LuxV Temperature Test

[modamag]

After reading this "What really causes (White) LED's to fail?" I finally tried to perform my own experiment to test the quality of Lumined's LuxV in Mag2D.

After 10 minutes the LED (not Tj) temperature stabilize to 122F (50C). After 30 minutes the hotest point on the Mag settled to 119F (48.3C). The coolest point was 100F (37.8C). At the end of the 30 minute run I turned off the light and held it in my hand, it was just extra warm, nothing unbearable. /ubbthreads/images/graemlins/grin.gif

The test was performed twice with very similar results. The reading error tolerance is +/- 3F (1.5C).

IMHO continuous run of LuxV @ 1000mA will be no problem. According to Lumileds' data sheet (DS40), the hottest that a LuxV can tolerate is 160F (70C), which corresponds to the maximum junction temperature of 275F (135C).

/ubbthreads/images/graemlins/happy14.gif for hotbeam, georges80, ElektroLumens, and ofcourse Lumineds for producing such high quality products.

=== Test Setup ===
Ambient room temp @ 75F (24C)
Pencil & Paper
Fluke 65 Infrared Thermometer
FatMag
- Copper Mag2D Host with cutdown reflector
- LuxV VY0? bin
- Hotlips
- Fatman driver @ 13.5KOhms for 1000mA output
- 2x EL 3AA-D battery adapter
- 1x AA dummy cell
- 5x AA Powerizer 2250mAh

 

[evan9162]

What point on the LED were you measuring?

 

[modamag]

I measure the LED temp at the edge of the emitter and the hotlips pedestal

 

[tvodrd]

The 5W is (4) 1W's on a common slug. I tortured a 1W once to right at 190 degF and it survived! (R2H, BB750, CR2) I doubt it saw 3/4A for very long (CR2) but I wasn't about to pick it up!

(Hey, it's Friday, have a laugh!)

Larry

 

[modamag]

Wow, I want one /ubbthreads/images/graemlins/bowdown.gif I'll hold it with my BBQ gripping tong /ubbthreads/images/graemlins/grin.gif

 

[Pinter]

ensig,

did you measure the Vf change over time? It can also be used to guess the temperature rise.

Here are my data. LuxV and LuxIII was mounted on a Pentium III CPU heatsink. There was no ventillation at all. Current was 1000mA.

Time Vf(LuxV) Vf(LuxIII)
0:00 6.58 3.69
0:02 6.53 3.66
0:04 6.47 3.65
0:07 6.45 3.65
0:09 6.43 3.65
0:12 6.42 3.65
0:14 6.41 3.64
0:17 6.40 3.64
0:19 6.40 3.63

[LuxV calculations]

Vf shifted by -0.18V in approx. 20 minutes.

Using formula from DS40 (-4mV/Celsius shift in Vf) it heated up by 45 Celsius. As Ta (room temp.) was 22 Celsius, starting Tj was 22[C] + 8[C/W]*6.58[W] = 75C Celsius. Ending Tj results in 75+45=120 Celsius. Just within limits.

When I made a 30 second ventillation (blowing the heatsink) Vf rose by 0.04V that is Tj was cooler by 10C.

[LuxIII calculations]

Vf shifted by -0.06V

Using formula from DS45 (-2mV/Celsius shift in Vf) it heated up by 30 Celsius. As room temp (Ta) was 22 Celsius, starting Tj was 22[C] + 13[C/W]*3.69[W] = 70C Celsius. Ending Tj results in 70+30=100 Celsius. Safe also.

Of course these are rough estimates (Vf shift over temp can vary and Lumiled datasheet specifies this for 700mA, not 1000mA) but can be used without having a thermometer.

 

[Doug S]

Pinter: a couple of things to consider. Those of us that have attempted to measure the actual tempco of Vf have found that it varies widely from sample to sample and can be much different from the datasheet values of -2mv/C and -4mv/C. Also, most of the junction to heatsink thermal resistance is at the die to slug interface. You need very fast measuring methods to see the Vf at T=0. I've measured the die thermal time constant to be around 20msec.

 

[Pinter]

Doug,

In my calculation I made the assumption that at T=0 (in reality T=2sec was approximately) T_j = T_a + R_th*P. I think at the beginning the ideal heatsinking case formula can be applied where T_slug=T_a. So I did not want to measure Vf at T_j=T_a.

Did you mean that the die temperature remains at T_a for 20ms?

How did you measure tempco? I have an idea, could you check it?
There was a steady state after 20 minutes at T_a = 22C. Vf was 6.40V. What if I put the whole assembly out in the window (temp is 7 Celsius now here), waiting for the steady state again. So I have the Vf difference for 15C and have the tempco for that specific emitter.

(I think the variation between Vf tempcos can be another explanation on why some lower binned LED can be brighter than a higher binned one in real life. Lower die temperature, brighter emission)

 

[Doug S]

[ QUOTE ]
Pinter said:
Doug,

In my calculation I made the assumption that at T=0 (in reality T=2sec was approximately) T_j = T_a + R_th*P. I think at the beginning the ideal heatsinking case formula can be applied where T_slug=T_a. So I did not want to measure Vf at T_j=T_a.



[/ QUOTE ]
OK, I understand. This makes sense. I assume your constant R_th is the datasheet thermal resistance.

[ QUOTE ]
Pinter said:
Did you mean that the die temperature remains at T_a for 20ms?


[/ QUOTE ]
No. I mean that it takes about 20msec for the die temperature to rise by (1-1/e) or about .63 times the amount due to the power times the die to slug thermal resistance.

[ QUOTE ]
Pinter said:
How did you measure tempco? I have an idea, could you check it?
There was a steady state after 20 minutes at T_a = 22C. Vf was 6.40V. What if I put the whole assembly out in the window (temp is 7 Celsius now here), waiting for the steady state again. So I have the Vf difference for 15C and have the tempco for that specific emitter.

[/ QUOTE ]
I think this would work fine assuming that the airflow over the heatsink was about the same in both cases. Not still inside and windy outside.

I have used several methods including attaching the emitter to a temperature controlled hotplate with thermocouple nearby.

[ QUOTE ]
Pinter said:
(I think the variation between Vf tempcos can be another explanation on why some lower binned LED can be brighter than a higher binned one in real life. Lower die temperature, brighter emission)

[/ QUOTE ]
I'm not sure I follow your reasoning. True, if the Vf tempco is larger, the Vf will drop more and hence the power dissapated will be a bit lower at a given current. The magnitude of the effect is small though. For a white lux a drop in temp by 1C only yields about 0.35% increase in flux.

 

[NewBie]

The 5W is (4) 1W's on a common slug. I tortured a 1W once to right at 190 degF and it survived! (R2H, BB750, CR2) I doubt it saw 3/4A for very long (CR2) but I wasn't about to pick it up!

(Hey, it's Friday, have a laugh!)

Larry

That is one of my favorite photos!

OUCH!