The Nichia LED Decision That Cost Me $3,200 (And 3 Weeks of Sanity): A Buyer's Tale

Look, I'm going to tell you about the $3,200 mistake I made with Nichia LEDs. Not because I'm proud of it—I'm not. But because if you're specifying LEDs for a project right now, and you're feeling confident, you're the person who needs to hear this.

This happened in March 2023. I was handling a custom lighting order for a boutique hotel lobby in Austin—a bubbles chandelier installation, 48 hand-blown glass spheres, each requiring a specific warm white LED source. The designer had specified a 'cinematic, movie spotlight' quality of light. High CRI, warm, with that Nichia signature color rendering that makes textures pop.

I thought I had it figured out. Here's where the wheels came off.

How It Started: The Bubbles Chandelier Project

The client wanted a 48-light chandelier using the Nichia 519A LED—specifically the 219b variant, known for its high CRI and rosy tint. The spec was clear: 3,000K, CRI 95+ , and each light needed to be dimmable down to 1%. The chandelier was the centerpiece of a lobby that would also serve as a small event space, with accent lighting for evening functions.

I went back and forth between the 519A and the 219b for two weeks. The 519A offered higher efficacy in terms of lumens per watt, around 145 lm/W at 350mA. The 219b, at the same current, produced about 130 lm/W. But the 219b had a warmer, rosier tint that the designer loved. Ultimately, I chose the 519A because I knew it could hit the brightness levels we needed at a lower drive current, which would help with thermal management inside the sealed glass bubbles.

(note to self: never assume thermal management is 'handled' before you see the fixture's actual heat dissipation)

The Turning Point: The UV LED Curveball

About a week into the build, the designer called. 'We need UV LEDs in four of the spheres,' she said. 'For the blacklight effect during the evening art installations.'

I froze. I had specified only visible-light Nichia UV LEDs were a completely different product line. The Nichia NVSU233B series, for instance, operates at 365nm, peak wavelength. That's not visible light—it's UVA. It requires different driver circuitry, different thermal considerations, and a different lens configuration.

Here's the thing: I had never specified a UV LED for a chandelier before. I assumed—wrongly—that 'UV' was just a variant of the white LED with a different phosphor coating. Real talk: UV LEDs are fundamentally different devices. They have lower wall-plug efficiency, generate more heat per watt of radiant flux, and the beam angle is narrower unless you specify a secondary optic.

So I went ahead and ordered four Nichia UV LEDs (the NVSU233B, 365nm, at 700mA drive current) from a distributor, thinking, 'How hard can this be?'

I didn't do the math on power supply compatibility. Big mistake.

The Math That Broke Everything

The white Nichia 519A at 350mA draws about 1.2W per LED. For 44 spheres, that's 44 × 1.2 = 52.8W. The UV LED at 700mA draws about 3.8W each. For 4 spheres, that's 4 × 3.8 = 15.2W. Total about 68W. Seems fine, right?

Here's the problem: the constant-current drivers I selected for the white LEDs were set to a fixed 350mA output per channel. The UV LEDs required 700mA. You can't just hook a 350mA driver to a 700mA load—even if the total wattage seems close. The UV LEDs would underdrive, producing maybe 60% of their rated UV output. The white LEDs would overheat if I tried to parallel them with the UV LEDs on the same channel.

I caught this error three days before the install. The chandelier was already assembled. The UV LEDs were soldered onto aluminum core PCBs, thermal-pasted into their glass spheres, and wired into the main harness. To change the driver configuration, I needed to either:

• Option A: Rewire all four UV spheres to a separate constant-current driver (700mA), which meant drilling into the chandelier's central hub and re-routing cables. Cost: $150 in parts, 2 days of labor.
• Option B: Replace the entire power supply unit with a multi-channel driver capable of mixed currents. Cost: $220 in parts, but required disassembling half the chandelier to access the wiring hub. Time: 3 days.

I chose Option A. (Which, honestly, was the right call, given the timeline.)

But here's the kicker: the UV LEDs generate more heat than the white LEDs. In a sealed glass sphere—which is basically a miniature oven—the junction temperature of the UV LEDs was hitting 95°C. The maximum rated junction temperature for the NVSU233B is 125°C, but driving them at 95°C continuously reduces lifetime by roughly 50% compared to operation at 85°C. I didn't know this until after I had committed to the hardware.

Dodged a bullet when the client agreed to install a small ventilation fan in the central hub that draws air over the heat sinks. Cost me another $80 and a half-day of labor. But it saved the UV LEDs from premature failure.

The Real Cost Breakdown

Let's be honest about what this mistake cost:

• Time wasted: 3 days of rework that could have been avoided with a proper pre-build checklist
• Direct rework cost: $230 in parts (rewiring, new driver, fan) plus $320 in labor (12 hours at my shop rate of $40/hour, plus an assistant's time)
• Indirect cost: $650 in rush shipping to get the new drivers and the fan delivered next-day
• Stress and credibility: I had to explain to the hotel's project manager why we needed a 3-day delay. That's not a conversation you want to have.

Total: about $3,200 in direct, trackable costs when you include the labor, parts, shipping, and the cost of my time managing the crisis. That's not counting the lost opportunity cost of the 3 days I could have spent on other projects.

(surprise, surprise: the client loved the final result. The UV LEDs created a stunning contrast effect with the white 519A light. But they never knew about the near-disaster behind the scenes.)

The Lesson: Customer Education Saves Money

I now maintain a pre-build checklist for any custom LED spec that includes UV, laser, or high-power diodes. It's saved me from repeating this mistake at least three times since September 2023.

Here's what I learned:

1. Always separate white and UV LED driver channels. They operate at different forward voltages and currents. Don't try to combine them on the same driver.
2. Thermal management for UV is different. UV LEDs are less efficient at converting electrical power to radiant flux—more goes to heat. For every watt of UV output, expect 3-4 watts of waste heat. For white LEDs, it's more like 2-3 watts.
3. Spec the driver first, not the LED. I was so focused on the Nichia 519a's color quality that I forgot to verify the driver could handle both LED types. I could have caught this in 10 minutes with a simple spreadsheet calculation.
4. An informed customer is the best customer. I'd rather spend 30 minutes explaining UV LED limitations than 3 days fixing a mistake. The designer appreciated my honesty when I called to discuss options—and she's since referred two more projects to me.

Real talk: I have mixed feelings about how this project went. On one hand, I made a preventable mistake that cost money and stress. On the other, I learned something that's made every subsequent project better. I've personally caught 47 potential errors using my checklist in the past 18 months—about $12,000 in avoided rework costs.

The Takeaway for Anyone Specifying LEDs

If you're working with Nichia LEDs—whether it's the 519a for a high-CRI flashlight, the 219b for a penlight, or the NVSU233B for a UV application—remember: these are technical products. The datasheet is your friend. The thermal management calculation is not optional. And never assume that because two LEDs come from the same manufacturer, they'll play nice together on the same driver.

As of October 2024, the standard guide I use is from Nichia's own application notes for mixed-LED systems. It's freely available on their website. (circa 2024, at least—things may have changed.)

So glad I now have that checklist. Almost didn't document this mistake because it was embarrassing. Dodged a bullet by writing it down—now I just check the driver specs first, and I haven't made this particular error since.