PowerSource

Several months ago I posted efficiency numbers indicating that 82% of the power to a HB LED is lost as heat, leaving only 18% converted to light. It turns out that these numbers are outdated. Thanks to Doug Leeper, a circuit design engineer with over 10 years experience working with LEDs in lighting, we have some real-world numbers on HB LED efficiencies: Expect power-to-light conversion of 32%, not 18%.

(These are production LEDs, not the lab-only ones that vendors put out press releases for their latest results that you can’t buy for another 2 years. His test numbers, tabulated below, were after temperatures stabilized for over 1 hour, and utilizing NIST traceable equipment.)

To quote Leeper:

“I ran some tests on one of CREE’s white LEDs, from their old Q5 brightness bin (107-114lm/W @ 350mA and die @ 20C): Note that there is a de-rating factor due to die temperature which I have not included. These LEDs were simply hooked up in series. On this older brightness bin they were still hitting on this older part, in reality, a 31% conversion efficiency @ 0.9894 Watts each, or only 69% of energy lost as heat -- not the 82% you mentioned.“

(Leeper included his data – it’s down at the bottom of this page.)

Sure enough, I checked with Mike King at Optek, the original source of the 82% number, and he agreed completely. “The 82% was old data for a 60 lm/W LED. As Doug’s tests show, today’s 100 lm/W LED converts 31% of the electrical power to optical power.” 

Leeper also included some gripes he has about published specs on production HB LEDs.

“One thing I find that’s a pain is when most LED manufactures state their lumens, they do it with a 0.2 second to 0.002 second light pulse. This doesn’t allow the LED to reach a steady state temperature, as it would in the real world. So, you need to test the available LEDs to find out the truth.

“For the consumer, most products state the pulse in lumens measured at the source, not the actual emitted lumens measured after the optics (mirror, lens, reflector, window), which is what the lumens actually are once everything has hit a steady state.”

Thanks, Doug, for setting the record straight.

Leeper listed some of the reasons LED efficiency has increased for production devices:

• Creating a rough surface on the LED (Helps extract light that is normally trapped due to TIR and re-absorbed
• Thinner die so the light has less to travel through (increases the chance of extraction)
• Mirror rear surface to reflect the light exiting towards the rear (reduces effect of rear absorption)
• Reducing ohmic resistance of contacts (reduces loss due to resistance)
• Techniques for current spreading (reduces losses due to non-uniform current)
• High index (1.54 or higher) encapsulents (reduces losses and reflections caused by index mis-match and alters TIR)
• Die bond and low thermal resistance slug/substrate to heatsink (reduces losses caused by heat)
• Larger die size to reduce overall current density, which increases efficiency (CREE and LumiLEDs die can pass 150 lm/W at low currents, OSRAM at 130 lm/W, though their new Golden Dragon plus is just starting to get close to the state of the industry, just when their competitors are beginning to take another jump)
• Drilling holes into the die to increase light extraction

Yet to be seen in a product:

• Reduction of polarization mismatches within the die
• Low index coating/material to match from dome to air (or variable index material for matching from die to air)

So LEDs have come a long way – and still have headroom for increasing efficiency even more.

To learn more about real-world design considerations for HB LED design, attend EDN’s free one-day “Designing with LEDs” Workshop in Chicago on October 6. Register now.

**Die Temp @ 5mA ~20C