Heat sink size for 10W LED chip?

primuspaul

Newly Enlightened
Joined
Jan 25, 2018
Messages
15
What sort of heat sink would I need for a 10W LED chip? Assume this is a heat sink without a fan.

Also, would it make more sense (price-wise) to underdrive an LED using a constant voltage power supply (say by running a 12V chip at 10 or 11V) and simply use more LED chips to get the same amount of lighting, while saving money on aluminum/heatsinks?
 

Genes

Newly Enlightened
Joined
Aug 24, 2003
Messages
85
Location
Kentucky, USA
Generally, X number of Lumens = Y number of watts of heat. As long as you are producing X lumens, whether it is with 1 led or 5 leds, the Y number of watts will be about the same. Being that leds are "current" driven devices, constant current supplies provide the best option for driving in most cases. Constant voltage drivers will drive leds just fine as long as you realize that the lumen output will not necessarily be constant.
 

iamlucky13

Flashlight Enthusiast
Joined
Oct 11, 2016
Messages
1,139
Conveniently, most household LED bulbs are right around 10W, so the heatsinks they use should be about the right size, if used in a place where air can circulate freely around the heat sink.

However, some LED's are less heat tolerant than others. It wouldn't be a bad idea to go slightly larger.

You will get a little bit less heat from several underdriven LED's vs. one designed for 10W driven at 10W, but not enough that I would say it is worth the extra complexity.
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
Generally, driving LEDs with fixed voltages gives unpredictable results.


The light output at various currents is well characterized. Temperature and age have an effect, but that's fairly predictable too.


What's not well characterized, nor controlled, is what current you will get at a particular voltage. You could easily get 20% variation from one batch to another of the same LED.


That said, if you are only building one or a few, it would not be unreasonable to build with a constant voltage supply, test, and adjust to get the results you want. With a fixed supply voltage, the easiest way to adjust is with dropping resistors. This cuts into your efficiency somewhat, but it doesn't have to be terrible. If that's what you decide to do, we can help you with that. Anything more complex than dropping resistors starts to look like a constant-current supply. Most LED displays, indicators, etc, are done with fixed voltage and dropping resistors. It's when power levels go up and efficiency becomes important that we begin to steer away from this.


As far as the economics, it's a far more complex question than it seems on the surface. It depends on your expected results, skills, tools and materials at hand, type of light you are making, etc. The value of your time probably exceeds the few dollars you could save, unless you are making a lot of them.
 
Last edited:

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
My apologies in advance for a long post. I hope someone finds this helpful.


To get an idea of the heatsink you need for various powers, you need to know a few thing:

1. Depending on the efficacy of the LED you choose, the LED will probably convert 50-75% of the electrical power it receives into heat.

2. Good quality LEDs will last a good while if run at 85 deg C die temperature. Crap LEDs won't last well at any temperature, but heat makes them much worse.

3. The heatsink will run cooler than the LED die. Running a good quality LED with proper assembly and copper parts, it's possible to keep this difference as low as a few degrees. Unless you really know what you are doing, I'd figure on 15-20 degrees.

4. Let's say you are using good parts, but you are a beginner, so your heatskink temp should be no higher than 65C. Obviously, the more power you put into the heatsink, the hotter it will get. Also, the cooler the air, the more power the heatsink can dissipate. It turns out the amount of power a heatsink can dissipate is based on the difference between the heatsink temp and the air temp. A heatsink that can dissipate 1W at 10 degrees difference could dissipate 2W at 20C, or 3W at 30C difference. Heatsinks are commonly rated in Degrees C per Watt. This one would be 10C/W . Sometimes K is used instead of C, as Kelvin and Celcius are interchangable for temperature differences, so this might be listed as 10K/W.

5. Since the heatsink performance depends on air temp, we need to know the worst case (hottest) air temp this will need to work in. "Room temperature" is generally considered to be 23C, and most elecronics are normalized at 25C. In an enclosed area, the temp could easily be 35 or 40C. In the sun it could be worse. Let's assume we are indoors, and the worst case is kind of warm but not too bad at 35.


Given a heatsink temp of 65C and a max air temp of 35C, we have a differential of 30C (or 30K). If we have a decent LED and are driving it at 10W, we might have 6W of heat to dissipate. So our ideal heatsink would be 30C/6W = 5C/W or 5K/W. A better heatsink would have lower temp rise at given power, so less than 5C/W is better. Since you don't want to use a fan, you want to look at the 'natural convection' numbers. Most heatsinks also have number for 'forced air' aka fans, which are much better (no surprise).


Armed with this knowlege, you can now go hunting for heatsinks. digikey.com is a great place to search for parts. They don't have the greatest prices and they don't have every part in the world, but they have, hands down, the best search tools. I've not seen anyone come CLOSE to them in search tools. You can search for heatsinks, select based on C/W, sort by price, look at pictures, and compare selections. They don't tout it as such, but as an engineer I find it a very educational tool. Even if you plan to build your own, you can see what experts have done in your area of concern.
 

Enderman

Enlightened
Joined
Nov 3, 2016
Messages
328
Location
Vancouver, Canada
Generally, X number of Lumens = Y number of watts of heat. As long as you are producing X lumens, whether it is with 1 led or 5 leds, the Y number of watts will be about the same. Being that leds are "current" driven devices, constant current supplies provide the best option for driving in most cases. Constant voltage drivers will drive leds just fine as long as you realize that the lumen output will not necessarily be constant.
This is 100% incorrect.
Driving an LED at lower currents results in a higher efficiency, lm/W.
So 10 LEDs running at 1W each will produce more lumens than 1 LED running at 10W, OR 1 LED producing X lumens will use more power than 10 LEDs producing the same amount of lumens combined.
The higher efficiency also results in less heat being produced.
 

DIWdiver

Flashlight Enthusiast
Joined
Jan 27, 2010
Messages
2,725
Location
Connecticut, USA
This is 100% incorrect.
Driving an LED at lower currents results in a higher efficiency, lm/W.
So 10 LEDs running at 1W each will produce more lumens than 1 LED running at 10W, OR 1 LED producing X lumens will use more power than 10 LEDs producing the same amount of lumens combined.
The higher efficiency also results in less heat being produced.

Just to dot the 'i's and cross the 't's, you only get this improvement if you compare 1 LED vs multiples of the same LED. Ten 1W LEDs running at 1W each would be about the same (generally) as one 10W LED running at 10W, while ten 10W LEDs running at 1W each would produce more light and less heat. It's an expensive way to gain efficacy.

...(author thinks a bit)... ...just how much efficacy?...

It's not a small difference! I just did some estimates from the XM-L2 datasheet (granted it's getting a bit long in the tooth, but it should be at least roughly representative). Running one 5000K T5 bin at 3A, 85C, you'd expect 845lm. Running ten of those at 300 mA, 85C you'd expect around 1300 lm. If you took advantage of the numbers to spread them out and dissipate the heat better, keeping them at 50C junction temp, you might get closer to 1430 lm, an impressive 69% increase.

Also, this is at 1/10th the current, not 1/10th the power. Because of the lower Vf at low current, the actual input power would be about 17% less, making the overall efficacy just over double.

This is certainly not the result I expected!

Hmm, what if we only used two LEDs?

one LED, 3000 mA, 85C, 845 lm, 3.3V, 9.9W, 85 lm/W
two LEDs, 1500 mA, 85C, 936 lm, 3.05V, 9.15W, 102 lm/W
two LEDs, 1500 mA, 65C, 983 lm, 3.05V, 9.15W, 107 lm/W
ten LEDs, 300 mA, 85C, 1300 lm, 2.75V, 8.25W, 158 lm/W
ten LEDs, 300 mA, 50C, 1430 lm, 2.75V, 8.25W, 173 lm/W
 
Last edited:

FRITZHID

Flashlight Enthusiast
Joined
Jun 20, 2011
Messages
2,500
Location
Icelandic wastelands of Monico, WI
Just to dot the 'i's and cross the 't's, you only get this improvement if you compare 1 LED vs multiples of the same LED. Ten 1W LEDs running at 1W each would be about the same (generally) as one 10W LED running at 10W, while ten 10W LEDs running at 1W each would produce more light and less heat. It's an expensive way to gain efficacy.

...(author thinks a bit)... ...just how much efficacy?...

It's not a small difference! I just did some estimates from the XM-L2 datasheet (granted it's getting a bit long in the tooth, but it should be at least roughly representative). Running one 5000K T5 bin at 3A, 85C, you'd expect 845lm. Running ten of those at 300 mA, 85C you'd expect around 1300 lm. If you took advantage of the numbers to spread them out and dissipate the heat better, keeping them at 50C junction temp, you might get closer to 1430 lm, an impressive 69% increase.

Also, this is at 1/10th the current, not 1/10th the power. Because of the lower Vf at low current, the actual input power would be about 17% less, making the overall efficacy just over double.

This is certainly not the result I expected!

Hmm, what if we only used two LEDs?

one LED, 3000 mA, 85C, 845 lm, 3.3V, 9.9W, 85 lm/W
two LEDs, 1500 mA, 85C, 936 lm, 3.05V, 9.15W, 102 lm/W
two LEDs, 1500 mA, 65C, 983 lm, 3.05V, 9.15W, 107 lm/W
ten LEDs, 300 mA, 85C, 1300 lm, 2.75V, 8.25W, 158 lm/W
ten LEDs, 300 mA, 50C, 1430 lm, 2.75V, 8.25W, 173 lm/W



:twothumbs:twothumbs:twothumbs
 

Enderman

Enlightened
Joined
Nov 3, 2016
Messages
328
Location
Vancouver, Canada
It's not a small difference!
Yeah, it makes a big difference.
Of course it does cost more when you need to buy many LEDs.
Another option is to buy a single more powerful LED and run it at much lower currents, also getting an efficiency gain for the same lumen output.
For example, running an XHP70.2 at 1A or 3A instead of 6A.
Down side is that more powerful LEDs also cost more.
 

iamlucky13

Flashlight Enthusiast
Joined
Oct 11, 2016
Messages
1,139
It's not a small difference! I just did some estimates from the XM-L2 datasheet (granted it's getting a bit long in the tooth, but it should be at least roughly representative). Running one 5000K T5 bin at 3A, 85C, you'd expect 845lm. Running ten of those at 300 mA, 85C you'd expect around 1300 lm. If you took advantage of the numbers to spread them out and dissipate the heat better, keeping them at 50C junction temp, you might get closer to 1430 lm, an impressive 69% increase.

I double checked your numbers, and yes, numerically it sounds impressive.

However, especially when keeping in mind the non-linear human perception of changes in light intensity, I consider that a rather small output increase compared to increasing your cost from $4.89 to $42.90 (Mountain's prices), not counting your MCPCB's, larger enclosure/heat sink, and more optics, and significantly increasing the size of the light, as well. Hypothetically you could at least run them all on the same driver in parallel, although there's some caveats to that, too.

If battery life is of an extremely high value to someone, it is a technique to be aware of, but very few lights are designed to never run an LED above a fraction of its rated output for good reason.
 
Top