New [to me] Luxeon behavior observed

Sometimes a bit of anal retentive behavior allows you to see things that you would otherwise miss. I have been maintaining traceability on Luxeon emitters that I purchase. I do a receipt inspection which consists of measuring the Vf at 90 and 400mA [which allows calculation of dynamic resistance] and measuring the relative brightness with a test fixture that I have built that fits on my photometer sensor. I recently mounted a 1W white emitter on a heatsink along with a NX-05 optic and holder. Wires were soldered to the emitter leads. The emitter was mounted with Arctic Alumina under the slug and JB Weld epoxy under the plastic portion of the emitter. I powered the assembly for testing purposes using a Li-ion cell and a resistor for current limiting. The assembly was powered several times for a minute or two to check out optic centering and general performance. It seemed a bit brighter than expected for the battery voltage, choice of ballast resistor, and the previously measured Vf. On a hunch, I remeasured the Vf at 90 and 400mA. They had shifted significantly! The receipt inspection values had been 3.30 and 3.69V at 90 and 400mA. The post mounting values were 3.13 and 3.52V. Interestingly, these two data sets give exactly the same dynamic resistance of 1.26 ohm. What caused the shift? I don't know. I do know that this is not a case of loss of traceability or inconsistancy in testing methodology. The only "stresses" that the emitter was exposed to are: heat of soldering the leads, several thermal cycles of being energized/deenergized, and pressure applied to the plastic outer portion of the emitter during mounting to spread the heatsink compound and epoxy. While I would really like to know if the relative output has also changed, the way the emitter is mounted precludes retesting it.
Has anyone else observed a similar shift in Vf in an emitter that has not been abused? One of the implications of this phenominon it that setting current by voltage control only is even more iffy than already widely recognized. Up until now, I have been willing to consider voltage only control in applications where the emitters are not heavily driven and ambient temperature ranges are not extremely large. Now I am less confident about that design approach.

Did you measure the temperature at both times, as close to the junction as possible? Seen similar results with other Luxeons, or with regular non-Luxeon LED's? Interesting.

[ QUOTE ]
paulr said:
Did you measure the temperature at both times, as close to the junction as possible

[/ QUOTE ]

No. Vf measurements made consistantly at 2 seconds after application of power. This produces consistant results since the thermal capacity of the slug seems to dominate the die temperature rise for the first three seconds if there is much thermal resistance between the slug and heatsink.

[ QUOTE ]
paulr said:
Seen similar results with other Luxeons, or with regular non-Luxeon LED's?

[/ QUOTE ]
No, but I haven't looked either. I am hoping that this thread might encourage others to look for this too.

There are a few things that I can see as a test engineer.

1. Your test setup is different than the voltage rating test equipment.
2. The LEDs you are checking have been through a thermal event in the soldering process.
3. By having a soldered connection instead of a mechanical connnection, a source of resistance is eliminated.

I admire your science and the detailed effort you put into your product. The primary data we give our customers is light output over time, since that appears to us as their primary interest. Thank you for providing a different and detailed view of other developers process.

This weekend we were at a show selling product and we tried out a new power supply without checking it's output first. Our attention was on sales, not science and after melting some product in the display case, we discovered that our new power supply was putting out 16V, not 12V as rated. After painful experience I can say definitively that overdriving Luxeons causes thermal breakdown of the surrounding material long before light output failure.

Dear Doug,

I had the same problem with white 5 mm InGaN LED. All started observing the behavior of a off-brand 6 LED flashlight, that was visibly dimming after less than 100 hrs of operation. It was powered from three AA cells for a total of 4.5 V, with a series resistor of 4.3 Ohm, drawing initially 180 mA, or 30 mA per LED. I started all over with a new one of these flashlight, and I measured the variation of the Vf over the time with the use of a constant current generator made with a FET. In about 200 hours, the Vf at 30 mA shifted from 2.9 to 3.4 Volt.
By further research, this behavior seems associated with bad doping process of type P gallium, due to poor control of the 3 diffusion flows.
I never observed this problem with 5 mm Nichia, and I wasn't aware that it may happen to Luxeon.
Have you measured a higher light output with higher Vf diodes? Otherwise, you may have stepped on lower efficiency Luxeons...

LOWER Vf = PURER SEMICONDUCTORS

Regards

Anthony

[ QUOTE ]
TheProphet said:

Have you measured a higher light output with higher Vf diodes? Anthony

[/ QUOTE ]

I once sorted a batch of 100 5mm Nichias from the same production lot by Vf. I then measured the relative intensity of the 6 LEDs with the lowest Vf and the 6 LEDs with the highest Vf. All of the 6 highest Vf LEDs were brighter than the 6 LEDs with the lowest Vf.

I note in your post that the off-brand LEDs that you tested had increasing Vf with use. The Luxeon I observed had the Vf drop instead.

Doug,

The heatslug of the luxeons are not electrically isolated. Could the mounting have involved metal-metal contact between the heatslug and heatsink, thus changing the electrical potential of the whole system?

Just a suggestion....

-leo

[ QUOTE ]
hotfoot said:
Doug,

The heatslug of the luxeons are not electrically isolated. Could the mounting have involved metal-metal contact between the heatslug and heatsink, thus changing the electrical potential of the whole system?

Just a suggestion....

-leo

[/ QUOTE ]
Leo, that is a good question. In this case I would say no. As mounted, the slug was not isolated from the heatsink but the heatsink was electrically isolated from both electrical terminals.

I gave a better read to my solid state physics books. Since you got an higher output with higher Vf, it is not matter of impurities in the doping process, but a variation of the quantity of doping agent during the delicate diffusion processes. To confirm this, make a capacitance measurement across the junction.
And you may want to upgrade your testing of Luxeons with a high current curve tracer.
From what I understand, you can use the Luxeon you have with no fear of sudden failure.

Anthony

Well I finally could not stand not knowing anymore! I disassembled the LED from the heatsink arrangement to be able to remeasure its luminous output. I want to know if whatever caused the shift in Vf characteristics also caused a shift in output. The answer is no. I did find that the Vf values at the 90 and 400mA test currents have decreased further. They now measure 3.05 and 3.37 volts respectively. The only stresses the LED experienced since last testing are a few dozen thermal cycles of operation well with design limits. My curiosity now further inflamed, I decided to test another LED from the same batch. I first retested Vf and luminosity values and found they matched the receipt inspection values. I then applied a wetted, 600F, temperature controlled tip, to each LED lead for 2 seconds each. Retested LED and found no change to Vf or luminosity. Repeated the above for 3 seconds each lead. Retested LED and found no changes. Remain baffled by subject phenomenon of this thread.

Doug,

I suspect that your Luxeon was subject to an uncommon case of P.W.E.P. (parallel world exchange program). This is typically not manifest in measurable form. I stand on this proposition until proven otherwise. /ubbthreads/images/graemlins/tongue.gif

- Don

PS. Fair winds off your bow.

Doug, I'd like you to know that you are making it much more difficult for us to sneak in and swap LEDs on you ... /ubbthreads/images/graemlins/rolleyes.gif

[ QUOTE ]
Tomas said:
Doug, I'd like you to know that you are making it much more difficult for us to sneak in and swap LEDs on you ... /ubbthreads/images/graemlins/rolleyes.gif

[/ QUOTE ]

Tomas, that may be true. On the otherhand, if I assume that you have been successful, that could account for all of my observations. /ubbthreads/images/graemlins/icon15.gif

it will probably be impossible to know for sure until you can reproduce the effect on a fairly regular basis. you could just have happened to test an odd LS. wierder stuff has happened.

Doug might just have found the equivalent of a luxeon perpetuum mobile - actually if he keeps on redoing the measurement the actual Vf at same brightness might keep going down increasing the efficiency until the LS starts to PRODUCE power instead of consuming it /ubbthreads/images/graemlins/tongue.gif

Klaus

Doug,

Are you actually measuring the current each measurement or calculating it based on the first-off supply voltage and assuming it is the same each test?

If calculating it, are you re-measuring the supply voltage each time you measure the LED Vf?

I'm curious about this, but don't want to go chasing red herings.

Chris

[ QUOTE ]
Burnt_Retinas said:
Doug,

Are you actually measuring the current each measurement or calculating it based on the first-off supply voltage and assuming it is the same each test?

If calculating it, are you re-measuring the supply voltage each time you measure the LED Vf?

I'm curious about this, but don't want to go chasing red herings.

Chris



[/ QUOTE ]

No to the above. There is no way constant voltage would yield sufficiently constant current [as I am sure that you know since you are asking this question]. The object of my testing is to measure Vf at a known constant current. Current is controlled by two different, fixed value, constant current limiters using LM317s configured as a constant current source. Actual current values are 91.9mA and 399mA. Proper function of current sources checked with ammeter.

Doug,

An interesting find indeed given you have a constant current supply. I am too new here to know how each person works. It was in my mind that perhaps it was battery powered and a resistor used to limit the current, yet merely measuring the Vf each time. It could have been something as simple as the batteries discharging. Obviously not the case. Now I know where you are at.

I have an unused 1W LED here due for a power-up soon. I'm now going to make sure I measure Vf new and after a burn-in period.

Chris

[ QUOTE ]
Burnt_Retinas said:
Doug,

I have an unused 1W LED here due for a power-up soon. I'm now going to make sure I measure Vf new and after a burn-in period.

Chris

[/ QUOTE ]

That's great. The more people we have looking at this the sooner we will know how common or rare this phenomenon is. While not likely, one possible outcome is that we may find that at least some high Vf luxeons can be manipulated in some way into becoming lower Vf ones.

One thing about your testing methodology: in addition of consistancy in test current values, it is important to have some consistancy in taking the Vf measurement quickly after powerup since Vf will drop due to the negative tempco of Vf. Measurements taken in the first 3 seconds are relatively insensitive to the degree of heatsinking. Therefore if you use a constant time after powerup that is less than 3 seconds it will be valid to compare bare emitter and post mounting measurements.

[ QUOTE ]
Doug S said:
One of the implications of this phenominon it that setting current by voltage control only is even more iffy than already widely recognized. Up until now, I have been willing to consider voltage only control in applications where the emitters are not heavily driven and ambient temperature ranges are not extremely large. Now I am less confident about that design approach.

[/ QUOTE ]

Current will vary quite a bit with temperature. This chart is from testing one of my outdoor fixtures. The voltage was regulated +-0.01 volts during testing. The green line is current in milliamps, the blue line is lux output. For this test, the fixture started at 0F and warmed to 84F. Note this is rather large heatsink, so the Luxeon never gets any warmer than the 84F (ambient ~76F).



The emitter is mounted with Artic Alumina. I did a similar test with regular epoxy mixed with some thermal paste and the current leveled off at 400ma! And the light output rose accordinly. The heatsink eventually reached about the same temperature (83F). I think the epoxy mix wasn't transfering the heat as well, so the Luxeon slug/die was getting much warmer. A hot Luxeon can really suck down some current. One reason I stick with Artic Silver/Alumina for Luxeons.