The Pathetic Life Span of Typical UV LEDs!

LEDs are supposed to last for tens of thousands of hours, right? Not the typical T1 3/4 "bulb style" LED [see Figure 1]. Many LED manufacturers use the same resin to manufacture their UV LEDs, as they use for all their other longer wavelength LED offerings. But, this resin doesn't stand up under the ionizing barrage of UV radiation -- it quickly yellows and limits the UV output from the LED chip, until the thing is dimmer than a sunset in a sandstorm!

T1 3/4 Bulb Style UV LED
Figure 1

To prove this I set up an experiment using two LM317Ls each wired as a Constant Current Source. Each has a 66.5Ω resistor across VREF to produce around 18.8mA out of the Adj pin. Each Adj output is fed to two UV LEDs in series, each from a different manufacturer. One of these LED pairs is on continuously, and the other is wired to a momentary pushbutton. Periodically I press the button to compare the two legs [i.e. to compare the two LEDs running continuously, against the two that run only when the button is pressed [the "Reference LEDs"]. Figure 2 is an animated GIF of me pushing and releasing the button [there's a glow-in-the-dark star over the LEDs, thus providing a more subjective indication—since much of the UV radiation is not only beyond the human visibility range, but also outside the camera's sensor range]. Notice how much brighter the flashing LEDs are. Those are the Reference LEDs—the one's NOT powered continuously, and thus, the one's NOT destroying the only path to full illumination! The others, as I'm sure you surmised, are the one's running 24/7 [597 hours as of the GIF's 'filming'].

Even over the VERY short time this test has been running, the difference is striking! And they're getting worse by the day!

It used to be that you could find a MTBF rating in a UV LED datasheet [I remember, in one case, the claim was 2000 hours]. But, apparently no longer! At least I haven't been able to find anything that gives any sort of clue as to the expected life span!

Figure 2

It smacks of larceny! I mean, I'm sure the in house engineers are aware of this deficiency, so ignorance isn't really an excuse. Is it expected to be just de facto knowledge, that this type of UV LED has such a short life span? I didn't know that when I used them in a production design. It wasn't long before customers started complaining! That was an expensive redo!! Maybe I was an idiot to expect them to stand up in an applicaton where they run 24/7!

But, it is possible to purchase LEDs that have the kind of amazing life span typically expected from an LED. There are three types that I know of:

  • Bulb style made from Silicone resin.
  • LEDs in a metal can with a quartz glass lens.
  • SMD LEDs with no lens -- typically with a much wider radiation pattern -- more like 120° vs something tighter like 30° or even 15°. Essentially, it's the lens that makes the tighter beam possible, and the wide dispersion pattern is due to the lack of a focusing lens.
Figure 3

The first two types [the ones with a lens] tend to be expensive. My latest estimate from Marubeni, for their L405R-04 UV LED was $4.15 a piece at quantity 100 [their minimum purchase quantity]. Yes, you heard that right: Each individual LED is just over $4!! That's $415 US Dollars for 100!

The surface mounted types without a lens are, typically, much less expensive. For instance: Mouser Electronics offers the Vishay VLMU3100-GS08 PLCC2 UV LED for a mere 35.1¢ at quantities of 100, and you can actually buy just one for 69¢!

Figure 3 shows the actual setup. In this photo the Test Button is pressed, showing how much brighter the intermittently powered B2 & V2 LEDs are!

  • B1 & B2 are Bivar UV5TZ 405nm LEDs
  • V1 & V2 are VCC VAOL-5EUV0T4 405nm LEDs

Both exhibited diminished brightness in less than 600 hours, but the Bivar LEDs are degrading far faster!

UV LED Test Jig Schematic
Figure 4

Figure 4 is a circuit diagram of the UV LED test circuit. The LM317L is being used, here, in it's "Constant Current" arrangment. The 66.5Ω resistor, bewteen the Output pin and the Adj. pin programs a constant current of around 18.8ma.