Interesting Facts (Cree's R5 XPG)

[clemence]

I never thought that the LED could emit (radiate) such heat up front. Recently I made my own very small (and ultra cheap, I custom ordered for $0.5 each) PCB for 3 XPGs to light an art object. It's an unexpected success for such a crowded configuration. I made 3 XPG in series, total 9.36V running at 680mA), using a DC power supply. The PSU has switches for polarity and option for 1.5V, 3V, 4.5V, 6V, 7.5V, 9V, and 12V output. It's very easy to select the brightness level by adjusting the output voltage from 7.5V, 9V, or 12V. The LED's won't light if the voltage is lower than 7.5VSince I wasn't planning to use MCPCB I asked the CNC operator to cut three slots in the PCB right in the thermal pads of the XPG's (1.2 - 1.3mm wide). It's a messy job since I ended up widen the slots manually using emery paper because he cut it less than 1.2mm. Then I simply flood the slots with solder to get a good thermal path to the base copper cylinder. I think it's better than using several small vias.The ugly but working PCB: The even uglier final PCB: There shall be light!!!!. I tested the LED nonstop for more than 10 hours without any problems. I measured the solder points right at the thermal pads, the highest was 43.4C after 3 hours nonstop test. I assume that was a stable readings. Ambient temperature at 20.3C Temperatur at approx. 1mm above the LED lens. It's a whopping 125.6C and it was still rising up. I placed an aluminum cylinder and measured it's surface temperature it was 36.4C and continue to climb. Based on that findings, It's wise to use lenses according to it's max. temp ratings, especially those which use cavity very close to the emitter. I also found that a closed reflectors ran hotter than open one.

 

[AnAppleSnail]

I never thought that the LED could emit (radiate) such heat up front.

That heat is just visible photons hitting objects and being absorbed. If you used an IR probe just above a high-power LED, the front surface of the probe will heat up from the light.

Neat project - what's the final going to look like?

 

[MikeAusC]

Hi Clemence, thanks for sharing this interesting build.

Tests that I've done, show that about one-quarter of the electrical input leaves out the front as visible or invisible radiation.

The reason the Thermocouple gets to such a high temperature, is because it's tiny and it's designed specifically to lose as little heat as possible by conduction. So it absorbs radiation and gets hotter until it gets hot enough so that the heat it radiates is equal to the heat or light it absorbs is equal. Heating a tiny bead doesn't take much energy.

The actual amount of heat isn't really that large - I can hold my hand in front of 100 watt LED and the heat is not a problem - partly because the blood in my hand can remove the heat effectively.

If you hold your hand in front of a filament light it feels very hot - and that demonstrates the fundamental difference between incandescent and LEDs. Incandescents get rid of waste heat by radiation because the filament is at 3000 degrees. LEDs get rid of waste heat by conduction, because at 100 degrees there is minimal heat radiation.

 

[clemence]

I see, so it won't heat up a transparent object? Now I learnt something. Thx snail!
It's an art object lighting. I must make a light source less in 2cm in diameter (including the reflector). The copper cylinder will be mounted to a 3 meter long parabolic shaped alum/copper pipe. It was designed with a heat pipe for the first 1 meter, but looking at the results it will be a waste of time (3 XPGs at 680mA already bright enough in a dark art gallery). I can pinch the warm copper cylinder with my bare fingers.

 

[MikeAusC]

An object that's transparent to light won't absorb light and convert it to heat . . . but you can't be sure if it's transparent to heat.

My glass doors are totally transparent to light - but the PIR movement sensor won't activate no matter how close I get or how much I move, because the glass absorbs infrared radiation from my body.

 

[clemence]

An object that's transparent to light won't absorb light and convert it to heat . . . but you can't be sure if it's transparent to heat.

My glass doors are totally transparent to light - but the PIR movement sensor won't activate no matter how close I get or how much I move, because the glass absorbs infrared radiation from my body.

Yes, the same goes to A/F sensor in cameras. So no need to worry about the TIR collimator then. =)

 

[Curt R]

Lead solder is not a good thermal conductor. Fill the slot with
copper wire of the same size and solder that in place under the
LED.

Curt

 

[clemence]

Lead solder is not a good thermal conductor. Fill the slot with
copper wire of the same size and solder that in place under the
LED.

Curt

Thx Curt, I thought about that too but my hands were too clumsy to do that at first (FYI, this is my FIRST LED soldering). I couldn't get the copper wire pieces soldered exactly at the center of the slots. I'll do some more practice before another attempts and I'll use pre-cut copper pieces made from copper sheet. But, I think it's still better than using MCPCB. I read somewhere in Cree's website that the average dielectric layer of MCPCB's thermal conductivity is 2.2W/MK. I used Pb60/Sn40 solder with TC approx. 40-50W/MK, far better.

 

[SemiMan]

Well to be accurate it absorbs the wavelengths specific to the PIR sensor ..... and no doubt many others.

 

[SemiMan]

Thx Curt, I thought about that too but my hands were too clumsy to do that at first (FYI, this is my FIRST LED soldering). I couldn't get the copper wire pieces soldered exactly at the center of the slots. I'll do some more practice before another attempts and I'll use pre-cut copper pieces made from copper sheet. But, I think it's still better than using MCPCB. I read somewhere in Cree's website that the average dielectric layer of MCPCB's thermal conductivity is 2.2W/MK. I used Pb60/Sn40 solder with TC approx. 40-50W/MK, far better.

Some of the good metal core substrates are 3-4 W/MK ... and they are 0.004 thick versus your board which looks like 0.064 .. or 16 times thicker. That shrinks the difference between the two considerably.

Of course, if you are doing low volume, I generally suggest Luxeon Rebel ES and an FR4 board with lots of vias. It sounds like you don't have a heat sink problem so there is no need to go overkill on the board. Any board house can make that.

Semiman

 

[MikeAusC]

Some of the good metal core substrates are 3-4 W/MK ... and they are 0.004 thick versus your board which looks like 0.064 .. or 16 times thicker. . . . . .

That's the key point. The thermal path is the dielectric in parallel with the solder. Virtually no heat flows through the dielectric because it's so thick, so there's no point comparing the conductive metal with the poorly conducting dielectric.

Any improvement in thermal conductivity of the metal will reduce the Die temperature and therefore improve reliability and lifetime of the LED.