How to Read a Flaw Detector Screen: A Field Inspector's Guide

Learning how to read a flaw detector means reading its A-scan: calibrated sound path runs across, echo height in percent of full screen height (FSH) runs up. An echo's position gives sound path, depth and surface distance. Its height, compared with a reference set at 80% FSH on a 1.5 mm hole, decides whether it is recorded.

Everything else on the screen, from gates and DAC curves to dB readouts and alarm lights, is there to make those two readings, position and height, repeatable. This guide works through the setup that makes the screen trustworthy, what each part of the display means, how to convert a reading into a location in the part, how amplitude is judged against DAC and against the AWS D1.1 rating, and what a rail flaw screen looks like when a 70° probe finds a transverse defect or the 0° channel loses its back-wall echo.

What does each part of a flaw detector screen show?

A conventional ultrasonic flaw detector shows an A-scan: echo amplitude plotted against time, with the time base converted to sound path once velocity and zero are calibrated. It does not display a picture of the flaw. Position in the part is worked out from the sound path and the probe angle, and size comes from scanning and measuring, not from the screen alone.

The screen has a small number of elements, and an inspector should be able to name every one before scanning.

Gain is measured in decibels, and 6 dB doubles or halves the echo amplitude, not its energy; halving energy is 3 dB. A rise of 20 dB is ten times the amplitude. Our guide to UT fundamentals: A-scan, B-scan and C-scan covers the display formats in more depth, and digital vs analog flaw detectors explains what the digital readouts add.

Key takeaway: horizontal position is distance along the beam, vertical height is amplitude relative to a reference, and every readout depends on the values you entered.

How do you set up a flaw detector before reading it?

Setup fixes the probe, frequency, couplant and surface condition the written procedure calls for, and confirms that the instrument itself is working, before any calibration is done. A screen read with the wrong probe angle entered, or through a surface that couples badly, gives precise-looking numbers that are wrong.

Start with the probe and cable. A loose connector or a damaged cable shows up as baseline noise or echoes that come and go when the cable is flexed, and that noise can look like an indication. Flex the cable with the probe on a block and watch the screen before you trust any reading.

Frequency follows the procedure. Higher frequency gives shorter wavelength and better resolution but more attenuation and scatter; lower frequency penetrates coarse-grained or thick material but resolves less. Structural weld UT to AWS D1.1 has traditionally used angle-beam probes of about 2 to 2.5 MHz, while thin-wall work and thickness gauging use higher frequencies. Probe size sets the near field, and readings taken inside the near field fluctuate; our article on near field vs far field in UT shows how to work it out for straight and angle-beam probes.

Surface preparation follows the procedure, not a blanket rule. Procedures normally require removal of weld spatter, loose scale, loose or flaking paint, dirt and grease so that the probe couples evenly. Tight mill scale and thin, well-bonded paint are often allowed to remain when the procedure permits it, and some procedures allow scanning through coatings only if a transfer correction is measured. Rust that has loosened, heavy pitting and thick coatings are what usually force grinding or cleaning. If the procedure is silent, ask the Level III who wrote it, and record what the surface looked like.

Couplant also follows the procedure: water, gel, oil or glycerin-based products are all used, with high-temperature couplants for hot surfaces. Use the same couplant on the calibration block as on the part.

Key takeaway: set up exactly as the procedure says, and check the instrument and cable before calibrating, because setup errors don't show themselves on the screen.

How is a flaw detector calibrated on the IIW block?

Calibration sets the time base (zero and velocity), confirms the probe's exit point and refracted angle, and then sets a sensitivity reference on a known reflector. The block most often used for angle-beam setup is the IIW Type 1 (V1) block.

The IIW V1 block does not contain a ladder of flat-bottomed holes, and it does not have reflectors at 20, 40 and 60 mm. In its metric form it is a steel block 300 × 100 × 25 mm with these features, as described on maker product pages:

The current block specification is ISO 2400:2025. ASME Section V work usually adds a basic calibration block of the same material and thickness range as the part, with side-drilled holes at fixed fractions of thickness, so that a DAC can be built.

The calibration steps for an angle-beam probe, in the usual order:

1. Beam index point. Peak the echo from the 100 mm radius and mark the wedge where it lines up with the block's center mark.
2. Zero and velocity (range). Use two known sound paths, such as the 100 mm radius and a second radius echo, so the instrument can solve both the wedge delay and the shear velocity. Steel's shear velocity is about 3,250 m/s (0.128 in./µs); calibrate rather than assume it.
3. Refracted angle. Peak the echo from the 1.5 mm or 50 mm hole and read the angle from the scale on the block, or calculate it from the probe position. AWS D1.1 requires the measured angle to be within ±2° of nominal.
4. Sensitivity reference. Peak the reference reflector and set it to the procedure's reference height. Record that gain as the reference level.
5. Check. Repeat the reference check at the intervals the procedure sets (start, end, change of operator, and at a maximum interval during the shift) and after any knock or power interruption. If the reference has shifted beyond the procedure's tolerance, re-examine the welds tested since the last good check.

Instrument performance, as opposed to calibration for a particular job, is checked separately. ASTM E317 covers this and is current: the store lists E317-21 as the active edition. Its scope covers horizontal and vertical linearity, resolution, sensitivity and noise, and the accuracy of calibrated gain controls. It describes the purpose this way:

"...by periodic repetition, for detecting long-term changes in the characteristics of a given instrument or system that may be indicative of impending failure..." — ASTM E317-21, Scope 1.1

E317 is not a calibration certificate standard, and it sets no acceptance limits of its own. Electronic measurements made with oscilloscopes and signal generators belong to its companion, ASTM E1324 (active edition E1324-21R26). Our UT calibration guide for rail inspection RFPs covers what to ask a contractor to show.

Key takeaway: the IIW V1 block calibrates distance, exit point, angle and a 1.5 mm hole reference; DAC needs a separate block with holes at several depths.

How do you read sound path, depth and surface distance?

Read the sound path at the leading edge or peak of the gated echo, as the procedure specifies, then convert it with the refracted angle: depth = sound path × cos θ and surface distance = sound path × sin θ. In the second leg, after the beam has bounced off the far surface, depth = 2t − sound path × cos θ, where t is the thickness.

First decide which leg the echo is in. The leg sound path is t ÷ cos θ. An echo shorter than that is in the first leg; one between one and two leg lengths is in the second leg. A depth that comes out greater than the plate thickness means the leg was misread, the angle is wrong, or the thickness entered is wrong.

Worked example: two indications on a 1 in. butt weld with a 70° probe

A 1.00 in. thick CJP butt weld is scanned with a 70° shear-wave probe. The screen range is 10 in. of sound path, so each of the ten major divisions is 1 in. Two indications are peaked from the same side of the weld.

The full skip distance on this plate is 2 × 1.00 × tan 70° = 2 × 2.747 = 5.49 in., so the probe must be moved back from the weld at least that far, plus half the cap width, to put the second leg through the whole weld.

To place each indication in the weld, measure the surface distance from the beam index point to the weld centerline. If the probe's index point sat 2.50 in. from the centerline when indication 1 peaked, the reflector is 0.24 in. short of the centerline on the probe side, at 0.82 in. deep. Plot both on a sketch of the joint before naming them. Our UT formulas article has the other time-of-flight checks a Level II uses.

Key takeaway: identify the leg first, then convert; any depth greater than the thickness means a reading or entry error.

What is the reference level, and how do DAC and TCG work?

The reference level is the gain, in dB, that brings a reference reflector to a set screen height; every indication is then judged as so many dB, or so many percent, above or below that reference. Because sound weakens with distance through attenuation and beam spread, codes that use a single reflector correct for distance either with a DAC curve or with a calculated attenuation factor.

DAC and TCG

A distance amplitude correction (DAC) curve is drawn by peaking the same size side-drilled hole at several sound paths, typically holes at 1/4, 1/2 and 3/4 of the block thickness read in the first and second legs, and joining the peaks on screen. Evident's tutorial describes the curve as one that "graphically compensates for material attenuation, nearfield effects, and beam spreading." Time-corrected gain (TCG, also called TVG) does the same job electronically: the instrument raises gain with time so that every reference hole peaks at the same height, typically 80% FSH, and the reference becomes a flat line.

A DAC tells you how an echo compares with a reference hole at the same distance. It does not give the flaw's size. Our DAC curve guide walks through building one.

The 20% and 50% DAC levels

Many ASME-based procedures use two evaluation levels below the reference curve. ASME Section VIII, Division 1, Mandatory Appendix 12, as reproduced in procedures and secondary sources, requires indications above 20% of the reference level to be investigated, and ASME-based procedures commonly record indications at or above 50% DAC. Check the levels in the referencing code and your written procedure, because they vary between codes and editions.

Two readings show how to use the curve. Indication A peaks at 40% FSH where the DAC sits at 60% FSH. That is 40 ÷ 60 = 67% DAC, or 20 × log(0.667) = −3.5 dB, which is above 50% DAC and is recorded and evaluated. Indication B peaks at 12% FSH where the DAC sits at 50% FSH, giving 24% DAC, or −12.4 dB. It is above the 20% investigation level but below the 50% recording level, so it is investigated and characterized, and recorded only if the procedure says so.

Scanning gain

Scanning is done at a higher gain than the reference so that small or badly oriented reflectors are not missed, then each indication is evaluated back at the reference level. ASME-based procedures state scanning "at a gain setting at least two times the primary reference level," that is, at least 6 dB above it. Older AWS D1.1 tables set scanning levels by sound path, with the minimum for short paths in statically loaded nontubular work commonly quoted as 14 dB above the zero reference and higher levels for longer paths. Confirm the values in the D1.1 edition named in your contract. Write the scanning gain on the report; an indication found at +14 dB and evaluated at the reference is a different claim from one found at the reference.

Key takeaway: find at scanning gain, evaluate at reference gain, and express each indication against the reference at the same sound path.

How do you read a weld indication under AWS D1.1?

AWS D1.1 does not use a DAC curve for its standard UT method; it rates each indication with d = a − b − c, where a is the indication level, b the reference level from the 1.5 mm hole, and c an attenuation factor of 2 dB per inch of sound path beyond the first inch. The rating, the weld thickness, the probe angle and the static or cyclic table together give the indication class.

Evident's AWS tutorial defines the terms as: A, the discontinuity indication level in dB; B, the reference indication level in dB; and C, the material attenuation factor, "(2x [soundpath length in inches – 1 in.]) expressed in dB". Because a is the gain needed to bring the indication to the reference line, a stronger reflector needs less gain and gives a lower, more negative d. Class A, the most serious, sits at the low end.

Worked example: rating the two indications

Using the 1.00 in. weld from the sound path example, the reference level from the 1.5 mm hole on the IIW block is 48 dB.

A widely reproduced older edition of the static table puts a 70° indication in a weld over 3/4 in. to 1-1/2 in. thick in Class A at a rating of +2 and lower. On that table, indication 1 at −5.8 is Class A, which is rejectable whatever its length. Indication 2 at +4.4 is not Class A, and whether it is B, C or D depends on the table for the edition, thickness and loading. Confirm the bands in your edition before reporting, and apply the edition's rounding rule for c. There are no fixed dB bands that apply to every weld.

Two cautions apply. The 2 dB per inch attenuation factor is an approximation, and published work has shown it can be off by about 6 dB at long sound paths with 60° and 70° probes. And D1.1 does not by itself require UT; the contract documents, or a referencing specification such as AISC 360, decide which welds get tested and to which criteria. Our AWS D1.1 weld inspection criteria article covers the acceptance side.

Key takeaway: under D1.1, read a, subtract the reference and the attenuation factor, then look up the class for the thickness, angle and loading.

What do common weld defects look like on the A-scan?

An A-scan echo can suggest what a reflector is, but characterization comes from how the echo behaves as you move the probe, where the reflector sits in the joint, and what other angles show. No single echo shape proves a defect type.

Lack of fusion is the defect most often described wrongly. It is not a "moderate-amplitude echo near the surface." It forms where weld metal failed to fuse with the sidewall or the previous run, at whatever depth that face lies, and because the face is smooth and flat it acts like a mirror. Evident describes the signature as a "strong reflection with fast rise and fall time in the second leg" when tested from the side of the fusion gap. TWI notes that "Ultrasonic inspection is normally more effective than radiography in detecting lack of inter-run fusion imperfections." The same mirror effect means a lack of fusion hit off-angle can give a weak echo and be under-rated, which is why procedures scan with more than one angle.

Geometry echoes are the other trap. Root reinforcement, cap edges, mismatch and backing bars all reflect sound, and they appear at predictable positions. Plotting each indication against the joint drawing, as in the worked example above, separates them from real flaws. Our guides to weld defect types and lack of fusion in welds go into each defect, and angle beam testing in UT covers the probe motions (swivel, orbit, along the weld) used to characterize them.

Key takeaway: an indication on the bevel line is lack of fusion until shown otherwise, whatever its amplitude.

How to read a flaw detector screen on rail

On rail, a 70° channel looks for transverse defects in the head, and a 0° channel watches the back-wall (base) echo; a reflector echo in the 70° head gate or a loss of the 0° back-wall echo is a suspect that must be hand-verified. Rail screens read the same physics as weld screens, but the geometry is fixed by the rail section and the defects of interest are different.

Rail test systems typically combine 70° probes for transverse defects, such as transverse fissures and detail fractures, with a 0° probe for horizontal defects and loss of back wall, plus probes at about 37° to 45° or side-looking probes for the web and bolt holes. A straight beam does not find a transverse defect in the head, because the defect face lies parallel to the beam.

What a 70° transverse defect looks like

A transverse defect lies across the head, roughly perpendicular to the rail length. A 70° probe aimed along the rail meets its face at a steep angle, and the echo appears inside the head gate at a sound path set by the defect's position. As the probe approaches, the echo walks along the baseline toward the initial pulse and rises to a peak, then falls away as the beam passes the defect. A system with forward- and rear-facing 70° probes sees the same defect twice, once from each side, which is a strong sign that the reflector is real.

For example, a 70° echo peaking at 2.0 in. of sound path is at a depth of 2.0 × cos 70° = 0.68 in. below the running surface and 2.0 × sin 70° = 1.88 in. along the rail from the probe's index point. That puts it inside the head, where transverse fissures start. A similar echo that appears only from one direction, sits right at the running surface, or repeats at every tie is more likely shelling, head checks, a weld geometry echo or a coupling artifact.

What loss of bottom looks like

On a 0° channel the back-wall echo from the base of the rail is the reference. On 136 RE rail, which is 7-5/16 in. (7.31 in.) high, a 0° longitudinal wave at about 5,920 m/s (0.2331 in./µs) returns from the base after 2 × 7.31 ÷ 0.2331 = 62.7 µs. A drop or loss of that echo, often called loss of bottom or loss of back wall, means something is stopping the sound on its way down or back.

The last row matters for compliance. Rail that could not be tested because coupling failed is unsearched rail under 49 CFR 213.237, and the location of track not tested must be recorded under 49 CFR 213.241(c)(6) on eCFR. A "valid search" under 49 CFR 213.237 requires equipment that performs as intended and responses interpreted by a qualified operator. FRA research on rolling contact fatigue notes that surface damage can prevent "the successful transmission of ultrasonic waves to the base of the rail and back," so a loss of bottom on damaged rail calls for a hand test, not a guess. Calling those sections false calls hides untested rail; our article on false calls in rail UT separates the two. For the defects themselves, see transverse fissure in rail and horizontal split head.

Key takeaway: on rail, a 70° echo seen from both directions points to a transverse defect, and a lost back wall with no reflector echo is a coupling or surface problem until a hand test says otherwise.

What do B-scan displays show, and where do they mislead?

A B-scan stacks many A-scans along a scan line and shows them as a cross-section, with position along the scan on one axis, sound path or depth on the other, and amplitude as color or brightness. It needs an encoder, a phased array or a test car's position system; a plain manual flaw detector without one only shows the A-scan.

In straight-beam work on plate, the back wall appears as a continuous line, and a reflector between the entry surface and that line is a candidate indication. A laminar flaw shows as a line parallel to the surface with the back wall shadowed beneath it.

That rule does not transfer to angle-beam weld examination. An angle beam does not strike the far surface squarely, so there is normally no back-wall echo to use as a boundary. Echoes in an angle-beam B-scan or S-scan come from geometry such as the root, cap and mismatch as well as from flaws, and each one must be placed on the joint drawing using the leg geometry above. In rail, test-car software shows a B-scan-style side profile channel by channel, and the operator still checks the A-scan behind any suspect.

Key takeaway: a B-scan is a map of A-scans; the "above the back wall" rule works only for straight-beam scans with a visible back wall.

Limits: what the screen can't tell you

A flaw detector screen reports how much sound came back and when. Several things it cannot do should be stated in every report:

• It doesn't show flaw size or shape. Amplitude depends on orientation, roughness and distance as well as size. A large, off-angle lack of fusion can give a smaller echo than a small, well-oriented pore. Length comes from scanning along the indication and a drop method; height needs techniques such as tip diffraction or TOFD.
• Closed or compressed cracks reflect less. A tight crack under compression can transmit much of the sound and give a weak echo.
• Near-surface zones are hard. Ringing from the initial pulse or wedge, and the near field, hide echoes close to the entry surface. Duals, delay lines and other angles help.
• Coarse grain and attenuation. Austenitic and cast materials scatter sound, raise the noise and bend the beam. The 2 dB per inch D1.1 factor and steel velocities don't apply to them.
• Readouts trust the inputs. Wrong thickness, angle or velocity gives confident, wrong depth and position readings.
• Coverage gaps. A reflector parallel to the beam, or outside the scanned zone, gives nothing at all. A clean screen means "nothing reflected from the volume covered," not "no defect."
• The operator matters. FRA report DOT/FRA/ORD-26/07 on handheld rail UT found that, with a small exception, no two operators produced statistically similar calibration results, and it warns that "calibration errors can cause excessive false positives and false negatives." Written procedures and reference checks exist for that reason.

How EMA Industries applies this

EMA Industries performs ultrasonic examination of AWS D1.1 structural welds and ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, from Southern California and nationwide. Reports state the calibration block, reference level, scanning gain, sound path, depth, surface distance and rating for each indication, so the reading can be checked. EMA's NDT Level III services can also write or review a UT procedure that sets surface preparation, DAC levels and scanning gain for a project. To discuss weld or rail testing, contact EMA Industries.

Frequently asked questions

How do you read the depth of a flaw on an angle-beam flaw detector?

Read the sound path of the peaked echo, decide whether it is in the first or second leg, and convert it. In the first leg, depth is sound path × cos θ; in the second leg, depth is 2t − sound path × cos θ. Most digital instruments do this automatically once the angle and thickness are entered, so check those entries first.

What is the difference between reference level and scanning gain?

The reference level is the gain that brings the reference reflector to its set height, and all indications are evaluated against it. Scanning gain is a higher setting, at least 6 dB above reference in ASME-based procedures and higher still in the older D1.1 tables, used only while searching so that weak reflectors are noticed. Each indication found at scanning gain is then evaluated at the reference.

What do 20% and 50% DAC mean?

They are evaluation levels below the reference curve: 50% DAC is 6 dB below it and 20% DAC is 14 dB below it. In many ASME-based procedures, an echo above 20% DAC is investigated and one at or above 50% DAC is recorded. The exact levels come from the referencing code and the written procedure.

Does AWS D1.1 use a DAC curve?

AWS D1.1's standard UT method does not use a DAC curve. It uses a single 1.5 mm side-drilled hole reference and the rating d = a − b − c, where c corrects for attenuation at 2 dB per inch of sound path beyond the first inch. Its PAUT annex and some contract specifications use other methods, so check the edition and contract.

Is ASTM E317 still a current standard?

ASTM E317 is current: the ASTM store lists E317-21 as the active edition. It covers evaluating the performance of pulse-echo instruments and systems, such as linearity, resolution and gain accuracy, without electronic test equipment. It is not a calibration certificate standard, and electronic measurements belong to the companion guide, ASTM E1324.

What does loss of back-wall echo mean on a rail flaw detector?

It means sound on the 0° channel is not returning from the base of the rail. Causes include a horizontal or tilted reflector in the head or web, surface damage such as shelling, or lost coupling. It needs a hand test, and if coupling was the cause, that length is untested rail that must be recorded and re-tested.

Sources

• ASTM E317-21, Standard Practice for Evaluating Performance Characteristics of Ultrasonic Pulse-Echo Testing Instruments and Systems without the Use of Electronic Measurement Instruments (store page, active edition)
• ASTM E1324-21R26, Standard Guide for Measuring Some Electronic Characteristics of Ultrasonic Testing Instruments (store page, active edition)
• Evident: AWS D1.1 indication rating tutorial
• Evident: DAC and TVG tutorial
• Evident: basic waveform interpretation
• TWI Job Knowledge 041: lack of sidewall and inter-run fusion
• Sonatest: IIW Type 1 calibration block
• 49 CFR 213.237, Inspection of rail (eCFR)
• 49 CFR 213.241, Inspection records (eCFR)
• Magel, Rolling Contact Fatigue: A Comprehensive Review, DOT/FRA/ORD-11/24 (2011), ROSA P
• Holloway, structural UT and D1.1 attenuation, CINDE Journal (2017)
• Bruzek, Einbinder and Tunna (ENSCO), handheld rail UT probability of detection, DOT/FRA/ORD-26/07 (2026), ROSA P
• ISO 2400:2025, ultrasonic testing standard block specification (AFNOR catalog listing)