Five technical questions about LEDs in general

[LMF5000]

Hope this is the right subforum. I have some generic questions about LEDs in general. Here they are in no particular order:

1. How much of the input power to an LED is lost as heat? For example, take a Cree XM-L running with 10W input power (3.2V x 3.0A). How many watts are actually dissipated in the heatsink?

2. For multi-LED setups, is the total number of lumens simply equal to the sum of the lumen output of each LED (eg. 4*1000lm LEDs = 4000lm total)?

3. If you needed, say, 500 lumens for a project, would it be best to use a single LED running at 500lm, or multiple LEDs running at less lumens (say, 2*250lm or 5*100lm LEDs) - assuming the LEDs used are running at less than their maximum lumen rating? (so for this example, say the LEDs available are all capable of 1000lm but running at reduced current).

4. What is the current "best" high-power (over 1000lm) single LED on the market? Is it the Cree XM-L2 or are there other LEDs with more lumens per watt or more lumens per dollar?

5. What is the single most powerful single LED currently on the market (in terms of lumens)?

 

[DIWdiver]

1. Semi-Man or someone will correct me if I'm wrong, but I think the best, newest, most efficient emitters, driven at 'nominal' current (which is often lots less than max), it can be as low as about 70% heat. Older versions, driven to max or beyond, it can be as high as 85-90%.

2. Yes,

3. They are more efficient at lower drive, but it's more cost and trouble to design and assemble them, so it's a tradeoff.

4. Arguably the XM-L2. I think some of the smaller ones from Cree (all gen2) might have a slight edge, but at lower power levels. If you are looking for good color rendition, the answer would be different, but it would likely still be a Cree. If you want excellent color rendition, I don't know.
5. Single die, Luminus CBT-140. If you can't get that yet, SST-90. If you want to go several dice in a small package, CBT-360. If you want to go bigger packages, probably something from Bridgelux, but that's many dice in a single package and can't be focused well (at least easily) because it's so large. Cheap no-name crap may match Bridgelux's specs, but won't last and may not even meet specs at the start.

 

[jtr1962]

1. Semi-Man or someone will correct me if I'm wrong, but I think the best, newest, most efficient emitters, driven at 'nominal' current (which is often lots less than max), it can be as low as about 70% heat. Older versions, driven to max or beyond, it can be as high as 85-90%.

The best whites dissipate about half their input power as heat when driven at nominal current. The top bin royal blues only dissipate about 40% as heat at 350 mA. At higher currents you may have 80% or more dissipated as heat. As a rough guide, for phosphor whites 100% efficient is ~300 lm/W. Just use the actual efficacy in lumens per watt and divide by 300 to obtain the rough efficiency. For example, if an LED achieves 120 lm/W, then efficiency = 120/300 = 0.4 = 40%. The remainder (60%) of the input power comes out as heat.

 

[MikeAusC]

. . . . As a rough guide, for phosphor whites 100% efficient is ~300 lm/W. . . .

From the definition of Lumens, the maximum efficacy for a white light is around 250 Lumens per Watt (http://en.wikipedia.org/wiki/Luminous_efficacy#Lighting_efficiency) , but the rest of your argument is wrong because this is OPTICAL WATTS, not electrical watts.

 

[Yoda4561]

250 has to be wrong, since cree has an LED that does more than that. Can't be breaking the laws of physics and all that http://finance.yahoo.com/news/cree-sets-r-d-performance-140000953.html (ahh I see it, the ideal black body number truncated to the visible spectrum is 251, that would be accurate perhaps for a 100CRI LED, but not for the more typical 70-92 CRI leds we commonly see) from that same chart under the LED row, "Theoretical limit (white LED with phosphorescence color mixing) 260.0–300.0[26]" I've seen numbers as high as 330 or so but that's probably with lots of green in the spectrum.

 

[SemiMan]

From the definition of Lumens, the maximum efficacy for a white light is around 250 Lumens per Watt (http://en.wikipedia.org/wiki/Luminous_efficacy#Lighting_efficiency) , but the rest of your argument is wrong because this is OPTICAL WATTS, not electrical watts.

Nope, everything JtR wrote is right. Only thing is the LER is above 300 on some of these LEDS.

Semiman

 

[DIWdiver]

Nope, everything JtR wrote is right. Only thing is the LER is above 300 on some of these LEDS.

Semiman

Absolutely. A 100% efficient LED would convert 1 electrical watt into 1 optical watt. This would generate some number of lumens (around 300?), which actual number would depend on the spectrum of light produced.

 

[jtr1962]

Nope, everything JtR wrote is right. Only thing is the LER is above 300 on some of these LEDS.

I picked 300 because that seems to be what the LER of many general lighting LEDs with CRIs in the low to mid 80s is around. I used to use 330 alot back when all we had were 6500K, ~70 CRI LEDs. As I'm sure you know you can get as high as 400 with reasonable color rendering by using RGB and carefully choosing the center wavelengths.

Any idea what the LER is for high-CRI (90-95) LEDs? I'm guessing about 260 to 270, or thereabouts.

 

[SemiMan]

I picked 300 because that seems to be what the LER of many general lighting LEDs with CRIs in the low to mid 80s is around. I used to use 330 alot back when all we had were 6500K, ~70 CRI LEDs. As I'm sure you know you can get as high as 400 with reasonable color rendering by using RGB and carefully choosing the center wavelengths.

Any idea what the LER is for high-CRI (90-95) LEDs? I'm guessing about 260 to 270, or thereabouts.

300 was a very reasonable number to use. Phosphor only 90-95CRI at low CCT are in the 250 range (maybe worse!). They generate a lot of deep red that contributes little to the lumen output.

Semiman

 

[LMF5000]

1. Semi-Man or someone will correct me if I'm wrong, but I think the best, newest, most efficient emitters, driven at 'nominal' current (which is often lots less than max), it can be as low as about 70% heat. Older versions, driven to max or beyond, it can be as high as 85-90%.

2. Yes,

3. They are more efficient at lower drive, but it's more cost and trouble to design and assemble them, so it's a tradeoff.

4. Arguably the XM-L2. I think some of the smaller ones from Cree (all gen2) might have a slight edge, but at lower power levels. If you are looking for good color rendition, the answer would be different, but it would likely still be a Cree. If you want excellent color rendition, I don't know.
5. Single die, Luminus CBT-140. If you can't get that yet, SST-90. If you want to go several dice in a small package, CBT-360. If you want to go bigger packages, probably something from Bridgelux, but that's many dice in a single package and can't be focused well (at least easily) because it's so large. Cheap no-name crap may match Bridgelux's specs, but won't last and may not even meet specs at the start.

Thanks to everyone for the replies so far. This one was particularly useful because of the answers to question #5 - I checked all the LEDs listed and although they do indeed make arrays with up to ~9000 lumens, none of them are better than the cree in terms of lumens per unit cost, at least online (including delivery cost to Malta). As of this moment, an XM-L2 sells on ebay for $6.44 delivered, including star-shaped heat sink.

Incidentally, since this is a thread about general technical questions, is there any reason reason to choose a particular shape of MCPCB (20mm star, 16mm round, 14mm round...) over the others for LED mounting - if considering an application where the MCPCB will be pasted onto a larger heatsink (not used in a flashlight pill)?

 

[LMF5000]

The best whites dissipate about half their input power as heat when driven at nominal current. The top bin royal blues only dissipate about 40% as heat at 350 mA. At higher currents you may have 80% or more dissipated as heat. As a rough guide, for phosphor whites 100% efficient is ~300 lm/W. Just use the actual efficacy in lumens per watt and divide by 300 to obtain the rough efficiency. For example, if an LED achieves 120 lm/W, then efficiency = 120/300 = 0.4 = 40%. The remainder (60%) of the input power comes out as heat.

Thanks for that tip. I always knew about lm/w but it never occurred to me to compare it to theoretical efficacy limit to calculate thermal efficiency - but now that I think about it it makes perfect sense.

So, if I were to look at an XM-L at max current (3A), we have 100lm/w (as per cree literature). That's one third of the 300lm/w efficacy limit, so the LED is 33.3% efficient. The power input is 10W, of which 66.6% is wasted as heat. So for heatsink calculations we can take the heat flux from the LED to be 6.7W. Think I'm getting the hang of this :thumbsup:

 

[jtr1962]

So, if I were to look at an XM-L at max current (3A), we have 100lm/w (as per cree literature). That's one third of the 300lm/w efficacy limit, so the LED is 33.3% efficient. The power input is 10W, of which 66.6% is wasted as heat. So for heatsink calculations we can take the heat flux from the LED to be 6.7W. Think I'm getting the hang of this :thumbsup:

Yep, you got the idea! Back in the (not so) good old days, the power dissipated by the heat sink could more or less be assumed to equal the input power to the LED. Now it's better to run the numbers because you'll often find you can get by with a somewhat smaller heat sink thanks to LED efficiency improvements. Note how even a relatively small increase in efficiency can have dramatic results in the amount of heat dissipated. Let's say we go from 100 to 150 lm/W with the same output as your example (1000 lumens). Now we're only dissipating 3.33 watts of heat. If we go to 200 lm/W, we're down to 1.67 watts.