What Emergency and Exit Lighting Have to Do

Why this matters

A building can pass its monthly emergency lighting check every month for seven years and still lose the path long before the rated duration is up. The monthly check proves the transfer relay works and the lamps light. It proves nothing at all about how long they stay lit, which is the only property anyone will ever need. That gap is the whole subject of this card, and the artifact that closes it is a timed discharge test with illuminance measured at the floor at the END of the rated duration, not at the start.

Two devices, two jobs, two different failures

Exit signs and emergency lighting get bought together, mounted together, and inspected together, and they do different work.

An exit sign marks. Its job is to be legible from a distance in the direction of travel and to identify which door is the one. 29 CFR 1910.37(b) is the federal floor for general industry: each exit must be marked by a sign reading "Exit," the lettering is not less than six inches high with a principal stroke of not less than three-quarters inch wide, and a door or passage along an exit access that could be mistaken for an exit has to be marked "Not an Exit" or with its actual use. That standard runs to your employees. The adopted building and fire codes reach the building and everyone in it, and they add sign spacing, directional indicators, and photometric performance.

Emergency lighting illuminates. Its job is to make the marked path walkable when normal power is gone. 29 CFR 1910.37(b) requires each exit route to be adequately lighted so that an employee with normal vision can see along it, which is a qualitative federal duty. The quantitative requirement, and the 90-minute duration everyone quotes, comes from NFPA 101, in the edition your authority having jurisdiction has adopted and amended. The AHJ is a named role with authority to interpret and enforce that adopted edition, not a synonym for whoever inspects.

The failures diverge too. Signs fail visibly: a dead legend, a sign painted over, a sign pointing at a door that is now a storeroom. Emergency lighting fails invisibly, because a unit with a half-dead battery looks exactly like a unit with a good one until the ninetieth minute of an event nobody has ever run.

A light level is not a measurement without its basis

This is where most reported numbers fall apart. An illuminance figure needs five things attached before it means anything, and the adopted NFPA 101 requirement states all five.

• The quantity and unit. Illuminance in foot-candles, which is lumens per square foot falling on a surface, not the output of the fixture.
• The plane and location. Measured at floor level, along the designated path of egress. Not at eye level, not on a desk, not under the fixture.
• The statistic. An average along the path AND a minimum at any single point. Both, not either.
• The time basis. Initial, meaning at the moment normal power is lost, or end-of-duration, meaning at 1.5 hours of battery operation. These are two different acceptance criteria with two different numbers.
• The duration itself. Not less than 1.5 hours.

With those attached, the adopted NFPA 101 criteria read: initially an average of not less than 1 foot-candle along the path at floor level and not less than 0.1 foot-candle at any point, permitted to decline over the 1.5 hours to an average of not less than 0.6 foot-candle and not less than 0.06 foot-candle at any point, with a maximum-to-minimum uniformity ratio not exceeding 40 to 1. Where in that window the uniformity ratio is evaluated belongs to the adopted edition, so confirm it with the AHJ rather than assuming.

A number reported as "we read 1.2" with none of that attached tells you nothing, and in practice it usually means somebody held a meter at chest height under a fixture with the normal lights on.

Why the 30-second test cannot find the failure

The adopted NFPA 101 edition requires both a monthly functional test of at least 30 seconds and an annual test for the full 1.5-hour duration, with written records; self-testing and self-diagnostic equipment is permitted in lieu of the manual test where the adopted edition allows it, with its results recorded and a visual inspection performed annually.

The 30-second test loads the battery for 30 seconds. A sealed lead-acid or nickel-cadmium cell that has lost most of its capacity to age and to cabinet heat will still deliver full lamp current for 30 seconds, because 30 seconds asks for almost no stored energy. Capacity fade is the dominant failure mode in this equipment and it is the one failure the monthly test structurally cannot detect. That is not a criticism of the monthly test; the monthly test is checking the transfer circuit, the charger, and the lamps, and it does that well. It is a warning against treating twelve monthly passes as evidence about duration.

Two consequences follow that people get wrong in the field. First, a unit that fails the 90-minute test almost never needs a fixture; it needs a battery, and the age of the head is a better predictor than anything you can see. Second, a full discharge leaves the batteries flat, and they need recharge time, commonly on the order of 24 hours, before the building has its rated duration back. A building tested to full discharge on Friday afternoon is a building with reduced emergency lighting capability on Friday evening, which is a bounded out-of-service condition and gets recorded as one.

Worked artifact: one annual duration test, filled in

Second floor of an office suite. The designated egress path from the far corner office to the stair door measures 120 feet. Sixteen measurement stations at 8-foot spacing along the path centerline, meter face flat on the floor, all normal lighting off, blinds closed, test performed at 8 pm with the floor unoccupied and temporary battery work lights staged at the stair door so anyone entering has a lit path while the test runs.

Before the test: ladder set on the level corridor floor at full extension of its rated working height with three points of contact maintained while resetting each head, per 29 CFR 1910.23 for ladders in general industry, or 1926 Subpart X on a construction site. Batteries in these heads are sealed lead-acid containing sulfuric acid; any head found with a swollen or leaking cell gets eye and face protection, no attempt to charge it, and disposal per the SDS, and terminals are worked with insulated tools with metal jewelry removed so a wrench cannot bridge them.

Corrections and qualifiers, printed.

Time basis applied separately. The initial row is compared against 1 fc average and 0.1 fc point minimum. The end-of-duration row is compared against 0.6 fc average and 0.06 fc point minimum. Nothing from one row is compared against the other row's criterion.

Meter error basis and character applied. The accuracy spec is percent of reading, so at 0.05 fc the plus or minus 3 percent band is 0.0485 to 0.0515 fc, entirely below the 0.06 fc floor. The spectral mismatch term is a fixed multiplicative calibration factor for the source type, not an additive offset, so it divides out of the max-to-min ratio taken with one meter on one lamp type but does NOT divide out of the absolute comparison against 0.06. The point fails on both treatments, which is what lets it be called a fail rather than an inconclusive.

Statistic applied as stated. The end-of-duration average of 0.71 fc passes the 0.6 fc average. The result is still a fail, because 0.05 fc at station 11 is below the 0.06 fc single-point minimum, and the criterion is an average AND a minimum, not an average or a minimum. This is the one line most field reports get wrong: they report the average, it passes, and station 11 never appears.

Uniformity: 52 to 1 exceeds the stated limit, and the edition question does not soften that. 21.3 to 1 initially passes. 52 to 1 at end of duration exceeds the 40 to 1 limit quoted above and is carried as a second fail pending the AHJ's answer on whether the limit is evaluated initially, at end of duration or both, because the conservative read of an unresolved edition question is the failing one. Whether the 40 to 1 limit is evaluated initially, at end of duration, or both is set by the adopted edition, so the record states both numbers and the question goes to the AHJ rather than being resolved on the report.

Finding. Two of the sixteen stations sit outside the cone of any head after the corridor was reconfigured; station 11 is the worst. The oldest heads are 7 years old and were carrying the shortfall. The corrective work is a battery replacement across the run plus one added head aimed at the inside corner, then a re-test of the full 1.5 hours after the new batteries have had their full recharge period.

Bounded out-of-service state. From the moment the test started until the batteries finished recharging, the floor's rated duration was not available. Start time, the temporary lighting staged at the stair, and the restoration time all go on the record with names, because that is a managed condition with a start and an end rather than an event.

Sibling-rule check. Nothing in this test blocked or narrowed the path, so the egress-path card's obstruction rule holds throughout; the ladder was worked from a side bay with the corridor clear. No exit door hardware was altered, so the locked-door card's one-motion gate is untouched. The recharge window is written as a dated state with a named owner and a named end, which the fire protection records card requires. Every level reported here carries its plane, its statistic and its time basis in the same row.

References

• 29 CFR 1910.37(b), exit route lighting and exit sign marking in general industry, including sign lettering height and stroke width; 29 CFR 1926.34 for means of egress and exit marking on construction work
• NFPA 101 Life Safety Code, in the edition adopted and amended by your authority having jurisdiction, for emergency illumination levels, the 1.5-hour duration, and the monthly and annual test requirements
• 29 CFR 1910.23 for ladders in general industry; 29 CFR 1926 Subpart X for ladders in construction
• Manufacturer documentation for battery type, recharge time, and listed replacement cells
• See related: What a Fire Protection Record Has to Show a Year Later; What a Means of Egress Cannot Pass Through