What a Light Loss Factor Is Accounting For

Why this matters

Every lighting calculation you will ever be handed carries one number that nobody on the job can source. It sits in the report as "LLF 0.72" and it is the difference between the light the fixtures make and the light the specification promises. When the space measures short four years later, that number is the first thing everyone points at and the last thing anyone can decompose, because it arrived as a single figure with no parts. A light loss factor is not one number. It is a product of up to nine separately-sourced terms, each owned by a different party, and knowing which term moved is the entire difference between a diagnosis and an argument.

It is a product, and it is a re-basing

Maintained illuminance equals initial illuminance multiplied by the light loss factor. That is the whole relationship, and two things follow that people get wrong constantly.

It multiplies, it does not subtract. Six terms at 0.95 each do not cost 30 percent; they cost 1 - 0.95^6 = 26.5 percent. That looks like a small distinction until the terms are uneven, and then the ordering intuition breaks entirely.

A maintained figure already contains it. Applying a loss factor to a number that is already maintained is the single commonest arithmetic error in this subject, and it always runs in the pessimistic direction on a design and the flattering direction on a complaint. If you need to move between the two, you divide to go from maintained back to initial and multiply to go the other way. It is a re-basing of one number, never a second number to stack a factor on.

The recoverable terms

These are the ones a maintenance program can get back. Their values are not properties of the equipment; they are properties of your maintenance interval, which is why the same fixture in the same room takes a different factor under a 12-month cleaning cycle than under a 48-month one.

• Luminaire dirt depreciation. Dirt on and in the optical assembly. Driven by the atmosphere category of the space and the cleaning interval, and the manufacturer's or the recommended practice's curve is the source. This is usually the largest single loss in an industrial space and the most under-estimated in a design.
• Room surface dirt depreciation. Dirt on ceiling and walls, which matters in proportion to how much of the delivered light arrives after a bounce. Large in a low, light-coloured room with indirect fixtures; nearly nothing in a high bay with dark surroundings and direct optics.
• Lamp or module lumen depreciation. Flux decay of the source at the chosen maintenance interval. For a solid-state luminaire that value comes from the projected maintenance curve, and the lumen-depreciation card owns where that projection comes from and what it is allowed to claim.
• Lamp burnout factor. The fraction still working at the interval. Setting this to 1.00 is a statement that failures get replaced promptly, not a property of the product.

The non-recoverable terms

These do not come back with a ladder and a rag. They are conditions of the installation, and they are decided at specification time.

• Ambient temperature at the luminaire. Output falls with ambient, and the derating curve is a separate manufacturer document you have to ask for. A high bay under an uninsulated roof deck in summer is not sitting at the 25 C bench ambient its test was run at.
• Supply voltage at the luminaire. A driver holds output steady across its input range far better than a magnetic ballast did, so this term is much closer to 1.00 than it used to be, but it is not automatically 1.00 at the far end of a long run.
• Driver or ballast factor. The ratio between the output the shipped driver produces and the output produced by whatever powered the photometric test. If the file is an absolute test of the finished luminaire with its own driver, this term is already inside the file and taking it again is double-counting.
• Luminaire surface depreciation. Permanent yellowing, hazing or degradation of lens and reflector materials, distinct from dirt because cleaning does not reverse it.
• Tilt. Applies to source types whose output depends on burning position. On a solid-state luminaire it is generally 1.00, and the photometric file carries a tilt field either way.

What a light loss factor does not account for

This is the useful half of the subject, because almost every "the loss factor must be wrong" argument is actually about something on this list.

• The room. Coefficient of utilization handles geometry and reflectances, and it is a separate multiplier in the same calculation. Folding a room problem into the loss factor double-counts the surfaces and then hides the double-count.
• Uniformity. A loss factor scales every point on the grid by the same amount, so it cannot change a max-to-min ratio at all. If a space went patchy rather than dim, the loss factor is arithmetically incapable of being the cause, and the uniformity card owns what is.
• Anything that changed in the space. Racking raised, a partition added, a wall repainted dark, a process moved under a fixture. None of that is depreciation and none of it is in the factor.
• Controls. A daylight sensor holding output down, a scene preset, a setback, a dimmer left at 70 percent. These produce exactly the symptom the loss factor is blamed for and none of them are in it.
• Fixtures that were never right. Wrong ordering code, wrong aiming, one bay never installed. A calculation assumes the layout it was given got built.
• Your meter. Spectral mismatch between the meter and the installed source is a property of the instrument, and no loss factor corrects for it. The repeatable-reading card owns that term.

Worked example: the same 0.72, built two ways

A distribution warehouse, 48 high bays, designed at 28 fc maintained, horizontal at 30 in above finished floor, with a light loss factor of 0.72.

Build A, decomposed. Lamp lumen depreciation 0.92 at the chosen 60,000 hour maintenance interval, luminaire dirt depreciation 0.84 for a dirty-category atmosphere on a 36-month cleaning interval, room surface dirt depreciation 0.96, ambient temperature factor 0.97 at the 40 C measured ceiling ambient, driver factor 1.00 because the photometric file is an absolute test of the finished luminaire with its shipping driver, burnout factor 1.00.

0.92 x 0.84 x 0.96 x 0.97 x 1.00 x 1.00 = 0.720.

Build B, typed in. Somebody entered 0.72 because that is what the last job used.

Both reports print 28 fc maintained. Both are the same design. Only one of them can be diagnosed.

Year four. The customer reports the aisles are dark. A survey on the same grid, plane and height reads 21.4 fc maintained. Before anything is compared, the design has to be put on one basis: 28 fc maintained at 0.72 re-bases to 28 / 0.72 = 38.9 fc initial. The measured 21.4 fc against 38.9 fc initial means the installation is running an actual in-service loss factor of 21.4 / 38.9 = 0.550, against a designed 0.720. That is where the shop with Build B stops, because there is nothing left to take apart.

Decomposing it. Two terms were measured directly rather than assumed.

Dirt. One luminaire's lens was cleaned and its station re-read: 20.8 fc before, 27.1 fc after. That fixture's actual dirt depreciation is 20.8 / 27.1 = 0.768, against the designed 0.84. The cleaning had never happened; 48 months had passed on a 36-month interval and a new process had moved the space into a dirtier category. Reaching the lens meant a scissor lift rated for the height with the fall protection the equipment requires under 29 CFR 1910 Subpart D for general industry, not a reach from a stepladder, and the lens was damp-wiped rather than dry-brushed after the process SDS was checked, because dry-brushing settled process dust puts it in the breathing zone and an inhalation route needs a respiratory control under a 29 CFR 1910.134 program, not a glove.

Burnout. A walking count found 8 of 48 luminaires dark, an actual burnout factor of 40 / 48 = 0.833 against the designed 1.00. Nobody had been replacing them, and each replacement is electrical work de-energized and locked out under 29 CFR 1910.333(b)(2) with live-dead-live proving per NFPA 70E-2021, 120.5, in the edition the employer's electrical safety program adopts.

Rebuild the product with the two measured terms. 0.92 x 0.768 x 0.96 x 0.97 x 1.00 x 0.833 = 0.548.

Against the measured 0.550, the rebuilt model sits 0.4 percent low, and the direction is the one the arithmetic predicts: the 0.92 depreciation term belongs to a 60,000 hour interval and these fixtures have run roughly 16,000 hours, so that term is more pessimistic than the real decay and pulls the model down. Two measured terms and one known-conservative term account for the whole gap, so no fourth cause needs inventing.

What it buys. Dirt cost 0.768 against 0.84 and burnout cost 0.833 against 1.00, so burnout is the larger of the two, and both are recoverable. Cleaning the run and replacing the eight dark units restores the product to 0.92 x 0.84 x 0.96 x 0.97 = 0.720, which is the design value, and 38.9 fc initial x 0.720 = 28.0 fc maintained. Nothing needs adding. A shop that had only the single 0.72 would have quoted more fixtures.

The caution the factor cannot give you. Eight dark units are not spread evenly. Wherever they cluster, the minimum is worse than the average shortfall implies, and no term in this product can tell you that, because the whole product scales every grid point equally.

Checking your own figures

• The product recomputes. 0.92 x 0.84 x 0.96 x 0.97 = 0.7196, reported as 0.720. Rebuilt: 0.92 x 0.768 x 0.96 x 0.97 x 0.833 = 0.5481, reported as 0.548.
• The factor was applied once, in one direction. 0.72 divides once (28 to 38.9 fc initial) and multiplies once (38.9 back to 28.0 fc maintained). It is never applied to a figure that already carries it.
• The residual closes without a fourth term. Model 0.548 against measured 0.550 is 0.4 percent, and the model sits low, which is the direction the 60,000 hour depreciation term at roughly 16,000 hours of run time predicts.
• Terms compared against their own design values. Dirt 0.768 against 0.84; burnout 0.833 against 1.00. Measured against their own design values, dirt lost 0.768 / 0.84 = 0.914 and burnout lost 0.833 / 1.00 = 0.833, so burnout is the larger departure, which is what "burnout is larger than dirt" is stating.
• Every illuminance carries plane, height and basis. 28 fc maintained, 38.9 fc initial, 21.4 fc maintained, all horizontal at 30 in above finished floor.
• Nine terms named, nine accounted for. Four recoverable and five non-recoverable are defined above. Six carry a value in the example (0.92, 0.84, 0.96, 0.97, 1.00 driver, 1.00 burnout) and the remaining three, tilt, supply voltage and luminaire surface depreciation, are stated as not moving in this case rather than dropped.

References

• IES recommended practice covering light loss factors and maintained illuminance, in the edition your specification or employer standard names, which binds only through that document and is where dirt depreciation categories and curves live
• Manufacturer ambient temperature derating curve and driver specification for the exact ordering code, which own the non-recoverable terms
• 29 CFR 1910 Subpart D, walking-working surfaces and fall protection for general industry, and 29 CFR 1926 Subpart M for construction, where luminaires are reached at height
• 29 CFR 1910.134, respiratory protection program requirements, where cleaning disturbs settled process contaminant identified on the SDS; 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5, for de-energized replacement work
• See related: Why a Lighting Design Is Computed at End of Life; What Lumen Depreciation and Lifetime Ratings Mean; Why Uniformity Matters More Than Average Level