Why a Retrofit Lamp in an Old Fixture Is a Different Fixture

Why this matters

The pitch is that you keep the housing and change the light source, so the only thing that changes is the energy bill. What actually changes is the shape of the source, its position relative to the optic, the heat path out of the housing, and whether anybody's listing still covers the thing on the ceiling. The assembly you leave behind has no photometric report, because no laboratory ever tested it, and every number in the submittal belongs to something else: the bare lamp's flux to a lamp on a test rig, the fixture's efficiency to the source that used to sit in it. Multiplying one by the other feels like arithmetic and is not.

The gate

Does a photometric report and a listing exist for the assembly exactly as it will leave your ladder?

That is one question with two halves, and both have to answer yes. The report tells you what the assembly will do. The listing tells you whether it is code-legal in that housing. Two jobs, same swap on paper, opposite answers.

What the swap changes, and who owns each part

Source geometry. A luminaire's distribution is a property of source plus optic, not of the source. The old reflector was cut for the emitting body it shipped with, and the new source has a different size, shape and position.

Original: compact arc tube near the reflector focus
     \                     /
      \       (o)         /     (o) arc tube
       \_________________/       at focus
              | |
        narrow, controlled beam

Retrofit: flat emitter plate across the neck
     \                     /
      \    [=========]    /     flat plate,
       \_________________/       below focus
              | |
        wide spill, reflector barely used

The thermal path. The old source radiated most of its waste heat; the new one conducts it into the housing. A housing that was a reflector is now also a heat sink, and it was not designed as one.

The listing. A replacement lamp that runs on the existing ballast and alters nothing else is a listed lamp in a listed luminaire, and the luminaire's listing is undisturbed. The moment the work rewires the luminaire, bypasses or removes the ballast, or puts a driver inside the housing, the luminaire has been modified, and the path that keeps a listing on the modified assembly is a field-installed conversion kit listed for that purpose under UL 1598C and installed per its instructions. That binds through the listing mark, through NFPA 70 (National Electrical Code) 110.3(B) in the edition your jurisdiction has adopted, and through 29 CFR 1910.303(b)(2) in general industry or 29 CFR 1926.403(b)(2) in construction, each requiring listed equipment to be installed and used per its listing instructions.

The photometric report. The old report is a measurement of the old assembly. The photometric file card owns what such a file contains; the point here is that nobody has one for what you are building unless a kit manufacturer paid for it.

Outcome one: a listed conversion kit with its own report

An office building, 180 recessed troffers. The kit is listed under UL 1598C for that housing family and ships with an absolute photometric report of the retrofitted assembly, measured as a whole with its own driver, plus a label the installer applies inside the housing.

Because the report is of the finished assembly, its output figure already contains the driver, so the driver factor in the light loss factor product is 1.00 and taking it again would double-count. The light loss factor card owns that product and the re-basing between initial and maintained figures.

Existing design basis: 50 fc maintained, horizontal at 30 in above finished floor, at a light loss factor of 0.80, which re-bases to 50 / 0.80 = 62.5 fc initial. The kit's report puts the retrofitted assembly at 0.86 of the original assembly's delivered flux in the same mounting condition, and the layout does not change, so predicted initial is 62.5 x 0.86 = 53.8 fc. At the new light loss factor of 0.84, which is the same maintenance interval with better lumen maintenance and a driver factor of 1.00, maintained becomes 53.8 x 0.84 = 45.2 fc, against a 50 fc requirement. The 0.84 light loss factor and the 0.86 flux ratio sit close enough together to get swapped on a spreadsheet, and they are unrelated quantities; neither stands in for the other. That is a 9.6 percent shortfall, it is known before anybody buys anything, and it is fixable by specifying the next lumen package up. Gate open means the answer arrives on paper.

Outcome two: a lamp into a 1990s high bay reflector

A distribution warehouse, 48 high bay reflectors, no kit exists for the housing, and the proposal is a screw-in replacement source. Nothing in the housing is rewired, so the luminaire's listing survives, but no report exists for the pair.

The arithmetic that gets offered. Old lamp bare flux 36,000 lm, luminaire efficiency from the original photometric report 0.72, delivered 25,920 lm. New source bare flux 18,000 lm. So, the pitch goes, 18,000 x 0.72 = 12,960 lm delivered, exactly half, and the space is over-lit anyway.

Why that multiplication is not allowed. The 0.72 is not a property of the housing. It is the fraction of a specific source's flux that a specific reflector geometry got out of the aperture, and the retrofit changes the geometry the number was derived under. Carrying it across is the same error as carrying a coefficient to a shape it was never measured on, and it can run either direction: a flat downward-facing array can beat 0.72 because less flux takes a bounce off an aged reflector, or it can fall well under it because the plate shades its own optic.

So the only honest instrument is a mock-up. One bay converted, everything else held: same grid, same plane and height, same control setting, same time of day with the skylights covered, meter unmoved between the two reads. The repeatable-reading card owns that procedure and it is what makes the two numbers comparable at all.

The mock-up that decided it

One bay, nine-point grid on 8 ft centers, horizontal at 30 in above finished floor, both reads initial-basis on the same fixtures with the lenses wiped before the first read so no dirt term moves between them.

The average outperformed the naive prediction. Bare flux fell to 18,000 / 36,000 = 0.500, and the delivered average fell only to 0.640, so the retrofitted assembly is getting 0.640 / 0.500 = 1.28 times as much of its source's flux onto the plane as the original did. The reflector was 20 years old and a downward-facing array skips it. The efficiency went up, not down, and the offered multiplication would have under-predicted the average by 12,960 against a measured equivalent of 25,920 x 0.640 = 16,589 lm of delivered flux, roughly 22 percent light.

And the space still failed. Average-to-minimum went from 27.5 / 18.0 = 1.53 to 17.6 / 9.0 = 1.96, because the flux that survived is concentrated under the units and the mid-span valleys took the whole loss. The uniformity card owns what those ratios catch and why the mid-span is what the pickers experience. The space could not clear its level requirement either, and the re-basing runs toward the requirement rather than toward the reading: the racking task is specified at 20 fc maintained on the same plane, which at the installation's light loss factor of 0.75 from the maintenance file, a different quantity from the 0.72 luminaire efficiency above and not to be confused with it, needs 20 / 0.75 = 26.7 fc initial. The measured 17.6 fc is 66 percent of that. The factor moves the requirement because the reading was taken on a new source behind a freshly wiped lens and is already an initial-basis figure; multiplying the reading by 0.75 as well would take the same loss twice.

Two safety items the mock-up itself created. Reaching the reflector is work at the mounting height from a lift rated for it with the fall protection the equipment requires under 29 CFR 1910 Subpart D in general industry or 29 CFR 1926 Subpart M in construction, and the fixture is de-energized, locked and tagged, and proved dead live-dead-live per NFPA 70E-2021, 120.5, in the edition the employer's electrical safety program adopts, under 29 CFR 1910.333(b)(2), before a hand goes into the socket. Removing a discharge lamp is a mercury exposure if it breaks: nobody sweeps or shop-vacuums the debris, the area is ventilated and cleared of people while the vapor disperses, and the fragments go into a sealed container, the control for that route being placement and ventilation rather than a glove, with mercury listed at 29 CFR 1910.1000 Table Z-2. Intact spent lamps are universal waste under 40 CFR 273 as adopted by your state, and a magnetic ballast predating 1979 may contain PCBs, so a stained or leaking one is handled under 40 CFR 761 rather than dropped in the skip.

What was done. The lamp swap was dropped and the bays were re-optic'd with complete luminaires carrying their own reports. The mock-up cost one bay and two hours of lift time; the alternative was finding the same 1.96 ratio across 48 bays after the invoice.

What flips this

A source whose emitting geometry genuinely resembles what the optic was cut for changes the picture, and the clearest case is a linear replacement in a linear housing whose optic was cut for a tube of the same diameter and length. There the naive multiplication is close, though still not a report. The other flip is a housing whose optic never did much work, such as an open strip with a white-painted channel: little flux was under optical control to lose, so the delivered ratio tracks the bare-lamp ratio.

Checking your own figures

• Every ratio recomputes from the table. 17.6 / 27.5 = 0.640; 9.0 / 18.0 = 0.500; 31.0 / 44.0 = 0.7045, reported as 0.705.
• The efficiency claim is derived, not asserted. 0.640 / 0.500 = 1.28, and the direction is up, which is what "the reflector was skipped" predicts.
• The under-prediction is stated against a like quantity. Offered 12,960 lm against a measured-equivalent 25,920 x 0.640 = 16,589 lm, and 12,960 / 16,589 = 0.781, so 22 percent light, both figures being delivered flux from the same assembly.
• The light loss factor is applied once, to one side. All six mock-up readings are initial-basis, on a new source behind wiped lenses. The requirement moves onto their basis: 20 / 0.75 = 26.7 fc initial, and 17.6 / 26.7 = 0.66, reported as 66 percent. The readings are never multiplied by 0.75.
• Outcome one's arithmetic closes too. 50 / 0.80 = 62.5 fc initial; 62.5 x 0.86 = 53.75, reported as 53.8; 53.8 x 0.84 = 45.2 fc maintained; (50 - 45.2) / 50 = 9.6 percent short.
• The driver factor is taken once. Outcome one's report is absolute for the finished assembly, so the driver factor is 1.00 in its light loss factor and appears nowhere else.
• Four things the swap changes, four named. Source geometry, thermal path, listing, photometric report, and each is picked up again later in the card.

References

• UL 1598C, listed field-installed conversion kits, and UL 1598 for the luminaire itself, both binding through the listing mark; NFPA 70 (National Electrical Code) 110.3(B), in the edition your authority having jurisdiction has adopted
• 29 CFR 1910.303(b)(2) and 29 CFR 1926.403(b)(2), listed equipment installed per its listing instructions; 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5; 29 CFR 1910 Subpart D and 29 CFR 1926 Subpart M for work at height
• 29 CFR 1910.1000 Table Z-2 for mercury; 40 CFR 273 universal waste and 40 CFR 761 for PCB-containing ballasts, each as adopted and administered by your state
• See related: What a Photometric File Actually Describes; How to Take a Light Level Reading Somebody Else Can Repeat; Why Uniformity Matters More Than Average Level; What a Light Loss Factor Is Accounting For