Nichia 519A CRI, UV LED Specs, and Antique Chandeliers: A 72-Hour Buyer's Notes

The 20-second version

If your deadline is inside 72 hours, three specs decide whether the order works: R9 (not Ra), peak wavelength (not wattage), and driver current (not claimed lumens). Everything else you can fix on site. These three you cannot.

I've handled 200-plus rush orders over nine years at a lighting components distributor, including same-day turnarounds for event production and fixture-refurb shops. The pattern is almost always the same: somebody ordered on lumen count, the order arrived, and the light looked wrong under a camera or a client's eye.

Below is what I actually check now, and where it has burned me before.

Why CRI alone won't save you

Here's something vendors won't tell you: a "Ra 90+" label can sit on top of an R9 of 40. Ra is an average across eight pastel test swatches. None of them are red. So a fixture can score beautifully on paper and still make a chandelier full of candle bulbs look slightly dead — because the deep reds are the thing that's missing.

This is where the Nichia 519A comes up in almost every conversation I have with refurb shops. In the 519A line, the R9080 variants are specified at both Ra and R9 levels, with CCTs running from warm to neutral. That R9 number is the one I ask for in writing, usually as a screenshot of the test report rather than a copy line.

Concrete example. In November 2024, a client called at 4:30 p.m. on a Thursday needing twelve tree chandelier fixtures relamped for a Saturday evening reception. Normal turnaround for that volume is seven to ten days. We found a local refurb partner with 519A stock in warm CCTs, paid roughly $600 in overtime labor on top of the $4,800 base cost, and had them hanging by Friday afternoon. The client's alternative was a rental set in 4000K that would have read cold on camera.

The number that mattered was not the CCT. It was the R9 on the datasheet. The 4000K rental would have been fine on paper too — same nominal Ra, different red rendering.

To be fair, R9 matters far less for a parking garage or a warehouse aisle. If nobody's photographing it and nothing needs to match a fabric sample, standard bins are cheaper and honestly fine. I don't push R9080 on those jobs.

UV LED: stop reading watts

Watts on a UV LED listing tell you how much electricity goes in. They tell you approximately nothing about how much UV comes out at the wavelength you need.

What I ask for instead, in this order:

• Peak wavelength — 365 nm and 395 nm behave differently against different photoinitiators and coatings. Match the chemistry, not the marketing.
• Radiant flux at that wavelength, typically in mW — not the electrical input.
• Irradiance at working distance — because a head that hits 2000 mW/cm² at 5 mm may hit 80 mW/cm² at 40 mm.

That last one is the whole LED mask vs panel question, by the way. It isn't a quality contest. It's geometry.

A mask-style array puts emitters close to a curved or small workpiece. High irradiance, small footprint, fast cure, and it's unforgiving if you move. A panel array stands further back to spread output evenly across a flat surface. You trade peak intensity for uniformity. If your workpiece is a contoured shell, a panel will undercure the edges no matter how much power you throw at it, and a mask will scorch the center if you crowd it.

I've tested both on the same 48-hour production run and gotten two different failure modes. That's not a vendor problem. That's picking the wrong tool for the shape.

One more thing, and I'll be blunt about it: UV-A is not a visual-safety footnote. Ask for the photobiological risk group classification and the eye and skin protection documentation with the quote, not after the install. I've watched a shop learn that lesson the expensive way in 2023, and we now write protective equipment into the scope of work by default.

Antique fixtures are their own category

The 'modern LED will slot right in' thinking comes from an era when most fixtures were simple E26 sockets behind a shade. A chandelier antique restoration rarely is. Rewiring, socket types, and heat paths in a brass 1920s frame predate every assumption the LED spec sheet makes.

Two things I no longer skip:

Thermal path first. A small-die emitter in a sealed glass cup with no convection is a derating problem, and derating changes color over the first hundred hours. If the fixture has no airflow, I spec lower drive current on purpose and accept less output.

Driver current over lumens. Two parts with identical lumens at different currents will look different in five years. Ask which current the lumen figure was measured at, and whether the binning was done at that current or extrapolated.

Granted, this is more paperwork than most buyers want on a 48-hour order. And sometimes the answer is a plain generic part — I've used plenty, and there's nothing wrong with them for plenty of jobs. The difference shows up when a client photographs the finished room and the red velvet reads brown.

On substantiation: per FTC advertising guidance (ftc.gov), claims like a specific CRI, wavelength, or output figure have to be truthful and backed by evidence. That's the standard I hold quotes to. If a supplier won't put the R9 value or the peak wavelength in writing, that silence is itself information. Verify current requirements and figures at the source before you commit, since specs and bin availability change.

There's something satisfying about a rush order that lands clean. Not because the scramble was fun — it wasn't — but because the twenty minutes of asking the right questions on Thursday morning is what made Friday boring. Boring is the goal.

Where this doesn't apply: decorative-only installs with no photography, no calibration, and no matching requirement. For those, buy on price and CCT, ship it, and move on. I do.