DIY LEDs - The write-up
- Thread starter JohnL
- Start date
- Status
der_wille_zur_macht
Team RC
And DWIZUM said, let there be light!
24v DC power supply in foreground, home-made driver in lower right, and 8 LEDs on a PC heatsink back by the soldering iron.
24v DC power supply in foreground, home-made driver in lower right, and 8 LEDs on a PC heatsink back by the soldering iron.
der_wille_zur_macht
Team RC
Dp, it's standard practice to use a thermal compound between the star and the heatsink. You can be cheap (like me) and use generic silicone-based metallic thermal grease from Radio Shack at $3 a tube, or you can be fancy and use Arctic Silver, which is about $10 a tube.
It's very important to have a good boundary between the star and the heat sink. Grease has always been the standard. A thin tape (5 mils) would be better than an air interface but the more I'm thinking of an adhesive tape (whether it has a thermal label or not) it will outgas with time especially at an elevated temperature thereby increasing the thermal resistance to the heat sink. If this statement is true then time will be 'time bomb' because the LED's junctin temperature will get gradually higher thereby decreaasing the lumen output and reduce the LED's life. I don't think it's worth the risk. Jim
stugray
Premium Member
The phase change material that I posted before:
http://www.kvms.com/Product/HSFPHASECM.aspx
seems to work as good as basic thermal grease ( based on touching each emitter with a thermocouple after running for an hour or so with no fan ).
It turns out that each pad does 4 Stars & I got 5 pads for $4 and 4 stars per sheet is = $4/20 = 20 cents each.
That is clearly more than grease and you still need the screws.
It seems to work but the only advantage I see so far is that there is no chance to drip in to your tank ( over grease ).
Stu
http://www.kvms.com/Product/HSFPHASECM.aspx
seems to work as good as basic thermal grease ( based on touching each emitter with a thermocouple after running for an hour or so with no fan ).
It turns out that each pad does 4 Stars & I got 5 pads for $4 and 4 stars per sheet is = $4/20 = 20 cents each.
That is clearly more than grease and you still need the screws.
It seems to work but the only advantage I see so far is that there is no chance to drip in to your tank ( over grease ).
Stu
der_wille_zur_macht
Team RC
Stu, what temps are your LEDs hitting? I have no test equipment on hand at the moment, but after running for an hour at 500mA and then an hour at 700mA, my little 8-LED rig barely got warm. The stars were a little warm but the heatsink stayed more or less within a few degrees of room temp. This with no fan, and crammed into a small steel enclosure.
BeanAnimal
Premium Member
No clue stu.... but compact CRT tubes with lenses (rear projection TVs, overhead projectors, etc. have glycerin coolant between the tube and lens.
stugray
Premium Member
der_wille_zur_macht,
"Stu, what temps are your LEDs hitting?"
I am running 6X 3W at 700mA or ~12.5 Watts.
The large heatsink 6X5 actually gets warm to the touch after an hour or so with no fan & pointing up ( fins on desktop ).
My TC temp sensor reads 105-112 F right at the emitter base and that only varies a few degs across all 6 emitters.
Bean good to see you
After running the new setup for a few hours burn-in, I turned off the array and inspected all the emitters.
Just one looked "foggy" and bulged out more than the rest, so I touched it with my fingernail to see what the "fogginess" was... the lens moved when I did this.
I did the same to other emitters & the LENS still moved, but not as much.
I figure this means that the foggy lens got hotter than the rest.
It has since cleared up ( must have baked off the residue ).
Stu
"Stu, what temps are your LEDs hitting?"
I am running 6X 3W at 700mA or ~12.5 Watts.
The large heatsink 6X5 actually gets warm to the touch after an hour or so with no fan & pointing up ( fins on desktop ).
My TC temp sensor reads 105-112 F right at the emitter base and that only varies a few degs across all 6 emitters.
Bean good to see you
After running the new setup for a few hours burn-in, I turned off the array and inspected all the emitters.
Just one looked "foggy" and bulged out more than the rest, so I touched it with my fingernail to see what the "fogginess" was... the lens moved when I did this.
I did the same to other emitters & the LENS still moved, but not as much.
I figure this means that the foggy lens got hotter than the rest.
It has since cleared up ( must have baked off the residue ).
Stu
kcress
New member
Hi Stu; unless your TC is gossamer like butterfly antenna it will give you a pretty substantial, lower than reality, reading. The TC will actually act as a heat sink and cool the point you're trying to measure, a very Heisenberg like result.
I try to use a temp gun as much as I can these days. although they have their limitations too.
I try to use a temp gun as much as I can these days. although they have their limitations too.
der_wille_zur_macht
Team RC
FWIW, I've seen people in other DIY LED threads mention the same phemonemon you are, stu. No clear explanantion or "fix" though.
kcress, don't most IR temp guns have a pretty big (inch or so) target area, and aren't we really concerned about the temperature of a really small spot (square mm or so) on the LED?
I'm still not recovered from how cool my rig ran. I really expected it to seem a lot hotter with 8 LEDs on a fairly small heatsink - at least it's small compared to the heatsinkusa.com monsters most people are using. I was planning on running these LEDs at 500mA over the tank, with fans on the sinks, but now I might just run them at 700mA without fans, or 1A with fans.
kcress, don't most IR temp guns have a pretty big (inch or so) target area, and aren't we really concerned about the temperature of a really small spot (square mm or so) on the LED?
I'm still not recovered from how cool my rig ran. I really expected it to seem a lot hotter with 8 LEDs on a fairly small heatsink - at least it's small compared to the heatsinkusa.com monsters most people are using. I was planning on running these LEDs at 500mA over the tank, with fans on the sinks, but now I might just run them at 700mA without fans, or 1A with fans.
BeanAnimal
Premium Member
der_wille_zur_macht
Don't be fooled by the temperature of the heatsink mass... If you pump the power into the LEDs, most of it comes out as heat. The better the heatsink, the faster the heat travels from the junction to the edge of the fins. Poorly designed heatsinks stay cool at the edges and instead amass heat at the core (keeping the junction temperature higher than it needs to be for the given mass it is attached to).
In other words, it can be very misleading to use the overall heatsink "feel" to determine acceptable power levels. If you have a way of measuring temperature close to the junction, you will get a much better idea of how hard you can push the LEDs.
Don't be fooled by the temperature of the heatsink mass... If you pump the power into the LEDs, most of it comes out as heat. The better the heatsink, the faster the heat travels from the junction to the edge of the fins. Poorly designed heatsinks stay cool at the edges and instead amass heat at the core (keeping the junction temperature higher than it needs to be for the given mass it is attached to).
In other words, it can be very misleading to use the overall heatsink "feel" to determine acceptable power levels. If you have a way of measuring temperature close to the junction, you will get a much better idea of how hard you can push the LEDs.
sammy113
New member
Yes, I used Arctic Silver Thermal Compound
<a href=showthread.php?s=&postid=15420734#post15420734 target=_blank>Originally posted</a> by der_wille_zur_macht
And DWIZUM said, let there be light!
24v DC power supply in foreground, home-made driver in lower right, and 8 LEDs on a PC heatsink back by the soldering iron.
How the hell did you fit 8 into such a small heatsink? Can we see a pic with lights off?
Aren't you worried that 8 LEDs for a small heat sink may be not enough cooling material to handle the 8 LEDs? Maybe I'm wrong. Nice work on the DIY driver. Can you make one of those to run like 12 with just one driver? How much did it cost you to build this driver? I might be trying some
der_wille_zur_macht
Team RC
There are only four stars on the heatsink, but 8 LEDs - two of the stars are three-up Endor stars, which have three Rebels per star. So, two stars with three LED on them, and two stars with one LED on them. The heatsink is maybe 2.5 x 3.5 inches, so the four stars are spaced a little farther apart than is the norm on some others' builds.
The particular chip I am using in this DIY driver can only handle 8 LED at a time, up to 1.5A current (peak, so probably 20 - 30% lower depending on component choice), and can operate in boost or buck topology with from 5 - 36v input. The IC itself was like a buck, and the other components a few bucks more. So maybe $5 - $6 in the driver - and it's basically equivalent to a $15 buckpuck, but it can handle 8 LED instead of 6. The company that makes this IC (ON semi) also make a pin-compatible chip that can drive 10 LED per circuit. And, there are other chips out there that can drive more. But, these drivers are so cheap to put together, IMHO there isn't as big a motivation to cram many LED on one driver as if you were spending $20+ on commercial drivers.
Half the reason I did the DIY driver was just "for the experience" of doing it, and for the flexibility. I can change components to get it to work any way I want, instead of being stuck with an off the shelf product. Heck, I've alread swapped out the sense resistor to change it from a 500mA to a 700mA driver. Also, the economy factored in to my choice. Often in DIY, the DIY solution isn't any cheaper than off-the-shelf. This is clearly a situation where DIY is actually cheaper. If I ever do build an LED array for a really big tank, doing these DIY drivers could save a few hundred bucks.
And yeah, I am worried about heat, but it doesn't seem to be an issue. The sink barely gets above room temp, and the stars themselves only a little warmer than that. I know that junction temp is what really matters and I don't have a way to measure that right now, but I can't imagine they're getting THAT hot, and I do have a fan on the sink if I need it.
Think about it this way - these CPU heatsinks were designed to cool CPUs that consume 50 - 60 watts, and turn most of that to heat. I've only got 15 - 20 watts of power in these LEDs.
I'm hoping to borrow an IR gun to try and get a reading on the LED emitter itself, but I'm less worried now that I have it running successfully than I was before I first fired them up.
I'm planning a detailed write up of the driver construction project, but that's a month or two away. I want to build several of them in different configurations to get a full set of experience to draw on before writing a ton about the subject.
The particular chip I am using in this DIY driver can only handle 8 LED at a time, up to 1.5A current (peak, so probably 20 - 30% lower depending on component choice), and can operate in boost or buck topology with from 5 - 36v input. The IC itself was like a buck, and the other components a few bucks more. So maybe $5 - $6 in the driver - and it's basically equivalent to a $15 buckpuck, but it can handle 8 LED instead of 6. The company that makes this IC (ON semi) also make a pin-compatible chip that can drive 10 LED per circuit. And, there are other chips out there that can drive more. But, these drivers are so cheap to put together, IMHO there isn't as big a motivation to cram many LED on one driver as if you were spending $20+ on commercial drivers.
Half the reason I did the DIY driver was just "for the experience" of doing it, and for the flexibility. I can change components to get it to work any way I want, instead of being stuck with an off the shelf product. Heck, I've alread swapped out the sense resistor to change it from a 500mA to a 700mA driver. Also, the economy factored in to my choice. Often in DIY, the DIY solution isn't any cheaper than off-the-shelf. This is clearly a situation where DIY is actually cheaper. If I ever do build an LED array for a really big tank, doing these DIY drivers could save a few hundred bucks.
And yeah, I am worried about heat, but it doesn't seem to be an issue. The sink barely gets above room temp, and the stars themselves only a little warmer than that. I know that junction temp is what really matters and I don't have a way to measure that right now, but I can't imagine they're getting THAT hot, and I do have a fan on the sink if I need it.
Think about it this way - these CPU heatsinks were designed to cool CPUs that consume 50 - 60 watts, and turn most of that to heat. I've only got 15 - 20 watts of power in these LEDs.
I'm hoping to borrow an IR gun to try and get a reading on the LED emitter itself, but I'm less worried now that I have it running successfully than I was before I first fired them up.
I'm planning a detailed write up of the driver construction project, but that's a month or two away. I want to build several of them in different configurations to get a full set of experience to draw on before writing a ton about the subject.
sammy113
New member
Great der_wille_zur_macht! I'm gonna try one of those. So how do you like those endo stars? Are they as powerful as 3 CREEs? Like you said, build it for the experience. I think there is nothing more satisfying than looking at your own DIY work and say to yourself: Damn! I made this! haha
This thread was automatically split due to performance issues. You can find the rest of the thread here: http://reefcentral.com/forums/showthread.php?threadid=1678127
- Status
