UV Curing: Measurement & validation, every question, answered

What is a UV radiometer?

A calibrated meter that measures UV irradiance (W/cm² or mW/cm²) at its sensor, many models also integrate over time to report dose in J/cm². It's the instrument that turns a lamp that looks bright into a controllable process number.

Do I really need a radiometer?

If you cure in production, yes. Output drifts with lamp age, guide wear, and fixturing, and a radiometer is the only way to see it before parts fail. Every dose-based recipe, alarm limit, and revalidation depends on one.

How do I choose a radiometer for my source?

Match its spectral band to what you're measuring: a UVA-band instrument for our systems' 365–405 nm outputs, with a range high enough for spot intensities, tens of W/cm² near a tip. For LED work, prefer a meter specified for LED sources at your wavelength.

Where should I take radiometer readings?

At the working plane, the same distance and angle the joint sees, using a jig so the sensor lands in the same place every time. A reading at the guide tip is fine for trending output, but the number your process runs on is at the part.

Why measure at the working plane, not the tip?

Because the part doesn't live at the tip. Divergence means intensity at 10 mm can be a fraction of the tip value, so a recipe computed from a tip reading over-promises dose. Trend at the tip if you like, but validate and alarm at the plane.

How often should I check UV output?

Set frequency by risk: many shops verify at the start of each shift, while weekly checks can suit stable, low-criticality work. Our PCB tacking note pairs weekly dose spot-checks with a first-article pull test each shift; regulated assembly usually justifies daily readings.

How do I baseline a new system or guide?

When hardware is new, lamp, guide, or lens, measure irradiance at your standard points (tip and working plane), record it with the date, hours, and radiometer ID, and treat that as 100%. Every future reading is judged against this baseline.

The replace-at-20%-drop guideline only works if you captured the baseline while the hardware was fresh.

At what output drop should I take action?

Our guidance: around a 20% drop from baseline with a known-good lamp means it's time to service the guide or swap the lamp. Set a warning limit before that point so the fix is planned maintenance, not a line-down reaction.

Why does band matching matter for radiometers?

A radiometer only responds within its spectral band, weighted by its response curve. Point a mismatched meter at a source and the number can be wildly off, a UVA meter reads an arc lamp's UVA only, and a broadband instrument may under- or over-weight a narrow LED line.

Can a broadband meter measure my LED source?

It will display a number, but accuracy depends on how its response curve treats your exact wavelength, errors can be large. Use an instrument specified for LED measurement at your wavelength, and whatever you choose, use the same meter consistently so trends stay valid.

Why do two radiometers give different readings?

Different spectral response curves, calibration references, sensor geometries, and drift, even meters of the same model disagree slightly. That's normal. Pick one instrument as your process standard, calibrate it on schedule, and never mix meters within a trend chart.

How often should a radiometer be calibrated?

Follow the manufacturer's interval, annual calibration is the common default, and recalibrate after any drop or suspected damage. If your quality system requires traceability, keep the calibration certificate filed with your process records.

What's the same-meter rule?

Always trend a given system with the same radiometer. Absolute accuracy between instruments varies, but a single meter's repeatability is much better, drift you see with one meter is real drift. Switch meters mid-trend and you can't tell drift from meter offset.

How do I measure dose rather than intensity?

Two ways: an integrating radiometer accumulates J/cm² during the exposure, or you multiply a steady irradiance reading by your timer setting. The integrating method also captures shutter and timer behavior, so it's the better audit of what a part actually receives.

How do I set alarm limits on UV output?

From your window: the hard floor is the output where minimum dose is no longer met at your validated time. Place the action limit above that floor with room to react, and a warning limit above the action limit, spaced by your measured drift rate between checks.

Should I run a control chart on output readings?

Yes, plot every scheduled reading against baseline with your warning and action limits drawn on. Trends announce a dying lamp weeks in advance, and a sudden step change flags a bumped fixture or damaged guide the day it happens, instead of when parts fail.

How do I verify splitter legs are balanced?

Measure each tip under identical conditions, same distance jig, same meter, and compare. Our splitters are factory-balanced to ±5%; if a leg reads well outside that against its siblings, inspect that leg's end faces and routing.

Recheck balance after any lamp or guide change, since coupling shifts can favor one leg.

When should I revalidate my cure process?

After any change that touches delivered dose: lamp or LED head replacement, light-guide swap or repair, lens change, fixture rebuild or relocation, a new adhesive formulation, or a substrate change. Re-measure at the working plane, re-run first-article strength checks in-house, and update the records.

What should I re-check after replacing a lamp?

Measure irradiance at every tip and at the working plane, compare to baseline (expect a rise over the tired lamp), confirm the timer still lands parts inside the validated window, and run first-article checks. Log the new lamp's install date and starting output as the fresh baseline.

Do I need to revalidate after a new light guide?

Yes, guides differ slightly in transmission, and a new one usually delivers more than the worn one it replaced. Re-baseline at the tip and working plane, confirm dose at your validated time, and note the guide part number in the record.

What should each output log entry include?

Date and time, system and channel, lamp or LED hours if available, radiometer model and serial, measurement position (tip or plane, with jig ID), the reading, and the operator. Consistent context is what makes the numbers comparable a year later.

How do I verify cure quality in production?

Layer your checks: every part gets a quick look, appearance, no wet edges, tack-free touch where accessible, periodic first articles get destructive strength checks, and scheduled radiometer readings guard the input side.

All of this runs in your own facility against your documented window; it's your data, tied to your parts.

What is a first-article pull test?

At the start of a shift or lot, you cure one part exactly per recipe and pull or shear the joint to destruction in-house, confirming strength before committing the run. Our PCB tacking note pairs it with weekly radiometer spot-checks, cheap insurance on the two things that matter.

How often should we pull-test joints?

A common rhythm is one first article per shift plus one after any process event, lamp change, new adhesive lot, fixture work. High-criticality or regulated products justify more. Set the interval from your failure cost, and write it into the SOP.

How can I check surface cure quickly?

A gentle wipe with a clean swab or a gloved-finger touch on a sacrificial spot: fully cured surfaces feel dry and glassy, inhibited ones drag or smear. It's a screen, not a proof, pair it with periodic strength checks.

How do fluorescing adhesives help inspection?

They glow under a low-power UV inspection lamp, so coverage, fillet size, and placement become a two-second visual check on every part. Fluorescence confirms where adhesive is, not whether it's cured, you still verify cure by feel, hardness, or strength checks.

What does a properly cured joint look like?

Glassy, dry surface; a clear or uniformly colored fillet without wrinkles, cracks, or wet edges; and no movement under a firm probe. Anything hazy, tacky, or liquid at the edges deserves a radiometer reading before the next part runs.

Can I use hardness to verify cure?

Yes, durometer or probe hardness on the fillet, or on a witness dot cured beside the part, correlates well with conversion for a given adhesive. Establish the expected hardness during qualification, then use it as a fast in-house check on production parts.

What's a witness coupon and why use one?

A small dot of the same adhesive cured alongside the part, in the same exposure. You can prod, scratch, or hardness-check it without risking product, a practical way to check every cycle's cure on parts that can't be touched.

Are there lab methods to measure conversion?

Yes, FTIR spectroscopy tracks the disappearing acrylate peak, and DSC measures residual cure exotherm; many quality labs run these in-house during qualification. For routine production, radiometer readings plus mechanical checks are usually sufficient and far faster.

How do I run a dose-response (ladder) study?

Hold intensity and distance fixed, cure sets of coupons at stepped exposure times, and test each set in-house for strength or hardness. Plot property against dose: the curve rises then flattens, the knee defines your minimum, and the flat region is your working territory.

How many parts per step in a window study?

Enough to see scatter, not just averages, five to ten per exposure step is a common starting point, more if strength data is noisy or the joint is critical. Let your own variation and quality requirements set the number rather than a universal rule.

How do I establish the minimum dose?

From the ladder: the lowest dose where every sample meets full requirements, including any environmental or aging tests your product needs. Set it from worst-case parts, thickest adhesive, largest gap, darkest substrate, not the friendliest ones.

How do I establish a maximum dose?

Keep stepping exposure up past the plateau until something degrades, yellowing, embrittlement, substrate heat damage, then set the max safely below that point. If nothing degrades within practical times, cap the window at a sensible multiple of minimum so cycle time stays honest.

What goes into the process window document?

System and channel, wavelength, guide and optics part numbers, working distance and angle, measured irradiance range, exposure time range, dose limits, adhesive product and lot policy, radiometer method, and the test data that justifies it all.

Sign and date it, it's the contract production runs against.

What does a daily verification routine look like?

A quick stack at shift start: radiometer reading at the standard point against limits, a visual on guide tips and fixtures, and a first-article cure with an in-house strength or hardness check, then log all three. Ten minutes that protects the whole shift's output.

What belongs in a weekly or monthly check?

The deeper items: dose measured at the working plane per channel or leg, splitter balance comparison, a timer and shutter spot-check, a fixture distance audit against the marked position, and a review of the trend chart for slope changes.

Monthly, reconcile lamp or LED hours against the output decline to keep replacement planning honest.

What UV process records do auditors ask for?

The validated window with supporting data, radiometer calibration certificates, scheduled output logs with limits and any excursions, revalidation records tied to maintenance events, and your in-house verification results such as first-article strength checks.

Traceability from a shipped part back to the output reading that bracketed it is the gold standard.

How do SkyBeam life logs support process control?

Each channel tracks its LED operating hours, so you can correlate hours against measured output and plan head replacements before drift threatens the window. The logs also give auditors a clean usage record per channel.

Output is low, is it the lamp or the guide?

Isolate it: if a known-good spare guide reads normal on the same lamp, the old guide is degraded; if every guide reads low, it's the lamp or its coupling. Guides typically outlast two to three lamp changes, so with an aging lamp the odds favor the lamp.

How do I measure substrate UV transmission?

Take an irradiance reading at your working plane, then repeat with the actual substrate coupon between tip and sensor at the real geometry. The ratio is your transmission factor, extend exposure accordingly and validate the through-substrate cure with strength checks.

How do I verify cure at the base of a deep pot?

Destructively, during qualification: cross-section cured samples or dig out the fill and probe the deepest adhesive for hardness and dryness. If depth is marginal, switch to layered fill-and-cure or a dual-cure product rather than stretching exposure.

How can I confirm nothing is uncured in shadows?

During qualification, disassemble representative parts and inspect every fillet and hidden region, wet adhesive is obvious on teardown. Fluorescing adhesive plus a mirror or borescope helps on geometry you can't easily open.

If shadows persist, fix the light path or the chemistry, inspection can't cure anything.

Should I record readings in mW/cm² or W/cm²?

Either, as long as you're consistent and label them: 1 W/cm² equals 1,000 mW/cm². Spot intensities often read in W/cm² near the tip while datasheets quote mW/cm², so unit slips are a classic source of thousand-fold errors, write the unit on every log entry.

Is intensity even across the whole spot?

No, spots are typically hottest at center and fall off toward the edge, and optics change the profile. Make sure the joint sits in the usable core of the spot, and validate with adhesive at the spot-edge position if placement tolerance can put it there.

How do I map the intensity profile of my spot?

Move a small-aperture sensor across the spot in steps at the working plane, logging position and reading, or step the spot across a fixed sensor using a positioning jig on the guide. The resulting profile tells you the usable spot diameter for your minimum intensity.

How do I audit fixture positioning repeatability?

Cycle the fixture as production would, load, clamp, unload, while measuring irradiance at a fixed sensor each cycle. The spread in readings is your positioning noise; if it eats meaningful margin, stiffen the fixture or add hard stops.

A reading is below the action limit, now what?

Stop the station, quarantine product made since the last passing check, find and fix the cause (lamp, guide, distance, or meter), re-measure, and disposition the quarantined parts with in-house strength checks against your window.

Then write the excursion up, the trend chart usually shows it was coming.

Why bracket production between output checks?

Because a passing reading only vouches for parts made after it, if the next check fails, everything between the two checks is suspect. Bracketing (pass before, pass after) is what lets you confidently release the interval.

That's also why check frequency should match how much product you can afford to quarantine.

How do I verify my exposure timer is accurate?

Use an integrating radiometer: dose divided by irradiance gives effective exposure seconds, which you compare with the setting. On arc systems this also captures shutter open and close behavior, which a stopwatch never will. Check it at commissioning and after service.

Should I verify LED intensity settings with a meter?

Yes, the panel percentage is a command, not a measurement. Confirm the delivered irradiance at each setting you actually use, and record the pairs. Matching recipes across machines by same-percent without measuring is a classic transfer mistake.

How should I re-baseline after maintenance?

Any service that touches the optical path, lamp, guide, lens, coupling, invalidates the old baseline. Take fresh tip and working-plane readings, log them as the new 100% with the maintenance event referenced, and restart your drift trend from there.

Do I need a jig for the radiometer sensor?

Strongly recommended. A simple nest that locates the sensor at the exact working distance and on-axis makes readings repeatable to a degree hand-holding never will, and repeatability is what trend charts live on. Label the jig and reference it in the SOP.

Does measuring at an angle skew the reading?

Yes, irradiance on the sensor falls off roughly with the cosine of the tilt angle, and an off-axis tip can miss the sensor's calibrated field. Keep the sensor perpendicular to the beam at the same geometry as the part, every time.

Can sensor heating affect radiometer readings?

Yes, intense spots heat sensor heads, and readings can drift within seconds under a high-power beam. Take short, consistent readings with the same dwell each time, let the head cool between measurements, and follow the meter maker's duty-cycle guidance.

What safety practice applies when measuring UV?

The same as curing: UV-blocking eyewear on, no line of sight into the tip or reflections, shields in place, and hands out of the beam, a measurement session means more open-beam time than production does. Our UV safety guide covers eyewear and shielding.

Should the arc lamp warm up before I measure?

Yes, measure only after the lamp stabilizes, a few minutes after strike, or your baseline captures warm-up sag instead of steady-state output. Trend charts built from mixed cold and warm readings look noisy for no real reason.

Should I keep cured reference samples?

It's cheap and useful: retain a few first articles or witness coupons per lot with their dose records. When a field question or audit arrives months later, you have physical parts tied to known exposure data that you can re-test in-house.

How repeatable is my measurement itself?

Check it like any gauge: same operator, same jig, ten consecutive readings, the spread is your measurement noise. Keep it small relative to your alarm-limit spacing; if it isn't, fix the jig or technique before tightening process limits the meter can't resolve.

Should I verify a brand-new light guide?

Yes, measure it on arrival against the incumbent. You confirm it's healthy, capture its baseline while it's fresh, and see exactly how much output the process gains, which matters because your validated time was tuned to the tired guide it replaces.

Is a 20% output drop really the replace point?

It's our recommended action point for guides and lamps: beyond that you're eating the margin the window was built on. A process with unusually generous dose margin may tolerate slightly more, but establish that deliberately in the window document, not by drifting into it.

How do I match two stations to one recipe?

Measure each station at the working plane with the same radiometer and adjust distance, intensity setting, or time until delivered dose matches within a tolerance you define, don't clone knob settings. Then validate first articles from each station against the same window.

How do I transfer a cure process between systems?

Reproduce the delivered dose, not the settings: match the wavelength, measure irradiance at the working plane on the new system, recompute time for the same J/cm², and confirm with in-house first-article checks. Document the new system's parameters as its own validated instance.

Moving from arc to LED, can I keep my recipe?

No, treat it as a new process. The spectrum changes from broadband to a single line, so the photoinitiator sees different energy even at equal measured UVA, and surface cure behavior can shift. Confirm the adhesive is LED-curable, then rebuild the window on the LED source.

Are identical settings proof of identical output?

No. Different lamp ages, guide conditions, and coupling mean two systems at the same settings can deliver meaningfully different irradiance. Output is something you measure at the working plane, settings are just how you ask for it.

Can measurements detect fixture distance creep?

Yes, that's a main job of the working-plane reading. Because intensity falls steeply with distance, a few millimeters of creep shows up as a clear step or slope in the trend chart even when the source itself is healthy. Pair the reading with a physical distance audit to confirm.

How should I trend output data over time?

Plot readings as percent of baseline with warning and action limits drawn in, and watch slope as much as level, a steepening decline forecasts when you'll cross the limit, which turns lamp changes into scheduled maintenance instead of line-down surprises.

How many readings should one check consist of?

Take at least three and record the average, and the spread if your chart tracks it, single readings carry positioning and meter noise. If the three disagree beyond your usual repeatability, don't average the problem away; find out why.

Do I re-measure after changing the focusing lens?

Yes, any optics or spot-size change is an intensity change by definition, since the same output spreads over a different area. Re-measure at the working plane, recompute your exposure time, and treat it as a process change with a first-article check.

Does sensor size need to match my spot size?

It matters: if the spot underfills the sensor, the meter averages beam and dark area and under-reads peak intensity; if it overfills, you read only the center. Know your meter's aperture, keep the geometry constant, and compare only like-for-like readings.

How do I verify all four SkyBeam channels?

Treat each channel as its own process: per-channel baseline, per-channel scheduled readings at that channel's working plane, and per-channel limits, wavelengths and guides differ, so numbers aren't interchangeable. The controller's per-channel life logs slot neatly into the same records.

Should the meter's ID appear in my records?

Yes, record the radiometer model, serial, and calibration due date with every log entry. When readings shift, the first question is whether the meter changed, and when auditors trace a release decision, the instrument's calibration status is part of the evidence.

How do we run shear or pull tests in-house?

Fixture the cured joint, load it to failure in a consistent direction and rate, and record the value and the failure mode, adhesive failure versus substrate break tells you as much as the number. A simple force gauge with a fixture covers most spot-cure joints; consistency matters more than fancy equipment.

What goes into a cure verification SOP?

The measurement point and jig, meter and calibration requirements, check frequency, pass limits, the first-article procedure, the reaction plan for failures (stop, quarantine, diagnose), and where results get logged. Keep it to one page an operator can actually follow at 6 a.m.

Can I check dose without stopping production?

Partially, a quick tip reading between cycles trends source health with seconds of downtime, and witness coupons ride along with real parts. Full working-plane checks need the fixture, so schedule them at shift boundaries; that's why layered checks exist.

What's a warning limit vs an action limit?

The action limit is where product is at risk and the station stops; the warning limit sits above it and exists to trigger planned maintenance while parts are still good. The gap between them should cover at least one check interval of expected drift.

My new lamp reads higher than baseline, problem?

It's expected, a baseline taken on an aged lamp will be beaten by a fresh one. Verify you're still inside your window's maximum for overdose, reset the baseline to the new lamp, and treat the extra output as the healthy start of the next decline curve.

Does a passing meter check guarantee good parts?

No, it verifies the light-source side only. A wrong adhesive lot, bad dispensing, shadowing, or contaminated surfaces all pass a radiometer check and still make bad joints. That's why output checks pair with first-article strength checks and visual inspection.

How do I verify dose through the actual fixture?

Where geometry allows, put the sensor exactly where the joint sits, through any windows, clamps, or apertures the light actually traverses, because fixtures can clip the beam edge. If the sensor can't fit, measure with the fixture but without the part, and correlate during qualification.

How do lamp hours help interpret readings?

Hours give context to output: a drop at high hours is normal aging headed toward replacement, while the same drop at low hours points at the guide, coupling, or fixture instead. Log hours with each reading so the trend chart can tell those stories apart.

What if my radiometer maxes out at the tip?

Spot tips can exceed a meter's range, the SunSpot 2 delivers over 18 W/cm² of UVA. Measure at the working plane instead, where the process actually runs, use an attenuating accessory from the meter maker, or select a high-range instrument specified for spot sources.

Do output checks replace strength checks?

No, they answer different questions. Radiometer data confirms the input, the dose delivered; strength and hardness checks confirm the outcome, a good joint. A robust in-house plan keeps light-side checks frequent and part-side checks periodic, so each covers the other's blind spots.

How does process capability apply to dose?

Treat delivered dose as the characteristic: its mean should sit centered in your window with the spread, drift plus fixture noise plus measurement noise, small relative to window width. If routine variation regularly flirts with the window edge, widen margin at the source, don't tighten paperwork.

Which events should trigger an extra dose check?

Any unexplained joint failure, tack complaint, or appearance shift; any maintenance or collision near the cure station; adhesive lot changes; and power sags on arc systems. An unscheduled reading is cheap, run one whenever the process gives you a reason to doubt it.

How exact must distance be when measuring?

As exact as your process claims, the jig should reproduce the validated distance within a fraction of a millimeter, because near-tip gradients are steep. If readings wobble more than the source plausibly could, the jig is usually what's moving, not the lamp.

Can I estimate dose from lamp hours instead?

No, hours predict aging on average, but individual lamps, guides, and fixtures vary too much to certify dose by arithmetic. Hours are for planning replacements; radiometer readings are for releasing product. Use each for its own job.

What is solarization and how do I detect it?

It's UV-induced degradation of light-guide materials that gradually reduces transmission, invisible from the outside. You detect it the only way possible: falling tip readings against baseline with a known-good lamp. That's a main reason the 20%-drop guideline exists.

Should each log entry note who took the reading?

Yes, initials or an operator ID belong on every line. It isn't about blame: when a reading looks odd, technique questions are the fastest to resolve, and audits expect attributable records. It also flags when a new operator needs a walkthrough of the jig procedure.

What's the fastest setup for a valid daily check?

A dedicated sensor nest at each station: drop the meter in, fire a timed exposure, read, log, done, under two minutes. The nest fixes distance and centering so the number is comparable day to day, which is the entire point of a daily check.

Can you help me plan a dose verification program?

The measurements and part checks happen in your facility, on your equipment and your product, that's what makes them valid. We're glad to advise on system choice, guide selection, and measurement practice, and our application notes lay out the routines we recommend.