Gidsen 

Start with four different numbers

Irradiance is incident optical power per unit area, commonly reported in mW/cm². It needs a location: a value measured at the center is a value at that position, not a description of the whole face. FDA’s US photobiomodulation submission draft distinguishes irradiance from other output parameters and asks for the relevant anatomical site and source-to-site distance. It remains a nonbinding draft, not a universal operating standard. [1]

These are statistical definitions, not clinical thresholds. Ask whether “peak” means a spatial maximum or the instantaneous maximum during a pulse. Those are different comparisons.

Define the area before interpreting the average. A report averaged over a small central rectangle answers a different question from one averaged over a larger rectangle including the edges. Neither is automatically a full-face measurement.

A simple example: the same peak, different coverage

The following numbers are invented only to demonstrate arithmetic. They are not device measurements, recommended outputs or safe exposure limits. Assume each value represents an equal-area cell in a three-by-three region; all values are in mW/cm².

A central reading of 60 would conceal the difference. In this equal-area example, Pattern A’s mean is (8 × 20 + 60) ÷ 9; Pattern B’s is (8 × 40 + 60) ÷ 9. This comparison describes output distribution only. It does not show that either pattern is preferable for skin.

An average can hide variation too

Two equal-area cells with values of 20 and 60 have the same mean as two cells with values of 40 and 40: 40 mW/cm². Their minima and maxima differ. A report giving only the average still leaves a coverage question unanswered.

A laboratory grid usually consists of sampled readings rather than known whole-cell averages. A simple arithmetic mean estimates an area mean only when the sampling and weighting justify it. Clustering many readings near the center should not give that small area disproportionate weight.

If a report gives a “uniformity percentage,” look for the equation. Minimum ÷ mean and minimum ÷ maximum are different ratios. Do not compare their percentages as if they were the same metric, or treat either as a universal pass mark.

A flat map does not measure facial curves

Optical radiation depends on position and direction. The receiving surface’s angle matters, and light from an extended source involves contributions from different source locations. The NBS radiometry manual explains these relationships and the limits of using a single distance in an inverse-square approximation. [2]

Applied to a face, this means a flat detector plane cannot by itself establish readings at the nose, curved cheeks and recessed areas. Those surfaces have different orientations and separations. An obstruction also removes direct paths from some emitters; other paths or reflected light are not quantified by simply labeling the face “covered.” This is a geometry explanation, not a measured estimate of how much light any facial region receives. [2]

Array geometry belongs in the report

Request a drawing of the array, its optics and the measured region. That lets you locate the readings relative to the emitters and identify what was actually tested. For a mask, ask whether the map used a flat fixture or a specified three-dimensional head form, and how fit and openings were represented. For a multipanel system, ask which panel angles and arrangement were used. These are documentation questions, not an assertion that every device needs the same fixture.

Do not calculate a missing map by assuming that doubling the distance quarters every reading. NIST’s work on limiting photometric distance shows that far-field applicability depends on source size, beam distribution and the accepted error. That is metrology context, not validation of a beauty-device distance rule. [3]

Compare maps only after comparing their labels

Use this reading checklist before placing two reports side by side. It is an editorial aid, not a certification procedure.

NIST’s spectral-irradiance calibration publication addresses measurement procedures, traceability and uncertainty. It supports asking for that basis; it does not certify a particular output map or provide a home measurement recipe. [4]

If one map lacks the test position or measured region, record it as not comparable from the supplied data. Avoid filling the gap with a calculated correction factor. Maps at different labeled positions can describe how a device was characterized without authorizing a user to operate at all of those positions.

Ask for separate bands in each relevant mode

A combined map contains less information than its components. For an arithmetic illustration, suppose one position receives 30 units from a red band and 10 from a near-infrared band, while another receives 10 and 30. Both totals are 40; neither band is spatially equal between the positions. These are hypothetical values, not measurements or an effect comparison.

Request the red and near-infrared contributions during the actual combined operating mode, with the wavelength ranges and separation method stated. Separate single-band-mode tests do not establish those contributions unless the report verifies that the operating outputs match.

Also ask whether values are continuous, time-averaged over pulsing, or pulse peaks. FDA’s draft treats output mode and pulsing parameters separately from irradiance. [1] For spatial comparison, keep the temporal basis consistent across the map.

Radiometric measurement concerns optical power; a visual brightness reading is a different quantity. NIST’s LED measurement work discusses spectral-response limitations in photometric LED measurements. A lux reading or phone photograph therefore needs more than a bright image to support a radiometric output claim. [5] A report should explain how its instrument handles the measured bands; do not infer a fixed meter error from the instrument’s appearance.

Three shortcuts to avoid

• “The center is the specification for the whole area.” It is one position unless the report establishes a broader distribution.
• “A uniform map proves uniform results.” An optical report does not contain a human outcome study. It cannot demonstrate equal visible changes across the face.
• “A dim region just needs extra time.” Multiplying exposure everywhere does not selectively correct that region. The map does not authorize a longer session or repositioning.

Frequently asked questions

Does a dark patch mean an LED has failed?

The map alone cannot diagnose a failed emitter. It describes the values under the reported conditions. Checking a suspected fault requires the device’s service information and an appropriate inspection; changing the setup or repairing an emitter is outside what the map permits.

Can a report leave eye openings out of its average?

It can define a region that excludes openings, provided those exclusions are explicit. The resulting average applies only to the included region. It says nothing about exposure through the openings or whether the eye-protection arrangement is adequate.

Does “minimum” guarantee there is no lower point?

A sampled minimum is the lowest observed reading. It does not guarantee that the grid captured every dip. Request the sampling and detector-aperture details; distinguish measured minima from minima predicted by a model.

Can a normalized map compare two output levels?

A map expressed as a percentage of its own maximum can compare relative patterns, if regions and methods match. It cannot establish absolute output without the maximum and units. Two identical normalized maps can represent different irradiances.

Read the distribution behind the headline number

The useful question is: “What was measured, where, and under which conditions?” Keep center, peak, mean and minimum separate, and leave missing evidence visible. That makes an output map useful without turning it into an unsupported treatment instruction.