Fiiiiine, Dave! I'll binge watch from the start. Maybe I'll catch up by 2030! Thank you for all the work you've done over the years! You've filled in many of the (many many) knowledge gaps in this physicist/software developer's education.
@jasonuscg It's called a conversational off-the-cuff style. If you think it's too long then you go redo it in 7 minutes. I'll give you a week, but remember - no script and no retakes - I'll trust you. Once you are done, post it as a video reply so we can all critique your style.
Great vids! Doing my best to avoid the young player traps haha - Ive built an SPD simulator in Excel that might be useful in any future emitter efficiency calculations if you're interested... - ua-cam.com/video/0dLx7-9xn_U/v-deo.html
Another issue with efficacy to efficiency calculations is that chromaticity of your light can make a really huge difference. Most likely the efficiency of your leds is actually a lot lower than you calculated, because they tend to be fairly high in the green/yellow region in order to get those amazingly good numbers. Also, they're probably measuring the efficacy at 350mA, and not 700. Regardless, they're pretty amazing, and unlike a lot of the other power leds they look reasonably natural.
@OldSkull87 Impossible to say unless you have the reliability specs for the LED in question and you can accurately measure the junction temperature. The generic answer is "as low as possible"
seeing that drop in the temperature graph got me thinking about the possibility of making a "windometer" or whatever based on this effect. of course after some googling it was already invented, check wikipedia "hot-wire anemometer", its pretty much the same thing and apparently those devices are used for studying turbulent flow. interesting
I don't know if someone has already noted this but the maximum possible efficacy of a light source is 683 lm/ W, for the case of monochromatic 555-nm green light. The maximum possible efficacy for an ideal white light source is 240 lm/W because the human eye is less sensitive to wave- lengths on each side of green.
I've got a Tesco brand 3W Cree LED torch which runs from two C cell batteries. It seems to use the torch body as a heatsink, and after a few hours running it gets "a bit warm". I'm guessing its batteries will run out before the torch heats up enough to damage the LED.
I think LEDs are fascinating. I believe LED lighting will become more popular as people start to look into other alternatives to incandescent lighting besides CFLs. Lumens per watt are going up and prices are coming down, very exciting. As far as Thermal Design goes, overkill has always worked for me.
Hey Dave, nice work again. I've been stuffing around with LEDs for many years and haven't gone into thermal calculations , instead just used PC heatsinks. It's amazing at what extents people go to in trying to get an 'optimal' cooling setup (candlepowerforums). Got a torch from dealextreme which doesn't seem to have much control circuitry. (SSC-P7) but the damn thing works great, maybe due to good Chinese engineering and much aluminium OR more likely just the "guess it'll work OK" factor.
thermal conductivity of aluminium is 237 W·m-1·K-1 according to wikipedia it's high, that's why you didn't get "hot spot" or large differences between your half quarter and under button if you'd used "angle iron" aluminium you would have gotten better convection cooling and even better with a couple vertical fins a thicker aluminium strip would ramp up slower but the plateau would be the same temp because the choke point is how much convection you're getting so only higher surface would cool it
I plan on using LEDs to light my room and of course I want to bulild my own :). As 5 years have passed now, I wonder what the best LEDs now are. They need to be warm white. I already tried some cheap 10W LEDs from ebay but they weren't that great.
3vIL VIrUs In terms of luminous efficiency, I think the best ones to get are still Cree, but they have some new product lines. They have multi-die packages for higher operating voltages (higher voltage means less power wasted in drivers and cables) and ones with larger footprints and much higher outputs like their XM-L2, MK-R and MT-G2. I built myself an LED desk lamp using a ~12w 4000K Bridgelux COB (Chip on board, a multi-chip module) and it's pretty nice. Keep in mind that you'll take a pretty serious efficiency hit for using warm white LEDs, especially at high CRI (Color rendering index, basically how well the light curve approximates a blackbody radiator like the sun or a tungsten filament, a perfect match is 100). For example, the Cree XM-L2 easywhite LED datasheet lists a maximum of 466 lumens at 700ma, 6v, 25c with a 5000K neutral/cool led. Same test conditions with a 5000K LED with a minimum CRI of 80 (normally 5000K is around 70-75) brings that down to 443. Same test with a 2700K 80-CRI warm white returns 396, and a 90-CRI 2700K gets 280, or only about 60% of the light output. Keep in mind that those are absolute maximum ratings with the best LEDs to come off the assembly line, and you see the problem with really nice warm LEDs: You're going to need about twice as many and twice as much power if you go for the nicest ones. Nichia makes some LEDs with really awesome light quality, but their power levels and efficiency aren't quite up to par with Cree. There is also Phillips Lumileds, Osram, Seoul Semiconductor, Bridgelux and Luminous Devices (maker of the largest and brightest single-chip LED), just to name a few of the large players. I recommend going to Digikey, Mouser or a similar electronics supplier and looking through their catalog, they have nice parametric search engines. Download datasheets for any likely candidates, and compare. If you have any questions, feel free to ask me.
Ryan Willis Wow that is a verry detailed answer, thank you verry much! I think I will test some cold white LEDs, but I really like the warm light my current halogen lights emit.
3vIL VIrUs No problem, glad I could help. I started short, and I just kept thinking of things I had to add. Something in the 4000-5000K range is reasonably neutral, and a nice tradeoff between acceptable efficiency and decent CRI and color temperature. If you're used to halogens, anything above 5000K is going to look really blue and cold until you get used to it.
In the U.S. some engineers didn't bother to do thermal calculations either and put LEDs into traffic lights. Now they don't generate enough heat anymore to melt any snow blocking them. Google for 'led traffic light snow' for more information. ;)
Or they did thermal calculations, realised there was no danger of overheating, but didn't even realise underheating was a problem. I can see that happening if all the engineering was done in a place where it rarely snows...
Your video was very interesting although at the end It didn't clear a lot of things up in terms of calculating the needs of dissipation or heatsink measures. I'm designing a led light panel for aquarium light with 4 modules of 7 led stars (14W each) and I'm trying to figure out the heatsink I need for passive refrigeration (no case or fans, all open for maximum air circulation) and it's getting quite tricky! :(
Seeing that these are so powerful... Could they sustain them self with soralr panels... And If not could on magnefigh the lifgt to do so... No matter what I Loved the show... I learned a bit... Thanks for taken the time to share your knowledge Peace!!!
been watching all videos from 1 to 50. Suddenly noticed he drops the words "you know" over and over again. great idea for a drinking game, you know ;-)
Good Day Dave Just watched video #50. It is 3+ years since you installed the Cree LEDs in your enclosed patio. Curious to know if you've had any degradation or other issues with the units you used? Cheers, Dave
I'm building a bike light with some cheapo Chinese LEDs for which I don't have datasheets and was quite worried about heatsinking.... I guess I'll just use one I have lying around and measure the temperature for a while :)
I was assigned to be the thermal engineer for a group capstone project. At the review, the lead from our sponsor asked who I pissed off in order to get the job. The basic calculations always suffer from a lot of potentially flawed assumptions about the system, and though a finite element model can help with more complex systems it's still GIGO. I don't suppose you ever ran it through a FEM?
Thermal design is a bastard sometimes. Convection, conservation of energy, emissivity. I do a lot of it, and it never ceases to amaze me. I'm getting into optics now, making portable spotlights. Shooting for 50 watts. I'm reverse engineering a Ryobi P717 LED spotlight. We're going for a 1.0 kilometer target in a reflector exercise.
I measure temperatures with an infra-red ear-temperature, used by caring parents. Choose one which has a wider temperature measurement than just human body temperatures. It can be used to measure hot spots on all sorts of gadgets: heatisink parts, sides of computer units, lighting, etc. Efficiency can be measured heating in a bottle distilled water, near external temperatures, as described by another comment here. If the water container is a silver-coated vacuum flask, the increase in water temperature would include light-turne-into-heat, as well. Moving the voltage from mains to CREE voltage would loose "efficiency". A simple current meter would measure this. If the power source was a rechargable or other battery, I'm not sure how to measure this. Depreciation costs on power-charging-storage-discharge would increase the cost of this power-source, compared to mains power.
The LEDs themselves (At least modern Cree ones, XP-G, XP-E, XM-L, etc, not sure about others) have an electrically isolated thermal path already. They have three pads underneath, two smallish ones for power and one big one in the middle for thermal relief. You can get LED stars with a direct electrical and thermal connection to that middle pad, and since there's no film, it conducts heat even better. They're often used in high-power LED flashlights since the lower you can keep the die temperature, the higher the lumen output for a given power input. Keeping the die temp low also improves the lifespan of the LED.
You are wacky lol, but good info in the vid. I don't have electronics background, but you can get decent led technology cheap these days. Definitely the future of lighting. I have a couple of led flashlights, with xm-l cree. They are awesome !
My server kept warning me that the CPU was over temperature and that the world was going to end, etc. So I opened the PC up, stuck my finger on the heatsink and after noticing it was mildly warm, put the lid back on and decided to ignore the temperature warnings.
Thanks (or no thanks) to medical involvement, i had to learn quick between efficiency and efficacy. I had no idea how prevalent the latter was until I learned it. A quite neat word actually, I like it.
Yeah, it was fresh compound. The machine had also been running for a few months quite happily before. I blame an uncalibrated sensor in the motherboard after upgrading the software it runs.
why? have you seen the ultra bright LED Flashlight video here on youtube? it's insane what those little beasts can do. Although a 100 watt array, will take some space so the light will be distributed over a certain area without optics
TheLukey21 That's 330 lumens per 100 watts of light energy, not 330 per watt of power in. You missed the part where he points out that this LED gets 130 lumens per watt of power input, or about 39 percent of the theoretical maximum. Keep in mind that to get that 130 lm/w figure you need to run the LEDs well below their maximum output capability and keep them really well cooled. At full power, their efficiency is about 90-105 lm/w, so for 100w that would be 9000-10500 lumens. Still a crazy amount of light.
also you say it's messy and not worth doing and that we shouldn't write because your model is wrong well.. that's why you think it's messy and full of uncontrollable variables, your model is wrong !!! I have a very hard time believing that those LEDs are 39% efficient from what I read LEDs are 22% efficient at best ! eye color response is irrelevant to the calculation of input power to luminous flux also a 39% efficient LED isn't outputing 39% of it's power as heat that would be 100-39=61%!!!
it's disingenuous to say XYZ doesn't work and then prove it's not worth doing by doing it several orders of magnitudes wrong !!! yes thermal design has a medium high learning curve but it's a useful consistent tool when you use it properly you don't HAVE to do trial and errors if in doubt overdesign , put fins and more mass and more powerful fan when you can but a good engineer will try to be as lean as possible (I'm NOT saying you're bad) on the same topic what do you think of those 100W LEDs
Somehow I watched the whole video... Talk about waste and efficiency this 21 minute video could have been done in 7, not very efficient. He sounds like Herbert form Family Guy!!!
Chill! The stress in your voice is way over the top. You'll get a heart attack if you don't learn to calm down. We'd have to heat-sink you if you started talking about something really exciting like politics or religion or football.
if in doubt, use a bigger heatsink
Lol... Did diodegonewild make his spotlight CFL video by chance? The lights he took apart look very similar to the ones Dave showed at 0:35 :D
agreed very similar diode gone wild's cfl
Fiiiiine, Dave! I'll binge watch from the start. Maybe I'll catch up by 2030! Thank you for all the work you've done over the years! You've filled in many of the (many many) knowledge gaps in this physicist/software developer's education.
One of the better videos I've seen of your's - Dave. Funny and highly efficient...
Interesting project, and interesting to have the results 11 year after ^^
@@Okurka. thanks ^^
@jasonuscg It's called a conversational off-the-cuff style. If you think it's too long then you go redo it in 7 minutes. I'll give you a week, but remember - no script and no retakes - I'll trust you. Once you are done, post it as a video reply so we can all critique your style.
Great vids! Doing my best to avoid the young player traps haha - Ive built an SPD simulator in Excel that might be useful in any future emitter efficiency calculations if you're interested... - ua-cam.com/video/0dLx7-9xn_U/v-deo.html
Another issue with efficacy to efficiency calculations is that chromaticity of your light can make a really huge difference.
Most likely the efficiency of your leds is actually a lot lower than you calculated, because they tend to be fairly high in the green/yellow region in order to get those amazingly good numbers. Also, they're probably measuring the efficacy at 350mA, and not 700.
Regardless, they're pretty amazing, and unlike a lot of the other power leds they look reasonably natural.
@OldSkull87 Impossible to say unless you have the reliability specs for the LED in question and you can accurately measure the junction temperature. The generic answer is "as low as possible"
seeing that drop in the temperature graph got me thinking about the possibility of making a "windometer" or whatever based on this effect. of course after some googling it was already invented, check wikipedia "hot-wire anemometer", its pretty much the same thing and apparently those devices are used for studying turbulent flow. interesting
+jonsykkel
That's how nowadays automotive air flow meters (MAF) work mate...
+krisbham heh, interesting indeed..
I don't know if someone has already noted this but the maximum possible efficacy of a light source is 683 lm/ W, for the case of monochromatic 555-nm green light. The maximum possible efficacy for an ideal white light source is 240 lm/W because the human eye is less sensitive to wave- lengths on each side of green.
I've got a Tesco brand 3W Cree LED torch which runs from two C cell batteries. It seems to use the torch body as a heatsink, and after a few hours running it gets "a bit warm". I'm guessing its batteries will run out before the torch heats up enough to damage the LED.
Did you check the heatsink compound undernearth? If that's not in good nick then you might indeed have a problem.
I think LEDs are fascinating. I believe LED lighting will become more popular as people start to look into other alternatives to incandescent lighting besides CFLs. Lumens per watt are going up and prices are coming down, very exciting.
As far as Thermal Design goes, overkill has always worked for me.
oh how i love Dave!! Thank you brother, i learn so much watching your videos.
Hey Dave, nice work again.
I've been stuffing around with LEDs for many years and haven't gone into thermal calculations , instead just used PC heatsinks.
It's amazing at what extents people go to in trying to get an 'optimal' cooling setup (candlepowerforums).
Got a torch from dealextreme which doesn't seem to have much control circuitry. (SSC-P7) but the damn thing works great, maybe due to good Chinese engineering and much aluminium OR more likely just the "guess it'll work OK" factor.
thermal conductivity of aluminium is 237 W·m-1·K-1 according to wikipedia
it's high, that's why you didn't get "hot spot" or large differences between your half quarter and under button
if you'd used "angle iron" aluminium you would have gotten better convection cooling and even better with a couple vertical fins
a thicker aluminium strip would ramp up slower but the plateau would be the same temp because the choke point is how much convection you're getting so only higher surface would cool it
I plan on using LEDs to light my room and of course I want to bulild my own :). As 5 years have passed now, I wonder what the best LEDs now are. They need to be warm white. I already tried some cheap 10W LEDs from ebay but they weren't that great.
***** Yes, but farnell has thousands of LEDs, I like to know what are the best LEDs right now to buy.
3vIL VIrUs In terms of luminous efficiency, I think the best ones to get are still Cree, but they have some new product lines. They have multi-die packages for higher operating voltages (higher voltage means less power wasted in drivers and cables) and ones with larger footprints and much higher outputs like their XM-L2, MK-R and MT-G2. I built myself an LED desk lamp using a ~12w 4000K Bridgelux COB (Chip on board, a multi-chip module) and it's pretty nice.
Keep in mind that you'll take a pretty serious efficiency hit for using warm white LEDs, especially at high CRI (Color rendering index, basically how well the light curve approximates a blackbody radiator like the sun or a tungsten filament, a perfect match is 100).
For example, the Cree XM-L2 easywhite LED datasheet lists a maximum of 466 lumens at 700ma, 6v, 25c with a 5000K neutral/cool led.
Same test conditions with a 5000K LED with a minimum CRI of 80 (normally 5000K is around 70-75) brings that down to 443.
Same test with a 2700K 80-CRI warm white returns 396, and a 90-CRI 2700K gets 280, or only about 60% of the light output.
Keep in mind that those are absolute maximum ratings with the best LEDs to come off the assembly line, and you see the problem with really nice warm LEDs: You're going to need about twice as many and twice as much power if you go for the nicest ones.
Nichia makes some LEDs with really awesome light quality, but their power levels and efficiency aren't quite up to par with Cree. There is also Phillips Lumileds, Osram, Seoul Semiconductor, Bridgelux and Luminous Devices (maker of the largest and brightest single-chip LED), just to name a few of the large players. I recommend going to Digikey, Mouser or a similar electronics supplier and looking through their catalog, they have nice parametric search engines. Download datasheets for any likely candidates, and compare. If you have any questions, feel free to ask me.
Ryan Willis
Wow that is a verry detailed answer, thank you verry much!
I think I will test some cold white LEDs, but I really like the warm light my current halogen lights emit.
3vIL VIrUs No problem, glad I could help. I started short, and I just kept thinking of things I had to add. Something in the 4000-5000K range is reasonably neutral, and a nice tradeoff between acceptable efficiency and decent CRI and color temperature. If you're used to halogens, anything above 5000K is going to look really blue and cold until you get used to it.
In the U.S. some engineers didn't bother to do thermal calculations either and put LEDs into traffic lights. Now they don't generate enough heat anymore to melt any snow blocking them.
Google for 'led traffic light snow' for more information. ;)
Or they did thermal calculations, realised there was no danger of overheating, but didn't even realise underheating was a problem.
I can see that happening if all the engineering was done in a place where it rarely snows...
Thanks for giving the update
Dave, @19:09 is it the high junction temperature causing the sensation on the finger? Or it's the ir spectrum of the light emission
Dave the way you were holding the whiteboard and pen reminds me of a violin player.
Your video was very interesting although at the end It didn't clear a lot of things up in terms of calculating the needs of dissipation or heatsink measures. I'm designing a led light panel for aquarium light with 4 modules of 7 led stars (14W each) and I'm trying to figure out the heatsink I need for passive refrigeration (no case or fans, all open for maximum air circulation) and it's getting quite tricky! :(
Seeing that these are so powerful... Could they sustain them self with soralr panels... And If not could on magnefigh the lifgt to do so... No matter what I Loved the show... I learned a bit... Thanks for taken the time to share your knowledge Peace!!!
been watching all videos from 1 to 50. Suddenly noticed he drops the words "you know" over and over again. great idea for a drinking game, you know ;-)
Good Day Dave
Just watched video #50.
It is 3+ years since you installed the Cree LEDs in your enclosed patio.
Curious to know if you've had any degradation or other issues with the units you used?
Cheers,
Dave
I'm building a bike light with some cheapo Chinese LEDs for which I don't have datasheets and was quite worried about heatsinking.... I guess I'll just use one I have lying around and measure the temperature for a while :)
How much better are led lights 6 years later?
just how fast do these cree leds go? how long until light is say 50% of what they do when new?
2:20 - How may lumens/second?
How much do the Cree LED's cost and where did you get them please?
You used the temperature differential from the small LEDs a friends gave him in the equation for the large led he had the data sheet on...
I was assigned to be the thermal engineer for a group capstone project. At the review, the lead from our sponsor asked who I pissed off in order to get the job. The basic calculations always suffer from a lot of potentially flawed assumptions about the system, and though a finite element model can help with more complex systems it's still GIGO.
I don't suppose you ever ran it through a FEM?
Thermal design is a bastard sometimes. Convection, conservation of energy, emissivity. I do a lot of it, and it never ceases to amaze me. I'm getting into optics now, making portable spotlights. Shooting for 50 watts. I'm reverse engineering a Ryobi P717 LED spotlight. We're going for a 1.0 kilometer target in a reflector exercise.
I measure temperatures with an infra-red ear-temperature, used by caring parents. Choose one which has a wider temperature measurement than just human body temperatures. It can be used to measure hot spots on all sorts of gadgets: heatisink parts, sides of computer units, lighting, etc.
Efficiency can be measured heating in a bottle distilled water, near external temperatures, as described by another comment here. If the water container is a silver-coated vacuum flask, the increase in water temperature would include light-turne-into-heat, as well.
Moving the voltage from mains to CREE voltage would loose "efficiency". A simple current meter would measure this. If the power source was a rechargable or other battery, I'm not sure how to measure this. Depreciation costs on power-charging-storage-discharge would increase the cost of this power-source, compared to mains power.
Absolute class as always
Totally mad, I can't say I understood it all, but it was interesting what I did learn and most entertaining watching your excitement :D
Do those little stars have a plate on the back for conducting heat but not electricity?
The LEDs themselves (At least modern Cree ones, XP-G, XP-E, XM-L, etc, not sure about others) have an electrically isolated thermal path already. They have three pads underneath, two smallish ones for power and one big one in the middle for thermal relief. You can get LED stars with a direct electrical and thermal connection to that middle pad, and since there's no film, it conducts heat even better. They're often used in high-power LED flashlights since the lower you can keep the die temperature, the higher the lumen output for a given power input. Keeping the die temp low also improves the lifespan of the LED.
What about a COB CXB3590 that puts out more than 50 watts.
700mA constant current. See my website for the links.
awesome shirt :D see you 100 years ago
Couldn't PWM come to the rescue in reducing wasted heat?
You are wacky lol, but good info in the vid.
I don't have electronics background, but you can get decent led technology cheap these days. Definitely the future of lighting.
I have a couple of led flashlights, with xm-l cree.
They are awesome !
arctic silver does make pastes.
another great video
My server kept warning me that the CPU was over temperature and that the world was going to end, etc.
So I opened the PC up, stuck my finger on the heatsink and after noticing it was mildly warm, put the lid back on and decided to ignore the temperature warnings.
Thanks (or no thanks) to medical involvement, i had to learn quick between efficiency and efficacy. I had no idea how prevalent the latter was until I learned it. A quite neat word actually, I like it.
Yeah, it was fresh compound. The machine had also been running for a few months quite happily before. I blame an uncalibrated sensor in the motherboard after upgrading the software it runs.
So a 100w array would put out around 33000 lumens, that can't be right.
why? have you seen the ultra bright LED Flashlight video here on youtube? it's insane what those little beasts can do. Although a 100 watt array, will take some space so the light will be distributed over a certain area without optics
TheLukey21 That's 330 lumens per 100 watts of light energy, not 330 per watt of power in. You missed the part where he points out that this LED gets 130 lumens per watt of power input, or about 39 percent of the theoretical maximum. Keep in mind that to get that 130 lm/w figure you need to run the LEDs well below their maximum output capability and keep them really well cooled. At full power, their efficiency is about 90-105 lm/w, so for 100w that would be 9000-10500 lumens. Still a crazy amount of light.
also you say it's messy and not worth doing and that we shouldn't write because your model is wrong
well.. that's why you think it's messy and full of uncontrollable variables, your model is wrong !!!
I have a very hard time believing that those LEDs are 39% efficient
from what I read LEDs are 22% efficient at best ! eye color response is irrelevant to the calculation of input power to luminous flux also a 39% efficient LED isn't outputing 39% of it's power as heat that would be 100-39=61%!!!
my said she doesn't understand most of what he's talking about but likes his accent!
Sounds like a job for matlab.
Acho que você se refere a frequencias ? Sim é a filosofia do movimento zeitgeist moving forward , mudânça de comportamento social , compreend?
👍👍
XD!!!! the conclusion is great!
@jasonuscg
well if you think you can produce a more 'efficient' video then off you go! link it to me when your done! I'll be waiting..
He sounds and looks so surprised with everything he says lol.
you remind me of the joker before he was evil/crazy
it's disingenuous to say XYZ doesn't work and then prove it's not worth doing by doing it several orders of magnitudes wrong !!!
yes thermal design has a medium high learning curve but it's a useful consistent tool when you use it properly you don't HAVE to do trial and errors
if in doubt overdesign , put fins and more mass and more powerful fan when you can
but a good engineer will try to be as lean as possible (I'm NOT saying you're bad)
on the same topic what do you think of those 100W LEDs
Such a great video :)
thank you for sharing it with us
you just saved me $250
Thanks again
Somehow I watched the whole video... Talk about waste and efficiency this 21 minute video could have been done in 7, not very efficient.
He sounds like Herbert form Family Guy!!!
Lol, aluminium strip, or Aluminum strip for you US people, lol that made me laugh
Chill! The stress in your voice is way over the top. You'll get a heart attack if you don't learn to calm down. We'd have to heat-sink you if you started talking about something really exciting like politics or religion or football.
But that's why this blog is fun!