Straight-beam work uses the same split. A lamination is planar and parallel to the surface: pulse-echo lights it as a mid-wall echo and a lost or shortened back-wall. A pore field is volumetric: mid-wall hash with a still-visible back-wall if the field is not dense enough to block the path.

Orientation versus amplitude

This is the sentence topic 5 exists to extract:

Amplitude is not size when orientation is unknown.

A small side-drilled hole or a small favorably oriented lack-of-fusion facet can plot above DAC. A large, smooth, poorly oriented crack can plot below the recording level. Candidates who treat %DAC as a crack-depth scale fail characterization items and then fail evaluation items in the next section.

Practical consequences:

• Do not rank two indications as "worse" and "better" from height alone.
• Rotate the problem: change angle, change skip (half-skip versus full skip), change side, or add a complementary technique the procedure allows.
• A planar echo that peaks hard in a narrow probe window and dies when you rock a few degrees is behaving like a mirror. A volumetric echo that stays in the gate over a wider rock is not a knife-edge plane.
• Recording "amplitude only" without orientation, skip, and scan direction is incomplete detection, even before anyone talks about accept or reject.

Echodynamics as the probe is scanned

Echodynamics is the envelope of amplitude (and sometimes of peak time) as you scan and orbit the probe. It is how a Level II tells a geometric corner from a crack that shares that corner's postcode.

Scan along the weld (or along the indication) to see length behavior. Scan transverse (toward and away from the weld) to see whether the peak sits on the plotted root, on a bevel, or in weld metal. Orbit or skew the probe to see whether the reflector is a line-like plane or a more compact body.

Typical patterns the exam expects you to recognize:

• Geometric corner (root bead, counterbore, plate edge). Amplitude rises smoothly to a peak at the predicted metal path and probe stand-off, then falls symmetrically as you leave that locus. Peak time stays where the plot says the corner is. The echo does not persist when you move the beam off that corner onto parent metal.
• Planar crack or lack of fusion. Often a sharper peak, a longer persistence as you scan along, a locus that extends beyond the geometric feature, or a second peak from a tip. Time may walk as the beam rides up a bevel. Amplitude can collapse with a small skew if the plane is smooth.
• Volumetric slag or porosity. Broader transverse envelope, more "hash," less of a single knife-edge peak, often still visible after a modest skew.
• Grain grass. A shifting forest with no repeatable peak at a fixed metal path when you reseat and rescan the same millimetres.

Echodynamics are data. Write what you did with the probe, not "it looked cracky."

Root geometry versus crack

Weld-root echoes are the most expensive mix-up in angle-beam work. A single-V or single-bevel root is a built-in corner. Incomplete penetration, a leftover land, and a root crack can occupy the same half-skip window.

Work the distinction in this order:

1. Plot the geometric root. Thickness, refracted angle, and skip give a metal path and a surface distance. If the echo is not on that plot, it is not "the root."
2. Watch echodynamics. A neat, symmetric peak that dies when you leave the root stand-off is geometry. A peak that continues along the weld past a stop-start, or that stays up when you move the beam slightly into the weld volume, is a candidate discontinuity.
3. Look for extras. A tip-diffracted echo above the root, a second peak at a different depth, or an echo from the opposite side that does not match the root bead, supports a crack or incomplete penetration rather than the bead itself.
4. Use a complementary view. A different angle, a full-skip approach, an ID creep / surface wave if the procedure allows it, or a straight-beam from the opposite face on accessible geometry.
5. Do not grind first. Characterization is not "remove metal until the echo goes away." If the procedure later allows a blend within remaining-thickness rules, that is a documented repair path, not a detection shortcut.

A stem that says "strong echo at the calculated half-skip root location, symmetric echodynamic, no extension, no tip" is asking for root geometry. A stem that says "echo at the root distance that persisted 18 mm past the weld stop and showed a second, shallower peak" is asking for a relevant planar call — crack or incomplete penetration — to be evaluated in topic 6.

Cap geometry, mismatch, and backing bars get the same treatment: predict, scan, look for the extra locus.

Coarse grain as a detection limit

Cast stainless, coarse austenitic weld metal, and some duplex or nickel alloys scatter and absorb shear waves aggressively. Two things happen at once:

• Attenuation. The back-wall, the DAC reference, and any deep reflector lose amplitude faster than carbon-steel calibration predicted. A discontinuity that would have been 80% DAC in fine-grain steel may sit in the noise.
• Backscatter (grass). The baseline fills with grain echoes. The signal-to-noise ratio collapses. Small planar facets hide in the grass.

That combination is a detection limit, not a character defect of the operator. Correct Level II responses:

• Lower frequency and, when the procedure allows, a larger crystal to put more energy into the metal and reduce scatter relative to wavelength.
• Prefer dual-element or transmit-receive longitudinal (TRL) techniques on austenitic welds when the written procedure names them. Straight-beam longitudinal often survives better than a high-angle shear path in cast stainless.
• Recalibrate on representative material. A IIW block in fine-grain carbon steel does not set a valid DAC for coarse cast stainless.
• If signal-to-noise still cannot support the required sensitivity, declare the volume uninspectable with that setup. That is a competent detection statement. Cranking gain until grass crosses the recording level and then rejecting "a hundred indications" is not.

Do not tell the exam that coarse grain "only matters to radiography." Do not switch to the highest frequency so the wavelength is shorter than the grains — that usually makes grass worse. Detection failed when the discontinuity you care about cannot stand above noise. Characterization never starts.

Realistic exam scenarios

An angle-beam peak sits exactly on the plotted counterbore corner, rises and falls symmetrically as you scan transversely, and vanishes when you move off that stand-off. Classification: nonrelevant geometry. If a second peak stays in the weld after you leave the corner, that second peak is the relevant candidate.

A 70° shear scan of a single-bevel weld shows almost nothing; a 45° scan from the same face lights a long, orientation-sensitive echo on the bevel. Characterization: planar, likely lack of fusion, missed by the first angle because of orientation, not because the weld is sound.

Cast stainless at 4 MHz shear is all grass; the back-wall is gone. Next action: lower frequency / change mode per procedure, restandardize on representative material, and stop evaluating grass. If the volume still cannot meet sensitivity, it is not inspected.

Topic 5 language is short: relevant or not, planar or volumetric, what the probe did when you moved it, and whether the metal would even let you see the target.