What Glare Is, and the Two Kinds That Matter

Why this matters

The most expensive mistake in lighting work is answering a visibility complaint by adding light. It is the obvious move, it is measurable, it is easy to sell, and in a space with a glare problem it delivers exactly nothing: the contrast the reader is actually working with comes out identical to four significant figures, while every meter reading doubles. The customer then has a brighter room, the same complaint, and a paid invoice, and the shop has no argument left. Knowing which of two complaints you are holding takes one extra reading and about five minutes of arithmetic.

Glare is a luminance ratio, so the meter cannot see it

An illuminance meter has a cosine-corrected cell on a plane. It sums arriving flux and discards where it came from. Glare is entirely about where it came from: the luminance of a source and its position relative to where the person is looking, against the luminance of everything else in the field of view. Two of those three quantities are invisible to a horizontal footcandle reading, which is why a retrofit can raise the measured average and make the space worse. The quantity split among flux, intensity, illuminance and luminance belongs to the units card; what matters here is that only the last of the four is what the eye is responding to when it complains.

Glare then divides in two, and the two are diagnosed differently because only one of them costs performance.

Disability glare: the veil that subtracts contrast

A bright source in the field of view scatters light inside the eye. That scattered light lands as a roughly uniform veiling luminance over the whole retinal image, adding to the task and to its background equally.

Adding the same quantity to both is not neutral, because contrast is a ratio. For a dark task on a light background:

Contrast = (L background - L task) / L background

Add a veiling luminance Lv to both terms and the numerator cancels it while the denominator does not:

Contrast with veil = (L background - L task) / (L background + Lv)

The numerator is untouched and the denominator grows, so contrast falls, always, and the amount it falls is set entirely by Lv against the background luminance.

Lv itself rises with the illuminance the glare source produces at the eye and falls with roughly the square of the angle between the source and the line of sight, holding the background and the source position otherwise fixed. The proportionality coefficient is not a constant: it rises substantially with observer age, roughly doubling between a young adult and someone in their seventh decade, and the angular-square form is an approximation good for small to moderate off-axis angles. The coefficient and its age dependence belong to the CIE treatment of disability glare in the edition your specification names; what you carry into the field is the shape of the relationship and the fact that a crew of mixed ages does not experience the same space.

Discomfort glare: a judgement with an index, not a measurement

The second kind is the sensation of annoyance from a bright source, and it can be strong where measured performance is untouched. It is indexed rather than measured. The interior index rises with source luminance and with the solid angle the source subtends, and falls with background luminance and with the source's angular displacement from the line of sight.

The consequence people get wrong: that index is a property of a room and an observer position, not of a luminaire. It cannot be printed on a cut sheet, because it needs the ceiling height, the surfaces, the layout and where the person is standing and looking. A manufacturer quoting one is quoting it for an assumed room, and the assumed room is the part that matters.

The exception is outdoor equipment, where a luminaire classification into backlight, uplight and glare zones is derived from the candela table alone, so it genuinely is a luminaire property and it does live in the photometric file the file card describes.

The gate

Will raising the illuminance improve this complaint?

Only if the limiting factor is adaptation luminance. Where a glare source sits in the field of view, a proportional increase in level scales the task luminance, the background luminance and the veiling luminance by the same factor, and a ratio does not change when you multiply every term by the same number. Two cases, same task, same materials.

Both use a matte pick ticket: paper reflectance 0.75, print reflectance 0.10. Illuminance is read in footcandles and converted once at entry; every luminance below is in cd/m2.

Outcome one: the parts counter, gate open

Horizontal task at the counter, 5 fc maintained, horizontal, at 30 in above finished floor, which is 53.8 lux. Indirect luminaires well above the line of sight, nothing bright in the field of view, so the veiling term is negligible.

• Paper: 53.8 x 0.75 / pi = 12.85 cd/m2
• Print: 53.8 x 0.10 / pi = 1.71 cd/m2
• Contrast = (12.85 - 1.71) / 12.85 = 0.867

Raise the level to 30 fc maintained, horizontal, at 30 in, which is 322.9 lux:

• Paper 77.09 cd/m2, print 10.28 cd/m2, contrast = 0.867

Contrast did not move, and it cannot, because both terms scaled by the same factor of 6.0. What did move is the adaptation luminance, from 12.85 to 77.09 cd/m2, a factor of 6.0. Small print at low adaptation luminance is slow and error-prone, and the improvement here is real: it comes from the eye operating higher on its performance curve, not from any change in the printed contrast. Gate open, add light, and say honestly why it will work.

Outcome two: the pick aisle, gate closed

Same ticket, now on a vertical pick face. The reading everyone takes is horizontal at the floor; the reading that matters is 12 fc maintained, vertical, at 48 in above finished floor, which is 129.2 lux. Take that reading with the aisle blocked to powered industrial truck traffic for the duration rather than relying on being seen, since 29 CFR 1910.178 governs how the truck is operated and imposes nothing on the pedestrian standing in its path.

• Paper: 129.2 x 0.75 / pi = 30.84 cd/m2
• Print: 129.2 x 0.10 / pi = 4.11 cd/m2
• Contrast, no veil = 0.867

Now the high bay. At the angle toward the picker's eye it emits 2,000 cd, and the eye is 6.71 m from it. The high bay's maximum luminous dimension is 0.40 m, so the five-times point-source boundary the inverse-square card sets is 2.0 m and the 6.71 m separation clears it. The illuminance at the eye is 2,000 / 6.71^2 = 44.4 lux. The source sits 8 degrees off the line of sight, which is what a face-height pick position in a tall aisle produces.

• Veiling luminance, young-observer coefficient: 6.94 cd/m2
• Contrast = 26.73 / (30.84 + 6.94) = 0.708, an 18.4 percent loss against 0.867
• Same geometry, older-observer coefficient roughly doubled: veil 13.88 cd/m2, contrast 0.598, a 31.0 percent loss

Now double the installed output, same luminaires, same positions. Vertical illuminance goes to 258.4 lux, the eye illuminance to 88.8 lux, and every term scales:

• Paper 61.68, print 8.22, veil 13.88 cd/m2
• Contrast = 53.46 / 75.56 = 0.708

Identical. The meter reads twice what it did, and the picker's task is exactly as hard as before, for the young observer and for the older one alike.

The move that actually worked

Geometry, at the original level. Shield the source, or relocate it so it sits farther off the picker's line of sight. Take the same luminaire from 8 degrees to 25 degrees off axis with the eye illuminance unchanged at 44.4 lux:

• Veil falls to 0.71 cd/m2
• Contrast = 26.73 / 31.55 = 0.847, against 0.708 before, a 19.7 percent recovery, at 12 fc vertical and not one watt added

That is the whole lesson in one comparison: doubling the light bought a contrast change of zero, and moving the source bought back most of what the glare had taken. What would flip it: if the aisle had no source within about 30 degrees of the line of sight and the complaint persisted, the veil term is small, the problem is back to adaptation luminance, and outcome one applies.

Checking your own figures

• Contrast recomputed at every level, not assumed constant. Counter: 0.867 at 5 fc and 0.867 at 30 fc. Aisle with veil: 0.708 at 12 fc and 0.708 at 24 fc. The invariance is a printed result, not a claim.
• The one thing that did change in outcome one is named. Adaptation luminance 12.85 to 77.09 cd/m2, a factor of 6.0, matching the illuminance factor of 6.0 exactly, as it must for a fixed reflectance.
• Every illuminance carries plane, height and basis. Counter 5 fc and 30 fc maintained, horizontal at 30 in above finished floor; aisle 12 fc and 24 fc maintained, vertical at 48 in above finished floor. The two are not compared, because they are different planes serving different tasks.
• The veil scaled with the level, and was not held fixed. Doubling took it from 6.94 to 13.88 cd/m2. Holding it fixed would have shown a fake improvement, which is exactly the error this article exists to prevent.
• The age coefficient was applied as a doubling to the veil only, giving 13.88 cd/m2 and contrast 0.598, and it was not applied to any surface luminance, because scatter inside the eye does not change what a surface emits.
• Unit conversion done once at entry and printed. 5 fc to 53.8 lux, 12 fc to 129.2 lux, 30 fc to 322.9 lux, all at 10.764 lux per footcandle, and every luminance thereafter in cd/m2.
• The relationship's held-constant conditions were stated before use. The inverse-square-of-angle form holds the background and source position otherwise fixed, and the coefficient is stated per observer age rather than as a universal constant.

References

• CIE treatment of disability glare and veiling luminance, and the interior discomfort glare index, in the editions your specification names, which bind through that specification rather than on their own
• IES TM-15, the backlight, uplight and glare classification for outdoor luminaires, in the edition your specification or ordinance names, which is derived from the candela table
• 29 CFR 1910.178, powered industrial truck operation, which governs the truck and not the pedestrian working in its aisle
• See related: What Light Is Measured In, and Why the Units Confuse People; What a Photometric File Actually Describes; Why Lumens and Lux Are Not Interchangeable