The Cree three-way LED light bulb, under not particularly rigorous testing, displayed a couple of odd behaviors.
A while back, I wrote a couple of pieces on several available LED three-way bulbs and their deficiencies. (See Let the 3-Way LED Lamp Race Begin and Why the Feit/Utilitech 3-Way LED Bulb Is Like a $9 Bill.) The conclusion: They are much like the three-way incandescent bulbs of the last 50 years, in that they are in effect two-way bulbs. There is almost no visible difference in output between the medium and high settings, and low is not really all that low.
A couple of months back, Cree announced its three-way entry, initially being sold only online. Now that these bulbs are available at retail, I picked one up at Home Depot for testing. It is a 30/60/100W equivalent, A21 size, labeled as 320/820/1,620 lumens at 2,700K. The price was $21.97.
I installed the bulb in a three-way socket in a floor lamp with a large shade. This lamp has worked fine over the years with three-way CFLs and three-way incandescents. I switched the Cree LED bulb through its three settings a couple of times -- after which the bulb went dead on all three levels. It was only the second or third time in my life a bulb has shown up DOA immediately after purchase.
Sample of one: one failure
My initial theory: In days of old, switching power supplies almost always had an NTC thermistor on input to reduce the surge current upon turn-on to a big input electrolytic capacitor. Having moderately high resistance, the thermistor limited input surge current but then heated up in a second or two and dropped to low resistance, allowing normal steady-state operating current. If one turned on the power supply but then turned it off and immediately back on, the thermistor may not have cooled sufficiently and could still have low resistance, allowing a potentially destructive capacitor-charging inrush current.
Every power supply engineer knew not to flip a switch quickly on/off/on/off/on, or bad things might happen. Did something like that happen here? Maybe. Does Cree employ input circuitry that is not as robust as it needs to be? Maybe. Maybe not. But nobody is paying me for a detailed analysis; I just wanted my Home Depot money back.
I returned the dead bulb and got another. No flicking on/off/on this time -- I just turned it to high and left it there.
Now here's an interesting observation: Weeks ago, in responding to a comment from Ron Amok, I said all or most of the LED bulbs I have observed drop a little in lumens after warmup. This is because virtually all lamps use a constant-current driver, which, when driving LEDs (with Vf dropping after warmup), must produce some drop in lamp wattage. That drop is about 5%. In fact, there are trade articles here and there stating that the drop is in the 4-6% range. This is consistent with what you will see on any LED data sheet for the Vf-versus-temperature curve from 25°C to maybe 80-90°C
What did I see with the Cree three-way bulb? About an 18% drop in lumens after 45 minutes and temperature equilibrium. Something else is going on here; the symptoms suggest LED driver efficiency dropping very significantly as it heats up.
The cause? Who knows, but it should be readily determinable if someone (not me) wants to spend a few hours testing the driver to identify what is shifting far more than expected.
The temperature story
After equilibrium, the exterior finned heatsink temperature is about 85°C, and that means the LED package inside is likely hotter than 92-95°C, and the LED junction is somewhere north of 100°C. This also means that the LED driver is operating inside a chamber with an ambient temperature of at least 85°C -- very different from the Philips SlimStyle, where there is appreciable thermal isolation between very hot LEDs and the driver components.
Who knows how important all these individual things are? But an engineer might say there is a way to do it, and maybe a way to do it "better." Is it OK to have the driver always in an 85°C ambient for 25,000 hours, or is it better if it's 50-60°C?
The package says "suitable for operation up to 45°C," under which conditions the power supply would be operating in an ambient of at least 105°C, and the LED junction would likely be at about 125°C. What might we expect for life expectancy at those temperatures? Harriet Homeowner probably assumes the same 25,000 hours, but that's not likely.
I will say that the lamp has clear-cut low, medium, and high settings -- it's probably the only three-way bulb of any kind that has ever accomplished this. However, only after a full warmup of up to an hour (with the very substantial drop in lumen output at the high setting) do the lamp's output readings closely match the differentiation percentages between low, medium, and high.
In all fairness, then, if we measured the light output of each setting only after an hour, we would see close correlation at that setting with the stated lumen values. In this bulb, unlike in incandescent or other three-way bulbs, the final low, medium, and high settings have a very long time constant, because of this pronounced lumen dropoff over time. This may be important only to a nitpicker who carries a light meter everywhere.