Digital vs Analog Flaw Detectors: Qualifying a Digital Replacement for Rail

The digital vs analog flaw detector question is mostly settled: portable ultrasonic instruments have used digital signal processing for years. The real job is qualifying the replacement. It must reproduce the existing procedure's response on the reference rail, and under 49 CFR 213.238 each operator must show proficiency "for each type of equipment" before authorization.

This guide is for the rail testing supervisor, transit track engineer or NDT Level III who has to retire an old instrument or rail test system and put a new one into service without breaking the procedure behind every valid search. It covers instrument evaluation, calibration transfer, procedure revision, operator qualification, records and side-by-side validation, and ends with a worked validation plan.

Is the digital vs analog flaw detector choice still a real decision?

For new equipment, no. Evident's flaw detection tutorial states that "most modern ultrasonic flaw detectors incorporate digital rather than analog signal processing", and that microprocessor-based processing is now the industry standard for portable flaw detectors (Evident, Digital Signal Processing). A buyer today chooses between digital instruments, not between digital and analog.

What remains is a migration problem. A rail program that still runs an older instrument, whether a true analog CRT unit or an early digital model the maker no longer supports, has a written procedure, a reference rail and a set of qualified operators built around it. Swapping the box changes none of the rules, but it can quietly change the sensitivity, the screen behavior and the records.

Evident's tutorial also notes that analog processing can sometimes offer faster screen refresh or data acquisition. That point matters for one validation check covered below: a digital unit samples and redraws the signal, so the screen update rate and any peak-hold behavior have to be good enough for the scanning speed the procedure allows.

Key takeaway: treat the change as qualifying a new instrument against an existing procedure, not as a technology debate.

What actually changes with a digital flaw detector, and what doesn't?

A digital flaw detector changes how the echo is displayed, stored and measured; it does not change the ultrasound in the rail. The probe, wedge, couplant, beam angle and the steel decide what echo comes back. The instrument decides what the operator sees and what gets saved.

Stored A-scans are the largest real gain. An operator can save the screen at a suspect location with its setup, so a Level III can review it later and an owner can see what the call was based on. Our A-scan, B-scan and C-scan guide explains how to read each view.

B-scan views and encoders matter mainly for carts and vehicles. An encoder ties each A-scan to a distance along the rail, so the software can draw a strip view and give each indication a position. On a hand instrument with a single probe, there is usually nothing to encode.

What doesn't change is just as important. The 70° beam still finds transverse defects in the head, the 0° beam still finds horizontal separations and loss of back echo, and a shell on the running surface still blocks sound from what lies beneath it. Our ultrasonic flaw detector guide covers instrument selection and the probe set in detail.

Key takeaway: budget the qualification effort around the features that really change, mainly sensitivity setting, stored data and positioning.

What does a replacement detector have to prove before rail service?

A replacement has to clear six gates: the instrument performs to specification, the calibration reproduces the procedure's sensitivity on the reference rail, the procedure is revised and approved, the operators are qualified on it, the records meet 213.241, and side-by-side runs show it finds what the old unit found.

Part 213 doesn't name any of the technical gates. It requires internal rail inspections and a valid search, and it makes the qualified operator responsible for determining that the equipment "is performing as intended" (213.238(b)(2)). The six gates are how a provider shows that it can make that determination with a new instrument.

Key takeaway: none of these gates is optional just because the new unit is newer or more capable.

How do you evaluate the new instrument under ASTM E317 and E1324?

ASTM E317-21 evaluates the pulse-echo instrument and system with blocks and the instrument itself; ASTM E1324-21(2026) measures its electronic characteristics with test equipment. Neither sets pass or fail limits, so the procedure has to.

E317 covers horizontal limit and linearity, vertical limit and linearity, entry-surface and far-surface resolution, sensitivity and noise, and the accuracy of calibrated gain controls (ASTM E317-21). It applies to systems with an A-scan presentation, works in shop or field conditions, and needs no extra electronic test gear. Its scope says the evaluation is intended "for comparing instruments and systems", which is exactly the migration case.

E1324 measures power supply, pulser, receiver, time base and gate or alarm sections with test instruments such as oscilloscopes (ASTM E1324-21(2026)). It describes itself as complementing E317, and it isn't meant for evaluating instruments on the production line. Most rail programs use E317 checks in the field and rely on the maker's E1324-type data for acceptance of the electronics.

Neither standard sets a schedule either. E317 5.7 states: "No frequency of system evaluation or calibration is recommended or implied." The interval is the using parties' decision, so the procedure has to name it.

E317 itself makes the point that not every check is needed for every use. In 5.6 it notes that vertical linearity may not matter for a go/no-go test with a gate alarm, while horizontal linearity matters for depth or thickness measurement. A rail procedure that both alarms on a gate and reports depth needs both.

In Europe, the equivalent instrument standard is EN ISO 22232-1:2020, which replaced EN 12668-1 and covers instruments for manual testing with single or dual probes. A US owner can cite it in a purchase specification alongside E317, but neither standard replaces the owner's acceptance limits.

Key takeaway: write the E317 limits into the procedure before the new unit arrives, so that the evaluation is a pass or a fail and not a judgment made afterward.

How do you re-verify calibration on the reference rail?

Calibrate the new instrument from scratch on the same reference rail and reflectors the procedure uses, then compare its response to the old unit's. The gain number on the screen does not transfer between instruments; the response on the reference reflector does.

Rail test sensitivity is normally set on a section of reference rail with machined reflectors, not on a generic weld block. AREMA's Manual for Railway Engineering Chapter 4 includes a section titled Recommended Calibration Rails for Rail Flaw Detection System (4.6, 2025 Chapter 4 TOC); the content is paywalled, and the reflectors in any given program come from the owner's or provider's procedure. Our rail UT calibration guide covers what an owner should ask for in an RFP.

Re-verification has four parts:

• Time base: set velocity and zero for each probe on a known block, then check sound path to known features on the reference rail.
• Angle: check the refracted angle of each angle probe, since a different wedge or probe on the new unit can shift it.
• Reference level: bring the reference reflector to the procedure's screen height and record the gain. Scanning gain is then the procedure's offset above it.
• DAC or TCG: rebuild the curve on the new unit from the same reflectors. Don't copy the old curve's points across.

Distance-amplitude correction works the same way on both instruments: a curve drawn through equal reflectors at different depths compensates for attenuation, beam spread and near-field effects, and TCG flattens the same correction electronically (Evident, DAC and TVG). The new unit's software builds the curve faster. It doesn't make the curve right if the reflectors were peaked badly. See our DAC curve guide for the construction steps.

The bigger risk is the person, not the box. FRA's 2026 study of handheld rail UT (DOT/FRA/ORD-26/07, ENSCO) found that, with a small exception, no two operators produced statistically similar calibration results, and it reported no standardized calibration procedure for handheld rail UT across North American railroads (ROSA P, ORD-26/07). A new digital unit inherits that variation unless the procedure fixes the steps.

Key takeaway: compare responses on the reference rail, not gain settings, and write each calibration step down tightly enough that two operators get the same answer.

How should the written procedure change, and who approves it?

Revise the procedure to name the new instrument, its stored setups and its checks, and have it approved by the person the employer's written practice assigns, usually the NDT Level III. The rule itself only requires that operators act "in accordance with the employer's procedures and instructions" (213.238(b)(3)).

A procedure revision for a new detector usually changes these items:

• the instrument make, model and firmware version, and how firmware updates are controlled
• the stored setup files for each probe and channel, with names, locks and who may edit them
• the calibration sequence, reference reflectors, screen heights and scanning gain offsets
• DAC or TCG construction and the evaluation and reporting levels
• the E317 checks, their limits and their interval
• the start-of-shift, periodic and end-of-shift checks, and what happens to rail tested since the last good check
• the files to save at each suspect and defect, and how they are named
• loss-of-coupling and screen-update requirements for the allowed scanning speed

Who approves the revision is set by the employer's written practice, not by FRA. Under SNT-TC-1A, the employer certifies its own NDT personnel, and a Level III is normally the person responsible for establishing and approving techniques and procedures. FRA doesn't certify UT technicians, and 213.238 uses no Level I, II or III at all.

The owner should see and accept the revision when a contractor provides the testing. Under 213.238(a), a track owner "shall not utilize a provider of rail flaw detection" that fails to keep a documented training program and to identify the equipment types each operator is qualified on. An owner who receives a new instrument without a revised procedure has a gap to close before the next test.

Key takeaway: the procedure, not the instrument manual, is what makes a new detector compliant; revise it before the first live run.

What does 49 CFR 213.238 require when operators move to new equipment?

Each operator must be trained on the new equipment and must demonstrate proficiency on it before the employer authorizes them to use it. Section 213.238(c) is explicit (49 CFR 213.238 on eCFR):

"The operator must demonstrate proficiency in the rail defect detection process, including the equipment to be utilized, prior to initial qualification and authorization by the employer for each type of equipment." — 49 CFR 213.238(c)

The rule hangs on the phrase "type of equipment". The provider must "identify the types of rail flaw detection equipment" each operator is qualified on (213.238(a)), and qualification records must name "the equipment to which the qualification applies" (213.238(e)). Part 213 doesn't define when a new model becomes a new type.

The safe reading is to treat any change of instrument family, display or setup logic as a new type, and to document the decision either way. An operator moving from one portable flaw detector to another with a different menu, gate logic and DAC tool is relearning the part of the job where mistakes happen.

The 160 hours of supervised experience in 213.238(c) is part of initial qualification, so an operator already qualified on rail flaw detection doesn't restart from zero. What the employer still needs is training on the new unit, a demonstration of proficiency on it, and, for reevaluation, "a recorded examination" on the specific equipment types (213.238(d)). The supervision may come from a qualified operator or the equipment manufacturer's representative.

A practical familiarization session covers the setup files, calibration on the reference rail, gate and alarm behavior, saving and naming files, recognizing lost coupling on the new screen, and making the same calls on the defect library as on the old unit. Keep the attendance list, the exam and the proficiency results; under 213.238(g) these records must be available to FRA.

Key takeaway: update each operator's 213.238 record with the new equipment and the date before that operator runs a valid search with it.

How do digital records help with 49 CFR 213.241?

Digital storage makes the 213.241 record easier to complete and harder to lose, but only if the system meets the rule's electronic-record conditions. A stored A-scan supports a record; it doesn't replace the required items.

For internal rail inspections, 213.241(c) requires the date, the track inspected with beginning and end points, the location and type of each defect found under 213.113, its size if not removed before the next train, the initial remedial action and its date, and the location of any track not tested under 213.237(g) (49 CFR 213.241 on eCFR). The owner keeps that record for at least two years after the inspection and one year after initial remedial action.

Electronic records are allowed under 213.241(j) if the system meets six conditions:

Rule (j)(4) is the one most portable instruments don't meet on their own. A flaw detector that lets any user overwrite a stored file is a data source, not a records system, so the export and the database that receive the files need the controls. Shared logins on the instrument fail (j)(3).

Item (c)(6) is where a digital detector helps the most. A cart or vehicle system with an encoder can log where coupling was lost, so untested rail is recorded by position rather than by memory. That untested rail is not a false call; it is rail without a valid search, and 213.237(h) sets what must happen before the deadline.

The record that matters to an owner also includes the remedial action, which a contractor can't complete alone. A report template should leave the (c)(5) field for the owner, as discussed in our 213.237 guide.

Key takeaway: check the export, logins and file locks against 213.241(j) during validation, not after an FRA records request.

How do you run a side-by-side validation?

Run the old and new instruments on the same reflectors, samples and track, with the same probes where possible, and compare detection and amplitude against differences agreed in writing before the runs. The aim is to show that the new unit finds and reports what the old one did, and to explain every case where it doesn't.

A useful validation has three parts:

1. Reference rail: every reflector the procedure uses, on every probe and channel, repeated by at least two operators.
2. Defect library: removed rail and weld samples with known indications, covering the defect types the procedure targets, such as transverse defects in the head, bolt-hole cracks, horizontal splits and weld indications.
3. Field locations: suspect locations from the owner's last test, re-verified with both instruments, plus a stretch of clean rail to compare noise and false indications.

Decide the acceptance rules first. A typical set is: every indication the old unit reported, the new unit must report; amplitude differences on reference reflectors stay within a stated band, such as ±2 dB; any indication only the new unit reports is investigated, not counted as a failure; and repeat runs fall within a stated spread.

Differences have three common causes, and each needs a different response. A probe, wedge or cable difference is fixed by matching the hardware. A setup difference, such as a gate start, a filter or a reject setting, is fixed in the stored setup and the procedure. A real performance difference, such as a noisier receiver, may mean the unit fails the evaluation.

Don't use the old instrument as the truth. If the new unit finds a real indication the old one missed, that is a gain, and the procedure should say how such finds are verified. If the old unit's response depended on an operator habit, the validation is the time to find out.

Key takeaway: an unexplained difference is a stop, even when the new unit looks better.

Worked example: a validation plan for a new digital flaw detector

This example qualifies a new portable digital flaw detector that replaces an older unit used for hand testing and for field-verifying test-car suspects. All numbers are illustrative; the limits are the example procedure's own choices, since E317 sets none.

Inputs:

• Probes: the procedure's 0°, 45° and 70° probes, used on both instruments
• Reference level: the 70° reference reflector at 80% full screen height (FSH)
• Scanning gain: reference + 6 dB; reporting level: −12 dB relative to reference
• Acceptance: amplitude difference within ±2 dB; no reportable indication missed; repeat-run range within 2 dB

Step 1: E317 checks on the new unit. Gain accuracy was checked from 80% FSH. A 6 dB cut should give 80 × 10^(−6/20) = 80 × 0.501 = 40.1% FSH; the unit showed 39%, an error of 20 log(39/40.1) = −0.24 dB. A 12 dB cut should give 80 × 0.251 = 20.1%; it showed 21%, an error of +0.38 dB. Both are within the procedure's ±1 dB limit.

Horizontal linearity was checked with five back echoes set at 2, 4, 6, 8 and 10 screen divisions. The fifth echo landed at 9.95 divisions, an error of 0.05 division, or 0.5% of a 10-division screen, within the procedure's 1% limit.

Step 2: reference rail and calibration transfer. The old unit brought the 70° reference reflector to 80% FSH at 42.0 dB; the new unit needed 45.5 dB. That 3.5 dB is a difference in gain scales, not an error, so the new unit's scanning gain is 45.5 + 6 = 51.5 dB.

The DAC was rebuilt on the new unit from the same three reflectors.

Both differences are inside ±2 dB, so the curve shapes agree.

Step 3: side-by-side runs. Two operators ran each item three times on each instrument. The table gives the mean amplitude relative to the reference level.

Item E's mean came from three new-unit runs of −4.0, −4.6 and −3.7 dB: (−4.0 − 4.6 − 3.7) ÷ 3 = −4.1 dB, with a range of 0.9 dB, inside the 2 dB repeat limit.

Item G failed on the first run. At −13.1 dB the indication fell below the −12 dB reporting level, so the new unit would have missed a call the old one made, and the difference of −2.6 dB was outside ±2 dB. The investigation found that the new kit's 70° probe had a flat wedge that rocked on the ground weld collar; with the procedure's contoured wedge it read −11.2 dB, reportable and 0.7 dB from the old unit.

With the re-run, the differences for items B to G are +0.6, −0.6, +0.8, +1.4, −0.6 and −0.7 dB. They sum to +0.9 dB, a mean of +0.15 dB, and the largest is 1.4 dB, so every item passes.

Step 4: field verification. Both units re-verified 10 suspect locations from the owner's last test-car run. Both confirmed the same 3 defects and called the same 7 locations non-relevant, an agreement of 10 out of 10. On 500 ft of clean rail, neither unit showed an indication above the reporting level.

Step 5: records and sign-off. The validation report lists each item, run, operator and file name, with the stored A-scans attached. Sign-off has four parts:

• The Level III approves the validation report and the revised procedure, here Revision D, which adds the contoured-wedge requirement found in item G.
• The employer updates each operator's 213.238 record with the new equipment type, the training date and the recorded proficiency exam.
• The records custodian confirms that the export meets the 213.241(j) conditions: individual logins, locked files and separate amendments.
• The owner accepts the revision before the new unit runs its first valid search.

Key takeaway: the example passed only because the failure on item G was investigated and fixed in the procedure; a plan that only averaged the differences would have hidden it.

What changes for multi-channel carts and test vehicles?

A multi-channel cart or test vehicle adds encoder accuracy, channel mapping and data review to the same six gates. The instrument evaluation becomes a system evaluation, and the B-scan or channel map becomes part of the record.

Check the encoder against a taped length before validation. In the same illustrative plan, a cart run over 100.0 ft of taped rail that reads 99.6 ft has an error of −0.4%, within a 1% procedure limit. An indication 37.5 in. from a rail end should then appear within the location tolerance the field crew needs to find it again.

Each channel's angle, direction and gate must match the procedure's channel list, and each must be shown to respond on the reference rail. In a rolling B-scan from a phased-array prototype, TTCI reported that bolt holes appear as diagonal lines and a bolt-hole crack as an extra line on the second hole (TTCI TD-16-050, 2016). The analyst has to know what normal rail geometry looks like on the new display before reading defects on it.

If the system is used for continuous testing, 49 CFR 213.240 adds its own duties: a person qualified to interpret the data, field verification of each suspect by a 213.238-qualified person, and time windows for verification (49 CFR 213.240 on eCFR). Phased array in rail remains mostly a verification and research tool; see our phased array guide for how it works.

Key takeaway: on a system, prove the position as well as the sensitivity, because a defect you can't find again can't be remediated.

Which common claims about digital detectors are wrong?

Several claims repeated in equipment articles, including this page's earlier version, don't hold up. Each matters because a buyer may quote it in a specification or a procedure.

"There are two types of ultrasonic testing: conventional and phased array." That's a false choice. UT techniques include pulse-echo contact testing, through-transmission, time-of-flight diffraction, phased array and guided wave, among others, and phased array is itself a pulse-echo variant. Rail in track is tested from the head, so in-track flaw detection is pulse-echo, with 0° and angle beams.

"There are four types of NDT: UT, MT, RT and eddy current." That list leaves out penetrant testing and visual testing, the two most widely used surface methods. Rail programs rely on visual track inspection under 213.233 and internal testing under 213.237, with MT and PT used on welds and components; see our pages on magnetic particle examination and visual testing.

"EN 13231 requires inspection documentation for flaw detection." EN 13231-1 covers the acceptance of track works on ballasted track, such as geometry, position and ballast profile (SS-EN 13231-1:2024 listing). The European standard for rail NDT in track is the EN 16729 series; EN 16729-1:2016 "applies to testing of rails installed in track for detecting internal discontinuities" (EN 16729-1 listing). For US track, the rules are 213.237, 213.238 and 213.241.

"Digital reduces human interpretation errors" and "automated calibration." A digital unit can store setups, draw DAC curves and alarm on gates, which removes some arithmetic and transcription errors. It doesn't remove interpretation; the FRA operator-calibration findings above show variation persisting on modern handheld equipment, and Part 213 assigns interpretation to the qualified operator under 213.238(b)(3).

Unsupported performance and data claims. The earlier version promised comparison data that it never gave. No public study comparing detection by analog and digital rail flaw detectors was found for this edit, so this page makes no detection-rate claim either way. A program's own side-by-side validation is the evidence that counts.

Key takeaway: if a claim about a detector can't be traced to a rule, a standard or a named study, don't put it in a specification.

What a digital flaw detector can't do

A digital flaw detector can't make a search valid on its own, and it can't see through conditions that block sound. Its limits are the method's limits plus a few of its own.

• Masking: shells, head checks and heavy surface damage still block or scatter the beam, so defects under them can be missed whatever the display.
• Coupling: a dry, lubricated or debris-covered head still gives a weak or lost signal. A loss-of-coupling alarm helps only if the procedure says what to do when it sounds.
• Interpretation: stored A-scans and alarms don't decide whether an indication is a defect, a weld geometry echo or a rail end. That remains the qualified operator's call.
• Sizing: an A-scan amplitude isn't a flaw size, and a DAC gives no size; sizing for 213.113 still comes from the procedure's method and, where needed, verification.
• Sampling and refresh: a digital screen redraws at a finite rate. If the scanning speed outruns it, a short echo can be missed, which is why the procedure should limit hand-scanning speed for the unit.
• Records: a stored file is not a 213.241 record until the system that holds it meets 213.241(j).
• Software and firmware: an uncontrolled firmware update can change filters or gate behavior; the procedure should require a reference-rail check after any update.

Key takeaway: the new detector inherits every limit of pulse-echo rail UT; plan the validation to show those limits haven't grown.

How EMA Industries applies this

EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, and NDT Level III services, from Southern California with nationwide mobilization. When an owner or provider replaces a flaw detector or test system, a Level III can write the validation plan, run the E317 and reference-rail checks, revise the procedure and document operator qualification under 213.238. See our ultrasonic examination services or contact EMA Industries to plan a qualification.

Frequently asked questions

Is ultrasonic testing digital or analog?

Ultrasonic testing is a method, and the instrument can process the signal either way. The echo from the transducer is an analog electrical signal; a digital flaw detector samples it and processes it in software, while an analog unit displays it directly. Most portable flaw detectors made today are digital.

Do I have to requalify operators when we buy a new flaw detector?

Each operator has to be trained on the new equipment and demonstrate proficiency on it before the employer authorizes its use, because 213.238(c) applies "for each type of equipment". An operator already qualified on rail flaw detection doesn't repeat the 160 supervised hours. The employer should update each operator's qualification record with the new equipment and the date.

Can I copy the old instrument's gain settings to the new one?

No, because gain scales differ between instruments, so the same dB number gives a different sensitivity. Calibrate the new unit on the procedure's reference rail and reflectors, record its own reference gain, and set scanning gain from that. Then compare the response on the reference reflectors against the old unit's.

What standard covers evaluating a new ultrasonic flaw detector?

In the US, ASTM E317-21 evaluates pulse-echo instruments and systems in the shop or field, and ASTM E1324-21(2026) measures their electronic characteristics with test equipment. EN ISO 22232-1:2020 is the European equivalent for instruments. None of these sets acceptance limits or check intervals, so the owner's or provider's procedure must.

Do digital flaw detectors satisfy FRA record requirements?

Not by themselves. A stored A-scan supports the record, but 49 CFR 213.241(c) still requires the date, track limits, defect location, type and size, remedial action and untested track. Electronic records must also meet 213.241(j), including individual logins, files locked once complete and storage started within 72 hours.

Sources

• 49 CFR 213.238, Qualified operators: eCFR (verbatim text read on Cornell LII, October 2026)
• 49 CFR 213.241, Inspection records: eCFR (read on Cornell LII, October 2026)
• 49 CFR 213.237, Inspection of rail: eCFR
• 49 CFR 213.240, Continuous rail testing: eCFR
• ASTM E317-21, Evaluating Performance Characteristics of Ultrasonic Pulse-Echo Testing Instruments and Systems without the Use of Electronic Measurement Instruments: ASTM store page
• ASTM E1324-21(2026), Measuring Some Electronic Characteristics of Ultrasonic Testing Instruments: ASTM store page
• Evident, Digital Signal Processing tutorial (vendor): ims.evidentscientific.com
• Evident, DAC and TVG tutorial (vendor): ims.evidentscientific.com
• FRA DOT/FRA/ORD-26/07, MAPOD for handheld rail UT (ENSCO, July 2026): ROSA P dot/92816
• TTCI Technology Digest TD-16-050, Characterizing Rail Flaws Using Phased Array Ultrasound (Witte and Poudel, November 2016): MxV Rail PDF
• AREMA Manual for Railway Engineering, Chapter 4 table of contents (2025): publications.arema.org
• EN 13231-1 (SS-EN 13231-1:2024), track acceptance of works: SIS listing
• EN 16729-1:2016, NDT on rails in track, ultrasonic inspection: genorma listing