Why does a corrected reading need a band around it?

Because every use of it is a comparison against a threshold. A dose of 1188 against a target of 1200 is a fail; 1188 against 1150 is a pass; and if the true value could be anywhere from 1129 to 1247 then neither comparison has actually been settled. Presenting a single corrected number hides that, and it hides it precisely in the region where the decision is hardest.

How do I estimate drift if I have no history?

Start from the difference between your last two certificates if you have two, and if you have one, be conservative. The point of the input is that it comes from your instrument rather than from a rate this page invents. Radiometers vary widely depending on how much UV they have seen and how hot they have been run. Once you have three or four certificates the trend is usually clear and remarkably linear.

Is drift really one-directional?

Largely, yes. UV detectors solarise and filters yellow, both of which reduce response, so an uncalibrated instrument tends to read progressively low rather than wandering either way. This page treats the band as symmetric, which is conservative above and realistic below. The lower bound is the half worth trusting, and it is the half the worst-case headroom uses.

My correction factor is 1.3. Is that a problem?

It is a signal. A working instrument needing a 30% correction has degraded substantially, and while the certificate makes the reading usable, the instrument's noise and its rate of further drift both tend to rise as its response falls. At that point the calibration is holding together an instrument that is nearing replacement, and treating a large correction as routine is how a meter stays in service past the point of being trustworthy.

Does correcting the instrument fix a band mismatch?

No, and this is the bigger error of the two. A calibration factor scales the instrument's response within the band it measures. If your photoinitiator absorbs at 254 nm and your radiometer measures UVA, then no factor makes the reading relevant. You are precisely measuring something that is not the thing that cures the material. Check the band before the calibration; it is the error that a corrected number is best at hiding.

Why do two radiometers disagree by more than their stated uncertainty?

Because their band responses differ. Two instruments both labelled UVA weight the wavelengths inside that band differently, so against a lamp with a peaky spectrum they integrate different amounts of it and legitimately disagree. This is why the practical advice is to trend one instrument rather than to compare across them, and why a process specification should name the instrument as well as the dose.

UV Curing calculator

UV Radiometer Calibration Correction Calculator (with Drift Band)

Correct a raw radiometer reading and see the uncertainty around it. Enter the reading the instrument displayed, the correction factor from its calibration certificate, the dose the process must deliver, and how far the instrument may have drifted since it was calibrated. The calculator returns the corrected dose, the band drift puts around it, and the headroom over target at both the nominal reading and the bottom of the band. A corrected reading is the one number in UV curing that should never be presented as a point, because every decision downstream is made against a threshold.

What this calculator does

Correct a raw radiometer reading with its calibration factor, and show the band the instrument's drift since calibration puts around it.
Use it for deciding whether a borderline process reading really passes, setting a calibration interval from observed drift rather than a default, reconciling a reading against a dose target before a customer audit, checking whether an instrument's correction has grown large enough to retire it, explaining why two radiometers on the same line disagree.
Correct a raw radiometer reading with its calibration factor, and show the band the instrument's drift since calibration puts around it.

Formula used

Corrected dose = raw reading × calibration factor
Lower bound = corrected dose × (1 − drift ÷ 100)
Upper bound = corrected dose × (1 + drift ÷ 100)
Headroom (nominal) = corrected dose − target dose
Headroom (worst case) = lower bound − target dose

Inputs explained

Raw radiometer reading: What the instrument displayed, before any correction.
Radiometer calibration correction factor: The multiplier from the calibration certificate. 1.08 means the instrument reads 8% low and the certificate says so.
Target cure dose: The dose the process must deliver, in the same band the instrument measures.
Drift since calibration: How far the instrument may have moved since its certificate, from your own calibration history. Set it to zero to see the nominal correction alone.

How to use the result

Best suited to deciding whether a borderline process reading really passes, setting a calibration interval from observed drift rather than a default, reconciling a reading against a dose target before a customer audit, checking whether an instrument's correction has grown large enough to retire it, explaining why two radiometers on the same line disagree.
Drift is treated as symmetric, and it usually is not: UV detector degradation reduces response, so the real distribution skews low. The lower bound is the reliable half of the band. Says nothing about band mismatch, which is a larger error than calibration for most processes and which no correction factor can fix. Does not model temperature effects, which are significant for instruments used near a hot mercury lamp. Two instruments from different makers can disagree well beyond either one's calibration uncertainty; this page cannot reconcile them. A corrected dose is still a process setting rather than cure evidence.

Common questions

Why does a corrected reading need a band around it? Because every use of it is a comparison against a threshold. A dose of 1188 against a target of 1200 is a fail; 1188 against 1150 is a pass; and if the true value could be anywhere from 1129 to 1247 then neither comparison has actually been settled. Presenting a single corrected number hides that, and it hides it precisely in the region where the decision is hardest.
How do I estimate drift if I have no history? Start from the difference between your last two certificates if you have two, and if you have one, be conservative. The point of the input is that it comes from your instrument rather than from a rate this page invents. Radiometers vary widely depending on how much UV they have seen and how hot they have been run. Once you have three or four certificates the trend is usually clear and remarkably linear.
Is drift really one-directional? Largely, yes. UV detectors solarise and filters yellow, both of which reduce response, so an uncalibrated instrument tends to read progressively low rather than wandering either way. This page treats the band as symmetric, which is conservative above and realistic below. The lower bound is the half worth trusting, and it is the half the worst-case headroom uses.
My correction factor is 1.3. Is that a problem? It is a signal. A working instrument needing a 30% correction has degraded substantially, and while the certificate makes the reading usable, the instrument's noise and its rate of further drift both tend to rise as its response falls. At that point the calibration is holding together an instrument that is nearing replacement, and treating a large correction as routine is how a meter stays in service past the point of being trustworthy.
Does correcting the instrument fix a band mismatch? No, and this is the bigger error of the two. A calibration factor scales the instrument's response within the band it measures. If your photoinitiator absorbs at 254 nm and your radiometer measures UVA, then no factor makes the reading relevant. You are precisely measuring something that is not the thing that cures the material. Check the band before the calibration; it is the error that a corrected number is best at hiding.

Last reviewed 2026-08-25.