5mm & flux LED fade tests with light meter

[JohnR66]

Now that I have a light meter, I retested some of the better LEDs and threw in an "ebay" 5mm LED for comparison. LEDs were driven at 30ma for 192 hours and tested on the light meter every 48 hours at 20ma. The light meter is a highly linear scale that outputs to a frequency so I used that in the Y axis until I can calibrate to a real light value. This however works fine for this test. The GS reads the highest due to its brightness and tight focus. Many LEDs shows a slight bump after a few hours of operation. Others have reported this phenomenon. Perhaps it is a change in Vf as the LED burns in.

Nichia GS. A highly stable, bright LED.:thumbsup::thumbsup:
RatShack LED: Wide beam "neutral warm" tint. Highly stable LED.:thumbsup::thumbsup:
Sylvania 3mm white: cool white. shows some droop
JELEDHK:Typical ebay 5mm LED severe fading. Junk :thumbsdow
Cree 24cd whiteretty stable and bright with tight focus :thumbsup:
JELEDHK cool white flux: disappointing fade for a flux LED.:thumbsdow
Light of Victory warm white flux: good brightness and stability for WW:thumbsup:

 

[JohnR66]

Updates: At 480 hours (graphs updated) The Nichia GS and Radio Shack 276-0017 show little change. Unfortunately, the Cree 24cd LED has started to fade, but is still above the 70% cut off point as is the LOV WW flux. I doubt these two will still be in contention at the 1000 hour mark. The lousy JELED 55,000 mcd LED is now at ~10% of its initial output although it has pulled out of that nose dive into darkness.

I don't expect much drop in output for the GS and RS LEDs when I check them at 1000 hours given their performance. I have one of the RS LEDs in continuous use for 1 year (8769h) at 10ma in a garage door button. It has not faded in the slightest. (I took the button apart and checked the LED). The Superbrightleds 18cd 15 Deg LED from 2006 that I had in the button initially had faded and turned purplish after only a couple months.

 

[PCC]

The problem with Nichia GS LEDs is that they're difficult to find. I'd use them if I could readily get them but I haven't really found a source for them so I'm using those cheap eBay LEds for now.

 

[JohnR66]

You can get Cree LEDs from several sellers like Digikey, They are not the brightest bin, but at least last and out throw the ebay LEDs. Given that the ebay LED faded to ~60% after 96 hours :shakehead I'd say why bother with them?

I am starting another test. I want to see how the ebay LEDs last at 12.5ma. It would seem that life would be extend exponentially at lower currents, but I see evidence that it is more linear. For example I found one of these LEDs severely faded after a month a 15ma. Nichia rates their new flux LED at 15,000 hrs at 50ma and 40,000hrs at 20ma. So it seems to be more linear below overdrive currents.

Anyone out there can spare a couple Cree C503 series 5mm LEDs for testing?

 

[JohnR66]

Test update: It has been 1056 hours now for the little 5mm LEDs cooking at 30ma. The others have dropped out of the test. The Nichia GS and Radio Shack 276-0017 shine on. The RS and GS are rock stable lighting class LEDs. From the horizontal lines, I can not estimate how long they will last, but it should be for several thousands of hours. (The RS LED is less intense because of its wide angle beam) I will keep the test running.

Unfortunately, the Cree 24cd has dropped below 70% initial light output at around 800 hrs so is out of the test. In fairness to Cree, this is an older LED sent to me from a CPF member about a year ago. Cree now has the C503 series. I recall reading a blurb from Cree that states that these new LEDs are designed to last longer. We shall see because I bought some and have started a new test with one. The good news is that they are brighter than the older tested one right out of the gate. Cree now uses a rectangular die like that of the nichia GS.

 

[jtr1962]

Excellent results for 30 mA drive current. These results make me feel Nichias are worth the premium if you can find them. Regarding the Cree, I think it would do much better at 20 mA. However, perhaps the newer version you have will do better.

 

[JohnR66]

Believe it or not, some Crees I have from 2005-06 are no better than the ebay LEDs in life time. Cree has been improving greatly here.

I still wonder who makes the RS LEDs. The lead frame shape and die cup stamping patterns indicate these are not Cree or Nichia. A CPF member found these at Mouser: http://www.mouser.com/ProductDetail/...O9HPTls82nE%3d
but not sure on the tint. I'm too tapped out right now to buy them. I'd be grateful to any CPF member to buy some and compare them to the RS 276-0017 LED.

 

[Fallingwater]

Impressive. I'm surprised that the GS - which, when it came out, was widely regarded as a fading risk - is performing so well.

 

[JohnR66]

Impressive. I'm surprised that the GS - which, when it came out, was widely regarded as a fading risk - is performing so well.

We must have read the same thread because I heard this too, so wanted to see for myself. Whoever said this mentioned that they heard it right from a Nichia sales person that the GS will not last like the DS. After dealing with sales people for years, I know that what comes out of their mouths is not always correct or just plain bull even if they work for the same company.

To me, the GS is likely to use the same phosphors and epoxy encapsulation as the DS. The GS just uses a rectangular die, so it should last as long or longer than the DS. One thing's for sure, according to my test, it isn't fading anytime soon!

 

[Cemoi]

JohnR66: :thanks: a lot for your tests.

Have you carried out similar tests on colour 5mm LEDs (I'm interested in red ones)?

 

[JohnR66]

I haven't tested the color LEDs yet. I expect the Cree LEDs I bought to perform well. Since they are only 21 cents each in a 100 pack for red and amber, I'm not spending the time trying the ebay stuff any longer.

 

[Cemoi]

I expect the Cree LEDs I bought to perform well.

Are you talking about these?
Would you be interested in testing these from BestHongKong, advertised as 25cd? I bought some following a recommendation in a thread in the Bicycle section, and find them very bright. I can send you a pair if you like.

 

[JohnR66]

Are you talking about these?
Would you be interested in testing these from BestHongKong, advertised as 25cd? I bought some following a recommendation in a thread in the Bicycle section, and find them very bright. I can send you a pair if you like.

Yes, those are the ones I was sent. C503B-RCN-CW0Z0AA1 was the number on the red 30 Deg ones.

Yes, I'd like to test them. They are more expensive, so they might be interesting. I'll send you a private message.

 

[JohnR66]

Relative light output and relative intensity of the Nichia GS, Cree and RS white LEDs. Note: (the Cree LED is from a new batch of brighter LEDs.)

Nichia NSPW500GS-K1 Lot 96CS14-c0W
Cree C503C-WAS-CBADA151
Radio Shack 276-0017 Neutral white 30 Deg beam

Relative light output------------------- Relative intensity

1=GS 2=RS 3=Cree

Notes:The Cree part number represents a wide range of brightness and tints of the cool white LED, although the ones I received have been pretty consistant. Relative light output was measured by filing down the LED to 1mm above the die cup and polishing the surface to make it a flat top. This allows a close approximation of the light output to be measured relative to one another since the light distribution has been normalized.

 

[balou]

The problem with Nichia GS LEDs is that they're difficult to find. I'd use them if I could readily get them but I haven't really found a source for them so I'm using those cheap eBay LEds for now.

www.Led1.de - 9.90 euro for the LEDs, 3.90 euro shipping worldwide, free shipping within EU. That's under $21 dollars shipped. They accept Paypal and credit cards.

Impressive performance of the GS LEDs... I will mod a Fauxton with it, will hand select my 10 pack for the one with the lowest forward voltage, and use it with just one CR2032. Wonder if it will be brighter than a normal Fauxton...

 

[Fallingwater]

Impressive performance of the GS LEDs... I will mod a Fauxton with it, will hand select my 10 pack for the one with the lowest forward voltage, and use it with just one CR2032. Wonder if it will be brighter than a normal Fauxton...

Be aware that you're very likely to have flickering problems. I've done what you suggest already, and so have others, and this always happens. It's a problem of resistance in the contact between the CR2032 and LED leg.

Also: no, it's not brighter than a stock fauxton at full power (fresh cells). It is, however, noticeably brighter than most fauxtons out there, that are typically run for most of their life on depleted cells - and, importantly, will retain the same brightness for a long time.

The only problem is the flickering. I tried to come up with a solution, but the only one I can think of is to redesign the way the cell is held in place, and there's no space in a fauxton to make up any sort of suitable contraption. It might be done with those fancier fauxtons with a click-on-click-off switch that DX sell.

 

[JohnR66]

More test results...

This graph shows how the brightness scales with current using the Nichia GS vs the Cree LED. The GS scales almost linear while the Cree does not scale quite as well. At 10-25ma drive currents, the Cree makes a fine LED given its price, availability and beam quality, however, for overdrive the GS scales well with current. It remains the 5mm LED champ except for beam quality

The second part of the test, I let the LEDs Idle at 60ma for 5 min. with a few seconds at 80ma. LEDs can suffer sudden lumen depriciation from overdrive. After letting them cool off, I rechecked their outputs at 20 ma and I'm happy to report no lumen depreciation was suffered with either LED.

Of the GS LEDs, I have warm and cool bins of the cool white LEDs. The warmer bin is the top brightness bin (W) and warmest tint (c0) The cooler GS is a W bin, but not sure of tint bin. Here is how the two LEDs look. They both have a cooler tint to the center relative to the fringe. The cooler tint meters slightly brighter although this is just one sample of each tint. By eye, they look the same in brightness.

 

[balou]

Also: no, it's not brighter than a stock fauxton at full power (fresh cells). It is, however, noticeably brighter than most fauxtons out there, that are typically run for most of their life on depleted cells - and, importantly, will retain the same brightness for a long time.

Ok, that means a DX Fauxton is brighter for the first 5-10 minutes. I can live with that

The only problem is the flickering. I tried to come up with a solution, but the only one I can think of is to redesign the way the cell is held in place, and there's no space in a fauxton to make up any sort of suitable contraption. It might be done with those fancier fauxtons with a click-on-click-off switch that DX sell.

Yes, I'm well aware of that problem. All my single CR2032 mods were unsuccessful so far. Some work more or less, and some are material for the trash (plastic body at least... always found a use for the other parts)
I've ordered those with a hopefully more reliable construction: http://www.dealextreme.com/details.dx/sku.16510
Exactly the same ones as the ones from Lighthound. Interestingly, they're more expensive than on Lighthound.

 

[JohnR66]

New Cree 5mm Whites are looking good.
This graph shows the relative intensity of the old cree vs the new one. Not only is the new Cree more intense, but it has a wider beam so it is making much more light. After 264 hours, the new Cree is not fading but is slightly on the increase in output where the old cree (from early 2009) is starting to fade. A few months ago Cree updated their 5mm LEDs with better performing dies to almost silent fanfare. Probably because of the interest on the new XP-G power LED. While the XP-G is amazing, their new white and color 5mm LEDs are impressive.

I will continue testing the new Cree for as long as it remains above 70% of initial output.

 

[jtr1962]

Great results! I think the new Cree LEDs are one of those hidden bargains. I look forward to seeing how long these take to hit 70%. Also, I assume this test is at 30 mA like the ones in your initial post? If so, that makes the results even more amazing!