LED Voltage Degrades with Time

[Overclocked]

I recently got these:
https://www.luxeonstar.com/item.php?id=377&link_str=202&partno=LXHL-PW03

No, I didnt get the chinese knock offs, these are the real deal, but a different bin (WWOS) I made a PCB so I can test these things and to give them a bit of a heatsink (its 1 inch by 1 inch in size). I soldered it and added some Arctic Silver to the bottom to make it thermally conductive. Both terminals are isolated from the center.

Now Im doing some testing, but LED voltage seems to degrade with time. It does seem to get warm, but is this why? Testing is performed in short periods, probably 30 seconds max to get voltage readings. Supply voltage is constant, and doesnt degrade. Do I need better heatsinking for even short periods?

Right now (as of this post) I am testing it with a current of 490 mA. It seemed to degrade at 383 mA also.

 

[Hero999]

What do you mean by degrade? Did it go up or down?

I take it you were using a resistor and constant voltage supply rather than a constant current supply, if so there will be some feadback between the LED's voltage and the current.

Also, did the LEDs revert back to their normal voltage after the power had been disconnected and they were allowed to cool? I have a feeling that it's temperature related.

When I first saw this thread I thought you were talking about aging LEDs, for example the strike voltage of an arc lamp tends to increase as it ages, I wonder if the same is true for LEDs.

 

[blueroomelectronics]

Try cooling it with a fan and seeing if the same thing happens. Your 1" heatsink sounds way too small.
**broken link removed**

 

[Overclocked]

What do you mean by degrade? Did it go up or down?

I take it you were using a resistor and constant voltage supply rather than a constant current supply, if so there will be some feadback between the LED's voltage and the current.

Also, did the LEDs revert back to their normal voltage after the power had been disconnected and they were allowed to cool? I have a feeling that it's temperature related.

When I first saw this thread I thought you were talking about aging LEDs, for example the strike voltage of an arc lamp tends to increase as it ages, I wonder if the same is true for LEDs.

Degrade = Go Down. It went down at a fairly quick rate ( .01V every second or so) Example: I first measured 6V across the LED, then it quickly went down to 5.99, 5.98, 5.97...etc..etc. At low currents I would measure I constant voltage.

I am using a Computer PSU connected to a LM317T Circuit to vary voltage. Max Voltage coming from the Reg was 9.25V. I dont know if this is important or not, but I painted the PCB black, to reflect heat (or atleast thats what I thought)

The max current I drove these with was 695mA, close to the max of 700mA. It was getting warm, But not HOT (then again, testing was only for 15 seconds or so).

I was going to mount it on a heatsink, but I dont know if FR4 is thermally conductive enough. I was thinking of drilling a hole where the bottom of the LED is (on the PCB) and filling it with thermal paste which would make contact with a heatsink.

But dam, these things are bright! If I could get a hold of the stars, it wouldnt be a thermal problem.

 

[audioguru]

LEDs operate with current, not voltage.
These ones operate at their limit of internal temperature and MUST have a huge heatsink for them to last only 500+ hours. the heatsink design guide is in the literature for them. They could quickly melt inside before the outside becomes warm.

 

[Roff]

LEDs have a negative temperature coefficient of voltage (see this graph), like all other (semiconductor) diodes. As they get hotter, the voltage across them will drop. Do they recover when they cool off?

 

[Overclocked]

LEDs have a negative temperature coefficient of voltage (see this graph), like all other (semiconductor) diodes. As they get hotter, the voltage across them will drop. Do they recover when they cool off?

Yes they do recover when they cool off.

Does anyone know where to get Metal Core PCB? Or atleast the blank stars So I can mount them on a heatsink?

 

[blueroomelectronics]

FR4 is not good heat sink material, black painting a PCB is not going to do much either. A nice large chunk of aluminum 18guage 6.5"x6.5" might help like the article I posted mentioned.

 

[Overclocked]

FR4 is not good heat sink material, black painting a PCB is not going to do much either. A nice large chunk of aluminum 18guage 6.5"x6.5" might help like the article I posted mentioned.

Yea, Ive been trying to figure out how to mount the emitters, but then I saw something like this:
https://saft7.com/?p=226

A Bit Crude, but it looks like it will work, and best of all, I have a HUGE bin of old heatsinks Ive been collecting over the years. I'll mount it on one of those tomorrow and remeasure.

Heres the data Ive collected:
5 W LED Data

R=33.2 Ohms
Vin VLED Vrlimit Current (mA)
5.2 V 5.12 0.08 2.41
6.2 5.33 0.87 26.21
7.19 5.42 1.77 55.31
8.21 5.47 2.74 82.53
9.22 5.51 3.71 111.75
9.48 5.51 3.97 119.58 Note: Max Voltage.

Vin Vled Vrlimit Current (mA) Resistance (Ohms)
9.03 5.72 3.31 300.91 11
9.23 5.8 3.43 385.39 8.9
9.25 5.8 3.43 490 7
9.13 6 3.13 695.57 4.5

Note: 2nd Test was set at max voltage. Duh. I forgot to load down then 5V Rail on the computer PSU, that would explain why Vin is low at high currents. Last 3 voltages were pretty much taken after I turned on power.

 

[picbits]

I've used a lot of Luxeon LEDs. I never run them at their full current - around 70% of their rated maximum current generally gave me the best output vs heat for the application I needed.

Use the LM317 in constant current mode and limit the current to 70% of the rated LED current and see how you go and how hot the heatsink gets.

 

[Sceadwian]

Why does this seem unusual to you? Semi conductor properties change with temperature, like the voltage drop of a diode as the junction heats, or it's resistance, high power LED's are going to heat, it's simply the nature of throwing current through a non superconducting material. Modules like this are usually best fed from constant current sources rather than voltage sources if you can a ford the power loss, keeps brightness level over the life of the device assuming it's operating parameters are met. You're at only slightly more than half the maximum current for the device so there's really no problem.

 

[Overclocked]

For Some Reason, I couldnt get the constant current circuit to work. For R I used 4 ohms or so. The LED lighted up, but I didnt read 1.25V Across the sense resistor. The LM317 Got hot pretty quickly, and so did the LED. I shut the circuit off almost after taking the reading, for fear that the regulator was pulling full current (1.5A !)

I now also have a properly heatsinked LED, Tomorrow morning I will take readings..But for now, take a look at the job I did on the heat sink from a old 486 proc. I shined it up to do two things (1) Increase Heat transfer (2) Be Reflective a bit. Its a technique some hard core overclocker's use to improve heat transfer in heat sinks.

The heat sink is Aluminum, despite it looking copperish. It is also round and has multiple horizontal layers protruding from a cylinder.

 

[Oznog]

Degrade = Go Down. It went down at a fairly quick rate ( .01V every second or so) Example: I first measured 6V across the LED, then it quickly went down to 5.99, 5.98, 5.97...etc..etc. At low currents I would measure I constant voltage.

This is part of their spec. Look at "Temperature Coefficient of Forward Voltage" or something. Voltage decreases as die temp increases in ALL LEDs. For the device you have, Vf decreases 4mv/deg C. In fact, you measure the Vf the moment it powers up with the heatsink and die stabilized at a known room temp and maintain a constant current, you can accurately measure the die temp itself just by looking at the change in Vf. You can calculate how effective your heatsink configuration is. You can verify how close the die is to its thermal limits.

This is why LEDs cannot be driven in parallel and cannot be driven off a constant voltage without a series resistor. If I set a vreg for the rated current at 5.99V at first, as the die warms up the voltage that produces that rated current drops to 5.88V. Wow, now we're giving 0.11V too much after going to so much trouble to make the voltage perfect. The current's too high. Heat generation increases, die temp increases, the Vf decreases and the gap between the appropriate voltage for this die temp and our set voltage increases. Soon it will runaway and destroy the device.

BTW, I think you picked up on this but just to be sre- the thermal slug in back is NOT electrically neutral and CANNOT be grounded not can it share the same heatsink as another non-isolated Luxeon emitter. This is a bummer. A Luxeon Star does not provide additional isolation either. Arctic Silver sometimes electrically insulates the pad and sometimes it doesn't, it is not reliable enough to be "effective" for this job.

I hope you're not applying Arctic thermal GREASE and just sticking it on the heatsink without pressure. This will not get the thermal resistance down low enough and it's very inconsistent. Arctic Silver ADHESIVE epoxy is quite acceptable (apply pressure while curing).

These guys are great hackers for power LEDs:
https://www.candlepowerforums.com/vb/index.php?

 

[Hero999]

For Some Reason, I couldnt get the constant current circuit to work. For R I used 4 ohms or so. The LED lighted up, but I didnt read 1.25V Across the sense resistor. The LM317 Got hot pretty quickly, and so did the LED. I shut the circuit off almost after taking the reading, for fear that the regulator was pulling full current (1.5A !)

Did you allow 4.2V of headroom between the power supply and the LED forward voltage?

Did you mount the LM315 on an adequate heatsink?

Was the voltage drop across the LM317 greater than 15V?

 

[Overclocked]

Before I used Arctic Silver mixed with Super glue to make some kind of Thermal Superglue. I think it worked, that is, until I drove the LED with too much current and the inevitable happened. I also went to get some epoxy, mixed it up with some arctic silver and then mounted another LED.

Did you allow 4.2V of headroom between the power supply and the LED forward voltage?

Did you mount the LM315 on an adequate heatsink?

Was the voltage drop across the LM317 greater than 15V?

Hmm, "Supposedly" The PSU is supposed to give 12V Out. I was getting around 9.3V, probably because I still didnt load down the 5V rail. LM317 Was on a heatsink, and it didnt get warm at all during testing (In constant voltage mode).

I really dont have a PSU greater than 12V, Oh wait! I do have a 13.5V Linear PSU that was used to power a CB radio...

 

[Oznog]

OK this mixing thermal grease with superglue or epoxy has got to stop. This will result in completely unpredictable mechanical strength and thermal resistance and probably destroy the device. There is Arctic Alumina/Silver (Silver is better) adhesive epoxy. Costs like $10 or so at Fry's. This is what you need.